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: provide server to client comms path
19 Details : server to client communications path via new LDLM_SET_INFO rpc
23 Description: truncate starts GFP_FS allocation under transaction causing deadlock
24 Details : ldiskfs_truncate calls grab_cache_page which may start page
25 allocation under an open transaction. This may lead to
26 calling prune_icache with consequent lustre reentrance.
30 Description: send correctly lsm on open replay
31 Details : MDS is trust to LSM size on replay open, but client can set wrong size
36 Description: enable client interface failover
37 Details : When a child reconnects from another NID, properly update export
38 nid hash position and ldlm reverse import.
40 Severity : enhancement
42 Description: Communicate OST degraded/readonly state via statfs to MDS
43 Details : Flags in the statfs returned from OSTs indicate whether the
44 OST is in a degraded RAID state, or if the filesystem has
45 turned read-only after a filesystem error is detected.
50 Description: don't panic if EPROTO was hit when reading symlink
51 Details : correctly handling request reference in error cases.
53 Severity : enhancement
55 Description: Add LustreNetLink, a kernel-userspace communcation path.
57 Severity : enhancement
59 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
64 Description: access to llog context before init.
65 Details : move handling CATALOGS file at osc layer and forbid access to llog
69 Frequency : rare, only when client is writing to OST on local node
71 Description: Avoid deadlock for local client writes
72 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
73 memory freeing context. This allows OST threads to set the
74 PF_MEMALLOC flag on task structures in order to allocate memory
75 from reserved pools and complete IO.
76 Use GFP_HIGHUSER for OST allocations for non-local client writes,
77 so that the OST threads generate memory pressure and allow
78 inactive pages to be reclaimed.
80 Severity : enhancement
82 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
87 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
88 Details : move ldlm namespace creation in setup phase to avoid grab
89 _lprocfs_lock with cli_sem held.
93 Description: Temporarily disable grant shrink.
94 Details : Disable the feature for debugging.
97 Frequency : only when formatting test filesystems
99 Description: Unable to run several mkfs.lustre on loop devices at the same time
100 Details : mkfs.lustre returns error 256 on the concurrent loop devices
101 formatting. The solution is to proper handle the error.
103 Severity : enhancement
105 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
107 Severity : enhancement
109 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
114 Description: resolve race between obd_disconnect and class_disconnect_exports
115 Details : if obd_disconnect will be called to already disconnected export he
116 forget release one reference and osc module can't unloaded.
118 Severity : enhancement
120 Description: Allow tuning service thread via /proc
121 Details : For each service a new
122 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
123 created that can be used to set min/max thread counts, and get the
124 current number of running threads.
129 Descriptoin: don't return error if have partially created objects for file.
130 Details : lov_update_create_set uses set->set_success as index for created
131 objects, so if some requests will be failed, they will have hole
132 at end of array and qos_shrink_lsm for allocate correct lsm.
134 Severity : enhancement
136 Description: Update OFED support to 1.4.1
141 Description: Reduce small size read RPC
142 Details : Set read-ahead limit for every file and only do read-ahead when
143 available read-ahead pages are bigger than 1M to avoid small size
146 Severity : enhancement
148 Description: move AT tunable parameters for more consistent usage
149 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
151 Severity : enhancement
153 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
154 Details : allow skip disconnected ost for send statfs request and hide error
160 Description: fix racy locking of mballoc block bitmaps causing BUG
161 Details : The locking of the mballoc buddy bitmap and the in-memory
162 block bitmap was using two different spin locks in some
163 cases. This made it possible to incorrectly access the
164 mballoc bitmap while another process was modifying it,
165 causing a sanity assertion to fail. While no on-disk corruption
166 was reported, there was some risk of this happening.
169 Frequency : rare, on llog test 6
171 Descriptoin: don't allow connect to already connected import
172 Details : allowing connect to already connected import is hide connecting problem.
177 Descriptoin: allow kill process which wait statahead result
178 Details : for some reasons 'ls' can stick in waiting result from statahead,
179 in this case need way for kill this process.
181 Severity : enhancement
183 Description: Add state history info file, enhance import info file
184 Details : Track import connection state changes in a new osc/mdc proc file;
185 add overview-type data to the osc/mdc import proc file.
187 Severity : enhancement
189 Description: MDS create should not wait for statfs RPC while holding DLM lock.
191 Severity : enhancement
193 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
198 Descriptoin: OST grant too much space to client even there are not enough space.
199 Details : Client will shrink its grant cache to OST if there are no write
200 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
201 this grant cache if there are already not enough avaible space
202 (left_space < total_clients * 32M).
205 Frequency : start MDS on uncleanly shutdowned MDS device
207 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
208 Details : stay in waiting mds<>ost recovery finished produce random bugs
209 due race between two ll_sync thread for one lov target. send
210 ACTIVATE event only if connect realy finished and import have
214 Frequency : rare, connect and disconnect target at same time
216 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
217 Details : don't call obd_disconnect under lov_lock. this long time
218 operation and can block ptlrpcd which answer to connect request.
223 Descriptoin: don't lose wakeup for imp_recovery_waitq
224 Details : recover_import_no_retry or invalidate_import and import_close can
225 both sleep on imp_recovery_waitq, but we was send only one wakeup
229 Frequency : always with long access acl
231 Descriptoin: mds can't pack reply with long acl.
232 Details : mds don't control size of acl but they limited by reint/getattr
235 Severity : enhancement
237 Description: Update to SLES10 kernel-2.6.16.60-0.33.
239 Severity : enhancement
241 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
244 Frequency : start MDS on uncleanly shutdowned MDS device
246 Descriptoin: aborting recovery hang on MDS
247 Details : don't throttle destroy RPCs for the MDT.
250 Frequency : on remount
252 Description: external journal device not working after the remount
253 Details : clear dev_rdonly flag for external journal devices in
259 Description: shutdown vs evict race
260 Details : client_disconnect_export vs connect request race.
261 if client will evicted at this time - we start invalidate
262 thread without referece to import and import can be freed
266 Frequency : rare, need acl's on inode.
268 Description: client can't handle ost additional correctly
269 Details : if ost was added after client connected to mds client can have
270 hit lnet_try_match_md ... to big messages to wide striped files.
271 in this case need teach client to handle config events about add
272 lov target and update client max ea size at that event.
274 Severity : enhancement
276 Description: Changelogs
277 Details : Changelogs are a lightweight mechanism to track filesystem
278 metadata and namespace changes. The changelog is recorded
279 permanently on the MDTs, and is periodically "consumed" / purged
280 when records are no longer needed.
282 Severity : enhancement
284 Description: compact fld format with extents
285 Details : Store range of seq rather than every seq in FLD. Seq
286 controller update FLD rather than clients. In Case of CMD, mdt0
287 has FLD, all other metadata server act as non persistent proxy
288 for FLD queries and cache fld entries in fld cache.
293 Description: don't skip ost target if they assigned to file
294 Details : Drop slow OSCs if we can, but not for requested start idx.
295 This means "if OSC is slow and it is not the requested
296 start OST, then it can be skipped, otherwise skip it only
297 if it is inactive/recovering/out-of-space.
301 Description: more cleanup in mds_lov
302 Details : not send LOV EA under replay, we can't know about they size at this
303 time. Don't allow client connect to mds before any ost connected,
304 for avoid problems with LOV EA size and returning EIO to client.
306 Severity : enhancement
308 Description: Interoperability at server side (Disk interoperability)
310 Severity : enhancement
312 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
314 Severity : enhancement
316 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
318 Severity : enhancement
320 Description: New client IO stack (CLIO).
322 Severity : enhancement
324 Description: Commit on sharing. Eliminate inter-client dependencies between
325 uncommitted transactions by doing transaction commits.
326 Thereby clients may recovery independently.
329 Frequency : Create a symlink file with a very long name
331 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
332 Details : If there is no extra space in the request for early cancels,
333 ldlm_req_handles_avail() returns 0 instead of a negative value.
335 Severity : enhancement
337 Description: Add /proc entry for import status
338 Details : The mdc, osc, and mgc import directories now have
339 an import directory that contains useful import data for debugging
342 Severity : enhancement
344 Description: Re-disable certain /proc logging
345 Details : Enable and disable client's offset_stats, extents_stats and
346 extents_stats_per_process stats logging on the fly.
349 Frequency : Only on FC kernels 2.6.22+
351 Description: oops in statahead
352 Details : Do not drop reference count for the dentry from VFS when lookup,
353 VFS will do that by itself.
355 Severity : enhancement
357 Description: Generic /proc file permissions
358 Details : Set /Proc file permissions in a more generic way to enable non-
359 root users operate on some /proc files.
363 Description: Hitting mdc_commit_close() ASSERTION
364 Details : Properly handle request reference release in
365 ll_release_openhandle().
369 Description: quota recovery deadlock during mds failover
370 Details : This patch includes att18982, att18236, att18237 in bz14840.
372 1. fix osts hang when mds does failover with quotaon
373 2. prevent watchdog storm when osts threads wait for the
378 Frequency : only patchless client
379 Description: add workaround for race between add/remove dentry from hash
381 Severity : enhancement
383 Description: Allow OST glimpses to return PW locks
387 Description: LBUG when llog conf file is full
388 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
392 Description: Prevent import from entering FULL state when server in recovery
396 Description: service mount cannot take device name with ":"
397 Details : Only when device name contains ":/" will mount treat it as
403 Description: replace ptlrpcd with the statahead thread to interpret the async
404 statahead RPC callback
408 Frequency : on recovery
409 Description: I/O failures after umount during fail back
410 Details : if client reconnected to restarted server we need join to recovery
411 instead of find server handler is changed and process self
412 eviction with cancel all locks.
414 Severity : enhancement
416 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
418 Severity : enhancement
420 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
422 Severity : enhancement
424 Description: Upcall on Lustre log has been dumped
425 Details : Allow for a user mode script to be called once a Lustre log has
426 been dumped. It passes the filename of the dumped log to the
427 script, the location of the script can be specified via
428 /proc/sys/lnet/debug_log_upcall.
433 Description: avoid idr_remove called for id which is not allocated.
434 Details : Move assigment s_dev for clustered nfs to end of initialization,
435 to avoid problem with error handling.
440 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
441 Details : When connection is reused this not moved from CONN_UNUSED_HASH
442 into CONN_USED_HASH and this prodice warning when put connection
443 again in unused hash.
446 Severity : enhancement
448 Description: File striping can now be set to use an arbitrary pool of OSTs.
450 Severity : enhancement
452 Description: Export bytes_read/bytes_write count on OSC/OST.
456 Description: Early reply size mismatch, MGC loses connection
457 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
458 the connect flags are properly negotiated.
463 Description: quotas are not honored with O_DIRECT
464 Details : all writes with the flag O_DIRECT will use grants which leads to
465 this problem. Now using OBD_BRW_SYNC to guard this.
469 Description: add quota statistics
470 Details : 1. sort out quota proc entries and proc code.
471 2. add quota statistics
473 Severity : enhancement
475 Description: enable quota support for HEAD.
479 Description: Properly propagate oinfo flags from lov to osc for statfs
480 Details : restore missing copy oi_flags to lov requests.
482 Severity : enhancement
484 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
485 : l_getgroups(8), lst(8), routerstat(8)
486 Details : included man pages for llobdstat(8), llstat(8),
487 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
489 Severity : enhancement
491 Description: configure's --enable-quota should check the
492 : kernel .config for CONFIG_QUOTA
493 Details : configure is terminated if --enable-quota is passed but
494 : no quota support is in kernel
498 Description: Remove portals compatibility
499 Details : Remove portals compatibility, not interoperable with releases
504 Description: Resolve device initialization race
505 Details : Prevent proc handler from accessing devices added to the
506 obd_devs array but yet be intialized.
508 Severity : enhancement
510 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
512 Severity : enhancement
514 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
519 Description: Using wrong pointer in osc_brw_prep_request
520 Details : Access to array[-1] can produce panic if kernel compiled with
521 CONFIG_PAGE_ALLOC enabled
525 Description: Client runs out of low memory
526 Details : Consider only lowmem when counting initial number of llap pages
530 Description: *optional* service tags registration
531 Details : if the "service tags" package is installed on a Lustre node
532 When the filesystem is mounted, a local-node service tag will
533 be created. See http://inventory.sun.com/ for more information
534 about the Service Tags asset management system.
538 Description: Kernel BUG tries to release flock
539 Details : Lustre does not destroy flock lock before last reference goes
540 away. So always drop flock locks when client is evicted and
541 perform unlock regardless of successfulness of speaking to MDS.
545 Description: add refcount protection for osc callbacks, avoid panic on shutdown
549 Description: sanity test 65a fails if stripecount of -1 is set
550 Details : handle -1 striping on filesystem in ll_dirstripe_verify
555 Description: ASSERTION(CheckWriteback(page,cmd)) failed
556 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
559 Severity : enhancement
561 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
565 Description: do not process already freed flock
566 Details : flock can possibly be freed by another thread before it reaches
567 to ldlm_flock_completion_ast.
571 Description: LBUG during stress test
572 Details : Need properly lock accesses the flock deadlock detection list.
576 Description: oops in page fault handler
577 Details : kernel page fault handler can return two special 'pages' in error
578 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
582 Description: timeout with invalidate import.
583 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
584 should be handled by ptlrpcd. This produce long age waiting and
585 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
587 Severity : enhancement
589 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
593 Description: enable MGS and MDT services start separately
594 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
595 a co-located MGS without starting the MGS, which is a complement
596 to 'nosvc' mount option.
600 Frequency : after recovery
601 Description: precreate to many object's after del orphan.
602 Details : del orphan st in oscc last_id == next_id and this triger growing
603 count of precreated objects. Set flag LOW to skip increase count
604 of precreated objects.
608 Frequency : rare, on clear nid stats
609 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
610 Details : when clean nid stats sometimes try destroy live entry,
611 and this produce panic in free.
615 Description: Stack overflow during MDS log replay
616 ease stack pressure by using a thread dealing llog_process.
620 Description: wait until IO finished before start new when do lock cancel.
621 Details : VM protocol want old IO finished before start new, in this case
622 need wait until PG_writeback is cleared until check dirty flag and
623 call writepages in lock cancel callback.
625 Severity : enhancement
627 Description: using special macro for print time and cleanup in includes.
631 Description: mds_mfd_close() ASSERTION(rc == 0)
632 Details : In mds_mfd_close(), we need protect inode's writecount change
633 within its orphan write semaphore to prevent possible races.
637 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
641 Frequency : rare, on shutdown ost
642 Description: don't hit live lock with umount ost.
643 Details : shrink_dcache_parent can be in long loop with destroy dentries,
644 use shrink_dcache_sb instead.
648 Description: don't panic with use echo client
649 Details : echo client pass NULL as client nid pointer and this produce null
654 Description: fix build on ppc32
655 Details : compile code with -m64 flag produce wrong object file for ppc32.
659 Description: add message levels for liblustreapi
663 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
664 Details : Move the 'good_osts' check before the 'total_bavail' check. This
665 will result in an -EAGAIN and in the exit call path we call
666 alloc_rr() which will with increasing aggressiveness attempt to
667 aquire precreated objects on the minimum number of required OSCs.
671 Description: Use old size assignment to avoid deadlock
672 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
673 the scheduling while holding a spinlock. We do not need locking
674 for size in ll_update_inode() because size is only updated from
675 the MDS for directories or files without objects, so there is no
676 other place to do the update, and concurrent access to such inodes
677 are protected by the inode lock.
681 Description: resolve "_IOWR redefined" build error on SLES10
685 Description: dump the memory debugging after all modules are unloaded to
686 suppress false negative in conf_sanity test 39
688 Severity : enhancement
690 Description: build kernel-ib packages for OFED 1.3 in our release cycle
695 Description: fix SLES kernel versioning
696 Details : the kernel version for our SLES 10 kernel did not include a "-"
697 before the "smp" at the end. while this was not a problem in
698 general, it did mean that software trying to use the kernel
699 version to try to detect a vendor specific kernel would fail.
700 this was most evident by the OFED build scripts.
704 Description: Don't update lov_desc members until making sure they are valid
705 Details : When updating lov_desc members via proc fs, need fix their
706 validities before doing the real update.
710 Description: don't put request into delay list while invalidate in flight.
711 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
712 import in flight. this produce timeout for invalidate and sometimes
713 can cause stale data.
717 Frequency : on ppc only
718 Description: not convert ost objects for directory because it's not exist.
719 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
723 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
725 Severity : enhancement
727 Description: Update to support 2.6.22.14 vanilla kernel.
731 Frequency : rare, on recovery
732 Description: read procfs can produce deadlock in some situation
733 Details : Holding lprocfs lock which send rpc can produce block for destroy
734 obd objects and this also block reconnect with -EALREADY. This
735 isn't fix all lprocfs bugs - but make it rare.
737 Severity : enhancement
739 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
742 Frequency : frequent on X2 node
744 Description: mdc_set_open_replay_data LBUG
745 Details : Set replay data for requests that are eligible for replay.
749 Description: lustre_mgs: operation 101 on unconnected MGS
750 Details : When MGC is disconnected from MGS long enough, MGS will evict the
751 MGC, and late on MGC cannot successfully connect to MGS and a lot
752 of the error messages complaining that MGS is not connected.
756 Frequency : on network error
757 Description: panic with double free request if network error
758 Details : mdc_finish_enqueue is finish request if any network error ocuring,
759 but it's true only for synchronus enqueue, for async enqueue
760 (via ptlrpcd) this incorrect and ptlrpcd want finish request
763 Severity : enhancement
765 Description: client-side metadata stat-ahead during readdir(directory readahead)
766 Details : perform client-side metadata stat-ahead when the client detects
767 readdir and sequential stat of dir entries therein
770 Frequency : on start mds
772 Description: Implement get_info(last_id) in obdfilter.
775 Frequency : occasional
777 Description: Correctly check stale fid, not start epoch if ost not support SOM
778 Details : open with flag O_CREATE need set old fid in op_fid3 because
779 op_fid2 was overwritten with new generated fid, but mds can answer
780 with one of these two fids and both is not stale. Setattr
781 incorrectly started an epoch and assume will be called
782 done_writing, but without SOM done_writing ever being called.
785 Frequency : rare, depends on device drivers and load
787 Description: MDS or OSS nodes crash due to stack overflow
788 Details : Code changes in 1.8.0 increased the stack usage of some functions.
789 In some cases, in conjunction with device drivers that use a lot
790 of stack the MDS (or possibly OSS) service threads could overflow
791 the stack. One change which was identified to consume additional
792 stack has been reworked to avoid the extra stack usage.
795 Frequency : occasional
797 Description: Do not fail import if osc_interpret_create gets -EAGAIN
798 Details : If osc_interpret_create got -EAGAIN it immediately exits and
799 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
800 oscc_has_objects and see OSC has no objests and call
801 oscc_internal_create to resend create request.
803 Severity : enhancement
805 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
807 Severity : enhancement
809 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
812 Frequency : very rare
814 Description: Fix replay if there is an un-replied request and open
815 Details : In some cases, older replay request will revert the
816 mcd->mcd_last_xid on MDS which is used to record the client's
819 Severity : enhancement
821 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
823 Severity : enhancement
825 Description: Add rhel5 support to HEAD.
827 Serverity : enhancement
829 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
834 Description: Don't allow skipping OSTs if index has been specified.
835 Details : Don't allow skipping OSTs if index has been specified, make
836 locking in internal create lots better.
840 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
841 Details : access to bitfield in structure is always rounded to long
842 and this produce problem with not atomic change any bit.
846 Description: Lustre make rpms failed.
847 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
848 builds Lustre rpms from tarball.
851 Frequency : rare on shutdown ost
853 Description: If llog cancel was not send before clean_exports phase, this can
854 produce deadlock in llog code.
855 Details : If llog thread has last reference to obd and call class_import_put
856 this produce deadlock because llog_cleanup_commit_master wait when
857 last llog_commit_thread exited, but this never success because was
858 called from llog_commit_thread.
862 Description: allow userland application know is lost one of stripes.
863 Details : fill lvb_blocks with error code on ost and return it to
864 application if error flag found.
868 Description: NULL lov_tgts causing MDS oops
869 Details : more safe checks for NULL lov_tgts for avoid oops.
871 Severity : enhancement
873 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
877 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
878 Details : If an OST is down the MDS will hang indefinitely in
879 obd_statfs_rqset() waiting for the statfs data. While for
880 MDS QOS usage of statfs, it should not stuck in waiting.
882 Severity : enhancement
884 Description: remote_acl support
885 Details : Support ACL-based permission check for remote user.
886 Support setfacl/getfacl for remote user with the utils
887 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
888 the system "{s,g}etfacl" utils.
890 Severity : enhancement
892 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
894 Severity : enhancement
896 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
900 Description: make mgs_setparam() handle fsname containing dash
901 Details : fsname containing a dash does not work with lctl conf_param
903 Severity : enhancement
905 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
907 Severity : enhancement
909 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
913 Frequency : rare, at shutdown
914 Description: access already free / zero obd_namespace.
915 Details : if client_disconnect_export was called without force flag set,
916 and exist connect request in flight, this can produce access to
917 NULL pointer (or already free pointer) when connect_interpret
918 store ocd flags in obd_namespace.
922 Frequency : only at startup
923 Description: not alloc memory with spinlock held.
924 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
925 if any spinlock held.
927 Severity : enhancement
929 Description: make lustre randomly fail allocating memory
930 Details : Make lustre randomly failed allocating memory for testing purpose.
932 Severity : enhancement
934 Description: lost problems with lov objid file
935 Details : Fixes some scability and access to not inited memory problems
936 in work with lov objdid file.
941 Description: lfs find does not continue on file error
942 Details : Continue other files processing when a file/dir is absent.
946 Description: Inconsistent usage of lustre_pack_reply()
947 Details : Standardize the usage of lustre_pack_reply() such that it
948 always generate a CERROR on failure.
951 Frequency : occasional
953 Description: MDS hang or stay in waiting lock
954 Details : If client receive lock with CBPENDING flag ldlm need send lock
955 cancel as separate rpc, to avoid situation when cancel request
956 can't processed due all i/o threads stay in wait lock.
960 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
964 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
968 Description: Update to SLES9 kernel-2.6.5-7.287.3.
972 Description: Update to RHEL4 latest kernel.
974 Severity : enhancement
976 Description: Build SLES10 patchless client fails
977 Details : The configure was broken by run ./configure with
978 --with-linux-obj=.... argument for patchless client. When the
979 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
980 can't search header adequately. Use absolute path such as
981 -I($LINUX)/include instead.
985 Description: interrupt oig_wait produce painc on resend.
986 Details : brw_redo_request can be used for resend requests from ptlrpcd and
987 private set, and this produce situation when rq_ptlrpcd_data not
988 copyed to new allocated request and triggered LBUG on assert
989 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
990 ptlrpcd set if request is changed and can be safety changed to use
995 Description: LASSERT_{REQ,REP}SWAB macros are buggy
996 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
997 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
998 these macros and replace them with their difinitions instead.
1002 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1003 Details : Update which_patch & target file for SLES10 latest kernel.
1006 Description: Remove client patches from SLES 10 kernel.
1007 Details : This causes SLES 10 clients to behave as patchless clients
1008 even on a Lustre-patched (server) kernel.
1010 Severity : enhancement
1012 Description: self-adjustable client's lru lists
1013 Details : use adaptive algorithm for managing client cached locks lru
1014 lists according to current server load, other client's work
1015 pattern, memory activities, etc. Both, server and client
1016 side namespaces provide number of proc tunables for controlling
1019 Severity : enhancement
1021 Description: light-weight GSS support
1022 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1023 and reduce performance overhead.
1025 Severity : enhancement
1027 Description: Linux keyring support
1028 Details : Support using service of Linux keyring for Lustre GSS internal
1029 context refresh/cache mechanism.
1033 Description: Fix errors in lfs documentation
1034 Details : Fixes man pages
1038 Description: don't use GFP_* in generic Lustre code.
1039 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1043 Description: obdclass is limited by single OBD_ALLOC(idarray)
1044 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1048 Description: warnings with build patchless client with vanila 2.6.19 and up
1049 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1050 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1054 Frequency : Only for SLES
1055 Description: sanity_quota fail test_1
1056 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1057 file, which makes TSTID[2] inunique.
1061 Frequency : Only for SLES9
1062 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1063 Details : Update target/ChangeLog/which_patch .
1067 Frequency : Only for test_10 in sanity-quota.sh
1068 Description: a bug in quota test code
1069 Details : checking a test flag in a wrong place
1073 Frequency : Only for test_13 in sanity-quota.sh
1074 Description: a bug in quota test script
1075 Details : shouldn't check overfull allocated quota
1079 Description: jbd statistics
1080 Details : Port older jbd statistics patch for sles10
1082 Severity : enhancement
1084 Description: Add printing inode info into message about error in writepage.
1089 Description: reply_lock_interpret crash due to race with it and lock cancel.
1090 Details : Do not replay locks that are being cancelled. Do not reference
1091 locks by their address during replay, just by their handle.
1095 Frequency : When flocks are used.
1096 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1097 Details : Flock locks might destroy just granted lock if it could be merged
1098 with another existing flock, this is done in completion handler,
1099 so teach ldlm_cli_enquque_fini that this is a valid case for
1105 Description: Oops in read and write path when failing to allocate lock.
1106 Details : Check if lock allocation failed and return error back.
1110 Description: lstripe command fails for valid OST index
1111 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1112 instead of lov->desc.ld_active_tgt_count.
1116 Description: sanity.sh failed test 103
1117 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1122 Description: df doesn't work properly if diskfs blocksize != 4K
1123 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1127 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1128 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1130 Severity : enhancement
1132 Description: Adaptive timeouts
1133 Details : RPC timeouts adapt to changing server load and network
1134 conditions to reduce resend attempts and improve recovery time.
1138 Description: llapi_file_create() does not allow some changes
1139 Details : add llapi_file_open() that allows specifying the mode and
1140 open flags, and also returns an open file handle.
1144 Description: merge and cleanup kernel patches.
1145 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1149 Description: Add journal checksum support.(Kernel part)
1150 Details : The journal checksum feature adds two new flags i.e
1151 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1152 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1153 indicates that the commit block contains the checksum for
1154 the blocks described by the descriptor blocks. Now commit
1155 record can be sent to disk without waiting for descriptor
1156 blocks to be written to disk. This behavior is controlled
1157 using JBD2_FEATURE_ASYNC_COMMIT flag.
1161 Description: OSS needs mutliple precreate threads
1162 Details : Add ability to start more than one create thread per OSS.
1166 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1167 Details : execute lfs setstripe on client
1171 Description: mds_obd_create error creating tmp object
1172 Details : When the user sets quota on root, llog will be affected and can't
1173 create files and write files.
1176 Frequency : Always on ia64 patchless client, and possibly others.
1178 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1179 Details : This allows the patchless client to be loaded on architectures
1180 without this export.
1184 Description: RedHat Update kernel for RHEL5
1185 Details : Add the kernel config file for RHEL5.
1189 Description: RedHat Update kernel for RHEL5
1190 Details : Modify the kernel config file more closer RHEL5.
1194 Description: Build failure against Centos5 (RHEL5)
1195 Details : Define PAGE_SIZE when it isn't present.
1199 Description: test_7 on sanity-quota.sh can't be used on separate machines
1200 Details : add facet to handle it
1204 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1205 Details : with NFS, the anon dentry's parent was set to itself in
1206 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1207 corresponding dentry other than use rec->ur_name.
1209 Severity : enhancement
1211 Description: lfs setstripe enhancement
1212 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1216 Description: enable data checksumming by default
1217 Details : enable checksum by default, allow --disable-checksum
1218 configure option and "-o nochecksum" mount option. Checksums
1219 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1223 Description: lustre support for RHEL5
1224 Details : Add support for RHEL5.
1228 Description: Client eviction due to failover config
1229 Details : after a connection loss, the lustre client should attempt to
1230 reconnect to the last active server first before trying the
1231 other potential connections.
1235 Description: when mds and osts use different quota unit(32bit and 64bit),
1236 quota will be released repeatly.
1237 Details : void sending multiple quota reqs to mds, which will keep the
1238 status between the reqs.
1242 Description: osts not allocated evenly to files
1243 Details : change the condition to increase offset_idx
1246 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1248 Description: Data corruption for OSTs that are formatted larger than 2TB
1250 Details : When generating the bio request for lustre file writes the
1251 sector number would overflow a temporary variable before being
1252 used for the IO. The data reads correctly from Lustre (which
1253 will overflow in a similar manner) but other file data or
1254 filesystem metadata may be corrupted in some cases.
1258 Description: Tune the kernel for good SCSI performance.
1259 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1260 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1265 Description: rewrite ext2-derived code in obdclass/uuid.c
1266 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1267 from scratch preserving functionality.
1271 Description: "lfs find -obd UUID" prints directories
1272 Details : "lfs find -obd UUID" will return all directory names instead
1273 of just file names. It is incorrect because the directories
1274 do not reside on the OSTs.
1278 Description: use i_size_read and i_size_write in 2.6 port
1279 Details : replace inode->i_size access with i_size_read/write()
1282 Frequency : when removing large files
1284 Description: scheduling issue during removal of large Lustre files
1285 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1286 It causes scheduling issues when removing large files (17TB in the
1290 Frequency : only with liblustre clients on XT3
1292 Description: evictions taking too long
1293 Details : allow llrd to evict clients directly on OSTs
1296 Frequency : only on ppc
1298 Description: /proc/fs/lustre/devices broken on ppc
1299 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1300 We should make sure the type of 'index' is loff_t and then cast
1301 explicitly as needed below. Do not assign an explicitly cast
1305 Frequency : only for rhel5
1307 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1308 Details : Modify the target file & which_kernel.
1310 Severity : enhancement
1312 Description: omit set fsid for export NFS
1313 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1314 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1318 Frequency : Always, for kernels after 2.6.16
1319 Description: Fix warning idr_remove called for id=.. which is not allocated.
1320 Details : Last kernels save old s_dev before kill super and not allow
1321 to restore from callback - restore it before call kill_anon_super.
1325 Description: buffer overruns could theoretically occur
1326 Details : llapi_semantic_traverse() modifies the "path" argument by
1327 appending values to the end of the origin string, and a
1328 overrun may occur. Adding buffer overrun check in liblustreapi.
1332 Description: Fix error on 'ls .' at the top of the Lustre mount.
1333 Details : Don't revalidate dentry if it is a root dentry.
1337 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1338 Details : Modify vm-tunables-rhel4.patch.
1342 Description: Kernel config for 2.6.18-vanilla.
1343 Details : Modify targets/2.6-vanilla.target.in.
1344 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1345 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1346 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1347 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1351 Description: improve handling recoverable errors
1352 Details : if request processig with error which can be recoverable on server
1353 request should be resend, otherwise page released from cache and
1358 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1359 Details : Modify target file & which_patch.
1360 A flaw was found in the IA32 system call emulation provided
1361 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1362 value could be stored in the %RAX register, which could trigger an
1363 out-of-bounds system call table access. An untrusted local user
1364 could exploit this flaw to run code in the kernel
1365 (ie a root privilege escalation). (CVE-2007-4573).
1369 Description: O_DIRECT bypasses client statistics.
1370 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1371 read_ahead_stats were updated. copied the stats section from
1372 osc_send_oap_rpc() into async_internal().
1376 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1380 Description: change order of libsysio includes
1381 Details : '#include sysio.h' should always come before '#include xtio.h'
1385 Description: handle "serious error: objid * already exists" more gracefully
1386 Details : If LAST_ID value on disk is smaller than the objects existing in
1387 the O/0/d* directories, it indicates disk corruption and causes an
1388 LBUG(). If the object is 0-length, then we should use the existing
1389 object. This will help to avoid a full fsck in most cases.
1391 Severity : enhancement
1393 Description: adapt the lustre_config script to support the upgrade case
1394 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1395 server targets to 1.6 in parallel.
1399 Description: To avoid grant space > avaible space when the disk is almost
1400 full. Without this patch you might see the error "grant XXXX >
1401 available" or some LBUG about grant, when the disk is almost
1403 Details : In filter_check_grant, for non_grant cache write, we should
1404 check the left space by if (*left > ungranted + bytes), instead
1405 of (*left > ungranted), because only we are sure the left space
1406 is enough for another "bytes", then the ungrant space should be
1407 increase. In client, we should update cl_avail_grant only there
1408 is OBD_MD_FLGRANT in the reply.
1412 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1413 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1414 described in CVE-2007-4573. This problem could allow untrusted
1415 local users to gain root access.
1418 Frequency : when using O_DIRECT and quotas
1420 Description: Incorrect file ownership on O_DIRECT output files
1421 Details : block usage reported by 'lfs quota' does not take into account
1422 files that have been written with O_DIRECT.
1427 Description: touch file failed when fs is not full
1428 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1429 otherwise we can get ENOSP while fs is not full.
1433 Description: parallel lock callbacks
1434 Details : Instead of sending blocking and completion callbacks as separated
1435 requests, adding them to a set and sending in parallel.
1438 Frequency : only for Cray XT3
1439 Bugzilla : 12829/13455
1440 Description: Changing primary group doesn't change the group lustre assigns to
1442 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1443 overridden with the primary group provided by the group upcall,
1444 whereas the supplied fsgid can be trusted if it is in the list of
1445 supplementary groups returned by the group upcall.
1447 Severity : enhancement
1449 Description: Allow masking D_WARNING, D_ERROR messages from console
1450 Details : Console messages can now be disabled via lnet.printk.
1454 Description: User code with malformed file open parameter crashes client node
1455 Details : Before packing join_file req, all the related reference should be
1456 checked carefully in case some malformed flags cause fake
1457 join_file req on client.
1461 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1462 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1463 mode is equal to the granted mode. However, it should also check
1464 that LDLM_FL_CP_REQD is not set.
1468 Description: Heavy nfs access might result in deadlocks
1469 Details : After ELC code landed, it is now improper to enqueue any mds
1470 locks under och_sem, because enqueue might want to decide to
1471 cancel open locks for same inode we are holding och_sem for.
1475 Description: Client eviction while running blogbench
1476 Details : A lot of unlink operations with concurrent I/O can lead to a
1477 deadlock causing evictions. To address the problem, the number of
1478 oustanding OST_DESTROY requests is now throttled to
1479 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1480 callbacks are processed in priority.
1484 Description: enable ACLs on MDS by default
1485 Details : ACLs must be enabled on MDS by default.
1488 Frequency : PPC/PPC64 only
1490 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1491 Details : fix duplicated definitions between asm-ppc64/types.h and
1492 lustre_types.h on PPC.
1495 Frequency : PPC/PPC64 only
1497 Description: asm-ppc/segment.h does not exist
1498 Details : fix compile issue on PPC.
1502 Description: better handle error messages in extents code
1505 Frequency : RHEL4 only
1507 Description: mkfs is very slow on IA64/RHEL4
1508 Details : A performance regression has been discovered in the MPT Fusion
1509 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1510 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1511 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1513 Severity : enhancement
1515 Description: SNMP support enhancement
1516 Details : Adding total number of sampled request for an MDS node in snmp
1519 Severity : enhancement
1521 Description: Optimize ldlm waiting list processing for PR extent locks
1522 Details : When processing waiting list for read extent lock and meeting read
1523 lock that is same or wider to it that is not contended, skip
1524 processing rest of the list and immediatelly return current
1525 status of conflictness, since we are guaranteed there are no
1526 conflicting locks in the rest of the list.
1530 Description: Time out and refuse to reconnect
1531 Details : When the failover node is the primary node, it is possible
1532 to have two identical connections in imp_conn_list. We must
1533 compare not conn's pointers but NIDs, otherwise we can defeat
1534 connection throttling.
1538 Description: port llog fixes from b1_6 into HEAD
1539 Details : Port llog reference couting and some llog cleanups from b1_6
1540 (bug 10800) into HEAD, for protect from panic and access to already
1541 free llog structures.
1545 Description: Detect stride IO mode in read-ahead
1546 Details : When a client does stride read, read-ahead should detect that and
1547 read-ahead pages according to the detected stride pattern.
1551 Description: data checksumming impacts single node performance
1552 Details : add support for several checksum algorithm. Currently, only CRC32
1553 and Adler-32 are supported. The checksum type can be changed on
1554 the fly via /proc/fs/lustre/osc/*/checksum_type.
1558 Description: use adler32 for page checksums
1559 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1564 Description: build for x2 fails
1565 Details : fix compile issue on Cray systems.
1569 Description: Properly match for duplicate locks
1570 Details : Due to different lock order from skiplists code, we need to
1571 traverse entire list for now
1574 Frequency : only on PPC/SLES10
1576 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1577 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1578 be defined. Add a hack in mkfs_lustre.c to work around this compile
1583 Description: LASSERT on MDS when client holding flock lock dies
1584 Details : ldlm pool logic depends on number of granted locks equal to
1585 number of released locks which is not true for flock locks, so
1586 just exclude such locks from consideration.
1590 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1591 Details : Use fsfilt_sync() for both the whole filesystem sync and
1592 individual file sync to eliminate dangerous inode locking
1593 with I_LOCK that can lead to a deadlock.
1597 Description: performance in 1.6.3
1598 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1599 sure that 1MB requests are not fragmented by the block layer.
1601 Severity : enhancement
1603 Description: organize the server-side client stats on per-nid basis
1604 Details : Change the structure of stats under obdfilter and mds to
1612 The "uuid"s file would list the uuids of _active_ exports.
1613 And the clear entry is to clear all stats and stale nids.
1615 Severity : enhancement
1617 Description: eliminate client locks in face of contention
1618 Details : file contention detection and lockless i/o implementation
1619 for contended files.
1623 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1626 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1628 Description: MDS may lose file striping (and hence file data) in some cases
1629 Details : If there are additional extended attributes stored on the MDS,
1630 in particular ACLs, SELinux, or user attributes (if user_xattr
1631 is specified for the client mount options) then there is a risk
1632 of attribute loss. Additionally, the Lustre file striping
1633 needs to be larger than default (e.g. striped over all OSTs),
1634 and an additional attribute must be stored initially in the
1635 inode and then increase in size enough to be moved to the
1636 external attribute block (e.g. ACL growing in size) for file
1641 Description: skiplist implementation simplification
1642 Details : skiplists are used to group compatible locks on granted list
1643 that was implemented as tracking first and last lock of each
1644 lock group the patch changes that to using doubly linked lists
1648 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1649 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1650 because LASSERT on both the data supplied by a client, and the
1651 data on disk is dangerous and incorrect.
1653 Severity : enhancement
1655 Description: Slow truncate/writes to huge files at high offsets.
1656 Details : Directly associate cached pages to lock that protect those pages,
1657 this allows us to quickly find what pages to write and remove
1658 once lock callback is received.
1662 Description: more ldlm soft lockups
1663 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1664 starve other threads from the resource lock for a long time in
1665 case of long waiting queue, so change the debug level from
1666 D_OTHER to the less frequently used D_INFO.
1668 Severity : enhancement
1670 Description: add -gid, -group, -uid, -user options to lfs find
1674 Description: Hung threads in invalidate_inode_pages2_range
1675 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1676 once one is completed. As a result, some RPCs are stuck in the
1677 queue and are never sent.
1681 Description: filter threads hungs on waiting journal commit
1682 Details : Cleanup filter group llog code, then only filter group llog will
1683 be only created in the MDS/OST syncing process.
1687 Description: Procfs and llog threads access destoryed import sometimes.
1688 Details : Sync the import destoryed process with procfs and llog threads by
1689 the import refcount and semaphore.
1691 Severity : enhancement
1693 Description: openlock cache of b1_6 port to HEAD
1698 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1699 Details : If insertion of an extent fails, then discard the inode
1700 preallocation and free data blocks else it can lead to duplicate
1705 Description: don't always update ctime in ext3_xattr_set_handle()
1706 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1707 In some cases the ctime should not be updated, for example for
1708 2.0->1.8 compatibility it is necessary to delete an xattr and it
1709 should not update the ctime.
1713 Bugzilla : 15713/16362
1714 Description: Assertion in iopen_connect_dentry in 1.6.3
1715 Details : looking up an inode via iopen with the wrong generation number can
1716 populate the dcache with a disconneced dentry while the inode
1717 number is in the process of being reallocated. This causes an
1718 assertion failure in iopen since the inode's dentry list contains
1719 both a connected and disconnected dentry.
1723 Description: assertion failure in ldlm_handle2lock()
1724 Details : fix a race between class_handle_unhash() and class_handle2object()
1725 introduced in lustre 1.6.5 by bug 13622.
1730 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1731 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1736 Description: oops in sd_iostats_seq_show()
1737 Details : unloading/reloading the scsi low level driver triggers a kernel
1738 bug when trying to access the sd iostat file.
1743 Description: Kernel panics during QLogic driver reload
1744 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1745 patch, causing memory corruption.
1750 Description: journal_dev option does not work in b1_6
1751 Details : pass mount option during pre-mount.
1753 Severity : enhancement
1755 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1756 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1757 extent information of a file. It can be used to map logical blocks
1758 in a file to physical blocks in the block device.
1762 Description: LDLM soft lockups - improvement
1763 Details : It is be possible to send the lock handle along with each read
1764 or write request because the client is already doing a lock match
1765 itself so there isn't any reason the OST should have to re-do that
1769 Frequency : only on Cray X2
1771 Description: X2 build failures
1772 Details : fix build failures on Cray X2.
1776 Description: xid & resent requests
1777 Details : Initialize RPC XID from clock at startup (randomly if clock is
1780 Severity : enhancement
1782 Description: Add lustre_start utility to start or stop multiple Lustre servers
1787 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1788 Details : In case of memory pressure, list_del() can be called twice on
1789 req->rq_history_list, causing a kernel oops.
1793 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1794 Details : fix stack overflow in the distributed lock manager by defering
1795 export eviction after a failed AST to the elt thread instead of
1796 handling it in the dlm interpret routine.
1800 Description: Convert some comments to new format.
1801 Details : Update documenting comments to match doxygen conventions.
1805 Description: Grammar fixes.
1806 Details : A couple of trivial spelling fixes.
1810 Description: OSD_COUNTERS-mandatory
1811 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1812 to simplify the code.
1816 Description: simplify cmm_device freeing logic.
1817 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1821 Description: Add lockdep support to dt_object_operations locking interface.
1822 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1823 parameter indicating lock ordering. Update mdd code to use new
1828 Description: Introduce failloc constants for lockless IO tests.
1829 Details : Add two new failloc constants to test lockless IO. Only one of
1830 them in implemented---another is checked in yet to be landed
1835 Description: Add lockdep support for inode mutex.
1836 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1837 in the situations where Lustre has to lock more than one inode
1842 Description: Add optional invariants checking support.
1843 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1844 configure switch. This macro is to be used for consistency and
1845 sanity checks that are too expensive to be left in `production'
1850 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1851 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1856 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1857 Details : Use spin_lock_nested() in (the only) situation where more than
1858 one ldlm_lock is locked simultaneously. Also, fix possible
1859 dead-lock in ldlm_lock_change_resource() by enforcing particular
1864 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1865 Details : Instead of specifying each ldlm_lock call-back through separate
1866 parameter, wrap them into struct ldlm_callback_suite.
1870 Description: Kill join_lru obd method and its callers.
1871 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1872 in memory---a requirement implemented through ->o_join_lru() obd
1873 method in HEAD. Remove this method and its users.
1877 Description: Add asynchronous ldlm ENQUEUE completion handler.
1878 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1879 case that stock ldlm_completion_ast() cannot handle as it waits
1880 until lock is granted. Introduce new ldlm_completion_ast_async()
1881 for this. Also comment ldlm_completion_ast().
1885 Description: ldlm_error <-> errno conversion.
1886 Details : Add functions to map (rather arbitrary) between LDLM error codes
1887 and standard errno values. CLIO needs this to prevent LDLM specific
1888 constants from escaping ldlm and osc.
1892 Description: Kill unused ldlm_handle2lock_ns() function.
1893 Details : Kill unused ldlm_handle2lock_ns() function.
1897 Description: Add lu_ref support to ldlm_lock
1898 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1899 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1900 and ldlm_resource. LDLM interface has to be changed, because code that
1901 releases a reference on a lock, has to "know" what reference this is.
1902 In the most frequent case
1904 lock = ldlm_handle2lock(handle);
1906 LDLM_LOCK_PUT(lock);
1908 no changes are required. When any other reference (received _not_ from
1909 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1910 instead of LDLM_LOCK_PUT().
1912 Arguably, changes are pervasive, and interface requires some discipline
1913 for proper use. On the other hand, it was very instrumental in finding
1914 a few leaked lock references.
1918 Description: Add ldlm_lock_addref_try().
1919 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1920 attempts to addref a lock that might be being canceled
1925 Description: Add ldlm_weigh_callback().
1926 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1927 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1928 hard-coded `number of pages' logic.
1932 Description: Add lockdep annotations to llog code.
1933 Details : Use appropriately tagged _nested() locking calls in the places
1934 where llog takes more than one ->lgh_lock lock.
1938 Description: Add loi_kms_set().
1939 Details : Wrap kms updates into a helper function.
1943 Description: Constify instances of struct lsm_operations.
1944 Details : Constify instances of struct lsm_operations.
1948 Description: lu_conf support.
1949 Details : On a server, a file system object is uniquely identified
1950 by a fid, which is sufficient to locate and load all object
1951 state (inode). On a client, on the other hand, more data are
1952 necessary instantiate an object. Change lu_object_find() and
1953 friends to take additional `lu_conf' argument describing object.
1954 Typically this includes layout information.
1958 Description: lu_context fixes.
1959 Details : Introduce new lu_context functions that are needed on the client
1960 side, where some system threads (ptlrpcd) are shared by multiple
1961 modules, and so cannot be stopped during module shutdown.
1965 Description: Add start and stop methods to lu_device_type_operations.
1966 Details : Introduce two new methods in lu_device_type_operations, that are
1967 invoked when first instance of a given type is created and last one
1968 is destroyed respectively. This is need by CLIO.
1972 Description: Add lu_ref support to struct lu_device.
1973 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1974 track leaked references.
1978 Description: Introduce lu_kmem_descr.
1979 Details : lu_kmem_descr and its companion interface allow to create
1980 and destroy a number of kmem caches at once.
1984 Description: Fix lu_object finalization race.
1985 Details : Fix a race between lu_object_find() finding an object and its
1986 concurrent finalization. This race is (most likely) not possible
1987 on the server, but might happen on the client.
1991 Description: Introduce lu_ref interface.
1992 Details : lu_ref is a debugging module allowing to track references to
1993 a given object. It is quite cpu expensive, and has to be
1994 explicitly enabled with --enable-lu_ref. See usage description
1999 Description: Factor lu_site procfs stats into a separate function.
2000 Details : Separate lu_site stats printing code into a separate function
2001 that can be reused on a client.
2005 Description: Constify instances of struct {lu,dt,md}_device_operations.
2006 Details : Constify instances of struct {lu,dt,md}_device_operations.
2010 Description: Introduce struct md_site and move meta-data specific parts of
2011 struct lu_site here.
2012 Details : Move md-specific fields out of struct lu_site into special struct
2013 md_site, so that lu_site can be used on a client.
2017 Description: Kill mdd_lov_destroy().
2018 Details : Remove unused mdd code.
2022 Description: Add st_block checking to multistat.c.
2023 Details : Add st_block checking to multistat.c.
2027 Description: Add lu_ref support to struct obd_device.
2028 Details : Add lu_ref tracking to obd_device.
2032 Description: Kill obd_set_fail_loc().
2033 Details : Remove unused code.
2037 Description: Add special type for ptlrpc_request interpret functions.
2038 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2039 there. Actual usage will be in CLIO.
2043 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2044 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2045 doesn't work with lockdep.
2049 Description: Add rwv.c test program.
2050 Details : New testing program exercising readv(2) and writev(2) (Qian).
2054 Description: Add sendfile.c test program.
2055 Details : New testing program exercising sendfile(2) (Jay).
2059 Description: Ratelimit a message that can be very frequent.
2060 Details : Ratelimit a memory allocation failure message that can
2065 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2066 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2070 Description: Liblustre build fixes.
2071 Details : Liblustre build fixes.
2075 Description: libcfs: add cfs_{need,cond}_resched() interface.
2076 Details : libcfs: add cfs_{need,cond}_resched() definition and
2077 implementations for Linux, NT, and liblustre.
2079 Severity : enhancement
2081 Description: More exported tunables for mballoc
2082 Details : Add support for tunable preallocation window and new tunables for
2083 large/small requests
2087 Description: Detect corruption of block bitmap and checking for preallocations
2088 Details : Checks validity of on-disk block bitmap. Also it does better
2089 checking of number of applied preallocations. When corruption is
2090 found, it turns filesystem readonly to prevent further corruptions.
2094 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2095 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2096 window to 0. If it does hit cache miss, and read-pattern is still
2097 stride-io mode, does not reset the stride window, but also does not
2098 increase the stride window length in this case.
2102 Frequency : only for big-endian servers
2103 Description: Check if system is big-endian while mounting fs with extents feature
2104 Details : Mounting a filesystem with extents feature will fail on big-endian
2105 systems since ext3-based ldiskfs is not supported on big-endian
2106 systems. This can be over-riden with "bigendian_extents" mount option.
2108 Severity : enhancement
2110 Description: The root squash functionality
2111 Details : A security feature, which is to prevent users from being able
2112 to mount lustre on their desktop, run as root, and delete
2113 all of the files in the filesystem. The goal is accomplished by
2114 remapping user id (UID) and group id (GID) of the root user to
2115 a UID and GID specified by the system administartor via Lustre
2116 configuration management server (MGS). The functionality also
2117 allows to specify sets of clients for which the remapping does
2122 Description: Excessive recovery window
2123 Details : With AT enabled, the recovery window can be excessively long (6000+
2124 seconds). To address this problem, we no longer use
2125 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2126 timeout no longer depends on the service time, it is set to
2127 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2128 time via pb_service_time.
2132 Description: Watchdog triggered on MDS failover
2133 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2134 the OSTs know that the MDS "UUID" can be reused for the same export
2135 from a different NID, so we do not need to wait for the export to be
2139 Frequency : rare, only if using MMP with Linux RAID
2141 Description: MMP doesn't work with Linux RAID
2142 Details : While using HA for Lustre servers with Linux RAID, it is possible
2143 that MMP will not detect multiple mounts. To make this work we
2144 need to unplug the device queue in RAID when the MMP block is being
2145 written. Also while reading the MMP block, we should read it from
2146 disk and not the cached one.
2148 Severity : enhancement
2150 Description: open file using fid
2151 Details : A file can be opened using just its fid, like
2152 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2155 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2158 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2159 Details : add proper locking for creation of jbd_4k slab cache
2163 Description: MMP check in ext3_remount() fails without displaying any error
2164 Details : When multiple mount protection fails during remount, proper error
2167 Severity : enhancement
2169 Description: Allow stripe size to be up to 4G-64k
2170 Details : Fix math logic to allow large stripe sizes.
2174 Description: add check for >8TB ldiskfs filesystems
2175 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2176 Don't allow >8TB ldiskfs filesystems to be mounted without
2177 force_over_8tb mount option
2182 Descriptoin: OSS complains "too many blocks" and crashes
2183 Details : In filter_commitrw_write():
2184 1. lquota_chkquota();
2185 2. obdo_from_inode();
2186 3. lquota_pending_commit();
2187 "2." has changed uid and gid in this case, then caused
2188 lquota_chkquota() and lquota_pending_commit() to use
2189 different uid and gid. That is the root of the bug.
2191 --------------------------------------------------------------------------------
2193 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2195 * Support for kernels:
2196 2.6.5-7.283 (SLES 9),
2197 2.6.9-55.EL (RHEL 4),
2198 2.6.16.46-0.14 (SLES 10),
2199 2.6.18.8 vanilla (kernel.org)
2200 * Client support for unpatched kernels:
2201 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2202 2.6.16 - 2.6.22 vanilla (kernel.org)
2203 * Due to recently discovered recovery problems, we do not recommend
2204 using patchless RHEL 4 clients with this or any earlier release.
2205 * Recommended e2fsprogs version: 1.39.cfs8
2206 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2207 * Starting with this release, the ldiskfs backing filesystem required
2208 by Lustre is now in its own package, lustre-ldiskfs. This package
2209 should be installed. It is versioned separately from Lustre and
2210 may be released separately in future.
2214 Description: block reactivating mgc import until all deactivates complete
2215 Details : Fix race when failing back MDT/MGS to itself (testing)
2218 Frequency : at statup only
2220 Description: mds_lov_synchronize race leads to various problems
2221 Details : simultaneous MDT->OST connections at startup can cause the
2222 sync to abort, leaving the OSC in a bad state.
2224 Severity : enhancement
2226 Description: add optional extra BUILD_VERSION info
2227 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2228 to override the lustre version.
2231 Frequency : 2.6.18 servers only
2233 Description: ll_kern_mount() doesn't release the module reference
2234 Details : The ldiskfs module reference count never drops down to 0
2235 because ll_kern_mount() doesn't release the module reference.
2240 Description: server LBUG when using old ost_num_threads parameter
2241 Details : Accept the old ost_num_threads parameter but warn that it
2242 is deprecated, and fix an off-by-one error that caused an LBUG.
2247 Description: Transient SCSI error results in persistent IO issue
2248 Details : iobuf->dr_error is not reinitialized to 0 between two
2252 Frequency : sometimes when underlying device returns I/O errors
2254 Description: OSTs not going read-only during write failures
2255 Details : OSTs are not remounted read-only when the journal commit threads
2256 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2257 instead of the ext3 wrappers.
2261 Description: poor connect scaling with increasing client count
2262 Details : Don't run filter_grant_sanity_check for more than 100 exports
2263 to improve scaling for large numbers of clients.
2266 Frequency : SLES10 only
2268 Description: sanity-quota.sh quotacheck failed: rc = -22
2269 Details : Quotas cannot be enabled on SLES10.
2272 Frequency : liblustre clients only
2274 Description: getdirentries does not give error when run on compute nodes
2275 Details : getdirentries does not fail when the size specified as an argument
2276 is too small to contain at least one entry
2278 Severity : enhancement
2280 Description: Add LNET router traceability for debug purposes
2281 Details : If a checksum failure occurs with a router as part of the
2282 IO path, the NID of the last router that forwarded the bulk data
2283 is printed so it can be identified.
2288 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2289 Details : Due to a race condition, liblustre clients were occasionally
2290 evicted incorrectly.
2292 Severity : enhancement
2294 Description: lfs setstripe use optional parameters instead of postional
2297 Severity : enhancement
2299 Description: Nanosecond timestamp support for ldiskfs
2300 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2301 resolution timestamps. There is not yet support for this at
2302 the Lustre filesystem level.
2305 Frequency : during server recovery
2307 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2308 Details : request with rq_no_resend flag not awake l_wait_event if they get
2312 Frequency : nfs export on patchless client
2314 Description: connectathon hang when test nfs export over patchless client
2315 Details : Disconnected dentry cannot be found with lookup, so we do not need
2316 to unhash it or make it invalid
2319 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2320 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2321 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2326 Description: open req refcounting wrong on reconnect
2327 Details : If reconnect happened between getting open reply from server and
2328 call to mdc_set_replay_data in ll_file_open, we will schedule
2329 replay for unreferenced request that we are about to free.
2330 Subsequent close will crash in variety of ways.
2331 Check that request is still eligible for replay in
2332 mdc_set_replay_data().
2337 Description: disable writes to filesystem when reading health_check file
2338 Details : the default for reading the health_check proc file has changed
2339 to NOT do a journal transaction and write to disk, because this
2340 can cause reads of the /proc file to hang and block HA state
2341 checking on a healthy but otherwise heavily loaded system. It
2342 is possible to return to the previous behaviour during configure
2343 with --enable-health-write.
2345 Severity : enhancement
2347 Description: 64-bit inode version
2348 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2349 to the inode. This will be required for version-based recovery.
2354 Description: MDS fails to start if a duplicate client export is detected
2355 Details : in some rare cases it was possible for a client to connect to
2356 an MDS multiple times. Upon recovery the MDS would detect this
2357 and fail during startup. Handle this more gracefully.
2359 Severity : enhancement
2361 Description: Add -o localflock option to simulate old noflock
2363 Details : This will achieve local-only flock/fcntl locks
2369 Description: log_commit_thread vs filter_destroy race leads to crash
2370 Details : Take import reference before releasing llog record semaphore
2375 Description: Wrong request locking in request set processing
2376 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2377 imp_delayed_list, in this place should be used imp_lock.
2380 Frequency : when reconnection
2382 Description: Grant Leak when osc reconnect to OST
2383 Details : When osc reconnect ost, OST(filter) should check whether it
2384 should grant more space to client by comparing fed_grant and
2385 cl_avail_grant, and return the granted space to client instead
2386 of "new granted" space, because client will call osc_init_grant
2387 to update the client grant space info.
2390 Frequency : when client reconnect to OST
2392 Description: Grant Leak when osc do resend and replay bulk write
2393 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2394 bulk write request, because the grant info while be sync between
2395 OSC and OST when reconnect, and we should ignore the grant info
2396 these of resend/replay write req.
2401 Description: Grant space more than avaiable left space sometimes.
2402 Details : When then OST is about to be full, if two bulk writing from
2403 different clients came to OST. Accord the avaliable space of the
2404 OST, the first req should be permitted, and the second one
2405 should be denied by ENOSPC. But if the seconde arrived before
2406 the first one is commited. The OST might wrongly permit second
2407 writing, which will cause grant space > avaiable space.
2410 Frequency : when client is evicted
2412 Description: Grant might be wrongly erased when osc is evicted by OST
2413 Details : when the import is evicted by server, it will fork another
2414 thread ptlrpc_invalidate_import_thread to invalidate the
2415 import, where the grant will be set to 0. While the original
2416 thread will update the grant it got when connecting. So if
2417 the former happened latter, the grant will be wrongly errased
2418 because of this race.
2423 Description: Checking Stale with correct fid
2424 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2425 to check whether it is stale, because sometimes, we want the
2426 enqueue happened anyway, and op_data.fid2 will not be initialized.
2428 Severity : enhancement
2430 Description: update patchless client
2431 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2434 Frequency : only with 2.4 kernel
2436 Description: random memory corruption
2437 Details : size of struct ll_inode_info is to big for union inode.u and this
2438 can be cause of random memory corruption.
2443 Description: Memory leak in recovery
2444 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2445 It should also check obd_fail before fsfilt_start, otherwise if
2446 fsfilt_start return -EROFS,(failover mds during mds recovery).
2447 then the req will return with repmsg->transno = 0 and rc = EROFS.
2448 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2449 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2450 be freed no matter whether fsfilt_commit success or not.
2453 Frequency : only with huge count clients
2455 Description: Prevents from taking the superblock lock in llap_from_page for
2457 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2458 for llap_from_page call in ll_removepage prevents from taking the
2459 superblock lock for a soon died page.
2464 Description: Not check open intent error before release open handle
2465 Details : in some rare cases, the open intent error is not checked before
2466 release open handle, which may cause
2467 ASSERTION(open_req->rq_transno != 0), because it tries to release
2468 the failed open handle.
2473 Description: Set cat log bitmap only after create log success.
2474 Details : in some rare cases, the cat log bitmap is set too early. and it
2475 should be set only after create log success.
2479 Description: Accessing a block bevice can re-enable I/O when Lustre is
2480 tearing down a device.
2481 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2486 Description: service threads may hog cpus when there are a lot of requests
2487 Details : Insert cond_resched to give other threads a chance to use some CPU
2492 Description: the cat log was not initialized in recovery
2493 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2494 unlink log on mds will not be initialized until mds post
2495 recovery. And also in mds post recovery, the unlink log will
2496 initialization will be done asynchronausly, so there will be race
2497 between add unlink log and unlink log initialization.
2501 Description: brw_stats were being printed incorrectly
2502 Details : brw_stats were being printed as log2 but all of them were not
2503 recorded as log2. Also remove some code duplication arising from
2504 filter_tally_{read,write}.
2508 Frequency : rare, only in recovery.
2509 Description: ASSERTION(req->rq_type != LI_POISON) failed
2510 Details : imp_lock should be held while iterating over imp_sending_list for
2511 prevent destroy request after get timeout in ptlrpc_queue_wait.
2515 Description: replay-single.sh test 52 fails
2516 Details : A lock's skiplist need to be cleanup when it being unlinked
2517 from its resource list.
2521 Description: Short directio read returns full requested size rather than
2523 Details : Direct I/O operations should return actual amount of bytes
2524 transferred rather than requested size.
2526 Severity : enhancement
2528 Description: metadata RPC reduction (e.g. for rm performance)
2529 Details : decrease the amount of synchronous RPC between clients and servers
2530 by canceling conflicing lock before the operation on the client
2531 and packing thier handles into the main operation RPC to server.
2533 Severity : enhancement
2535 Description: Async OSC create to avoid the blocking unnecessarily.
2536 Details : If a OST has no remain object, system will block on the creating
2537 when need to create a new object on this OST. Now, ways use
2538 pre-created objects when available, instead of blocking on an
2539 empty osc while others are not empty. If we must block, we block
2540 for the shortest possible period of time.
2544 Description: improve handling recoverable errors
2545 Details : if request processig with error which can be recoverable on server
2546 request should be resend, otherwise page released from cache and
2549 Severity : enhancement
2551 Description: refine locking for avoid write wrong info into lov_objid file
2552 Details : fix possible races with add new target and write/update data in
2555 --------------------------------------------------------------------------------
2557 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2562 Frequency : on some architectures
2564 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2565 Details : Uninitialized flags sometimes cause configuration commands to
2569 Frequency : patchless clients only
2571 Description: missing __iget() symbol export
2572 Details : The __iget() symbol export is missing. To avoid the need for
2573 this on patchless clients the deathrow inode reaper is turned
2574 off, and we depend on the VM to clean up old inodes. This
2575 dependency was during via the fix for bug 12181.
2577 --------------------------------------------------------------------------------
2579 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2581 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2582 INTEROPERATE with older versions automatically. In many cases a
2583 special upgrade step is needed. Please read the
2584 user documentation before upgrading any part of a live system.
2585 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2586 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2587 * WARNING: Lustre configuration and startup changes are required with
2588 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2592 Severity : enhancement
2594 Description: MountConf
2595 Details : Lustre configuration is now managed via mkfs and mount
2596 commands instead of lmc and lconf. New obd types (MGS, MGC)
2597 are added for dynamic configuration management. See
2598 https://mail.clusterfs.com/wikis/lustre/MountConf for
2601 Severity : enhancement
2603 Description: dynamic OST addition
2604 Details : OSTs can now be added to a live filesystem
2606 Severity : enhancement
2608 Description: startup order invariance
2609 Details : MDTs and OSTs can be started in any order. Clients only
2610 require the MDT to complete startup.
2612 Severity : enhancement
2614 Description: parallel, asynchronous orphan cleanup
2615 Details : orphan cleanup is now performed in separate threads for each
2616 OST, allowing parallel non-blocking operation.
2618 Severity : enhancement
2620 Description: optimized stripe assignment
2621 Details : stripe assignments are now made based on ost space available,
2622 ost previous usage, and OSS previous usage, in order to try
2623 to optimize storage space and networking resources.
2625 Severity : enhancement
2627 Description: Permanently set tunables
2628 Details : All writable /proc/fs/lustre tunables can now be permanently
2629 set on a per-server basis, at mkfs time or on a live system.
2631 Severity : enhancement
2633 Description: Lustre message v2
2634 Details : Add lustre message format v2.
2636 Severity : enhancement
2638 Description: client OST exclusion list
2639 Details : Clients can be started with a list of OSTs that should be
2640 declared "inactive" for known non-responsive OSTs.
2644 Description: SPEC SFS validation failure on NFS v2 over lustre.
2645 Details : Changes the blocksize for regular files to be 2x RPC size,
2646 and not depend on stripe size.
2648 Severity : enhancement
2650 Description: Multiple MD RPCs in flight.
2651 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2652 To avoid overly-overloading MDS, introduce a limit on number of
2653 MDS RPCs in flight for a single client and add /proc controls
2654 to adjust this limit.
2656 Severity : enhancement
2658 Description: client read/write statistics
2659 Details : Add client read/write call usage stats for performance
2660 analysis of user processes.
2661 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2662 file access. extents_stats shows chunk size distribution.
2663 extents_stats_per_process show chunk size distribution per
2666 Severity : enhancement
2668 Description: mds statistics
2669 Details : Add detailed mds operations statistics in
2670 /proc/fs/lustre/mds/*/stats.
2674 Description: Failure of copying files with lustre special EAs.
2675 Details : Client side always return success for setxattr call for lustre
2676 special xattr (currently only "trusted.lov").
2681 Description: Refcount LNET uuids
2682 Details : The global LNET uuid list grew linearly with every startup;
2683 refcount repeated list entries instead of always adding to
2686 Severity : enhancement
2688 Description: Dynamic service threads
2689 Details : Within a small range, start extra service threads
2690 automatically when the request queue builds up.
2693 Frequency : mixed-endian client/server environments
2695 Description: mixed-endian crashes
2696 Details : The new msg_v2 system had some failures in mixed-endian
2699 Severity : enhancement
2701 Description: Easy OST removal
2702 Details : OSTs can be permanently deactivated with e.g. 'lctl
2703 conf_param lustre-OST0001.osc.active=0'
2705 Severity : enhancement
2707 Description: MGS proc entries
2708 Details : Added basic proc entries for the MGS showing what filesystems
2711 Severity : enhancement
2713 Description: provide MGS failover
2714 Details : Added config lock reacquisition after MGS server failover.
2716 Severity : enhancement
2718 Description: add Linux 2.4 support
2719 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2723 Description: a large application tries to do I/O to the same resource and dies
2724 in the middle of it.
2725 Details : Check the req->rq_arrival time after the call to
2726 ost_brw_lock_get(), but before we do anything about
2727 processing it & sending the BULK transfer request. This
2728 should help move old stale pending locks off the queue as
2729 quickly as obd_timeout.
2732 Frequency : when an incorrect nid is specified during startup
2734 Description: ptlrpc connect to non-existant node causes kernel crash
2735 Details : LNET can't be re-entered from an event callback, which
2736 happened when we expire a message after the export has been
2737 cleaned up. Instead, hand the zombie cleanup off to another
2740 Severity : enhancement
2742 Description: plain/inodebits lock performance improvement
2743 Details : Grouping plain/inodebits in granted list by their request modes
2744 and bits policy, thus improving the performance of search through
2748 Frequency : only if OST filesystem is corrupted
2750 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2751 Details : for a short time RPCs with bulk IO are in the replay list,
2752 but replay of bulk IOs is unimplemented. If the OST filesystem
2753 is corrupted due to disk cache incoherency and then replay is
2754 started it is possible to trip an assertion. Avoid putting
2755 committed RPCs into the replay list at all to avoid this issue.
2758 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2760 Description: System hang on startup
2761 Details : This bug allowed the liblustre (e.g. catamount) client to
2762 return to the app before handling all startup RPCs. This
2763 could leave the node unresponsive to lustre network traffic
2764 and manifested as a server ptllnd timeout.
2766 Severity : enhancement
2768 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2769 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2770 boolean to control whether to print peer debug info when a
2771 client's RPC times out.
2774 Frequency : only for kernels with patches from Lustre below 1.4.3
2776 Description: Remove old rdonly API
2777 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2780 Frequency : only for devices with external journals
2782 Description: Set external device read-only also
2783 Details : During a commanded failover stop, we set the disk device
2784 read-only while the server shuts down. We now also set any
2785 external journal device read-only at the same time.
2788 Frequency : when upgrading from 1.4 while trying to change parameters
2790 Description: The wrong (new) MDC name was used when setting parameters for
2791 upgraded MDT's. Also allows changing of OSC (and MDC)
2792 parameters if --writeconf is specified at tunefs upgrade time.
2795 Frequency : when setting specific ost indicies
2797 Description: QOS code breaks on skipped indicies
2798 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2799 created with --index need not be sequential.
2801 Severity : enhancement
2803 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2804 Details : The uninit_groups feature works in conjunction with the kernel
2805 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2806 up the pass1 processing of e2fsck. This is a read-only feature
2807 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2808 mount filesystems that have had this feature enabled.
2810 Severity : enhancement
2812 Description: Improve multi-block allocation algorithm to avoid fragmentation
2813 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2814 allocation locality and avoid filesystem fragmentation.
2816 ------------------------------------------------------------------------------
2818 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2820 * Support for kernels:
2821 2.4.21-47.0.1.EL (RHEL 3)
2822 2.6.5-7.283 (SLES 9)
2823 2.6.9-42.0.10.EL (RHEL 4)
2824 2.6.12.6 vanilla (kernel.org)
2825 2.6.16.27-0.9 (SLES 10)
2826 * Recommended e2fsprogs version: 1.39.cfs5
2828 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2832 Frequency : occasional, depends on client load and configuration
2833 Bugzilla : 12181, 12203
2834 Description: data loss for recently-modified files
2836 Details : In some cases it is possible that recently written or created
2837 files may not be written to disk in a timely manner (this should
2838 normally be within 30s unless client IO load is very high).
2839 The problem appears as zero-length files or files that are a
2840 multiple of 1MB in size after a client crash or client eviction
2841 that are missing data at the end of the file.
2843 This problem is more likely to be hit on clients where files are
2844 repeatedly created and unlinked in the same directory, clients
2845 have a large amount of RAM, have many CPUs, the filesystem has
2846 many OSTs, the clients are rebooted frequently, and/or the files
2847 are not accessed by other nodes after being written.
2849 The presence of the problem can be detected by looking at
2850 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2851 smaller than the second (nr_unused) then dirty files will not
2852 be flushed automatically to disk. "sync; sleep 10" should be
2853 run several times on the node before unmounting it to update
2854 Lustre (this is also safe to run on nodes without this problem).
2856 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2857 that can cause this same problem, so customers using that kernel
2858 also need to update the kernel in addition to Lustre. In order
2859 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2861 It is normal that files written just before a client crash (less
2862 than 30s) may not yet have been flushed to disk, even for local
2866 Frequency : frequent on thin XT3 nodes
2868 Description: UUID collision on thin XT3 Linux nodes
2869 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2870 randomly, since we relied on an insufficiently-seeded PRNG.
2875 Description: OSS hangs after "All ost request buffers busy"
2876 Details : A deadlock between quota and journal operations caused OSS
2877 hangs after printing "All ost request buffers busy."
2880 Frequency : always on liblustre builds
2882 Description: Cleanup compiler warnings on liblustre
2885 Frequency : always on liblustre builds on XT3
2887 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2891 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2892 > 32000 subdirectories
2893 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2894 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2895 subdirectory count crosses 32000. This will aid e2fsck to
2896 correctly handle more than 32000 subdirectories.
2899 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2901 Description: System hang on startup
2902 Details : This bug allowed the liblustre (e.g. catamount) client to
2903 return to the app before handling all startup RPCs. This
2904 could leave the node unresponsive to lustre network traffic
2905 and manifested as a server ptllnd timeout.
2907 Severity : enhancement
2909 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2910 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2911 boolean to control whether to print peer debug info when a
2912 client's RPC times out.
2917 Description: make O_SYNC working on 2.6 kernels
2918 Details : 2.6 kernels use different method for mark pages for write,
2919 so need add a code to lustre for O_SYNC work.
2924 Description: Failure to close file and release space on NFS
2925 Details : Put inode details into lock acquired in ll_intent_file_open.
2926 Use mdc_intent_lock in ll_intent_open to properly
2927 detect all kind of errors unhandled by mdc_enqueue.
2932 Description: proc file read during shutdown sometimes raced obd removal,
2934 Details : Add lock to prevent obd access after proc file removal.
2937 Frequency : Only for files larger than 4GB on 32-bit clients.
2939 Description: improperly doing page alignment of locks
2940 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2941 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2946 Description: RPCs being resent when they shouldn't be
2947 Details : Some RPCs that should not be resent are being resent. This
2948 can cause inconsistencies in the RPC state machine. Do not
2949 resend such requests.
2952 Frequency : rare, only with NFS export
2954 Description: Crash on NFS re-export node
2955 Details : under very unusual load conditions an assertion is hit in
2956 ll_intent_file_open()
2959 Frequency : only if OST filesystem is corrupted
2961 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2962 Details : for a short time RPCs with bulk IO are in the replay list,
2963 but replay of bulk IOs is unimplemented. If the OST filesystem
2964 is corrupted due to disk cache incoherency and then replay is
2965 started it is possible to trip an assertion. Avoid putting
2966 committed RPCs into the replay list at all to avoid this issue.
2971 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2972 Details : Large single O_DIRECT read and write calls can fail to allocate
2973 a sufficiently large buffer to process the request. In case of
2974 allocation failure the allocation is retried with a smaller
2975 buffer and broken into smaller requests.
2977 Severity : enhancement
2979 Description: Add -o localflock option to simulate old noflock behaviour.
2980 Details : This will achieve local-only flock/fcntl locks coherentness.
2985 Description: versioning check is incomplete
2986 Details : Checking the version difference of client vs. server, report
2987 error if the gap is too big.
2992 Description: add support PG_writeback bit
2993 Details : add support for PG_writeback bit for Lustre, for more carefull
2994 work with page cache in 2.6 kernel. This also fix some deadlocks
2995 and remove hack for work O_SYNC with 2.6 kernel.
2997 Severity : enhancement
2999 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3000 Details : The uninit_groups feature works in conjunction with the kernel
3001 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3002 up the pass1 processing of e2fsck. This is a read-only feature
3003 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3004 mount filesystems that have had this feature enabled.
3006 Severity : enhancement
3008 Description: Improve multi-block allocation algorithm to avoid fragmentation
3009 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3010 allocation locality and avoid filesystem fragmentation.
3012 ------------------------------------------------------------------------------
3014 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3016 * Support for kernels:
3017 2.6.9-42.0.3.EL (RHEL 4)
3018 2.6.5-7.276 (SLES 9)
3019 2.4.21-47.0.1.EL (RHEL 3)
3020 2.6.12.6 vanilla (kernel.org)
3021 2.6.16.21-0.8 (SLES10)
3022 * Recommended e2fsprogs version: 1.39.cfs2-0
3024 * The backwards-compatible /proc/sys/portals symlink has been removed
3025 in this release. Before upgrading, please ensure that you change
3026 any configuration scripts or /etc/sysctl.conf files that access
3027 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3028 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3029 in advance of the upgrade on any system running Lustre 1.4.6 or
3030 newer, since /proc/sys/lnet was added in that version.
3031 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3035 Frequency : only when quota is used
3037 Description: avoid scanning export list for quota master
3038 Details : Change the algorithms to avoid scanning export list in order
3039 to improve the efficiency.
3042 Frequency : MDS failover only, very rarely
3044 Description: "went back in time" messages on mds failover
3045 Details : The greatest transno may be lost when the current operation
3046 finishes with an error (transno==0) and the client's last_rcvd
3047 record is over-written. Save the greatest transno in the
3048 mds_last_transno for this case.
3051 Frequency : always for specific kernels and striping counts
3053 Description: client may get "Matching packet too big" without ACL support
3054 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3055 when trying to access files in certain configurations. The
3056 clients should in fact be denied when mounting because they do
3057 not understand ACLs.
3060 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3062 Description: many clients connecting with IO in progress causes connect timeouts
3063 Details : Avoid synchronous journal commits to avoid delays caused by many
3064 clients connecting/disconnecting when bulk IO is in progress.
3065 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3066 OST_IO_PORTAL to avoid delays behind potentially many pending
3070 Frequency : occasionally with multiple writers to a single file
3072 Description: shared writes to file may result in wrong size reported by stat()
3073 Details : Allow growing of kms when extent lock is cancelled
3076 Frequency : always with random mmap IO to multi-striped file
3078 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3079 Details : Only if the hole is at the end of OST object so that kms is too
3080 small. Fix is to increase kms accordingly in ll_nopage.
3083 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3085 Description: writes to a missing object would leak memory on the OST
3086 Details : If there is an inconsistency between the MDS and OST filesystems,
3087 such that the MDS references an object that doesn't exist, writes
3088 to that object will leak memory due to incorrect cleanup in the
3089 error handling path, eventually running out of memory on the OST.
3094 Description: Creating too long symlink causes lustre errors
3095 Details : Check symlink and name lengths before sending requests to MDS.
3098 Frequency : only if flock is enabled (not on by default)
3100 Description: posix locks not released on fd closure on 2.6.9+
3101 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3102 kernels, this caused such locks not to be released on fd close and
3103 then assertions on fs unmount about still used locks.
3106 Frequency : MDS failover only, very rarely
3108 Description: clients may get ASSERTION(granted_lock != NULL)
3109 Details : When request was taking a long time, and a client was resending
3110 a getattr by name lock request. The were multiple lock requests
3111 with the same client lock handle and
3112 mds_getattr_name->fixup_handle_for_resent_request found one of the
3113 lock handles but later failed with ASSERTION(granted_lock != NULL).
3118 Description: handle->h_buffer_credits > 0, assertion failure
3119 Details : h_buffer_credits is zero after truncate, causing assertion
3120 failure. This patch extends the transaction or creates a new
3124 Frequency : NFS re-export or patchless client
3125 Bugzilla : 11179, 10796
3126 Description: Crash on NFS re-export node (__d_move)
3127 Details : We do not want to hash the dentry if we don't have a lock.
3128 But if this dentry is later used in d_move, we'd hit uninitialised
3129 list head d_hash, so we just do this to init d_hash field but
3130 leave dentry unhashed.
3133 Frequency : NFS re-export or patchless client
3135 Description: NFS exports has problem with symbolic link
3136 Details : lustre client didn't properly install dentry when re-exported
3137 to NFS or running patchless client.
3140 Frequency : NFS re-export or patchless client
3142 Description: Various nfs/patchless fixes.
3143 Details : fixes reuse disconected alias for lookup process - this fixes
3144 warning "find_exported_dentry: npd != pd",
3145 fix permission error with open files at nfs.
3146 fix apply umask when do revalidate.
3149 Frequency : occasional
3151 Description: Crash on NFS re-export node
3152 Details : calling clear_page() on the wrong pointer triggered oops in
3153 generic_mapping_read().
3156 Frequency : rarely, using O_DIRECT IO
3158 Description: unaligned directio crashes client with LASSERT
3159 Details : check for unaligned buffers before trying any requests.
3162 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3164 Description: stale data returned from RAID cache
3165 Details : If only a small amount of IO is done to the RAID device before
3166 reading it again it is possible to get stale data from the RAID
3167 cache instead of reading it from disk.
3170 Frequency : always for sles10 kernel
3172 Description: sles10 support
3173 Details : ll_follow_link: compile fixes and using of nd_set_link
3174 under newer kernels.
3177 Frequency : depends on arch, kernel and compiler version, always on sles10
3180 Description: recursive or deep enough symlinks cause stack overflow
3181 Details : getting rid of large stack-allocated variable in
3185 Frequency : depends on hardware
3187 Description: lustre write performance loss in the SLES10 kernel
3188 Details : the performance loss is caused by using of write barriers in the
3189 ext3 code. The SLES10 kernel turns barrier support on by
3190 default. The fix is to undo that change for ldiskfs.
3192 ------------------------------------------------------------------------------
3194 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3196 * Support for kernels:
3197 2.6.9-42.0.3EL (RHEL 4)
3198 2.6.5-7.276 (SLES 9)
3199 2.4.21-47.0.1.EL (RHEL 3)
3200 2.6.12.6 vanilla (kernel.org)
3204 Frequency : quota enabled and large files being deleted
3206 Description: releasing more than 4GB of quota at once hangs OST
3207 Details : If a user deletes more than 4GB of files on a single OST it
3208 will cause the OST to spin in an infinite loop. Release
3209 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3214 Description: statfs data retrieved from /proc may be stale or zero
3215 Details : When reading per-device statfs data from /proc, in the
3216 {kbytes,files}_{total,free,avail} files, it may appear
3217 as zero or be out of date.
3220 Frequency : systems with MD RAID1 external journal devices
3222 Description: lconf's call to blkid is confused by RAID1 journal devices
3223 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3224 of returning all block devices that match the journal UUID.
3227 Frequency : always, for aggregate stripe size over 4GB
3229 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3230 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3231 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3234 Frequency : always if "lfs find" used on a local file/directory
3236 Description: "lfs find" segfaults if used on a local file/directory
3237 Details : The case where a directory component was not specified wasn't
3238 handled correctly. Handle this properly.
3241 Frequency : always on ppc64
3243 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3244 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3245 perform endianness conversion.
3248 Frequency : rarely (truncate to non-zero file size after write under load)
3249 Bugzilla : 10730, 10687
3250 Description: Files padded with zeros to next 4K multiple
3251 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3252 it is possible that files that are truncated to a non-zero size
3253 immediately after being written are filled with zero bytes beyond
3254 the truncated size. No file data is lost.
3256 Severity : enhancement
3258 Description: Allow recovery/failover for liblustre clients.
3259 Details : liblustre clients were unaware of failover configurations until
3262 Severity : enhancement
3264 Description: user file locks should fail when not mounting with flock option
3265 Details : Set up an error-returning stub in ll_file_operations.lock field
3266 to prevent incorrect behaviour when client is mounted without
3267 flock option. Also, set up properly f_op->flock field for
3271 Frequency : always on ia64
3273 Description: "lfs df" loops on printing out MDS statfs information
3274 Details : The obd_ioctl_data was not initialized and in some systems
3275 this caused a failure during the ioctl that did not return
3276 an error. Initialize the struct and return an error on failure.
3279 Frequency : SLES 9 only
3281 Description: Error of copying files with lustre special EAs as root
3282 Details : Client side always return success for setxattr call for lustre
3283 special xattr (currently only "trusted.lov").
3286 Frequency : rarely on clusters with both ia64+i386 clients
3288 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3289 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3290 clients are doing concurrent writes to the same file it is
3291 possible that the ia64 clients may LASSERT because the OST
3292 extent locks are not PAGE_SIZE aligned. Ensure that grown
3293 locks are always aligned on the request boundary.
3296 Frequency : specific use, occasional
3298 Description: Overwriting in use executable truncates on-disk binary image
3299 Details : If one node attempts to overwrite an executable in use by
3300 another node, we now correctly return ETXTBSY instead of
3301 truncating the file.
3303 Severity : enhancement
3305 Description: Async OSC create to avoid the blocking unnecessarily.
3306 Details : If a OST has no remain object, system will block on the creating
3307 when need to create a new object on this OST. Now, ways use
3308 pre-created objects when available, instead of blocking on an
3309 empty osc while others are not empty. If we must block, we block
3310 for the shortest possible period of time.
3315 Description: chmod on Lustre root is propagated to other clients
3316 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3317 invalid by the follow_mount time.
3320 Frequency : liblustre clients only
3322 Description: Race in 'instant cancel' lock handling could lead to such locks
3323 never to be granted in case of SMP MDS
3324 Details : Do not destroy not yet granted but cbpending locks in
3328 Frequency : replay/resend of open
3330 Description: non null lock assetion failure in mds_intent_policy
3331 Details : Trying to replay/resend lockless open requests resulted in
3332 mds_open() returning 0 with no lock. Now it sets a flag if
3333 a lock is going to be returned.
3335 Severity : enhancement
3337 Description: Checksum enhancements
3338 Details : New checksum enhancements allow for resending RPCs that failed
3341 Severity : enhancement
3343 Description: Tunables on number of dirty pages in cacche
3344 Details : Allow to set limit on number of dirty pages cached.
3349 Description: client crash on unmount - lock still has references
3350 Details : In some error handling cases it was possible to leak a lock
3351 reference on a client while accessing a file. This was not
3352 harmful to the client during operation, but would cause the
3353 client to crash when the filesystem is unmounted.
3356 Frequency : specific case, rare
3358 Description: ETXTBSY on mds though file not in use by client
3359 Details : ETXTBSY is no longer incorrectly returned when attempting to
3360 chmod or chown a directory that the user previously tried to
3361 execute or a currently-executing binary.
3364 Frequency : extremely rare except on liblustre-based clients
3366 Description: Lustre space not freed when files are deleted
3367 Details : Clean up open-unlinked files after client eviction. Previously
3368 the unlink was skipped and the files remained as orphans.
3373 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3374 Details : In some cases it was possible to send a blocking callback to
3375 a client doing a glimpse, even though that client didn't get
3376 a lock granted. When the glimpse lock is cancelled on the OST
3377 the freed lock is left on the waiting list and corrupted the list.
3380 Frequency : all core dumps
3382 Description: Broke core dumps to lustre
3383 Details : Negative dentry may be unhashed if parent does not have UPDATE
3384 lock, but some callers, e.g. do_coredump, expect dentry to be
3385 hashed after successful create, hash it in ll_create_it.
3387 ------------------------------------------------------------------------------
3389 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3391 * Support for kernels:
3392 2.6.9-42.0.2.EL (RHEL 4)
3393 2.6.5-7.276 (SLES 9)
3394 2.4.21-40.EL (RHEL 3)
3395 2.6.12.6 vanilla (kernel.org)
3399 Frequency : always on RHEL 3
3401 Description: Number of open files grows over time
3402 Details : The number of open files grows over time, whether or not
3403 Lustre is started. This was due to a filp leak introduced
3404 by one of our kernel patches.
3406 ------------------------------------------------------------------------------
3408 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3410 * Support for kernels:
3411 2.6.9-42.EL (RHEL 4)
3412 2.6.5-7.276 (SLES 9)
3413 2.4.21-40.EL (RHEL 3)
3414 2.6.12.6 vanilla (kernel.org)
3419 Bugzilla : 5719, 9635, 9792, 9684
3420 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3421 Details : If a server is under heavy load and cannot reply to new
3422 connection requests before the client resends the (re)connect,
3423 the connection handling code can behave badly if two service
3424 threads are concurrently handing separate (re)connections from
3425 the same client. Add better locking to the connection handling
3426 code, and ensure that only a single connection will be processed
3427 for a given client UUID, even if the lock is dropped.
3429 Severity : enhancement
3431 Description: add TCP zero-copy support to kernel
3432 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3433 sends if the hardware supports scatter-gather and checksumming.
3434 This allows socklnd to do client-write and server-read more
3435 efficiently and reduce CPU utilization from skbuf copying.
3438 Frequency : only if NFS exporting from client
3440 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3441 Details : If setting attributes on a file created under NFS that had
3442 never been opened it would be possible to oops the client
3443 if the file had no objects.
3446 Frequency : always for liblustre
3448 Description: liblustre client does MDS+OSTs setattr RPC for each write
3449 Details : When doing a write from a liblustre client, the client
3450 incorrectly issued an RPC to the MDS and each OST the file was
3451 striped over in order to update the timestamps. When writing
3452 with small chunks and many clients this could overwhelm the MDS
3453 with RPCs. In all cases it would slow down the write because
3454 these RPCs are unnecessary.
3456 Severity : enhancement
3458 Description: allow number of MDS service threads to be changed at module load
3459 Details : It is now possible to change the number of MDS service threads
3460 running. Adding "options mds mds_num_threads={N}" to the MDS's
3461 /etc/modprobe.conf will set the number of threads for the next
3462 time Lustre is restarted (assuming the "mds" module is also
3463 reloaded at that time). The default number of threads will
3464 stay the same, 32 for most systems.
3469 Description: OST crash if filesystem is unformatted or corrupt
3470 Details : If an OST is started on a device that has never been formatted
3471 or if the filesystem is corrupt and cannot even mount then the
3472 error handling cleanup routines would dereference a NULL pointer.
3477 Description: NULL pointer deref in llap_from_page.
3478 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3479 incorrect) mapping if the page was truncated/reclaimed while it was
3480 searched for. Check for this condition and skip such pages when
3481 doing readahead. Introduce extra check to llap_from_page() to
3482 verify page->mapping->host is non-NULL (so page is not anonymous).
3485 Frequency : Sometimes when using sys_sendfile
3487 Description: "page not covered by a lock" warnings from ll_readpage
3488 Details : sendfile called ll_readpage without right page locks present.
3489 Now we introduced ll_file_sendfile that does necessary locking
3490 around call to generic_file_sendfile() much like we do in
3494 Frequency : with certain MDS communication failures at client mount time
3496 Description: NULL pointer deref after failed client mount
3497 Details : a client connection request may delayed by the network layer
3498 and not be sent until after the PTLRPC layer has timed out the
3499 request. If the client fails the mount immediately it will try
3500 to clean up before the network times out the request. Add a
3501 reference from the request import to the obd device and delay
3502 the cleanup until the network drops the request.
3505 Frequency : occasionally during client (re)connect
3507 Description: assertion failure during client (re)connect
3508 Details : processing a client connection request may be delayed by the
3509 client or server longer than the client connect timeout. This
3510 causes the client to resend the connection request. If the
3511 original connection request is replied in this interval, the
3512 client may trip an assertion failure in ptlrpc_connect_interpret()
3513 which thought it would be the only running connect process.
3516 Frequency : only with obd_echo servers and clients that are rebooted
3518 Description: kernel BUG accessing uninitialized data structure
3519 Details : When running an obd_echo server it did not start the ping_evictor
3520 thread, and when a client was evicted an uninitialized data
3521 structure was accessed. Start the ping_evictor in the RPC
3522 service startup instead of the OBD startup.
3524 Severity : enhancement
3525 Bugzilla : 10193 (patchless)
3526 Description: Remove dependency on various unexported kernel interfaces.
3527 Details : No longer need reparent_to_init, exit_mm, exit_files,
3528 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3531 Frequency : rare (only users of deprecated and unsupported LDAP config)
3533 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3534 Details : LDAP apparently contains 'lustreName' attributes instead of
3535 'name'. A simple remapping of the name is sufficient.
3538 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3540 Description: waiting_locks_callback trips kernel BUG if client is evicted
3541 Details : Running with the dump_on_timeout debug flag turned on makes
3542 it possible that the waiting_locks_callback() can try to dump
3543 the Lustre kernel debug logs from an interrupt handler. Defer
3544 this log dumping to the expired_lock_main() thread.
3546 Severity : enhancement
3548 Description: Support NFS exporting on 2.6 kernels.
3549 Details : Implement non-rawops metadata methods for NFS server to use without
3550 changing NFS server code.
3553 Frequency : very rare (synthetic metadata workload only)
3555 Description: two racing renames might cause an MDS thread to deadlock
3556 Details : Running the "racer" program may cause one MDS thread to rename
3557 a file from being the source of a rename to being the target of
3558 a rename at exactly the same time that another thread is doing
3559 so, and the second thread has already enqueued these locks after
3560 doing a lookup of the target and is trying to relock them in
3561 order. Ensure that we don't try to re-lock the same resource.
3564 Frequency : only very large systems with liblustre clients
3566 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3567 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3568 all OSTs in parallel. This allows the eviction of stale liblustre
3569 clients to proceed much faster than if they were done in series,
3570 and also offers similar improvements for other set_info RPCs.
3575 Description: excessive CPU usage during initial read phase on client
3576 Details : During the initial read phase on a client, it would agressively
3577 retry readahead on the file, consuming too much CPU and impacting
3578 performance (since 1.4.5.8). Improve the readahead algorithm
3579 to avoid this, and also improve some other common cases (read
3580 of small files in particular, where "small" is files smaller than
3581 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3586 Description: MDS crash when receiving packet with unknown intent.
3587 Details : Do not LBUG in unknown intent case, just return -EFAULT
3589 Severity : enhancement
3590 Bugzilla : 9293, 9385
3591 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3592 Details : Do not serialize getattr (non-intent version) and statfs.
3595 Frequency : occasional, when OST network is overloaded/intermittent
3597 Description: client evicted by OST after bulk IO timeout
3598 Details : If a client sends a bulk IO request (read or write) the OST
3599 may evict the client if it is unresposive to its data GET/PUT
3600 request. This is incorrect if the network is overloaded (takes
3601 too long to transfer the RPC data) or dropped the OST GET/PUT
3602 request. There is no need to evict the client at all, since
3603 the pinger and/or lock callbacks will handle this, and the
3604 client can restart the bulk request.
3607 Frequency : Always when mmapping file with no objects
3609 Description: client crashes when mmapping file with no objects
3610 Details : Check that we actually have objects in a file before doing any
3611 operations on objects in ll_vm_open, ll_vm_close and
3617 Description: Request leak when working with deleted CWD
3618 Details : Introduce advanced request refcount tracking for requests
3619 referenced from lustre intent.
3621 Severity : Enhancement
3623 Description: Cache open file handles on client.
3624 Details : MDS now will return special lock along with openhandle, if
3625 requested and client is allowed to hold openhandle, even if unused,
3626 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3627 closing files for every read/write openration.
3629 Severity : Enhancement
3631 Description: Cache open negative dentries on client when possible.
3632 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3633 negative dentries on lock revocation.
3638 Description: Remounting a client read-only wasn't possible with a zconf mount
3639 Details : It wasn't possible to remount a client read-only with llmount.
3641 Severity : enhancement
3642 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3643 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3644 MPICH is included to add Lustre-specific ADIO interfaces.
3645 This is based closely on the UFS ADIO layer and only differs
3646 in file creation, in order to allow the OST striping to be set.
3647 This is user-contributed code and not supported by CFS.
3652 Description: extended inode attributes (immutable, append-only) work improperly
3653 when 2.4 and 2.6 kernels are used on client/server or vice versa
3654 Details : Introduce kernel-independent values for these flags.
3656 Severity : enhancement
3659 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3660 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3661 granularity. Now, Fractional MB granularity is be supported,
3662 this is very useful for low memory system.
3664 Severity : enhancement
3666 Description: Getattr by fid
3667 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3668 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3669 have locks and avoids getattr rpc requests for it on every stat.
3672 Frequency : Always, for filesystems larger than 2TB
3674 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3675 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3676 filesystems larger than 2TB to be mounted with mballoc enabled.
3679 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3681 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3682 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3683 filesystems larger than 2TB is mounted then the kernel will
3684 silently corrupt the start of the filesystem. CONFIG_LBD is
3685 enabled for all CFS-supported kernels, but the possibility of
3686 this happening with a modified kernel config exists.
3688 Severity : enhancement
3690 Description: add client O_DIRECT support for 2.6 kernels
3691 Details : It is now possible to do O_DIRECT reads and writes to files
3692 in the Lustre client mountpoint on 2.6 kernel clients.
3694 Severity : enhancement
3696 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3697 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3698 parallel, not waiting for response from every OST before sending
3699 a rpc to the next OST.
3704 Description: setattr vs write race when updating file timestamps
3705 Details : Client processes that update a file timestamp into the past
3706 right after writing to the file (e.g. tar) it is possible that
3707 the updated file modification time can be reset to the current
3708 time due to a race between processing the setattr and write RPC.
3710 Severity : enhancement
3712 Description: Bring 'lfs find' closer in line with regular Linux find.
3713 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3714 -print0 options and obtains all the needed info through the lustre
3717 Severity : enhancement
3719 Description: support up to 1024 configured devices on one node
3720 Details : change obd_dev array from statically allocated to dynamically
3721 allocated structs as they are first used to reduce memory usage
3726 Description: Flush dirty partially truncated pages during truncate
3727 Details : Immediatelly flush partially truncated pages in filter_setattr,
3728 this way we completely avoid having any pages in page cache on OST
3729 and can retire ugly workarounds during writes to flush such pages.
3734 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3735 Details : protect inode from truncation within vfs_unlink() context
3736 just take a reference before calling vfs_unlink() and release it
3737 when parent's i_sem is free.
3742 Description: last_id value checked outside lock on OST caused LASSERT failure
3743 Details : If there were multiple MDS->OST object precreate requests in
3744 flight, it was possible that the OST's last object id was checked
3745 outside a lock and incorrectly tripped an assertion. Move checks
3746 inside locks, and discard old precreate requests.
3749 Frequency : always, if extents are used on OSTs
3751 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3752 Details : OSTs using the extents format would not zero the high 16 bits of
3753 the index physical block number. This is not a problem for any
3754 OST filesystems smaller than 16TB, and no kernels support ext3
3755 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3756 new/modified files) and can be fixed for existing filesystems
3757 with e2fsprogs-1.39-cfs1.
3762 Description: import connection selection may be incorrect if timer wraps
3763 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3764 import connections to be ignored if the 32-bit jiffies counter
3765 wraps. Use a 64-bit jiffies counter.
3768 Frequency : very large clusters immediately after boot
3770 Description: LNET request buffers exhausted under heavy short-term load
3771 Details : If a large number of client requests are generated on a service
3772 that has previously never seen so many requests it is possible
3773 that the request buffer growth cannot keep up with the spike in
3774 demand. Instead of dropping incoming requests, they are held in
3775 the LND until the RPC service can accept more requests.
3778 Frequency : Sometimes during replay
3780 Description: Assertion failure in ll_local_open after replay.
3781 Details : If replay happened on an open request reply before we were able
3782 to set replay handler, reply will become not swabbed tripping the
3783 assertion in ll_local_open. Now we set the handler right after
3784 recognising of open request
3787 Frequency : very rare
3789 Description: kernel reports "badness in vsnprintf"
3790 Details : Reading from the "recovery_status" /proc file in small chunks
3791 may cause a negative length in lprocfs_obd_rd_recovery_status()
3792 call to vsnprintf() (which is otherwise harmless). Exit early
3793 if there is no more space in the output buffer.
3795 Severity : enhancement
3797 Description: clear OBD RPC statistics by writing to them
3798 Details : It is now possible to clear the OBD RPC statistics by writing
3799 to the "stats" file.
3804 Description: Client mtime is not the same on different clients after utimes
3805 Details : In some cases, the client was using the utimes() syscall on
3806 a file cached on another node. The clients now validate the
3807 ctime from the MDS + OSTs to determine which one is right.
3812 Description: Inability to activate failout mode
3813 Details : lconf script incorrectly assumed that in pythong string's numeric
3814 value is used in comparisons.
3817 Frequency : always with multiple stripes per file
3819 Description: Inefficient object allocation for mutli-stripe files
3820 Details : When selecting which OSTs to stripe files over, for files with
3821 a stripe count that divides evenly into the number of OSTs,
3822 the MDS is always picking the same starting OST for each file.
3823 Return the OST selection heuristic to the original design.
3828 Description: mount failures may take full timeout to return an error
3829 Details : Under some heavy load conditions it is possible that a
3830 failed mount can wait for the full obd_timeout interval,
3831 possibly several minutes, before reporting an error.
3832 Instead return an error as soon as the status is known.
3834 Frequency : quota enabled and large files being deleted
3836 Description: releasing more than 4GB of quota at once hangs OST
3837 Details : If a user deletes more than 4GB of files on a single OST it
3838 will cause the OST to spin in an infinite loop. Release
3839 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3844 Description: statfs data retrieved from /proc may be stale or zero
3845 Details : When reading per-device statfs data from /proc, in the
3846 {kbytes,files}_{total,free,avail} files, it may appear
3847 as zero or be out of date.
3850 Frequency : systems with MD RAID1 external journal devices
3852 Description: lconf's call to blkid is confused by RAID1 journal devices
3853 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3854 of returning all block devices that match the journal UUID.
3857 Frequency : always, for aggregate stripe size over 4GB
3859 Description: assertion fails when trying to use 4GB stripe size
3860 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3861 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3864 Frequency : always on ppc64
3866 Description: the first write on an ext3 filesystem with mballoc got stuck
3867 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3868 perform endianness conversion.
3870 ------------------------------------------------------------------------------
3872 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3874 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3875 INTEROPERATE with older versions automatically. Please read the
3876 user documentation before upgrading any part of a live system.
3877 * WARNING: Lustre networking configuration changes are required with
3878 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3881 * Support for kernels:
3882 2.6.9-22.0.2.EL (RHEL 4)
3883 2.6.5-7.244 (SLES 9)
3884 2.6.12.6 vanilla (kernel.org)
3887 Severity : enhancement
3888 Bugzilla : 7981/8208
3889 Description: Introduced Lustre Networking (LNET)
3890 Details : LNET is new networking infrastructure for Lustre, it includes
3891 a reorganized network configuration mode (see the user
3892 documentation for full details) as well as support for routing
3893 between different network fabrics. Lustre Networking Devices
3894 (LNDs) for the supported network fabrics have also been
3895 created for this new infrastructure.
3897 Severity : enhancement
3898 Description: Introduced Access control lists
3899 Details : clients can set ACLs on files and directories in order to have
3900 more fine-grained permissions than the standard Unix UGO+RWX.
3901 The MDS must be started with the "-o acl" mount option.
3903 Severity : enhancement
3904 Description: Introduced filesystem quotas
3905 Details : Administrators may now establish per-user quotas on the
3908 Severity : enhancement
3910 Description: Configuration change for the XT3
3911 The PTLLND is now used to run Lustre over Portals on the XT3
3912 The configure option(s) --with-cray-portals are no longer used.
3913 Rather --with-portals=<path-to-portals-includes> is used to
3914 enable building on the XT3. In addition to enable XT3 specific
3915 features the option --enable-cray-xt3 must be used.
3920 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3921 Details : A race between allocating a new llcd and re-getting the llcd_lock
3922 allowed another thread to grab newly-allocated llcd.
3924 Severity : enhancement
3926 Description: 2.6 OST async journal commit and locking fix to improve performance
3927 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3928 2.6 kernels are now async as they already were in 2.4 kernels so
3929 that they can commit concurrently with the network bulk transfer.
3930 For block-allocated files the filter allocation semaphore is held
3931 to avoid filesystem fragmentation during allocation. BKL lock
3932 removed for 2.6 xattr operations where it is no longer needed.
3937 Description: lconf incorrectly determined whether two IP networks could talk
3938 Details : In some more complicated routing and multiple-network
3939 configurations, lconf will avoid trying to make a network
3940 connection to a disjoint part of the IP space. It was doing the
3941 math incorrectly for one set of cases.
3946 Description: Fix for potential infinite loop processing records in an llog.
3947 Details : If an llog record is corrupted/zeroed, it is possible to loop
3948 forever in llog_process(). Validate the llog record length
3949 and skip the remainder of the block on error.
3952 Frequency : occasional (liblustre only)
3954 Description: liblustre could not open files whose last component is a symlink
3955 Details : sysio_path_walk() would incorrectly pass the open intent to
3956 intermediate path components.
3959 Frequency : rare (liblustre only with non-standard tuning)
3960 Bugzilla : 7201 (7350)
3961 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3962 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3963 after it has been released, only do this for referenced locks
3965 Severity : enhancement
3967 Description: specifying an (invalid) directory default stripe_size of -1
3968 would reset the directory default striping
3969 Details : stripe_size -1 was used internally to signal directory stripe
3970 removal, now use "all default" to signal dir stripe removal
3971 as a directory striping of "all default" is not useful
3974 Frequency : common for large clusters running liblustre clients
3976 Description: doing an ls when liblustre clients are running is slow
3977 Details : sending a glimpse AST to a liblustre client waits for every AST
3978 to time out, as liblustre clients will not respond. Since they
3979 cannot cache data we refresh the OST lock LVB from disk instead.
3981 Severity : enhancement
3983 Description: doing an ls at the same time as file IO can be slow
3984 Details : enqueue and other "small" requests can be blocked behind many
3985 large IO requests. Create a new OST IO portal for non-IO
3986 requests so they can be processed faster.
3989 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3991 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3992 created file with mode 000
3993 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3994 allow an ftruncate() as a non-root user to a file with mode 000.
3995 Lustre now respects this flag to disable mode checks when
3996 truncating a file owned by the user
3999 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4002 Description: Revoking locks from clients that went dead or catatonic might take
4004 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4005 cancellation of such locks instant on servers without waiting for
4006 any reply from clients. Clients drops these locks when cancel
4007 notification from server is received without replying.
4010 Frequency : liblustre-only, when liblustre client dies or becomes busy
4012 Description: Doing ls on Linux clients can take a long time with active
4014 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4015 granting such locks to it in the first place if possible. Locks
4016 are taken by proxy on the OST during the read or write and
4017 dropped immediately afterward. Add connect flags handling, do
4018 not grant locks to liblustre clients for glimpse ASTs.
4020 Severity : enhancement
4022 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4023 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4024 sequential reads and behaves badly with applications doing random
4025 reads. Improve it by reading ahead at least read region, and
4026 avoiding excessive large RPC for small reads.
4028 Severity : enhancement
4030 Description: Creating more than 1000 files for a single job may cause a load
4031 imbalance on the OSTs if there are also a large number of OSTs.
4032 Details : qos_prep_create() uses an OST index reseed value that is an
4033 even multiple of the number of available OSTs so that if the
4034 reseed happens in the middle of the object allocation it will
4035 still utilize the OSTs as uniformly as possible.
4040 Description: OST or MDS may oops in ping_evictor_main()
4041 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4042 but doesn't restart at beginning of obd_exports_timed list
4045 Severity : enhancement
4047 Description: improve by-nid export eviction on the MDS and OST
4048 Details : allow multiple exports with the same NID to be evicted at one
4049 time without re-searching the exports list.
4052 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4054 Description: MDS may oops in groups_free()
4055 Details : in rare race conditions a newly allocated group_info struct is
4056 freed again, and this can be NULL. The 2.4 compatibility code
4057 for groups_free() checked for a NULL pointer, but 2.6 did not.
4060 Frequency : common for liblustre clients doing little filesystem IO
4061 Bugzilla : 9352, 7313
4062 Description: server may evict liblustre clients accessing contended locks
4063 Details : if a client is granted a lock or receives a completion AST
4064 with a blocking AST already set it would not reply to the AST
4065 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4068 Frequency : lfs setstripe, only systems with more than 160 OSTs
4070 Description: unable to set striping with a starting offset beyond OST 160
4071 Details : llapi_create_file() incorrectly limited the starting stripe
4072 index to the maximum single-file stripe count.
4075 Frequency : LDAP users only
4077 Description: lconf did not handle in-kernel recovery with LDAP properly
4078 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4080 Severity : enhancement
4082 Description: bind OST threads to NUMA nodes to improve performance
4083 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4084 node and do their allocations there to localize memory access
4086 Severity : enhancement
4088 Description: llmount can determine client NID directly from Myrinet (GM)
4089 Details : the client NID code from gmnalnid was moved directly into
4090 llmount, removing the need to use this or specifying the
4091 client NID explicitly when mounting GM clients with zeroconf
4094 Frequency : if client is started with down MDS
4096 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4097 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4098 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4099 ptlrpc_import_delayed_req() because we didn't check if the
4100 request was interrupted, and we also didn't break out of the
4101 event loop if there was no timeout
4106 Description: data loss during non-page-aligned writes to a single file from
4107 both multiple nodes and multiple threads on one node at same time
4108 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4109 inconsistency led to false short-reads, that were cached and later
4110 used by ->prepare_write() to fill in partially written page,
4111 leading to data loss.
4114 Frequency : always, if lconf --abort_recovery used
4116 Description: lconf --abort_recovery fails with 'Operation not supported'
4117 Details : lconf was attempting to abort recovery on the MDT device and not
4120 Severity : enhancement
4122 Description: remove cleanup logs
4123 Details : replace lconf-generated cleanup logs with lustre internal
4124 cleanup routines. Eliminates the need for client-cleanup and
4127 Severity : enhancement
4129 Description: add support for EAs (user and system) on lustre filesystems
4130 Details : it is now possible to store extended attributes in the Lustre
4131 client filesystem, and with the user_xattr mount option it
4132 is possible to allow users to store EAs on their files also
4134 Severity : enhancement
4136 Description: Add possibility (config option) to show minimal available OST free
4138 Details : When compiled with --enable-mindf configure option, statfs(2)
4139 (and so, df) will return least minimal free space available from
4140 all OSTs as amount of free space on FS, instead of summary of
4141 free spaces of all OSTs.
4143 Severity : enhancement
4145 Description: do not expand extent locks acquired on OST-side
4146 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4147 by server: they are not cached anyway.
4150 Frequency : when mmap is used/binaries executed from Lustre
4152 Description: Unmmap pages before throwing them away from read cache.
4153 Details : llap_shrink cache now attempts to unmap pages before discarding
4154 them (if unmapping failed - do not discard). SLES9 kernel has
4155 extra checks that trigger if this unmapping is not done first.
4160 Description: lconf didn't resolve symlinks before checking to see whether a
4161 given mountpoint was already in use
4164 Frequency : when migrating failover services
4165 Bugzilla : 6395, 9514
4166 Description: When migrating a subset of services from a node (e.g. failback
4167 from a failover service node) the remaining services would
4168 time out and evict clients.
4169 Details : lconf --force (implied by --failover) sets the global obd_timeout
4170 to 5 seconds in order to quickly disconnect, but this caused
4171 other RPCs to time out too quickly. Do not change the global
4172 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4174 Severity : enhancement
4175 Frequency : if MDS is started with down OST
4176 Bugzilla : 9439,5706
4177 Description: Allow startup/shutdown of an MDS without depending on the
4178 availability of the OSTs.
4179 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4180 Add appropriate locking and lov-osc refcounts for safe
4181 cleaning. Add osc abort_inflight calls in case the
4182 synchronize never started.
4185 Frequency : occasional (Cray XT3 only)
4187 Description: root not authorized to access files in CRAY_PORTALS environment
4188 Details : The client process capabilities were not honoured on the MDS in
4189 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4190 been accessed by an authorized user then root was able to access
4191 the file on the local client also. The root user capabilities
4192 are now allowed on the MDS, as this environment has secure UID.
4195 Frequency : occasional
4197 Description: ldiskfs "too long searching" message happens too often
4198 Details : A debugging message (otherwise harmless) prints too often on
4199 the OST console. This has been reduced to only happen when
4200 there are fragmentation problems on the filesystem.
4205 Description: Division by zero in statfs when all OSCs are inactive
4206 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4207 lov_statfs divides by value returned by lov_get_stripecnt().
4211 Bugzilla : 9489, 3273
4212 Description: First write from each client to each OST was only 4kB in size,
4213 to initialize client writeback cache, which caused sub-optimal
4214 RPCs and poor layout on disk for the first writen file.
4215 Details : Clients now request an initial cache grant at (re)connect time
4216 and so that they can start streaming writes to the cache right
4217 away and always do full-sized RPCs if there is enough data.
4218 If the OST is rebooted the client also re-establishes its grant
4219 so that client cached writes will be honoured under the grant.
4224 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4225 Details : Now I/O RPCs go to different portal number and (presumably) fast
4226 lock requests (and glimses) and other RPCs get their own service
4227 threads pool that should be able to service those RPCs
4230 Severity : enhancement
4232 Description: Ability to exchange lustre version between client and servers and
4233 issue warnings at client side if client is too old. Also for
4234 liblustre clients there is ability to refuse connection of too old
4236 Details : New 'version' field is added to connect data structure that is
4237 filled with version info. That info is later checked by server and
4241 Frequency : rare, liblustre only.
4242 Bugzilla : 9296, 9581
4243 Description: Two simultaneous writes from liblustre at offset within same page
4244 might proceed at the same time overwriting eachother with stale
4246 Details : I/O lock withing llu_file_prwv was released too early, before data
4247 actually was hitting the wire. Extended lock-holding time until
4248 server acknowledges receiving data.
4251 Frequency : extremely rare. Never observed in practice.
4253 Description: avoid generating lustre_handle cookie of 0.
4254 Details : class_handle_hash() generates handle cookies by incrementing
4255 global counter, and can hit 0 occasionaly (this is unlikely, but
4256 not impossible, because initial value of cookie counter is
4257 selected randonly). Value of 0 is used as a sentinel meaning
4258 "unassigned handle" --- avoid it. Also coalesce two critical
4259 sections in this function into one.
4261 Severity : enhancement
4263 Description: allow liblustre clients to delegate truncate locking to OST
4264 Details : To avoid overhead of locking, liblustre client instructs OST to
4265 take extent lock in ost_punch() on client's behalf. New connection
4266 flag is added to handle backward compatibility.
4268 Severity : enhancement
4269 Bugzilla : 4928, 7341, 9758
4270 Description: allow number of OST service threads to be specified
4271 Details : a module parameter allows the number of OST service threads
4272 to be specified via "options ost ost_num_threads={N}" in the
4273 OSS's /etc/modules.conf or /etc/modprobe.conf.
4277 Bugzilla : 6146, 9635, 9895
4278 Description: servers crash with bad pointer in target_handle_connect()
4279 Details : In rare cases when a client is reconnecting it was possible that
4280 the connection request was the last reference for that export.
4281 We would temporarily drop the export reference and get a new
4282 one, but this may have been the last reference and the export
4283 was just destroyed. Get new reference before dropping old one.
4285 Severity : enhancement
4286 Frequency : if client is started with failover MDS
4288 Description: Allow multiple MDS hostnames in the mount command
4289 Details : Try to read the configuration from all specified MDS
4290 hostnames during a client mount in case the "primary"
4293 Severity : enhancement
4295 Description: Stop sending data to evicted clients as soon as possible.
4296 Details : Check if the client we are about to send or are sending data to
4297 was evicted already. (Check is done every second of waiting,
4298 for which l_wait_event interface was extended to allow checking
4299 of exit condition at specified intervals).
4302 Frequency : rare, normally only when NFS exporting is done from client
4304 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4306 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4309 Frequency : always, if srand() is called before liblustre initialization
4311 Description: Liblustre uses system PRNG disturbing its usage by user application
4312 Details : Introduce internal to lustre fast and high-quality PRNG for
4313 lustre usage and make liblustre and some other places in generic
4314 lustre code to use it.
4316 Severity : enhancement
4317 Bugzilla : 9477, 9557, 9870
4318 Description: Verify that the MDS configuration logs are updated when xml is
4319 Details : Check if the .xml configuration logs are newer than the config
4320 logs stored on the MDS and report an error if this is the case.
4321 Request --write-conf, or allow starting with --old_conf.
4323 Severity : enhancement
4325 Description: Handle symlinks in the path when checking if Lustre is mounted.
4326 Details : Resolve intermediate symlinks when checking if a client has
4327 mounted a filesystem to avoid duplicate client mounts.
4332 Description: lconf can hit an error exception but still return success.
4333 Details : The lconf command catches the Command error exception at the top
4334 level script context and will exit with the associated exit
4335 status, but doesn't ensure that this exit status is non-zero.
4340 Description: failure of ptlrpc thread startup can cause oops
4341 Details : Starting a ptlrpc service thread can fail if there are a large
4342 number of threads or the server memory is very fragmented.
4343 Handle this without oopsing.
4346 Frequency : always, only if liblustre and non-default acceptor port was used
4348 Description: liblustre cannot connect to servers with non-default acceptor port
4349 Details : tcpnal_set_default_params() was not called and was therefore
4350 ignoring the environment varaible TCPNAL_PORT, as well as other
4351 TCPNAL_ environment variables
4356 Description: two objects could be created on the same OST for a single file
4357 Details : If an OST is down, in some cases it was possible to create two
4358 objects on a single OST for a single file. No problems other
4359 than potential performance impact and spurious error messages.
4363 Bugzilla : 5681, 9562
4364 Description: Client may oops in ll_unhash_aliases
4365 Details : Client dcache may become inconsistent in race condition.
4366 In some cases "getcwd" can fail if the current directory is
4372 Description: Inode refcounting problems in NFS export code
4373 Details : link_raw functions used to call d_instantiate without obtaining
4374 extra inode reference first.
4378 Bugzilla : 9942, 9903
4379 Description: Referencing freed requests leading to crash, memleaks with NFS.
4380 Details : We used to require that call to ll_revalidate_it was always
4381 followed by ll_lookup_it. Also with revalidate_special() it is
4382 possible to call ll_revalidate_it() twice for the same dentry
4383 even if first occurence returned success. This fix changes semantic
4384 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4385 reference on a request referred from the intent.
4386 ll_intent_release() then releases such a request.
4389 Frequency : rare, normally benchmark loads only
4391 Description: unlinked inodes were kept in memory on the client
4392 Details : If a client is repeatedly creating and unlinking files it
4393 can accumulate a lot of stale inodes in the inode slab cache.
4394 If there is no other client load running this can cause the
4395 client node to run out of memory. Instead flush old inodes
4396 from client cache that have the same inode number as a new inode.
4399 Frequency : SLES9 2.6.5 kernel and long filenames only
4400 Bugzilla : 9969, 10379
4401 Description: utime reports stale NFS file handle
4402 Details : SLES9 uses out-of-dentry names in some cases, which confused
4403 the lustre dentry revalidation. Change it to always use the
4407 Frequency : rare, unless heavy write-truncate concurrency is continuous
4408 Bugzilla : 4180, 6984, 7171, 9963, 9331
4409 Description: OST becomes very slow and/or deadlocked during object unlink
4410 Details : filter_destroy() was holding onto the parent directory lock
4411 while truncating+unlinking objects. For very large objects this
4412 may block other threads for a long time and slow overall OST
4413 responsiveness. It may also be possible to get a lock ordering
4414 deadlock in this case, or run out of journal credits because of
4415 the combined truncate+unlink. Solution is to do object truncate
4416 first in one transaction without parent lock, and then do the
4417 final unlink in a new transaction with the parent lock. This
4418 reduces the lock hold time dramatically.
4421 Frequency : rare, 2.4 kernels only
4423 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4424 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4425 before unmounting their filesystem in order to avoid deadlock.
4426 The kernel_locked() function in 2.4 kernels only checks whether
4427 the kernel lock is held, not whether it is this process that is
4428 holding it as 2.6 kernels do.
4433 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4434 Details : The client ptlrpc code may be trying to reconnect to a down
4435 server before a previous connection attempt has timed out.
4436 Increase the reconnect interval to be longer than the connection
4437 timeout interval to avoid sending duplicate connections to
4441 Frequency : echo_client brw_test command
4443 Description: fix echo_client to work with OST preallocated code
4444 Details : OST preallocation code (5137) didn't take echo_client IO path
4445 into account: echo_client calls filter methods outside of any
4446 OST thread and, hence, there is no per-thread preallocated
4447 pages and buffers to use. Solution: hijack pga pages for IO. As
4448 a byproduct, this avoids unnecessary data copying.
4452 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4453 Description: Client can oops in mdc_commit_close() after open replay
4454 Details : It was possible for the MDS to return an open request with no
4455 transaction number in mds_finish_transno() if the client was
4456 evicted, but without actually returning an error. Clients
4457 would later try to replay that open and may trip an assertion
4458 Simplify the client close codepath, and always return an error
4459 from the MDS in case the open is not successful.
4462 Frequency : rare, 2.6 OSTs only
4464 Description: OST may deadlock under high load on fragmented files
4465 Details : If there was a heavy load and highly-fragmented OST filesystems
4466 it was possible to have all the OST threads deadlock waiting on
4467 allocation of biovecs, because the biovecs were not released
4468 until the entire RPC IO was completed. Instead, release biovecs
4469 as soon as they are complete to ensure forward IO progress.
4471 Severity : enhancement
4473 Description: Support for specifying external journal device at mount
4474 Details : If an OST or MDS device is formatted with an external journal
4475 device, this device major/minor is stored in the ext3 superblock
4476 and may not be valid for failover. Allow detecting and
4477 specifying the external journal at mount time.
4482 Description: Mounting an MDS with pending unlinked files may cause oops
4483 Details : target_finish_recovery() calls mds_postrecov() which returned
4484 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4485 considers this an error and immediately begins cleaning up the
4486 lov, just after starting the mds_lov process
4488 Severity : enhancement
4490 Description: Implement 'lfs df' to report actual free space on per-OST basis
4491 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4492 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4493 usage of objects; '-h' to report in human readable format.
4495 ------------------------------------------------------------------------------
4497 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4504 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4505 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4506 at startup during orphan unlinking. Instead, skip buddy update
4507 before setup, as it will be regenerated after recovery is complete.
4512 Description: If an OST is inactive, its locks might reference stale inodes.
4513 Details : lov_change_cbdata() must iterate over all namespaces, even if
4514 they are inactive to clear inode references from the lock.
4516 Severity : enhancement
4517 Frequency : occasional, if non-standard max_dirty_mb used
4519 Description: Client will block write RPCs if not enough grant
4520 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4521 then the client will block writes while waiting for another RPC
4522 to complete instead of consuming its dirty limit. With change
4523 we get improved performance when max_dirty_mb is small.
4525 Severity : enhancement
4526 Bugzilla : 3389, 6253
4527 Description: Add support for supplementary groups on the MDS.
4528 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4529 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4530 lookups for user supplementary groups into a cache.
4534 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4535 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4536 for liblustre O_EXCL is enforced by MDS.
4540 Description: readdir never returns NULL in liblustre.
4541 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4542 dentry in struct dirent.
4546 Frequency : liblustre only, depends on application IO pattern
4547 Description: liblustre clients evicted if not contacting servers
4548 Details : Don't put liblustre clients into the ping_evictor list, so
4549 they will not be evicted by the pinger ever.
4551 Severity : enhancement
4553 Description: Add ability to evict clients by NID from MDS.
4554 Details : By echoing "nid:$NID" string into
4555 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4556 $NID would be instantly evicted from this MDS and from all active
4557 OSTs connected to it.
4561 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4562 Details : lookup_it_finish() used to query file size from OSTs that was not
4567 Description: service threads change working directory to that of init
4568 Details : Starting lustre service threads may pin the working directory
4569 of the parent thread, making that filesystem busy. Threads
4570 now change to the working directory of init to avoid this.
4574 Frequency : during shutdown only
4575 Description: shutdown with a failed MDS or OST can cause unmount to hang
4576 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4581 Frequency : 2.6 only
4582 Description: chmod/chown may include an extra supplementary group
4583 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4584 supplementary group and if none is specified this is used.
4587 Bugzilla : 5479 (6816)
4589 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4590 Details : If lookup is in progress on a file that is unlinked we might try
4591 to revalidate the inode and fail in revalidate after lookup is
4592 complete and ll_file_open() enqueues the open again but
4593 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4596 Frequency : always, if lconf --abort_recovery used
4598 Description: lconf --abort_recovery fails with 'Operation not supported'
4599 Details : lconf was attempting to abort recovery on the MDT device and not
4602 ------------------------------------------------------------------------------
4604 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4609 Frequency : rare (only unsupported configurations with a node running as an
4611 Bugzilla : 6514, 5137
4612 Description: Mounting a Lustre file system on a node running as an OST could
4614 Details : OSTs now preallocates memory needed to write out data at
4615 startup, instead of when needed, to avoid having to
4616 allocate memory in possibly low memory situations.
4617 Specifically, if the file system is mounted on on OST,
4618 memory pressure could force it to try to write out data,
4619 which it needed to allocate memory to do. Due to the low
4620 memory, it would be unable to do so and the node would
4621 become unresponsive.
4623 Severity : enhancement
4625 Description: Addition of lconf --service command line option
4626 Details : lconf now accepts a '--service <arg>' option, which is
4627 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4629 Severity : enhancement
4631 Description: Failover mode is now the default for OSTs.
4632 Details : By default, OSTs will now run in failover mode. To return to
4633 the old behaviour, add '--failout' to the lmc line for OSTs.
4635 Severity : enhancement
4637 Description: Health checks are now provided for MDS and OSTs
4638 Details : Additional detailed health check information on MSD and OSTs
4639 is now provided through the procfs health_check value.
4642 Frequency : occasional, depends on IO load
4644 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4645 after numerous create/delete cycles
4646 Details : The ext3 inode allocation policy would not allocate new inodes
4647 very well on the OSTs because there are no new directories
4648 being created. Instead we look for groups with free space if
4649 the parent directories are nearly full.
4654 Description: Network or server problems during mount may cause partially
4655 mounted clients instead of returning an error.
4656 Details : The config llog parsing code may overwrite the error return
4657 code during mount error handling, returning success instead
4663 Description: MDS can fail to allocate large reply buffers
4664 Details : After long uptimes the MDS can fail to allocate large reply
4665 buffers (e.g. zconf client mount config records) due to memory
4666 fragmentation or consumption by the buffer cache. Preallocate
4667 some large reply buffers so that these replies can be sent even
4668 under memory pressure.
4672 Frequency : rare (liblustre)
4673 Description: fsx running with liblustre complained that using truncate() to
4674 extend the file doesn't work. This patch corrects that issue.
4675 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4676 ATTR_SIZE and lsm use in llu_setattr_raw.
4680 Frequency : rare, only 2.6 kernels
4681 Description: Unusual file access patterns on the MDS may result in inode
4682 data being lost in very rare circumstances.
4683 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4684 code in the 2.6 kernel caused the inode and/or EA data not to
4685 be read from disk, causing single-file corruption.
4689 Frequency : rare, only 2.6 filesystems using extents
4690 Description: Heavy concurrent write and delete load may cause data corruption.
4691 Details : It was possible under high-load situations to have an extent
4692 metadata block in the block device cache from a just-unlinked
4693 file overwrite a newly-allocated data block. We now unmap any
4694 metadata buffers that alias just-allocated data blocks.
4698 Frequency : filesystems with default stripe_count larger than 77
4699 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4700 Details : lconf specifies an inode size of 4096 bytes when the default
4701 stripe_count is larger than 77. This conflicts with the default
4702 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4703 this case to avoid pinning too much memory for large EAs.
4705 ------------------------------------------------------------------------------
4707 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4712 Frequency : rare (extremely heavy IO load with hundreds of clients)
4714 Description: Client is evicted, gets IO error writing to file
4715 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4716 caused clients to be evicted for AST timeouts. The fixes in
4717 bug 5192 mean we no longer need to have such short AST timeouts
4718 so ldlm_timeout has been increased.
4721 Frequency : occasional during --force or --failover shutdown under load
4722 Bugzilla : 5949, 4834
4723 Description: Server oops/LBUG if stopped with --force or --failover under load
4724 Details : a collection of import/export refcount and cleanup ordering
4725 issues fixed for safer force cleanup
4728 Frequency : only filesystems larger than 120 OSTs
4729 Bugzilla : 5990, 6223
4730 Description: lfs getstripe would oops on a very large filesystem
4731 Details : lov_getconfig used kfree on vmalloc'd memory
4734 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4735 Bugzilla : 6242, 6243
4736 Description: reading from files that had been truncated to a non-zero size
4737 but never opened returned no data
4738 Details : ll_file_read() reads zeros from no-object files to EOF
4743 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4744 Details : There's little harm in aborting MDS/OSS recovery and letting it
4745 try again, so I removed the LASSERT and return an error instead.
4747 Severity : enhancement
4749 Description: New debugging infrastructure for tracking down data corruption
4750 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4751 (b) cover more of the client-side code path, and (c) try to narrow
4752 down where problems occurred
4756 Bugzilla : 3819, 4364, 4397, 6313
4757 Description: Racing close and eviction MDS could cause assertion in mds_close
4758 Details : It was possible to get multiple mfd references during close and
4759 client eviction, leading to one thread referencing a freed mfd.
4761 Severity: : enhancement
4762 Bugzilla : 3262, 6359
4763 Description: Attempts to reconnect to servers are now more aggressive.
4764 Details : This builds on the enhanced upcall-less recovery that was added
4765 in 1.4.2. When trying to reconnect to servers, clients will
4766 now try each server in the failover group every 10 seconds. By
4767 default, clients would previously try one server every 25 seconds.
4772 Description: After recovery, certain operations trigger a failed
4773 assertion on a client.
4774 Details : Failing over an mds, using lconf -d --failover, while a
4775 client was doing a readdir() call would cause the client to
4776 LBUG after recovery completed and the readdir() was resent.
4778 Severity : enhancement
4780 Description: Default groups are now added by lconf
4781 Details : You can now run lconf --group <servicename> without having to
4782 manually add groups with lmc.
4785 Frequency : occasional
4787 Description: Nodes with an elan id of 0 trigger a failed assertion
4790 Frequency : always when accessing e.g. tty/console device nodes
4792 Description: tty and some other devices nodes cannot be used on lustre
4793 Details : file's private_data field is used by device data and lustre
4794 values in there got lost. New field was added to struct file to
4795 store fs-specific private data.
4798 Frequency : when exporting Lustre via NFS
4800 Description: NFSD failed occasionally when looking up a path component
4801 Details : NFSD is looking up ".." which was broken in ext3 directories
4802 that had grown large enough to become hashed.
4805 Frequency : Clusters with multiple interfaces not on the same subnet
4807 Description: Nodes will repeatedly try to reconnect to an interface which it
4808 cannot reach and report an error to the log.
4809 Details : Extra peer list entries will be created by lconf with some peers
4810 unreachable. lconf now validates the peer before adding it.
4813 Frequency : Only if a default stripe is set on the filesystem root.
4815 Description: Setting a default stripe on the filesystem root prevented the
4816 filesystem from being remounted.
4817 Details : The client was sending extra request flags in the root getattr
4818 request and did not allocate a reply buffer for the dir EA.
4821 Frequency : occasional, higher if lots of files are accessed by one client
4822 Bugzilla : 6159, 6097
4823 Description: Client trips assertion regarding lsm mismatch/magic
4824 Details : While revalidating inodes the VFS looks up inodes with ifind()
4825 and in rare cases can find an inode that is being freed.
4826 The ll_test_inode() code will free the lsm during ifind()
4827 when it finds an existing inode and then the VFS later attaches
4828 this free lsm to a new inode.
4832 Bugzilla : 6422, 7030
4833 Description: MDS deadlock between mkdir and client eviction
4834 Details : Creating a new file via mkdir or mknod (starting a transaction
4835 and getting the ns lock) can deadlock with client eviction
4836 (gets ns lock and trying to finish a synchronous transaction).
4839 Frequency : occasional
4840 Description: While starting a server, the fsfilt_ext3 module could not be
4842 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4843 kernels. Previously, lconf would still use the ext3 name
4844 when trying to load modules. Now, it will correctly use
4845 ext3 on 2.4 and ldiskfs on 2.6.
4847 Severity : enhancement
4848 Description: The default stripe count has been changed to 1
4849 Details : The interpretation of the default stripe count (0, to lfs
4850 or lmc) has been changed to mean striping across a single
4851 OST, rather than all available. For general usage we have
4852 found a stripe count of 1 or 2 works best.
4854 Severity : enhancement
4855 Description: Add support for compiling against Cray portals.
4856 Details : Conditional compiling for some areas that are different
4860 Frequency : occasional
4861 Bugzilla : 6409, 6834
4862 Description: Creating files with an explicit stripe count may lead to
4863 a failed assertion on the MDS
4864 Details : If some OSTs are full or unavailable, creating files may
4865 trigger a failed assertion on the MDS. Now, Lustre will
4866 try to use other servers or return an error to the
4870 Frequency : occasional
4872 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4873 caused lock timeout during readahead (since 1.4.2).
4874 Details : Processes doing readahead might match a lock that hasn't been
4875 granted yet if there are overlapping and conflicting lock
4876 requests. The readahead process waits on ungranted lock
4877 (original lock is CBPENDING), while OST waits for that process
4878 to cancel CBPENDING read lock and eventually evicts client.
4880 Severity : enhancement
4882 Description: Initial enabling of flock support for clients
4883 Details : Implements fcntl advisory locking and file status functions.
4884 This feature is provided as an optional mount flag (default
4885 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4886 locking are implemented yet, and those that are are not guaranteed
4887 to be completely correct in production environments.
4888 mount -t lustre -o [flock|noflock] ...
4891 Frequency : occasional
4893 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4894 assertion in the ext3 JBD (journaling) layer.
4895 Details : The b_committed_data struct is protected by the big kernel lock
4896 in 2.4 kernels, serializing journal_commit_transaction() and
4897 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4898 access to this struct. In 2.6 kernels there is finer grained
4899 locking to improve SMP performance of the JBD layer.
4903 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4908 Description: merge and cleanup kernel patches.
4911 -----------------------------------------------------------------------------
4913 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4915 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4916 versions of Lustre. Please update all servers and clients to
4917 version 1.4.2 or later at the same time. You must also run
4918 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4919 to update the configuration logs.
4921 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4922 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4923 - swab llog records of type '0' so we get proper header size/idx (5861)
4924 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4925 - fix rename of one directory over another leaking an inode (5953)
4926 - avoid SetPageDirty on 2.6 (5981)
4927 - don't re-add just-being-destroyed locks to the waiting list (5653)
4928 - when creating new directories, inherit the parent's custom
4929 striping settings if present parent (3048)
4930 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4931 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4932 - don't hold client locks on temporary worklist from l_lru (5666)
4933 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4934 - automatically evict dead clients (5921)
4935 - Update file size properly in create+truncate+fstat case (6196)
4936 - Do not unhash mountpoint dentries, do not allow removal of
4938 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4939 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4940 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4941 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4942 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4943 - Correctly handle reads of files with no objects (6243)
4944 - lctl recover will also mark a device active if deactivate used (5933)
4946 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4947 - allow --write-conf on an MDS with different nettype than client (5619)
4948 - don't write config llogs to MDS for mounts not from that MDS (5617)
4949 - lconf should create multiple TCP connections from a client (5201)
4950 - init scripts are now turned off by default; run chkconfig --on
4951 lustre and chkconfig --on lustrefs to use them
4952 - upcalls are no longer needed for clients to recover to failover
4954 - add --abort-recovery option to lconf to abort recovery on device
4956 - add support for an arbitrary number of OSTs (3026)
4957 - Quota support protocol changes.
4958 - forward compatibility changes to wire structs (6007)
4959 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4960 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4961 - improved "lustre status" script
4962 - initialize blocksize for non-regular files (6062)
4963 - added --disable-server and --disable-client configure options (5782)
4964 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4965 - Vanilla 2.4.29 support
4966 - increase maximum number of obd devices to 520 (6242)
4967 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4968 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4969 SLES 9 2.6.5 kernels
4970 - limit stripes per file to 160 (the maximum EA size) (6093)
4972 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4975 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4976 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4977 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4978 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4979 - small ext3 extents cleanups and fixes (5733)
4980 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4981 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4982 - avoid needless client->OST connect, fix handle mismatch (5317)
4983 - fix DLM error path that led to out-of-sync client, long delays (5779)
4984 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4985 - fix several locking issues related to i_size (5492,5624,5654,5672)
4986 - don't move pending lock onto export if it is already evicted (5683)
4987 - fix kernel oops when creating .foo in unlinked directory (5548)
4988 - fix deadlock in obdfilter statistics vs. object create (5811)
4989 - use time_{before,after} to avoid timer jiffies wrap (5882)
4990 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4991 - Do not leak request if server was not able to process it (5154)
4992 - If mds_open unable to find parent dir, make that negative lookup(5154)
4993 - don't create new directories with extent-mapping (5909, 5936)
4995 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4996 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4997 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4998 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4999 - support for RHEL4 kernel on i686 (5773)
5000 - provide error messages when incompatible logs are encountered (5898)
5002 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5003 * version 1.4.0.10 (1.4.1 release candidate 1)
5005 - don't keep a lock reference when lock is not granted (4238)
5006 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5007 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5008 - drop import inflight refcount on signal_completed_replay error (5255)
5009 - unlock page after async write error during send (3677)
5010 - handle missing objects in filter_preprw_read properly (5265)
5011 - no transno return for symlink open, don't save no-trasno open (3440)
5012 - don't try to complete elan receive that already failed (4012)
5013 - free RPC server reply state on error (5406)
5014 - clean up thread from ptlrpc_start_thread() on error (5160)
5015 - readahead could read extra page into cache that wasn't ejected (5388)
5016 - prevent races in class_attach/setup/cleanup/detach (5260)
5017 - don't dereference de->d_inode after l_dput of de (5458)
5018 - use "int" for stripe value returned from lock_to_stripe (5544)
5019 - mballoc allocation and error-checking fixes in 2.6 (5504)
5020 - block device patches to fix I/O request sizes in 2.6 (5482)
5021 - look up hostnames for IB nals (5602)
5022 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5023 - don't start multiple acceptors for the same port (5277)
5024 - fix incorrect LASSERT in mds_getattr_name (5635)
5025 - export a proc file for general "ping" checking (5628)
5026 - fix "lfs check" to not block when the MDS is down (5628)
5028 - service request history (4965)
5029 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5030 - create an "evict_client" /proc entry on OSTs, like the MDS has
5031 - fix mount usage message, return errors per mount(8) (5168)
5032 - change grep [] to grep "[]" in tests so they work in more UMLs
5033 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5034 - remove ancient LOV_MAGIC_V0 EA support (5047)
5035 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5036 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5037 - allow readahead allocations to fail when low on memory (5383)
5038 - mmap locking landed again, after considerable improvement (2828)
5039 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5041 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5044 - send OST transaction number in read/write reply to free req (4966)
5045 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5046 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5047 - the watchdog thread now runs as interruptible (5246)
5048 - flock/lockf fixes (but it's still disabled, pending 5135)
5049 - don't use EXT3 constants in llite code (5094)
5050 - memory shortage at startup could cause assertion (5176)
5052 - reorganization of lov code
5053 - single portals codebase
5055 - add extents/mballoc support (5025)
5056 - direct I/O reads in the obdfilter (4048)
5057 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5059 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5062 - send OST transaction number in read/write reply to free req (4966)
5063 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5064 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5065 - the watchdog thread now runs as interruptible (5246)
5066 - handle missing objects in filter_preprw_read properly (5265)
5067 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5068 - drop import inflight refcount on signal_completed_replay error (5255)
5069 - unlock page after async write error during send (3677)
5070 - return original error code on reconstructed replies (3761)
5071 - no transno return for symlink open, don't save no-trasno open (3440)
5073 - add pid to ldlm debugging output (4922)
5074 - bump the watchdog timeouts -- we can't handle 30sec yet
5075 - extra debugging for orphan dentry/inode bug (5259)
5077 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5080 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5081 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5082 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5083 - fix NULL dereference in /proc/sys/portals/routes (4827)
5084 - allow failed mdc_close() operations to be interrupted (4561)
5085 - stop precreate on OST before MDS would time out on it (4778)
5086 - don't send partial-page writes before EOF from client (4410)
5087 - discard client grant for sub-page writes on large-page clients (4520)
5088 - don't free dentries not owned by NFS code, check generation (4806)
5089 - fix lsm leak if mds_create_objects() fails (4801)
5090 - limit debug_daemon file size, always print CERROR messages (4789)
5091 - use transno after validating reply (3892)
5092 - process timed out requests if import state changes (3754)
5093 - update mtime on OST during writes, return in glimpse (4829)
5094 - add mkfsoptions to LDAP (4679)
5095 - use ->max_readahead method instead of zapping global ra (5039)
5096 - don't interrupt __l_wait_event() during strace
5098 - add software watchdogs to catch hung threads quickly (4941)
5099 - make lustrefs init script start after nfs is mounted
5100 - fix CWARN/ERROR duplication (4930)
5101 - return async write errors to application if possible (2248)
5102 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5103 - print NAL number in %x format (4645)
5104 - update barely-supported suse-2.4.21-171 series (4842)
5105 - support for sles 9 %post scripts
5106 - support for building 2.6 kernel-source packages
5107 - support for sles km_* packages
5109 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5112 - ignore -ENOENT errors in osc_destroy (3639)
5113 - notify osc create thread that OSC is being cleaned up (4600)
5114 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5115 - reconstruct ost_handle() like mds_handle() (4657)
5116 - create a new thread to do import eviction to avoid deadlock (3969)
5117 - let lconf resolve symlinked-to devices (4629)
5118 - don't unlink "objects" from directory with default EA (4554)
5119 - hold socknal file ref over connect in case target is down (4394)
5120 - allow more than 32000 subdirectories in a single directory (3244)
5121 - fix blocks count for O_DIRECT writes (3751)
5122 - OST returns ENOSPC from object create when no space left (4539)
5123 - don't send truncate RPC if file size isn't changing (4410)
5124 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5125 - bind to privileged port in socknal and tcpnal (3689)
5127 - rate limit CERROR/CWARN console message to avoid overload (4519)
5128 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5129 - basic mmap support (3918)
5130 - kernel patch series update from b1_4 (4711)
5132 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5135 - avoid crash during MDS cleanup with OST shut down (2775)
5136 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5137 - don't use bad inodes on the MDS (3744)
5138 - dynamic object preallocation to improve recovery speed (4236)
5139 - don't hold spinlock over lock dumping or change debug flags (4401)
5140 - don't zero obd_dev when it is force cleaned (3651)
5141 - print grants to console if they go negative (4431)
5142 - "lctl deactivate" will stop automatic recovery attempts (3406)
5143 - look for existing locks in ldlm_handle_enqueue() (3764)
5144 - don't resolve lock handle twice in recovery avoiding race (4401)
5145 - revalidate should check working dir is a directory (4134)
5147 - don't always mark "slow" obdfilter messages as errors (4418)
5149 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5152 - don't close LustreDB during write_conf until it is done (3860)
5153 - fix typo in lconf for_each_profile (3821)
5154 - allow dumping logs from multiple threads at one time (3820)
5155 - don't allow multiple threads in OSC recovery (3812)
5156 - fix debug_size parameters (3864)
5157 - fix mds_postrecov to initialize import for llog ctxt (3121)
5158 - replace config semaphore with spinlock (3306)
5159 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5160 - don't allow enqueue to complete on a destroyed export (3822)
5161 - down write_lock before checking llog header bitmap (3825)
5162 - recover from lock replay timeout (3764)
5163 - up llog sem before sending rpc (3652)
5164 - reduce ns lock hold times when setting kms (3267)
5165 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5166 - fix NULL deref and obd_dev leak on setup error (3312)
5167 - replace some LBUG about llog ops with error handling (3841)
5168 - don't match INVALID dentries from d_lookup and spin (3784)
5169 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5170 - fix invalid assertion in ptlrpc_set_wait (3880)
5172 - add libwrap support for the TCP acceptor (3996)
5173 - add /proc/sys/portals/routes for non-root route listing (3994)
5174 - allow setting MDS UUID in .xml (2580)
5175 - print the stack of a process that LBUGs (4228)
5177 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5180 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5181 - make sure to unset replay flag from failed open requests (3440)
5182 - if default stripe count is 0, use OST count for inode size (3636)
5183 - update parent mtime/ctime on client for create/unlink (2611)
5184 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5185 - free recovery state on server during a forced cleanup (3571)
5186 - unregister_reply for resent reqs (3063)
5187 - loop back devices mounting and status check on 2.6 (3563)
5188 - fix resource-creation race that can provoke i_size == 0 (3513)
5189 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5190 - more debugging for page-accounting assertion (3746)
5191 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5192 - avoid deadlock after precreation failure (3758)
5193 - fix race and lock order deadlock in orphan handling (3450, 3750)
5194 - add validity checks when grabbing inodes from l_ast_data (3599)
5196 - add /proc/.../recovery_status to obdfilter (3428)
5197 - lightweight CDEBUG infrastructure, debug daemon (3668)
5198 - change default OSC RPC parameters to be better on small clusters
5199 - turn off OST read cache for files smaller than 32MB
5200 - install man pages and include them in rpms (3100)
5201 - add new init script for (un)mounting lustre filesystems (2593)
5202 - run chkconfig in %post for init scripts (3701)
5203 - drop scimac NAL (unmaintained)
5205 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5208 - clean kiobufs before and after use (3485)
5209 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5210 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5211 - change init script to fail more clearly if not run as root (1528)
5212 - allow clients to reconnect during replay (1742)
5213 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5214 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5215 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5216 - fix cond_resched (3554)
5217 - only evict unfinished clients after recovery (3515)
5218 - allow bulk resend, prevent data loss (3570)
5219 - dynamic ptlrpc request buffer allocation (2102)
5220 - don't allow unlinking open directory if it isn't empty (2904)
5221 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5222 - remove extraneous obd dereference causing LASSERT failure (3334)
5223 - don't use get_cycles() when creating temp. files on the mds (3156)
5224 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5225 - handle EEXIST for set-stripe, set proper directory name (3336)
5227 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5228 - fix ksocknal_fmb_callback() error messages (2918)
5230 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5233 - don't copy lvb into (possibly NULL) reply on error (2983)
5234 - don't deref dentry after dput, don't free lvb on error (2922)
5235 - use the kms to determine writeback rpc length (2947)
5236 - increment oti_logcookies when osc is inactive (2948)
5237 - update client's i_blocks count via lvb messages (2543)
5238 - handle intent open/close of special files properly (1557)
5239 - mount MDS with errors=remount-ro, like obdfilter (2009)
5240 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5241 - don't use cancelling-locks' kms values (2947)
5242 - use highest lock extent for kms, not last one (2925)
5243 - don't dereference ERR_PTR() dentry in error handling path (3107)
5244 - fix thread race in portals_debug_dumplog() (3122)
5245 - create lprocfs device entries at setup instead of at attach (1519)
5246 - common AST error handler, don't evict client on completion race (3145)
5247 - zero nameidata in detach_mnt in 2.6 (3118)
5248 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5249 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5250 - initialize RPC timeout timer earlier for 2.6 (3219)
5251 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5252 - print nal/nid for unknown nid (3258)
5253 - additional checks for oscc recovery before doing precreate (3284)
5254 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5255 - don't crash in mdc_close for bad permissions on open (3285)
5256 - zero i_rdev for non-device files (3147)
5257 - clear page->private before handing to FS, better assertion (3119)
5258 - tune the read pipeline (3236)
5259 - fix incorrect decref of invalidated dentry (2350)
5260 - provide read-ahead stats and refine rpc in flight stats (3328)
5261 - don't hold journal transaction open across create RPC (3313)
5262 - update atime on MDS at close time (3265)
5263 - close LDAP connection when recovering to avoid server load (3315)
5264 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5265 - don't leak open file on MDS after open resend (3325)
5266 - serialize filter_precreate and filter_destroy_precreated (3329)
5267 - loop device shouldn't call sync_dev() for nul device (3092)
5268 - clear page cache after eviction (2766)
5269 - resynchronize MDS->OST in background (2824)
5270 - refuse to mount the same filesystem twice on same mountpoint (3394)
5271 - allow llmount to create routes for mounting behind routers (3320)
5272 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5273 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5274 - fix rare problem with rename on htree directories (3417)
5276 - allow default OST striping configuration per directory (1414)
5277 - fix compilation for qswnal for 2.6 kernels (3125)
5278 - increase maximum number of MDS request buffers for large systems
5279 - change liblustreapi to be useful for external progs like lfsck (3098)
5280 - increase local configuration timeout for slow disks (3353)
5281 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5283 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5286 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5287 - don't overwrite extent policy data in reply if lock was blocked (2901)
5288 - drop filter export grants atomically with removal from device (2663)
5289 - del obd_self_export from work_list in class_disconnect_exports (2908)
5290 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5291 - swab reply message in mdc_close, other PPC fixes (2464)
5292 - fix destroying of named logs (2325)
5293 - overwrite old logs when running lconf --write_conf (2264)
5294 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5295 - fix race in target_handle_connect (2898)
5296 - mds_reint_create() should take same inode create lock (2926)
5297 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5298 - don't close files for self_export to avoid uninitialized obd (2936)
5299 - allow MDS with the same name as client node (2939)
5300 - hold dentry reference for closed log files for unlink (2325)
5301 - reserve space for all logs during transactions (2059)
5302 - don't evict page beyond end of stripe extent (2925)
5303 - don't oops on a deleted current working directory (2399)
5304 - handle hard links to targets without a parent properly (2517)
5305 - don't dereference NULL lock when racing during eviction (2867)
5306 - don't grow lock extents when lots of conflicting locks (2919)
5308 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5311 - account for cache space usage on clients to avoid data loss (974)
5312 - lfsck support in lustre kernel code (2349)
5313 - reduce journal credits needed for BRW writes (2370)
5314 - orphan handling to avoid losing space on client/server crashes
5315 - ptlrpcd can be blocked, stopping ALL progress (2477)
5316 - use lock value blocks to assist in proper KMS, faster stat (1021)
5317 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5318 - recovery for initial connections (2355)
5319 - fixes for mds_cleanup_orphans (1934)
5320 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5321 - block all file creations until orphan recovery completes (1901)
5322 - client remove rq_connection from request struct (2423)
5323 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5324 - recovery timer race (2670)
5325 - mdc_close recovey bug (2532)
5326 - ptlrpc cleanup bug (2710)
5327 - mds timeout on local locks (2588)
5328 - namespace lock held during RPCs (2431)
5329 - handle interrupted sync write properly (2503)
5330 - don't try to handle a message that hasn't been replied to (2699)
5331 - client assert failure during cleanup after abort recovery (2701)
5332 - leak mdc device after failed mount (2712)
5333 - ptlrpc_check_set allows timedout requests to complete (2714)
5334 - wait for inflight reqs when ptlrpcd finishes (2710)
5335 - make sure unregistered services are removed from the srv_list
5336 - reset bulk XID's when resending them (caught by 1138 test)
5337 - unregister_bulk after timeout
5338 - fix lconf error (2694)
5339 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5340 - readahead locks pages, leaves pending causing memory pressure (2673)
5341 - increase OST request buffers to 4096 on large machines (2729)
5342 - fix up permission of existing directories in simple_mkdir (2661)
5343 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5344 - don't assign transno to errored transactions (2742)
5345 - don't delete objects on OST if given a bogus objid from MDS (2751)
5346 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5347 - if rq_no_resend, then timeout request after recovery (2432)
5348 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5349 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5350 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5351 - MDS recovery shouldn't delete objects that it has given out (2730)
5352 - if enqueue arrives after completion, don't clobber LVB (2819)
5353 - don't unlock pages twice when trigger_group_io returns error (2814)
5354 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5355 - don't write pages to disk if there was an error (1450)
5356 - don't ping imports that have recovery disabled (2676)
5357 - take buffered bytes into account when balancing socknal conn (2817)
5358 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5359 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5360 - remove little-endian swabbing of llog records (1987)
5361 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5362 - retry reposting request buffers if they fail (1191)
5363 - grow extent at grant time to avoid granting a revoked lock (2809)
5364 - lock revoke doesn't evict page if covered by a second lock (2765)
5365 - disable VM readahead to avoid reading outside lock extents (2805)
5367 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5368 - updated kernel patches for hp-2.4.20 kernel (2681)
5370 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5373 - fix truncated write corruption (2366)
5374 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5376 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5377 - ensure len(uuid) < 37 in lmc (1171)
5378 - fix ia64 OOPS in llog_test (2255)
5379 - zero end of page at obdfilter for partial page writes (2648)
5380 - don't leave stale dentries around after renames (bug 2428)
5381 - fix timeouts when evicting a client with a single lock held (2642)
5382 - set deadline for the initial HELLO message to drain (2634)
5383 - print out dotted-quad IP addresses in the socknal (2302)
5385 - additional debugging for MDS client eviction problem (2443)
5386 - fix mkfsoptions support for osts (2603, 2604)
5388 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5391 - add series for the vanilla 2.6.0 kernel
5392 - add series for the vanilla 2.4.24 kernel
5393 - add series for a cray x86/64 UL kernel drop
5394 - fix xattr patches for the vanilla 2.4.19 series
5396 - generate true UUIDs in lmc (1171)
5397 - have portals stack dumping break in UML (2466)
5398 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5399 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5400 - wake callback waiting threads on client eviction (2460)
5401 - Add --ptldebug and --subsystem to lmc (1719)
5402 - update assertion to allow safe interrupt allocation
5403 - set rq_no_resend for cancel requests (2432)
5404 - recalculate ptlrpcd timeout after resend (2494)
5405 - call vfs_rmdir when removing pending directories (2368)
5406 - fix renaming a file to itself (2429)
5407 - lmc creates a default one-stripe lov (2454)
5408 - expand procfs space to handle large clusters (2326)
5409 - increase UML stack to avoid overflow
5410 - update lconf's list of debug and subsystem masks
5411 - fix lfs find --obd (2510)
5412 - /proc tunable for disabling filter read caching (2591)
5413 - stop rpm packages from altering slapd.conf (2301)
5414 - disable nagle in the socknal under 0conf (2578)
5415 - choose mds inode size based on stripe count (2572)
5416 - fix kernel-source rpm problems (2516)
5418 - add --disable-doc to avoid pdf generation (2421)
5419 - update documentation, tests, type-os, comments
5420 - avoid format warnings on ia64
5421 - remove the TOE NAL
5422 - tiny code cleanups by removing unused fields
5424 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5427 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5428 - remove the most blatant lies from BUILDING file (2371)
5429 - change default debug level to reasonable production setting
5430 - reduce client side cache size to reduce cache flush time
5431 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5432 - make TCP zerocopy and pinger support enabled by default (2476)
5433 - sync writes completed after process exits caused crashes (2319)
5434 - maintain correct mount count on the MDS (2356)
5435 - backout 1557, because 2316 wasn't really fixed
5436 - better file I/O statistics gathering in /proc
5437 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5438 - another kernel patch to fix zero-copy TCP function export
5439 - don't take duplicate lock when processing re-sent getattr (2420)
5440 - lctl uses obd_self_export instead of creating new conn (2353)
5441 - MDS/OST recovery case which requires object creation asserted (2425)
5442 - move lfs from /usr/sbin to /usr/bin in packages
5443 - fix race between mds_client_add and mds_client_free (2417)
5444 - use kmalloc instead of slabs in portals (2430)
5445 - don't create duplicate records when a failover MDS is present (2442)
5446 - remove unnecessary mount age check (2332)
5447 - don't remove directory inodes from locks prematurely (2451)
5448 - don't break if MDS service name is the same as hostname (2103)
5449 - fix races in client write RPC generation when cache full (2482)
5451 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5454 - remove now-unused request->rq_obd (278)
5455 - if an allocation fails, print out how much memory we've used (1933)
5456 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5457 - add the "configurable stack size" patch to most series files (1256)
5458 - ability to write large log records, for 100+ OST configs (2306)
5459 - fix NULL deref when filter_prep fails (2314)
5460 - fix operator precedence error in filter_sync
5461 - dynamic allocation of socknal TX descriptors (2315)
5462 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5463 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5464 - ensure that CWARN messages in Portals always get to the syslog
5465 - __init/__exit are not for prototype decls (ldlm_init/exit)
5466 - x86-64 compile warning fixes
5467 - fix gateway LMC keyword conflict (2318)
5468 - fix MDS lock inversions in getattr/reint paths (1844)
5469 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5470 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5471 - DLM race condition prevented some lock evictions (2328)
5472 - ENOMEM detection and retry on socknal sends (2230)
5473 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5474 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5475 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5476 - better error messages when a client is rejected during recovery (1505)
5477 - avoid cancelling locks which were never granted, after failure (2330)
5478 - fix i_sem/journal inversion in mds_client_add (2333)
5479 - fix truncate/getattr lock cycle deadlock (2334)
5480 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5481 - fix two rare exit paths which could leak an l_lock() ref (2321)
5482 - fencepost error in MDS/OST orphan recovery (2226)
5483 - make log record alignment 8 bytes (1988)
5484 - lstripe now fails when requested offset > ost_count (2237)
5485 - ensure that all kernel series have a complete list.h (1607)
5486 - fix crashes in special-file operations (2316)
5487 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5489 - allow configurable automake binary, for testing new versions
5490 - small update to the lfs documentation
5492 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5494 * fix negative export reference count in fsfilt_sync (2312)
5496 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5497 * release candidate 0.9.1
5499 - orphans are moved into the PENDING directory for possible recovery
5500 - replayed opens now open by fid for orphan/rename safety (1042)
5501 - last close of an orphan inode generates a transno (683)
5502 - chdir() and mount() now pin the directory entry (1020)
5503 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5504 - discard very old requests without processing them (1502)
5505 - remove some common, well-understood CERRORs (1505)
5506 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5507 - clear "grant" flags in OST replies until OST grant code lands (1644)
5508 - fix read performance by not clobbering i_blksize on client (1598)
5509 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5510 - make LRU size a /proc tunable, clears locks when reduced (707)
5511 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5512 - server locks take references on exports now (1558)
5513 - build fixes for 2.4.20-rh trees (1663)
5514 - return an error from lov_create if all OSCs are inactive (1751)
5515 - fix import levels when a reconnect happens without a timeout (1597)
5516 - exit early from mds_open if we get a lookup error (1749)
5517 - partial page read at EOF wouldn't wait for disk before sending (1642)
5518 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5519 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5520 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5521 - write/truncate lock inversion (1639)
5522 - Don't auto-load obdclass, portals modules during cleanup (1495)
5523 - fix timestamps from jumping to "now" (1763)
5524 - extra journal assertions (1648)
5525 - add an extra multiunlink test (1771)
5526 - fix read_record/write_record API (1776)
5527 - fix leak of offset_extent, possible incorrect i_size later (1772)
5528 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5529 - Debugging for the kqswnal_get_idle_tx problems (1820)
5530 - Allow recovery to be attempted multiple times (1536)
5531 - Write out MDS last_rcvd file after it is first created (1600)
5532 - Fix tx_descriptor leak in failed transmit situations (1827)
5533 - ext3 journaling fixes for assertion failure after IO error (1871)
5534 - class_export_put() on freed export after completion AST error (1896)
5535 - Fix revalidate looping in VFS (1322)
5536 - Don't access a freed export during MDS_REINT timeout (1521)
5537 - Add open-unlink recovery support on the MDS (1673,1764)
5538 - Return an error if no MDS data was read from last_rcvd (1946)
5539 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5540 - Don't setup a disk device that doesn't match exported UUID (317)
5541 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5542 - avoid committing NULL handle in force close
5543 - local.sh is now a one-stripe LOV configuration
5544 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5545 - fix ext3 htree duplicate directory entry corruption (1516)
5546 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5547 - update to new LOV EA format (2097)
5548 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5550 2003-06-15 Phil Schwan <phil@clusterfs.com>
5553 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5554 - per-import/export recovery handling (958, 931, 959)
5555 - multiple last-rcvd slots, for serving multiple FSes (949)
5556 - connections are again shared between multiple imp/exports (963, 964)
5557 - "umount -f" would hang if any requests needed to be sent (393, 978)
5558 - avoid pinning large req buffer by copying for queued messages (989)
5559 - add "uuid" to "lctl device" command to help upcalls (991)
5560 - "open" RPCs with transnos would confuse recovery counters (1037)
5561 - do proper endian conversion of all wire messages (288, 340, 891)
5562 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5563 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5564 - fix ptlrpc_connection leak in target_handle_connect (1174)
5565 - fix import refcounting bug in OST and MDS cleanup (1134)
5566 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5567 - fix very unlikely obd_types race condition (501)
5568 - remove osc_open hack for echo_client (1187)
5569 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5570 - a failure in read_inode2 leads to deadlock (1139)
5571 - cancel ack-locks as soon as transaction is committed (1072)
5572 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5573 - make sure to commitrw after any preprw to avoid deadlock (1162)
5574 - failing to execute a file in a lustre FS would lock inode (1203)
5575 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5576 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5577 - obd_uuid2tgt would walk off the end of the list (1255)
5578 - on IA64 the osc would give portals incorrect bulk size (1258)
5579 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5580 - fix lock inversion caused by new llite matching code (1282)
5581 - limit the number of dirty pages on a client to 10MB (1286)
5582 - timed out locks were not being corrected cancelled (1289)
5583 - fix O_DIRECT above 4GB on IA-32 (1292)
5584 * major user-visible changes
5585 - fail out/fail over policy now controlled by the upcall (993)
5587 - add OBD_PING to check server availability and failure (954)
5588 - lustre messages are now sent in sending host order (288, 340, 891)
5589 - add eadatalen to MDS getattr reply (340)
5590 - OST read replies may contain second buffer, with per-page status (593)
5592 2003-03-11 Phil Schwan <phil@clusterfs.com>
5595 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5596 - failed open()s could cause deadlock; fixed (867, 869)
5597 - stop cancelling OST locks when files are closed (481)
5598 - overlapping XID spaces caused network corruption (851, 853)
5599 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5600 - setattr_raw added, to do single-RPC, server-side setattrs
5601 - lmc/lconf syntax change for OST UUIDs
5602 - fix crashy race condition between ptlrpc_free_req and osc_close
5603 - don't use request in mdc_enqueue if we hit a timeout (889)
5604 - don't set the inode i_size for regular files from the MDS (896)
5605 - handle out of order completion AST (842)
5606 - don't LBUG if a lock request times out after receiving AST (913)
5607 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5608 - if a bad lock AST arrives, send an error instead of dropping entirely
5609 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5610 - fix leak in bulk IO when only partially completed (899, 900, 926)
5611 - fix O_DIRECT for ia64 (55)
5612 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5613 - C-z after timeout could hang a process forever; fixed (977)
5615 - client-side I/O cache (678, 924, 929, 941, 970)
5617 - READPAGE and SETATTRs which don't take server-side locks get
5620 2003-02-11 Phil Schwan <phil@clusterfs.com>
5623 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5624 - Fix fsfilt_extN_readpage() to read a full page of directory
5625 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5626 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5627 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5628 - Most intent operations were removed, in favour of a new RPC mode
5629 that does a single RPC to the server and bypasses most of the VFS
5630 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5631 - Aggressively cancel local locks on DLM servers
5632 - mds_reint_unlink sends EA to the client if it's the last nlink.
5633 client uses that EA to unlink OST objects.
5634 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5635 - recursive symlinks were fixed (439)
5636 - fixed NULL deref in DEBUG_REQ
5637 - filter_update_lastobjid no longer calls sync, which annoyed extN
5638 - fixed multi-client small-writes to a single file problem (445)
5639 - fixed mtime updates during file writes (607)
5640 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5641 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5642 - fixed Linux OST size reporting problem (444, 656)
5643 - OST now updates object mtime with writes or setattr (607, 619)
5644 - client verifies file size before zeroing page past EOF (445)
5645 - OST now writes last allocated objid to disk with allocation (108)
5646 - LOV on echo now works (409)
5648 - mds_reint_unlink sends a new buffer, with the EA included. this
5649 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5650 set if a regular file was being unlinked, and it was the last link
5651 - use PtlGet from the target for bulk writes (315)
5652 - OST now updates object mtime with writes or setattr (607, 619)
5653 - LDLM now has a grant-time callback to revalidate locked items, if
5655 - Many MDS operations were reorganized to combat race conditions
5657 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5658 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5660 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5663 - Fully reactivate OST imports after reconnection (512, others)
5664 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5665 - More graceful error handling for truncating on dead OST (515)
5666 - Don't error out unless we're actually accessing dead stripes (474)
5667 - Fix garbage sizes when stripes are missing (410)
5668 - LRU counters were broken, causing constant lock purge (433, 432)
5669 - garbage on read from stripes with failed OSTs (441)
5670 - mark OSCs as active before reconnecting during recovery (438)
5671 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5672 - lfind did not preserve OST order in output (443)
5673 - symlinks cause hung clients, incorrect data (439)
5674 - stop dereferencing request after dropping refcount (457)
5675 - don't LASSERT(spin_is_locked) on non-SMP (455)
5676 - fixes for many rename() bugs
5677 - fstat didn't correctly synchronize attributes (399)
5678 - server must handle lock cancellation during blocking AST prep (487)
5679 - bulk descriptors were free()d too soon (511)
5680 - fix paths in lconf, which would load incorrect modules (451, 507)
5681 - fix confusing lconf 'host not found' error message (386)
5682 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5683 - fix race condition in mdc_blocking_ast() for inode access (526)
5684 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5685 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5686 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5687 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5688 - fix lstripe in several ways (kernel side) (527)
5689 - fix request leak in ldlm_cli_enqueue (262)
5690 - incorrect OSC was marked inactive after OST failure
5691 - call mds_fs_cleanup before unmounting filesystem (524)
5692 - fix races between taking ns_lock and ldlm_lock_change_resource
5693 - fix races updating LOV export open file list
5694 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5695 - fix recovery NULL deref in ldlm_cli_cancel_unused
5696 - fix some DLM races by using new hash table for lock handles (419)
5697 - permit the client to specify desired inodes, at replay
5698 - duplicate requests when we queue them for replay reintegration
5699 - fix last_rcvd offset calculation
5700 - sync after each recovered transaction, so we always make progress
5701 - never, not always, ERESTART requests without transnos
5702 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5704 - skip replay if the MDS says that the client is already connected
5705 - don't check for a recovery-enabled export to match lctl's UUID
5706 - don't INC_USE_COUNT for phantom exports
5707 - don't crash when cleaning up phantom exports (567)
5708 - don't double-finish or set replay data for errored mdc_open requests
5709 - abort requests when they time out, so we don't get old replies
5710 - send/receive replies for AST messages again
5711 - if the client says that it doesn't have the lock, cancel it on the
5713 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5714 mark it as destroyed, it will all work out when decref is called
5715 - fix module use counts (22, 581)
5717 - ASTs now expect a reply (server cancels lock on error reply)
5719 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5722 - fix many simultaneous client startup (392)
5723 - fix dentry->d_it clobbering
5724 - credentials weren't being shipped for readdir/getattr operations
5725 - remove invalid assertions triggered during some concurrent MD
5727 - proper Lustre versions added (336, 389)
5728 - fix memory leak for create error case (398)
5729 - fix LOV locking bug that would get cli/srv out of sync
5730 - fix echo client over LOV (409)
5731 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5732 - fix double-O_EXCL intent crash (424)
5733 - avoid sending multiple lock CANCELs (352)
5735 - MDS can do multi-client recovery (modulo bugs in new code)
5737 - many updates, edits, cleanups
5739 2002-11-18 Phil Schwan <phil@clusterfs.com>
5742 - fix null d_it dereference (346)
5743 - fix full OST/dbench hang (333)
5744 - fix permission problem with file removal (286)
5745 - fix removal of OSCs from LOV when they fail
5746 - fix NULL deref during bulk timeout (214)
5747 - fix problems related to multiple filesystems on one MDS (241)
5748 - fixed serious subtle metadata locking bugs
5749 - free locks on clients when inodes are removed due to memory
5751 - fix inode pointer in lock data (285)
5752 - partial support for multiple MDS on a single host (241)
5753 - data locks weren't cancelled at clear_inode time (290, 311)
5754 - intent locks could lead to unbounded lock growth (205)
5755 - added a maximum lock count, an LRU list, and a flusher
5756 - fix multiple rename (365)
5757 - properly abstracted the echo client
5758 - OSC locked 1 byte too many; fixed
5759 - rewrote brw callback code:
5760 - fixed recovery bugs related to LOVs (306)
5761 - fixed too-many-pages-in-one-write crash (191)
5762 - fixed (again) crash in sync_io_timeout (214)
5763 - probably fixed callback-related race (385)
5765 - Add capability to MDS protocol
5766 - LDLM cancellations and callbacks on different portals
5768 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5771 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5772 - defer open object destruction to close time (601981, 138)
5773 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5774 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5775 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5776 - don't call obd_setattr in ll_file_release for destroyed objects
5777 * protocol change to lustre_msg: move |version| and add |flags|
5778 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5779 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5780 * added replay of create, unlink, link and rename operations during
5781 MDS failover; recovery should be much more robust now
5782 * remove failed OSCs from LOVs (only lov_create uses this so far)
5783 * the lustre-HOWTO was brought (more) up to date (582544)
5785 2002-10-23 Phil Schwan <phil@clusterfs.com>
5788 - in-use dentries weren't being reused properly (617851)
5789 - prevent multiple LDLM setup (599178)
5790 - fix LOV size calculations for truncate (617853)
5791 - fix client handling of MDS intent errors (POSIX)
5792 - fix permission bug in lovstripe.c test (624321)
5793 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5794 - truncate past end of file could corrupt data
5795 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5796 - a race in recovery could return ETIMEDOUT to apps (623947)
5797 - building outside the source directory was fixed
5798 * the lustre-HOWTO was brought (more) up to date (582544)
5799 * major progress was made on recovery functionality
5801 2002-10-10 Phil Schwan <phil@clusterfs.com>
5804 - recovery deadlock fix
5805 - rm -rf causes LBUG fix (617817)
5806 - file open by multiple tasks fix (618962)
5807 - directory permissions bugs (602707 and 620007)
5808 - journal_stop fixed with locking (611313)
5809 - O_APPEND failures resolved (618273, perhaps 614459)
5810 - lconf PATH fix (619770)
5811 - IA64 build fix (621450)
5812 - RPC buffer sizes scale with amount of memory
5814 2002-10-01 Phil Schwan <phil@clusterfs.com>
5817 - locks would be cancelled without throwing away data pages,
5818 resulting in inconsistent data (605627)
5819 - inode attributes were not always being refreshed (605627, 612449)
5820 - lconf now continues to cleanup after lctl reports an error
5821 - MDS now enforces user permissions (602707)
5822 - lprocfs cleanup fixed, but not yet enabled (614157)
5823 - fixed infinite server hang, should a client not respond to an AST
5824 - avoid going into recovery if user calls readlink() with a buffer
5825 that's too small (613941)
5826 - AST RPCs no longer require replies (614867) -- this may be changed
5827 - don't crash server if client sends an IOV that's too big (611336)
5828 - fixed lock conversion deadlock (611892)
5829 - fixed the following of symlinks (614622)
5830 * recovery: the server can remove locks from a client that dies, other
5831 clients can make progress
5832 * more extN patch fixes
5833 * compile-time configurable ptlrpc buffer allocations
5835 - collaborative read cache document
5836 - Lustre Lite Performance CDR document-in-progress
5838 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5841 - fix typo in patch-2.4.18
5843 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5846 - clear ptlrpc request each time in handle_incoming_request()
5847 - unlink of files now destroys the object on the OST
5849 2002-09-19 Peter Braam <braam@clusterfs.com>
5851 * add hard link support
5852 * change obdfile creation method
5853 * kernel patch changed
5855 2002-09-19 Peter Braam <braam@clusterfs.com>
5858 - stack overflow bug in extN fixed
5860 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5862 * documentation updates
5863 - add man pages for config tools
5864 - update tests/README to describe testing with new config tools
5865 - finish metadata API descriptions
5866 * bug fixes and cleanups
5867 - statfs workaround for 16TB limit
5868 - LOV stripe allocation improved, can stripe on subset of OSTs
5869 - LOV file size/IO offset was wrong for files > 4GB in size
5870 - object EA data was being dropped, caused files to be unreadable
5871 - memory overflow with non-LOV OST caused memory corruption
5872 - fixed regression tests to work with new config tools, obdfilter
5873 - fixed bug when directory size became larger than 1 block
5874 - fixed bug (for single client case) when PWD was deleted
5875 - invalidate local directory pages when doing intent-based ops
5876 - avoid LDLM oops when lock callback contained bad data
5878 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5880 * documentation updates
5881 * bug fixes and cleanups
5882 - configuration tools
5885 - 64-bit compile warnings
5886 - 64-bit internal statfs data
5888 * test_brw on persistent OST devices
5890 * lprocfs (disabled)
5892 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5894 * documentation updates
5895 * bug fixes and cleanups
5896 * configuration tools
5898 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5901 * many small fixes to 0_5_4
5902 * io/network handling
5903 * thinkos in MDS operations
5905 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5908 * crucial basic fixes to 0.5.3
5909 * IOR, Iozone work over Elan
5912 2002-08-07 Phil Schwan <phil@clusterfs.com>
5913 * version 0_5_3, our first alpha
5914 * we use the new Portals iovs
5915 * documentation updates
5916 * bug fixes and cleanups
5917 * small changes in the DLM wire protocol
5919 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5920 * version 0_5_1 with some initial stability,
5921 * locking on MD and file I/O.
5922 * documentation updates
5923 * several bug fixes since 0.5.0
5924 * small changes in wire protocol
5926 2002-07-18 Phil Schwan <phil@clusterfs.com>
5928 * delivered as Lustre Light Alpha
5929 * fixed a crash after handling invalid MDS requests
5930 * fixed directory pages for architectures with non-4k pages sizes
5932 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5933 * release version v0_4_4
5934 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5936 * Fixes a number of interruption problems with OST operations.
5937 * Update documentation for portals header changes
5938 * Move all wire protocol structs/defines to lustre_idl.h
5939 * Fixes symlink length bug.
5940 * Add tcpdump to repository.
5942 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5943 * release version v0_4_3
5944 * Fixes statfs for inodes on extN.
5945 * Fixes bug in runtests which would delete /etc/hosts.
5946 * Use 64-bit object IDs wherever possible (not into VFS though)
5947 Remove ost_get_info, which is unused by lustre, and out of date.
5949 2002-07-03 Peter Braam <braam@clusterfs.com>
5950 * release version v0_4_2 Fixes a lookup error (type not passed)
5951 * move forward to head of Portals
5952 * move forward to latest Lustre kernel
5954 2002-06-25 Peter Braam <braam@clusterfs.com>
5955 * release version v0_4_1. Hopefully stable on single node use.