1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.6.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: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
23 Description: access to llog context before init.
24 Details : move handling CATALOGS file at osc layer and forbid access to llog
29 Description: Avoid deadlock for local client writes
30 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
31 memory freeing context. This allows OST threads to set the
32 PF_MEMALLOC flag on task structures in order to allocate memory
33 from reserved pools and complete IO.
34 Use GFP_HIGHUSER for OST allocations for non-local client writes,
35 so that the OST threads generate memory pressure and allow
36 inactive pages to be reclaimed.
41 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
42 Details : move ldlm namespace creation in setup phase to avoid grab
43 _lprocfs_lock with cli_sem held.
47 Description: Temporarily disable grant shrink.
48 Details : Disable the feature for debugging.
52 Description: Unable to run several mkfs.lustre on loop devices at the same
54 Details : mkfs.lustre returns error 256 on the concurrent loop devices
55 formatting. The solution is to proper handle the error.
57 Severity : enhancement
59 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
61 Severity : enhancement
63 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
67 Description: resolve race between obd_disconnect and class_disconnect_exports
68 Details : if obd_disconnect will be called to already disconnected export he
69 forget release one reference and osc module can't unloaded.
71 Severity : enhancement
73 Description: Allow tuning service thread via /proc
74 Details : For each service a new
75 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
76 created that can be used to set min/max thread counts, and get the
77 current number of running threads.
81 Descriptoin: don't return error if have particaly created objects for file.
82 Details : lov_update_create_set uses set->set_success as index for created objects,
83 so if some requests will be failed, they will have hole at end of
84 array and we can use qos_shrink_lsm for allocate correct lsm.
86 Severity : enhancement
88 Description: Update OFED support to 1.4.1
92 Description: Reduce small size read RPC
93 Details : Set read-ahead limite for every file and only do read-ahead when
94 available read-ahead pages are bigger than 1M to avoid small size
97 Severity : enhancement
99 Description: move AT tunable parameters for more consistent usage
100 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
102 Severity : enhancement
104 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
105 Details : allow skip disconnected ost for send statfs request and hide error
111 Description: fix racy locking of mballoc block bitmaps causing BUG
112 Details : The locking of the mballoc buddy bitmap and the in-memory
113 block bitmap was using two different spin locks in some
114 cases. This made it possible to incorrectly access the
115 mballoc bitmap while another process was modifying it,
116 causing a sanity assertion to fail. While no on-disk corruption
117 was reported, there was some risk of this happening.
120 Frequency : rare, on llog test 6
122 Descriptoin: don't allow connect to already connected import
123 Details : allowing connect to already connected import is hide connecting problem.
128 Descriptoin: allow kill process which wait statahead result
129 Details : for some reasons 'ls' can stick in waiting result from statahead,
130 in this case need way for kill this process.
132 Severity : enhancement
134 Description: Add state history info file, enhance import info file
135 Details : Track import connection state changes in a new osc/mdc proc file;
136 add overview-type data to the osc/mdc import proc file.
138 Severity : enhancement
140 Description: MDS create should not wait for statfs RPC while holding DLM lock.
142 Severity : enhancement
144 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
149 Descriptoin: OST grant too much space to client even there are not enough space.
150 Details : Client will shrink its grant cache to OST if there are no write
151 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
152 this grant cache if there are already not enough avaible space
153 (left_space < total_clients * 32M).
156 Frequency : start MDS on uncleanly shutdowned MDS device
158 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
159 Details : stay in waiting mds<>ost recovery finished produce random bugs
160 due race between two ll_sync thread for one lov target. send
161 ACTIVATE event only if connect realy finished and import have
165 Frequency : rare, connect and disconnect target at same time
167 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
168 Details : don't call obd_disconnect under lov_lock. this long time
169 operation and can block ptlrpcd which answer to connect request.
174 Descriptoin: don't lose wakeup for imp_recovery_waitq
175 Details : recover_import_no_retry or invalidate_import and import_close can
176 both sleep on imp_recovery_waitq, but we was send only one wakeup
180 Frequency : always with long access acl
182 Descriptoin: mds can't pack reply with long acl.
183 Details : mds don't control size of acl but they limited by reint/getattr
186 Severity : enhancement
188 Description: Update to SLES10 kernel-2.6.16.60-0.33.
190 Severity : enhancement
192 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
195 Frequency : start MDS on uncleanly shutdowned MDS device
197 Descriptoin: aborting recovery hang on MDS
198 Details : don't throttle destroy RPCs for the MDT.
201 Frequency : on remount
203 Description: external journal device not working after the remount
204 Details : clear dev_rdonly flag for external journal devices in
210 Description: shutdown vs evict race
211 Details : client_disconnect_export vs connect request race.
212 if client will evicted at this time - we start invalidate
213 thread without referece to import and import can be freed
217 Frequency : rare, need acl's on inode.
219 Description: client can't handle ost additional correctly
220 Details : if ost was added after client connected to mds client can have
221 hit lnet_try_match_md ... to big messages to wide striped files.
222 in this case need teach client to handle config events about add
223 lov target and update client max ea size at that event.
225 Severity : enhancement
227 Description: Changelogs
228 Details : Changelogs are a lightweight mechanism to track filesystem
229 metadata and namespace changes. The changelog is recorded
230 permanently on the MDTs, and is periodically "consumed" / purged
231 when records are no longer needed.
233 Severity : enhancement
235 Description: compact fld format with extents
236 Details : Store range of seq rather than every seq in FLD. Seq
237 controller update FLD rather than clients. In Case of CMD, mdt0
238 has FLD, all other metadata server act as non persistent proxy
239 for FLD queries and cache fld entries in fld cache.
244 Description: don't skip ost target if they assigned to file
245 Details : Drop slow OSCs if we can, but not for requested start idx.
246 This means "if OSC is slow and it is not the requested
247 start OST, then it can be skipped, otherwise skip it only
248 if it is inactive/recovering/out-of-space.
252 Description: more cleanup in mds_lov
253 Details : not send LOV EA under replay, we can't know about they size at this
254 time. Don't allow client connect to mds before any ost connected,
255 for avoid problems with LOV EA size and returning EIO to client.
257 Severity : enhancement
259 Description: Interoperability at server side (Disk interoperability)
261 Severity : enhancement
263 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
265 Severity : enhancement
267 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
269 Severity : enhancement
271 Description: New client IO stack (CLIO).
273 Severity : enhancement
275 Description: Commit on sharing. Eliminate inter-client dependencies between
276 uncommitted transactions by doing transaction commits.
277 Thereby clients may recovery independently.
280 Frequency : Create a symlink file with a very long name
282 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
283 Details : If there is no extra space in the request for early cancels,
284 ldlm_req_handles_avail() returns 0 instead of a negative value.
286 Severity : enhancement
288 Description: Add /proc entry for import status
289 Details : The mdc, osc, and mgc import directories now have
290 an import directory that contains useful import data for debugging
293 Severity : enhancement
295 Description: Re-disable certain /proc logging
296 Details : Enable and disable client's offset_stats, extents_stats and
297 extents_stats_per_process stats logging on the fly.
300 Frequency : Only on FC kernels 2.6.22+
302 Description: oops in statahead
303 Details : Do not drop reference count for the dentry from VFS when lookup,
304 VFS will do that by itself.
306 Severity : enhancement
308 Description: Generic /proc file permissions
309 Details : Set /Proc file permissions in a more generic way to enable non-
310 root users operate on some /proc files.
314 Description: Hitting mdc_commit_close() ASSERTION
315 Details : Properly handle request reference release in
316 ll_release_openhandle().
320 Description: quota recovery deadlock during mds failover
321 Details : This patch includes att18982, att18236, att18237 in bz14840.
323 1. fix osts hang when mds does failover with quotaon
324 2. prevent watchdog storm when osts threads wait for the
329 Frequency : only patchless client
330 Description: add workaround for race between add/remove dentry from hash
332 Severity : enhancement
334 Description: Allow OST glimpses to return PW locks
338 Description: LBUG when llog conf file is full
339 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
343 Description: Prevent import from entering FULL state when server in recovery
347 Description: service mount cannot take device name with ":"
348 Details : Only when device name contains ":/" will mount treat it as
354 Description: replace ptlrpcd with the statahead thread to interpret the async
355 statahead RPC callback
359 Frequency : on recovery
360 Description: I/O failures after umount during fail back
361 Details : if client reconnected to restarted server we need join to recovery
362 instead of find server handler is changed and process self
363 eviction with cancel all locks.
365 Severity : enhancement
367 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
369 Severity : enhancement
371 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
373 Severity : enhancement
375 Description: Upcall on Lustre log has been dumped
376 Details : Allow for a user mode script to be called once a Lustre log has
377 been dumped. It passes the filename of the dumped log to the
378 script, the location of the script can be specified via
379 /proc/sys/lnet/debug_log_upcall.
384 Description: avoid idr_remove called for id which is not allocated.
385 Details : Move assigment s_dev for clustered nfs to end of initialization,
386 to avoid problem with error handling.
391 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
392 Details : When connection is reused this not moved from CONN_UNUSED_HASH
393 into CONN_USED_HASH and this prodice warning when put connection
394 again in unused hash.
397 Severity : enhancement
399 Description: File striping can now be set to use an arbitrary pool of OSTs.
401 Severity : enhancement
403 Description: Export bytes_read/bytes_write count on OSC/OST.
407 Description: Early reply size mismatch, MGC loses connection
408 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
409 the connect flags are properly negotiated.
414 Description: quotas are not honored with O_DIRECT
415 Details : all writes with the flag O_DIRECT will use grants which leads to
416 this problem. Now using OBD_BRW_SYNC to guard this.
420 Description: add quota statistics
421 Details : 1. sort out quota proc entries and proc code.
422 2. add quota statistics
424 Severity : enhancement
426 Description: enable quota support for HEAD.
430 Description: Properly propagate oinfo flags from lov to osc for statfs
431 Details : restore missing copy oi_flags to lov requests.
433 Severity : enhancement
435 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
436 : l_getgroups(8), lst(8), routerstat(8)
437 Details : included man pages for llobdstat(8), llstat(8),
438 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
440 Severity : enhancement
442 Description: configure's --enable-quota should check the
443 : kernel .config for CONFIG_QUOTA
444 Details : configure is terminated if --enable-quota is passed but
445 : no quota support is in kernel
449 Description: Remove portals compatibility
450 Details : Remove portals compatibility, not interoperable with releases
455 Description: Resolve device initialization race
456 Details : Prevent proc handler from accessing devices added to the
457 obd_devs array but yet be intialized.
459 Severity : enhancement
461 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
463 Severity : enhancement
465 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
470 Description: Using wrong pointer in osc_brw_prep_request
471 Details : Access to array[-1] can produce panic if kernel compiled with
472 CONFIG_PAGE_ALLOC enabled
476 Description: Client runs out of low memory
477 Details : Consider only lowmem when counting initial number of llap pages
481 Description: *optional* service tags registration
482 Details : if the "service tags" package is installed on a Lustre node
483 When the filesystem is mounted, a local-node service tag will
484 be created. See http://inventory.sun.com/ for more information
485 about the Service Tags asset management system.
489 Description: Kernel BUG tries to release flock
490 Details : Lustre does not destroy flock lock before last reference goes
491 away. So always drop flock locks when client is evicted and
492 perform unlock regardless of successfulness of speaking to MDS.
496 Description: add refcount protection for osc callbacks, avoid panic on shutdown
500 Description: sanity test 65a fails if stripecount of -1 is set
501 Details : handle -1 striping on filesystem in ll_dirstripe_verify
506 Description: ASSERTION(CheckWriteback(page,cmd)) failed
507 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
510 Severity : enhancement
512 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
516 Description: do not process already freed flock
517 Details : flock can possibly be freed by another thread before it reaches
518 to ldlm_flock_completion_ast.
522 Description: LBUG during stress test
523 Details : Need properly lock accesses the flock deadlock detection list.
527 Description: oops in page fault handler
528 Details : kernel page fault handler can return two special 'pages' in error
529 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
533 Description: timeout with invalidate import.
534 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
535 should be handled by ptlrpcd. This produce long age waiting and
536 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
538 Severity : enhancement
540 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
544 Description: enable MGS and MDT services start separately
545 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
546 a co-located MGS without starting the MGS, which is a complement
547 to 'nosvc' mount option.
551 Frequency : after recovery
552 Description: precreate to many object's after del orphan.
553 Details : del orphan st in oscc last_id == next_id and this triger growing
554 count of precreated objects. Set flag LOW to skip increase count
555 of precreated objects.
559 Frequency : rare, on clear nid stats
560 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
561 Details : when clean nid stats sometimes try destroy live entry,
562 and this produce panic in free.
566 Description: Stack overflow during MDS log replay
567 ease stack pressure by using a thread dealing llog_process.
571 Description: wait until IO finished before start new when do lock cancel.
572 Details : VM protocol want old IO finished before start new, in this case
573 need wait until PG_writeback is cleared until check dirty flag and
574 call writepages in lock cancel callback.
576 Severity : enhancement
578 Description: using special macro for print time and cleanup in includes.
582 Description: mds_mfd_close() ASSERTION(rc == 0)
583 Details : In mds_mfd_close(), we need protect inode's writecount change
584 within its orphan write semaphore to prevent possible races.
588 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
592 Frequency : rare, on shutdown ost
593 Description: don't hit live lock with umount ost.
594 Details : shrink_dcache_parent can be in long loop with destroy dentries,
595 use shrink_dcache_sb instead.
599 Description: don't panic with use echo client
600 Details : echo client pass NULL as client nid pointer and this produce null
605 Description: fix build on ppc32
606 Details : compile code with -m64 flag produce wrong object file for ppc32.
610 Description: add message levels for liblustreapi
614 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
615 Details : Move the 'good_osts' check before the 'total_bavail' check. This
616 will result in an -EAGAIN and in the exit call path we call
617 alloc_rr() which will with increasing aggressiveness attempt to
618 aquire precreated objects on the minimum number of required OSCs.
622 Description: Use old size assignment to avoid deadlock
623 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
624 the scheduling while holding a spinlock. We do not need locking
625 for size in ll_update_inode() because size is only updated from
626 the MDS for directories or files without objects, so there is no
627 other place to do the update, and concurrent access to such inodes
628 are protected by the inode lock.
632 Description: resolve "_IOWR redefined" build error on SLES10
636 Description: dump the memory debugging after all modules are unloaded to
637 suppress false negative in conf_sanity test 39
639 Severity : enhancement
641 Description: build kernel-ib packages for OFED 1.3 in our release cycle
646 Description: fix SLES kernel versioning
647 Details : the kernel version for our SLES 10 kernel did not include a "-"
648 before the "smp" at the end. while this was not a problem in
649 general, it did mean that software trying to use the kernel
650 version to try to detect a vendor specific kernel would fail.
651 this was most evident by the OFED build scripts.
655 Description: Don't update lov_desc members until making sure they are valid
656 Details : When updating lov_desc members via proc fs, need fix their
657 validities before doing the real update.
661 Description: don't put request into delay list while invalidate in flight.
662 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
663 import in flight. this produce timeout for invalidate and sometimes
664 can cause stale data.
668 Frequency : on ppc only
669 Description: not convert ost objects for directory because it's not exist.
670 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
674 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
676 Severity : enhancement
678 Description: Update to support 2.6.22.14 vanilla kernel.
682 Frequency : rare, on recovery
683 Description: read procfs can produce deadlock in some situation
684 Details : Holding lprocfs lock which send rpc can produce block for destroy
685 obd objects and this also block reconnect with -EALREADY. This
686 isn't fix all lprocfs bugs - but make it rare.
688 Severity : enhancement
690 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
693 Frequency : frequent on X2 node
695 Description: mdc_set_open_replay_data LBUG
696 Details : Set replay data for requests that are eligible for replay.
700 Description: lustre_mgs: operation 101 on unconnected MGS
701 Details : When MGC is disconnected from MGS long enough, MGS will evict the
702 MGC, and late on MGC cannot successfully connect to MGS and a lot
703 of the error messages complaining that MGS is not connected.
707 Frequency : on network error
708 Description: panic with double free request if network error
709 Details : mdc_finish_enqueue is finish request if any network error ocuring,
710 but it's true only for synchronus enqueue, for async enqueue
711 (via ptlrpcd) this incorrect and ptlrpcd want finish request
714 Severity : enhancement
716 Description: client-side metadata stat-ahead during readdir(directory readahead)
717 Details : perform client-side metadata stat-ahead when the client detects
718 readdir and sequential stat of dir entries therein
721 Frequency : on start mds
723 Description: Implement get_info(last_id) in obdfilter.
726 Frequency : occasional
728 Description: Correctly check stale fid, not start epoch if ost not support SOM
729 Details : open with flag O_CREATE need set old fid in op_fid3 because
730 op_fid2 was overwritten with new generated fid, but mds can answer
731 with one of these two fids and both is not stale. Setattr
732 incorrectly started an epoch and assume will be called
733 done_writing, but without SOM done_writing ever being called.
736 Frequency : rare, depends on device drivers and load
738 Description: MDS or OSS nodes crash due to stack overflow
739 Details : Code changes in 1.8.0 increased the stack usage of some functions.
740 In some cases, in conjunction with device drivers that use a lot
741 of stack the MDS (or possibly OSS) service threads could overflow
742 the stack. One change which was identified to consume additional
743 stack has been reworked to avoid the extra stack usage.
746 Frequency : occasional
748 Description: Do not fail import if osc_interpret_create gets -EAGAIN
749 Details : If osc_interpret_create got -EAGAIN it immediately exits and
750 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
751 oscc_has_objects and see OSC has no objests and call
752 oscc_internal_create to resend create request.
754 Severity : enhancement
756 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
758 Severity : enhancement
760 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
763 Frequency : very rare
765 Description: Fix replay if there is an un-replied request and open
766 Details : In some cases, older replay request will revert the
767 mcd->mcd_last_xid on MDS which is used to record the client's
770 Severity : enhancement
772 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
774 Severity : enhancement
776 Description: Add rhel5 support to HEAD.
778 Serverity : enhancement
780 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
785 Description: Don't allow skipping OSTs if index has been specified.
786 Details : Don't allow skipping OSTs if index has been specified, make
787 locking in internal create lots better.
791 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
792 Details : access to bitfield in structure is always rounded to long
793 and this produce problem with not atomic change any bit.
797 Description: Lustre make rpms failed.
798 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
799 builds Lustre rpms from tarball.
802 Frequency : rare on shutdown ost
804 Description: If llog cancel was not send before clean_exports phase, this can
805 produce deadlock in llog code.
806 Details : If llog thread has last reference to obd and call class_import_put
807 this produce deadlock because llog_cleanup_commit_master wait when
808 last llog_commit_thread exited, but this never success because was
809 called from llog_commit_thread.
813 Description: allow userland application know is lost one of stripes.
814 Details : fill lvb_blocks with error code on ost and return it to
815 application if error flag found.
819 Description: NULL lov_tgts causing MDS oops
820 Details : more safe checks for NULL lov_tgts for avoid oops.
822 Severity : enhancement
824 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
828 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
829 Details : If an OST is down the MDS will hang indefinitely in
830 obd_statfs_rqset() waiting for the statfs data. While for
831 MDS QOS usage of statfs, it should not stuck in waiting.
833 Severity : enhancement
835 Description: remote_acl support
836 Details : Support ACL-based permission check for remote user.
837 Support setfacl/getfacl for remote user with the utils
838 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
839 the system "{s,g}etfacl" utils.
841 Severity : enhancement
843 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
845 Severity : enhancement
847 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
851 Description: make mgs_setparam() handle fsname containing dash
852 Details : fsname containing a dash does not work with lctl conf_param
854 Severity : enhancement
856 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
858 Severity : enhancement
860 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
864 Frequency : rare, at shutdown
865 Description: access already free / zero obd_namespace.
866 Details : if client_disconnect_export was called without force flag set,
867 and exist connect request in flight, this can produce access to
868 NULL pointer (or already free pointer) when connect_interpret
869 store ocd flags in obd_namespace.
873 Frequency : only at startup
874 Description: not alloc memory with spinlock held.
875 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
876 if any spinlock held.
878 Severity : enhancement
880 Description: make lustre randomly fail allocating memory
881 Details : Make lustre randomly failed allocating memory for testing purpose.
883 Severity : enhancement
885 Description: lost problems with lov objid file
886 Details : Fixes some scability and access to not inited memory problems
887 in work with lov objdid file.
892 Description: lfs find does not continue on file error
893 Details : Continue other files processing when a file/dir is absent.
897 Description: Inconsistent usage of lustre_pack_reply()
898 Details : Standardize the usage of lustre_pack_reply() such that it
899 always generate a CERROR on failure.
902 Frequency : occasional
904 Description: MDS hang or stay in waiting lock
905 Details : If client receive lock with CBPENDING flag ldlm need send lock
906 cancel as separate rpc, to avoid situation when cancel request
907 can't processed due all i/o threads stay in wait lock.
911 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
915 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
919 Description: Update to SLES9 kernel-2.6.5-7.287.3.
923 Description: Update to RHEL4 latest kernel.
925 Severity : enhancement
927 Description: Build SLES10 patchless client fails
928 Details : The configure was broken by run ./configure with
929 --with-linux-obj=.... argument for patchless client. When the
930 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
931 can't search header adequately. Use absolute path such as
932 -I($LINUX)/include instead.
936 Description: interrupt oig_wait produce painc on resend.
937 Details : brw_redo_request can be used for resend requests from ptlrpcd and
938 private set, and this produce situation when rq_ptlrpcd_data not
939 copyed to new allocated request and triggered LBUG on assert
940 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
941 ptlrpcd set if request is changed and can be safety changed to use
946 Description: LASSERT_{REQ,REP}SWAB macros are buggy
947 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
948 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
949 these macros and replace them with their difinitions instead.
953 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
954 Details : Update which_patch & target file for SLES10 latest kernel.
957 Description: Remove client patches from SLES 10 kernel.
958 Details : This causes SLES 10 clients to behave as patchless clients
959 even on a Lustre-patched (server) kernel.
961 Severity : enhancement
963 Description: self-adjustable client's lru lists
964 Details : use adaptive algorithm for managing client cached locks lru
965 lists according to current server load, other client's work
966 pattern, memory activities, etc. Both, server and client
967 side namespaces provide number of proc tunables for controlling
970 Severity : enhancement
972 Description: light-weight GSS support
973 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
974 and reduce performance overhead.
976 Severity : enhancement
978 Description: Linux keyring support
979 Details : Support using service of Linux keyring for Lustre GSS internal
980 context refresh/cache mechanism.
984 Description: Fix errors in lfs documentation
985 Details : Fixes man pages
989 Description: don't use GFP_* in generic Lustre code.
990 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
994 Description: obdclass is limited by single OBD_ALLOC(idarray)
995 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
999 Description: warnings with build patchless client with vanila 2.6.19 and up
1000 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1001 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1005 Frequency : Only for SLES
1006 Description: sanity_quota fail test_1
1007 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1008 file, which makes TSTID[2] inunique.
1012 Frequency : Only for SLES9
1013 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1014 Details : Update target/ChangeLog/which_patch .
1018 Frequency : Only for test_10 in sanity-quota.sh
1019 Description: a bug in quota test code
1020 Details : checking a test flag in a wrong place
1024 Frequency : Only for test_13 in sanity-quota.sh
1025 Description: a bug in quota test script
1026 Details : shouldn't check overfull allocated quota
1030 Description: jbd statistics
1031 Details : Port older jbd statistics patch for sles10
1033 Severity : enhancement
1035 Description: Add printing inode info into message about error in writepage.
1040 Description: reply_lock_interpret crash due to race with it and lock cancel.
1041 Details : Do not replay locks that are being cancelled. Do not reference
1042 locks by their address during replay, just by their handle.
1046 Frequency : When flocks are used.
1047 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1048 Details : Flock locks might destroy just granted lock if it could be merged
1049 with another existing flock, this is done in completion handler,
1050 so teach ldlm_cli_enquque_fini that this is a valid case for
1056 Description: Oops in read and write path when failing to allocate lock.
1057 Details : Check if lock allocation failed and return error back.
1061 Description: lstripe command fails for valid OST index
1062 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1063 instead of lov->desc.ld_active_tgt_count.
1067 Description: sanity.sh failed test 103
1068 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1073 Description: df doesn't work properly if diskfs blocksize != 4K
1074 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1078 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1079 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1081 Severity : enhancement
1083 Description: Adaptive timeouts
1084 Details : RPC timeouts adapt to changing server load and network
1085 conditions to reduce resend attempts and improve recovery time.
1089 Description: llapi_file_create() does not allow some changes
1090 Details : add llapi_file_open() that allows specifying the mode and
1091 open flags, and also returns an open file handle.
1095 Description: merge and cleanup kernel patches.
1096 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1100 Description: Add journal checksum support.(Kernel part)
1101 Details : The journal checksum feature adds two new flags i.e
1102 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1103 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1104 indicates that the commit block contains the checksum for
1105 the blocks described by the descriptor blocks. Now commit
1106 record can be sent to disk without waiting for descriptor
1107 blocks to be written to disk. This behavior is controlled
1108 using JBD2_FEATURE_ASYNC_COMMIT flag.
1112 Description: OSS needs mutliple precreate threads
1113 Details : Add ability to start more than one create thread per OSS.
1117 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1118 Details : execute lfs setstripe on client
1122 Description: mds_obd_create error creating tmp object
1123 Details : When the user sets quota on root, llog will be affected and can't
1124 create files and write files.
1127 Frequency : Always on ia64 patchless client, and possibly others.
1129 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1130 Details : This allows the patchless client to be loaded on architectures
1131 without this export.
1135 Description: RedHat Update kernel for RHEL5
1136 Details : Add the kernel config file for RHEL5.
1140 Description: RedHat Update kernel for RHEL5
1141 Details : Modify the kernel config file more closer RHEL5.
1145 Description: Build failure against Centos5 (RHEL5)
1146 Details : Define PAGE_SIZE when it isn't present.
1150 Description: test_7 on sanity-quota.sh can't be used on separate machines
1151 Details : add facet to handle it
1155 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1156 Details : with NFS, the anon dentry's parent was set to itself in
1157 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1158 corresponding dentry other than use rec->ur_name.
1160 Severity : enhancement
1162 Description: lfs setstripe enhancement
1163 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1167 Description: enable data checksumming by default
1168 Details : enable checksum by default, allow --disable-checksum
1169 configure option and "-o nochecksum" mount option. Checksums
1170 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1174 Description: lustre support for RHEL5
1175 Details : Add support for RHEL5.
1179 Description: Client eviction due to failover config
1180 Details : after a connection loss, the lustre client should attempt to
1181 reconnect to the last active server first before trying the
1182 other potential connections.
1186 Description: when mds and osts use different quota unit(32bit and 64bit),
1187 quota will be released repeatly.
1188 Details : void sending multiple quota reqs to mds, which will keep the
1189 status between the reqs.
1193 Description: osts not allocated evenly to files
1194 Details : change the condition to increase offset_idx
1197 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1199 Description: Data corruption for OSTs that are formatted larger than 2TB
1201 Details : When generating the bio request for lustre file writes the
1202 sector number would overflow a temporary variable before being
1203 used for the IO. The data reads correctly from Lustre (which
1204 will overflow in a similar manner) but other file data or
1205 filesystem metadata may be corrupted in some cases.
1209 Description: Tune the kernel for good SCSI performance.
1210 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1211 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1216 Description: rewrite ext2-derived code in obdclass/uuid.c
1217 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1218 from scratch preserving functionality.
1222 Description: "lfs find -obd UUID" prints directories
1223 Details : "lfs find -obd UUID" will return all directory names instead
1224 of just file names. It is incorrect because the directories
1225 do not reside on the OSTs.
1229 Description: use i_size_read and i_size_write in 2.6 port
1230 Details : replace inode->i_size access with i_size_read/write()
1233 Frequency : when removing large files
1235 Description: scheduling issue during removal of large Lustre files
1236 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1237 It causes scheduling issues when removing large files (17TB in the
1241 Frequency : only with liblustre clients on XT3
1243 Description: evictions taking too long
1244 Details : allow llrd to evict clients directly on OSTs
1247 Frequency : only on ppc
1249 Description: /proc/fs/lustre/devices broken on ppc
1250 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1251 We should make sure the type of 'index' is loff_t and then cast
1252 explicitly as needed below. Do not assign an explicitly cast
1256 Frequency : only for rhel5
1258 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1259 Details : Modify the target file & which_kernel.
1261 Severity : enhancement
1263 Description: omit set fsid for export NFS
1264 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1265 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1269 Frequency : Always, for kernels after 2.6.16
1270 Description: Fix warning idr_remove called for id=.. which is not allocated.
1271 Details : Last kernels save old s_dev before kill super and not allow
1272 to restore from callback - restore it before call kill_anon_super.
1276 Description: buffer overruns could theoretically occur
1277 Details : llapi_semantic_traverse() modifies the "path" argument by
1278 appending values to the end of the origin string, and a
1279 overrun may occur. Adding buffer overrun check in liblustreapi.
1283 Description: Fix error on 'ls .' at the top of the Lustre mount.
1284 Details : Don't revalidate dentry if it is a root dentry.
1288 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1289 Details : Modify vm-tunables-rhel4.patch.
1293 Description: Kernel config for 2.6.18-vanilla.
1294 Details : Modify targets/2.6-vanilla.target.in.
1295 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1296 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1297 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1298 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1302 Description: improve handling recoverable errors
1303 Details : if request processig with error which can be recoverable on server
1304 request should be resend, otherwise page released from cache and
1309 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1310 Details : Modify target file & which_patch.
1311 A flaw was found in the IA32 system call emulation provided
1312 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1313 value could be stored in the %RAX register, which could trigger an
1314 out-of-bounds system call table access. An untrusted local user
1315 could exploit this flaw to run code in the kernel
1316 (ie a root privilege escalation). (CVE-2007-4573).
1320 Description: O_DIRECT bypasses client statistics.
1321 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1322 read_ahead_stats were updated. copied the stats section from
1323 osc_send_oap_rpc() into async_internal().
1327 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1331 Description: change order of libsysio includes
1332 Details : '#include sysio.h' should always come before '#include xtio.h'
1336 Description: handle "serious error: objid * already exists" more gracefully
1337 Details : If LAST_ID value on disk is smaller than the objects existing in
1338 the O/0/d* directories, it indicates disk corruption and causes an
1339 LBUG(). If the object is 0-length, then we should use the existing
1340 object. This will help to avoid a full fsck in most cases.
1342 Severity : enhancement
1344 Description: adapt the lustre_config script to support the upgrade case
1345 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1346 server targets to 1.6 in parallel.
1350 Description: To avoid grant space > avaible space when the disk is almost
1351 full. Without this patch you might see the error "grant XXXX >
1352 available" or some LBUG about grant, when the disk is almost
1354 Details : In filter_check_grant, for non_grant cache write, we should
1355 check the left space by if (*left > ungranted + bytes), instead
1356 of (*left > ungranted), because only we are sure the left space
1357 is enough for another "bytes", then the ungrant space should be
1358 increase. In client, we should update cl_avail_grant only there
1359 is OBD_MD_FLGRANT in the reply.
1363 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1364 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1365 described in CVE-2007-4573. This problem could allow untrusted
1366 local users to gain root access.
1369 Frequency : when using O_DIRECT and quotas
1371 Description: Incorrect file ownership on O_DIRECT output files
1372 Details : block usage reported by 'lfs quota' does not take into account
1373 files that have been written with O_DIRECT.
1378 Description: touch file failed when fs is not full
1379 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1380 otherwise we can get ENOSP while fs is not full.
1384 Description: parallel lock callbacks
1385 Details : Instead of sending blocking and completion callbacks as separated
1386 requests, adding them to a set and sending in parallel.
1389 Frequency : only for Cray XT3
1390 Bugzilla : 12829/13455
1391 Description: Changing primary group doesn't change the group lustre assigns to
1393 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1394 overridden with the primary group provided by the group upcall,
1395 whereas the supplied fsgid can be trusted if it is in the list of
1396 supplementary groups returned by the group upcall.
1398 Severity : enhancement
1400 Description: Allow masking D_WARNING, D_ERROR messages from console
1401 Details : Console messages can now be disabled via lnet.printk.
1405 Description: User code with malformed file open parameter crashes client node
1406 Details : Before packing join_file req, all the related reference should be
1407 checked carefully in case some malformed flags cause fake
1408 join_file req on client.
1412 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1413 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1414 mode is equal to the granted mode. However, it should also check
1415 that LDLM_FL_CP_REQD is not set.
1419 Description: Heavy nfs access might result in deadlocks
1420 Details : After ELC code landed, it is now improper to enqueue any mds
1421 locks under och_sem, because enqueue might want to decide to
1422 cancel open locks for same inode we are holding och_sem for.
1426 Description: Client eviction while running blogbench
1427 Details : A lot of unlink operations with concurrent I/O can lead to a
1428 deadlock causing evictions. To address the problem, the number of
1429 oustanding OST_DESTROY requests is now throttled to
1430 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1431 callbacks are processed in priority.
1435 Description: enable ACLs on MDS by default
1436 Details : ACLs must be enabled on MDS by default.
1439 Frequency : PPC/PPC64 only
1441 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1442 Details : fix duplicated definitions between asm-ppc64/types.h and
1443 lustre_types.h on PPC.
1446 Frequency : PPC/PPC64 only
1448 Description: asm-ppc/segment.h does not exist
1449 Details : fix compile issue on PPC.
1453 Description: better handle error messages in extents code
1456 Frequency : RHEL4 only
1458 Description: mkfs is very slow on IA64/RHEL4
1459 Details : A performance regression has been discovered in the MPT Fusion
1460 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1461 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1462 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1464 Severity : enhancement
1466 Description: SNMP support enhancement
1467 Details : Adding total number of sampled request for an MDS node in snmp
1470 Severity : enhancement
1472 Description: Optimize ldlm waiting list processing for PR extent locks
1473 Details : When processing waiting list for read extent lock and meeting read
1474 lock that is same or wider to it that is not contended, skip
1475 processing rest of the list and immediatelly return current
1476 status of conflictness, since we are guaranteed there are no
1477 conflicting locks in the rest of the list.
1481 Description: Time out and refuse to reconnect
1482 Details : When the failover node is the primary node, it is possible
1483 to have two identical connections in imp_conn_list. We must
1484 compare not conn's pointers but NIDs, otherwise we can defeat
1485 connection throttling.
1489 Description: port llog fixes from b1_6 into HEAD
1490 Details : Port llog reference couting and some llog cleanups from b1_6
1491 (bug 10800) into HEAD, for protect from panic and access to already
1492 free llog structures.
1496 Description: Detect stride IO mode in read-ahead
1497 Details : When a client does stride read, read-ahead should detect that and
1498 read-ahead pages according to the detected stride pattern.
1502 Description: data checksumming impacts single node performance
1503 Details : add support for several checksum algorithm. Currently, only CRC32
1504 and Adler-32 are supported. The checksum type can be changed on
1505 the fly via /proc/fs/lustre/osc/*/checksum_type.
1509 Description: use adler32 for page checksums
1510 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1515 Description: build for x2 fails
1516 Details : fix compile issue on Cray systems.
1520 Description: Properly match for duplicate locks
1521 Details : Due to different lock order from skiplists code, we need to
1522 traverse entire list for now
1525 Frequency : only on PPC/SLES10
1527 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1528 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1529 be defined. Add a hack in mkfs_lustre.c to work around this compile
1534 Description: LASSERT on MDS when client holding flock lock dies
1535 Details : ldlm pool logic depends on number of granted locks equal to
1536 number of released locks which is not true for flock locks, so
1537 just exclude such locks from consideration.
1541 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1542 Details : Use fsfilt_sync() for both the whole filesystem sync and
1543 individual file sync to eliminate dangerous inode locking
1544 with I_LOCK that can lead to a deadlock.
1548 Description: performance in 1.6.3
1549 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1550 sure that 1MB requests are not fragmented by the block layer.
1552 Severity : enhancement
1554 Description: organize the server-side client stats on per-nid basis
1555 Details : Change the structure of stats under obdfilter and mds to
1563 The "uuid"s file would list the uuids of _active_ exports.
1564 And the clear entry is to clear all stats and stale nids.
1566 Severity : enhancement
1568 Description: eliminate client locks in face of contention
1569 Details : file contention detection and lockless i/o implementation
1570 for contended files.
1574 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1577 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1579 Description: MDS may lose file striping (and hence file data) in some cases
1580 Details : If there are additional extended attributes stored on the MDS,
1581 in particular ACLs, SELinux, or user attributes (if user_xattr
1582 is specified for the client mount options) then there is a risk
1583 of attribute loss. Additionally, the Lustre file striping
1584 needs to be larger than default (e.g. striped over all OSTs),
1585 and an additional attribute must be stored initially in the
1586 inode and then increase in size enough to be moved to the
1587 external attribute block (e.g. ACL growing in size) for file
1592 Description: skiplist implementation simplification
1593 Details : skiplists are used to group compatible locks on granted list
1594 that was implemented as tracking first and last lock of each
1595 lock group the patch changes that to using doubly linked lists
1599 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1600 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1601 because LASSERT on both the data supplied by a client, and the
1602 data on disk is dangerous and incorrect.
1604 Severity : enhancement
1606 Description: Slow truncate/writes to huge files at high offsets.
1607 Details : Directly associate cached pages to lock that protect those pages,
1608 this allows us to quickly find what pages to write and remove
1609 once lock callback is received.
1613 Description: more ldlm soft lockups
1614 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1615 starve other threads from the resource lock for a long time in
1616 case of long waiting queue, so change the debug level from
1617 D_OTHER to the less frequently used D_INFO.
1619 Severity : enhancement
1621 Description: add -gid, -group, -uid, -user options to lfs find
1625 Description: Hung threads in invalidate_inode_pages2_range
1626 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1627 once one is completed. As a result, some RPCs are stuck in the
1628 queue and are never sent.
1632 Description: filter threads hungs on waiting journal commit
1633 Details : Cleanup filter group llog code, then only filter group llog will
1634 be only created in the MDS/OST syncing process.
1638 Description: Procfs and llog threads access destoryed import sometimes.
1639 Details : Sync the import destoryed process with procfs and llog threads by
1640 the import refcount and semaphore.
1642 Severity : enhancement
1644 Description: openlock cache of b1_6 port to HEAD
1649 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1650 Details : If insertion of an extent fails, then discard the inode
1651 preallocation and free data blocks else it can lead to duplicate
1656 Description: don't always update ctime in ext3_xattr_set_handle()
1657 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1658 In some cases the ctime should not be updated, for example for
1659 2.0->1.8 compatibility it is necessary to delete an xattr and it
1660 should not update the ctime.
1664 Bugzilla : 15713/16362
1665 Description: Assertion in iopen_connect_dentry in 1.6.3
1666 Details : looking up an inode via iopen with the wrong generation number can
1667 populate the dcache with a disconneced dentry while the inode
1668 number is in the process of being reallocated. This causes an
1669 assertion failure in iopen since the inode's dentry list contains
1670 both a connected and disconnected dentry.
1674 Description: assertion failure in ldlm_handle2lock()
1675 Details : fix a race between class_handle_unhash() and class_handle2object()
1676 introduced in lustre 1.6.5 by bug 13622.
1681 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1682 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1687 Description: oops in sd_iostats_seq_show()
1688 Details : unloading/reloading the scsi low level driver triggers a kernel
1689 bug when trying to access the sd iostat file.
1694 Description: Kernel panics during QLogic driver reload
1695 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1696 patch, causing memory corruption.
1701 Description: journal_dev option does not work in b1_6
1702 Details : pass mount option during pre-mount.
1704 Severity : enhancement
1706 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1707 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1708 extent information of a file. It can be used to map logical blocks
1709 in a file to physical blocks in the block device.
1713 Description: LDLM soft lockups - improvement
1714 Details : It is be possible to send the lock handle along with each read
1715 or write request because the client is already doing a lock match
1716 itself so there isn't any reason the OST should have to re-do that
1720 Frequency : only on Cray X2
1722 Description: X2 build failures
1723 Details : fix build failures on Cray X2.
1727 Description: xid & resent requests
1728 Details : Initialize RPC XID from clock at startup (randomly if clock is
1731 Severity : enhancement
1733 Description: Add lustre_start utility to start or stop multiple Lustre servers
1738 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1739 Details : In case of memory pressure, list_del() can be called twice on
1740 req->rq_history_list, causing a kernel oops.
1744 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1745 Details : fix stack overflow in the distributed lock manager by defering
1746 export eviction after a failed AST to the elt thread instead of
1747 handling it in the dlm interpret routine.
1751 Description: Convert some comments to new format.
1752 Details : Update documenting comments to match doxygen conventions.
1756 Description: Grammar fixes.
1757 Details : A couple of trivial spelling fixes.
1761 Description: OSD_COUNTERS-mandatory
1762 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1763 to simplify the code.
1767 Description: simplify cmm_device freeing logic.
1768 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1772 Description: Add lockdep support to dt_object_operations locking interface.
1773 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1774 parameter indicating lock ordering. Update mdd code to use new
1779 Description: Introduce failloc constants for lockless IO tests.
1780 Details : Add two new failloc constants to test lockless IO. Only one of
1781 them in implemented---another is checked in yet to be landed
1786 Description: Add lockdep support for inode mutex.
1787 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1788 in the situations where Lustre has to lock more than one inode
1793 Description: Add optional invariants checking support.
1794 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1795 configure switch. This macro is to be used for consistency and
1796 sanity checks that are too expensive to be left in `production'
1801 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1802 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1807 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1808 Details : Use spin_lock_nested() in (the only) situation where more than
1809 one ldlm_lock is locked simultaneously. Also, fix possible
1810 dead-lock in ldlm_lock_change_resource() by enforcing particular
1815 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1816 Details : Instead of specifying each ldlm_lock call-back through separate
1817 parameter, wrap them into struct ldlm_callback_suite.
1821 Description: Kill join_lru obd method and its callers.
1822 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1823 in memory---a requirement implemented through ->o_join_lru() obd
1824 method in HEAD. Remove this method and its users.
1828 Description: Add asynchronous ldlm ENQUEUE completion handler.
1829 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1830 case that stock ldlm_completion_ast() cannot handle as it waits
1831 until lock is granted. Introduce new ldlm_completion_ast_async()
1832 for this. Also comment ldlm_completion_ast().
1836 Description: ldlm_error <-> errno conversion.
1837 Details : Add functions to map (rather arbitrary) between LDLM error codes
1838 and standard errno values. CLIO needs this to prevent LDLM specific
1839 constants from escaping ldlm and osc.
1843 Description: Kill unused ldlm_handle2lock_ns() function.
1844 Details : Kill unused ldlm_handle2lock_ns() function.
1848 Description: Add lu_ref support to ldlm_lock
1849 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1850 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1851 and ldlm_resource. LDLM interface has to be changed, because code that
1852 releases a reference on a lock, has to "know" what reference this is.
1853 In the most frequent case
1855 lock = ldlm_handle2lock(handle);
1857 LDLM_LOCK_PUT(lock);
1859 no changes are required. When any other reference (received _not_ from
1860 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1861 instead of LDLM_LOCK_PUT().
1863 Arguably, changes are pervasive, and interface requires some discipline
1864 for proper use. On the other hand, it was very instrumental in finding
1865 a few leaked lock references.
1869 Description: Add ldlm_lock_addref_try().
1870 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1871 attempts to addref a lock that might be being canceled
1876 Description: Add ldlm_weigh_callback().
1877 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1878 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1879 hard-coded `number of pages' logic.
1883 Description: Add lockdep annotations to llog code.
1884 Details : Use appropriately tagged _nested() locking calls in the places
1885 where llog takes more than one ->lgh_lock lock.
1889 Description: Add loi_kms_set().
1890 Details : Wrap kms updates into a helper function.
1894 Description: Constify instances of struct lsm_operations.
1895 Details : Constify instances of struct lsm_operations.
1899 Description: lu_conf support.
1900 Details : On a server, a file system object is uniquely identified
1901 by a fid, which is sufficient to locate and load all object
1902 state (inode). On a client, on the other hand, more data are
1903 necessary instantiate an object. Change lu_object_find() and
1904 friends to take additional `lu_conf' argument describing object.
1905 Typically this includes layout information.
1909 Description: lu_context fixes.
1910 Details : Introduce new lu_context functions that are needed on the client
1911 side, where some system threads (ptlrpcd) are shared by multiple
1912 modules, and so cannot be stopped during module shutdown.
1916 Description: Add start and stop methods to lu_device_type_operations.
1917 Details : Introduce two new methods in lu_device_type_operations, that are
1918 invoked when first instance of a given type is created and last one
1919 is destroyed respectively. This is need by CLIO.
1923 Description: Add lu_ref support to struct lu_device.
1924 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1925 track leaked references.
1929 Description: Introduce lu_kmem_descr.
1930 Details : lu_kmem_descr and its companion interface allow to create
1931 and destroy a number of kmem caches at once.
1935 Description: Fix lu_object finalization race.
1936 Details : Fix a race between lu_object_find() finding an object and its
1937 concurrent finalization. This race is (most likely) not possible
1938 on the server, but might happen on the client.
1942 Description: Introduce lu_ref interface.
1943 Details : lu_ref is a debugging module allowing to track references to
1944 a given object. It is quite cpu expensive, and has to be
1945 explicitly enabled with --enable-lu_ref. See usage description
1950 Description: Factor lu_site procfs stats into a separate function.
1951 Details : Separate lu_site stats printing code into a separate function
1952 that can be reused on a client.
1956 Description: Constify instances of struct {lu,dt,md}_device_operations.
1957 Details : Constify instances of struct {lu,dt,md}_device_operations.
1961 Description: Introduce struct md_site and move meta-data specific parts of
1962 struct lu_site here.
1963 Details : Move md-specific fields out of struct lu_site into special struct
1964 md_site, so that lu_site can be used on a client.
1968 Description: Kill mdd_lov_destroy().
1969 Details : Remove unused mdd code.
1973 Description: Add st_block checking to multistat.c.
1974 Details : Add st_block checking to multistat.c.
1978 Description: Add lu_ref support to struct obd_device.
1979 Details : Add lu_ref tracking to obd_device.
1983 Description: Kill obd_set_fail_loc().
1984 Details : Remove unused code.
1988 Description: Add special type for ptlrpc_request interpret functions.
1989 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1990 there. Actual usage will be in CLIO.
1994 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1995 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1996 doesn't work with lockdep.
2000 Description: Add rwv.c test program.
2001 Details : New testing program exercising readv(2) and writev(2) (Qian).
2005 Description: Add sendfile.c test program.
2006 Details : New testing program exercising sendfile(2) (Jay).
2010 Description: Ratelimit a message that can be very frequent.
2011 Details : Ratelimit a memory allocation failure message that can
2016 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2017 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2021 Description: Liblustre build fixes.
2022 Details : Liblustre build fixes.
2026 Description: libcfs: add cfs_{need,cond}_resched() interface.
2027 Details : libcfs: add cfs_{need,cond}_resched() definition and
2028 implementations for Linux, NT, and liblustre.
2030 Severity : enhancement
2032 Description: More exported tunables for mballoc
2033 Details : Add support for tunable preallocation window and new tunables for
2034 large/small requests
2038 Description: Detect corruption of block bitmap and checking for preallocations
2039 Details : Checks validity of on-disk block bitmap. Also it does better
2040 checking of number of applied preallocations. When corruption is
2041 found, it turns filesystem readonly to prevent further corruptions.
2045 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2046 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2047 window to 0. If it does hit cache miss, and read-pattern is still
2048 stride-io mode, does not reset the stride window, but also does not
2049 increase the stride window length in this case.
2053 Frequency : only for big-endian servers
2054 Description: Check if system is big-endian while mounting fs with extents feature
2055 Details : Mounting a filesystem with extents feature will fail on big-endian
2056 systems since ext3-based ldiskfs is not supported on big-endian
2057 systems. This can be over-riden with "bigendian_extents" mount option.
2059 Severity : enhancement
2061 Description: The root squash functionality
2062 Details : A security feature, which is to prevent users from being able
2063 to mount lustre on their desktop, run as root, and delete
2064 all of the files in the filesystem. The goal is accomplished by
2065 remapping user id (UID) and group id (GID) of the root user to
2066 a UID and GID specified by the system administartor via Lustre
2067 configuration management server (MGS). The functionality also
2068 allows to specify sets of clients for which the remapping does
2073 Description: Excessive recovery window
2074 Details : With AT enabled, the recovery window can be excessively long (6000+
2075 seconds). To address this problem, we no longer use
2076 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2077 timeout no longer depends on the service time, it is set to
2078 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2079 time via pb_service_time.
2083 Description: Watchdog triggered on MDS failover
2084 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2085 the OSTs know that the MDS "UUID" can be reused for the same export
2086 from a different NID, so we do not need to wait for the export to be
2090 Frequency : rare, only if using MMP with Linux RAID
2092 Description: MMP doesn't work with Linux RAID
2093 Details : While using HA for Lustre servers with Linux RAID, it is possible
2094 that MMP will not detect multiple mounts. To make this work we
2095 need to unplug the device queue in RAID when the MMP block is being
2096 written. Also while reading the MMP block, we should read it from
2097 disk and not the cached one.
2099 Severity : enhancement
2101 Description: open file using fid
2102 Details : A file can be opened using just its fid, like
2103 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2106 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2109 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2110 Details : add proper locking for creation of jbd_4k slab cache
2114 Description: MMP check in ext3_remount() fails without displaying any error
2115 Details : When multiple mount protection fails during remount, proper error
2118 Severity : enhancement
2120 Description: Allow stripe size to be up to 4G-64k
2121 Details : Fix math logic to allow large stripe sizes.
2125 Description: add check for >8TB ldiskfs filesystems
2126 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2127 Don't allow >8TB ldiskfs filesystems to be mounted without
2128 force_over_8tb mount option
2130 --------------------------------------------------------------------------------
2132 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2134 * Support for kernels:
2135 2.6.5-7.283 (SLES 9),
2136 2.6.9-55.EL (RHEL 4),
2137 2.6.16.46-0.14 (SLES 10),
2138 2.6.18.8 vanilla (kernel.org)
2139 * Client support for unpatched kernels:
2140 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2141 2.6.16 - 2.6.22 vanilla (kernel.org)
2142 * Due to recently discovered recovery problems, we do not recommend
2143 using patchless RHEL 4 clients with this or any earlier release.
2144 * Recommended e2fsprogs version: 1.39.cfs8
2145 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2146 * Starting with this release, the ldiskfs backing filesystem required
2147 by Lustre is now in its own package, lustre-ldiskfs. This package
2148 should be installed. It is versioned separately from Lustre and
2149 may be released separately in future.
2153 Description: block reactivating mgc import until all deactivates complete
2154 Details : Fix race when failing back MDT/MGS to itself (testing)
2157 Frequency : at statup only
2159 Description: mds_lov_synchronize race leads to various problems
2160 Details : simultaneous MDT->OST connections at startup can cause the
2161 sync to abort, leaving the OSC in a bad state.
2163 Severity : enhancement
2165 Description: add optional extra BUILD_VERSION info
2166 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2167 to override the lustre version.
2170 Frequency : 2.6.18 servers only
2172 Description: ll_kern_mount() doesn't release the module reference
2173 Details : The ldiskfs module reference count never drops down to 0
2174 because ll_kern_mount() doesn't release the module reference.
2179 Description: server LBUG when using old ost_num_threads parameter
2180 Details : Accept the old ost_num_threads parameter but warn that it
2181 is deprecated, and fix an off-by-one error that caused an LBUG.
2186 Description: Transient SCSI error results in persistent IO issue
2187 Details : iobuf->dr_error is not reinitialized to 0 between two
2191 Frequency : sometimes when underlying device returns I/O errors
2193 Description: OSTs not going read-only during write failures
2194 Details : OSTs are not remounted read-only when the journal commit threads
2195 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2196 instead of the ext3 wrappers.
2200 Description: poor connect scaling with increasing client count
2201 Details : Don't run filter_grant_sanity_check for more than 100 exports
2202 to improve scaling for large numbers of clients.
2205 Frequency : SLES10 only
2207 Description: sanity-quota.sh quotacheck failed: rc = -22
2208 Details : Quotas cannot be enabled on SLES10.
2211 Frequency : liblustre clients only
2213 Description: getdirentries does not give error when run on compute nodes
2214 Details : getdirentries does not fail when the size specified as an argument
2215 is too small to contain at least one entry
2217 Severity : enhancement
2219 Description: Add LNET router traceability for debug purposes
2220 Details : If a checksum failure occurs with a router as part of the
2221 IO path, the NID of the last router that forwarded the bulk data
2222 is printed so it can be identified.
2227 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2228 Details : Due to a race condition, liblustre clients were occasionally
2229 evicted incorrectly.
2231 Severity : enhancement
2233 Description: lfs setstripe use optional parameters instead of postional
2236 Severity : enhancement
2238 Description: Nanosecond timestamp support for ldiskfs
2239 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2240 resolution timestamps. There is not yet support for this at
2241 the Lustre filesystem level.
2244 Frequency : during server recovery
2246 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2247 Details : request with rq_no_resend flag not awake l_wait_event if they get
2251 Frequency : nfs export on patchless client
2253 Description: connectathon hang when test nfs export over patchless client
2254 Details : Disconnected dentry cannot be found with lookup, so we do not need
2255 to unhash it or make it invalid
2258 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2259 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2260 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2265 Description: open req refcounting wrong on reconnect
2266 Details : If reconnect happened between getting open reply from server and
2267 call to mdc_set_replay_data in ll_file_open, we will schedule
2268 replay for unreferenced request that we are about to free.
2269 Subsequent close will crash in variety of ways.
2270 Check that request is still eligible for replay in
2271 mdc_set_replay_data().
2276 Description: disable writes to filesystem when reading health_check file
2277 Details : the default for reading the health_check proc file has changed
2278 to NOT do a journal transaction and write to disk, because this
2279 can cause reads of the /proc file to hang and block HA state
2280 checking on a healthy but otherwise heavily loaded system. It
2281 is possible to return to the previous behaviour during configure
2282 with --enable-health-write.
2284 Severity : enhancement
2286 Description: 64-bit inode version
2287 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2288 to the inode. This will be required for version-based recovery.
2293 Description: MDS fails to start if a duplicate client export is detected
2294 Details : in some rare cases it was possible for a client to connect to
2295 an MDS multiple times. Upon recovery the MDS would detect this
2296 and fail during startup. Handle this more gracefully.
2298 Severity : enhancement
2300 Description: Add -o localflock option to simulate old noflock
2302 Details : This will achieve local-only flock/fcntl locks
2308 Description: log_commit_thread vs filter_destroy race leads to crash
2309 Details : Take import reference before releasing llog record semaphore
2314 Description: Wrong request locking in request set processing
2315 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2316 imp_delayed_list, in this place should be used imp_lock.
2319 Frequency : when reconnection
2321 Description: Grant Leak when osc reconnect to OST
2322 Details : When osc reconnect ost, OST(filter) should check whether it
2323 should grant more space to client by comparing fed_grant and
2324 cl_avail_grant, and return the granted space to client instead
2325 of "new granted" space, because client will call osc_init_grant
2326 to update the client grant space info.
2329 Frequency : when client reconnect to OST
2331 Description: Grant Leak when osc do resend and replay bulk write
2332 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2333 bulk write request, because the grant info while be sync between
2334 OSC and OST when reconnect, and we should ignore the grant info
2335 these of resend/replay write req.
2340 Description: Grant space more than avaiable left space sometimes.
2341 Details : When then OST is about to be full, if two bulk writing from
2342 different clients came to OST. Accord the avaliable space of the
2343 OST, the first req should be permitted, and the second one
2344 should be denied by ENOSPC. But if the seconde arrived before
2345 the first one is commited. The OST might wrongly permit second
2346 writing, which will cause grant space > avaiable space.
2349 Frequency : when client is evicted
2351 Description: Grant might be wrongly erased when osc is evicted by OST
2352 Details : when the import is evicted by server, it will fork another
2353 thread ptlrpc_invalidate_import_thread to invalidate the
2354 import, where the grant will be set to 0. While the original
2355 thread will update the grant it got when connecting. So if
2356 the former happened latter, the grant will be wrongly errased
2357 because of this race.
2362 Description: Checking Stale with correct fid
2363 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2364 to check whether it is stale, because sometimes, we want the
2365 enqueue happened anyway, and op_data.fid2 will not be initialized.
2367 Severity : enhancement
2369 Description: update patchless client
2370 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2373 Frequency : only with 2.4 kernel
2375 Description: random memory corruption
2376 Details : size of struct ll_inode_info is to big for union inode.u and this
2377 can be cause of random memory corruption.
2382 Description: Memory leak in recovery
2383 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2384 It should also check obd_fail before fsfilt_start, otherwise if
2385 fsfilt_start return -EROFS,(failover mds during mds recovery).
2386 then the req will return with repmsg->transno = 0 and rc = EROFS.
2387 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2388 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2389 be freed no matter whether fsfilt_commit success or not.
2392 Frequency : only with huge count clients
2394 Description: Prevents from taking the superblock lock in llap_from_page for
2396 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2397 for llap_from_page call in ll_removepage prevents from taking the
2398 superblock lock for a soon died page.
2403 Description: Not check open intent error before release open handle
2404 Details : in some rare cases, the open intent error is not checked before
2405 release open handle, which may cause
2406 ASSERTION(open_req->rq_transno != 0), because it tries to release
2407 the failed open handle.
2412 Description: Set cat log bitmap only after create log success.
2413 Details : in some rare cases, the cat log bitmap is set too early. and it
2414 should be set only after create log success.
2418 Description: Accessing a block bevice can re-enable I/O when Lustre is
2419 tearing down a device.
2420 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2425 Description: service threads may hog cpus when there are a lot of requests
2426 Details : Insert cond_resched to give other threads a chance to use some CPU
2431 Description: the cat log was not initialized in recovery
2432 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2433 unlink log on mds will not be initialized until mds post
2434 recovery. And also in mds post recovery, the unlink log will
2435 initialization will be done asynchronausly, so there will be race
2436 between add unlink log and unlink log initialization.
2440 Description: brw_stats were being printed incorrectly
2441 Details : brw_stats were being printed as log2 but all of them were not
2442 recorded as log2. Also remove some code duplication arising from
2443 filter_tally_{read,write}.
2447 Frequency : rare, only in recovery.
2448 Description: ASSERTION(req->rq_type != LI_POISON) failed
2449 Details : imp_lock should be held while iterating over imp_sending_list for
2450 prevent destroy request after get timeout in ptlrpc_queue_wait.
2454 Description: replay-single.sh test 52 fails
2455 Details : A lock's skiplist need to be cleanup when it being unlinked
2456 from its resource list.
2460 Description: Short directio read returns full requested size rather than
2462 Details : Direct I/O operations should return actual amount of bytes
2463 transferred rather than requested size.
2465 Severity : enhancement
2467 Description: metadata RPC reduction (e.g. for rm performance)
2468 Details : decrease the amount of synchronous RPC between clients and servers
2469 by canceling conflicing lock before the operation on the client
2470 and packing thier handles into the main operation RPC to server.
2472 Severity : enhancement
2474 Description: Async OSC create to avoid the blocking unnecessarily.
2475 Details : If a OST has no remain object, system will block on the creating
2476 when need to create a new object on this OST. Now, ways use
2477 pre-created objects when available, instead of blocking on an
2478 empty osc while others are not empty. If we must block, we block
2479 for the shortest possible period of time.
2483 Description: improve handling recoverable errors
2484 Details : if request processig with error which can be recoverable on server
2485 request should be resend, otherwise page released from cache and
2488 Severity : enhancement
2490 Description: refine locking for avoid write wrong info into lov_objid file
2491 Details : fix possible races with add new target and write/update data in
2494 --------------------------------------------------------------------------------
2496 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2501 Frequency : on some architectures
2503 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2504 Details : Uninitialized flags sometimes cause configuration commands to
2508 Frequency : patchless clients only
2510 Description: missing __iget() symbol export
2511 Details : The __iget() symbol export is missing. To avoid the need for
2512 this on patchless clients the deathrow inode reaper is turned
2513 off, and we depend on the VM to clean up old inodes. This
2514 dependency was during via the fix for bug 12181.
2516 --------------------------------------------------------------------------------
2518 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2520 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2521 INTEROPERATE with older versions automatically. In many cases a
2522 special upgrade step is needed. Please read the
2523 user documentation before upgrading any part of a live system.
2524 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2525 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2526 * WARNING: Lustre configuration and startup changes are required with
2527 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2531 Severity : enhancement
2533 Description: MountConf
2534 Details : Lustre configuration is now managed via mkfs and mount
2535 commands instead of lmc and lconf. New obd types (MGS, MGC)
2536 are added for dynamic configuration management. See
2537 https://mail.clusterfs.com/wikis/lustre/MountConf for
2540 Severity : enhancement
2542 Description: dynamic OST addition
2543 Details : OSTs can now be added to a live filesystem
2545 Severity : enhancement
2547 Description: startup order invariance
2548 Details : MDTs and OSTs can be started in any order. Clients only
2549 require the MDT to complete startup.
2551 Severity : enhancement
2553 Description: parallel, asynchronous orphan cleanup
2554 Details : orphan cleanup is now performed in separate threads for each
2555 OST, allowing parallel non-blocking operation.
2557 Severity : enhancement
2559 Description: optimized stripe assignment
2560 Details : stripe assignments are now made based on ost space available,
2561 ost previous usage, and OSS previous usage, in order to try
2562 to optimize storage space and networking resources.
2564 Severity : enhancement
2566 Description: Permanently set tunables
2567 Details : All writable /proc/fs/lustre tunables can now be permanently
2568 set on a per-server basis, at mkfs time or on a live system.
2570 Severity : enhancement
2572 Description: Lustre message v2
2573 Details : Add lustre message format v2.
2575 Severity : enhancement
2577 Description: client OST exclusion list
2578 Details : Clients can be started with a list of OSTs that should be
2579 declared "inactive" for known non-responsive OSTs.
2583 Description: SPEC SFS validation failure on NFS v2 over lustre.
2584 Details : Changes the blocksize for regular files to be 2x RPC size,
2585 and not depend on stripe size.
2587 Severity : enhancement
2589 Description: Multiple MD RPCs in flight.
2590 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2591 To avoid overly-overloading MDS, introduce a limit on number of
2592 MDS RPCs in flight for a single client and add /proc controls
2593 to adjust this limit.
2595 Severity : enhancement
2597 Description: client read/write statistics
2598 Details : Add client read/write call usage stats for performance
2599 analysis of user processes.
2600 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2601 file access. extents_stats shows chunk size distribution.
2602 extents_stats_per_process show chunk size distribution per
2605 Severity : enhancement
2607 Description: mds statistics
2608 Details : Add detailed mds operations statistics in
2609 /proc/fs/lustre/mds/*/stats.
2613 Description: Failure of copying files with lustre special EAs.
2614 Details : Client side always return success for setxattr call for lustre
2615 special xattr (currently only "trusted.lov").
2620 Description: Refcount LNET uuids
2621 Details : The global LNET uuid list grew linearly with every startup;
2622 refcount repeated list entries instead of always adding to
2625 Severity : enhancement
2627 Description: Dynamic service threads
2628 Details : Within a small range, start extra service threads
2629 automatically when the request queue builds up.
2632 Frequency : mixed-endian client/server environments
2634 Description: mixed-endian crashes
2635 Details : The new msg_v2 system had some failures in mixed-endian
2638 Severity : enhancement
2640 Description: Easy OST removal
2641 Details : OSTs can be permanently deactivated with e.g. 'lctl
2642 conf_param lustre-OST0001.osc.active=0'
2644 Severity : enhancement
2646 Description: MGS proc entries
2647 Details : Added basic proc entries for the MGS showing what filesystems
2650 Severity : enhancement
2652 Description: provide MGS failover
2653 Details : Added config lock reacquisition after MGS server failover.
2655 Severity : enhancement
2657 Description: add Linux 2.4 support
2658 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2662 Description: a large application tries to do I/O to the same resource and dies
2663 in the middle of it.
2664 Details : Check the req->rq_arrival time after the call to
2665 ost_brw_lock_get(), but before we do anything about
2666 processing it & sending the BULK transfer request. This
2667 should help move old stale pending locks off the queue as
2668 quickly as obd_timeout.
2671 Frequency : when an incorrect nid is specified during startup
2673 Description: ptlrpc connect to non-existant node causes kernel crash
2674 Details : LNET can't be re-entered from an event callback, which
2675 happened when we expire a message after the export has been
2676 cleaned up. Instead, hand the zombie cleanup off to another
2679 Severity : enhancement
2681 Description: plain/inodebits lock performance improvement
2682 Details : Grouping plain/inodebits in granted list by their request modes
2683 and bits policy, thus improving the performance of search through
2687 Frequency : only if OST filesystem is corrupted
2689 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2690 Details : for a short time RPCs with bulk IO are in the replay list,
2691 but replay of bulk IOs is unimplemented. If the OST filesystem
2692 is corrupted due to disk cache incoherency and then replay is
2693 started it is possible to trip an assertion. Avoid putting
2694 committed RPCs into the replay list at all to avoid this issue.
2697 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2699 Description: System hang on startup
2700 Details : This bug allowed the liblustre (e.g. catamount) client to
2701 return to the app before handling all startup RPCs. This
2702 could leave the node unresponsive to lustre network traffic
2703 and manifested as a server ptllnd timeout.
2705 Severity : enhancement
2707 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2708 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2709 boolean to control whether to print peer debug info when a
2710 client's RPC times out.
2713 Frequency : only for kernels with patches from Lustre below 1.4.3
2715 Description: Remove old rdonly API
2716 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2719 Frequency : only for devices with external journals
2721 Description: Set external device read-only also
2722 Details : During a commanded failover stop, we set the disk device
2723 read-only while the server shuts down. We now also set any
2724 external journal device read-only at the same time.
2727 Frequency : when upgrading from 1.4 while trying to change parameters
2729 Description: The wrong (new) MDC name was used when setting parameters for
2730 upgraded MDT's. Also allows changing of OSC (and MDC)
2731 parameters if --writeconf is specified at tunefs upgrade time.
2734 Frequency : when setting specific ost indicies
2736 Description: QOS code breaks on skipped indicies
2737 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2738 created with --index need not be sequential.
2740 Severity : enhancement
2742 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2743 Details : The uninit_groups feature works in conjunction with the kernel
2744 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2745 up the pass1 processing of e2fsck. This is a read-only feature
2746 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2747 mount filesystems that have had this feature enabled.
2749 Severity : enhancement
2751 Description: Improve multi-block allocation algorithm to avoid fragmentation
2752 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2753 allocation locality and avoid filesystem fragmentation.
2755 ------------------------------------------------------------------------------
2757 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2759 * Support for kernels:
2760 2.4.21-47.0.1.EL (RHEL 3)
2761 2.6.5-7.283 (SLES 9)
2762 2.6.9-42.0.10.EL (RHEL 4)
2763 2.6.12.6 vanilla (kernel.org)
2764 2.6.16.27-0.9 (SLES 10)
2765 * Recommended e2fsprogs version: 1.39.cfs5
2767 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2771 Frequency : occasional, depends on client load and configuration
2772 Bugzilla : 12181, 12203
2773 Description: data loss for recently-modified files
2775 Details : In some cases it is possible that recently written or created
2776 files may not be written to disk in a timely manner (this should
2777 normally be within 30s unless client IO load is very high).
2778 The problem appears as zero-length files or files that are a
2779 multiple of 1MB in size after a client crash or client eviction
2780 that are missing data at the end of the file.
2782 This problem is more likely to be hit on clients where files are
2783 repeatedly created and unlinked in the same directory, clients
2784 have a large amount of RAM, have many CPUs, the filesystem has
2785 many OSTs, the clients are rebooted frequently, and/or the files
2786 are not accessed by other nodes after being written.
2788 The presence of the problem can be detected by looking at
2789 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2790 smaller than the second (nr_unused) then dirty files will not
2791 be flushed automatically to disk. "sync; sleep 10" should be
2792 run several times on the node before unmounting it to update
2793 Lustre (this is also safe to run on nodes without this problem).
2795 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2796 that can cause this same problem, so customers using that kernel
2797 also need to update the kernel in addition to Lustre. In order
2798 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2800 It is normal that files written just before a client crash (less
2801 than 30s) may not yet have been flushed to disk, even for local
2805 Frequency : frequent on thin XT3 nodes
2807 Description: UUID collision on thin XT3 Linux nodes
2808 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2809 randomly, since we relied on an insufficiently-seeded PRNG.
2814 Description: OSS hangs after "All ost request buffers busy"
2815 Details : A deadlock between quota and journal operations caused OSS
2816 hangs after printing "All ost request buffers busy."
2819 Frequency : always on liblustre builds
2821 Description: Cleanup compiler warnings on liblustre
2824 Frequency : always on liblustre builds on XT3
2826 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2830 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2831 > 32000 subdirectories
2832 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2833 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2834 subdirectory count crosses 32000. This will aid e2fsck to
2835 correctly handle more than 32000 subdirectories.
2838 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2840 Description: System hang on startup
2841 Details : This bug allowed the liblustre (e.g. catamount) client to
2842 return to the app before handling all startup RPCs. This
2843 could leave the node unresponsive to lustre network traffic
2844 and manifested as a server ptllnd timeout.
2846 Severity : enhancement
2848 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2849 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2850 boolean to control whether to print peer debug info when a
2851 client's RPC times out.
2856 Description: make O_SYNC working on 2.6 kernels
2857 Details : 2.6 kernels use different method for mark pages for write,
2858 so need add a code to lustre for O_SYNC work.
2863 Description: Failure to close file and release space on NFS
2864 Details : Put inode details into lock acquired in ll_intent_file_open.
2865 Use mdc_intent_lock in ll_intent_open to properly
2866 detect all kind of errors unhandled by mdc_enqueue.
2871 Description: proc file read during shutdown sometimes raced obd removal,
2873 Details : Add lock to prevent obd access after proc file removal.
2876 Frequency : Only for files larger than 4GB on 32-bit clients.
2878 Description: improperly doing page alignment of locks
2879 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2880 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2885 Description: RPCs being resent when they shouldn't be
2886 Details : Some RPCs that should not be resent are being resent. This
2887 can cause inconsistencies in the RPC state machine. Do not
2888 resend such requests.
2891 Frequency : rare, only with NFS export
2893 Description: Crash on NFS re-export node
2894 Details : under very unusual load conditions an assertion is hit in
2895 ll_intent_file_open()
2898 Frequency : only if OST filesystem is corrupted
2900 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2901 Details : for a short time RPCs with bulk IO are in the replay list,
2902 but replay of bulk IOs is unimplemented. If the OST filesystem
2903 is corrupted due to disk cache incoherency and then replay is
2904 started it is possible to trip an assertion. Avoid putting
2905 committed RPCs into the replay list at all to avoid this issue.
2910 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2911 Details : Large single O_DIRECT read and write calls can fail to allocate
2912 a sufficiently large buffer to process the request. In case of
2913 allocation failure the allocation is retried with a smaller
2914 buffer and broken into smaller requests.
2916 Severity : enhancement
2918 Description: Add -o localflock option to simulate old noflock behaviour.
2919 Details : This will achieve local-only flock/fcntl locks coherentness.
2924 Description: versioning check is incomplete
2925 Details : Checking the version difference of client vs. server, report
2926 error if the gap is too big.
2931 Description: add support PG_writeback bit
2932 Details : add support for PG_writeback bit for Lustre, for more carefull
2933 work with page cache in 2.6 kernel. This also fix some deadlocks
2934 and remove hack for work O_SYNC with 2.6 kernel.
2936 Severity : enhancement
2938 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2939 Details : The uninit_groups feature works in conjunction with the kernel
2940 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2941 up the pass1 processing of e2fsck. This is a read-only feature
2942 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2943 mount filesystems that have had this feature enabled.
2945 Severity : enhancement
2947 Description: Improve multi-block allocation algorithm to avoid fragmentation
2948 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2949 allocation locality and avoid filesystem fragmentation.
2951 ------------------------------------------------------------------------------
2953 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2955 * Support for kernels:
2956 2.6.9-42.0.3.EL (RHEL 4)
2957 2.6.5-7.276 (SLES 9)
2958 2.4.21-47.0.1.EL (RHEL 3)
2959 2.6.12.6 vanilla (kernel.org)
2960 2.6.16.21-0.8 (SLES10)
2961 * Recommended e2fsprogs version: 1.39.cfs2-0
2963 * The backwards-compatible /proc/sys/portals symlink has been removed
2964 in this release. Before upgrading, please ensure that you change
2965 any configuration scripts or /etc/sysctl.conf files that access
2966 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2967 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2968 in advance of the upgrade on any system running Lustre 1.4.6 or
2969 newer, since /proc/sys/lnet was added in that version.
2970 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2974 Frequency : only when quota is used
2976 Description: avoid scanning export list for quota master
2977 Details : Change the algorithms to avoid scanning export list in order
2978 to improve the efficiency.
2981 Frequency : MDS failover only, very rarely
2983 Description: "went back in time" messages on mds failover
2984 Details : The greatest transno may be lost when the current operation
2985 finishes with an error (transno==0) and the client's last_rcvd
2986 record is over-written. Save the greatest transno in the
2987 mds_last_transno for this case.
2990 Frequency : always for specific kernels and striping counts
2992 Description: client may get "Matching packet too big" without ACL support
2993 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2994 when trying to access files in certain configurations. The
2995 clients should in fact be denied when mounting because they do
2996 not understand ACLs.
2999 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3001 Description: many clients connecting with IO in progress causes connect timeouts
3002 Details : Avoid synchronous journal commits to avoid delays caused by many
3003 clients connecting/disconnecting when bulk IO is in progress.
3004 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3005 OST_IO_PORTAL to avoid delays behind potentially many pending
3009 Frequency : occasionally with multiple writers to a single file
3011 Description: shared writes to file may result in wrong size reported by stat()
3012 Details : Allow growing of kms when extent lock is cancelled
3015 Frequency : always with random mmap IO to multi-striped file
3017 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3018 Details : Only if the hole is at the end of OST object so that kms is too
3019 small. Fix is to increase kms accordingly in ll_nopage.
3022 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3024 Description: writes to a missing object would leak memory on the OST
3025 Details : If there is an inconsistency between the MDS and OST filesystems,
3026 such that the MDS references an object that doesn't exist, writes
3027 to that object will leak memory due to incorrect cleanup in the
3028 error handling path, eventually running out of memory on the OST.
3033 Description: Creating too long symlink causes lustre errors
3034 Details : Check symlink and name lengths before sending requests to MDS.
3037 Frequency : only if flock is enabled (not on by default)
3039 Description: posix locks not released on fd closure on 2.6.9+
3040 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3041 kernels, this caused such locks not to be released on fd close and
3042 then assertions on fs unmount about still used locks.
3045 Frequency : MDS failover only, very rarely
3047 Description: clients may get ASSERTION(granted_lock != NULL)
3048 Details : When request was taking a long time, and a client was resending
3049 a getattr by name lock request. The were multiple lock requests
3050 with the same client lock handle and
3051 mds_getattr_name->fixup_handle_for_resent_request found one of the
3052 lock handles but later failed with ASSERTION(granted_lock != NULL).
3057 Description: handle->h_buffer_credits > 0, assertion failure
3058 Details : h_buffer_credits is zero after truncate, causing assertion
3059 failure. This patch extends the transaction or creates a new
3063 Frequency : NFS re-export or patchless client
3064 Bugzilla : 11179, 10796
3065 Description: Crash on NFS re-export node (__d_move)
3066 Details : We do not want to hash the dentry if we don't have a lock.
3067 But if this dentry is later used in d_move, we'd hit uninitialised
3068 list head d_hash, so we just do this to init d_hash field but
3069 leave dentry unhashed.
3072 Frequency : NFS re-export or patchless client
3074 Description: NFS exports has problem with symbolic link
3075 Details : lustre client didn't properly install dentry when re-exported
3076 to NFS or running patchless client.
3079 Frequency : NFS re-export or patchless client
3081 Description: Various nfs/patchless fixes.
3082 Details : fixes reuse disconected alias for lookup process - this fixes
3083 warning "find_exported_dentry: npd != pd",
3084 fix permission error with open files at nfs.
3085 fix apply umask when do revalidate.
3088 Frequency : occasional
3090 Description: Crash on NFS re-export node
3091 Details : calling clear_page() on the wrong pointer triggered oops in
3092 generic_mapping_read().
3095 Frequency : rarely, using O_DIRECT IO
3097 Description: unaligned directio crashes client with LASSERT
3098 Details : check for unaligned buffers before trying any requests.
3101 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3103 Description: stale data returned from RAID cache
3104 Details : If only a small amount of IO is done to the RAID device before
3105 reading it again it is possible to get stale data from the RAID
3106 cache instead of reading it from disk.
3109 Frequency : always for sles10 kernel
3111 Description: sles10 support
3112 Details : ll_follow_link: compile fixes and using of nd_set_link
3113 under newer kernels.
3116 Frequency : depends on arch, kernel and compiler version, always on sles10
3119 Description: recursive or deep enough symlinks cause stack overflow
3120 Details : getting rid of large stack-allocated variable in
3124 Frequency : depends on hardware
3126 Description: lustre write performance loss in the SLES10 kernel
3127 Details : the performance loss is caused by using of write barriers in the
3128 ext3 code. The SLES10 kernel turns barrier support on by
3129 default. The fix is to undo that change for ldiskfs.
3131 ------------------------------------------------------------------------------
3133 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3135 * Support for kernels:
3136 2.6.9-42.0.3EL (RHEL 4)
3137 2.6.5-7.276 (SLES 9)
3138 2.4.21-47.0.1.EL (RHEL 3)
3139 2.6.12.6 vanilla (kernel.org)
3143 Frequency : quota enabled and large files being deleted
3145 Description: releasing more than 4GB of quota at once hangs OST
3146 Details : If a user deletes more than 4GB of files on a single OST it
3147 will cause the OST to spin in an infinite loop. Release
3148 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3153 Description: statfs data retrieved from /proc may be stale or zero
3154 Details : When reading per-device statfs data from /proc, in the
3155 {kbytes,files}_{total,free,avail} files, it may appear
3156 as zero or be out of date.
3159 Frequency : systems with MD RAID1 external journal devices
3161 Description: lconf's call to blkid is confused by RAID1 journal devices
3162 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3163 of returning all block devices that match the journal UUID.
3166 Frequency : always, for aggregate stripe size over 4GB
3168 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3169 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3170 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3173 Frequency : always if "lfs find" used on a local file/directory
3175 Description: "lfs find" segfaults if used on a local file/directory
3176 Details : The case where a directory component was not specified wasn't
3177 handled correctly. Handle this properly.
3180 Frequency : always on ppc64
3182 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3183 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3184 perform endianness conversion.
3187 Frequency : rarely (truncate to non-zero file size after write under load)
3188 Bugzilla : 10730, 10687
3189 Description: Files padded with zeros to next 4K multiple
3190 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3191 it is possible that files that are truncated to a non-zero size
3192 immediately after being written are filled with zero bytes beyond
3193 the truncated size. No file data is lost.
3195 Severity : enhancement
3197 Description: Allow recovery/failover for liblustre clients.
3198 Details : liblustre clients were unaware of failover configurations until
3201 Severity : enhancement
3203 Description: user file locks should fail when not mounting with flock option
3204 Details : Set up an error-returning stub in ll_file_operations.lock field
3205 to prevent incorrect behaviour when client is mounted without
3206 flock option. Also, set up properly f_op->flock field for
3210 Frequency : always on ia64
3212 Description: "lfs df" loops on printing out MDS statfs information
3213 Details : The obd_ioctl_data was not initialized and in some systems
3214 this caused a failure during the ioctl that did not return
3215 an error. Initialize the struct and return an error on failure.
3218 Frequency : SLES 9 only
3220 Description: Error of copying files with lustre special EAs as root
3221 Details : Client side always return success for setxattr call for lustre
3222 special xattr (currently only "trusted.lov").
3225 Frequency : rarely on clusters with both ia64+i386 clients
3227 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3228 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3229 clients are doing concurrent writes to the same file it is
3230 possible that the ia64 clients may LASSERT because the OST
3231 extent locks are not PAGE_SIZE aligned. Ensure that grown
3232 locks are always aligned on the request boundary.
3235 Frequency : specific use, occasional
3237 Description: Overwriting in use executable truncates on-disk binary image
3238 Details : If one node attempts to overwrite an executable in use by
3239 another node, we now correctly return ETXTBSY instead of
3240 truncating the file.
3242 Severity : enhancement
3244 Description: Async OSC create to avoid the blocking unnecessarily.
3245 Details : If a OST has no remain object, system will block on the creating
3246 when need to create a new object on this OST. Now, ways use
3247 pre-created objects when available, instead of blocking on an
3248 empty osc while others are not empty. If we must block, we block
3249 for the shortest possible period of time.
3254 Description: chmod on Lustre root is propagated to other clients
3255 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3256 invalid by the follow_mount time.
3259 Frequency : liblustre clients only
3261 Description: Race in 'instant cancel' lock handling could lead to such locks
3262 never to be granted in case of SMP MDS
3263 Details : Do not destroy not yet granted but cbpending locks in
3267 Frequency : replay/resend of open
3269 Description: non null lock assetion failure in mds_intent_policy
3270 Details : Trying to replay/resend lockless open requests resulted in
3271 mds_open() returning 0 with no lock. Now it sets a flag if
3272 a lock is going to be returned.
3274 Severity : enhancement
3276 Description: Checksum enhancements
3277 Details : New checksum enhancements allow for resending RPCs that failed
3280 Severity : enhancement
3282 Description: Tunables on number of dirty pages in cacche
3283 Details : Allow to set limit on number of dirty pages cached.
3288 Description: client crash on unmount - lock still has references
3289 Details : In some error handling cases it was possible to leak a lock
3290 reference on a client while accessing a file. This was not
3291 harmful to the client during operation, but would cause the
3292 client to crash when the filesystem is unmounted.
3295 Frequency : specific case, rare
3297 Description: ETXTBSY on mds though file not in use by client
3298 Details : ETXTBSY is no longer incorrectly returned when attempting to
3299 chmod or chown a directory that the user previously tried to
3300 execute or a currently-executing binary.
3303 Frequency : extremely rare except on liblustre-based clients
3305 Description: Lustre space not freed when files are deleted
3306 Details : Clean up open-unlinked files after client eviction. Previously
3307 the unlink was skipped and the files remained as orphans.
3312 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3313 Details : In some cases it was possible to send a blocking callback to
3314 a client doing a glimpse, even though that client didn't get
3315 a lock granted. When the glimpse lock is cancelled on the OST
3316 the freed lock is left on the waiting list and corrupted the list.
3319 Frequency : all core dumps
3321 Description: Broke core dumps to lustre
3322 Details : Negative dentry may be unhashed if parent does not have UPDATE
3323 lock, but some callers, e.g. do_coredump, expect dentry to be
3324 hashed after successful create, hash it in ll_create_it.
3326 ------------------------------------------------------------------------------
3328 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3330 * Support for kernels:
3331 2.6.9-42.0.2.EL (RHEL 4)
3332 2.6.5-7.276 (SLES 9)
3333 2.4.21-40.EL (RHEL 3)
3334 2.6.12.6 vanilla (kernel.org)
3338 Frequency : always on RHEL 3
3340 Description: Number of open files grows over time
3341 Details : The number of open files grows over time, whether or not
3342 Lustre is started. This was due to a filp leak introduced
3343 by one of our kernel patches.
3345 ------------------------------------------------------------------------------
3347 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3349 * Support for kernels:
3350 2.6.9-42.EL (RHEL 4)
3351 2.6.5-7.276 (SLES 9)
3352 2.4.21-40.EL (RHEL 3)
3353 2.6.12.6 vanilla (kernel.org)
3358 Bugzilla : 5719, 9635, 9792, 9684
3359 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3360 Details : If a server is under heavy load and cannot reply to new
3361 connection requests before the client resends the (re)connect,
3362 the connection handling code can behave badly if two service
3363 threads are concurrently handing separate (re)connections from
3364 the same client. Add better locking to the connection handling
3365 code, and ensure that only a single connection will be processed
3366 for a given client UUID, even if the lock is dropped.
3368 Severity : enhancement
3370 Description: add TCP zero-copy support to kernel
3371 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3372 sends if the hardware supports scatter-gather and checksumming.
3373 This allows socklnd to do client-write and server-read more
3374 efficiently and reduce CPU utilization from skbuf copying.
3377 Frequency : only if NFS exporting from client
3379 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3380 Details : If setting attributes on a file created under NFS that had
3381 never been opened it would be possible to oops the client
3382 if the file had no objects.
3385 Frequency : always for liblustre
3387 Description: liblustre client does MDS+OSTs setattr RPC for each write
3388 Details : When doing a write from a liblustre client, the client
3389 incorrectly issued an RPC to the MDS and each OST the file was
3390 striped over in order to update the timestamps. When writing
3391 with small chunks and many clients this could overwhelm the MDS
3392 with RPCs. In all cases it would slow down the write because
3393 these RPCs are unnecessary.
3395 Severity : enhancement
3397 Description: allow number of MDS service threads to be changed at module load
3398 Details : It is now possible to change the number of MDS service threads
3399 running. Adding "options mds mds_num_threads={N}" to the MDS's
3400 /etc/modprobe.conf will set the number of threads for the next
3401 time Lustre is restarted (assuming the "mds" module is also
3402 reloaded at that time). The default number of threads will
3403 stay the same, 32 for most systems.
3408 Description: OST crash if filesystem is unformatted or corrupt
3409 Details : If an OST is started on a device that has never been formatted
3410 or if the filesystem is corrupt and cannot even mount then the
3411 error handling cleanup routines would dereference a NULL pointer.
3416 Description: NULL pointer deref in llap_from_page.
3417 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3418 incorrect) mapping if the page was truncated/reclaimed while it was
3419 searched for. Check for this condition and skip such pages when
3420 doing readahead. Introduce extra check to llap_from_page() to
3421 verify page->mapping->host is non-NULL (so page is not anonymous).
3424 Frequency : Sometimes when using sys_sendfile
3426 Description: "page not covered by a lock" warnings from ll_readpage
3427 Details : sendfile called ll_readpage without right page locks present.
3428 Now we introduced ll_file_sendfile that does necessary locking
3429 around call to generic_file_sendfile() much like we do in
3433 Frequency : with certain MDS communication failures at client mount time
3435 Description: NULL pointer deref after failed client mount
3436 Details : a client connection request may delayed by the network layer
3437 and not be sent until after the PTLRPC layer has timed out the
3438 request. If the client fails the mount immediately it will try
3439 to clean up before the network times out the request. Add a
3440 reference from the request import to the obd device and delay
3441 the cleanup until the network drops the request.
3444 Frequency : occasionally during client (re)connect
3446 Description: assertion failure during client (re)connect
3447 Details : processing a client connection request may be delayed by the
3448 client or server longer than the client connect timeout. This
3449 causes the client to resend the connection request. If the
3450 original connection request is replied in this interval, the
3451 client may trip an assertion failure in ptlrpc_connect_interpret()
3452 which thought it would be the only running connect process.
3455 Frequency : only with obd_echo servers and clients that are rebooted
3457 Description: kernel BUG accessing uninitialized data structure
3458 Details : When running an obd_echo server it did not start the ping_evictor
3459 thread, and when a client was evicted an uninitialized data
3460 structure was accessed. Start the ping_evictor in the RPC
3461 service startup instead of the OBD startup.
3463 Severity : enhancement
3464 Bugzilla : 10193 (patchless)
3465 Description: Remove dependency on various unexported kernel interfaces.
3466 Details : No longer need reparent_to_init, exit_mm, exit_files,
3467 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3470 Frequency : rare (only users of deprecated and unsupported LDAP config)
3472 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3473 Details : LDAP apparently contains 'lustreName' attributes instead of
3474 'name'. A simple remapping of the name is sufficient.
3477 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3479 Description: waiting_locks_callback trips kernel BUG if client is evicted
3480 Details : Running with the dump_on_timeout debug flag turned on makes
3481 it possible that the waiting_locks_callback() can try to dump
3482 the Lustre kernel debug logs from an interrupt handler. Defer
3483 this log dumping to the expired_lock_main() thread.
3485 Severity : enhancement
3487 Description: Support NFS exporting on 2.6 kernels.
3488 Details : Implement non-rawops metadata methods for NFS server to use without
3489 changing NFS server code.
3492 Frequency : very rare (synthetic metadata workload only)
3494 Description: two racing renames might cause an MDS thread to deadlock
3495 Details : Running the "racer" program may cause one MDS thread to rename
3496 a file from being the source of a rename to being the target of
3497 a rename at exactly the same time that another thread is doing
3498 so, and the second thread has already enqueued these locks after
3499 doing a lookup of the target and is trying to relock them in
3500 order. Ensure that we don't try to re-lock the same resource.
3503 Frequency : only very large systems with liblustre clients
3505 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3506 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3507 all OSTs in parallel. This allows the eviction of stale liblustre
3508 clients to proceed much faster than if they were done in series,
3509 and also offers similar improvements for other set_info RPCs.
3514 Description: excessive CPU usage during initial read phase on client
3515 Details : During the initial read phase on a client, it would agressively
3516 retry readahead on the file, consuming too much CPU and impacting
3517 performance (since 1.4.5.8). Improve the readahead algorithm
3518 to avoid this, and also improve some other common cases (read