3 # add uml1 uml2 uml3 in your /etc/hosts
5 # FIXME - there is no reason to use all of these different
6 # return codes, espcially when most of them are mapped to something
7 # else anyway. The combination of test number and return code
8 # figure out what failed.
14 # bug number for skipped test:
15 ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT"
16 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
19 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
21 PTLDEBUG=${PTLDEBUG:--1}
23 LUSTRE=${LUSTRE:-`dirname $0`/..}
24 RLUSTRE=${RLUSTRE:-$LUSTRE}
26 . $LUSTRE/tests/test-framework.sh
28 # STORED_MDSSIZE is used in test_18
29 if [ -n "$MDSSIZE" ]; then
30 STORED_MDSSIZE=$MDSSIZE
32 # use small MDS + OST size to speed formatting time
35 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
37 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
38 remote_ost_nodsh && skip "remote OST with nodsh" && exit 0
41 [ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a 45"
50 local facet=$SINGLEMDS
51 local dev=${facet}_dev
55 # who knows if/where $TUNEFS is installed? Better reformat if it fails...
56 do_facet ${facet} "$TUNEFS --writeconf ${!dev}" || echo "tunefs failed, reformatting instead" && reformat
62 # The MGS must be started before the OSTs for a new fs, so start
63 # and stop to generate the startup logs.
66 wait_osc_import_state mds ost FULL
71 reformat_and_config() {
77 local facet=$SINGLEMDS
78 # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
79 local num=$(echo $facet | tr -d "mds")
80 local dev=$(mdsdevname $num)
81 echo "start mds service on `facet_active_host $facet`"
82 start $facet ${dev} $MDS_MOUNT_OPTS || return 94
86 echo "stop mds service on `facet_active_host $SINGLEMDS`"
87 # These tests all use non-failover stop
88 stop $SINGLEMDS -f || return 97
92 echo "start ost1 service on `facet_active_host ost1`"
93 start ost1 `ostdevname 1` $OST_MOUNT_OPTS || return 95
97 echo "stop ost1 service on `facet_active_host ost1`"
98 # These tests all use non-failover stop
99 stop ost1 -f || return 98
103 echo "start ost2 service on `facet_active_host ost2`"
104 start ost2 `ostdevname 2` $OST_MOUNT_OPTS || return 92
108 echo "stop ost2 service on `facet_active_host ost2`"
109 # These tests all use non-failover stop
110 stop ost2 -f || return 93
115 echo "mount $FSNAME on ${MOUNTPATH}....."
116 zconf_mount `hostname` $MOUNTPATH || return 96
120 local SAVEMOUNTOPT=$MOUNTOPT
121 MOUNTOPT="remount,$1"
123 echo "remount '$1' lustre on ${MOUNTPATH}....."
124 zconf_mount `hostname` $MOUNTPATH || return 96
125 MOUNTOPT=$SAVEMOUNTOPT
130 echo "umount lustre on ${MOUNTPATH}....."
131 zconf_umount `hostname` $MOUNTPATH || return 97
134 manual_umount_client(){
137 echo "manual umount lustre on ${MOUNT}...."
138 do_facet client "umount -d ${FORCE} $MOUNT"
156 stop_mds || return 201
157 stop_ost || return 202
158 unload_modules || return 203
162 umount_client $MOUNT || return 200
163 cleanup_nocli || return $?
167 do_facet client "cp /etc/passwd $DIR/a" || return 71
168 do_facet client "rm $DIR/a" || return 72
169 # make sure lustre is actually mounted (touch will block,
170 # but grep won't, so do it after)
171 do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73
172 echo "setup single mount lustre success"
176 do_facet client "touch $DIR/a" || return 71
177 do_facet client "rm $DIR/a" || return 72
178 do_facet client "touch $DIR2/a" || return 73
179 do_facet client "rm $DIR2/a" || return 74
180 echo "setup double mount lustre success"
185 if [ "$ONLY" == "setup" ]; then
190 if [ "$ONLY" == "cleanup" ]; then
197 #create single point mountpoint
203 check_mount || return 41
206 run_test 0 "single mount setup"
210 echo "start ost second time..."
212 check_mount || return 42
215 run_test 1 "start up ost twice (should return errors)"
220 echo "start mds second time.."
223 check_mount || return 43
226 run_test 2 "start up mds twice (should return err)"
230 #mount.lustre returns an error if already in mtab
231 mount_client $MOUNT && return $?
232 check_mount || return 44
235 run_test 3 "mount client twice (should return err)"
239 touch $DIR/$tfile || return 85
243 # ok for ost to fail shutdown
244 if [ 202 -ne $eno ]; then
249 run_test 4 "force cleanup ost, then cleanup"
251 test_5a() { # was test_5
253 touch $DIR/$tfile || return 1
254 fuser -m -v $MOUNT && echo "$MOUNT is in use by user space process."
256 stop_mds -f || return 2
258 # cleanup may return an error from the failed
259 # disconnects; for now I'll consider this successful
260 # if all the modules have unloaded.
264 echo "killing umount"
265 kill -TERM $UMOUNT_PID
266 echo "waiting for umount to finish"
268 if grep " $MOUNT " /proc/mounts; then
269 echo "test 5: /proc/mounts after failed umount"
273 echo "killing umount"
274 kill -TERM $UMOUNT_PID
275 echo "waiting for umount to finish"
277 grep " $MOUNT " /proc/mounts && echo "test 5: /proc/mounts after second umount" && return 11
281 # stop_mds is a no-op here, and should not fail
282 cleanup_nocli || return $?
283 # df may have lingering entry
285 # mtab may have lingering entry
289 while [ "$WAIT" -ne "$MAX_WAIT" ]; do
291 grep -q $MOUNT" " /etc/mtab || break
292 echo "Waiting /etc/mtab updated ... "
293 WAIT=$(( WAIT + sleep))
295 [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs"
296 echo "/etc/mtab updated in $WAIT secs"
298 run_test 5a "force cleanup mds, then cleanup"
302 [ -d $MOUNT ] || mkdir -p $MOUNT
303 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab before mount" && return 10
304 mount_client $MOUNT && return 1
305 grep " $MOUNT " /etc/mtab && echo "test 5b: mtab after failed mount" && return 11
307 # stop_mds is a no-op here, and should not fail
308 cleanup_nocli || return $?
311 run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)"
316 [ -d $MOUNT ] || mkdir -p $MOUNT
317 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10
318 local oldfs="${FSNAME}"
319 FSNAME="wrong.${FSNAME}"
320 mount_client $MOUNT || :
322 grep " $MOUNT " /etc/mtab && echo "test 5c: mtab after failed mount" && return 11
324 cleanup_nocli || return $?
326 run_test 5c "cleanup after failed mount (bug 2712) (should return errs)"
332 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab before mount" && return 10
333 mount_client $MOUNT || return 1
335 grep " $MOUNT " /etc/mtab && echo "test 5d: mtab after unmount" && return 11
338 run_test 5d "mount with ost down"
344 #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506
345 do_facet client "lctl set_param fail_loc=0x80000506"
346 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab before mount" && return 10
347 mount_client $MOUNT || echo "mount failed (not fatal)"
349 grep " $MOUNT " /etc/mtab && echo "test 5e: mtab after unmount" && return 11
352 run_test 5e "delayed connect, don't crash (bug 10268)"
357 mount_client ${MOUNT} || return 87
358 touch $DIR/a || return 86
361 run_test 6 "manual umount, then mount again"
366 cleanup_nocli || return $?
368 run_test 7 "manual umount, then cleanup"
373 check_mount2 || return 45
374 umount_client $MOUNT2
377 run_test 8 "double mount setup"
382 do_facet ost1 lctl set_param debug=\'inode trace\' || return 1
383 do_facet ost1 lctl set_param subsystem_debug=\'mds ost\' || return 1
385 CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`"
386 if [ "$CHECK_PTLDEBUG" ] && { \
387 [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] ||
388 [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then
389 echo "lnet.debug success"
391 echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'"
394 CHECK_SUBSYS="`do_facet ost1 lctl get_param -n subsystem_debug`"
395 if [ "$CHECK_SUBSYS" ] && [ "$CHECK_SUBSYS" = "mds ost" ]; then
396 echo "lnet.subsystem_debug success"
398 echo "lnet.subsystem_debug: want 'mds ost', have '$CHECK_SUBSYS'"
401 stop_ost || return $?
404 run_test 9 "test ptldebug and subsystem for mkfs"
406 # LOGS/PENDING do not exist anymore since CMD3
408 local TMPMTPT="${TMP}/conf16"
409 local dev=${SINGLEMDS}_dev
411 if [ ! -e "$MDSDEV" ]; then
412 log "no $MDSDEV existing, so mount Lustre to create one"
414 check_mount || return 41
418 [ -f "$MDSDEV" ] && LOOPOPT="-o loop"
420 log "change the mode of $MDSDEV/OBJECTS to 555"
421 do_facet $SINGLEMDS "mkdir -p $TMPMTPT &&
422 mount $LOOPOPT -t $FSTYPE $MDSDEV $TMPMTPT &&
423 chmod 555 $TMPMTPT/OBJECTS &&
424 umount $TMPMTPT" || return $?
426 log "mount Lustre to change the mode of OBJECTS, then umount Lustre"
428 check_mount || return 41
431 log "read the mode of OBJECTS and check if they has been changed properly"
432 EXPECTEDOBJECTSMODE=`do_facet $SINGLEMDS "$DEBUGFS -R 'stat OBJECTS' $MDSDEV 2> /dev/null" | grep 'Mode: ' | sed -e "s/.*Mode: *//" -e "s/ *Flags:.*//"`
434 if [ "$EXPECTEDOBJECTSMODE" = "0777" ]; then
435 log "Success:Lustre change the mode of OBJECTS correctly"
437 error "Lustre does not change mode of OBJECTS properly"
440 run_test 16 "verify that lustre will correct the mode of OBJECTS"
443 local dev=${SINGLEMDS}_dev
446 if [ ! -e "$MDSDEV" ]; then
447 echo "no $MDSDEV existing, so mount Lustre to create one"
449 check_mount || return 41
453 echo "Remove mds config log"
454 do_facet $SINGLEMDS "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MDSDEV || return \$?" || return $?
457 start_mds && return 42
460 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
463 [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return
465 local dev=${SINGLEMDS}_dev
471 # check if current MDSSIZE is large enough
472 [ $MDSSIZE -ge $MIN ] && OK=1 && myMDSSIZE=$MDSSIZE && \
473 log "use MDSSIZE=$MDSSIZE"
475 # check if the global config has a large enough MDSSIZE
476 [ -z "$OK" -a ! -z "$STORED_MDSSIZE" ] && [ $STORED_MDSSIZE -ge $MIN ] && \
477 OK=1 && myMDSSIZE=$STORED_MDSSIZE && \
478 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
480 # check if the block device is large enough
481 [ -z "$OK" -a -b $MDSDEV ] && \
482 [ "$(dd if=$MDSDEV of=/dev/null bs=1k count=1 skip=$MIN 2>&1 |
483 awk '($3 == "in") { print $1 }')" = "1+0" ] && OK=1 && \
484 myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
486 # check if a loopback device has enough space for fs metadata (5%)
487 [ -z "$OK" ] && [ -f $MDSDEV -o ! -e $MDSDEV ] &&
488 SPACE=$(df -P $(dirname $MDSDEV) |
489 awk '($1 != "Filesystem") {print $4}') &&
490 [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \
491 log "use file $MDSDEV with MIN=$MIN"
493 [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return
496 echo "mount mds with large journal..."
497 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
499 MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT"
502 echo "mount lustre system..."
504 check_mount || return 41
506 echo "check journal size..."
507 local FOUNDSIZE=`do_facet mds "$DEBUGFS -c -R 'stat <8>' $MDSDEV" | awk '/Size: / { print $NF; exit;}'`
508 if [ $FOUNDSIZE -gt $((32 * 1024 * 1024)) ]; then
509 log "Success: mkfs creates large journals. Size: $((FOUNDSIZE >> 20))M"
511 error "expected journal size > 32M, found $((FOUNDSIZE >> 20))M"
516 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
519 run_test 18 "check mkfs creates large journals"
522 start_mds || return 1
523 stop_mds -f || return 2
525 run_test 19a "start/stop MDS without OSTs"
528 start_ost || return 1
529 stop_ost -f || return 2
531 run_test 19b "start/stop OSTs without MDS"
534 # first format the ost/mdt
538 check_mount || return 43
540 remount_client ro $MOUNT || return 44
541 touch $DIR/$tfile && echo "$DIR/$tfile created incorrectly" && return 45
542 [ -e $DIR/$tfile ] && echo "$DIR/$tfile exists incorrectly" && return 46
543 remount_client rw $MOUNT || return 47
545 [ ! -f $DIR/$tfile ] && echo "$DIR/$tfile missing" && return 48
546 MCNT=`grep -c $MOUNT /etc/mtab`
547 [ "$MCNT" -ne 1 ] && echo "$MOUNT in /etc/mtab $MCNT times" && return 49
552 run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab"
557 wait_osc_import_state mds ost FULL
561 run_test 21a "start mds before ost, stop ost first"
566 wait_osc_import_state mds ost FULL
570 run_test 21b "start ost before mds, stop mds first"
576 wait_osc_import_state mds ost2 FULL
580 #writeconf to remove all ost2 traces for subsequent tests
583 run_test 21c "start mds between two osts, stop mds last"
588 echo Client mount with ost in logs, but none running
590 # wait until mds connected to ost and open client connection
591 wait_osc_import_state mds ost FULL
594 # check_mount will block trying to contact ost
598 echo Client mount with a running ost
601 # if gss enabled, wait full time to let connection from
602 # mds to ost be established, due to the mismatch between
603 # initial connect timeout and gss context negotiation timeout.
604 # This perhaps could be remove after AT landed.
605 echo "sleep $((TIMEOUT + TIMEOUT + TIMEOUT))s"
606 sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
609 check_mount || return 41
614 run_test 22 "start a client before osts (should return errs)"
616 test_23a() { # was test_23
620 # force down client so that recovering mds waits for reconnect
621 local running=$(grep -c $MOUNT /proc/mounts) || true
622 if [ $running -ne 0 ]; then
623 echo "Stopping client $MOUNT (opts: -f)"
627 # enter recovery on mds
629 # try to start a new client
630 mount_client $MOUNT &
632 MOUNT_PID=$(ps -ef | grep "t lustre" | grep -v grep | awk '{print $2}')
633 MOUNT_LUSTRE_PID=`ps -ef | grep mount.lustre | grep -v grep | awk '{print $2}'`
634 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID}
636 ps --ppid $MOUNT_LUSTRE_PID
637 # FIXME why o why can't I kill these? Manual "ctrl-c" works...
638 kill -TERM $MOUNT_LUSTRE_PID
639 echo "waiting for mount to finish"
641 # we can not wait $MOUNT_PID because it is not a child of this shell
647 while [ "$WAIT" -lt "$MAX_WAIT" ]; do
649 PID1=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_PID)
650 PID2=$(ps -ef | awk '{print $2}' | grep -w $MOUNT_LUSTRE_PID)
653 [ -z "$PID1" -a -z "$PID2" ] && break
654 echo "waiting for mount to finish ... "
655 WAIT=$(( WAIT + sleep))
657 [ "$WAIT" -eq "$MAX_WAIT" ] && error "MOUNT_PID $MOUNT_PID and \
658 MOUNT__LUSTRE_PID $MOUNT__LUSTRE_PID still not killed in $WAIT secs"
663 run_test 23a "interrupt client during recovery mount delay"
668 test_23b() { # was test_23
671 # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD
672 lctl set_param fail_loc=0x80000313
676 run_test 23b "Simulate -EINTR during mount"
678 fs2mds_HOST=$mds_HOST
679 fs2ost_HOST=$ost_HOST
683 echo "umount $MOUNT2 ..."
684 umount $MOUNT2 || true
685 echo "stopping fs2mds ..."
686 stop fs2mds -f || true
687 echo "stopping fs2ost ..."
688 stop fs2ost -f || true
695 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
696 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
697 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
698 skip_env "mixed loopback and real device not working" && return
701 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
702 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
704 # test 8-char fsname as well
705 local FSNAME2=test1234
706 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10
708 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10
711 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
712 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
714 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
716 check_mount || return 2
717 # files written on 1 should not show up on 2
718 cp /etc/passwd $DIR/$tfile
720 [ -e $MOUNT2/$tfile ] && error "File bleed" && return 7
723 cp /etc/passwd $MOUNT2/b || return 3
724 rm $MOUNT2/b || return 4
725 # 2 is actually mounted
726 grep $MOUNT2' ' /proc/mounts > /dev/null || return 5
728 facet_failover fs2mds
729 facet_failover fs2ost
732 # the MDS must remain up until last MDT
734 MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
735 [ -z "$MDS" ] && error "No MDT" && return 8
737 cleanup_nocli || return 6
739 run_test 24a "Multiple MDTs on a single node"
742 if [ -z "$fs2mds_DEV" ]; then
743 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
744 skip_env "mixed loopback and real device not working" && return
747 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
749 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10
751 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
754 run_test 24b "Multiple MGSs on a single node (should return err)"
758 check_mount || return 2
759 local MODULES=$($LCTL modules | awk '{ print $2 }')
760 rmmod $MODULES 2>/dev/null || true
763 run_test 25 "Verify modules are referenced"
767 # we need modules before mount for sysctl, so make sure...
768 do_facet $SINGLEMDS "lsmod | grep -q lustre || modprobe lustre"
769 #define OBD_FAIL_MDS_FS_SETUP 0x135
770 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000135"
771 start_mds && echo MDS started && return 1
772 lctl get_param -n devices
773 DEVS=$(lctl get_param -n devices | wc -l)
774 [ $DEVS -gt 0 ] && return 2
775 unload_modules || return 203
777 run_test 26 "MDT startup failure cleans LOV (should return errs)"
783 local ORIG=$(do_facet $myfacet "$TEST")
784 if [ $# -gt 3 ]; then
790 echo "Setting $PARAM from $ORIG to $FINAL"
791 do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
793 wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
797 start_ost || return 1
798 start_mds || return 2
799 echo "Requeue thread should have started: "
800 ps -e | grep ll_cfg_requeue
801 set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
804 run_test 27a "Reacquire MGS lock if OST started first"
809 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
811 facet_failover $SINGLEMDS
812 set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
813 set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
817 run_test 27b "Reacquire MGS lock after failover"
821 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
822 PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
825 set_and_check client "$TEST" "$PARAM" $FINAL || return 3
826 FINAL=$(($FINAL + 1))
827 set_and_check client "$TEST" "$PARAM" $FINAL || return 4
828 umount_client $MOUNT || return 200
831 if [ $RESULT -ne $FINAL ]; then
832 echo "New config not seen: wanted $FINAL got $RESULT"
835 echo "New config success: got $RESULT"
837 set_and_check client "$TEST" "$PARAM" $ORIG || return 5
840 run_test 28 "permanent parameter setting"
843 [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
844 setup > /dev/null 2>&1
848 local PARAM="$FSNAME-OST0001.osc.active"
849 local PROC_ACT="osc.$FSNAME-OST0001-osc-[^M]*.active"
850 local PROC_UUID="osc.$FSNAME-OST0001-osc-[^M]*.ost_server_uuid"
852 ACTV=$(lctl get_param -n $PROC_ACT)
854 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
855 # also check ost_server_uuid status
856 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
857 if [ -z "$RESULT" ]; then
858 echo "Live client not deactivated: $(lctl get_param -n $PROC_UUID)"
861 echo "Live client success: got $RESULT"
865 local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active"
870 RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
871 [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
872 if [ $RESULT -eq $DEAC ]; then
873 echo "MDT deactivated also after $WAIT sec (got $RESULT)"
877 if [ $WAIT -eq $MAX ]; then
878 echo "MDT not deactivated: wanted $DEAC got $RESULT"
881 echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
884 # test new client starts deactivated
885 umount_client $MOUNT || return 200
887 RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW)
888 if [ -z "$RESULT" ]; then
889 echo "New client not deactivated from start: $(lctl get_param -n $PROC_UUID)"
892 echo "New client success: got $RESULT"
895 # make sure it reactivates
896 set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
901 #writeconf to remove all ost2 traces for subsequent tests
904 run_test 29 "permanently remove an OST"
909 TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb"
911 LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
912 for i in ${LIST[@]}; do
913 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
915 # make sure client restart still works
917 mount_client $MOUNT || return 4
918 [ "$($TEST)" -ne "$i" ] && return 5
919 set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6
922 run_test 30 "Big config llog"
924 test_31() { # bug 10734
925 # ipaddr must not exist
926 mount -t lustre 4.3.2.1@tcp:/lustre $MOUNT || true
929 run_test 31 "Connect to non-existent node (shouldn't crash)"
931 # Use these start32/stop32 fn instead of t-f start/stop fn,
932 # for local devices, to skip global facet vars init
936 echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
937 umount -d $@ ${MOUNT%/*}/${facet}
946 mkdir -p ${MOUNT%/*}/${facet}
948 echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
949 mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
951 if [ $RC -ne 0 ]; then
952 echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
953 echo "Start of ${device} of local ${facet} failed ${RC}"
967 echo "Cleanup test_32 umount $MOUNT ..."
968 umount -f $MOUNT || true
969 echo "Cleanup local mds ost1 ..."
975 # this test is totally useless on a client-only system
976 [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0
977 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
978 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
980 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
981 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
982 local tmpdir=$TMP/conf32a
984 tar xjvf $DISK1_8 -C $tmpdir || \
985 { skip_env "Cannot untar $DISK1_8" && return 0; }
988 $LCTL set_param debug=$PTLDEBUG
990 $TUNEFS $tmpdir/mds || error "tunefs failed"
992 # nids are wrong, so client wont work, but server should start
993 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
994 trap cleanup_32 EXIT INT || return 3
996 local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
998 [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1000 $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
1001 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1002 UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
1004 [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1006 local NID=$($LCTL list_nids | head -1)
1008 echo "OSC changes should return err:"
1009 $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
1010 $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
1013 echo "MDC changes should succeed:"
1014 $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1015 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
1020 # mount a second time to make sure we didnt leave upgrade flag on
1022 $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
1023 start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
1024 trap cleanup_32 EXIT INT || return 12
1028 rm -rf $tmpdir || true # true is only for TMP on NFS
1030 run_test 32a "Upgrade from 1.8 (not live)"
1033 # this test is totally useless on a client-only system
1034 [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0
1035 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
1036 [ -z "$TUNEFS" ] && skip_env "No tunefs" && return
1038 local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
1039 [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
1040 local tmpdir=$TMP/conf32b
1042 tar xjvf $DISK1_8 -C $tmpdir || \
1043 { skip_env "Cannot untar $DISK1_8" && return ; }
1046 $LCTL set_param debug=$PTLDEBUG
1047 local NEWNAME=lustre
1049 # writeconf will cause servers to register with their current nids
1050 $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
1051 start32 mds1 $tmpdir/mds "-o loop" && \
1052 trap cleanup_32 EXIT INT || return 3
1054 local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
1056 [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
1058 $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 || error "tunefs failed"
1059 start32 ost1 $tmpdir/ost1 "-o loop" || return 5
1060 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
1062 [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
1064 local NID=$($LCTL list_nids | head -1)
1066 echo "OSC changes should succeed:"
1068 $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
1069 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
1072 echo "MDC changes should succeed:"
1073 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
1076 # MDT and OST should have registered with new nids, so we should have
1077 # a fully-functioning client
1078 echo "Check client and old fs contents"
1080 local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
1081 echo "Starting local client: $HOSTNAME: $device $MOUNT"
1082 mount -t lustre $device $MOUNT || return 1
1084 local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
1085 local new=$((old + 5))
1086 $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
1087 wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
1089 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
1094 rm -rf $tmpdir || true # true is only for TMP on NFS
1096 run_test 32b "Upgrade from 1.8 with writeconf"
1098 test_33a() { # bug 12333, was test_33
1100 local FSNAME2=test-123
1101 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST
1103 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then
1104 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1105 skip_env "mixed loopback and real device not working" && return
1108 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1109 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1110 add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1111 add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1113 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
1114 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1115 do_facet $SINGLEMDS "$LCTL conf_param $FSNAME2.sys.timeout=200" || rc=1
1117 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2
1120 cp /etc/hosts $MOUNT2/ || rc=3
1121 $LFS getstripe $MOUNT2/hosts
1126 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev
1127 cleanup_nocli || rc=6
1130 run_test 33a "Mount ost with a large index number"
1132 test_33b() { # was test_34
1135 do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
1136 # Drop lock cancelation reply during umount
1137 #define OBD_FAIL_LDLM_CANCEL 0x304
1138 do_facet client lctl set_param fail_loc=0x80000304
1139 #lctl set_param debug=-1
1140 umount_client $MOUNT
1143 run_test 33b "Drop cancel during umount"
1147 do_facet client "sh runmultiop_bg_pause $DIR/file O_c"
1148 manual_umount_client
1150 do_facet client killall -USR1 multiop
1151 if [ $rc -eq 0 ]; then
1152 error "umount not fail!"
1157 run_test 34a "umount with opened file should be fail"
1162 touch $DIR/$tfile || return 1
1163 stop_mds --force || return 2
1165 manual_umount_client --force
1167 if [ $rc -ne 0 ]; then
1168 error "mtab after failed umount - rc $rc"
1174 run_test 34b "force umount with failed mds should be normal"
1178 touch $DIR/$tfile || return 1
1179 stop_ost --force || return 2
1181 manual_umount_client --force
1183 if [ $rc -ne 0 ]; then
1184 error "mtab after failed umount - rc $rc"
1190 run_test 34c "force umount with failed ost should be normal"
1192 test_35() { # bug 12459
1195 DBG_SAVE="`lctl get_param -n debug`"
1196 lctl set_param debug="ha"
1198 log "Set up a fake failnode for the MDS"
1200 local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
1201 do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
1203 log "Wait for RECONNECT_INTERVAL seconds (10s)"
1206 MSG="conf-sanity.sh test_35 `date +%F%kh%Mm%Ss`"
1209 log "Stopping the MDT:"
1210 stop_mds || return 5
1212 df $MOUNT > /dev/null 2>&1 &
1214 log "Restarting the MDT:"
1215 start_mds || return 6
1216 log "Wait for df ($DFPID) ... "
1219 lctl set_param debug="$DBG_SAVE"
1221 # retrieve from the log the first server that the client tried to
1222 # contact after the connection loss
1223 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1224 NEXTCONN=`awk "/${MSG}/ {start = 1;}
1225 /import_select_connection.*$device-mdc.* using connection/ {
1227 if (\\\$NF ~ /$FAKENID/)
1233 }" $TMP/lustre-log-$TESTNAME.log`
1234 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
1237 run_test 35 "Reconnect to the last active server first"
1241 local FSNAME2=test1234
1242 local fs3ost_HOST=$ost_HOST
1244 [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST
1247 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then
1248 do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \
1249 skip_env "mixed loopback and real device not working" && return
1251 [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return
1253 [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \
1254 { skip "remote OST" && return 0; }
1256 local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2}
1257 local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2}
1258 local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2}
1259 add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10
1260 # XXX after we support non 4K disk blocksize, change following --mkfsoptions with
1262 add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10
1263 add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10
1265 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
1266 start fs2ost $fs2ostdev $OST_MOUNT_OPTS
1267 start fs3ost $fs3ostdev $OST_MOUNT_OPTS
1269 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || return 1
1271 sleep 5 # until 11778 fixed
1273 dd if=/dev/zero of=$MOUNT2/$tfile bs=1M count=7 || return 2
1275 BKTOTAL=`lctl get_param -n obdfilter.*.kbytestotal | awk 'BEGIN{total=0}; {total+=$1}; END{print total}'`
1276 BKFREE=`lctl get_param -n obdfilter.*.kbytesfree | awk 'BEGIN{free=0}; {free+=$1}; END{print free}'`
1277 BKAVAIL=`lctl get_param -n obdfilter.*.kbytesavail | awk 'BEGIN{avail=0}; {avail+=$1}; END{print avail}'`
1278 STRING=`df -P $MOUNT2 | tail -n 1 | awk '{print $2","$3","$4}'`
1279 DFTOTAL=`echo $STRING | cut -d, -f1`
1280 DFUSED=`echo $STRING | cut -d, -f2`
1281 DFAVAIL=`echo $STRING | cut -d, -f3`
1282 DFFREE=$(($DFTOTAL - $DFUSED))
1284 ALLOWANCE=$((64 * $OSTCOUNT))
1286 if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] ||
1287 [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then
1288 echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)"
1291 if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] ||
1292 [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then
1293 echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)"
1296 if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] ||
1297 [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then
1298 echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)"
1303 stop fs3ost -f || return 200
1304 stop fs2ost -f || return 201
1305 stop fs2mds -f || return 202
1306 rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev
1307 unload_modules || return 203
1310 run_test 36 "df report consistency on OSTs with different block size"
1313 [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0
1314 LOCAL_MDSDEV="$TMP/mdt.img"
1315 SYM_MDSDEV="$TMP/sym_mdt.img"
1317 echo "MDS : $LOCAL_MDSDEV"
1318 echo "SYMLINK : $SYM_MDSDEV"
1322 mkfs.lustre --reformat --fsname=lustre --mdt --mgs --device-size=9000 $LOCAL_MDSDEV ||
1323 error "mkfs.lustre $LOCAL_MDSDEV failed"
1324 ln -s $LOCAL_MDSDEV $SYM_MDSDEV
1326 echo "mount symlink device - $SYM_MDSDEV"
1328 mount_op=`mount -v -t lustre -o loop $SYM_MDSDEV ${MOUNT%/*}/mds 2>&1 | grep "unable to set tunable"`
1329 umount -d ${MOUNT%/*}/mds
1330 rm -f $LOCAL_MDSDEV $SYM_MDSDEV
1332 if [ -n "$mount_op" ]; then
1333 error "**** FAIL: set tunables failed for symlink device"
1337 run_test 37 "verify set tunables works for symlink device"
1339 test_38() { # bug 14222
1344 FILES=`find $SRC -type f -mtime +1 | head -n $COUNT`
1345 log "copying $(echo $FILES | wc -w) files to $DIR/$tdir"
1347 tar cf - $FILES | tar xf - -C $DIR/$tdir || \
1348 error "copying $SRC to $DIR/$tdir"
1350 umount_client $MOUNT
1352 log "rename lov_objid file on MDS"
1353 rm -f $TMP/lov_objid.orig
1355 local dev=${SINGLEMDS}_dev
1356 local MDSDEV=${!dev}
1357 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV"
1358 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1360 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.orig"
1361 # check create in mds_lov_connect
1365 [ $V ] && log "verifying $DIR/$tdir/$f"
1366 diff -q $f $DIR/$tdir/$f || ERROR=y
1368 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new\\\" $MDSDEV"
1369 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new"
1370 [ "$ERROR" = "y" ] && error "old and new files are different after connect" || true
1372 # check it's updates in sync
1373 umount_client $MOUNT
1376 do_facet $SINGLEMDS dd if=/dev/zero of=$TMP/lov_objid.clear bs=4096 count=1
1377 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV"
1378 do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"write $TMP/lov_objid.clear lov_objid\\\" $MDSDEV "
1383 [ $V ] && log "verifying $DIR/$tdir/$f"
1384 diff -q $f $DIR/$tdir/$f || ERROR=y
1386 do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.new1\\\" $MDSDEV"
1387 do_facet $SINGLEMDS "od -Ax -td8 $TMP/lov_objid.new1"
1388 umount_client $MOUNT
1390 [ "$ERROR" = "y" ] && error "old and new files are different after sync" || true
1392 log "files compared the same"
1395 run_test 38 "MDS recreates missing lov_objid file from OST data"
1401 perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' &&
1402 error "memory leak detected" || true
1404 run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages"
1406 test_40() { # bug 15759
1408 #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706
1409 do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706"
1413 run_test 40 "race during service thread startup"
1415 test_41() { #bug 14134
1417 local dev=${SINGLEMDS}_dev
1418 local MDSDEV=${!dev}
1420 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n
1421 start ost1 `ostdevname 1` $OST_MOUNT_OPTS
1422 start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force
1424 mount_client $MOUNT || return 1
1427 echo "blah blah" > $MOUNT/$tfile
1430 umount_client $MOUNT
1431 stop ost1 -f || return 201
1432 stop_mds -f || return 202
1433 stop_mds -f || return 203
1434 unload_modules || return 204
1437 run_test 41 "mount mds with --nosvc and --nomgs"
1439 test_42() { #bug 14693
1441 check_mount || return 2
1442 do_facet client lctl conf_param lustre.llite.some_wrong_param=10
1443 umount_client $MOUNT
1444 mount_client $MOUNT || return 1
1448 run_test 42 "invalid config param should not prevent client from mounting"
1451 [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
1453 chmod ugo+x $DIR || error "chmod 0 failed"
1455 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1456 "$FSNAME.mdt.root_squash" \
1459 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1460 "$FSNAME.mdt.nosquash_nids" \
1464 # create set of test files
1466 echo "111" > $DIR/$tfile-userfile || error "write 1 failed"
1467 chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed"
1468 chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed"
1470 echo "222" > $DIR/$tfile-rootfile || error "write 2 failed"
1471 chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield"
1473 mkdir $DIR/$tdir-rootdir -p || error "mkdir failed"
1474 chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed"
1475 touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed"
1478 # check root_squash:
1479 # set root squash UID:GID to RUNAS_ID
1480 # root should be able to access only files owned by RUNAS_ID
1483 "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \
1484 "$FSNAME.mdt.root_squash" \
1485 "$RUNAS_ID:$RUNAS_ID"
1487 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
1488 dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1489 error "$ST: root read permission is denied"
1490 echo "$ST: root read permission is granted - ok"
1493 dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
1494 error "$ST: root write permission is denied"
1495 echo "$ST: root write permission is granted - ok"
1497 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1498 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1499 error "$ST: root read permission is granted"
1500 echo "$ST: root read permission is denied - ok"
1503 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \
1504 error "$ST: root write permission is granted"
1505 echo "$ST: root write permission is denied - ok"
1507 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1508 rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \
1509 error "$ST: root unlink permission is granted"
1510 echo "$ST: root unlink permission is denied - ok"
1512 touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \
1513 error "$ST: root create permission is granted"
1514 echo "$ST: root create permission is denied - ok"
1517 # check nosquash_nids:
1518 # put client's NID into nosquash_nids list,
1519 # root should be able to access root file after that
1521 local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
1522 NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
1523 NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
1525 "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
1526 "$FSNAME-MDTall.mdt.nosquash_nids" \
1529 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
1530 dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1531 error "$ST: root read permission is denied"
1532 echo "$ST: root read permission is granted - ok"
1535 dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
1536 error "$ST: root write permission is denied"
1537 echo "$ST: root write permission is granted - ok"
1539 ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir)
1540 rm $DIR/$tdir-rootdir/tfile-1 || \
1541 error "$ST: root unlink permission is denied"
1542 echo "$ST: root unlink permission is granted - ok"
1543 touch $DIR/$tdir-rootdir/tfile-2 || \
1544 error "$ST: root create permission is denied"
1545 echo "$ST: root create permission is granted - ok"
1549 run_test 43 "check root_squash and nosquash_nids"
1551 umount_client $MOUNT
1556 check_mount || return 2
1557 UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2)
1559 UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid")
1560 for VAL in $UUIDS; do
1561 NID=$(echo $VAL | cut -d= -f1)
1562 CLUUID=$(echo $VAL | cut -d= -f2)
1563 [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break
1565 [ "$STATS_FOUND" = "no" ] && error "stats not found for client"
1569 run_test 44 "mounted client proc entry exists"
1573 check_mount || return 2
1578 #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f
1579 do_facet client "lctl set_param fail_loc=0x50f"
1582 manual_umount_client --force || return 3
1583 do_facet client "lctl set_param fail_loc=0x0"
1585 mount_client $MOUNT || return 4
1589 run_test 45 "long unlink handling in ptlrpcd"
1596 umount_client $MOUNT2 || rc=$?
1597 umount_client $MOUNT || rc=$?
1598 while [ $count -gt 0 ]; do
1599 stop ost${count} -f || rc=$?
1603 # writeconf is needed after the test, otherwise,
1604 # we might end up with extra OSTs
1606 cleanup_nocli || rc=$?
1611 [ $OSTCOUNT -lt 5 ] && skip_env "too few OSTs" && return
1613 start_mds || return 1
1614 #first client should see only one ost
1615 start_ost || return 2
1616 wait_osc_import_state mds ost FULL
1618 mount_client $MOUNT || return 3
1619 trap cleanup_46a EXIT ERR
1621 start_ost2 || return 4
1622 start ost3 `ostdevname 3` $OST_MOUNT_OPTS || return 5
1623 start ost4 `ostdevname 4` $OST_MOUNT_OPTS || return 6
1624 start ost5 `ostdevname 5` $OST_MOUNT_OPTS || return 7
1625 # wait until ost2-5 is sync
1627 wait_osc_import_state mds ost2 FULL
1628 wait_osc_import_state mds ost3 FULL
1629 wait_osc_import_state mds ost4 FULL
1630 wait_osc_import_state mds ost5 FULL
1631 #second client see all ost's
1633 mount_client $MOUNT2 || return 8
1634 $LFS setstripe $MOUNT2 -c -1 || return 9
1635 $LFS getstripe $MOUNT2 || return 10
1637 echo "ok" > $MOUNT2/widestripe
1638 $LFS getstripe $MOUNT2/widestripe || return 11
1639 # fill acl buffer for avoid expand lsm to them
1640 awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do
1641 setfacl -m $acl $MOUNT2/widestripe
1645 stat $MOUNT/widestripe || return 12
1647 cleanup_46a || { echo "cleanup_46a failed!" && return 13; }
1650 run_test 46a "handle ost additional - wide striped file"
1655 check_mount || return 2
1656 $LCTL set_param ldlm.namespaces.$FSNAME-*-*-*.lru_size=100
1660 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1661 if echo $ns | grep "MDT[[:digit:]]*"; then
1664 lrs=$(echo $ns | sed 's/.*lru_size=//')
1665 lru_size[count]=$lrs
1670 facet_failover $SINGLEMDS
1671 df -h $MOUNT || return 3
1674 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do
1675 if echo $ns | grep "MDT[[:digit:]]*"; then
1678 lrs=$(echo $ns | sed 's/.*lru_size=//')
1679 if ! test "$lrs" -eq "${lru_size[count]}"; then
1680 n=$(echo $ns | sed -e 's/ldlm.namespaces.//' -e 's/.lru_size=.*//')
1681 error "$n has lost lru_size: $lrs vs. ${lru_size[count]}"
1689 run_test 47 "server restart does not make client loss lru_resize settings"
1694 # reformat after this test is needed - if test will failed
1695 # we will have unkillable file at FS
1699 test_48() { # bug 17636
1702 check_mount || return 2
1704 $LFS setstripe $MOUNT -c -1 || return 9
1705 $LFS getstripe $MOUNT || return 10
1707 echo "ok" > $MOUNT/widestripe
1708 $LFS getstripe $MOUNT/widestripe || return 11
1710 trap cleanup_48 EXIT ERR
1712 # fill acl buffer for avoid expand lsm to them
1713 getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do
1714 setfacl -m $acl $MOUNT/widestripe
1717 stat $MOUNT/widestripe || return 12
1722 run_test 48 "too many acls on file"
1724 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
1725 test_49() { # bug 17710
1726 local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS
1727 local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS
1728 local LOCAL_TIMEOUT=20
1731 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT"
1737 check_mount || return 1
1739 echo "check ldlm_timout..."
1740 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1741 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1742 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1744 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1745 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1748 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
1749 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1752 umount_client $MOUNT
1753 stop_ost || return 2
1754 stop_mds || return 3
1756 OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT"
1759 start_mds || return 4
1760 start_ost || return 5
1761 mount_client $MOUNT || return 6
1762 check_mount || return 7
1764 LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`"
1765 LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
1766 LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
1768 if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
1769 error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
1772 if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
1773 error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
1776 cleanup || return $?
1778 MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS
1779 OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS
1781 run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
1784 # Test both statfs and lfs df and fail if either one fails
1785 multiop_bg_pause $1 f_
1788 killall -USR1 multiop
1789 [ $RC1 -ne 0 ] && log "lazystatfs multiop failed"
1790 wait $PID || { RC1=$?; log "multiop return error "; }
1797 if [ $RC2 -eq 0 ]; then
1799 log "lazystatfs df failed"
1803 [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1
1809 lctl set_param llite.$FSNAME-*.lazystatfs=1
1812 lazystatfs $MOUNT || error "lazystatfs failed but no down servers"
1814 cleanup || return $?
1816 run_test 50a "lazystatfs all servers available =========================="
1820 lctl set_param llite.$FSNAME-*.lazystatfs=1
1823 # Wait for client to detect down OST
1824 stop_ost || error "Unable to stop OST1"
1825 wait_osc_import_state mds ost DISCONN
1827 lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO"
1829 umount_client $MOUNT || error "Unable to unmount client"
1830 stop_mds || error "Unable to stop MDS"
1832 run_test 50b "lazystatfs all servers down =========================="
1835 start_mds || error "Unable to start MDS"
1836 start_ost || error "Unable to start OST1"
1837 start_ost2 || error "Unable to start OST2"
1838 mount_client $MOUNT || error "Unable to mount client"
1839 lctl set_param llite.$FSNAME-*.lazystatfs=1
1842 # Wait for client to detect down OST
1843 stop_ost || error "Unable to stop OST1"
1844 wait_osc_import_state mds ost DISCONN
1845 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1847 umount_client $MOUNT || error "Unable to unmount client"
1848 stop_ost2 || error "Unable to stop OST2"
1849 stop_mds || error "Unable to stop MDS"
1851 run_test 50c "lazystatfs one server down =========================="
1854 start_mds || error "Unable to start MDS"
1855 start_ost || error "Unable to start OST1"
1856 start_ost2 || error "Unable to start OST2"
1857 mount_client $MOUNT || error "Unable to mount client"
1858 lctl set_param llite.$FSNAME-*.lazystatfs=1
1861 # Issue the statfs during the window where the client still
1862 # belives the OST to be available but it is in fact down.
1863 # No failure just a statfs which hangs for a timeout interval.
1864 stop_ost || error "Unable to stop OST1"
1865 lazystatfs $MOUNT || error "lazystatfs failed with one down server"
1867 umount_client $MOUNT || error "Unable to unmount client"
1868 stop_ost2 || error "Unable to stop OST2"
1869 stop_mds || error "Unable to stop MDS"
1871 run_test 50d "lazystatfs client/server conn race =========================="
1878 start_mds || return 1
1879 #first client should see only one ost
1880 start_ost || return 2
1881 wait_osc_import_state mds ost FULL
1883 # Wait for client to detect down OST
1884 stop_ost || error "Unable to stop OST1"
1885 wait_osc_import_state mds ost DISCONN
1887 mount_client $MOUNT || error "Unable to mount client"
1888 lctl set_param llite.$FSNAME-*.lazystatfs=0
1890 multiop_bg_pause $MOUNT _f
1894 if [ $RC1 -ne 0 ]; then
1895 log "multiop failed $RC1"
1898 sleep $(( $TIMEOUT+1 ))
1900 [ $? -ne 0 ] && error "process isn't sleep"
1901 start_ost || error "Unable to start OST1"
1902 wait $pid || error "statfs failed"
1905 umount_client $MOUNT || error "Unable to unmount client"
1906 stop_ost || error "Unable to stop OST1"
1907 stop_mds || error "Unable to stop MDS"
1909 run_test 50e "normal statfs all servers down =========================="
1914 CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid"
1916 start_mds || error "Unable to start mds"
1917 #first client should see only one ost
1918 start_ost || error "Unable to start OST1"
1919 wait_osc_import_state mds ost FULL
1921 start_ost2 || error "Unable to start OST2"
1922 wait_osc_import_state mds ost2 FULL
1924 # Wait for client to detect down OST
1925 stop_ost2 || error "Unable to stop OST2"
1927 wait_osc_import_state mds ost2 DISCONN
1928 mount_client $MOUNT || error "Unable to mount client"
1929 lctl set_param llite.$FSNAME-*.lazystatfs=0
1931 multiop_bg_pause $MOUNT _f
1935 if [ $RC1 -ne 0 ]; then
1936 log "lazystatfs multiop failed $RC1"
1939 sleep $(( $TIMEOUT+1 ))
1941 [ $? -ne 0 ] && error "process isn't sleep"
1942 start_ost2 || error "Unable to start OST2"
1943 wait $pid || error "statfs failed"
1944 stop_ost2 || error "Unable to stop OST2"
1947 umount_client $MOUNT || error "Unable to unmount client"
1948 stop_ost || error "Unable to stop OST1"
1949 stop_mds || error "Unable to stop MDS"
1952 run_test 50f "normal statfs one server in down =========================="
1955 local LOCAL_TIMEOUT=20
1961 check_mount || return 1
1964 $LFS setstripe -c -1 $MOUNT/d1
1965 #define OBD_FAIL_MDS_REINT_DELAY 0x142
1966 do_facet $SINGLEMDS "lctl set_param fail_loc=0x142"
1967 touch $MOUNT/d1/f1 &
1970 start_ost2 || return 2
1972 stop_ost2 || return 3
1975 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
1984 do_node $node mkdir -p $dest
1985 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
1987 do_node $node 'tar cf - '$@' | tar xf - -C '$dest';
1988 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
1989 [ $? -eq 0 ] || { error "Unable to tar files"; return 2; }
1991 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs
1992 [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; }
2002 local backup2=${TMP}/backup2
2004 do_node $node mkdir -p $backup2
2005 [ $? -eq 0 ] || { error "Unable to create directory"; return 1; }
2007 do_node $node 'tar cf - '$@' | tar xf - -C '$backup2';
2008 [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1'
2009 [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; }
2011 do_node $node "diff -rq $backup $backup2"
2012 [ $? -eq 0 ] || { error "contents differ"; return 3; }
2014 local xattrs2=${TMP}/xattrs2
2015 do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2
2016 [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; }
2018 do_node $node "diff $xattrs $xattrs2"
2019 [ $? -eq 0 ] || { error "xattrs differ"; return 5; }
2021 do_node $node "rm -rf $backup2 $xattrs2"
2022 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; }
2027 [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
2029 [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; }
2031 [ $? -eq 0 ] || { error "Unable to mount client"; return 3; }
2034 local ost1mnt=${MOUNT%/*}/ost1
2035 local ost1node=$(facet_active_host ost1)
2036 local ost1tmp=$TMP/conf52
2039 [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; }
2040 touch $TMP/modified_first
2041 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; }
2042 do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first"
2043 [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; }
2046 $LFS setstripe $DIR/$tdir -c -1 -s 1M
2047 [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; }
2049 for (( i=0; i < nrfiles; i++ )); do
2050 multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c
2051 [ $? -eq 0 ] || { error "multiop failed"; return 8; }
2057 echo backup files to $TMP/files
2058 local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first)
2059 copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2060 [ $? -eq 0 ] || { error "Unable to copy files"; return 9; }
2062 umount_client $MOUNT
2063 [ $? -eq 0 ] || { error "Unable to umount client"; return 10; }
2065 [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; }
2067 echo mount ost1 as ldiskfs
2068 do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS
2069 [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; }
2072 echo backup objects to $ost1tmp/objects
2073 local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
2074 '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
2075 copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2076 [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
2078 # move objects to lost+found
2079 do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found'
2080 [ $? -eq 0 ] || { error "Unable to move objects"; return 14; }
2083 do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found"
2084 [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; }
2086 # compare restored objects against saved ones
2087 diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
2088 [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; }
2090 do_node $ost1node "umount $ost1_dev"
2091 [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; }
2094 [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; }
2096 [ $? -eq 0 ] || { error "Unable to mount client"; return 19; }
2099 diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files
2100 [ $? -eq 0 ] || { error "Unable to diff files"; return 20; }
2102 rm -rf $TMP/files $TMP/file_xattrs
2103 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; }
2104 do_node $ost1node "rm -rf $ost1tmp"
2105 [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; }
2108 run_test 52 "check recovering objects from lost+found"
2111 equals_msg `basename $0`: test complete
2112 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true