X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=32ddff14ef8e336f2cb045e63b3f205a4f05ab0e;hp=27e3e62aeff3a9d966adbb9ba94d3a2b5461e7e0;hb=c41d198c5ce4e4917edf1207e34f8a4811b77d46;hpb=467c1efa023d402205e9bbc72a840d2da4d35023

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 27e3e62..32ddff1 100644
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -9,16 +9,39 @@ set -e
 
 ONLY=${ONLY:-"$*"}
 
-# bug number for skipped test:
-#               15977
+# bug number for skipped test:     LU-2828
+ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT 59 64"
 # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
 
+is_sles11()						# LU-2181
+{
+	if [ -r /etc/SuSE-release ]
+	then
+		local vers=`grep VERSION /etc/SuSE-release | awk '{print $3}'`
+		local patchlev=`grep PATCHLEVEL /etc/SuSE-release \
+			| awk '{print $3}'`
+		if [ $vers -eq 11 ] && [ $patchlev -eq 2 ]
+		then
+			return 0
+		fi
+	fi
+	return 1
+}
+
+if is_sles11; then					# LU-2181
+	ALWAYS_EXCEPT="$ALWAYS_EXCEPT 23a 34b"
+fi
+
 if [ "$FAILURE_MODE" = "HARD" ]; then
 	CONFIG_EXCEPTIONS="24a " && \
 	echo "Except the tests: $CONFIG_EXCEPTIONS for FAILURE_MODE=$FAILURE_MODE, bug 23573" && \
 	ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
 fi
 
+# bug number for skipped test:
+# a tool to create lustre filesystem images
+ALWAYS_EXCEPT="32newtarball $ALWAYS_EXCEPT"
+
 SRCDIR=`dirname $0`
 PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH
 
@@ -52,6 +75,10 @@ if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then
 	LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init)
 fi
 
+[ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
+# bug number for skipped test:        LU-2778 LU-2059
+	ALWAYS_EXCEPT="$ALWAYS_EXCEPT 57b     50h"
+
 init_logging
 
 #
@@ -63,31 +90,6 @@ require_dsh_ost || exit 0
 
 assert_DIR
 
-reformat() {
-        formatall
-}
-
-writeconf1() {
-	local facet=$1
-	local dev=$2
-
-	stop ${facet} -f
-	rm -f ${facet}active
-	# who knows if/where $TUNEFS is installed?  Better reformat if it fails...
-	do_facet ${facet} "$TUNEFS --quiet --writeconf $dev" ||
-		{ echo "tunefs failed, reformatting instead" && reformat_and_config && return 1; }
-	return 0
-}
-
-writeconf() {
-	# we need ldiskfs
-	load_modules
-	# if writeconf fails anywhere, we reformat everything
-	writeconf1 mds `mdsdevname 1` || return 0
-	writeconf1 ost1 `ostdevname 1` || return 0
-	writeconf1 ost2 `ostdevname 2` || return 0
-}
-
 gen_config() {
 	# The MGS must be started before the OSTs for a new fs, so start
 	# and stop to generate the startup logs.
@@ -106,20 +108,54 @@ reformat_and_config() {
 	gen_config
 }
 
+writeconf_or_reformat() {
+	# There are at most 2 OSTs for write_conf test
+	# who knows if/where $TUNEFS is installed?
+	# Better reformat if it fails...
+	writeconf_all $MDSCOUNT 2 ||
+		{ echo "tunefs failed, reformatting instead" &&
+		  reformat_and_config && return 0; }
+	return 0
+}
+
+reformat() {
+        formatall
+}
+
 start_mgs () {
 	echo "start mgs"
 	start mgs $MGSDEV $MGS_MOUNT_OPTS
 }
 
-start_mds() {
-	local facet=$SINGLEMDS
-	# we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only
-	local num=$(echo $facet | tr -d "mds")
+start_mdt() {
+	local num=$1
+	local facet=mds$num
 	local dev=$(mdsdevname $num)
+	shift 1
+
 	echo "start mds service on `facet_active_host $facet`"
 	start $facet ${dev} $MDS_MOUNT_OPTS $@ || return 94
 }
 
+stop_mdt() {
+	local num=$1
+	local facet=mds$num
+	local dev=$(mdsdevname $num)
+	shift 1
+
+	echo "stop mds service on `facet_active_host $facet`"
+	# These tests all use non-failover stop
+	stop $facet -f  || return 97
+}
+
+start_mds() {
+	local num
+
+	for num in $(seq $MDSCOUNT); do
+		start_mdt $num $@ || return 94
+	done
+}
+
 start_mgsmds() {
 	if ! combined_mgs_mds ; then
 		start_mgs
@@ -128,9 +164,10 @@ start_mgsmds() {
 }
 
 stop_mds() {
-	echo "stop mds service on `facet_active_host $SINGLEMDS`"
-	# These tests all use non-failover stop
-	stop $SINGLEMDS -f  || return 97
+	local num
+	for num in $(seq $MDSCOUNT); do
+		stop_mdt $num || return 97
+	done
 }
 
 stop_mgs() {
@@ -277,9 +314,9 @@ test_1() {
 run_test 1 "start up ost twice (should return errors)"
 
 test_2() {
-	start_mds
+	start_mdt 1
 	echo "start mds second time.."
-	start_mds && error "2nd MDT start should fail"
+	start_mdt 1 && error "2nd MDT start should fail"
 	start_ost
 	mount_client $MOUNT
 	check_mount || return 43
@@ -414,6 +451,9 @@ test_5d() {
 	grep " $MOUNT " /etc/mtab && \
 		error false "unexpected entry in mtab before mount" && return 10
 
+	[ "$(facet_fstype ost1)" = "zfs" ] &&
+		skip "LU-2059: no local config for ZFS OSTs" && return
+
 	local rc=0
 	start_ost
 	start_mds
@@ -558,24 +598,30 @@ is_blkdev () {
 #
 
 test_17() {
-        setup
-        check_mount || return 41
-        cleanup || return $?
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
 
-        echo "Remove mds config log"
-        if ! combined_mgs_mds ; then
-                stop mgs
-        fi
+	setup
+	check_mount || return 41
+	cleanup || return $?
 
-        do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' $MGSDEV || return \$?" || return $?
+	echo "Remove mds config log"
+	if ! combined_mgs_mds ; then
+		stop mgs
+	fi
 
-        if ! combined_mgs_mds ; then
-                start_mgs
-        fi
+	do_facet mgs "$DEBUGFS -w -R 'unlink CONFIGS/$FSNAME-MDT0000' \
+		$(mgsdevname) || return \$?" || return $?
 
-        start_ost
-        start_mds && return 42
-        reformat_and_config
+	if ! combined_mgs_mds ; then
+		start_mgs
+	fi
+
+	start_ost
+	start_mds && return 42
+	reformat_and_config
 }
 run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)"
 
@@ -600,8 +646,14 @@ test_18() {
                 log "use STORED_MDSSIZE=$STORED_MDSSIZE"
 
         # check if the block device is large enough
-        [ -z "$OK" ] && $(is_blkdev $SINGLEMDS $MDSDEV $MIN) && OK=1 &&
-                myMDSSIZE=$MIN && log "use device $MDSDEV with MIN=$MIN"
+	is_blkdev $SINGLEMDS $MDSDEV $MIN
+	local large_enough=$?
+	if [ -n "$OK" ]; then
+		[ $large_enough -ne 0 ] && OK=""
+	else
+		[ $large_enough -eq 0 ] && OK=1 && myMDSSIZE=$MIN &&
+			log "use device $MDSDEV with MIN=$MIN"
+	fi
 
         # check if a loopback device has enough space for fs metadata (5%)
 
@@ -648,6 +700,9 @@ test_19a() {
 run_test 19a "start/stop MDS without OSTs"
 
 test_19b() {
+	[ "$(facet_fstype ost1)" = "zfs" ] &&
+		skip "LU-2059: no local config for ZFS OSTs" && return
+
 	start_ost || return 1
 	stop_ost -f || return 2
 }
@@ -684,6 +739,9 @@ test_21a() {
 run_test 21a "start mds before ost, stop ost first"
 
 test_21b() {
+	[ "$(facet_fstype ost1)" = "zfs" ] &&
+		skip "LU-2059: no local config for ZFS OSTs" && return
+
         start_ost
 	start_mds
         wait_osc_import_state mds ost FULL
@@ -701,7 +759,7 @@ test_21c() {
 	stop_ost2
 	stop_mds
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 21c "start mds between two osts, stop mds last"
 
@@ -724,18 +782,22 @@ test_21d() {
         stop_mds
         stop_mgs
         #writeconf to remove all ost2 traces for subsequent tests
-        writeconf
+	writeconf_or_reformat
         start_mgs
 }
 run_test 21d "start mgs then ost and then mds"
 
 test_22() {
+	local num
+
 	start_mds
 
 	echo Client mount with ost in logs, but none running
 	start_ost
 	# wait until mds connected to ost and open client connection
-        wait_osc_import_state mds ost FULL
+	for num in $(seq 1 $MDSCOUNT); do
+		wait_osc_import_state mds${num} ost FULL
+	done
 	stop_ost
 	mount_client $MOUNT
 	# check_mount will block trying to contact ost
@@ -755,8 +817,10 @@ test_22() {
 		sleep $((TIMEOUT + TIMEOUT + TIMEOUT))
 	fi
 	mount_client $MOUNT
-        wait_osc_import_state mds ost FULL
-        wait_osc_import_state client ost FULL
+	for num in $(seq 1 $MDSCOUNT); do
+		wait_osc_import_state mds${num} ost FULL
+	done
+	wait_osc_import_state client ost FULL
 	check_mount || return 41
 	pass
 
@@ -864,11 +928,12 @@ test_24a() {
 
 	# test 8-char fsname as well
 	local FSNAME2=test1234
-	add fs2mds $(mkfs_opts mds1) --nomgs --mgsnode=$MGSNID \
+
+	add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --nomgs --mgsnode=$MGSNID \
 		--fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10
 
-	add fs2ost $(mkfs_opts ost1) --fsname=${FSNAME2} --reformat \
-		$fs2ostdev $fs2ostvdev || exit 10
+	add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --fsname=${FSNAME2} \
+		--reformat $fs2ostdev $fs2ostvdev || exit 10
 
 	setup
 	start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
@@ -914,8 +979,8 @@ test_24b() {
 	local fs2mdsdev=$(mdsdevname 1_2)
 	local fs2mdsvdev=$(mdsvdevname 1_2)
 
-	add fs2mds $(mkfs_opts mds1) --mgs --fsname=${FSNAME}2 --reformat \
-		$fs2mdsdev $fs2mdsvdev || exit 10
+	add fs2mds $(mkfs_opts mds1 ${fs2mdsdev} ) --mgs --fsname=${FSNAME}2 \
+		--reformat $fs2mdsdev $fs2mdsvdev || exit 10
 	setup
 	start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2
 	cleanup || return 6
@@ -948,41 +1013,34 @@ test_26() {
 }
 run_test 26 "MDT startup failure cleans LOV (should return errs)"
 
-set_and_check() {
-	local myfacet=$1
-	local TEST=$2
-	local PARAM=$3
-	local ORIG=$(do_facet $myfacet "$TEST")
-	if [ $# -gt 3 ]; then
-	    local FINAL=$4
-	else
-	    local -i FINAL
-	    FINAL=$(($ORIG + 5))
-	fi
-	echo "Setting $PARAM from $ORIG to $FINAL"
-	do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed
-
-	wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed!
-}
-
 test_27a() {
+	[ "$(facet_fstype ost1)" = "zfs" ] &&
+		skip "LU-2059: no local config for ZFS OSTs" && return
+
 	start_ost || return 1
 	start_mds || return 2
 	echo "Requeue thread should have started: "
 	ps -e | grep ll_cfg_requeue
-	set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
+	set_conf_param_and_check ost1					      \
+	   "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \
+	   "$FSNAME-OST0000.ost.client_cache_seconds" || return 3
 	cleanup_nocli
 }
 run_test 27a "Reacquire MGS lock if OST started first"
 
 test_27b() {
 	# FIXME. ~grev
-        setup
-        local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }')
+	setup
+	local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" |
+			awk '($3 ~ "mdt" && $4 ~ "MDT0000") { print $4 }')
 
 	facet_failover $SINGLEMDS
-	set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3
-	set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4
+	set_conf_param_and_check $SINGLEMDS				\
+		"lctl get_param -n mdt.$device.identity_acquire_expire"	\
+		"$device.mdt.identity_acquire_expire" || return 3
+	set_conf_param_and_check client					\
+		"lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\
+		"$device.mdc.max_rpcs_in_flight" || return 4
 	check_mount
 	cleanup
 }
@@ -994,9 +1052,9 @@ test_28() {
 	PARAM="$FSNAME.llite.max_read_ahead_whole_mb"
 	ORIG=$($TEST)
 	FINAL=$(($ORIG + 1))
-	set_and_check client "$TEST" "$PARAM" $FINAL || return 3
+	set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3
 	FINAL=$(($FINAL + 1))
-	set_and_check client "$TEST" "$PARAM" $FINAL || return 4
+	set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4
 	umount_client $MOUNT || return 200
 	mount_client $MOUNT
 	RESULT=$($TEST)
@@ -1006,7 +1064,7 @@ test_28() {
 	else
 	    echo "New config success: got $RESULT"
 	fi
-	set_and_check client "$TEST" "$PARAM" $ORIG || return 5
+	set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5
 	cleanup
 }
 run_test 28 "permanent parameter setting"
@@ -1023,7 +1081,8 @@ test_29() {
 
         ACTV=$(lctl get_param -n $PROC_ACT)
 	DEAC=$((1 - $ACTV))
-	set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
+	set_conf_param_and_check client \
+		"lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2
         # also check ost_server_uuid status
 	RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV)
 	if [ -z "$RESULT" ]; then
@@ -1033,31 +1092,30 @@ test_29() {
 	    echo "Live client success: got $RESULT"
 	fi
 
-	# check MDT too
-	local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0001)
-	mdtosc=${mdtosc/-MDT*/-MDT\*}
-	local MPROC="osc.$mdtosc.active"
-	local MAX=30
-	local WAIT=0
-	while [ 1 ]; do
-	    sleep 5
-	    RESULT=`do_facet $SINGLEMDS " lctl get_param -n $MPROC"`
-	    [ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
-	    if [ $RESULT -eq $DEAC ]; then
-		echo "MDT deactivated also after $WAIT sec (got $RESULT)"
-		break
-	    fi
-	    WAIT=$((WAIT + 5))
-	    if [ $WAIT -eq $MAX ]; then
-		echo "MDT not deactivated: wanted $DEAC got $RESULT"
-		return 4
-	    fi
-	    echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated"
+	# check MDTs too
+	for num in $(seq $MDSCOUNT); do
+		local mdtosc=$(get_mdtosc_proc_path mds${num} $FSNAME-OST0001)
+		local MPROC="osc.$mdtosc.active"
+		local MAX=30
+		local WAIT=0
+		while [ 1 ]; do
+			sleep 5
+			RESULT=$(do_facet mds${num} " lctl get_param -n $MPROC")
+			[ ${PIPESTATUS[0]} = 0 ] || error "Can't read $MPROC"
+			if [ $RESULT -eq $DEAC ]; then
+				echo -n "MDT deactivated also after"
+				echo "$WAIT sec (got $RESULT)"
+				break
+			fi
+			WAIT=$((WAIT + 5))
+			if [ $WAIT -eq $MAX ]; then
+				echo -n "MDT not deactivated: wanted $DEAC"
+				echo  "got $RESULT"
+				return 4
+			fi
+			echo "Waiting $(($MAX - $WAIT))secs for MDT deactivated"
+		done
 	done
-
-        # quotacheck should not fail immediately after deactivate
-	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
-
         # test new client starts deactivated
 	umount_client $MOUNT || return 200
 	mount_client $MOUNT
@@ -1069,17 +1127,15 @@ test_29() {
 	    echo "New client success: got $RESULT"
 	fi
 
-        # quotacheck should not fail after umount/mount operation
-	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
-
 	# make sure it reactivates
-	set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
+	set_conf_param_and_check client \
+		"lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6
 
 	umount_client $MOUNT
 	stop_ost2
 	cleanup_nocli
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 29 "permanently remove an OST"
 
@@ -1091,7 +1147,8 @@ test_30a() {
 	ORIG=$($TEST)
 	LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5)
 	for i in ${LIST[@]}; do
-	    set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
+		set_conf_param_and_check client "$TEST" \
+			"$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3
 	done
 	# make sure client restart still works
 	umount_client $MOUNT
@@ -1125,7 +1182,9 @@ test_30b() {
 	echo "Using fake nid $NEW"
 
 	TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
-	set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW"
+	set_conf_param_and_check client "$TEST" \
+		"$FSNAME-OST0000.failover.node" $NEW ||
+		error "didn't add failover nid $NEW"
 	NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
 	echo $NIDS
 	NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1))
@@ -1152,192 +1211,669 @@ test_31() { # bug 10734
 }
 run_test 31 "Connect to non-existent node (shouldn't crash)"
 
-# Use these start32/stop32 fn instead of t-f start/stop fn,
-# for local devices, to skip global facet vars init
-stop32 () {
-	local facet=$1
-	shift
-	echo "Stopping local ${MOUNT%/*}/${facet} (opts:$@)"
-	umount -d $@ ${MOUNT%/*}/${facet}
-	losetup -a
-}
-
-start32 () {
-	local facet=$1
-	shift
-	local device=$1
-	shift
-	mkdir -p ${MOUNT%/*}/${facet}
-
-	echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}"
-	mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}
-	local RC=$?
-	if [ $RC -ne 0 ]; then
-		echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}"
-		echo "Start of ${device} of local ${facet} failed ${RC}"
-	fi
-	losetup -a
-	return $RC
-}
 
-cleanup_nocli32 () {
-	stop32 mds1 -f
-	stop32 ost1 -f
-	wait_exit_ST client
-}
+T32_QID=60000
+T32_BLIMIT=20480 # Kbytes
+T32_ILIMIT=2
 
-cleanup_32() {
-	trap 0
-	echo "Cleanup test_32 umount $MOUNT ..."
-	umount -f $MOUNT || true
-	echo "Cleanup local mds ost1 ..."
-	cleanup_nocli32
-	combined_mgs_mds || start_mgs
-	unload_modules_conf
+#
+# This is not really a test but a tool to create new disk
+# image tarballs for the upgrade tests.
+#
+# Disk image tarballs should be created on single-node
+# clusters by running this test with default configurations
+# plus a few mandatory environment settings that are verified
+# at the beginning of the test.
+#
+test_32newtarball() {
+	local version
+	local dst=.
+	local src=/etc/rc.d
+	local tmp=$TMP/t32_image_create
+
+	if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o								\
+		 \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o			\
+		 -z "$OSTDEV1" ]; then
+		error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>"	\
+			  "(or MDSDEV, in the case of b1_8) OSTCOUNT=1"					\
+			  "OSTDEV1=<nonexistent_file>"
+	fi
+
+	mkdir $tmp || {
+		echo "Found stale $tmp"
+		return 1
+	}
+
+	mkdir $tmp/src
+	tar cf - -C $src . | tar xf - -C $tmp/src
+	dd if=/dev/zero of=$tmp/src/t32_qf_old bs=1M \
+		count=$(($T32_BLIMIT / 1024 / 2))
+	chown $T32_QID.$T32_QID $tmp/src/t32_qf_old
+
+	formatall
+
+	setupall
+
+	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
+		$LFS quotacheck -ug /mnt/$FSNAME
+	$LFS setquota -u $T32_QID -b 0 -B $T32_BLIMIT -i 0 -I $T32_ILIMIT \
+		/mnt/$FSNAME
+
+	tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME
+	stopall
+
+	mkdir $tmp/img
+
+	setupall
+	pushd /mnt/$FSNAME
+	ls -Rni --time-style=+%s >$tmp/img/list
+	find . ! -name .lustre -type f -exec sha1sum {} \; |
+		sort -k 2 >$tmp/img/sha1sums
+	popd
+	$LCTL get_param -n version | head -n 1 |
+		sed -e 's/^lustre: *//' >$tmp/img/commit
+
+	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ] &&
+		$LFS quotaon -ug /mnt/$FSNAME
+	$LFS quota -u $T32_QID -v /mnt/$FSNAME
+	$LFS quota -v -u $T32_QID /mnt/$FSNAME |
+		awk 'BEGIN { num='1' } { if ($1 == "'/mnt/$FSNAME'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*" > $tmp/img/bspace
+	$LFS quota -v -u $T32_QID /mnt/$FSNAME |
+		awk 'BEGIN { num='5' } { if ($1 == "'/mnt/$FSNAME'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*" > $tmp/img/ispace
+
+	stopall
+
+	pushd $tmp/src
+	find -type f -exec sha1sum {} \; | sort -k 2 >$tmp/sha1sums.src
+	popd
+
+	if ! diff -u $tmp/sha1sums.src $tmp/img/sha1sums; then
+		echo "Data verification failed"
+	fi
+
+	uname -r >$tmp/img/kernel
+	uname -m >$tmp/img/arch
+
+	mv ${MDSDEV1:-$MDSDEV} $tmp/img
+	mv $OSTDEV1 $tmp/img
+
+	version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
+			  sed -e 's/\./_/g')	# E.g., "1.8.7" -> "1_8"
+	dst=$(cd $dst; pwd)
+	pushd $tmp/img
+	tar cjvf $dst/disk$version-$(facet_fstype $SINGLEMDS).tar.bz2 -S *
+	popd
+
+	rm -r $tmp
 }
+#run_test 32newtarball "Create a new test_32 disk image tarball for this version"
 
-test_32a() {
-	client_only && skip "client only testing" && return 0
-	[ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
-	[ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0
+#
+# The list of applicable tarballs is returned via the caller's
+# variable "tarballs".
+#
+t32_check() {
+	local node=$(facet_active_host $SINGLEMDS)
+	local r="do_node $node"
 
-	local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
-	[ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
-	local tmpdir=$TMP/conf32a
-	mkdir -p $tmpdir
-	tar xjvf $DISK1_8 -C $tmpdir || \
-		{ skip_env "Cannot untar $DISK1_8" && return 0; }
+	if [ "$CLIENTONLY" ]; then
+		skip "Client-only testing"
+		exit 0
+	fi
 
-	load_modules
-	$LCTL set_param debug="$PTLDEBUG"
+	if ! $r which $TUNEFS; then
+		skip_env "tunefs.lustre required on $node"
+		exit 0
+	fi
 
-	$TUNEFS $tmpdir/mds || error "tunefs failed"
+	if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
+		skip "LU-2200: Test cannot run over Infiniband"
+		exit 0
+	fi
 
-	combined_mgs_mds || stop mgs
+	local IMGTYPE=$(facet_fstype $SINGLEMDS)
 
-	# nids are wrong, so client wont work, but server should start
-	start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
-		trap cleanup_32 EXIT INT || return 3
+	tarballs=$($r find $RLUSTRE/tests -maxdepth 1 -name \'disk*-$IMGTYPE.tar.bz2\')
 
-	local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid)
-	echo MDS uuid $UUID
-	[ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
+	if [ -z "$tarballs" ]; then
+		skip "No applicable tarballs found"
+		exit 0
+	fi
+}
 
-	$TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed"
-	start32 ost1 $tmpdir/ost1 "-o loop" || return 5
-	UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid)
-	echo OST uuid $UUID
-	[ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID"
+t32_test_cleanup() {
+	local node=$(facet_active_host $SINGLEMDS)
+	local r="do_node $node"
+	local tmp=$TMP/t32
+	local rc=$?
 
-	local NID=$($LCTL list_nids | head -1)
+	if $shall_cleanup_lustre; then
+		umount $tmp/mnt/lustre || rc=$?
+	fi
+	if $shall_cleanup_mdt; then
+		$r umount -d $tmp/mnt/mdt || rc=$?
+	fi
+	if $shall_cleanup_mdt1; then
+		$r umount -d $tmp/mnt/mdt1 || rc=$?
+	fi
+	if $shall_cleanup_ost; then
+		$r umount -d $tmp/mnt/ost || rc=$?
+	fi
+	$r rm -rf $tmp || rc=$?
+	rm -rf $tmp || rc=$?
+	return $rc
+}
 
-	echo "OSC changes should succeed:"
-	$LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7
-	$LCTL conf_param lustre-OST0000.failover.node=$NID || return 8
-	echo "ok."
+t32_bits_per_long() {
+	#
+	# Yes, this is not meant to be perfect.
+	#
+	case $1 in
+		ppc64|x86_64)
+			echo -n 64;;
+		i*86)
+			echo -n 32;;
+	esac
+}
 
-	echo "MDC changes should succeed:"
-	$LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
-	$LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10
-	echo "ok."
+t32_reload_modules() {
+	local node=$1
+	local all_removed=false
+	local i=0
+
+	while ((i < 20)); do
+		echo "Unloading modules on $node: Attempt $i"
+		do_rpc_nodes $node $LUSTRE_RMMOD $(facet_fstype $SINGLEMDS) &&
+			all_removed=true
+		do_rpc_nodes $node check_mem_leak || return 1
+		if $all_removed; then
+			load_modules
+			return 0
+		fi
+		sleep 5
+		i=$((i + 1))
+	done
+	echo "Unloading modules on $node: Given up"
+	return 1
+}
 
-	echo "LOV changes should succeed:"
-	$LCTL pool_new lustre.interop || return 11
-	$LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12
-	echo "ok."
+t32_wait_til_devices_gone() {
+	local node=$1
+	local devices
+	local i=0
+
+	echo wait for devices to go
+	while ((i < 20)); do
+		devices=$(do_rpc_nodes $node $LCTL device_list | wc -l)
+		echo $device
+		((devices == 0)) && return 1
+		sleep 5
+		i=$((i + 1))
+	done
+	echo "waiting for devices on $node: Given up"
+	return 1
+}
 
-	cleanup_32
+t32_verify_quota() {
+	local node=$1
+	local fsname=$2
+	local mnt=$3
+	local fstype=$(facet_fstype $SINGLEMDS)
+	local qval
+	local cmd
+
+	$LFS quota -u $T32_QID -v $mnt
+
+	qval=$($LFS quota -v -u $T32_QID $mnt |
+		awk 'BEGIN { num='1' } { if ($1 == "'$mnt'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*")
+	[ $qval -eq $img_bspace ] || {
+		echo "bspace, act:$qval, exp:$img_bspace"
+		return 1
+	}
+
+	qval=$($LFS quota -v -u $T32_QID $mnt |
+		awk 'BEGIN { num='5' } { if ($1 == "'$mnt'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*")
+	[ $qval -eq $img_ispace ] || {
+		echo "ispace, act:$qval, exp:$img_ispace"
+		return 1
+	}
+
+	qval=$($LFS quota -v -u $T32_QID $mnt |
+		awk 'BEGIN { num='3' } { if ($1 == "'$mnt'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*")
+	[ $qval -eq $T32_BLIMIT ] || {
+		echo "blimit, act:$qval, exp:$T32_BLIMIT"
+		return 1
+	}
+
+	qval=$($LFS quota -v -u $T32_QID $mnt |
+		awk 'BEGIN { num='7' } { if ($1 == "'$mnt'") \
+		{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
+		| tr -d "*")
+	[ $qval -eq $T32_ILIMIT ] || {
+		echo "ilimit, act:$qval, exp:$T32_ILIMIT"
+		return 1
+	}
 
-	# mount a second time to make sure we didnt leave upgrade flag on
-	load_modules
-	$TUNEFS --dryrun $tmpdir/mds || error "tunefs failed"
+	do_node $node $LCTL conf_param $fsname.quota.mdt=ug
+	cmd="$LCTL get_param -n osd-$fstype.$fsname-MDT0000"
+	cmd=$cmd.quota_slave.enabled
+	wait_update $node "$cmd" "ug" || {
+		echo "Enable mdt quota failed"
+		return 1
+	}
 
-	combined_mgs_mds || stop mgs
+	do_node $node $LCTL conf_param $fsname.quota.ost=ug
+	cmd="$LCTL get_param -n osd-$fstype.$fsname-OST0000"
+	cmd=$cmd.quota_slave.enabled
+	wait_update $node "$cmd" "ug" || {
+		echo "Enable ost quota failed"
+		return 1
+	}
 
-	start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \
-		trap cleanup_32 EXIT INT || return 12
+	chmod 0777 $mnt
+	runas -u $T32_QID -g $T32_QID dd if=/dev/zero of=$mnt/t32_qf_new \
+		bs=1M count=$(($T32_BLIMIT / 1024)) oflag=sync && {
+		echo "Write succeed, but expect -EDQUOT"
+		return 1
+	}
+	rm -f $mnt/t32_qf_new
 
-	cleanup_32
+	runas -u $T32_QID -g $T32_QID createmany -m $mnt/t32_qf_ \
+		$T32_ILIMIT && {
+		echo "Create succeed, but expect -EDQUOT"
+		return 1
+	}
+	unlinkmany $mnt/t32_qf_ $T32_ILIMIT
 
-	rm -rf $tmpdir || true	# true is only for TMP on NFS
+	return 0
 }
-run_test 32a "Upgrade from 1.8 (not live)"
 
-test_32b() {
-	client_only && skip "client only testing" && return 0
-	[ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; }
-	[ -z "$TUNEFS" ] && skip_env "No tunefs" && return
+t32_test() {
+	local tarball=$1
+	local writeconf=$2
+	local dne_upgrade=${dne_upgrade:-"no"}
+	local shall_cleanup_mdt=false
+	local shall_cleanup_mdt1=false
+	local shall_cleanup_ost=false
+	local shall_cleanup_lustre=false
+	local node=$(facet_active_host $SINGLEMDS)
+	local r="do_node $node"
+	local node2=$(facet_active_host mds2)
+	local r2="do_node $node2"
+	local tmp=$TMP/t32
+	local img_commit
+	local img_kernel
+	local img_arch
+	local img_bspace
+	local img_ispace
+	local fsname=t32fs
+	local nid=$($r $LCTL list_nids | head -1)
+	local mopts
+	local uuid
+	local nrpcs_orig
+	local nrpcs
+	local list
+	local fstype=$(facet_fstype $SINGLEMDS)
+
+	trap 'trap - RETURN; t32_test_cleanup' RETURN
+
+	mkdir -p $tmp/mnt/lustre
+	$r mkdir -p $tmp/mnt/{mdt,ost}
+	$r tar xjvf $tarball -S -C $tmp || {
+		error_noexit "Unpacking the disk image tarball"
+		return 1
+	}
+	img_commit=$($r cat $tmp/commit)
+	img_kernel=$($r cat $tmp/kernel)
+	img_arch=$($r cat $tmp/arch)
+	img_bspace=$($r cat $tmp/bspace)
+	img_ispace=$($r cat $tmp/ispace)
+	echo "Upgrading from $(basename $tarball), created with:"
+	echo "  Commit: $img_commit"
+	echo "  Kernel: $img_kernel"
+	echo "    Arch: $img_arch"
+
+	$r $LCTL set_param debug="$PTLDEBUG"
+
+	$r $TUNEFS --dryrun $tmp/mdt || {
+		error_noexit "tunefs.lustre before mounting the MDT"
+		return 1
+	}
+	if [ "$writeconf" ]; then
+		mopts=loop,writeconf
+		if [ $fstype == "ldiskfs" ]; then
+			$r $TUNEFS --quota $tmp/mdt || {
+				error_noexit "Enable mdt quota feature"
+				return 1
+			}
+		fi
+	else
+		mopts=loop,exclude=$fsname-OST0000
+	fi
+
+	t32_wait_til_devices_gone $node
 
-	local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2
-	[ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0
-	local tmpdir=$TMP/conf32b
-	mkdir -p $tmpdir
-	tar xjvf $DISK1_8 -C $tmpdir || \
-		{ skip_env "Cannot untar $DISK1_8" && return ; }
+	$r mount -t lustre -o $mopts $tmp/mdt $tmp/mnt/mdt || {
+		error_noexit "Mounting the MDT"
+		return 1
+	}
+	shall_cleanup_mdt=true
+
+	if [ "$dne_upgrade" != "no" ]; then
+		echo "mkfs new MDT...."
+		add mds2 $(mkfs_opts mds2 $(mdsdevname 2) $fsname) --reformat \
+			$(mdsdevname 2) $(mdsvdevname 2) > /dev/null || {
+			error_noexit "Mkfs new MDT failed"
+			return 1
+		}
+
+		$r2 $TUNEFS --dryrun $(mdsdevname 2) || {
+			error_noexit "tunefs.lustre before mounting the MDT"
+			return 1
+		}
+
+		echo "mount new MDT...."
+		$r2 mkdir -p $tmp/mnt/mdt1
+		$r2 mount -t lustre -o $mopts $(mdsdevname 2) $tmp/mnt/mdt1 || {
+			error_noexit "mount mdt1 failed"
+			return 1
+		}
+		shall_cleanup_mdt1=true
+	fi
 
-	load_modules
-	$LCTL set_param debug="+config"
-	local NEWNAME=lustre
+	uuid=$($r $LCTL get_param -n mdt.$fsname-MDT0000.uuid) || {
+		error_noexit "Getting MDT UUID"
+		return 1
+	}
+	if [ "$uuid" != $fsname-MDT0000_UUID ]; then
+		error_noexit "Unexpected MDT UUID: \"$uuid\""
+		return 1
+	fi
 
-	# writeconf will cause servers to register with their current nids
-	$TUNEFS --writeconf --erase-params \
-        --param mdt.identity_upcall=$L_GETIDENTITY \
-        --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed"
-	combined_mgs_mds || stop mgs
+	$r $TUNEFS --dryrun $tmp/ost || {
+		error_noexit "tunefs.lustre before mounting the OST"
+		return 1
+	}
+	if [ "$writeconf" ]; then
+		mopts=loop,mgsnode=$nid,$writeconf
+		if [ $fstype == "ldiskfs" ]; then
+			$r $TUNEFS --quota $tmp/ost || {
+				error_noexit "Enable ost quota feature"
+				return 1
+			}
+		fi
+	else
+		mopts=loop,mgsnode=$nid
+	fi
+	$r mount -t lustre -o $mopts $tmp/ost $tmp/mnt/ost || {
+		error_noexit "Mounting the OST"
+		return 1
+	}
+	shall_cleanup_ost=true
+
+	uuid=$($r $LCTL get_param -n obdfilter.$fsname-OST0000.uuid) || {
+		error_noexit "Getting OST UUID"
+		return 1
+	}
+	if [ "$uuid" != $fsname-OST0000_UUID ]; then
+		error_noexit "Unexpected OST UUID: \"$uuid\""
+		return 1
+	fi
 
-	start32 mds1 $tmpdir/mds "-o loop" && \
-		trap cleanup_32 EXIT INT || return 3
+	$r $LCTL conf_param $fsname-OST0000.osc.max_dirty_mb=15 || {
+		error_noexit "Setting \"max_dirty_mb\""
+		return 1
+	}
+	$r $LCTL conf_param $fsname-OST0000.failover.node=$nid || {
+		error_noexit "Setting OST \"failover.node\""
+		return 1
+	}
+	$r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=9 || {
+		error_noexit "Setting \"max_rpcs_in_flight\""
+		return 1
+	}
+	$r $LCTL conf_param $fsname-MDT0000.failover.node=$nid || {
+		error_noexit "Setting MDT \"failover.node\""
+		return 1
+	}
+	$r $LCTL pool_new $fsname.interop || {
+		error_noexit "Setting \"interop\""
+		return 1
+	}
+	$r $LCTL conf_param $fsname-MDT0000.lov.stripesize=4M || {
+		error_noexit "Setting \"lov.stripesize\""
+		return 1
+	}
+
+	if [ "$dne_upgrade" != "no" ]; then
+		$r $LCTL conf_param \
+				$fsname-MDT0001.mdc.max_rpcs_in_flight=9 || {
+			error_noexit "Setting MDT1 \"max_rpcs_in_flight\""
+			return 1
+		}
+		$r $LCTL conf_param $fsname-MDT0001.failover.node=$nid || {
+			error_noexit "Setting MDT1 \"failover.node\""
+			return 1
+		}
+		$r $LCTL conf_param $fsname-MDT0001.lov.stripesize=4M || {
+			error_noexit "Setting MDT1 \"lov.stripesize\""
+			return 1
+		}
 
-	local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid)
-	echo MDS uuid $UUID
-	[ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID"
+	fi
 
-	$TUNEFS  --writeconf --erase-params \
-        --mgsnode=$HOSTNAME --fsname=$NEWNAME $tmpdir/ost1 ||\
-	    error "tunefs failed"
-	start32 ost1 $tmpdir/ost1 "-o loop" || return 5
-	UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid)
-	echo OST uuid $UUID
-	[ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID"
+	if [ "$writeconf" ]; then
+		mount -t lustre $nid:/$fsname $tmp/mnt/lustre || {
+			error_noexit "Mounting the client"
+			return 1
+		}
+		shall_cleanup_lustre=true
+		$LCTL set_param debug="$PTLDEBUG"
+
+		t32_verify_quota $node $fsname $tmp/mnt/lustre || {
+			error_noexit "verify quota failed"
+			return 1
+		}
+
+		if [ "$dne_upgrade" != "no" ]; then
+			$LFS mkdir -i 1 $tmp/mnt/lustre/remote_dir || {
+				error_noexit "set remote dir failed"
+				return 1
+			}
+
+			pushd $tmp/mnt/lustre
+			tar -cf - . --exclude=./remote_dir |
+				tar -xvf - -C remote_dir 1>/dev/null || {
+				error_noexit "cp to remote dir failed"
+				return 1
+			}
+			popd
+		fi
 
-	local NID=$($LCTL list_nids | head -1)
+		dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
+			error_noexit "dd failed"
+			return 1
+		}
+		rm -rf $tmp/mnt/lustre/tmp_file || {
+			error_noexit "rm failed"
+			return 1
+		}
+
+		if $r test -f $tmp/sha1sums; then
+			# LU-2393 - do both sorts on same node to ensure locale
+			# is identical
+			$r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
+			if [ "$dne_upgrade" != "no" ]; then
+				pushd $tmp/mnt/lustre/remote_dir
+			else
+				pushd $tmp/mnt/lustre
+			fi
+
+			find ! -name .lustre -type f -exec sha1sum {} \; |
+				sort -k 2 >$tmp/sha1sums || {
+				error_noexit "sha1sum"
+				return 1
+			}
+			popd
+			if ! diff -ub $tmp/sha1sums.orig $tmp/sha1sums; then
+				error_noexit "sha1sum verification failed"
+				return 1
+			fi
+		else
+			echo "sha1sum verification skipped"
+		fi
 
-	echo "OSC changes should succeed:"
-	$LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7
-	$LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8
-	echo "ok."
+		if [ "$dne_upgrade" != "no" ]; then
+			rm -rf $tmp/mnt/lustre/remote_dir || {
+				error_noexit "remove remote dir failed"
+				return 1
+			}
+		fi
 
-	echo "MDC changes should succeed:"
-	$LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9
-	$LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10
-	echo "ok."
+		if $r test -f $tmp/list; then
+			#
+			# There is not a Test Framework API to copy files to or
+			# from a remote node.
+			#
+			# LU-2393 - do both sorts on same node to ensure locale
+			# is identical
+			$r cat $tmp/list | sort -k 6 >$tmp/list.orig
+			pushd $tmp/mnt/lustre
+			ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || {
+				error_noexit "ls"
+				return 1
+			}
+			popd
+			#
+			# 32-bit and 64-bit clients use different algorithms to
+			# convert FIDs into inode numbers.  Hence, remove the inode
+			# numbers from the lists, if the original list was created
+			# on an architecture with different number of bits per
+			# "long".
+			#
+			if [ $(t32_bits_per_long $(uname -m)) !=						\
+				 $(t32_bits_per_long $img_arch) ]; then
+				echo "Different number of bits per \"long\" from the disk image"
+				for list in list.orig list; do
+					sed -i -e 's/^[0-9]\+[ \t]\+//' $tmp/$list
+				done
+			fi
+			if ! diff -ub $tmp/list.orig $tmp/list; then
+				error_noexit "list verification failed"
+				return 1
+			fi
+		else
+			echo "list verification skipped"
+		fi
 
-	echo "LOV changes should succeed:"
-	$LCTL pool_new ${NEWNAME}.interop || return 11
-	$LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12
-	echo "ok."
+		#
+		# When adding new data verification tests, please check for
+		# the presence of the required reference files first, like
+		# the "sha1sums" and "list" tests above, to avoid the need to
+		# regenerate every image for each test addition.
+		#
+
+		nrpcs_orig=$($LCTL get_param \
+				-n mdc.*MDT0000*.max_rpcs_in_flight) || {
+			error_noexit "Getting \"max_rpcs_in_flight\""
+			return 1
+		}
+		nrpcs=$((nrpcs_orig + 5))
+		$r $LCTL conf_param $fsname-MDT0000.mdc.max_rpcs_in_flight=$nrpcs || {
+			error_noexit "Changing \"max_rpcs_in_flight\""
+			return 1
+		}
+		wait_update $HOSTNAME "$LCTL get_param \
+			-n mdc.*MDT0000*.max_rpcs_in_flight" $nrpcs || {
+			error_noexit "Verifying \"max_rpcs_in_flight\""
+			return 1
+		}
+
+		umount $tmp/mnt/lustre || {
+			error_noexit "Unmounting the client"
+			return 1
+		}
+		shall_cleanup_lustre=false
+	else
+		$r umount -d $tmp/mnt/mdt || {
+			error_noexit "Unmounting the MDT"
+			return 1
+		}
+		shall_cleanup_mdt=false
+		$r umount -d $tmp/mnt/ost || {
+			error_noexit "Unmounting the OST"
+			return 1
+		}
+		shall_cleanup_ost=false
+
+		t32_reload_modules $node || {
+			error_noexit "Reloading modules"
+			return 1
+		}
+
+		# mount a second time to make sure we didnt leave upgrade flag on
+		$r $TUNEFS --dryrun $tmp/mdt || {
+			error_noexit "tunefs.lustre before remounting the MDT"
+			return 1
+		}
+		$r mount -t lustre -o loop,exclude=$fsname-OST0000 $tmp/mdt			\
+				 $tmp/mnt/mdt || {
+			error_noexit "Remounting the MDT"
+			return 1
+		}
+		shall_cleanup_mdt=true
+	fi
+}
 
-	# MDT and OST should have registered with new nids, so we should have
-	# a fully-functioning client
-	echo "Check client and old fs contents"
+test_32a() {
+	local tarballs
+	local tarball
+	local rc=0
 
-	local device=`h2$NETTYPE $HOSTNAME`:/$NEWNAME
-	echo "Starting local client: $HOSTNAME: $device $MOUNT"
-	mount -t lustre $device $MOUNT || return 1
+	t32_check
+	for tarball in $tarballs; do
+		t32_test $tarball || rc=$?
+	done
+	return $rc
+}
+run_test 32a "Upgrade (not live)"
 
-	local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight)
-	local new=$((old + 5))
-	$LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new
-	wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11
+test_32b() {
+	local tarballs
+	local tarball
+	local rc=0
 
-	[ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12
-	echo "ok."
+	t32_check
+	for tarball in $tarballs; do
+		t32_test $tarball writeconf || rc=$?
+	done
+	return $rc
+}
+run_test 32b "Upgrade with writeconf"
 
-	cleanup_32
+test_32c() {
+	local tarballs
+	local tarball
+	local rc=0
 
-	rm -rf $tmpdir || true  # true is only for TMP on NFS
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	t32_check
+	for tarball in $tarballs; do
+		dne_upgrade=yes t32_test $tarball writeconf || rc=$?
+	done
+	return $rc
 }
-run_test 32b "Upgrade from 1.8 with writeconf"
+run_test 32c "dne upgrade test"
 
 test_33a() { # bug 12333, was test_33
         local rc=0
@@ -1363,10 +1899,11 @@ test_33a() { # bug 12333, was test_33
 		mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931.
 	fi
 
-	add fs2mds $(mkfs_opts mds1) --fsname=${FSNAME2} --reformat \
-		$mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
-	add fs2ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \
-		--index=8191 --reformat $fs2ostdev $fs2ostvdev || exit 10
+	add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
+		--reformat $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10
+	add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
+		--fsname=${FSNAME2} --index=8191 --reformat $fs2ostdev \
+		$fs2ostvdev || exit 10
 
         start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT
         start fs2ost $fs2ostdev $OST_MOUNT_OPTS
@@ -1391,7 +1928,7 @@ test_33b() {	# was test_34
 
         do_facet client dd if=/dev/zero of=$MOUNT/24 bs=1024k count=1
         # Drop lock cancelation reply during umount
-	#define OBD_FAIL_LDLM_CANCEL             0x304
+	#define OBD_FAIL_LDLM_CANCEL_NET			0x304
         do_facet client lctl set_param fail_loc=0x80000304
         #lctl set_param debug=-1
         umount_client $MOUNT
@@ -1455,7 +1992,8 @@ test_35a() { # bug 12459
 	log "Set up a fake failnode for the MDS"
 	FAKENID="127.0.0.2"
 	local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
-	do_facet mgs $LCTL conf_param ${device}.failover.node=$FAKENID || return 4
+	do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
+		"$(h2$NETTYPE $FAKENID)" || return 4
 
 	log "Wait for RECONNECT_INTERVAL seconds (10s)"
 	sleep 10
@@ -1463,13 +2001,13 @@ test_35a() { # bug 12459
 	MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`"
 	$LCTL clear
 	log "$MSG"
-	log "Stopping the MDT:"
-	stop_mds || return 5
+	log "Stopping the MDT: $device"
+	stop_mdt 1 || return 5
 
 	df $MOUNT > /dev/null 2>&1 &
 	DFPID=$!
-	log "Restarting the MDT:"
-	start_mds || return 6
+	log "Restarting the MDT: $device"
+	start_mdt 1 || return 6
 	log "Wait for df ($DFPID) ... "
 	wait $DFPID
 	log "done"
@@ -1491,7 +2029,7 @@ test_35a() { # bug 12459
 	[ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
 	cleanup
 	# remove nid settings
-	writeconf
+	writeconf_or_reformat
 }
 run_test 35a "Reconnect to the last active server first"
 
@@ -1509,8 +2047,8 @@ test_35b() { # bug 18674
 	FAKENID="127.0.0.2"
 	local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \
 			awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1)
-	do_facet mgs "$LCTL conf_param ${device}.failover.node=$FAKENID" || \
-		return 1
+	do_facet mgs "$LCTL conf_param ${device}.failover.node=" \
+		"$(h2$NETTYPE $FAKENID)" || return 1
 
 	local at_max_saved=0
 	# adaptive timeouts may prevent seeing the issue
@@ -1571,7 +2109,7 @@ test_35b() { # bug 18674
 
 	cleanup
 	# remove nid settings
-	writeconf
+	writeconf_or_reformat
 }
 run_test 35b "Continue reconnection retries, if the active server is busy"
 
@@ -1600,14 +2138,14 @@ test_36() { # 12743
 	local fs2ostvdev=$(ostvdevname 1_2)
 	local fs3ostvdev=$(ostvdevname 2_2)
 
-	add fs2mds $(mkfs_opts mds1) --fsname=${FSNAME2} --reformat \
-		$fs2mdsdev $fs2mdsvdev || exit 10
+	add fs2mds $(mkfs_opts mds1 ${fs2mdsdev}) --mgs --fsname=${FSNAME2} \
+		--reformat $fs2mdsdev $fs2mdsvdev || exit 10
 	# XXX after we support non 4K disk blocksize in ldiskfs, specify a
 	#     different one than the default value here.
-	add fs2ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \
-		--reformat $fs2ostdev $fs2ostvdev || exit 10
-	add fs3ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \
-		--reformat $fs3ostdev $fs3ostvdev || exit 10
+	add fs2ost $(mkfs_opts ost1 ${fs2ostdev}) --mgsnode=$MGSNID \
+		--fsname=${FSNAME2} --reformat $fs2ostdev $fs2ostvdev || exit 10
+	add fs3ost $(mkfs_opts ost1 ${fs3ostdev}) --mgsnode=$MGSNID \
+		--fsname=${FSNAME2} --reformat $fs3ostdev $fs3ostvdev || exit 10
 
         start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS
         start fs2ost $fs2ostdev $OST_MOUNT_OPTS
@@ -1697,6 +2235,11 @@ test_37() {
 run_test 37 "verify set tunables works for symlink device"
 
 test_38() { # bug 14222
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
 	setup
 	# like runtests
 	COUNT=10
@@ -1832,28 +2375,36 @@ test_41b() {
 run_test 41b "mount mds with --nosvc and --nomgs on first mount"
 
 test_42() { #bug 14693
-        setup
-        check_mount || return 2
-        do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10
-        umount_client $MOUNT
-        mount_client $MOUNT || return 1
-        cleanup
-        return 0
+	setup
+	check_mount || error "client was not mounted"
+
+	do_facet mgs $LCTL conf_param $FSNAME.llite.some_wrong_param=10
+	umount_client $MOUNT ||
+		error "unmounting client failed with invalid llite param"
+	mount_client $MOUNT ||
+		error "mounting client failed with invalid llite param"
+
+	do_facet mgs $LCTL conf_param $FSNAME.sys.some_wrong_param=20
+	cleanup || error "stopping $FSNAME failed with invalid sys param"
+	setup
+	check_mount || "client was not mounted with invalid sys param"
+	cleanup || error "stopping $FSNAME failed with invalid sys param"
+	return 0
 }
-run_test 42 "invalid config param should not prevent client from mounting"
+run_test 42 "allow client/server mount/unmount with invalid config param"
 
 test_43() {
-    [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
-    setup
-    chmod ugo+x $DIR || error "chmod 0 failed"
-    set_and_check mds                                        \
-        "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash"  \
-        "$FSNAME.mdt.root_squash"                            \
-        "0:0"
-    set_and_check mds                                        \
-       "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
-       "$FSNAME.mdt.nosquash_nids"                           \
-       "NONE"
+	[ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root"
+	setup
+	chmod ugo+x $DIR || error "chmod 0 failed"
+	set_conf_param_and_check mds					\
+		"lctl get_param -n mdt.$FSNAME-MDT0000.root_squash"	\
+		"$FSNAME.mdt.root_squash"				\
+		"0:0"
+	set_conf_param_and_check mds					\
+		"lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids"	\
+		"$FSNAME.mdt.nosquash_nids"				\
+		"NONE"
 
     #
     # create set of test files
@@ -1869,15 +2420,15 @@ test_43() {
     chmod go-rwx $DIR/$tdir-rootdir   || error "chmod 3 failed"
     touch $DIR/$tdir-rootdir/tfile-1  || error "touch failed"
 
-    #
-    # check root_squash:
-    #   set root squash UID:GID to RUNAS_ID
-    #   root should be able to access only files owned by RUNAS_ID
-    #
-    set_and_check mds                                        \
-       "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash"   \
-       "$FSNAME.mdt.root_squash"                             \
-       "$RUNAS_ID:$RUNAS_ID"
+	#
+	# check root_squash:
+	#   set root squash UID:GID to RUNAS_ID
+	#   root should be able to access only files owned by RUNAS_ID
+	#
+	set_conf_param_and_check mds					\
+		"lctl get_param -n mdt.$FSNAME-MDT0000.root_squash"	\
+		"$FSNAME.mdt.root_squash"				\
+		"$RUNAS_ID:$RUNAS_ID"
 
     ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile)
     dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \
@@ -1908,18 +2459,18 @@ test_43() {
         error "$ST: root create permission is granted"
     echo "$ST: root create permission is denied - ok"
 
-    #
-    # check nosquash_nids:
-    #   put client's NID into nosquash_nids list,
-    #   root should be able to access root file after that
-    #
-    local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
-    NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
-    NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
-    set_and_check mds                                        \
-       "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
-       "$FSNAME-MDTall.mdt.nosquash_nids"                    \
-       "$NIDLIST"
+	#
+	# check nosquash_nids:
+	#   put client's NID into nosquash_nids list,
+	#   root should be able to access root file after that
+	#
+	local NIDLIST=$(lctl list_nids all | tr '\n' ' ')
+	NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
+	NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
+	set_conf_param_and_check mds					\
+		"lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids"	\
+		"$FSNAME-MDTall.mdt.nosquash_nids"			\
+		"$NIDLIST"
 
     ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile)
     dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \
@@ -1997,7 +2548,7 @@ cleanup_46a() {
 	stop_mds || rc=$?
 	cleanup_nocli || rc=$?
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 	return $rc
 }
 
@@ -2239,7 +2790,7 @@ test_50c() {
 	stop_ost2 || error "Unable to stop OST2"
 	stop_mds || error "Unable to stop MDS"
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 50c "lazystatfs one server down =========================="
 
@@ -2261,7 +2812,7 @@ test_50d() {
 	stop_ost2 || error "Unable to stop OST2"
 	stop_mds || error "Unable to stop MDS"
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 50d "lazystatfs client/server conn race =========================="
 
@@ -2343,7 +2894,7 @@ test_50f() {
 	stop_ost || error "Unable to stop OST1"
 	stop_mds || error "Unable to stop MDS"
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 50f "normal statfs one server in down =========================="
 
@@ -2371,10 +2922,46 @@ test_50g() {
 	stop_ost || error "Unable to stop OST1"
 	stop_mds || error "Unable to stop MDS"
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 50g "deactivated OST should not cause panic====================="
 
+# LU-642
+test_50h() {
+	# prepare MDT/OST, make OSC inactive for OST1
+	[ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return
+	do_facet ost1 "$TUNEFS --param osc.active=0 `ostdevname 1`" ||
+		error "tunefs OST1 failed"
+	start_mds  || error "Unable to start MDT"
+	start_ost  || error "Unable to start OST1"
+	start_ost2 || error "Unable to start OST2"
+	mount_client $MOUNT || error "client start failed"
+
+	mkdir -p $DIR/$tdir
+
+	# activatate OSC for OST1
+	local TEST="$LCTL get_param -n osc.${FSNAME}-OST0000-osc-[!M]*.active"
+	set_conf_param_and_check client					\
+		"$TEST" "${FSNAME}-OST0000.osc.active" 1 ||
+		error "Unable to activate OST1"
+
+	mkdir -p $DIR/$tdir/2
+	$LFS setstripe -c -1 -i 0 $DIR/$tdir/2
+	sleep 1 && echo "create a file after OST1 is activated"
+	# create some file
+	createmany -o $DIR/$tdir/2/$tfile-%d 1
+
+	# check OSC import is working
+	stat $DIR/$tdir/2/* >/dev/null 2>&1 ||
+		error "some OSC imports are still not connected"
+
+	# cleanup
+	umount_client $MOUNT || error "Unable to umount client"
+	stop_ost2 || error "Unable to stop OST2"
+	cleanup_nocli
+}
+run_test 50h "LU-642: activate deactivated OST  ==="
+
 test_51() {
 	local LOCAL_TIMEOUT=20
 
@@ -2394,7 +2981,7 @@ test_51() {
 	stop_ost2 || return 3
 	cleanup
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added"
 
@@ -2447,6 +3034,11 @@ diff_files_xattrs()
 }
 
 test_52() {
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
 	start_mds
 	[ $? -eq 0 ] || { error "Unable to start MDS"; return 1; }
 	start_ost
@@ -2503,9 +3095,10 @@ test_52() {
 
 	# backup objects
 	echo backup objects to $ost1tmp/objects
-	local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\
-			'-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
-	copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects
+	local objects=$(do_node $ost1node 'find '$ost1mnt'/O/[0-9]* -type f'\
+		'-size +0 -newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"')
+	copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs \
+			$objects
 	[ $? -eq 0 ] || { error "Unable to copy objects"; return 13; }
 
 	# move objects to lost+found
@@ -2561,7 +3154,6 @@ thread_sanity() {
 	local nthrs
         shift 4
 
-        setup
         check_mount || return 41
 
         # We need to expand $parampat, but it may match multiple parameters, so
@@ -2633,25 +3225,37 @@ thread_sanity() {
         lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $?
         cleanup
 
-        # Workaround a YALA bug where YALA expects that modules will remain
-        # loaded on the servers
-        LOAD_MODULES_REMOTE=false
         load_modules
         setup
-        cleanup
 }
 
 test_53a() {
+	setup
 	thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16'
+	cleanup
 }
 run_test 53a "check OSS thread count params"
 
 test_53b() {
-	thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' '16'
+	setup
+	local mds=$(do_facet $SINGLEMDS "lctl get_param -N mds.*.*.threads_max \
+		    2>/dev/null")
+	if [ -z "$mds" ]; then
+		#running this on an old MDT
+		thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' 16
+	else
+		thread_sanity MDT $SINGLEMDS 'mds.*.*.' 'mds_num_threads' 16
+	fi
+	cleanup
 }
-run_test 53b "check MDT thread count params"
+run_test 53b "check MDS thread count params"
 
 test_54a() {
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
     do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p
     [ $? -eq 0 ] || error "llverdev failed!"
     reformat_and_config
@@ -2659,6 +3263,11 @@ test_54a() {
 run_test 54a "test llverdev and partial verify of device"
 
 test_54b() {
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
     setup
     run_llverfs $MOUNT -p
     [ $? -eq 0 ] || error "llverfs failed!"
@@ -2673,15 +3282,20 @@ lov_objid_size()
 }
 
 test_55() {
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
 	local mdsdev=$(mdsdevname 1)
 	local mdsvdev=$(mdsvdevname 1)
 
 	for i in 1023 2048
 	do
-		add mds1 $(mkfs_opts mds1) --reformat $mdsdev $mdsvdev ||
-			exit 10
-		add ost1 $(mkfs_opts ost1) --index=$i --reformat \
-			$(ostdevname 1) $(ostvdevname 1)
+		add mds1 $(mkfs_opts mds1 ${mdsdev}) --reformat $mdsdev \
+			$mdsvdev || exit 10
+		add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=$i \
+			--reformat $(ostdevname 1) $(ostvdevname 1)
 		setup_noconfig
 		stopall
 		setup_noconfig
@@ -2705,10 +3319,14 @@ test_56() {
 	local mds_journal_size_orig=$MDSJOURNALSIZE
 
 	MDSJOURNALSIZE=16
-	add mds1 $(mkfs_opts mds1) --reformat $(mdsdevname 1) $(mdsvdevname 1)
-	add ost1 $(mkfs_opts ost1) --index=1000 --reformat \
+
+	for num in $(seq 1 $MDSCOUNT); do
+		add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
+			--reformat $(mdsdevname $num) $(mdsvdevname $num)
+	done
+	add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=1000 --reformat \
 		$(ostdevname 1) $(ostvdevname 1)
-	add ost2 $(mkfs_opts ost2) --index=10000 --reformat \
+	add ost2 $(mkfs_opts ost2 $(ostdevname 2)) --index=10000 --reformat \
 		$(ostdevname 2) $(ostvdevname 2)
 
 	start_mgsmds
@@ -2717,7 +3335,6 @@ test_56() {
 	mount_client $MOUNT || error "Unable to mount client"
 	echo ok
 	$LFS osts
-	[ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; }
 	stopall
 	MDSJOURNALSIZE=$mds_journal_size_orig
 	reformat
@@ -2726,7 +3343,7 @@ run_test 56 "check big indexes"
 
 test_57a() { # bug 22656
 	local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
-	writeconf
+	writeconf_or_reformat
 	do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
 	start_mgsmds
 	start_ost && error "OST registration from failnode should fail"
@@ -2736,7 +3353,7 @@ run_test 57a "initial registration from failnode should fail (should return errs
 
 test_57b() {
 	local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}')
-	writeconf
+	writeconf_or_reformat
 	do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed"
 	start_mgsmds
 	start_ost || error "OST registration from servicenode should not fail"
@@ -2749,8 +3366,8 @@ count_osts() {
 }
 
 test_58() { # bug 22658
-	if [ $(facet_fstype mds) == zfs ]; then
-		skip "Does not work with ZFS-based MDTs yet"
+	if [ $(facet_fstype mds) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
 		return
 	fi
 	setup_noconfig
@@ -2801,19 +3418,24 @@ test_59() {
 	stop_ost2 >> /dev/null
 	cleanup_nocli >> /dev/null
 	#writeconf to remove all ost2 traces for subsequent tests
-	writeconf
+	writeconf_or_reformat
 }
 run_test 59 "writeconf mount option"
 
 test_60() { # LU-471
+	local num
+
 	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
 		skip "Only applicable to ldiskfs-based MDTs"
 		return
 	fi
 
-	add mds1 $(mkfs_opts mds1) \
-		--mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' --reformat \
-		$(mdsdevname 1) $(mdsvdevname 1) || exit 10
+	for num in $(seq $MDSCOUNT); do
+		add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
+			--mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' \
+			--reformat $(mdsdevname $num) $(mdsvdevname $num) ||
+			exit 10
+	done
 
 	dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1))
 	rc=${PIPESTATUS[0]}
@@ -2838,10 +3460,13 @@ test_61() { # LU-80
 	if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
 	   ! large_xattr_enabled; then
 		reformat=true
-		local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
 		LDISKFS_MKFS_OPTS+=" -O large_xattr"
-		add $SINGLEMDS $(mkfs_opts $SINGLEMDS) --reformat $mds_dev ||
-			error "reformatting $mds_dev failed"
+
+		for num in $(seq $MDSCOUNT); do
+			add mds${num} $(mkfs_opts mds$num $(mdsdevname $num)) \
+			--reformat $(mdsdevname $num) $(mdsvdevname $num) ||
+			error "add mds $num failed"
+		done
 	fi
 
     setup_noconfig || error "setting up the filesystem failed"
@@ -2897,6 +3522,11 @@ test_61() { # LU-80
 run_test 61 "large xattr"
 
 test_62() {
+	if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then
+		skip "Only applicable to ldiskfs-based MDTs"
+		return
+	fi
+
 	# MRP-118
 	local mdsdev=$(mdsdevname 1)
 	local ostdev=$(ostdevname 1)
@@ -2935,11 +3565,418 @@ test_63() {
 }
 run_test 63 "Verify each page can at least hold 3 ldisk inodes"
 
+test_64() {
+	start_mds
+	start_ost
+	start_ost2 || error "Unable to start second ost"
+	mount_client $MOUNT || error "Unable to mount client"
+	stop_ost2 || error "Unable to stop second ost"
+	echo "$LFS df"
+	$LFS df --lazy || error "lfs df failed"
+	cleanup || return $?
+	#writeconf to remove all ost2 traces for subsequent tests
+	writeconf_or_reformat
+}
+run_test 64 "check lfs df --lazy "
+
+test_65() { # LU-2237
+	# Currently, the test is only valid for ldiskfs backend
+	[ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
+		skip "non-ldiskfs backend" && return
+
+	local devname=$(mdsdevname ${SINGLEMDS//mds/})
+	local brpt=$(facet_mntpt brpt)
+	local opts=""
+
+	if ! do_facet $SINGLEMDS "test -b $devname"; then
+		opts="-o loop"
+	fi
+
+	stop_mds
+	local obj=$(do_facet $SINGLEMDS \
+		    "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" |
+		    grep Inode)
+	if [ -z "$obj" ]; then
+		# The MDT may be just re-formatted, mount the MDT for the
+		# first time to guarantee the "last_rcvd" file is there.
+		start_mds || error "fail to mount the MDS for the first time"
+		stop_mds
+	fi
+
+	# remove the "last_rcvd" file
+	do_facet $SINGLEMDS "mkdir -p $brpt"
+	do_facet $SINGLEMDS \
+		"mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
+	do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
+	do_facet $SINGLEMDS "umount $brpt"
+
+	# restart MDS, the "last_rcvd" file should be recreated.
+	start_mds || error "fail to restart the MDS"
+	stop_mds
+	obj=$(do_facet $SINGLEMDS \
+	      "$DEBUGFS -c -R \\\"stat last_rcvd\\\" $devname" | grep Inode)
+	[ -n "$obj" ] || error "fail to re-create the last_rcvd"
+}
+run_test 65 "re-create the lost last_rcvd file when server mount"
+
+test_66() {
+	[[ $(lustre_version_code mgs) -ge $(version_code 2.3.59) ]] ||
+		{ skip "Need MGS version at least 2.3.59"; return 0; }
+
+	setup
+	local OST1_NID=$(do_facet ost1 $LCTL list_nids | head -1)
+	local MDS_NID=$(do_facet $SINGLEMDS $LCTL list_nids | head -1)
+
+	echo "replace_nids should fail if MDS, OSTs and clients are UP"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
+		error "replace_nids fail"
+
+	umount_client $MOUNT || error "unmounting client failed"
+	echo "replace_nids should fail if MDS and OSTs are UP"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
+		error "replace_nids fail"
+
+	stop_ost
+	echo "replace_nids should fail if MDS is UP"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID &&
+		error "replace_nids fail"
+
+	stop_mds || error "stopping mds failed"
+
+	if combined_mgs_mds; then
+		start_mds "-o nosvc" ||
+			error "starting mds with nosvc option failed"
+	fi
+
+	echo "command should accept two parameters"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 &&
+		error "command should accept two params"
+
+	echo "correct device name should be passed"
+	do_facet mgs $LCTL replace_nids $FSNAME-WRONG0000 $OST1_NID &&
+		error "wrong devname"
+
+	echo "wrong nids list should not destroy the system"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 "wrong nids list" &&
+		error "wrong parse"
+
+	echo "replace OST nid"
+	do_facet mgs $LCTL replace_nids $FSNAME-OST0000 $OST1_NID ||
+		error "replace nids failed"
+
+	echo "command should accept two parameters"
+	do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 &&
+		error "command should accept two params"
+
+	echo "wrong nids list should not destroy the system"
+	do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 "wrong nids list" &&
+		error "wrong parse"
+
+	echo "replace MDS nid"
+	do_facet mgs $LCTL replace_nids $FSNAME-MDT0000 $MDS_NID ||
+		error "replace nids failed"
+
+	if ! combined_mgs_mds ; then
+		stop_mgs
+	else
+		stop_mds
+	fi
+
+	setup_noconfig
+	check_mount || error "error after nid replace"
+	cleanup || error "cleanup failed"
+	reformat
+}
+run_test 66 "replace nids"
+
+test_70a() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	local MDTIDX=1
+
+	start_mdt 1 || error "MDT0 start fail"
+
+	start_ost || error "OST0 start fail"
+
+	start_mdt 2 || error "MDT1 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "create dir fail"
+
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+					error "create remote dir fail"
+
+	rm -rf $DIR/$tdir || error "delete dir fail"
+	cleanup || return $?
+}
+run_test 70a "start MDT0, then OST, then MDT1"
+
+test_70b() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	local MDTIDX=1
+
+	start_ost || error "OST0 start fail"
+
+	start_mdt 1 || error "MDT0 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "create dir fail"
+
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+					error "create remote dir fail"
+
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	cleanup || return $?
+}
+run_test 70b "start OST, MDT1, MDT0"
+
+test_70c() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	local MDTIDX=1
+
+	start_mdt 1 || error "MDT0 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+	start_ost || error "OST0 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+	stop_mdt 1 || error "MDT1 start fail"
+
+	local mdc_for_mdt1=$($LCTL dl | grep MDT0000-mdc | awk '{print $4}')
+	echo "deactivate $mdc_for_mdt1"
+        $LCTL --device $mdc_for_mdt1 deactivate || return 1
+
+	mkdir -p $DIR/$tdir && error "mkdir succeed"
+
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
+					error "create remote dir succeed"
+
+	cleanup || return $?
+}
+run_test 70c "stop MDT0, mkdir fail, create remote dir fail"
+
+test_70d() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	local MDTIDX=1
+
+	start_mdt 1 || error "MDT0 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+	start_ost || error "OST0 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	stop_mdt 2 || error "MDT1 start fail"
+
+	local mdc_for_mdt2=$($LCTL dl | grep MDT0001-mdc |
+			     awk '{print $4}')
+	echo "deactivate $mdc_for_mdt2"
+        $LCTL --device $mdc_for_mdt2 deactivate ||
+			error "set $mdc_for_mdt2 deactivate failed"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir &&
+			error "create remote dir succeed"
+
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	cleanup || return $?
+}
+run_test 70d "stop MDT1, mkdir succeed, create remote dir fail"
+
+test_71a() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	if combined_mgs_mds; then
+		skip "needs separate MGS/MDT" && return
+	fi
+	local MDTIDX=1
+
+	start_mdt 1 || error "MDT0 start fail"
+	start_ost || error "OST0 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+	start_ost2 || error "OST1 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+			error "create remote dir succeed"
+
+	mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	umount_client $MOUNT
+	stop_mdt 1 || error "MDT0 stop fail"
+	stop_mdt 2 || error "MDT1 stop fail"
+	stop_ost || error "OST0 stop fail"
+	stop_ost2 || error "OST1 stop fail"
+}
+run_test 71a "start MDT0 OST0, MDT1, OST1"
+
+test_71b() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	if combined_mgs_mds; then
+		skip "needs separate MGS/MDT" && return
+	fi
+	local MDTIDX=1
+
+	start_mdt 2 || error "MDT1 start fail"
+	start_ost || error "OST0 start fail"
+	start_mdt 1 || error "MDT0 start fail"
+	start_ost2 || error "OST1 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+			error "create remote dir succeed"
+
+	mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	umount_client $MOUNT
+	stop_mdt 1 || error "MDT0 stop fail"
+	stop_mdt 2 || error "MDT1 stop fail"
+	stop_ost || error "OST0 stop fail"
+	stop_ost2 || error "OST1 stop fail"
+}
+run_test 71b "start MDT1, OST0, MDT0, OST1"
+
+test_71c() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	if combined_mgs_mds; then
+		skip "needs separate MGS/MDT" && return
+	fi
+	local MDTIDX=1
+
+	start_ost || error "OST0 start fail"
+	start_ost2 || error "OST1 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+	start_mdt 1 || error "MDT0 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+			error "create remote dir succeed"
+
+	mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	umount_client $MOUNT
+	stop_mdt 1 || error "MDT0 stop fail"
+	stop_mdt 2 || error "MDT1 stop fail"
+	stop_ost || error "OST0 stop fail"
+	stop_ost2 || error "OST1 stop fail"
+
+}
+run_test 71c "start OST0, OST1, MDT1, MDT0"
+
+test_71d() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	if combined_mgs_mds; then
+		skip "needs separate MGS/MDT" && return
+	fi
+	local MDTIDX=1
+
+	start_ost || error "OST0 start fail"
+	start_mdt 2 || error "MDT0 start fail"
+	start_mdt 1 || error "MDT0 start fail"
+	start_ost2 || error "OST1 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+			error "create remote dir succeed"
+
+	mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	umount_client $MOUNT
+	stop_mdt 1 || error "MDT0 stop fail"
+	stop_mdt 2 || error "MDT1 stop fail"
+	stop_ost || error "OST0 stop fail"
+	stop_ost2 || error "OST1 stop fail"
+
+}
+run_test 71d "start OST0, MDT1, MDT0, OST1"
+
+test_71e() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+	if combined_mgs_mds; then
+		skip "needs separate MGS/MDT" && return
+	fi
+	local MDTIDX=1
+
+	start_ost || error "OST0 start fail"
+	start_mdt 2 || error "MDT1 start fail"
+	start_ost2 || error "OST1 start fail"
+	start_mdt 1 || error "MDT0 start fail"
+
+	mount_client $MOUNT || error "mount client fails"
+
+	mkdir -p $DIR/$tdir || error "mkdir fail"
+	$LFS mkdir -i $MDTIDX $DIR/$tdir/remote_dir ||
+			error "create remote dir succeed"
+
+	mcreate $DIR/$tdir/remote_dir/$tfile || error "create file failed"
+	rm -rf $DIR/$tdir || error "delete dir fail"
+
+	umount_client $MOUNT
+	stop_mdt 1 || error "MDT0 stop fail"
+	stop_mdt 2 || error "MDT1 stop fail"
+	stop_ost || error "OST0 stop fail"
+	stop_ost2 || error "OST1 stop fail"
+
+}
+run_test 71e "start OST0, MDT1, OST1, MDT0"
+
+test_72() { #LU-2634
+	local mdsdev=$(mdsdevname 1)
+	local ostdev=$(ostdevname 1)
+	local cmd="$E2FSCK -fnvd $mdsdev"
+	local fn=3
+
+	[ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
+		skip "ldiskfs only test" && return
+
+	#tune MDT with "-O extents"
+	add $SINGLEMDS \
+		$(mkfs_opts $SINGLEMDS ${mdsdev}) --reformat $mdsdev ||
+			error "add $SINGLEMDS failed"
+	$TUNE2FS -O extents $mdsdev
+	add ost1 $(mkfs_opts ost1 $ostdev) --reformat $ostdev ||
+		error "add $ostdev failed"
+	start_mgsmds || error "start mds failed"
+	start_ost || error "start ost failed"
+	mount_client $MOUNT || error "mount client failed"
+
+	#create some short symlinks
+	mkdir -p $DIR/$tdir
+	createmany -o $DIR/$tdir/$tfile-%d $fn
+	echo "create $fn short symlinks"
+	for i in $(seq -w 1 $fn); do
+		ln -s $DIR/$tdir/$tfile-$i $MOUNT/$tfile-$i
+	done
+	ls -al $MOUNT
+
+	#umount
+	umount_client $MOUNT || error "umount client failed"
+	stop_mds || error "stop mds failed"
+	stop_ost || error "stop ost failed"
+
+	#run e2fsck
+	run_e2fsck $(facet_active_host $SINGLEMDS) $mdsdev "-n"
+}
+run_test 72 "test fast symlink with extents flag enabled"
+
 if ! combined_mgs_mds ; then
 	stop mgs
 fi
 
 cleanup_gss
 
-complete $(basename $0) $SECONDS
+complete $SECONDS
 exit_status