X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=31e7ec35a4fe5841c7f1061af86670acee3cd2ea;hp=7b0fe56953847de3aa8b2afd7238ff164de78235;hb=5a911faae25784a91fc085debeb6dbe8512d80b6;hpb=fde0e290a8cc370f6eb7986e9ada8b5bcc41fef7

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 7b0fe56..31e7ec3 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -22,6 +22,7 @@ export SK_UNIQUE_NM=${SK_UNIQUE_NM:-false}
 export SK_S2S=${SK_S2S:-false}
 export SK_S2SNM=${SK_S2SNM:-TestFrameNM}
 export SK_S2SNMCLI=${SK_S2SNMCLI:-TestFrameNMCli}
+export SK_SKIPFIRST=${SK_SKIPFIRST:-true}
 export IDENTITY_UPCALL=default
 export QUOTA_AUTO=1
 export FLAKEY=${FLAKEY:-true}
@@ -94,13 +95,14 @@ usage() {
 }
 
 print_summary () {
-    trap 0
-	[ -z "$DEFAULT_SUITES"] && return 0
-    [ -n "$ONLY" ] && echo "WARNING: ONLY is set to $(echo $ONLY)"
-    local details
-    local form="%-13s %-17s %-9s %s %s\n"
-    printf "$form" "status" "script" "Total(sec)" "E(xcluded) S(low)"
-    echo "------------------------------------------------------------------------------------"
+	trap 0
+	[ -z "$DEFAULT_SUITES" ] && return 0
+	[ -n "$ONLY" ] && echo "WARNING: ONLY is set to $(echo $ONLY)"
+	local details
+	local form="%-13s %-17s %-9s %s %s\n"
+
+	printf "$form" "status" "script" "Total(sec)" "E(xcluded) S(low)"
+	echo "---------------------------------------------------------------"
     for O in $DEFAULT_SUITES; do
         O=$(echo $O  | tr "-" "_" | tr "[:lower:]" "[:upper:]")
         [ "${!O}" = "no" ] && continue || true
@@ -147,6 +149,27 @@ print_summary () {
     fi
 }
 
+# Get information about the Lustre environment. The information collected
+# will be used in Lustre tests.
+# usage: get_lustre_env
+# input: No required or optional arguments
+# output: No return values, environment variables are exported
+
+get_lustre_env() {
+
+	export mds1_FSTYPE=${mds1_FSTYPE:-$(facet_fstype mds1)}
+	export ost1_FSTYPE=${ost1_FSTYPE:-$(facet_fstype ost1)}
+
+	export MGS_VERSION=$(lustre_version_code mgs)
+	export MDS1_VERSION=$(lustre_version_code mds1)
+	export OST1_VERSION=$(lustre_version_code ost1)
+	export CLIENT_VERSION=$(lustre_version_code client)
+
+	# Prefer using "mds1" directly instead of SINGLEMDS.
+	# Keep this for compat until it is removed from scripts.
+	export SINGLEMDS=${SINGLEMDS:-mds1}
+}
+
 init_test_env() {
 	export LUSTRE=$(absolute_path $LUSTRE)
 	export TESTSUITE=$(basename $0 .sh)
@@ -156,6 +179,7 @@ init_test_env() {
 	export DO_CLEANUP=${DO_CLEANUP:-true}
 	export KEEP_ZPOOL=${KEEP_ZPOOL:-false}
 	export CLEANUP_DM_DEV=false
+	export PAGE_SIZE=$(get_page_size client)
 
 	export MKE2FS=$MKE2FS
 	if [ -z "$MKE2FS" ]; then
@@ -399,6 +423,14 @@ init_test_env() {
 	fi
 
 	export TF_FAIL=${TF_FAIL:-$TMP/tf.fail}
+
+	# Constants used in more than one test script
+	export LOV_MAX_STRIPE_COUNT=2000
+
+	export MACHINEFILE=${MACHINEFILE:-$TMP/$(basename $0 .sh).machines}
+	. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+	get_lustre_env
+
 }
 
 check_cpt_number() {
@@ -608,6 +640,24 @@ load_modules_local() {
 		return 0
 	fi
 
+	# Create special udev test rules on every node
+	if [ -f $LUSTRE/lustre/conf/99-lustre.rules ]; then {
+		sed -e 's|/usr/sbin/lctl|$LCTL|g' $LUSTRE/lustre/conf/99-lustre.rules > /etc/udev/rules.d/99-lustre-test.rules
+	} else {
+		echo "SUBSYSTEM==\"lustre\", ACTION==\"change\", ENV{PARAM}==\"?*\", RUN+=\"$LCTL set_param '\$env{PARAM}=\$env{SETTING}'\"" > /etc/udev/rules.d/99-lustre-test.rules
+	} fi
+	udevadm control --reload-rules
+	udevadm trigger
+
+	# For kmemleak-enabled kernels we need clear all past state
+	# that obviously has nothing to do with this Lustre run
+	# Disable automatic memory scanning to avoid perf hit.
+	if [ -f /sys/kernel/debug/kmemleak ] ; then
+		echo scan=off > /sys/kernel/debug/kmemleak
+		echo scan > /sys/kernel/debug/kmemleak
+		echo clear > /sys/kernel/debug/kmemleak
+	fi
+
 	echo Loading modules from $LUSTRE
 
 	local ncpus
@@ -753,12 +803,20 @@ unload_modules() {
 
 	$LUSTRE_RMMOD ldiskfs || return 2
 
+	[ -f /etc/udev/rules.d/99-lustre-test.rules ] &&
+		rm /etc/udev/rules.d/99-lustre-test.rules
+	udevadm control --reload-rules
+	udevadm trigger
+
 	if $LOAD_MODULES_REMOTE; then
 		local list=$(comma_list $(remote_nodes_list))
 		if [ -n "$list" ]; then
 			echo "unloading modules on: '$list'"
 			do_rpc_nodes "$list" $LUSTRE_RMMOD ldiskfs
 			do_rpc_nodes "$list" check_mem_leak
+			do_rpc_nodes "$list" "rm -f /etc/udev/rules.d/99-lustre-test.rules"
+			do_rpc_nodes "$list" "udevadm control --reload-rules"
+			do_rpc_nodes "$list" "udevadm trigger"
 		fi
 	fi
 
@@ -926,6 +984,16 @@ init_gss() {
 		return
 	fi
 
+	case $LUSTRE in
+	/usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
+	/usr/lib64/lustre )
+		from_build_tree=false
+		;;
+	*)
+		from_build_tree=true
+		;;
+	esac
+
 	if ! module_loaded ptlrpc_gss; then
 		load_module ptlrpc/gss/ptlrpc_gss
 		module_loaded ptlrpc_gss ||
@@ -942,26 +1010,41 @@ init_gss() {
 
 		# security ctx config for keyring
 		SK_NO_KEY=false
-		mkdir -p $SK_OM_PATH
-		mount -o bind $SK_OM_PATH /etc/request-key.d/
-		local lgssc_conf_line='create lgssc * * '
-		lgssc_conf_line+=$(which lgss_keyring)
-		lgssc_conf_line+=' %o %k %t %d %c %u %g %T %P %S'
-
 		local lgssc_conf_file="/etc/request-key.d/lgssc.conf"
-		echo "$lgssc_conf_line" > $lgssc_conf_file
+
+		if $from_build_tree; then
+			mkdir -p $SK_OM_PATH
+			if grep -q request-key /proc/mounts > /dev/null; then
+				echo "SSK: Request key already mounted."
+			else
+				mount -o bind $SK_OM_PATH /etc/request-key.d/
+			fi
+			local lgssc_conf_line='create lgssc * * '
+			lgssc_conf_line+=$(which lgss_keyring)
+			lgssc_conf_line+=' %o %k %t %d %c %u %g %T %P %S'
+			echo "$lgssc_conf_line" > $lgssc_conf_file
+		fi
+
 		[ -e $lgssc_conf_file ] ||
 			error_exit "Could not find key options in $lgssc_conf_file"
+		echo "$lgssc_conf_file content is:"
+		cat $lgssc_conf_file
 
 		if ! local_mode; then
-			do_nodes $(comma_list $(all_nodes)) "mkdir -p \
-				$SK_OM_PATH"
-			do_nodes $(comma_list $(all_nodes)) "mount \
-				-o bind $SK_OM_PATH \
-				/etc/request-key.d/"
-			do_nodes $(comma_list $(all_nodes)) "rsync -aqv \
-				$HOSTNAME:$lgssc_conf_file \
-				$lgssc_conf_file >/dev/null 2>&1"
+			if $from_build_tree; then
+				do_nodes $(comma_list $(all_nodes)) "mkdir -p \
+					$SK_OM_PATH"
+				do_nodes $(comma_list $(all_nodes)) "mount \
+					-o bind $SK_OM_PATH \
+					/etc/request-key.d/"
+				do_nodes $(comma_list $(all_nodes)) "rsync \
+					-aqv $HOSTNAME:$lgssc_conf_file \
+					$lgssc_conf_file >/dev/null 2>&1"
+			else
+				do_nodes $(comma_list $(all_nodes)) \
+					"echo $lgssc_conf_file: ; \
+					cat $lgssc_conf_file"
+			fi
 		fi
 
 		# create shared key on all nodes
@@ -990,8 +1073,9 @@ init_gss() {
 		done
 		# Distribute keys
 		if ! local_mode; then
-			do_nodes $(comma_list $(all_nodes)) "rsync -av \
-				$HOSTNAME:$SK_PATH/ $SK_PATH >/dev/null 2>&1"
+			for lnode in $(all_nodes); do
+				scp -r $SK_PATH ${lnode}:$(dirname $SK_PATH)/
+			done
 		fi
 		# Set client keys to client type to generate prime P
 		if local_mode; then
@@ -1044,6 +1128,16 @@ cleanup_gss() {
 
 cleanup_sk() {
 	if $GSS_SK; then
+		case $LUSTRE in
+		/usr/lib/lustre/* | /usr/lib64/lustre/* | /usr/lib/lustre | \
+		/usr/lib64/lustre )
+			from_build_tree=false
+			;;
+		*)
+			from_build_tree=true
+			;;
+		esac
+
 		if $SK_S2S; then
 			do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNM"
 			do_node $(mgs_node) "$LCTL nodemap_del $SK_S2SNMCLI"
@@ -1054,12 +1148,18 @@ cleanup_sk() {
 		$RPC_MODE || echo "Cleaning up Shared Key.."
 		do_nodes $(comma_list $(all_nodes)) "rm -f \
 			$SK_PATH/$FSNAME*.key $SK_PATH/nodemap/$FSNAME*.key"
-		# Remove the mount and clean up the files we added to SK_PATH
-		do_nodes $(comma_list $(all_nodes)) "umount \
-			/etc/request-key.d/"
-		do_nodes $(comma_list $(all_nodes)) "rm -f \
-			$SK_OM_PATH/lgssc.conf"
-		do_nodes $(comma_list $(all_nodes)) "rmdir $SK_OM_PATH"
+		do_nodes $(comma_list $(all_nodes)) "keyctl show | \
+		  awk '/lustre/ { print \\\$1 }' | xargs -IX keyctl unlink X"
+		if $from_build_tree; then
+			# Remove the mount and clean up the files we added to
+			# SK_PATH
+			do_nodes $(comma_list $(all_nodes)) "while grep -q \
+				request-key.d /proc/mounts; do umount \
+				/etc/request-key.d/; done"
+			do_nodes $(comma_list $(all_nodes)) "rm -f \
+				$SK_OM_PATH/lgssc.conf"
+			do_nodes $(comma_list $(all_nodes)) "rmdir $SK_OM_PATH"
+		fi
 		SK_NO_KEY=true
 	fi
 }
@@ -1236,7 +1336,7 @@ running_in_vm() {
 	virt=$(dmidecode -s system-product-name | awk '{print $1}')
 
 	case $virt in
-		VMware|KVM|VirtualBox|Parallels)
+		VMware|KVM|VirtualBox|Parallels|Bochs)
 			echo $virt | tr '[A-Z]' '[a-z]' ;;
 		*) ;;
 	esac
@@ -1920,11 +2020,6 @@ mount_facet() {
 
 	set_default_debug_facet $facet
 
-	if [[ $facet == mds* ]]; then
-		do_facet $facet \
-		lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 2>/dev/null
-	fi
-
 	if [[ $opts =~ .*nosvc.* ]]; then
 		echo "Start $dm_dev without service"
 	else
@@ -1986,12 +2081,6 @@ start() {
 	mount_facet ${facet}
 	RC=$?
 
-	if [[ $facet == mds* ]]; then
-		do_facet $facet \
-			lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
-				2>/dev/null
-	fi
-
 	return $RC
 }
 
@@ -2044,7 +2133,7 @@ ost_quota_type() {
 # restore old quota type settings
 restore_quota() {
 	if [ "$old_MDT_QUOTA_TYPE" ]; then
-		if [[ $PERM_CMD = *"set_param -P"* ]]; then
+		if [[ $PERM_CMD == *"set_param -P"* ]]; then
 			do_facet mgs $PERM_CMD \
 				osd-*.$FSNAME-MDT*.quota_slave.enable = \
 				$old_MDT_QUOTA_TYPE
@@ -2054,7 +2143,7 @@ restore_quota() {
 		fi
 	fi
 	if [ "$old_OST_QUOTA_TYPE" ]; then
-		if [[ $PERM_CMD = *"set_param -P"* ]]; then
+		if [[ $PERM_CMD == *"set_param -P"* ]]; then
 			do_facet mgs $PERM_CMD \
 				osd-*.$FSNAME-OST*.quota_slave.enable = \
 				$old_OST_QUOTA_TYPE
@@ -2114,7 +2203,7 @@ setup_quota(){
 	export old_MDT_QUOTA_TYPE=$mdt_qtype
 	export old_OST_QUOTA_TYPE=$ost_qtype
 
-	if [[ $PERM_CMD = *"set_param -P"* ]]; then
+	if [[ $PERM_CMD == *"set_param -P"* ]]; then
 		do_facet mgs $PERM_CMD \
 			osd-*.$FSNAME-MDT*.quota_slave.enable=$QUOTA_TYPE
 		do_facet mgs $PERM_CMD \
@@ -2204,43 +2293,31 @@ zconf_mount() {
 }
 
 zconf_umount() {
-    local client=$1
-    local mnt=$2
-    local force
-    local busy
-    local need_kill
-
-    [ "$3" ] && force=-f
-    local running=$(do_node $client "grep -c $mnt' ' /proc/mounts") || true
-    if [ $running -ne 0 ]; then
-        echo "Stopping client $client $mnt (opts:$force)"
-        do_node $client lsof -t $mnt || need_kill=no
-        if [ "x$force" != "x" -a "x$need_kill" != "xno" ]; then
-            pids=$(do_node $client lsof -t $mnt | sort -u);
-            if [ -n $pids ]; then
-                do_node $client kill -9 $pids || true
-            fi
-        fi
-
-        busy=$(do_node $client "umount $force $mnt 2>&1" | grep -c "busy") || true
-        if [ $busy -ne 0 ] ; then
-            echo "$mnt is still busy, wait one second" && sleep 1
-            do_node $client umount $force $mnt
-        fi
-    fi
-}
-
-# Mount the file system on the MGS
-mount_mgs_client() {
-	do_facet mgs "mkdir -p $MOUNT"
-	zconf_mount $mgs_HOST $MOUNT $MOUNT_OPTS ||
-		error "unable to mount $MOUNT on MGS"
-}
+	local client=$1
+	local mnt=$2
+	local force
+	local busy
+	local need_kill
+	local running=$(do_node $client "grep -c $mnt' ' /proc/mounts") || true
+
+	[ "$3" ] && force=-f
+	[ $running -eq 0 ] && return 0
+
+	echo "Stopping client $client $mnt (opts:$force)"
+	do_node $client lsof -t $mnt || need_kill=no
+	if [ "x$force" != "x" ] && [ "x$need_kill" != "xno" ]; then
+		pids=$(do_node $client lsof -t $mnt | sort -u);
+		if [ -n "$pids" ]; then
+			do_node $client kill -9 $pids || true
+		fi
+	fi
 
-# Unmount the file system on the MGS
-umount_mgs_client() {
-	zconf_umount $mgs_HOST $MOUNT
-	do_facet mgs "rm -rf $MOUNT"
+	busy=$(do_node $client "umount $force $mnt 2>&1" | grep -c "busy") ||
+		true
+	if [ $busy -ne 0 ] ; then
+		echo "$mnt is still busy, wait one second" && sleep 1
+		do_node $client umount $force $mnt
+	fi
 }
 
 # nodes is comma list
@@ -2316,7 +2393,8 @@ zconf_mount_clients() {
 	fi
 
 	echo "Starting client $clients: $flags $opts $device $mnt"
-	if [ -n "$FILESET" -a ! -n "$SKIP_FILESET" ]; then
+	do_nodes $clients mkdir -p $mnt
+	if [ -n "$FILESET" -a -z "$SKIP_FILESET" ]; then
 		if $GSS_SK && ($SK_UNIQUE_NM || $SK_S2S); then
 			# Mount with own nodemap key
 			local i=0
@@ -2623,6 +2701,18 @@ start_client_load() {
 			LFS=$LFS \
 			LCTL=$LCTL \
 			FSNAME=$FSNAME \
+			MPIRUN=$MPIRUN \
+			MPIRUN_OPTIONS=\\\"$MPIRUN_OPTIONS\\\" \
+			MACHINEFILE_OPTION=\\\"$MACHINEFILE_OPTION\\\" \
+			num_clients=$(get_node_count ${CLIENTS//,/ }) \
+			ior_THREADS=$ior_THREADS ior_iteration=$ior_iteration \
+			ior_blockSize=$ior_blockSize \
+			ior_blockUnit=$ior_blockUnit \
+			ior_xferSize=$ior_xferSize ior_type=$ior_type \
+			ior_DURATION=$ior_DURATION \
+			ior_stripe_params=\\\"$ior_stripe_params\\\" \
+			ior_custom_params=\\\"$ior_custom_param\\\" \
+			mpi_ior_custom_threads=$mpi_ior_custom_threads \
 			run_${load}.sh" &
 	local ppid=$!
 	log "Started client load: ${load} on $client"
@@ -2907,6 +2997,142 @@ wait_zfs_commit() {
 	fi
 }
 
+fill_ost() {
+	local filename=$1
+	local ost_idx=$2
+	local lwm=$3  #low watermark
+	local size_mb #how many MB should we write to pass watermark
+	local ost_name=$(ostname_from_index $ost_idx)
+
+	free_kb=$($LFS df $MOUNT | awk "/$ost_name/ { print \$4 }")
+	size_mb=0
+	if (( $free_kb / 1024 > lwm )); then
+		size_mb=$((free_kb / 1024 - lwm))
+	fi
+	#If 10% of free space cross low watermark use it
+	if (( $free_kb / 10240 > size_mb )); then
+		size_mb=$((free_kb / 10240))
+	else
+		#At least we need to store 1.1 of difference between
+		#free space and low watermark
+		size_mb=$((size_mb + size_mb / 10))
+	fi
+	if (( lwm <= $free_kb / 1024 )) ||
+	   [ ! -f $DIR/${filename}.fill_ost$ost_idx ]; then
+		$LFS setstripe -i $ost_idx -c1 $DIR/${filename}.fill_ost$ost_idx
+		dd if=/dev/zero of=$DIR/${filename}.fill_ost$ost_idx bs=1M \
+			count=$size_mb oflag=append conv=notrunc
+	fi
+
+	sleep_maxage
+
+	free_kb=$($LFS df $MOUNT | awk "/$ost_name/ { print \$4 }")
+	echo "OST still has $((free_kb / 1024)) MB free"
+}
+
+# This checks only the primary MDS
+ost_watermarks_get() {
+	local ost_idx=$1
+	local ost_name=$(ostname_from_index $ost_idx)
+	local mdtosc_proc=$(get_mdtosc_proc_path $SINGLEMDS $ost_name)
+
+	local hwm=$(do_facet $SINGLEMDS $LCTL get_param -n \
+			osp.$mdtosc_proc.reserved_mb_high)
+	local lwm=$(do_facet $SINGLEMDS $LCTL get_param -n \
+			osp.$mdtosc_proc.reserved_mb_low)
+
+	echo "$lwm $hwm"
+}
+
+# Note that we set watermarks on all MDSes (necessary for striped dirs)
+ost_watermarks_set() {
+	local ost_idx=$1
+	local lwm=$2
+	local hwm=$3
+	local ost_name=$(ostname_from_index $ost_idx)
+	local facets=$(get_facets MDS)
+
+	do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param -n \
+		osp.*$ost_name*.reserved_mb_low=$lwm \
+		osp.*$ost_name*.reserved_mb_high=$hwm > /dev/null
+
+	# sleep to ensure we see the change
+	sleep_maxage
+}
+
+ost_watermarks_set_low_space() {
+	local ost_idx=$1
+	local wms=$(ost_watermarks_get $ost_idx)
+	local ost_name=$(ostname_from_index $ost_idx)
+
+	local old_lwm=$(echo $wms | awk '{ print $1 }')
+	local old_hwm=$(echo $wms | awk '{ print $2 }')
+
+	local blocks=$($LFS df $MOUNT | awk "/$ost_name/ { print \$4 }")
+	# minimal extension size is 64M
+	local new_lwm=50
+	if (( $blocks / 1024 > 50 )); then
+		new_lwm=$((blocks / 1024 - 50))
+	fi
+	local new_hwm=$((new_lwm + 5))
+
+	ost_watermarks_set $ost_idx $new_lwm $new_hwm
+	echo "watermarks: $old_lwm $old_hwm $new_lwm $new_hwm"
+}
+
+# Set watermarks to ~current available space & then write data to fill it
+# Note OST is not *actually* full after this, it just reports ENOSPC in the
+# internal statfs used by the stripe allocator
+#
+# first parameter is the filename-prefix, which must get under t-f cleanup
+# requirements (rm -rf $DIR/[Rdfs][0-9]*), i.e. $tfile work fine
+ost_watermarks_set_enospc() {
+	local filename=$1
+	local ost_idx=$2
+	# on the mdt's osc
+	local ost_name=$(ostname_from_index $ost_idx)
+	local facets=$(get_facets MDS)
+	local wms
+	local MDS
+
+	for MDS in ${facets//,/ }; do
+		local mdtosc_proc=$(get_mdtosc_proc_path $MDS $ost_name)
+
+		do_facet $MDS $LCTL get_param -n \
+			osp.$mdtosc_proc.reserved_mb_high ||
+			skip  "remote MDS does not support reserved_mb_high"
+	done
+
+	wms=$(ost_watermarks_set_low_space $ost_idx)
+	local new_lwm=$(echo $wms | awk '{ print $4 }')
+	fill_ost $filename $ost_idx $new_lwm
+	#First enospc could execute orphan deletion so repeat
+	fill_ost $filename $ost_idx $new_lwm
+	echo $wms
+}
+
+ost_watermarks_enospc_delete_files() {
+	local filename=$1
+	local ost_idx=$2
+
+	rm -f $DIR/${filename}.fill_ost$ost_idx
+
+	wait_delete_completed
+	wait_mds_ost_sync
+}
+
+# clean up from "ost_watermarks_set_enospc"
+ost_watermarks_clear_enospc() {
+	local filename=$1
+	local ost_idx=$2
+	local old_lwm=$4
+	local old_hwm=$5
+
+	ost_watermarks_enospc_delete_files $filename $ost_idx
+	ost_watermarks_set $ost_idx $old_lwm $old_hwm
+	echo "set OST$ost_idx lwm back to $old_lwm, hwm back to $old_hwm"
+}
+
 wait_delete_completed_mds() {
 	local max_wait=${1:-20}
 	local mds2sync=""
@@ -3283,6 +3509,10 @@ facet_failover() {
 		change_active ${affecteds[index]}
 
 		wait_for_facet ${affecteds[index]}
+		if $GSS_SK; then
+			init_gss
+			init_facets_vars_simple
+		fi
 		# start mgs first if it is affected
 		if ! combined_mgs_mds &&
 			list_member ${affecteds[index]} mgs; then
@@ -3292,6 +3522,12 @@ facet_failover() {
 		affected=$(exclude_items_from_list ${affecteds[index]} mgs)
 		echo mount facets: ${affecteds[index]}
 		mount_facets ${affecteds[index]}
+		if $GSS_SK; then
+			do_nodes $(comma_list $(all_nodes)) \
+				"keyctl show | grep lustre | cut -c1-11 |
+				sed -e 's/ //g;' |
+				xargs -IX keyctl setperm X 0x3f3f3f3f"
+		fi
 	done
 }
 
@@ -3385,7 +3621,12 @@ fail() {
 	local facets=$1
 	local clients=${CLIENTS:-$HOSTNAME}
 
+	SK_NO_KEY_save=$SK_NO_KEY
+	if $GSS_SK; then
+		export SK_NO_KEY=false
+	fi
 	facet_failover $* || error "failover: $?"
+	export SK_NO_KEY=$SK_NO_KEY_save
 	# to initiate all OSC idling connections
 	clients_up
 	wait_clients_import_state "$clients" "$facets" "\(FULL\|IDLE\)"
@@ -4108,40 +4349,44 @@ unmount_fstype() {
 ## MountConf setup
 
 stopall() {
-    # make sure we are using the primary server, so test-framework will
-    # be able to clean up properly.
-    activemds=`facet_active mds1`
-    if [ $activemds != "mds1" ]; then
-        fail mds1
-    fi
+	# make sure we are using the primary server, so test-framework will
+	# be able to clean up properly.
+	activemds=`facet_active mds1`
+	if [ $activemds != "mds1" ]; then
+		fail mds1
+	fi
 
-    local clients=$CLIENTS
-    [ -z $clients ] && clients=$(hostname)
+	local clients=$CLIENTS
+	[ -z $clients ] && clients=$(hostname)
 
-    zconf_umount_clients $clients $MOUNT "$*" || true
-    [ -n "$MOUNT2" ] && zconf_umount_clients $clients $MOUNT2 "$*" || true
+	zconf_umount_clients $clients $MOUNT "$*" || true
+	[ -n "$MOUNT2" ] && zconf_umount_clients $clients $MOUNT2 "$*" || true
 
-    [ -n "$CLIENTONLY" ] && return
+	[ -n "$CLIENTONLY" ] && return
 
-    # The add fn does rm ${facet}active file, this would be enough
-    # if we use do_facet <facet> only after the facet added, but
-    # currently we use do_facet mds in local.sh
-    for num in `seq $MDSCOUNT`; do
-        stop mds$num -f
-        rm -f ${TMP}/mds${num}active
-    done
-    combined_mgs_mds && rm -f $TMP/mgsactive
+	# The add fn does rm ${facet}active file, this would be enough
+	# if we use do_facet <facet> only after the facet added, but
+	# currently we use do_facet mds in local.sh
+	for num in `seq $MDSCOUNT`; do
+		stop mds$num -f
+		rm -f ${TMP}/mds${num}active
+	done
+	combined_mgs_mds && rm -f $TMP/mgsactive
 
-    for num in `seq $OSTCOUNT`; do
-        stop ost$num -f
-        rm -f $TMP/ost${num}active
-    done
+	for num in `seq $OSTCOUNT`; do
+		stop ost$num -f
+		rm -f $TMP/ost${num}active
+	done
 
-    if ! combined_mgs_mds ; then
-        stop mgs
-    fi
+	if ! combined_mgs_mds ; then
+		stop mgs
+	fi
 
-    return 0
+	if $SHARED_KEY; then
+		export SK_MOUNTED=false
+	fi
+
+	return 0
 }
 
 cleanup_echo_devs () {
@@ -4180,6 +4425,48 @@ upper() {
 	echo -n "$1" | tr '[:lower:]' '[:upper:]'
 }
 
+squash_opt() {
+	local var="$*"
+	local other=""
+	local opt_o=""
+	local opt_e=""
+	local first_e=0
+	local first_o=0
+	local take=""
+
+	var=$(echo "$var" | sed -e 's/,\( \)*/,/g')
+	for i in $(echo "$var"); do
+		if [ "$i" == "-O" ]; then
+			take="o";
+			first_o=$(($first_o + 1))
+			continue;
+		fi
+		if [ "$i" == "-E" ]; then
+			take="e";
+			first_e=$(($first_e + 1 ))
+			continue;
+		fi
+		case $take in
+			"o")
+				[ $first_o -gt 1 ] && opt_o+=",";
+				opt_o+="$i";
+				;;
+			"e")
+				[ $first_e -gt 1 ] && opt_e+=",";
+				opt_e+="$i";
+				;;
+			*)
+				other+=" $i";
+				;;
+		esac
+		take=""
+	done
+
+	echo -n "$other"
+	[ -n "$opt_o" ] && echo " -O $opt_o"
+	[ -n "$opt_e" ] && echo " -E $opt_e"
+}
+
 mkfs_opts() {
 	local facet=$1
 	local dev=$2
@@ -4220,11 +4507,7 @@ mkfs_opts() {
 		opts+=${L_GETIDENTITY:+" --param=mdt.identity_upcall=$L_GETIDENTITY"}
 
 		if [ $fstype == ldiskfs ]; then
-			# Check for wide striping
-			if [ $OSTCOUNT -gt 160 ]; then
-				MDSJOURNALSIZE=${MDSJOURNALSIZE:-4096}
-				fs_mkfs_opts+="-O ea_inode"
-			fi
+			fs_mkfs_opts+="-O ea_inode,large_dir"
 
 			var=${facet}_JRN
 			if [ -n "${!var}" ]; then
@@ -4262,6 +4545,8 @@ mkfs_opts() {
 	var=${type}_FS_MKFS_OPTS
 	fs_mkfs_opts+=${!var:+" ${!var}"}
 
+	[ $fstype == ldiskfs ] && fs_mkfs_opts=$(squash_opt $fs_mkfs_opts)
+
 	if [ -n "${fs_mkfs_opts## }" ]; then
 		opts+=" --mkfsoptions=\\\"${fs_mkfs_opts## }\\\""
 	fi
@@ -4326,6 +4611,9 @@ __touch_device()
 	[[ ! "$device" =~ ^/dev/ ]] || [[ "$device" =~ ^/dev/shm/ ]] ||
 		error "$facet: device '$device' does not exist"
 
+	# zpool create doesn't like empty files
+	[[ $(facet_fstype $facet) == zfs ]] && return 0
+
 	do_facet $facet "touch \"${device}\""
 }
 
@@ -4386,7 +4674,7 @@ formatall() {
 	# (Assumes MDS version is also OSS version)
 	if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ];
 	then
-	    do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+		do_rpc_nodes "$(comma_list $(all_server_nodes))" set_hostid
 	fi
 
 	# We need ldiskfs here, may as well load them all
@@ -4631,8 +4919,17 @@ setupall() {
 		if $GSS_SK; then
 			set_rule $FSNAME any cli2mdt $SK_FLAVOR
 			set_rule $FSNAME any cli2ost $SK_FLAVOR
-			wait_flavor cli2mdt $SK_FLAVOR
-			wait_flavor cli2ost $SK_FLAVOR
+			if $SK_SKIPFIRST; then
+				export SK_SKIPFIRST=false
+
+				sleep 30
+				do_nodes $CLIENTS \
+					 "lctl set_param osc.*.idle_connect=1"
+				return
+			else
+				wait_flavor cli2mdt $SK_FLAVOR
+				wait_flavor cli2ost $SK_FLAVOR
+			fi
 		else
 			set_flavor_all $SEC
 		fi
@@ -4738,6 +5035,31 @@ init_facets_vars () {
 	fi
 }
 
+init_facets_vars_simple () {
+	local devname
+
+	if ! remote_mds_nodsh; then
+		for num in $(seq $MDSCOUNT); do
+			devname=$(mdsdevname $num)
+			eval export mds${num}_dev=${devname}
+			eval export mds${num}_opt=\"${MDS_MOUNT_OPTS}\"
+		done
+	fi
+
+	if ! combined_mgs_mds ; then
+		eval export mgs_dev=$(mgsdevname)
+		eval export mgs_opt=\"${MGS_MOUNT_OPTS}\"
+	fi
+
+	if ! remote_ost_nodsh; then
+		for num in $(seq $OSTCOUNT); do
+			devname=$(ostdevname $num)
+			eval export ost${num}_dev=${devname}
+			eval export ost${num}_opt=\"${OST_MOUNT_OPTS}\"
+		done
+	fi
+}
+
 osc_ensure_active () {
     local facet=$1
     local timeout=$2
@@ -4789,7 +5111,7 @@ set_persistent_param() {
 		final=$((orig + 5))
 	fi
 
-	if [[ $PERM_CMD = *"set_param -P"* ]]; then
+	if [[ $PERM_CMD == *"set_param -P"* ]]; then
 		echo "Setting $test_param from $orig to $final"
 		do_facet mgs "$PERM_CMD $test_param='$final'" ||
 			error "$PERM_CMD $test_param failed"
@@ -4823,7 +5145,7 @@ init_param_vars () {
 	TIMEOUT=$(lctl get_param -n timeout)
 	TIMEOUT=${TIMEOUT:-20}
 
-	if [ -n $arg1 ]; then
+	if [ -n "$arg1" ]; then
 		[ "$arg1" = "server_only" ] && return
 	fi
 
@@ -5043,22 +5365,12 @@ check_and_setup_lustre() {
 		fi
 	fi
 
-	init_gss
 	if $GSS_SK; then
 		set_flavor_all null
 	elif $GSS; then
 		set_flavor_all $SEC
 	fi
 
-	if [ -z "$CLIENTONLY" ]; then
-		# Enable remote MDT create for testing
-		for num in $(seq $MDSCOUNT); do
-			do_facet mds$num \
-				lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
-					2>/dev/null
-		done
-	fi
-
 	if [ "$ONLY" == "setup" ]; then
 		exit 0
 	fi
@@ -5488,7 +5800,7 @@ at_max_set() {
 drop_request() {
 # OBD_FAIL_MDS_ALL_REQUEST_NET
     RC=0
-    do_facet $SINGLEMDS lctl set_param fail_loc=0x123
+    do_facet $SINGLEMDS lctl set_param fail_val=0 fail_loc=0x123
     do_facet client "$1" || RC=$?
     do_facet $SINGLEMDS lctl set_param fail_loc=0
     return $RC
@@ -5573,11 +5885,11 @@ drop_bl_callback() {
 	return $rc
 }
 
-drop_ldlm_reply() {
-#define OBD_FAIL_LDLM_REPLY              0x30c
+drop_mdt_ldlm_reply() {
+#define OBD_FAIL_MDS_LDLM_REPLY_NET	0x157
     RC=0
-    local list=$(comma_list $(mdts_nodes) $(osts_nodes))
-    do_nodes $list lctl set_param fail_loc=0x30c
+    local list=$(comma_list $(mdts_nodes))
+    do_nodes $list lctl set_param fail_loc=0x157
 
     do_facet client "$@" || RC=$?
 
@@ -5585,11 +5897,11 @@ drop_ldlm_reply() {
     return $RC
 }
 
-drop_ldlm_reply_once() {
-#define OBD_FAIL_LDLM_REPLY              0x30c
+drop_mdt_ldlm_reply_once() {
+#define OBD_FAIL_MDS_LDLM_REPLY_NET	0x157
     RC=0
-    local list=$(comma_list $(mdts_nodes) $(osts_nodes))
-    do_nodes $list lctl set_param fail_loc=0x8000030c
+    local list=$(comma_list $(mdts_nodes))
+    do_nodes $list lctl set_param fail_loc=0x80000157
 
     do_facet client "$@" || RC=$?
 
@@ -5751,7 +6063,7 @@ report_error() {
 # usage: stack_trap arg sigspec
 #
 # stack_trap() behaves like bash's built-in trap, except that it "stacks" the
-# command ``arg`` on top of previously defined commands for ``sigspec`` instead
+# command "arg" on top of previously defined commands for "sigspec" instead
 # of overwriting them.
 # stacked traps are executed in reverse order of their registration
 #
@@ -5759,7 +6071,7 @@ report_error() {
 stack_trap()
 {
 	local arg="$1"
-	local sigspec="$2"
+	local sigspec="${2:-EXIT}"
 
 	# Use "trap -p" to get the quoting right
 	local old_trap="$(trap -p "$sigspec")"
@@ -5786,7 +6098,7 @@ exit_status () {
 	local status=0
 	local log=$TESTSUITELOG
 
-	[ -f "$log" ] && grep -q FAIL $log && status=1
+	[ -f "$log" ] && grep -qw FAIL $log && status=1
 	exit $status
 }
 
@@ -5828,6 +6140,12 @@ error_not_in_vm() {
 	fi
 }
 
+#
+# Function: skip_env()
+# Purpose:  to skip a test during developer testing because some tool
+#           is missing, but fail the test in release testing because the test
+#           environment is not configured properly".
+#
 skip_env () {
 	$FAIL_ON_SKIP_ENV && error false $@ || skip $@
 }
@@ -5904,63 +6222,78 @@ export ALWAYS_SKIPPED=
 #
 run_test() {
 	assert_DIR
-
 	export base=$(basetest $1)
-	if [ -n "$ONLY" ]; then
-		testname=ONLY_$1
-		if [ ${!testname}x != x ]; then
-			[ -n "$LAST_SKIPPED" ] && echo "" && LAST_SKIPPED=
-			run_one_logged $1 "$2"
-			return $?
-		fi
-		testname=ONLY_$base
-		if [ ${!testname}x != x ]; then
-			[ -n "$LAST_SKIPPED" ] && echo "" && LAST_SKIPPED=
-			run_one_logged $1 "$2"
-			return $?
-		fi
-		LAST_SKIPPED="y"
-		return 0
-	fi
+	TESTNAME=test_$1
+	LAST_SKIPPED=
+	ALWAYS_SKIPPED=
 
-	LAST_SKIPPED="y"
-	ALWAYS_SKIPPED="y"
-	testname=EXCEPT_$1
+	# Check the EXCEPT, ALWAYS_EXCEPT and SLOW lists to see if we
+	# need to skip the current test. If so, set the ALWAYS_SKIPPED flag.
+	local testname=EXCEPT_$1
+	local testname_base=EXCEPT_$base
 	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping excluded test $1"
-		return 0
-	fi
-	testname=EXCEPT_$base
-	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping excluded test $1 (base $base)"
-		return 0
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping excluded test $1"
+	elif [ ${!testname_base}x != x ]; then
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping excluded test $1 (base $base)"
 	fi
+
 	testname=EXCEPT_ALWAYS_$1
+	testname_base=EXCEPT_ALWAYS_$base
 	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping ALWAYS excluded test $1"
-		return 0
-	fi
-	testname=EXCEPT_ALWAYS_$base
-	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping ALWAYS excluded test $1 (base $base)"
-		return 0
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping ALWAYS excluded test $1"
+	elif [ ${!testname_base}x != x ]; then
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping ALWAYS excluded test $1 (base $base)"
 	fi
+
 	testname=EXCEPT_SLOW_$1
+	testname_base=EXCEPT_SLOW_$base
 	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping SLOW test $1"
-		return 0
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping SLOW test $1"
+	elif [ ${!testname_base}x != x ]; then
+		ALWAYS_SKIPPED="y"
+		skip_message="skipping SLOW test $1 (base $base)"
 	fi
-	testname=EXCEPT_SLOW_$base
-	if [ ${!testname}x != x ]; then
-		TESTNAME=test_$1 skip_noexit "skipping SLOW test $1 (base $base)"
-		return 0
+
+	# If there are tests on the ONLY list, check if the current test
+	# is on that list and, if so, check if the test is to be skipped
+	# and if we are supposed to honor the skip lists.
+	if [ -n "$ONLY" ]; then
+		testname=ONLY_$1
+		testname_base=ONLY_$base
+		if [[ ${!testname}x != x || ${!testname_base}x != x ]]; then
+
+			if [[ -n "$ALWAYS_SKIPPED" && -n "$HONOR_EXCEPT" ]]; then
+				LAST_SKIPPED="y"
+				skip_noexit "$skip_message"
+				return 0
+			else
+				[ -n "$LAST_SKIPPED" ] &&
+					echo "" && LAST_SKIPPED=
+				ALWAYS_SKIPPED=
+				run_one_logged $1 "$2"
+				return $?
+			fi
+
+		else
+			LAST_SKIPPED="y"
+			return 0
+		fi
 	fi
 
-	LAST_SKIPPED=
-	ALWAYS_SKIPPED=
-	run_one_logged $1 "$2"
+	if [ -n "$ALWAYS_SKIPPED" ]; then
+		LAST_SKIPPED="y"
+		skip_noexit "$skip_message"
+		return 0
+	else
+		run_one_logged $1 "$2"
+		return $?
+	fi
 
-	return $?
 }
 
 log() {
@@ -6399,7 +6732,7 @@ remote_nodes_list () {
 all_mdts_nodes () {
 	local host
 	local failover_host
-	local nodes="${mds_HOST} ${mdsfailover_HOST}"
+	local nodes
 	local nodes_sort
 	local i
 
@@ -6409,6 +6742,7 @@ all_mdts_nodes () {
 		nodes="$nodes ${!host} ${!failover_host}"
 	done
 
+	[ -n "$nodes" ] || nodes="${mds_HOST} ${mdsfailover_HOST}"
 	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
 	echo -n $nodes_sort
 }
@@ -6417,7 +6751,7 @@ all_mdts_nodes () {
 all_osts_nodes () {
 	local host
 	local failover_host
-	local nodes="${ost_HOST} ${ostfailover_HOST}"
+	local nodes=
 	local nodes_sort
 	local i
 
@@ -6427,6 +6761,7 @@ all_osts_nodes () {
 		nodes="$nodes ${!host} ${!failover_host}"
 	done
 
+	[ -n "$nodes" ] || nodes="${ost_HOST} ${ostfailover_HOST}"
 	nodes_sort=$(for i in $nodes; do echo $i; done | sort -u)
 	echo -n $nodes_sort
 }
@@ -6814,23 +7149,25 @@ delayed_recovery_enabled () {
 ########################
 
 convert_facet2label() {
-    local facet=$1
+	local facet=$1
 
-    if [ x$facet = xost ]; then
-       facet=ost1
-    fi
+	if [ x$facet = xost ]; then
+		facet=ost1
+	elif [ x$facet = xmgs ] && combined_mgs_mds ; then
+		facet=mds1
+	fi
 
-    local varsvc=${facet}_svc
+	local varsvc=${facet}_svc
 
-    if [ -n ${!varsvc} ]; then
-        echo ${!varsvc}
-    else
-        error "No lablel for $facet!"
-    fi
+	if [ -n ${!varsvc} ]; then
+		echo ${!varsvc}
+	else
+		error "No label for $facet!"
+	fi
 }
 
 get_clientosc_proc_path() {
-	echo "${1}-osc-ffff*"
+	echo "${1}-osc-[-0-9a-f]*"
 }
 
 # If the 2.0 MDS was mounted on 1.8 device, then the OSC and LOV names
@@ -6974,7 +7311,7 @@ _wait_osc_import_state() {
 	local ost_facet=$2
 	local expected=$3
 	local target=$(get_osc_import_name $facet $ost_facet)
-	local param="osc.${target}.ost_server_uuid"
+	local param="os[cp].${target}.ost_server_uuid"
 	local params=$param
 	local i=0
 
@@ -7184,7 +7521,7 @@ wait_osp_active() {
 
 			[ ${PIPESTATUS[0]} = 0 ] || error "Can't read $mproc"
 			if [ $result -eq $expected ]; then
-				echo -n "target updated after"
+				echo -n "target updated after "
 				echo "$wait sec (got $result)"
 				break
 			fi
@@ -7286,6 +7623,8 @@ destroy_pool_int() {
 	for ost in $OSTS; do
 		do_facet mgs lctl pool_remove $1 $ost
 	done
+	wait_update_facet $SINGLEMDS "lctl pool_list $1 | wc -l" "1" ||
+		error "MDS: pool_list $1 failed"
 	do_facet mgs lctl pool_destroy $1
 }
 
@@ -7609,6 +7948,8 @@ flvr_cnt_cli2ost()
     local clients=${CLIENTS:-$HOSTNAME}
 
     for c in ${clients//,/ }; do
+	# reconnect if idle
+	do_node $c lctl set_param osc.*.idle_connect=1 >/dev/null 2>&1
 	local output=$(do_node $c lctl get_param -n \
 		 osc.*OST*-osc-[^M][^D][^T]*.$PROC_CLI 2>/dev/null)
 	local tmpcnt=$(count_flvr "$output" $flavor)
@@ -7664,12 +8005,12 @@ flvr_cnt_mdt2ost()
         mdtosc=$(get_mdtosc_proc_path mds$num)
         mdtosc=${mdtosc/-MDT*/-MDT\*}
 	local output=$(do_facet mds$num lctl get_param -n \
-		 osc.$mdtosc.$PROC_CLI 2>/dev/null)
+		       os[cp].$mdtosc.$PROC_CLI 2>/dev/null)
 	local tmpcnt=$(count_flvr "$output" $flavor)
 	if $GSS_SK && [ $flavor != "null" ]; then
 		# tmpcnt=min(contexts,flavors) to ensure SK context is on
 		output=$(do_facet mds$num lctl get_param -n \
-			 osc.$mdtosc.$PROC_CON 2>/dev/null)
+			 os[cp].$mdtosc.$PROC_CON 2>/dev/null)
 		local outcon=$(count_contexts "$output")
 		if [ "$outcon" -lt "$tmpcnt" ]; then
 			tmpcnt=$outcon
@@ -8093,13 +8434,15 @@ get_obd_size() {
 
 #
 # Get the page size (bytes) on a given facet node.
+# The local client page_size is directly available in PAGE_SIZE.
 #
 get_page_size() {
 	local facet=$1
-	local size=$(getconf PAGE_SIZE 2>/dev/null)
+	local page_size=$(getconf PAGE_SIZE 2>/dev/null)
 
-	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet getconf PAGE_SIZE)
-	echo -n ${size:-4096}
+	[ -z "$CLIENTONLY" -a "$facet" != "client" ] &&
+		page_size=$(do_facet $facet getconf PAGE_SIZE)
+	echo -n ${page_size:-4096}
 }
 
 #
@@ -8144,8 +8487,7 @@ max_xattr_size() {
     local size
 
     if large_xattr_enabled; then
-        # include/linux/limits.h: #define XATTR_SIZE_MAX 65536
-        size=65536
+	size=$($LCTL get_param -n llite.*.max_easize)
     else
         local mds_dev=$(mdsdevname ${SINGLEMDS//mds/})
         local block_size=$(get_block_size $SINGLEMDS $mds_dev)
@@ -8294,14 +8636,14 @@ generate_logname() {
 test_mkdir() {
 	local path
 	local p_option
-	local stripe_count=2
-	local stripe_index=-1
+	local dirstripe_count=${DIRSTRIPE_COUNT:-"2"}
+	local dirstripe_index=${DIRSTRIPE_INDEX:-$((base % $MDSCOUNT))}
 	local OPTIND=1
 
 	while getopts "c:i:p" opt; do
 		case $opt in
-			c) stripe_count=$OPTARG;;
-			i) stripe_index=$OPTARG;;
+			c) dirstripe_count=$OPTARG;;
+			i) dirstripe_index=$OPTARG;;
 			p) p_option="-p";;
 			\?) error "only support -i -c -p";;
 		esac
@@ -8324,17 +8666,25 @@ test_mkdir() {
 	if [ $MDSCOUNT -le 1 ]; then
 		mkdir $path || error "mkdir '$path' failed"
 	else
-		local test_num=$(echo $testnum | sed -e 's/[^0-9]*//g')
 		local mdt_index
 
-		if [ $stripe_index -eq -1 ]; then
-			mdt_index=$((test_num % MDSCOUNT))
+		if [ $dirstripe_index -eq -1 ]; then
+			mdt_index=$((base % MDSCOUNT))
+		else
+			mdt_index=$dirstripe_index
+		fi
+
+		if (($MDS1_VERSION >= $(version_code 2.8.0))); then
+			if [ $dirstripe_count -eq -1 ]; then
+				dirstripe_count=$((RANDOM % MDSCOUNT + 1))
+			fi
 		else
-			mdt_index=$stripe_index
+			dirstripe_count=1
 		fi
-		echo "striped dir -i$mdt_index -c$stripe_count $path"
-		$LFS mkdir -i$mdt_index -c$stripe_count $path ||
-			error "mkdir -i $mdt_index -c$stripe_count $path failed"
+
+		echo "striped dir -i$mdt_index -c$dirstripe_count $path"
+		$LFS mkdir -i$mdt_index -c$dirstripe_count $path ||
+			error "mkdir -i $mdt_index -c$dirstripe_count $path failed"
 	fi
 }
 
@@ -8420,7 +8770,11 @@ pool_add_targets() {
 	local last=$3
 	local step=${4:-1}
 
-	local list=$(seq $first $step $last)
+	if [ -z $last ]; then
+		local list=$first
+	else
+		local list=$(seq $first $step $last)
+	fi
 
 	local t=$(for i in $list; do printf "$FSNAME-OST%04x_UUID " $i; done)
 	do_facet mgs $LCTL pool_add \
@@ -8669,15 +9023,15 @@ check_stripe_count() {
 	[[ -z "$file" || -z "$expected" ]] &&
 		error "check_stripe_count: invalid argument"
 
-	local cmd="$GETSTRIPE -c $file"
+	local cmd="$LFS getstripe -c $file"
 	actual=$($cmd) || error "$cmd failed"
 	actual=${actual%% *}
 
 	if [[ $actual -ne $expected ]]; then
-		[[ $expected -eq -1 ]] ||
-			error "$cmd wrong: found $actual, expected $expected"
-		[[ $actual -eq $OSTCOUNT ]] ||
-			error "$cmd wrong: found $actual, expected $OSTCOUNT"
+		[[ $expected -eq -1 ]] || { $LFS getstripe $file;
+			error "$cmd not expected ($expected): found $actual"; }
+		[[ $actual -eq $OSTCOUNT ]] || { $LFS getstripe $file;
+			error "$cmd not OST count ($OSTCOUNT): found $actual"; }
 	fi
 }
 
@@ -9341,3 +9695,580 @@ save_layout_restore_at_exit() {
 
 	stack_trap "restore_layout $dir $layout" EXIT
 }
+
+verify_yaml_layout() {
+	local src=$1
+	local dst=$2
+	local temp=$3
+	local msg_prefix=$4
+
+	echo "getstripe --yaml $src"
+	$LFS getstripe --yaml $src > $temp || error "getstripe $src failed"
+	echo "setstripe --yaml=$temp $dst"
+	$LFS setstripe --yaml=$temp $dst|| error "setstripe $dst failed"
+
+	echo "compare"
+	local layout1=$(get_layout_param $src)
+	local layout2=$(get_layout_param $dst)
+	# compare their layout info
+	[ "$layout1" == "$layout2" ] ||
+		error "$msg_prefix $src/$dst layouts are not equal"
+}
+
+is_project_quota_supported() {
+	$ENABLE_PROJECT_QUOTAS || return 1
+	[ "$(facet_fstype $SINGLEMDS)" == "ldiskfs" ] &&
+		[ $(lustre_version_code $SINGLEMDS) -gt \
+		$(version_code 2.9.55) ] &&
+		lfs --help | grep project >&/dev/null &&
+		egrep -q "7." /etc/redhat-release && return 0
+
+	if [ "$(facet_fstype $SINGLEMDS)" == "zfs" ]; then
+		[ $(lustre_version_code $SINGLEMDS) -le \
+			$(version_code 2.10.53) ] && return 1
+
+		do_facet mds1 $ZPOOL upgrade -v |
+			grep project_quota && return 0
+	fi
+
+	return 1
+}
+
+enable_project_quota() {
+	is_project_quota_supported || return 0
+	[ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && return 0
+	stopall || error "failed to stopall (1)"
+
+	for num in $(seq $MDSCOUNT); do
+		do_facet mds$num $TUNE2FS -O project $(mdsdevname $num) ||
+			error "tune2fs $(mdsdevname $num) failed"
+	done
+
+	for num in $(seq $OSTCOUNT); do
+		do_facet ost$num $TUNE2FS -O project $(ostdevname $num) ||
+			error "tune2fs $(ostdevname $num) failed"
+	done
+
+	mount
+	setupall
+}
+
+disable_project_quota() {
+	is_project_quota_supported || return 0
+	[ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] && return 0
+	stopall || error "failed to stopall (1)"
+
+	for num in $(seq $MDSCOUNT); do
+		do_facet mds$num $TUNE2FS -Q ^prj $(mdsdevname $num) ||
+			error "tune2fs $(mdsdevname $num) failed"
+	done
+
+	for num in $(seq $OSTCOUNT); do
+		do_facet ost$num $TUNE2FS -Q ^prj $(ostdevname $num) ||
+			error "tune2fs $(ostdevname $num) failed"
+	done
+
+	mount
+	setupall
+}
+
+#
+# In order to test multiple remote HSM agents, a new facet type named "AGT" and
+# the following associated variables are added:
+#
+# AGTCOUNT: number of agents
+# AGTDEV{N}: target HSM mount point (root path of the backend)
+# agt{N}_HOST: hostname of the agent agt{N}
+# SINGLEAGT: facet of the single agent
+#
+# The number of agents is initialized as the number of remote client nodes.
+# By default, only single copytool is started on a remote client/agent. If there
+# was no remote client, then the copytool will be started on the local client.
+#
+init_agt_vars() {
+	local n
+	local agent
+
+	export AGTCOUNT=${AGTCOUNT:-$((CLIENTCOUNT - 1))}
+	[[ $AGTCOUNT -gt 0 ]] || AGTCOUNT=1
+
+	export SHARED_DIRECTORY=${SHARED_DIRECTORY:-$TMP}
+	if [[ $CLIENTCOUNT -gt 1 ]] &&
+		! check_shared_dir $SHARED_DIRECTORY $CLIENTS; then
+		skip_env "SHARED_DIRECTORY should be accessible"\
+			 "on all client nodes"
+		exit 0
+	fi
+
+	# We used to put the HSM archive in $SHARED_DIRECTORY but that
+	# meant NFS issues could hose sanity-hsm sessions. So now we
+	# use $TMP instead.
+	for n in $(seq $AGTCOUNT); do
+		eval export AGTDEV$n=\$\{AGTDEV$n:-"$TMP/arc$n"\}
+		agent=CLIENT$((n + 1))
+		if [[ -z "${!agent}" ]]; then
+			[[ $CLIENTCOUNT -eq 1 ]] && agent=CLIENT1 ||
+				agent=CLIENT2
+		fi
+		eval export agt${n}_HOST=\$\{agt${n}_HOST:-${!agent}\}
+		local var=agt${n}_HOST
+		[[ ! -z "${!var}" ]] || error "agt${n}_HOST is empty!"
+	done
+
+	export SINGLEAGT=${SINGLEAGT:-agt1}
+
+	export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
+	export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
+	export HSMTOOL_UPDATE_INTERVAL=${HSMTOOL_UPDATE_INTERVAL:=""}
+	export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
+	export HSMTOOL_TESTDIR
+	export HSMTOOL_BASE=$(basename "$HSMTOOL" | cut -f1 -d" ")
+
+	HSM_ARCHIVE_NUMBER=2
+
+	# The test only support up to 10 MDTs
+	MDT_PREFIX="mdt.$FSNAME-MDT000"
+	HSM_PARAM="${MDT_PREFIX}0.hsm"
+
+	# archive is purged at copytool setup
+	HSM_ARCHIVE_PURGE=true
+
+	# Don't allow copytool error upon start/setup
+	HSMTOOL_NOERROR=false
+}
+
+# Get the backend root path for the given agent facet.
+copytool_device() {
+	local facet=$1
+	local dev=AGTDEV$(facet_number $facet)
+
+	echo -n ${!dev}
+}
+
+get_mdt_devices() {
+	local mdtno
+	# get MDT device for each mdc
+	for mdtno in $(seq 1 $MDSCOUNT); do
+		local idx=$(($mdtno - 1))
+		MDT[$idx]=$($LCTL get_param -n \
+			mdc.$FSNAME-MDT000${idx}-mdc-*.mds_server_uuid |
+			awk '{gsub(/_UUID/,""); print $1}' | head -n1)
+	done
+}
+
+search_copytools() {
+	local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+	do_nodesv $hosts "pgrep -x $HSMTOOL_BASE"
+}
+
+kill_copytools() {
+	local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+
+	echo "Killing existing copytools on $hosts"
+	do_nodesv $hosts "killall -q $HSMTOOL_BASE" || true
+}
+
+wait_copytools() {
+	local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+	local wait_timeout=200
+	local wait_start=$SECONDS
+	local wait_end=$((wait_start + wait_timeout))
+	local sleep_time=100000 # 0.1 second
+
+	while ((SECONDS < wait_end)); do
+		if ! search_copytools $hosts; then
+			echo "copytools stopped in $((SECONDS - wait_start))s"
+			return 0
+		fi
+
+		echo "copytools still running on $hosts"
+		usleep $sleep_time
+		[ $sleep_time -lt 32000000 ] && # 3.2 seconds
+			sleep_time=$(bc <<< "$sleep_time * 2")
+	done
+
+	# try to dump Copytool's stack
+	do_nodesv $hosts "echo 1 >/proc/sys/kernel/sysrq ; " \
+			 "echo t >/proc/sysrq-trigger"
+
+	echo "copytools failed to stop in ${wait_timeout}s"
+
+	return 1
+}
+
+copytool_monitor_cleanup() {
+	local facet=${1:-$SINGLEAGT}
+	local agent=$(facet_active_host $facet)
+
+	if [ -n "$HSMTOOL_MONITOR_DIR" ]; then
+		# Should die when the copytool dies, but just in case.
+		local cmd="kill \\\$(cat $HSMTOOL_MONITOR_DIR/monitor_pid)"
+		cmd+=" 2>/dev/null || true"
+		do_node $agent "$cmd"
+		do_node $agent "rm -fr $HSMTOOL_MONITOR_DIR"
+		export HSMTOOL_MONITOR_DIR=
+	fi
+
+	# The pdsh should die on its own when the monitor dies. Just
+	# in case, though, try to clean up to avoid any cruft.
+	if [ -n "$HSMTOOL_MONITOR_PDSH" ]; then
+		kill $HSMTOOL_MONITOR_PDSH 2>/dev/null || true
+		export HSMTOOL_MONITOR_PDSH=
+	fi
+}
+
+copytool_logfile()
+{
+	local host="$(facet_host "$1")"
+	local prefix=$TESTLOG_PREFIX
+	[ -n "$TESTNAME" ] && prefix+=.$TESTNAME
+
+	printf "${prefix}.copytool${archive_id}_log.${host}.log"
+}
+
+__lhsmtool_rebind()
+{
+	do_facet $facet $HSMTOOL -p "$hsm_root" --rebind "$@" "$mountpoint"
+}
+
+__lhsmtool_import()
+{
+	mkdir -p "$(dirname "$2")" ||
+		error "cannot create directory '$(dirname "$2")'"
+	do_facet $facet $HSMTOOL -p "$hsm_root" --import "$@" "$mountpoint"
+}
+
+__lhsmtool_setup()
+{
+	local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --hsm-root \"$hsm_root\""
+	[ -n "$bandwidth" ] && cmd+=" --bandwidth $bandwidth"
+	[ -n "$archive_id" ] && cmd+=" --archive $archive_id"
+	[ ${#misc_options[@]} -gt 0 ] &&
+		cmd+=" $(IFS=" " echo "$@")"
+	cmd+=" \"$mountpoint\""
+
+	echo "Starting copytool $facet on $(facet_host $facet)"
+	stack_trap "do_facet $facet libtool execute pkill -x '$HSMTOOL' || true" EXIT
+	do_facet $facet "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
+}
+
+hsm_root() {
+	local facet="${1:-$SINGLEAGT}"
+
+	printf "$(copytool_device "$facet")/${TESTSUITE}.${TESTNAME}/"
+}
+
+# Main entry point to perform copytool related operations
+#
+# Sub-commands:
+#
+#	setup	setup a copytool to run in the background, that copytool will be
+#		killed on EXIT
+#	import	import a file from an HSM backend
+#	rebind	rebind an archived file to a new fid
+#
+# Although the semantics might suggest otherwise, one does not need to 'setup'
+# a copytool before a call to 'copytool import' or 'copytool rebind'.
+#
+copytool()
+{
+	local action=$1
+	shift
+
+	# Parse arguments
+	local fail_on_error=true
+	local -a misc_options
+	while [ $# -gt 0 ]; do
+		case "$1" in
+		-f|--facet)
+			shift
+			local facet="$1"
+			;;
+		-m|--mountpoint)
+			shift
+			local mountpoint="$1"
+			;;
+		-a|--archive-id)
+			shift
+			local archive_id="$1"
+			;;
+		-h|--hsm-root)
+			shift
+			local hsm_root="$1"
+			;;
+		-b|--bwlimit)
+			shift
+			local bandwidth="$1" # in MB/s
+			;;
+		-n|--no-fail)
+			local fail_on_error=false
+			;;
+		*)
+			# Uncommon(/copytool dependent) option
+			misc_options+=("$1")
+			;;
+		esac
+		shift
+	done
+
+	# Use default values if needed
+	local facet=${facet:-$SINGLEAGT}
+	local mountpoint="${mountpoint:-${MOUNT2:-$MOUNT}}"
+	local hsm_root="${hsm_root:-$(hsm_root "$facet")}"
+
+	stack_trap "do_facet $facet rm -rf '$hsm_root'" EXIT
+	do_facet $facet mkdir -p "$hsm_root" ||
+		error "mkdir '$hsm_root' failed"
+
+	case "$HSMTOOL" in
+	lhsmtool_posix)
+		local copytool=lhsmtool
+		;;
+	esac
+
+	__${copytool}_${action} "${misc_options[@]}"
+	if [ $? -ne 0 ]; then
+		local error_msg
+
+		case $action in
+		setup)
+			local host="$(facet_host $facet)"
+			error_msg="Failed to start copytool $facet on '$host'"
+			;;
+		import)
+			local src="${misc_options[0]}"
+			local dest="${misc_options[1]}"
+			error_msg="Failed to import '$src' to '$dest'"
+			;;
+		rebind)
+			error_msg="could not rebind file"
+			;;
+		esac
+
+		$fail_on_error && error "$error_msg" || echo "$error_msg"
+	fi
+}
+
+needclients() {
+	local client_count=$1
+	if [[ $CLIENTCOUNT -lt $client_count ]]; then
+		skip "Need $client_count or more clients, have $CLIENTCOUNT"
+		return 1
+	fi
+	return 0
+}
+
+path2fid() {
+	$LFS path2fid $1 | tr -d '[]'
+	return ${PIPESTATUS[0]}
+}
+
+get_hsm_flags() {
+	local f=$1
+	local u=$2
+	local st
+
+	if [[ $u == "user" ]]; then
+		st=$($RUNAS $LFS hsm_state $f)
+	else
+		u=root
+		st=$($LFS hsm_state $f)
+	fi
+
+	[[ $? == 0 ]] || error "$LFS hsm_state $f failed (run as $u)"
+
+	st=$(echo $st | cut -f 2 -d" " | tr -d "()," )
+	echo $st
+}
+
+check_hsm_flags() {
+	local f=$1
+	local fl=$2
+
+	local st=$(get_hsm_flags $f)
+	[[ $st == $fl ]] || error "hsm flags on $f are $st != $fl"
+}
+
+mdts_set_param() {
+	local arg=$1
+	local key=$2
+	local value=$3
+	local mdtno
+	local rc=0
+	if [[ "$value" != "" ]]; then
+		value="=$value"
+	fi
+	for mdtno in $(seq 1 $MDSCOUNT); do
+		local idx=$(($mdtno - 1))
+		local facet=mds${mdtno}
+		# if $arg include -P option, run 1 set_param per MDT on the MGS
+		# else, run set_param on each MDT
+		[[ $arg = *"-P"* ]] && facet=mgs
+		do_facet $facet $LCTL set_param $arg mdt.${MDT[$idx]}.$key$value
+		[[ $? != 0 ]] && rc=1
+	done
+	return $rc
+}
+
+wait_result() {
+	local facet=$1
+	shift
+	wait_update --verbose $(facet_active_host $facet) "$@"
+}
+
+mdts_check_param() {
+	local key="$1"
+	local target="$2"
+	local timeout="$3"
+	local mdtno
+	for mdtno in $(seq 1 $MDSCOUNT); do
+		local idx=$(($mdtno - 1))
+		wait_result mds${mdtno} \
+			"$LCTL get_param -n $MDT_PREFIX${idx}.$key" "$target" \
+			$timeout ||
+			error "$key state is not '$target' on mds${mdtno}"
+	done
+}
+
+cdt_set_mount_state() {
+	mdts_set_param "-P" hsm_control "$1"
+	# set_param -P is asynchronous operation and could race with set_param.
+	# In such case configs could be retrieved and applied at mgc after
+	# set_param -P completion. Sleep here to avoid race with set_param.
+	# We need at least 20 seconds. 10 for mgc_requeue_thread to wake up
+	# MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC(5 + 5)
+	# and 10 seconds to retrieve config from server.
+	sleep 20
+}
+
+cdt_check_state() {
+	mdts_check_param hsm_control "$1" 20
+}
+
+cdt_set_sanity_policy() {
+	if [[ "$CDT_POLICY_HAD_CHANGED" ]]
+	then
+		# clear all
+		mdts_set_param "" hsm.policy "+NRA"
+		mdts_set_param "" hsm.policy "-NBR"
+		CDT_POLICY_HAD_CHANGED=
+	fi
+}
+
+set_hsm_param() {
+	local param=$1
+	local value=$2
+	local opt=$3
+	mdts_set_param "$opt -n" "hsm.$param" "$value"
+	return $?
+}
+
+wait_request_state() {
+	local fid=$1
+	local request=$2
+	local state=$3
+	# 4th arg (mdt index) is optional
+	local mdtidx=${4:-0}
+	local mds=mds$(($mdtidx + 1))
+
+	local cmd="$LCTL get_param -n ${MDT_PREFIX}${mdtidx}.hsm.actions"
+	cmd+=" | awk '/'$fid'.*action='$request'/ {print \\\$13}' | cut -f2 -d="
+
+	wait_result $mds "$cmd" "$state" 200 ||
+		error "request on $fid is not $state on $mds"
+}
+
+
+rmultiop_start() {
+	local client=$1
+	local file=$2
+	local cmds=$3
+	local WAIT_MAX=${4:-60}
+	local wait_time=0
+
+	# We need to run do_node in bg, because pdsh does not exit
+	# if child process of run script exists.
+	# I.e. pdsh does not exit when runmultiop_bg_pause exited,
+	# because of multiop_bg_pause -> $MULTIOP_PROG &
+	# By the same reason we need sleep a bit after do_nodes starts
+	# to let runmultiop_bg_pause start muliop and
+	# update /tmp/multiop_bg.pid ;
+	# The rm /tmp/multiop_bg.pid guarantees here that
+	# we have the updated by runmultiop_bg_pause
+	# /tmp/multiop_bg.pid file
+
+	local pid_file=$TMP/multiop_bg.pid.$$
+
+	do_node $client "MULTIOP_PID_FILE=$pid_file LUSTRE= \
+			runmultiop_bg_pause $file $cmds" &
+	local pid=$!
+	local multiop_pid
+
+	while [[ $wait_time -lt $WAIT_MAX ]]; do
+		sleep 3
+		wait_time=$((wait_time + 3))
+		multiop_pid=$(do_node $client cat $pid_file)
+		if [ -n "$multiop_pid" ]; then
+			break
+		fi
+	done
+
+	[ -n "$multiop_pid" ] ||
+		error "$client : Can not get multiop_pid from $pid_file "
+
+	eval export $(node_var_name $client)_multiop_pid=$multiop_pid
+	eval export $(node_var_name $client)_do_node_pid=$pid
+	local var=$(node_var_name $client)_multiop_pid
+	echo client $client multiop_bg started multiop_pid=${!var}
+	return $?
+}
+
+rmultiop_stop() {
+	local client=$1
+	local multiop_pid=$(node_var_name $client)_multiop_pid
+	local do_node_pid=$(node_var_name $client)_do_node_pid
+
+	echo "Stopping multiop_pid=${!multiop_pid} (kill ${!multiop_pid} on $client)"
+	do_node $client kill -USR1 ${!multiop_pid}
+
+	wait ${!do_node_pid}
+}
+
+sleep_maxage() {
+	local delay=$(do_facet $SINGLEMDS lctl get_param -n lo[vd].*.qos_maxage |
+		      awk '{ print $1 * 2; exit; }')
+	sleep $delay
+}
+
+check_component_count() {
+	local comp_cnt=$($LFS getstripe --component-count $1)
+	[ $comp_cnt -eq $2 ] || error "$1, component count $comp_cnt != $2"
+}
+
+# Verify there are no init components with "extension" flag
+verify_no_init_extension() {
+	local flg_opts="--component-flags init,extension"
+	local found=$($LFS find $flg_opts $1 | wc -l)
+	[ $found -eq 0 ] || error "$1 has component with initialized extension"
+}
+
+# Verify there is at least one component starting at 0
+verify_comp_at_zero() {
+	flg_opts="--component-flags init"
+	found=$($LFS find --component-start 0M $flg_opts $1 | wc -l)
+	[ $found -eq 1 ] ||
+		error "No component starting at zero(!)"
+}
+
+#TODO: This version is a placeholder, to be replaced before final commit
+SEL_VER="2.12.52"
+
+sel_layout_sanity() {
+	local file=$1
+	local comp_cnt=$2
+
+	verify_no_init_extension $file
+	verify_comp_at_zero $file
+	check_component_count $file $comp_cnt
+}
+