X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;ds=sidebyside;f=lustre%2Ftests%2Ftest-framework.sh;h=b7ef7f4185c29b74c70e6c8d33c270b038630ef5;hb=0a338970c2c73e14cc9be65d360de85be28ca488;hp=d26cb56c44c1912c55892f91daf986d1a534c188;hpb=7936ec229c7f62cfbf5a0e4ff5933149967e7c8f;p=fs%2Flustre-release.git

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index d26cb56..b7ef7f4 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -6,11 +6,9 @@ set -e
 #set -x
 
 export LANG=en_US
-export EJOURNAL=${EJOURNAL:-""}
 export REFORMAT=${REFORMAT:-""}
 export WRITECONF=${WRITECONF:-""}
 export VERBOSE=${VERBOSE:-false}
-export CATASTROPHE=${CATASTROPHE:-/proc/sys/lnet/catastrophe}
 export GSS=false
 export GSS_KRB5=false
 export GSS_PIPEFS=false
@@ -25,6 +23,11 @@ export LOAD_LLOOP=${LOAD_LLOOP:-false}
 
 #export PDSH="pdsh -S -Rssh -w"
 export MOUNT_CMD=${MOUNT_CMD:-"mount -t lustre"}
+export UMOUNT=${UMOUNT:-"umount -d"}
+# sles12 umount has a issue with -d option
+[ -e /etc/SuSE-release ] && grep -w VERSION /etc/SuSE-release | grep -wq 12 && {
+	export UMOUNT="umount"
+}
 
 # function used by scripts run on remote nodes
 LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
@@ -51,16 +54,25 @@ fi
 [ -z "$MODPROBECONF" -a -f /etc/modprobe.conf ] &&
     MODPROBECONF=/etc/modprobe.conf
 
+sanitize_parameters() {
+	for i in DIR DIR1 DIR2 MOUNT MOUNT1 MOUNT2
+	do
+		local path=${!i}
+		if [ -d "$path" ]; then
+			eval export $i=$(echo $path | sed -E 's/\/+$//g')
+		fi
+	done
+}
 assert_DIR () {
-    local failed=""
-    [[ $DIR/ = $MOUNT/* ]] || \
-        { failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
-    [[ $DIR1/ = $MOUNT1/* ]] || \
-        { failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
-    [[ $DIR2/ = $MOUNT2/* ]] || \
-        { failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
+	local failed=""
+	[[ $DIR/ = $MOUNT/* ]] ||
+		{ failed=1 && echo "DIR=$DIR not in $MOUNT. Aborting."; }
+	[[ $DIR1/ = $MOUNT1/* ]] ||
+		{ failed=1 && echo "DIR1=$DIR1 not in $MOUNT1. Aborting."; }
+	[[ $DIR2/ = $MOUNT2/* ]] ||
+		{ failed=1 && echo "DIR2=$DIR2 not in $MOUNT2. Aborting"; }
 
-    [ -n "$failed" ] && exit 99 || true
+	[ -n "$failed" ] && exit 99 || true
 }
 
 usage() {
@@ -72,7 +84,7 @@ usage() {
 
 print_summary () {
     trap 0
-    [ "$TESTSUITE" == "lfsck" ] && return 0
+	[ -z "$DEFAULT_SUITES"] && return 0
     [ -n "$ONLY" ] && echo "WARNING: ONLY is set to $(echo $ONLY)"
     local details
     local form="%-13s %-17s %-9s %s %s\n"
@@ -194,9 +206,8 @@ init_test_env() {
 		fi
 	fi
 
-    export LFSCK_BIN=${LFSCK_BIN:-lfsck}
-    export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after each test suite
-    export FSCK_MAX_ERR=4   # File system errors left uncorrected
+	export LFSCK_ALWAYS=${LFSCK_ALWAYS:-"no"} # check fs after test suite
+	export FSCK_MAX_ERR=4   # File system errors left uncorrected
 
 	export ZFS=${ZFS:-zfs}
 	export ZPOOL=${ZPOOL:-zpool}
@@ -226,9 +237,11 @@ init_test_env() {
     [ ! -f "$LST" ] && export LST=$(which lst)
     export SGPDDSURVEY=${SGPDDSURVEY:-"$LUSTRE/../lustre-iokit/sgpdd-survey/sgpdd-survey")}
     [ ! -f "$SGPDDSURVEY" ] && export SGPDDSURVEY=$(which sgpdd-survey)
+	export MCREATE=${MCREATE:-mcreate}
     # Ubuntu, at least, has a truncate command in /usr/bin
     # so fully path our truncate command.
     export TRUNCATE=${TRUNCATE:-$LUSTRE/tests/truncate}
+	export FSX=${FSX:-$LUSTRE/tests/fsx}
     export MDSRATE=${MDSRATE:-"$LUSTRE/tests/mpi/mdsrate"}
     [ ! -f "$MDSRATE" ] && export MDSRATE=$(which mdsrate 2> /dev/null)
     if ! echo $PATH | grep -q $LUSTRE/tests/racer; then
@@ -267,6 +280,9 @@ init_test_env() {
     export LFS_MIGRATE=${LFS_MIGRATE:-$LUSTRE/scripts/lfs_migrate}
     [ ! -f "$LFS_MIGRATE" ] &&
         export LFS_MIGRATE=$(which lfs_migrate 2> /dev/null)
+    export LR_READER=${LR_READER:-"$LUSTRE/utils/lr_reader"}
+    [ ! -f "$LR_READER" ] && export LR_READER=$(which lr_reader 2> /dev/null)
+    [ -z "$LR_READER" ] && export LR_READER="/usr/sbin/lr_reader"
     export NAME=${NAME:-local}
     export LGSSD=${LGSSD:-"$LUSTRE/utils/gss/lgssd"}
     [ "$GSS_PIPEFS" = "true" ] && [ ! -f "$LGSSD" ] && \
@@ -338,6 +354,8 @@ init_test_env() {
 	if ! $RPC_MODE; then
 		rm -f $TMP/*active
 	fi
+
+	export TF_FAIL=${TF_FAIL:-$TMP/tf.fail}
 }
 
 check_cpt_number() {
@@ -354,23 +372,62 @@ check_cpt_number() {
 	fi
 }
 
+# Return a numeric version code based on a version string.  The version
+# code is useful for comparison two version strings to see which is newer.
 version_code() {
-    # split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
-    eval set -- $(tr "[:punct:]" " " <<< $*)
+	# split arguments like "1.8.6-wc3" into "1", "8", "6", "wc3"
+	eval set -- $(tr "[:punct:]" " " <<< $*)
 
-    echo -n "$((($1 << 16) | ($2 << 8) | $3))"
+	echo -n "$((($1 << 16) | ($2 << 8) | $3))"
 }
 
 export LINUX_VERSION=$(uname -r | sed -e "s/\([0-9]*\.[0-9]*\.[0-9]*\).*/\1/")
 export LINUX_VERSION_CODE=$(version_code ${LINUX_VERSION//\./ })
 
+# Report the Lustre build version string (e.g. 1.8.7.3 or 2.4.1).
+#
+# usage: lustre_build_version
+#
+# All Lustre versions support "lctl get_param" to report the version of the
+# code running in the kernel (what our tests are interested in), but it
+# doesn't work without modules loaded.  If that fails, use "lctl version"
+# instead, which is easy to parse and works without the kernel modules,
+# but was only added in 2.6.50.  If that also fails, fall back to calling
+# "lctl lustre_build_version" which prints either (or both) the userspace
+# and kernel build versions, but is deprecated and should eventually be
+# removed.
+#
+# output: prints version string to stdout in dotted-decimal format
+lustre_build_version() {
+	local facet=${1:-client}
+
+	# lustre: 2.6.52
+	# kernel: patchless_client
+	# build: v2_6_92_0-gadb3ee4-2.6.32-431.29.2.el6_lustre.x86_64
+	local VER=$(do_facet $facet $LCTL get_param -n version 2> /dev/null |
+		    awk '/lustre: / { print $2 }')
+	# lctl 2.6.50
+	[ -z "$VER" ] && VER=$(do_facet $facet $LCTL --version 2>/dev/null |
+			       awk '{ print $2 }')
+	# Lustre version: 2.5.3-gfcfd782-CHANGED-2.6.32.26-175.fc12.x86_64
+	# lctl   version: 2.5.3-gfcfd782-CHANGED-2.6.32.26-175.fc12.x86_64
+	[ -z "$VER" ] && VER=$(do_facet $facet $LCTL lustre_build_version |
+			       awk '/version:/ { print $3; exit; }')
+	sed -e 's/^v//' -e 's/-.*//' -e 's/_/./g' <<<$VER
+}
+
+# Report the Lustre numeric build version code for the supplied facet.
+lustre_version_code() {
+	version_code $(lustre_build_version $1)
+}
+
 module_loaded () {
-   /sbin/lsmod | grep -q "^\<$1\>"
+	/sbin/lsmod | grep -q "^\<$1\>"
 }
 
 # Load a module on the system where this is running.
 #
-# Synopsis: load_module module_name [module arguments for insmod/modprobe]
+# usage: load_module module_name [module arguments for insmod/modprobe]
 #
 # If module arguments are not given but MODOPTS_<MODULE> is set, then its value
 # will be used as the arguments.  Otherwise arguments will be obtained from
@@ -381,12 +438,12 @@ load_module() {
     EXT=".ko"
     module=$1
     shift
-    BASE=`basename $module $EXT`
+    BASE=$(basename $module $EXT)
 
     module_loaded ${BASE} && return
 
-    # If no module arguments were passed, get them from $MODOPTS_<MODULE>, else from
-    # modprobe.conf
+    # If no module arguments were passed, get them from $MODOPTS_<MODULE>,
+    # else from modprobe.conf
     if [ $# -eq 0 ]; then
         # $MODOPTS_<MODULE>; we could use associative arrays, but that's not in
         # Bash until 4.x, so we resort to eval.
@@ -416,22 +473,24 @@ load_module() {
 
     [ $# -gt 0 ] && echo "${module} options: '$*'"
 
-    # Note that insmod will ignore anything in modprobe.conf, which is why we're
-    # passing options on the command-line.
-    if [ "$BASE" == "lnet_selftest" ] && \
-            [ -f ${LUSTRE}/../lnet/selftest/${module}${EXT} ]; then
-        insmod ${LUSTRE}/../lnet/selftest/${module}${EXT}
-    elif [ -f ${LUSTRE}/${module}${EXT} ]; then
-        insmod ${LUSTRE}/${module}${EXT} "$@"
-    else
-        # must be testing a "make install" or "rpm" installation
-        # note failed to load ptlrpc_gss is considered not fatal
-        if [ "$BASE" == "ptlrpc_gss" ]; then
-            modprobe $BASE "$@" 2>/dev/null || echo "gss/krb5 is not supported"
-        else
-            modprobe $BASE "$@"
-        fi
-    fi
+	# Note that insmod will ignore anything in modprobe.conf, which is why
+	# we're passing options on the command-line.
+	if [[ "$BASE" == "lnet_selftest" ]] &&
+		[[ -f ${LUSTRE}/../lnet/selftest/${module}${EXT} ]]; then
+		insmod ${LUSTRE}/../lnet/selftest/${module}${EXT}
+	elif [[ -f ${LUSTRE}/${module}${EXT} ]]; then
+		[[ "$BASE" != "ptlrpc_gss" ]] || modprobe sunrpc
+		insmod ${LUSTRE}/${module}${EXT} "$@"
+	else
+		# must be testing a "make install" or "rpm" installation
+		# note failed to load ptlrpc_gss is considered not fatal
+		if [[ "$BASE" == "ptlrpc_gss" ]]; then
+			modprobe $BASE "$@" 2>/dev/null ||
+				echo "gss/krb5 is not supported"
+		else
+			modprobe $BASE "$@"
+		fi
+	fi
 }
 
 llite_lloop_enabled() {
@@ -490,6 +549,13 @@ load_modules_local() {
     [ "$PTLDEBUG" ] && lctl set_param debug="$PTLDEBUG"
     [ "$SUBSYSTEM" ] && lctl set_param subsystem_debug="${SUBSYSTEM# }"
     load_module ../lnet/lnet/lnet
+	case $NETTYPE in
+	o2ib)
+		LNETLND="o2iblnd/ko2iblnd"
+		;;
+	*)
+		;;
+	esac
     LNETLND=${LNETLND:-"socklnd/ksocklnd"}
     load_module ../lnet/klnds/$LNETLND
     load_module obdclass/obdclass
@@ -503,13 +569,15 @@ load_modules_local() {
     load_module lov/lov
     load_module mgc/mgc
     load_module obdecho/obdecho
-    if ! client_only; then
-        SYMLIST=/proc/kallsyms
-        grep -q crc16 $SYMLIST || { modprobe crc16 2>/dev/null || true; }
-        grep -q -w jbd $SYMLIST || { modprobe jbd 2>/dev/null || true; }
-        grep -q -w jbd2 $SYMLIST || { modprobe jbd2 2>/dev/null || true; }
+	if ! client_only; then
+		SYMLIST=/proc/kallsyms
+		grep -q crc16 $SYMLIST ||
+			{ modprobe crc16 2>/dev/null || true; }
+		grep -q -w jbd2 $SYMLIST ||
+			{ modprobe jbd2 2>/dev/null || true; }
 		load_module lfsck/lfsck
-		[ "$LQUOTA" != "no" ] && load_module quota/lquota $LQUOTAOPTS
+		[ "$LQUOTA" != "no" ] &&
+			load_module quota/lquota $LQUOTAOPTS
 		if [[ $(node_fstypes $HOSTNAME) == *zfs* ]]; then
 			modprobe zfs
 			load_module osd-zfs/osd_zfs
@@ -517,10 +585,11 @@ load_modules_local() {
 		if [[ $(node_fstypes $HOSTNAME) == *ldiskfs* ]]; then
 			grep -q exportfs_decode_fh $SYMLIST ||
 				{ modprobe exportfs 2> /dev/null || true; }
+			grep -q -w mbcache $SYMLIST ||
+				{ modprobe mbcache 2>/dev/null || true; }
 			load_module ../ldiskfs/ldiskfs
 			load_module osd-ldiskfs/osd_ldiskfs
 		fi
-		load_module nodemap/nodemap
 		load_module mgs/mgs
 		load_module mdd/mdd
 		load_module mdt/mdt
@@ -529,7 +598,7 @@ load_modules_local() {
 		load_module osp/osp
 		load_module ofd/ofd
 		load_module osp/osp
-    fi
+	fi
 
 	load_module llite/lustre
 	llite_lloop_enabled && load_module llite/llite_lloop
@@ -539,15 +608,27 @@ load_modules_local() {
 	$LCTL modules > $OGDB/ogdb-$HOSTNAME
 
 	# 'mount' doesn't look in $PATH, just sbin
-	local MOUNT_LUSTRE=$LUSTRE/utils/mount.lustre
-	if [ -f $MOUNT_LUSTRE ]; then
-		if grep -qe "/sbin/mount\.lustre " /proc/mounts; then
-			cmp $MOUNT_LUSTRE /sbin/mount.lustre ||
-				umount /sbin/mount.lustre
+	local mount_lustre=$LUSTRE/utils/mount.lustre
+	if [ -f $mount_lustre ]; then
+		local sbin_mount=$(readlink -f /sbin)/mount.lustre
+		if grep -qw "$sbin_mount" /proc/mounts; then
+			cmp -s $mount_lustre $sbin_mount || umount $sbin_mount
 		fi
-		if ! grep -qe "/sbin/mount\.lustre " /proc/mounts; then
-			[ ! -f /sbin/mount.lustre ] && touch /sbin/mount.lustre
-			mount --bind $MOUNT_LUSTRE /sbin/mount.lustre
+		if ! grep -qw "$sbin_mount" /proc/mounts; then
+			[ ! -f "$sbin_mount" ] && touch "$sbin_mount"
+			if [ ! -s "$sbin_mount" -a -w "$sbin_mount" ]; then
+				cat <<- EOF > "$sbin_mount"
+				#!/bin/sh
+				#STUB MARK
+				echo "This $sbin_mount just a mountpoint." 1>&2
+				echo "It is never supposed to be run." 1>&2
+				logger -p emerg -- "using stub $sbin_mount $@"
+				exit 1
+				EOF
+				chmod a+x $sbin_mount
+			fi
+			mount --bind $mount_lustre $sbin_mount ||
+				error "can't bind $mount_lustre to $sbin_mount"
 		fi
 	fi
 }
@@ -593,16 +674,17 @@ unload_modules() {
 		fi
 	fi
 
-    if grep -qe "/sbin/mount\.lustre" /proc/mounts; then
-        umount /sbin/mount.lustre || true
-        [ -w /sbin/mount.lustre -a ! -s /sbin/mount.lustre ] && \
-            rm -f /sbin/mount.lustre || true
-    fi
+	local sbin_mount=$(readlink -f /sbin)/mount.lustre
+	if grep -qe "$sbin_mount " /proc/mounts; then
+		umount $sbin_mount || true
+		[ -s $sbin_mount ] && ! grep -q "STUB MARK" $sbin_mount ||
+			rm -f $sbin_mount
+	fi
 
-    check_mem_leak || return 254
+	check_mem_leak || return 254
 
-    echo "modules unloaded."
-    return 0
+	echo "modules unloaded."
+	return 0
 }
 
 fs_log_size() {
@@ -718,7 +800,8 @@ init_gss() {
         fi
 
         if [ -n "$LGSS_KEYRING_DEBUG" ]; then
-            echo $LGSS_KEYRING_DEBUG > /proc/fs/lustre/sptlrpc/gss/lgss_keyring/debug_level
+		lctl set_param -n \
+		    sptlrpc.gss.lgss_keyring.debug_level=$LGSS_KEYRING_DEBUG
         fi
     fi
 }
@@ -747,7 +830,7 @@ facet_type() {
 facet_number() {
 	local facet=$1
 
-	if [ $facet == mgs ]; then
+	if [ $facet == mgs ] || [ $facet == client ]; then
 		return 1
 	fi
 
@@ -802,6 +885,25 @@ node_fstypes() {
 	echo -n $fstypes
 }
 
+facet_index() {
+	local facet=$1
+	local num=$(facet_number $facet)
+	local index
+
+	if [[ $(facet_type $facet) = OST ]]; then
+		index=OSTINDEX${num}
+		if [[ -n "${!index}" ]]; then
+			echo -n ${!index}
+			return
+		fi
+
+		index=${OST_INDICES[num - 1]}
+	fi
+
+	[[ -n "$index" ]] || index=$((num - 1))
+	echo -n $index
+}
+
 devicelabel() {
 	local facet=$1
 	local dev=$2
@@ -1027,7 +1129,7 @@ get_osd_param() {
 	local name=$3
 
 	do_nodes $nodes "$LCTL get_param -n obdfilter.$device.$name \
-		osd-*.$device.$name 2>&1" | grep -v 'Found no match'
+		osd-*.$device.$name 2>&1" | grep -v 'error:'
 }
 
 set_osd_param() {
@@ -1037,7 +1139,7 @@ set_osd_param() {
 	local value=$4
 
 	do_nodes $nodes "$LCTL set_param -n obdfilter.$device.$name=$value \
-		osd-*.$device.$name=$value 2>&1" | grep -v 'Found no match'
+		osd-*.$device.$name=$value 2>&1" | grep -v 'error:'
 }
 
 set_debug_size () {
@@ -1134,6 +1236,10 @@ mount_facet() {
 	local opt=${facet}_opt
 	local mntpt=$(facet_mntpt $facet)
 	local opts="${!opt} $@"
+	local fstype=$(facet_fstype $facet)
+	local devicelabel
+
+	module_loaded lustre || load_modules
 
 	if [ $(facet_fstype $facet) == ldiskfs ] &&
 	   ! do_facet $facet test -b ${!dev}; then
@@ -1145,6 +1251,16 @@ mount_facet() {
 		import_zpool $facet || return ${PIPESTATUS[0]}
 	fi
 
+	case $fstype in
+	ldiskfs)
+		devicelabel=$(do_facet ${facet} "$E2LABEL ${!dev}");;
+	zfs)
+		devicelabel=$(do_facet ${facet} "$ZFS get -H -o value \
+						lustre:svname ${!dev}");;
+	*)
+		error "unknown fstype!";;
+	esac
+
 	echo "Starting ${facet}: $opts ${!dev} $mntpt"
 	# for testing LU-482 error handling in mount_facets() and test_0a()
 	if [ -f $TMP/test-lu482-trigger ]; then
@@ -1154,34 +1270,68 @@ mount_facet() {
 		                   ${!dev} $mntpt"
 		RC=${PIPESTATUS[0]}
 	fi
+
 	if [ $RC -ne 0 ]; then
 		echo "Start of ${!dev} on ${facet} failed ${RC}"
-    else
-        set_default_debug_facet $facet
+		return $RC
+	fi
 
-		label=$(devicelabel ${facet} ${!dev})
-        [ -z "$label" ] && echo no label for ${!dev} && exit 1
-        eval export ${facet}_svc=${label}
-        echo Started ${label}
-    fi
-    return $RC
+	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 ${!dev} without service"
+	else
+
+		case $fstype in
+		ldiskfs)
+			wait_update_facet ${facet} "$E2LABEL ${!dev} \
+				2>/dev/null | grep -E ':[a-zA-Z]{3}[0-9]{4}'" \
+				"" || error "${!dev} failed to initialize!";;
+		zfs)
+			wait_update_facet ${facet} "$ZFS get -H -o value \
+				lustre:svname ${!dev} 2>/dev/null | \
+				grep -E ':[a-zA-Z]{3}[0-9]{4}'" "" ||
+				error "${!dev} failed to initialize!";;
+
+		*)
+			error "unknown fstype!";;
+		esac
+	fi
+
+	# commit the device label change to disk
+	if [[ $devicelabel =~ (:[a-zA-Z]{3}[0-9]{4}) ]]; then
+		do_facet $facet "sync; sync; sync"
+	fi
+
+
+	label=$(devicelabel ${facet} ${!dev})
+	[ -z "$label" ] && echo no label for ${!dev} && exit 1
+	eval export ${facet}_svc=${label}
+	echo Started ${label}
+
+	return $RC
 }
 
 # start facet device options
 start() {
-    local facet=$1
-    shift
-    local device=$1
-    shift
-    eval export ${facet}_dev=${device}
-    eval export ${facet}_opt=\"$@\"
+	local facet=$1
+	shift
+	local device=$1
+	shift
+	eval export ${facet}_dev=${device}
+	eval export ${facet}_opt=\"$@\"
 
-    local varname=${facet}failover_dev
-    if [ -n "${!varname}" ] ; then
-        eval export ${facet}failover_dev=${!varname}
-    else
-        eval export ${facet}failover_dev=$device
-    fi
+	local varname=${facet}failover_dev
+	if [ -n "${!varname}" ] ; then
+		eval export ${facet}failover_dev=${!varname}
+	else
+		eval export ${facet}failover_dev=$device
+	fi
 
 	local mntpt=$(facet_mntpt $facet)
 	do_facet ${facet} mkdir -p $mntpt
@@ -1209,7 +1359,7 @@ stop() {
     running=$(do_facet ${facet} "grep -c $mntpt' ' /proc/mounts") || true
     if [ ${running} -ne 0 ]; then
         echo "Stopping $mntpt (opts:$@) on $HOST"
-        do_facet ${facet} umount -d $@ $mntpt
+	do_facet ${facet} $UMOUNT $@ $mntpt
     fi
 
 	# umount should block, but we should wait for unrelated obd's
@@ -1262,70 +1412,6 @@ quota_type() {
 	return $rc
 }
 
-# XXX This function is kept for interoperability with old server (< 2.3.50),
-#     it should be removed whenever we drop the interoperability for such
-#     server.
-restore_quota_old() {
-	local mntpt=${1:-$MOUNT}
-	local quota_type=$(quota_type $FSNAME | grep MDT | cut -d "=" -f2)
-	if [ ! "$old_QUOTA_TYPE" ] ||
-		[ "$quota_type" = "$old_QUOTA_TYPE" ]; then
-		return
-	fi
-	quota_save_version $old_QUOTA_TYPE
-}
-
-# XXX This function is kept for interoperability with old server (< 2.3.50),
-#     it should be removed whenever we drop the interoperability for such
-#     server.
-setup_quota_old(){
-	local mntpt=$1
-
-	# no quota enforcement for now and accounting works out of the box
-	return
-
-    # We need save the original quota_type params, and restore them after testing
-
-    # Suppose that quota type the same on mds and ost
-    local quota_type=$(quota_type | grep MDT | cut -d "=" -f2)
-    [ ${PIPESTATUS[0]} -eq 0 ] || error "quota_type failed!"
-    echo "[HOST:$HOSTNAME] [old_quota_type:$quota_type] [new_quota_type:$QUOTA_TYPE]"
-    if [ "$quota_type" != "$QUOTA_TYPE" ]; then
-        export old_QUOTA_TYPE=$quota_type
-        quota_save_version $QUOTA_TYPE
-    else
-        qtype=$(tr -c -d "ug" <<< $QUOTA_TYPE)
-        $LFS quotacheck -$qtype $mntpt || error "quotacheck has failed for $type"
-    fi
-
-    local quota_usrs=$QUOTA_USERS
-
-    # get_filesystem_size
-    local disksz=$(lfs_df $mntpt | grep "summary"  | awk '{print $2}')
-    local blk_soft=$((disksz + 1024))
-    local blk_hard=$((blk_soft + blk_soft / 20)) # Go 5% over
-
-    local Inodes=$(lfs_df -i $mntpt | grep "summary"  | awk '{print $2}')
-    local i_soft=$Inodes
-    local i_hard=$((i_soft + i_soft / 20))
-
-    echo "Total disk size: $disksz  block-softlimit: $blk_soft block-hardlimit:
-        $blk_hard inode-softlimit: $i_soft inode-hardlimit: $i_hard"
-
-    local cmd
-    for usr in $quota_usrs; do
-        echo "Setting up quota on $HOSTNAME:$mntpt for $usr..."
-        for type in u g; do
-            cmd="$LFS setquota -$type $usr -b $blk_soft -B $blk_hard -i $i_soft -I $i_hard $mntpt"
-            echo "+ $cmd"
-            eval $cmd || error "$cmd FAILED!"
-        done
-        # display the quota status
-        echo "Quota settings for $usr : "
-        $LFS quota -v -u $usr $mntpt || true
-    done
-}
-
 # get mdt quota type
 mdt_quota_type() {
 	local varsvc=${SINGLEMDS}_svc
@@ -1343,11 +1429,6 @@ ost_quota_type() {
 
 # restore old quota type settings
 restore_quota() {
-	if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
-		restore_quota_old
-		return
-	fi
-
 	if [ "$old_MDT_QUOTA_TYPE" ]; then
 		do_facet mgs $LCTL conf_param \
 			$FSNAME.quota.mdt=$old_MDT_QUOTA_TYPE
@@ -1383,11 +1464,6 @@ mdt_free_inodes() {
 }
 
 setup_quota(){
-	if [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.3.50) ]; then
-		setup_quota_old $1
-		return
-	fi
-
 	local mntpt=$1
 
 	# save old quota type & set new quota type
@@ -1436,23 +1512,37 @@ setup_quota(){
 }
 
 zconf_mount() {
-    local client=$1
-    local mnt=$2
-    local OPTIONS=${3:-$MOUNTOPT}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
-        exit 1
-    fi
+	local client=$1
+	local mnt=$2
+	local opts=${3:-$MOUNT_OPTS}
+	opts=${opts:+-o $opts}
+	local flags=${4:-$MOUNT_FLAGS}
+
+	local device=$MGSNID:/$FSNAME$FILESET
+	if [ -z "$mnt" -o -z "$FSNAME" ]; then
+		echo "Bad mount command: opt=$flags $opts dev=$device " \
+		     "mnt=$mnt"
+		exit 1
+	fi
 
-    echo "Starting client: $client: $OPTIONS $device $mnt"
-    do_node $client mkdir -p $mnt
-    do_node $client $MOUNT_CMD $OPTIONS $device $mnt || return 1
+	echo "Starting client: $client: $flags $opts $device $mnt"
+	do_node $client mkdir -p $mnt
+	if [ -n "$FILESET" -a -z "$SKIP_FILESET" ];then
+		do_node $client $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+			$mnt || return 1
+		#disable FILESET if not supported
+		do_nodes $client lctl get_param -n \
+			mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+				device=$MGSNID:/$FSNAME
+		do_node $client mkdir -p $mnt/$FILESET
+		do_node $client "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+	fi
+	do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
 
-    set_default_debug_nodes $client
+	set_default_debug_nodes $client
 
-    return 0
+	return 0
 }
 
 zconf_umount() {
@@ -1542,34 +1632,50 @@ sanity_mount_check () {
 
 # mount clients if not mouted
 zconf_mount_clients() {
-    local clients=$1
-    local mnt=$2
-    local OPTIONS=${3:-$MOUNTOPT}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$OPTIONS dev=$device mnt=$mnt
-        exit 1
-    fi
+	local clients=$1
+	local mnt=$2
+	local opts=${3:-$MOUNT_OPTS}
+	opts=${opts:+-o $opts}
+	local flags=${4:-$MOUNT_FLAGS}
+
+	local device=$MGSNID:/$FSNAME$FILESET
+	if [ -z "$mnt" -o -z "$FSNAME" ]; then
+		echo "Bad conf mount command: opt=$flags $opts dev=$device " \
+		     "mnt=$mnt"
+		exit 1
+	fi
 
-    echo "Starting client $clients: $OPTIONS $device $mnt"
+	echo "Starting client $clients: $flags $opts $device $mnt"
+	if [ -n "$FILESET" -a ! -n "$SKIP_FILESET" ]; then
+		do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+		do_nodes $clients $MOUNT_CMD $flags $opts $MGSNID:/$FSNAME \
+			$mnt || return 1
+		#disable FILESET if not supported
+		do_nodes $clients lctl get_param -n \
+			mdc.$FSNAME-MDT0000*.import | grep -q subtree ||
+				device=$MGSNID:/$FSNAME
+		do_nodes $clients mkdir -p $mnt/$FILESET
+		do_nodes $clients "! grep -q $mnt' ' /proc/mounts ||
+			umount $mnt"
+	fi
 
-    do_nodes $clients "
+	do_nodes $clients "
 running=\\\$(mount | grep -c $mnt' ');
 rc=0;
 if [ \\\$running -eq 0 ] ; then
     mkdir -p $mnt;
-    $MOUNT_CMD $OPTIONS $device $mnt;
+    $MOUNT_CMD $flags $opts $device $mnt;
     rc=\\\$?;
 fi;
 exit \\\$rc" || return ${PIPESTATUS[0]}
 
-    echo "Started clients $clients: "
-    do_nodes $clients "mount | grep $mnt' '"
+	echo "Started clients $clients: "
+	do_nodes $clients "mount | grep $mnt' '"
 
-    set_default_debug_nodes $clients
+	set_default_debug_nodes $clients
 
-    return 0
+	return 0
 }
 
 zconf_umount_clients() {
@@ -1942,8 +2048,12 @@ stop_client_loads() {
 
 # verify that lustre actually cleaned up properly
 cleanup_check() {
-	[ -f "$CATASTROPHE" ] && [[ $(< $CATASTROPHE) -ne 0 ]] &&
-		error "LBUG/LASSERT detected"
+	VAR=$(lctl get_param -n catastrophe 2>&1)
+	if [ $? = 0 ] ; then
+		if [ $VAR != 0 ]; then
+			error "LBUG/LASSERT detected"
+		fi
+	fi
 	BUSY=$(dmesg | grep -i destruct || true)
 	if [ -n "$BUSY" ]; then
 		echo "$BUSY" 1>&2
@@ -2007,21 +2117,26 @@ wait_update () {
 }
 
 wait_update_facet() {
+	local verbose=
+	[ "$1" = "--verbose" ] && verbose="$1" && shift
+
 	local facet=$1
 	shift
-	wait_update $(facet_active_host $facet) "$@"
+	wait_update $verbose $(facet_active_host $facet) "$@"
 }
 
 sync_all_data() {
 	do_nodes $(comma_list $(mdts_nodes)) \
-	    "lctl set_param -n osd*.*MDT*.force_sync 1"
+	    "lctl set_param -n osd*.*MDT*.force_sync=1"
 	do_nodes $(comma_list $(osts_nodes)) \
-	    "lctl set_param -n osd*.*OS*.force_sync 1" 2>&1 |
+	    "lctl set_param -n osd*.*OS*.force_sync=1" 2>&1 |
 		grep -v 'Found no match'
 }
 
 wait_delete_completed_mds() {
 	local MAX_WAIT=${1:-20}
+	# for ZFS, waiting more time for DMUs to be committed
+	local ZFS_WAIT=${2:-5}
 	local mds2sync=""
 	local stime=$(date +%s)
 	local etime
@@ -2057,6 +2172,14 @@ wait_delete_completed_mds() {
 		if [[ $changes -eq 0 ]]; then
 			etime=$(date +%s)
 			#echo "delete took $((etime - stime)) seconds"
+
+			# the occupied disk space will be released
+			# only after DMUs are committed
+			if [[ $(facet_fstype $SINGLEMDS) == zfs ]]; then
+				echo "sleep $ZFS_WAIT for ZFS OSD"
+				sleep $ZFS_WAIT
+			fi
+
 			return
 		fi
 		sleep 1
@@ -2322,6 +2445,13 @@ affected_facets () {
 }
 
 facet_failover() {
+	local E2FSCK_ON_MDT0=false
+	if [ "$1" == "--fsck" ]; then
+		shift
+		[ $(facet_fstype $SINGLEMDS) == ldiskfs ] &&
+			E2FSCK_ON_MDT0=true
+	fi
+
 	local facets=$1
 	local sleep_time=$2
 	local -a affecteds
@@ -2355,6 +2485,9 @@ facet_failover() {
 		shutdown_facet $facet
 	done
 
+	$E2FSCK_ON_MDT0 && (run_e2fsck $(facet_active_host $SINGLEMDS) \
+		$(mdsdevname 1) "-n" || error "Running e2fsck")
+
 	for ((index=0; index<$total; index++)); do
 		facet=$(echo ${affecteds[index]} | tr -s " " | cut -d"," -f 1)
 		echo reboot facets: ${affecteds[index]}
@@ -2619,7 +2752,7 @@ hostlist_expand() {
     done
     myList="${myList%* }";
 
-    # We can select an object at a offset in the list
+    # We can select an object at an offset in the list
     [ $# -eq 2 ] && {
         cnt=0
         for item in $myList; do
@@ -2697,13 +2830,7 @@ facet_active() {
 }
 
 facet_active_host() {
-    local facet=$1
-    local active=`facet_active $facet`
-    if [ "$facet" == client ]; then
-        echo $HOSTNAME
-    else
-        echo `facet_host $active`
-    fi
+	facet_host $(facet_active $1)
 }
 
 # Get the passive failover partner host of facet.
@@ -2813,7 +2940,7 @@ get_env_vars() {
 	local facet
 
 	for var in ${!MODOPTS_*}; do
-		value=${!var}
+		value=${!var//\"/\\\"}
 		echo -n " ${var}=\"$value\""
 	done
 
@@ -2830,9 +2957,21 @@ get_env_vars() {
 		fi
 	done
 
+	for var in VERBOSE; do
+		if [ -n "${!var}" ]; then
+			echo -n " $var=${!var}"
+		fi
+	done
+
 	if [ -n "$FSTYPE" ]; then
 		echo -n " FSTYPE=$FSTYPE"
 	fi
+
+	for var in LNETLND NETTYPE; do
+		if [ -n "${!var}" ]; then
+			echo -n " $var=${!var}"
+		fi
+	done
 }
 
 do_nodes() {
@@ -2877,12 +3016,19 @@ do_nodes() {
     return ${PIPESTATUS[0]}
 }
 
+##
+# Execute commands on a single service's host
+#
+# The \a facet (service) may be on a local or remote node, which is
+# determined at the time the command is run.
+#
+# usage: do_facet $facet command [arg ...]
 do_facet() {
-    local facet=$1
-    shift
-    local HOST=`facet_active_host $facet`
-    [ -z $HOST ] && echo No host defined for facet ${facet} && exit 1
-    do_node $HOST "$@"
+	local facet=$1
+	shift
+	local HOST=$(facet_active_host $facet)
+	[ -z $HOST ] && echo "No host defined for facet ${facet}" && exit 1
+	do_node $HOST "$@"
 }
 
 # Function: do_facet_random_file $FACET $FILE $SIZE
@@ -2967,7 +3113,7 @@ ostvdevname() {
 			eval VDEVPTR="";;
 		zfs )
 			#if $OSTDEVn isn't defined, default is $OSTDEVBASE{n}
-			# Device formated by zfs
+			# Device formatted by zfs
 			DEVNAME=OSTDEV$num
 			eval VDEVPTR=${!DEVNAME:=${OSTDEVBASE}${num}};;
 		* )
@@ -2977,7 +3123,7 @@ ostvdevname() {
 	echo -n $VDEVPTR
 }
 
-# Logical device formated for lustre
+# Logical device formatted for lustre
 mdsdevname() {
 	local num=$1
 	local DEVNAME=MDSDEV$num
@@ -3011,7 +3157,7 @@ mdsvdevname() {
 			eval VDEVPTR="";;
 		zfs )
 			# if $MDSDEVn isn't defined, default is $MDSDEVBASE{n}
-			# Device formated by ZFS
+			# Device formatted by ZFS
 			local DEVNAME=MDSDEV$num
 			eval VDEVPTR=${!DEVNAME:=${MDSDEVBASE}${num}};;
 		* )
@@ -3099,7 +3245,7 @@ unmount_ldiskfs() {
 	local dev=$(facet_device $facet)
 	local mnt=$(facet_mntpt $facet)
 
-	do_facet $facet umount -d $mnt
+	do_facet $facet $UMOUNT $mnt
 }
 
 var_name() {
@@ -3210,6 +3356,7 @@ cleanup_echo_devs () {
 
 cleanupall() {
     nfs_client_mode && return
+	cifs_client_mode && return
 
     stopall $*
     cleanup_echo_devs
@@ -3236,17 +3383,13 @@ mkfs_opts() {
 	local dev=$2
 	local fsname=${3:-"$FSNAME"}
 	local type=$(facet_type $facet)
-	local index=$(($(facet_number $facet) - 1))
+	local index=$(facet_index $facet)
 	local fstype=$(facet_fstype $facet)
 	local host=$(facet_host $facet)
 	local opts
 	local fs_mkfs_opts
 	local var
 
-	if [ $type == MGS ] && combined_mgs_mds; then
-		return 1
-	fi
-
 	if [ $type == MGS ] || ( [ $type == MDS ] &&
                                  [ "$dev" == $(mgsdevname) ] &&
 				 [ "$host" == "$(facet_host mgs)" ] ); then
@@ -3282,9 +3425,11 @@ mkfs_opts() {
 				fs_mkfs_opts+="-O large_xattr"
 			fi
 
-			fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
-			if [ ! -z $EJOURNAL ]; then
-				fs_mkfs_opts+=${MDSJOURNALSIZE:+" device=$EJOURNAL"}
+			var=${facet}_JRN
+			if [ -n "${!var}" ]; then
+				fs_mkfs_opts+=" -J device=${!var}"
+			else
+				fs_mkfs_opts+=${MDSJOURNALSIZE:+" -J size=$MDSJOURNALSIZE"}
 			fi
 			fs_mkfs_opts+=${MDSISIZE:+" -i $MDSISIZE"}
 		fi
@@ -3295,7 +3440,12 @@ mkfs_opts() {
 		opts+=${OSSCAPA:+" --param=ost.capa=$OSSCAPA"}
 
 		if [ $fstype == ldiskfs ]; then
-			fs_mkfs_opts+=${OSTJOURNALSIZE:+" -J size=$OSTJOURNALSIZE"}
+			var=${facet}_JRN
+			if [ -n "${!var}" ]; then
+				fs_mkfs_opts+=" -J device=${!var}"
+			else
+				fs_mkfs_opts+=${OSTJOURNALSIZE:+" -J size=$OSTJOURNALSIZE"}
+			fi
 		fi
 	fi
 
@@ -3322,37 +3472,81 @@ mkfs_opts() {
 	echo -n "$opts"
 }
 
-formatall() {
+check_ost_indices() {
+	local index_count=${#OST_INDICES[@]}
+	[[ $index_count -eq 0 || $OSTCOUNT -le $index_count ]] && return 0
+
+	# OST count is greater than the index count in $OST_INDEX_LIST.
+	# We need check whether there are duplicate indices.
+	local i
+	local j
+	local index
+	for i in $(seq $((index_count + 1)) $OSTCOUNT); do
+		index=$(facet_index ost$i)
+		for j in $(seq 0 $((index_count - 1))); do
+			[[ $index -ne ${OST_INDICES[j]} ]] ||
+			error "ost$i has the same index $index as ost$((j+1))"
+		done
+	done
+}
+
+format_mgs() {
+	local quiet
+
+	if ! $VERBOSE; then
+		quiet=yes
+	fi
+	echo "Format mgs: $(mgsdevname)"
+	reformat_external_journal mgs
+	add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
+		$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} || exit 10
+}
+
+format_mdt() {
+	local num=$1
 	local quiet
 
 	if ! $VERBOSE; then
 		quiet=yes
 	fi
+	echo "Format mds$num: $(mdsdevname $num)"
+	reformat_external_journal mds$num
+	add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
+		--reformat $(mdsdevname $num) $(mdsvdevname $num) \
+		${quiet:+>/dev/null} || exit 10
+}
+
+format_ost() {
+	local num=$1
+
+	if ! $VERBOSE; then
+		quiet=yes
+	fi
+	echo "Format ost$num: $(ostdevname $num)"
+	reformat_external_journal ost$num
+	add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
+		--reformat $(ostdevname $num) $(ostvdevname ${num}) \
+		${quiet:+>/dev/null} || exit 10
+}
 
+formatall() {
 	stopall
 	# We need ldiskfs here, may as well load them all
 	load_modules
 	[ "$CLIENTONLY" ] && return
 	echo Formatting mgs, mds, osts
 	if ! combined_mgs_mds ; then
-		echo "Format mgs: $(mgsdevname)"
-		add mgs $(mkfs_opts mgs $(mgsdevname)) --reformat \
-			$(mgsdevname) $(mgsvdevname) ${quiet:+>/dev/null} ||
-			exit 10
+		format_mgs
 	fi
 
 	for num in $(seq $MDSCOUNT); do
-		echo "Format mds$num: $(mdsdevname $num)"
-		add mds$num $(mkfs_opts mds$num $(mdsdevname ${num})) \
-			--reformat $(mdsdevname $num) $(mdsvdevname $num) \
-			${quiet:+>/dev/null} || exit 10
+		format_mdt $num
 	done
 
+	export OST_INDICES=($(hostlist_expand "$OST_INDEX_LIST"))
+	check_ost_indices
 	for num in $(seq $OSTCOUNT); do
-		echo "Format ost$num: $(ostdevname $num)"
-		add ost$num $(mkfs_opts ost$num $(ostdevname ${num})) \
-			--reformat $(ostdevname $num) $(ostvdevname ${num}) \
-			${quiet:+>/dev/null} || exit 10
+		format_ost $num
 	done
 }
 
@@ -3430,6 +3624,7 @@ writeconf_all () {
 
 setupall() {
     nfs_client_mode && return
+	cifs_client_mode && return
 
     sanity_mount_check ||
         error "environments are insane!"
@@ -3695,6 +3890,11 @@ nfs_client_mode () {
     return 1
 }
 
+cifs_client_mode () {
+	[ x$CIFSCLIENT = xyes ] &&
+		echo "CIFSCLIENT=$CIFSCLIENT mode: setup, cleanup, check config skipped"
+}
+
 check_config_client () {
     local mntpt=$1
 
@@ -3742,6 +3942,7 @@ check_config_clients () {
 	local mntpt=$1
 
 	nfs_client_mode && return
+	cifs_client_mode && return
 
 	do_rpc_nodes "$clients" check_config_client $mntpt
 
@@ -3787,11 +3988,13 @@ is_empty_fs() {
 }
 
 check_and_setup_lustre() {
-    nfs_client_mode && return
+	sanitize_parameters
+	nfs_client_mode && return
+	cifs_client_mode && return
 
-    local MOUNTED=$(mounted_lustre_filesystems)
+	local MOUNTED=$(mounted_lustre_filesystems)
 
-    local do_check=true
+	local do_check=true
     # 1.
     # both MOUNT and MOUNT2 are not mounted
     if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
@@ -4000,96 +4203,92 @@ check_shared_dir() {
 	return 0
 }
 
-# Run e2fsck on MDT and OST(s) to generate databases used for lfsck.
-generate_db() {
-	local i
-	local ostidx
-	local dev
+run_lfsck() {
+	do_nodes $(comma_list $(mdts_nodes) $(osts_nodes)) \
+		$LCTL set_param printk=+lfsck
+	do_facet $SINGLEMDS "$LCTL lfsck_start -M $FSNAME-MDT0000 -r -A -t all"
+
+	for k in $(seq $MDSCOUNT); do
+		# wait up to 10+1 minutes for LFSCK to complete
+		wait_update_facet --verbose mds${k} "$LCTL get_param -n \
+			mdd.$(facet_svc mds${k}).lfsck_layout |
+			awk '/^status/ { print \\\$2 }'" "completed" 600 ||
+			error "MDS${k} layout isn't the expected 'completed'"
+		wait_update_facet --verbose mds${k} "$LCTL get_param -n \
+			mdd.$(facet_svc mds${k}).lfsck_namespace |
+			awk '/^status/ { print \\\$2 }'" "completed" 60 ||
+			error "MDS${k} namespace isn't the expected 'completed'"
+	done
+	local rep_mdt=$(do_nodes $(comma_list $(mdts_nodes)) \
+			$LCTL get_param -n mdd.$FSNAME-*.lfsck_* |
+			awk '/repaired/ { print $2 }' | calc_sum)
+	local rep_ost=$(do_nodes $(comma_list $(osts_nodes)) \
+			$LCTL get_param -n obdfilter.$FSNAME-*.lfsck_* |
+			awk '/repaired/ { print $2 }' | calc_sum)
+	local repaired=$((rep_mdt + rep_ost))
+	[ $repaired -eq 0 ] ||
+		error "lfsck repaired $rep_mdt MDT and $rep_ost OST errors"
+}
+
+dump_file_contents() {
+	local nodes=$1
+	local dir=$2
+	local logname=$3
 	local node
 
-	[[ $(lustre_version_code $SINGLEMDS) -ne $(version_code 2.2.0) ]] ||
-		{ skip "Lustre 2.2.0 lacks the patch for LU-1255"; exit 0; }
-
-	check_shared_dir $SHARED_DIRECTORY ||
-		error "$SHARED_DIRECTORY isn't a shared directory"
-
-	export MDSDB=$SHARED_DIRECTORY/mdsdb
-	export OSTDB=$SHARED_DIRECTORY/ostdb
-
-	# DNE is not supported, so when running e2fsck on a DNE filesystem,
-	# we only pass master MDS parameters.
-	run_e2fsck $MDTNODE $MDTDEV "-n --mdsdb $MDSDB"
-
-    i=0
-    ostidx=0
-    OSTDB_LIST=""
-    for node in $(osts_nodes); do
-        for dev in ${OSTDEVS[i]}; do
-            run_e2fsck $node $dev "-n --mdsdb $MDSDB --ostdb $OSTDB-$ostidx"
-            OSTDB_LIST="$OSTDB_LIST $OSTDB-$ostidx"
-            ostidx=$((ostidx + 1))
-        done
-        i=$((i + 1))
-    done
+	if [ -z "$nodes" -o -z "$dir" -o -z "$logname" ]; then
+		error_noexit false \
+			"Invalid parameters for dump_file_contents()"
+		return 1
+	fi
+	for node in ${nodes}; do
+		do_node $node "for i in \\\$(find $dir -type f); do
+				echo ====\\\${i}=======================;
+				cat \\\${i};
+				done" >> ${logname}.${node}.log
+	done
 }
 
-# Run lfsck on server node if lfsck can't be found on client (LU-2571)
-run_lfsck_remote() {
-	local cmd="$LFSCK_BIN -c -l --mdsdb $MDSDB --ostdb $OSTDB_LIST $MOUNT"
-	local client=$1
-	local mounted=true
-	local rc=0
+dump_command_output() {
+	local nodes=$1
+	local cmd=$2
+	local logname=$3
+	local node
 
-	#Check if lustre is already mounted
-	do_rpc_nodes $client is_mounted $MOUNT || mounted=false
-	if ! $mounted; then
-		zconf_mount $client $MOUNT ||
-			error "failed to mount Lustre on $client"
-	fi
-	#Run lfsck
-	echo $cmd
-	do_node $client $cmd || rc=$?
-	#Umount if necessary
-	if ! $mounted; then
-		zconf_umount $client $MOUNT ||
-			error "failed to unmount Lustre on $client"
+	if [ -z "$nodes" -o -z "$cmd" -o -z "$logname" ]; then
+		error_noexit false \
+			"Invalid parameters for dump_command_output()"
+		return 1
 	fi
 
-	[ $rc -le $FSCK_MAX_ERR ] ||
-		error "$cmd returned $rc, should be <= $FSCK_MAX_ERR"
-	echo "lfsck finished with rc=$rc"
-
-	return $rc
+	for node in ${nodes}; do
+		do_node $node "echo ====${cmd}=======================;
+				$cmd" >> ${logname}.${node}.log
+	done
 }
 
-run_lfsck() {
-	local facets="client $SINGLEMDS"
-	local found=false
-	local facet
-	local node
-	local rc=0
-
-	for facet in $facets; do
-		node=$(facet_active_host $facet)
-		if check_progs_installed $node $LFSCK_BIN; then
-			found=true
-			break
-		fi
-	done
-	! $found && error "None of \"$facets\" supports lfsck"
+log_zfs_info() {
+	local logname=$1
 
-	run_lfsck_remote $node || rc=$?
+	# dump file contents from /proc/spl in case of zfs test
+	if [ "$(facet_fstype ost1)" = "zfs" ]; then
+		dump_file_contents "$(osts_nodes)" "/proc/spl" "${logname}"
+		dump_command_output \
+			"$(osts_nodes)" "zpool events -v" "${logname}"
+	fi
 
-	rm -rvf $MDSDB* $OSTDB* || true
-	return $rc
+	if [ "$(facet_fstype $SINGLEMDS)" = "zfs" ]; then
+		dump_file_contents "$(mdts_nodes)" "/proc/spl" "${logname}"
+		dump_command_output \
+			"$(mdts_nodes)" "zpool events -v" "${logname}"
+	fi
 }
 
 check_and_cleanup_lustre() {
-    if [ "$LFSCK_ALWAYS" = "yes" -a "$TESTSUITE" != "lfsck" ]; then
-        get_svr_devs
-        generate_db
-        run_lfsck
-    fi
+	if [ "$LFSCK_ALWAYS" = "yes" -a "$TESTSUITE" != "sanity-lfsck" -a \
+	     "$TESTSUITE" != "sanity-scrub" ]; then
+		run_lfsck
+	fi
 
 	if is_mounted $MOUNT; then
 		[ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
@@ -4109,10 +4308,6 @@ check_and_cleanup_lustre() {
 		cleanupall -f || error "cleanup failed"
 		unset I_MOUNTED
 	fi
-
-	if grep -qe "/sbin/mount\.lustre " /proc/mounts; then
-		umount /sbin/mount.lustre
-	fi
 }
 
 #######
@@ -4310,24 +4505,24 @@ drop_request() {
 
 drop_reply() {
 # OBD_FAIL_MDS_ALL_REPLY_NET
-    RC=0
-    do_facet $SINGLEMDS lctl set_param fail_loc=0x122
-    do_facet client "$@" || RC=$?
-    do_facet $SINGLEMDS lctl set_param fail_loc=0
-    return $RC
+	RC=0
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0x122
+	eval "$@" || RC=$?
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+	return $RC
 }
 
 drop_reint_reply() {
 # OBD_FAIL_MDS_REINT_NET_REP
-    RC=0
-    do_facet $SINGLEMDS lctl set_param fail_loc=0x119
-    do_facet client "$@" || RC=$?
-    do_facet $SINGLEMDS lctl set_param fail_loc=0
-    return $RC
-}
+	RC=0
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0x119
+	eval "$@" || RC=$?
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+	return $RC
+}
 
 drop_update_reply() {
-# OBD_FAIL_UPDATE_OBJ_NET_REP
+# OBD_FAIL_OUT_UPDATE_NET_REP
 	local index=$1
 	shift 1
 	RC=0
@@ -4363,21 +4558,51 @@ drop_ldlm_cancel() {
 	return $RC
 }
 
-drop_bl_callback() {
+drop_bl_callback_once() {
+	local rc=0
+	do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=0
 #define OBD_FAIL_LDLM_BL_CALLBACK_NET			0x305
-	RC=0
 	do_facet client lctl set_param fail_loc=0x80000305
-	do_facet client "$@" || RC=$?
+	do_facet client "$@" || rc=$?
 	do_facet client lctl set_param fail_loc=0
-	return $RC
+	do_facet client lctl set_param fail_val=0
+	do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
+	return $rc
+}
+
+drop_bl_callback() {
+	rc=0
+	do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=0
+#define OBD_FAIL_LDLM_BL_CALLBACK_NET			0x305
+	do_facet client lctl set_param fail_loc=0x305
+	do_facet client "$@" || rc=$?
+	do_facet client lctl set_param fail_loc=0
+	do_facet client lctl set_param fail_val=0
+	do_facet client lctl set_param ldlm.namespaces.*.early_lock_cancel=1
+	return $rc
 }
 
 drop_ldlm_reply() {
 #define OBD_FAIL_LDLM_REPLY              0x30c
     RC=0
-    do_facet $SINGLEMDS lctl set_param fail_loc=0x30c
+    local list=$(comma_list $(mdts_nodes) $(osts_nodes))
+    do_nodes $list lctl set_param fail_loc=0x30c
+
+    do_facet client "$@" || RC=$?
+
+    do_nodes $list lctl set_param fail_loc=0
+    return $RC
+}
+
+drop_ldlm_reply_once() {
+#define OBD_FAIL_LDLM_REPLY              0x30c
+    RC=0
+    local list=$(comma_list $(mdts_nodes) $(osts_nodes))
+    do_nodes $list lctl set_param fail_loc=0x8000030c
+
     do_facet client "$@" || RC=$?
-    do_facet $SINGLEMDS lctl set_param fail_loc=0
+
+    do_nodes $list lctl set_param fail_loc=0
     return $RC
 }
 
@@ -4394,12 +4619,10 @@ set_nodes_failloc () {
 }
 
 cancel_lru_locks() {
-    $LCTL mark "cancel_lru_locks $1 start"
-    for d in `lctl get_param -N ldlm.namespaces.*.lru_size | egrep -i $1`; do
-        $LCTL set_param -n $d=clear
-    done
-    $LCTL get_param ldlm.namespaces.*.lock_unused_count | egrep -i $1 | grep -v '=0'
-    $LCTL mark "cancel_lru_locks $1 stop"
+	#$LCTL mark "cancel_lru_locks $1 start"
+	$LCTL set_param -n ldlm.namespaces.*$1*.lru_size=clear
+	$LCTL get_param ldlm.namespaces.*$1*.lock_unused_count | grep -v '=0'
+	#$LCTL mark "cancel_lru_locks $1 stop"
 }
 
 default_lru_size()
@@ -4475,21 +4698,18 @@ stop_full_debug_logging() {
 }
 
 # prints bash call stack
-log_trace_dump() {
+print_stack_trace() {
+	local skip=${1:-1}
 	echo "  Trace dump:"
-	for (( i=1; i < ${#BASH_LINENO[*]} ; i++ )) ; do
-		local s=${BASH_SOURCE[$i]}
-		local l=${BASH_LINENO[$i-1]}
-		local f=${FUNCNAME[$i]}
-		echo "  = $s:$l:$f()"
+	for (( i=$skip; i < ${#BASH_LINENO[*]} ; i++ )) ; do
+		local src=${BASH_SOURCE[$i]}
+		local lineno=${BASH_LINENO[$i-1]}
+		local funcname=${FUNCNAME[$i]}
+		echo "  = $src:$lineno:$funcname()"
 	done
 }
 
-##################################
-# Test interface
-##################################
-
-error_noexit() {
+report_error() {
 	local TYPE=${TYPE:-"FAIL"}
 
 	local dump=true
@@ -4499,10 +4719,8 @@ error_noexit() {
 		dump=false
 	fi
 
-
 	log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
-	log_trace_dump
-
+	(print_stack_trace 2) >&2
 	mkdir -p $LOGDIR
 	# We need to dump the logs on all nodes
 	if $dump; then
@@ -4521,6 +4739,17 @@ error_noexit() {
 			echo "$@" > $LOGDIR/err
 		fi
 	fi
+
+	# cleanup the env for failed tests
+	reset_fail_loc
+}
+
+##################################
+# Test interface
+##################################
+
+error_noexit() {
+	report_error "$@"
 }
 
 exit_status () {
@@ -4532,12 +4761,13 @@ exit_status () {
 }
 
 error() {
-	error_noexit "$@"
+	report_error "$@"
 	exit 1
 }
 
 error_exit() {
-	error "$@"
+	report_error "$@"
+	exit 1
 }
 
 # use only if we are ignoring failures for this test, bugno required.
@@ -4547,11 +4777,11 @@ error_exit() {
 error_ignore() {
 	local TYPE="IGNORE ($1)"
 	shift
-	error_noexit "$@"
+	report_error "$@"
 }
 
 error_and_remount() {
-	error_noexit "$@"
+	report_error "$@"
 	remount_client $MOUNT
 	exit 1
 }
@@ -4687,8 +4917,8 @@ run_test() {
 }
 
 log() {
-    echo "$*"
-    module_loaded lnet || load_modules
+	echo "$*" >&2
+	load_module ../libcfs/libcfs/libcfs
 
     local MSG="$*"
     # Get rid of '
@@ -4742,7 +4972,8 @@ check_mds() {
 
 reset_fail_loc () {
     echo -n "Resetting fail_loc on all nodes..."
-    do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 2>/dev/null || true"
+    do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 \
+	    fail_val=0 2>/dev/null || true"
     echo done.
 }
 
@@ -4762,6 +4993,18 @@ banner() {
     log "$msg== $(date +"%H:%M:%S (%s)")"
 }
 
+check_dmesg_for_errors() {
+	local res
+	local errors="VFS: Busy inodes after unmount of\|\
+ldiskfs_check_descriptors: Checksum for group 0 failed\|\
+group descriptors corrupted"
+
+	res=$(do_nodes $(comma_list $(nodes_list)) "dmesg" | grep "$errors")
+	[ -z "$res" ] && return 0
+	echo "Kernel error detected: $res"
+	return 1
+}
+
 #
 # Run a single test function and cleanup after it.
 #
@@ -4777,12 +5020,17 @@ run_one() {
 	local SAVE_UMASK=`umask`
 	umask 0022
 
+	if ! grep -q $DIR /proc/mounts; then
+		$SETUP
+	fi
+
 	banner "test $testnum: $message"
 	test_${testnum} || error "test_$testnum failed with $?"
 	cd $SAVE_PWD
 	reset_fail_loc
 	check_grant ${testnum} || error "check_grant $testnum failed with $?"
 	check_catastrophe || error "LBUG/LASSERT detected"
+	check_dmesg_for_errors || error "Error in dmesg detected"
 	if [ "$PARALLEL" != "yes" ]; then
 		ps auxww | grep -v grep | grep -q multiop &&
 					error "multiop still running"
@@ -4791,6 +5039,7 @@ run_one() {
 	unset tdir
 	unset tfile
 	umask $SAVE_UMASK
+	$CLEANUP
 	return 0
 }
 
@@ -4805,6 +5054,8 @@ run_one_logged() {
 	local TEST_ERROR
 	local name=${TESTSUITE}.test_${1}.test_log.$(hostname -s).log
 	local test_log=$LOGDIR/$name
+	local zfs_log_name=${TESTSUITE}.test_${1}.zfs_log
+	local zfs_debug_log=$LOGDIR/$zfs_log_name
 	rm -rf $LOGDIR/err
 	rm -rf $LOGDIR/ignore
 	rm -rf $LOGDIR/skip
@@ -4836,6 +5087,7 @@ run_one_logged() {
 	fi
 
 	if [ -f $LOGDIR/err ]; then
+		log_zfs_info "$zfs_debug_log"
 		$FAIL_ON_ERROR && exit $RC
 	fi
 
@@ -5015,6 +5267,7 @@ require_dsh_ost()
 
 remote_mgs_nodsh()
 {
+	[ "$CLIENTONLY" ] && return 0 || true
     local MGS 
     MGS=$(facet_host mgs)
     remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
@@ -5478,50 +5731,15 @@ restore_lustre_params() {
 }
 
 check_catastrophe() {
-	local rnodes=${1:-$(comma_list $(remote_nodes_list))}
-	local C=$CATASTROPHE
-	[ -f $C ] && [ $(cat $C) -ne 0 ] && return 1
-
-	[ -z "$rnodes" ] && return 0
-
-	local data
-	data=$(do_nodes "$rnodes" "rc=\\\$([ -f $C ] &&
-		echo \\\$(< $C) || echo 0);
-		if [ \\\$rc -ne 0 ]; then echo \\\$(hostname): \\\$rc; fi
-		exit \\\$rc")
-	local rc=$?
-	if [ -n "$data" ]; then
-	    echo $data
-	    return $rc
-	fi
-	return 0
-}
-
-# CMD: determine mds index where directory inode presents
-get_mds_dir () {
-    local dir=$1
-    local file=$dir/f0.get_mds_dir_tmpfile
-
-    mkdir -p $dir
-    rm -f $file
-    sleep 1
-    local iused=$(lfs df -i $dir | grep MDT | awk '{print $3}')
-    local -a oldused=($iused)
+	local nodes=${1:-$(comma_list $(nodes_list))}
 
-    openfile -f O_CREAT:O_LOV_DELAY_CREATE -m 0644 $file > /dev/null
-    sleep 1
-    iused=$(lfs df -i $dir | grep MDT | awk '{print $3}')
-    local -a newused=($iused)
-
-    local num=0
-    for ((i=0; i<${#newused[@]}; i++)); do
-         if [ ${oldused[$i]} -lt ${newused[$i]} ];  then
-             echo $(( i + 1 ))
-             rm -f $file
-             return 0
-         fi
-    done
-    error "mdt-s : inodes count OLD ${oldused[@]} NEW ${newused[@]}"
+	do_nodes $nodes "rc=0;
+val=\\\$($LCTL get_param -n catastrophe 2>&1);
+if [[ \\\$? -eq 0 && \\\$val -ne 0 ]]; then
+	echo \\\$(hostname -s): \\\$val;
+	rc=\\\$val;
+fi;
+exit \\\$rc"
 }
 
 mdsrate_cleanup () {
@@ -5556,17 +5774,7 @@ convert_facet2label() {
 }
 
 get_clientosc_proc_path() {
-    echo "${1}-osc-*"
-}
-
-get_lustre_version () {
-    local facet=${1:-"$SINGLEMDS"}    
-    do_facet $facet $LCTL get_param -n version | awk '/^lustre:/ {print $2}'
-}
-
-lustre_version_code() {
-    local facet=${1:-"$SINGLEMDS"}
-    version_code $(get_lustre_version $1)
+	echo "${1}-osc-*"
 }
 
 # If the 2.0 MDS was mounted on 1.8 device, then the OSC and LOV names
@@ -5584,19 +5792,21 @@ mds_on_old_device() {
 }
 
 get_mdtosc_proc_path() {
-    local mds_facet=$1
-    local ost_label=${2:-"*OST*"}
-
-    [ "$mds_facet" = "mds" ] && mds_facet=$SINGLEMDS
-    local mdt_label=$(convert_facet2label $mds_facet)
-    local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
-
-    if [ $(lustre_version_code $mds_facet) -le $(version_code 1.8.0) ] ||
-       mds_on_old_device $mds_facet; then
-        echo "${ost_label}-osc"
-    else
-        echo "${ost_label}-osc-${mdt_index}"
-    fi
+	local mds_facet=$1
+	local ost_label=${2:-"*OST*"}
+
+	[ "$mds_facet" = "mds" ] && mds_facet=$SINGLEMDS
+	local mdt_label=$(convert_facet2label $mds_facet)
+	local mdt_index=$(echo $mdt_label | sed -e 's/^.*-//')
+
+	if [ $(lustre_version_code $mds_facet) -le $(version_code 1.8.0) ] ||
+	   mds_on_old_device $mds_facet; then
+		echo "${ost_label}-osc"
+	elif [[ $ost_label = *OST* ]]; then
+		echo "${ost_label}-osc-${mdt_index}"
+	else
+		echo "${ost_label}-osp-${mdt_index}"
+	fi
 }
 
 get_osc_import_name() {
@@ -5617,6 +5827,7 @@ _wait_import_state () {
     local expected=$1
     local CONN_PROC=$2
     local maxtime=${3:-$(max_recovery_time)}
+    local error_on_failure=${4:-1}
     local CONN_STATE
     local i=0
 
@@ -5629,9 +5840,12 @@ _wait_import_state () {
             # reconnect timeout and test can't see real disconnect
             [ "${CONN_STATE}" == "CONNECTING" ] && return 0
         fi
-        [ $i -ge $maxtime ] && \
-            error "can't put import for $CONN_PROC into ${expected} state after $i sec, have ${CONN_STATE}" && \
+	if [ $i -ge $maxtime ]; then
+	    [ $error_on_failure -ne 0 ] && \
+		error "can't put import for $CONN_PROC into ${expected}" \
+		      "state after $i sec, have ${CONN_STATE}"
             return 1
+	fi
         sleep 1
 	# Add uniq for multi-mount case
 	CONN_STATE=$($LCTL get_param -n $CONN_PROC 2>/dev/null | cut -f2 | uniq)
@@ -5646,10 +5860,11 @@ wait_import_state() {
     local state=$1
     local params=$2
     local maxtime=${3:-$(max_recovery_time)}
+    local error_on_failure=${4:-1}
     local param
 
     for param in ${params//,/ }; do
-        _wait_import_state $state $param $maxtime || return
+	_wait_import_state $state $param $maxtime $error_on_failure || return
     done
 }
 
@@ -5705,8 +5920,8 @@ _wait_osc_import_state() {
 	local facet=$1
 	local ost_facet=$2
 	local expected=$3
-	local ost=$(get_osc_import_name $facet $ost_facet)
-	local param="osc.${ost}.ost_server_uuid"
+	local target=$(get_osc_import_name $facet $ost_facet)
+	local param="osc.${target}.ost_server_uuid"
 	local params=$param
 	local i=0
 
@@ -5730,6 +5945,16 @@ _wait_osc_import_state() {
 			params=$($LCTL list_param $param 2>/dev/null || true)
 		done
 	fi
+
+	if [[ $ost_facet = mds* ]]; then
+		# no OSP connection to itself
+		if [[ $facet = $ost_facet ]]; then
+			return 0
+		fi
+		param="osp.${target}.mdt_server_uuid"
+		params=$param
+	fi
+
 	if ! do_rpc_nodes "$(facet_active_host $facet)" \
 			wait_import_state $expected "$params" $maxtime; then
 		error "import is not in ${expected} state"
@@ -5754,10 +5979,81 @@ wait_osc_import_state() {
 	fi
 }
 
+_wait_mgc_import_state() {
+	local facet=$1
+	local expected=$2
+	local error_on_failure=${3:-1}
+	local param="mgc.*.mgs_server_uuid"
+	local params=$param
+	local i=0
+
+	# 1. wait the deadline of client 1st request (it could be skipped)
+	# 2. wait the deadline of client 2nd request
+	local maxtime=$(( 2 * $(request_timeout $facet)))
+
+	if [[ $facet == client* ]]; then
+		# During setup time, the osc might not be setup, it need wait
+		# until list_param can return valid value. And also if there
+		# are mulitple osc entries we should list all of them before
+		# go to wait.
+		params=$($LCTL list_param $param 2>/dev/null || true)
+		while [ -z "$params" ]; do
+			if [ $i -ge $maxtime ]; then
+				echo "can't get $param in $maxtime secs"
+				return 1
+			fi
+			sleep 1
+			i=$((i + 1))
+			params=$($LCTL list_param $param 2>/dev/null || true)
+		done
+	fi
+	if ! do_rpc_nodes "$(facet_active_host $facet)" \
+			wait_import_state $expected "$params" $maxtime \
+					  $error_on_failure; then
+		if [ $error_on_failure -ne 0 ]; then
+		    error "import is not in ${expected} state"
+		fi
+		return 1
+	fi
+
+	return 0
+}
+
+wait_mgc_import_state() {
+	local facet=$1
+	local expected=$2
+	local error_on_failure=${3:-1}
+	local num
+
+	if [[ $facet = mds ]]; then
+		for num in $(seq $MDSCOUNT); do
+			_wait_mgc_import_state mds$num "$expected" \
+					       $error_on_failure || return
+		done
+	else
+		_wait_mgc_import_state "$facet" "$expected"
+				       $error_on_failure || return
+	fi
+}
+
+wait_dne_interconnect() {
+	local num
+
+	if [ $MDSCOUNT -gt 1 ]; then
+		for num in $(seq $MDSCOUNT); do
+			wait_osc_import_state mds mds$num FULL
+		done
+	fi
+}
+
 get_clientmdc_proc_path() {
     echo "${1}-mdc-*"
 }
 
+get_clientmgc_proc_path() {
+    echo "*"
+}
+
 do_rpc_nodes () {
 	local list=$1
 	shift
@@ -5772,28 +6068,34 @@ do_rpc_nodes () {
 }
 
 wait_clients_import_state () {
-    local list=$1
-    local facet=$2
-    local expected=$3
+	local list=$1
+	local facet=$2
+	local expected=$3
 
-    local facets=$facet
+	local facets=$facet
 
-    if [ "$FAILURE_MODE" = HARD ]; then
-        facets=$(facets_on_host $(facet_active_host $facet))
-    fi
+	if [ "$FAILURE_MODE" = HARD ]; then
+		facets=$(facets_on_host $(facet_active_host $facet))
+	fi
 
-    for facet in ${facets//,/ }; do
-    local label=$(convert_facet2label $facet)
-    local proc_path
-    case $facet in
-        ost* ) proc_path="osc.$(get_clientosc_proc_path $label).ost_server_uuid" ;;
-        mds* ) proc_path="mdc.$(get_clientmdc_proc_path $label).mds_server_uuid" ;;
-        *) error "unknown facet!" ;;
-    esac
-    local params=$(expand_list $params $proc_path)
-    done
+	for facet in ${facets//,/ }; do
+		local label=$(convert_facet2label $facet)
+		local proc_path
+		case $facet in
+		ost* ) proc_path="osc.$(get_clientosc_proc_path \
+				  $label).ost_server_uuid" ;;
+		mds* ) proc_path="mdc.$(get_clientmdc_proc_path \
+				  $label).mds_server_uuid" ;;
+		mgs* ) proc_path="mgc.$(get_clientmgc_proc_path \
+				  $label).mgs_server_uuid" ;;
+		*) error "unknown facet!" ;;
+		esac
+
+		local params=$(expand_list $params $proc_path)
+	done
 
-	if ! do_rpc_nodes "$list" wait_import_state_mount $expected $params; then
+	if ! do_rpc_nodes "$list" wait_import_state_mount $expected $params;
+	then
 		error "import is not in ${expected} state"
 		return 1
 	fi
@@ -5850,19 +6152,21 @@ create_pool() {
 }
 
 add_pool_to_list () {
-    local fsname=${1%%.*}
-    local poolname=${1##$fsname.}
+	local fsname=${1%%.*}
+	local poolname=${1##$fsname.}
 
-    local listvar=${fsname}_CREATED_POOLS
-    eval export ${listvar}=$(expand_list ${!listvar} $poolname)
+	local listvar=${fsname}_CREATED_POOLS
+	local temp=${listvar}=$(expand_list ${!listvar} $poolname)
+	eval export $temp
 }
 
 remove_pool_from_list () {
-    local fsname=${1%%.*}
-    local poolname=${1##$fsname.}
+	local fsname=${1%%.*}
+	local poolname=${1##$fsname.}
 
-    local listvar=${fsname}_CREATED_POOLS
-    eval export ${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+	local listvar=${fsname}_CREATED_POOLS
+	local temp=${listvar}=$(exclude_items_from_list ${!listvar} $poolname)
+	eval export $temp
 }
 
 destroy_pool_int() {
@@ -5980,22 +6284,48 @@ do_ls () {
     return $rc
 }
 
-# target_start_and_reset_recovery_timer()
-#        service_time = at_est2timeout(service_time);
-#        service_time += 2 * (CONNECTION_SWITCH_MAX + CONNECTION_SWITCH_INC +
-#                             INITIAL_CONNECT_TIMEOUT);
-# CONNECTION_SWITCH_MAX : min(25U, max(CONNECTION_SWITCH_MIN,obd_timeout))
-#define CONNECTION_SWITCH_INC 1
-#define INITIAL_CONNECT_TIMEOUT max(CONNECTION_SWITCH_MIN,obd_timeout/20)
-#define CONNECTION_SWITCH_MIN 5U
+# check_and_start_recovery_timer()
+#	service_time = at_est2timeout(service_time);
+#	service_time += 2 * INITIAL_CONNECT_TIMEOUT;
+#	service_time += 2 * (CONNECTION_SWITCH_MAX + CONNECTION_SWITCH_INC);
 
-max_recovery_time () {
-    local init_connect_timeout=$(( TIMEOUT / 20 ))
-    [[ $init_connect_timeout -ge 5 ]] || init_connect_timeout=5
+#define INITIAL_CONNECT_TIMEOUT max(CONNECTION_SWITCH_MIN, obd_timeout/20)
+#define CONNECTION_SWITCH_MAX min(50, max(CONNECTION_SWITCH_MIN, obd_timeout))
+#define CONNECTION_SWITCH_MIN 5
+#define CONNECTION_SWITCH_INC 5
+max_recovery_time() {
+	local init_connect_timeout=$((TIMEOUT / 20))
+	((init_connect_timeout >= 5)) || init_connect_timeout=5
 
-    local service_time=$(( $(at_max_get client) + $(( 2 * $(( 25 + 1  + init_connect_timeout)) )) ))
+	local service_time=$(($(at_max_get client) * 9 / 4 + 5))
+	service_time=$((service_time + 2 * (init_connect_timeout + 50 + 5)))
 
-    echo $service_time 
+	echo -n $service_time
+}
+
+recovery_time_min() {
+	local connection_switch_min=5
+	local connection_switch_inc=5
+	local connection_switch_max
+	local reconnect_delay_max
+	local initial_connect_timeout
+	local max
+	local timout_20
+
+	#connection_switch_max=min(50, max($connection_switch_min,$TIMEOUT)
+	(($connection_switch_min > $TIMEOUT)) &&
+		max=$connection_switch_min || max=$TIMEOUT
+	(($max < 50)) && connection_switch_max=$max || connection_switch_max=50
+
+	#initial_connect_timeout = max(connection_switch_min, obd_timeout/20)
+	timeout_20=$((TIMEOUT/20))
+	(($connection_switch_min > $timeout_20)) &&
+		initial_connect_timeout=$connection_switch_min ||
+		initial_connect_timeout=$timeout_20
+
+	reconnect_delay_max=$((connection_switch_max + connection_switch_inc + \
+			       initial_connect_timeout))
+	echo $((2 * reconnect_delay_max))
 }
 
 get_clients_mount_count () {
@@ -6624,12 +6954,14 @@ generate_string() {
 
 reformat_external_journal() {
 	local facet=$1
+	local var
 
-	if [ ! -z ${EJOURNAL} ]; then
+	var=${facet}_JRN
+	if [ -n "${!var}" ]; then
 		local rcmd="do_facet $facet"
 
-		echo "reformat external journal on $facet:${EJOURNAL}"
-		${rcmd} mke2fs -O journal_dev ${EJOURNAL} || return 1
+		echo "reformat external journal on $facet:${!var}"
+		${rcmd} mke2fs -O journal_dev ${!var} || return 1
 	fi
 }
 
@@ -6669,14 +7001,10 @@ mds_backup_restore() {
 	echo "backup data"
 	${rcmd} tar zcf $metadata -C $mntpt/ . > /dev/null 2>&1 || return 3
 	# step 6: umount
-	${rcmd} umount -d $mntpt || return 4
-	# step 7: reformat external journal if needed
-	reformat_external_journal $facet || return 5
+	${rcmd} $UMOUNT $mntpt || return 4
 	# step 8: reformat dev
 	echo "reformat new device"
-	add $facet $(mkfs_opts $facet ${devname}) --backfstype ldiskfs \
-		--reformat ${devname} $(mdsvdevname $(facet_number $facet)) \
-		> /dev/null || exit 6
+	format_mdt $(facet_number $facet)
 	# step 9: mount dev
 	${rcmd} mount -t ldiskfs $opts $devname $mntpt || return 7
 	# step 10: restore metadata
@@ -6689,7 +7017,7 @@ mds_backup_restore() {
 	echo "remove recovery logs"
 	${rcmd} rm -fv $mntpt/OBJECTS/* $mntpt/CATALOGS
 	# step 13: umount dev
-	${rcmd} umount -d $mntpt || return 10
+	${rcmd} $UMOUNT $mntpt || return 10
 	# step 14: cleanup tmp backup
 	${rcmd} rm -f $metaea $metadata
 	# step 15: reset device label - it's not virgin on
@@ -6729,7 +7057,7 @@ mds_remove_ois() {
 		done
 	fi
 	# step 4: umount
-	${rcmd} umount -d $mntpt || return 2
+	${rcmd} $UMOUNT $mntpt || return 2
 	# OI files will be recreated when mounted as lustre next time.
 }
 
@@ -6743,62 +7071,46 @@ generate_logname() {
 
 # make directory on different MDTs
 test_mkdir() {
-	local option
-	local parent
-	local child
 	local path
 	local p_option
-	local option2
 	local stripe_count=2
-	local rc=0
-
-	case $# in
-		1) path=$1;;
-		2) option=$1
-		   path=$2;;
-		3) option=$1
-		   option2=$2
-		   path=$3;;
-		*) error "Only creating single directory is supported";;
-	esac
-
-	child=$(basename $path)
-	parent=$(dirname $path)
-
-	if [ "$option" == "-p" -o "$option2" == "-p" ]; then
-		if [ -d $parent/$child ]; then
-			return $rc
-		fi
-		p_option="-p"
-	fi
+	local stripe_index=-1
+	local OPTIND=1
+
+	while getopts "c:i:p" opt; do
+		case $opt in
+			c) stripe_count=$OPTARG;;
+			i) stripe_index=$OPTARG;;
+			p) p_option="-p";;
+			\?) error "only support -i -c -p";;
+		esac
+	done
 
-	if [ "${option:0:2}" == "-c" ]; then
-		stripe_count=$(echo $option | sed 's/^-c//')
-	fi
+	shift $((OPTIND - 1))
+	[ $# -eq 1 ] || error "Only creating single directory is supported"
+	path="$*"
 
-	if [ "${option2:0:2}" == "-c" ]; then
-		stripe_count=$(echo $option2 | sed 's/^-c//')
-	fi
+	if [ "$p_option" == "-p" ]; then
+		local parent=$(dirname $path)
 
-	if [ ! -d ${parent} ]; then
-		if [ "$p_option" == "-p" ]; then
-			mkdir -p ${parent}
-		else
-			return 1
-		fi
+		[ -d $path ] && return 0
+		[ ! -d ${parent} ] && mkdir -p ${parent}
 	fi
 
 	if [ $MDSCOUNT -le 1 ]; then
-		mkdir $p_option $parent/$child || rc=$?
+		mkdir $path
 	else
-		local mdt_idx=$($LFS getstripe -M $parent)
 		local test_num=$(echo $testnum | sed -e 's/[^0-9]*//g')
+		local mdt_index
 
-		mdt_idx=$((test_num % MDSCOUNT))
-		echo "striped dir -i$mdt_idx -c$stripe_count $path"
-		$LFS setdirstripe -i$mdt_idx -c$stripe_count $path || rc=$?
+		if [ $stripe_index -eq -1 ]; then
+			mdt_index=$((test_num % MDSCOUNT))
+		else
+			mdt_index=$stripe_index
+		fi
+		echo "striped dir -i$mdt_index -c$stripe_count $path"
+		$LFS setdirstripe -i$mdt_index -c$stripe_count $path
 	fi
-	return $rc
 }
 
 # find the smallest and not in use file descriptor
@@ -6819,6 +7131,10 @@ check_mount_and_prep()
 
 	rm -rf $DIR/[df][0-9]* || error "Fail to cleanup the env!"
 	mkdir $DIR/$tdir || error "Fail to mkdir $DIR/$tdir."
+	for idx in $(seq $MDSCOUNT); do
+		local name="MDT$(printf '%04x' $((idx - 1)))"
+		rm -rf $MOUNT/.lustre/lost+found/$name/*
+	done
 }
 
 # calcule how many ost-objects to be created.
@@ -6829,10 +7145,320 @@ precreated_ost_obj_count()
 	local mdt_name="MDT$(printf '%04x' $mdt_idx)"
 	local ost_name="OST$(printf '%04x' $ost_idx)"
 	local proc_path="${FSNAME}-${ost_name}-osc-${mdt_name}"
-	local last_id=$(do_facet mds${mdt_idx} lctl get_param -n \
+	local last_id=$(do_facet mds$((mdt_idx + 1)) lctl get_param -n \
 			osp.$proc_path.prealloc_last_id)
-	local next_id=$(do_facet mds${mdt_idx} lctl get_param -n \
+	local next_id=$(do_facet mds$((mdt_idx + 1)) lctl get_param -n \
 			osp.$proc_path.prealloc_next_id)
-
 	echo $((last_id - next_id + 1))
 }
+
+check_file_in_pool()
+{
+	local file=$1
+	local pool=$2
+	local tlist="$3"
+	local res=$($GETSTRIPE $file | grep 0x | cut -f2)
+	for i in $res
+	do
+		for t in $tlist ; do
+			[ "$i" -eq "$t" ] && continue 2
+		done
+
+		echo "pool list: $tlist"
+		echo "striping: $res"
+		error_noexit "$file not allocated in $pool"
+		return 1
+	done
+	return 0
+}
+
+pool_add() {
+	echo "Creating new pool"
+	local pool=$1
+
+	create_pool $FSNAME.$pool ||
+		{ error_noexit "No pool created, result code $?"; return 1; }
+	[ $($LFS pool_list $FSNAME | grep -c "$FSNAME.${pool}\$") -eq 1 ] ||
+		{ error_noexit "$pool not in lfs pool_list"; return 2; }
+}
+
+pool_add_targets() {
+	echo "Adding targets to pool"
+	local pool=$1
+	local first=$2
+	local last=$3
+	local step=${4:-1}
+
+	local list=$(seq $first $step $last)
+
+	local t=$(for i in $list; do printf "$FSNAME-OST%04x_UUID " $i; done)
+	do_facet mgs $LCTL pool_add \
+			$FSNAME.$pool $FSNAME-OST[$first-$last/$step]
+	wait_update $HOSTNAME "lctl get_param -n lov.$FSNAME-*.pools.$pool \
+			| sort -u | tr '\n' ' ' " "$t" || {
+		error_noexit "Add to pool failed"
+		return 1
+	}
+	local lfscount=$($LFS pool_list $FSNAME.$pool | grep -c "\-OST")
+	local addcount=$(((last - first) / step + 1))
+	[ $lfscount -eq $addcount ] || {
+		error_noexit "lfs pool_list bad ost count" \
+						"$lfscount != $addcount"
+		return 2
+	}
+}
+
+pool_set_dir() {
+	local pool=$1
+	local tdir=$2
+	echo "Setting pool on directory $tdir"
+
+	$SETSTRIPE -c 2 -p $pool $tdir && return 0
+
+	error_noexit "Cannot set pool $pool to $tdir"
+	return 1
+}
+
+pool_check_dir() {
+	local pool=$1
+	local tdir=$2
+	echo "Checking pool on directory $tdir"
+
+	local res=$($GETSTRIPE --pool $tdir | sed "s/\s*$//")
+	[ "$res" = "$pool" ] && return 0
+
+	error_noexit "Pool on '$tdir' is '$res', not '$pool'"
+	return 1
+}
+
+pool_dir_rel_path() {
+	echo "Testing relative path works well"
+	local pool=$1
+	local tdir=$2
+	local root=$3
+
+	mkdir -p $root/$tdir/$tdir
+	cd $root/$tdir
+	pool_set_dir $pool $tdir          || return 1
+	pool_set_dir $pool ./$tdir        || return 2
+	pool_set_dir $pool ../$tdir       || return 3
+	pool_set_dir $pool ../$tdir/$tdir || return 4
+	rm -rf $tdir; cd - > /dev/null
+}
+
+pool_alloc_files() {
+	echo "Checking files allocation from directory pool"
+	local pool=$1
+	local tdir=$2
+	local count=$3
+	local tlist="$4"
+
+	local failed=0
+	for i in $(seq -w 1 $count)
+	do
+		local file=$tdir/file-$i
+		touch $file
+		check_file_in_pool $file $pool "$tlist" || \
+			failed=$((failed + 1))
+	done
+	[ "$failed" = 0 ] && return 0
+
+	error_noexit "$failed files not allocated in $pool"
+	return 1
+}
+
+pool_create_files() {
+	echo "Creating files in pool"
+	local pool=$1
+	local tdir=$2
+	local count=$3
+	local tlist="$4"
+
+	mkdir -p $tdir
+	local failed=0
+	for i in $(seq -w 1 $count)
+	do
+		local file=$tdir/spoo-$i
+		$SETSTRIPE -p $pool $file
+		check_file_in_pool $file $pool "$tlist" || \
+			failed=$((failed + 1))
+	done
+	[ "$failed" = 0 ] && return 0
+
+	error_noexit "$failed files not allocated in $pool"
+	return 1
+}
+
+pool_lfs_df() {
+	echo "Checking 'lfs df' output"
+	local pool=$1
+
+	local t=$($LCTL get_param -n lov.$FSNAME-clilov-*.pools.$pool |
+			tr '\n' ' ')
+	local res=$($LFS df --pool $FSNAME.$pool |
+			awk '{print $1}' |
+			grep "$FSNAME-OST" |
+			tr '\n' ' ')
+	[ "$res" = "$t" ] && return 0
+
+	error_noexit "Pools OSTs '$t' is not '$res' that lfs df reports"
+	return 1
+}
+
+pool_file_rel_path() {
+	echo "Creating files in a pool with relative pathname"
+	local pool=$1
+	local tdir=$2
+
+	mkdir -p $tdir ||
+		{ error_noexit "unable to create $tdir"; return 1 ; }
+	local file="/..$tdir/$tfile-1"
+	$SETSTRIPE -p $pool $file ||
+		{ error_noexit "unable to create $file" ; return 2 ; }
+
+	cd $tdir
+	$SETSTRIPE -p $pool $tfile-2 || {
+		error_noexit "unable to create $tfile-2 in $tdir"
+		return 3
+	}
+}
+
+pool_remove_first_target() {
+	echo "Removing first target from a pool"
+	local pool=$1
+
+	local pname="lov.$FSNAME-*.pools.$pool"
+	local t=$($LCTL get_param -n $pname | head -1)
+	do_facet mgs $LCTL pool_remove $FSNAME.$pool $t
+	wait_update $HOSTNAME "lctl get_param -n $pname | grep $t" "" || {
+		error_noexit "$t not removed from $FSNAME.$pool"
+		return 1
+	}
+}
+
+pool_remove_all_targets() {
+	echo "Removing all targets from pool"
+	local pool=$1
+	local file=$2
+	local pname="lov.$FSNAME-*.pools.$pool"
+	for t in $($LCTL get_param -n $pname | sort -u)
+	do
+		do_facet mgs $LCTL pool_remove $FSNAME.$pool $t
+	done
+	wait_update $HOSTNAME "lctl get_param -n $pname" "" || {
+		error_noexit "Pool $FSNAME.$pool cannot be drained"
+		return 1
+	}
+	# striping on an empty/nonexistant pool should fall back
+	# to "pool of everything"
+	touch $file || {
+		error_noexit "failed to use fallback striping for empty pool"
+		return 2
+	}
+	# setstripe on an empty pool should fail
+	$SETSTRIPE -p $pool $file 2>/dev/null && {
+		error_noexit "expected failure when creating file" \
+							"with empty pool"
+		return 3
+	}
+	return 0
+}
+
+pool_remove() {
+	echo "Destroying pool"
+	local pool=$1
+	local file=$2
+
+	do_facet mgs $LCTL pool_destroy $FSNAME.$pool
+
+	sleep 2
+	# striping on an empty/nonexistant pool should fall back
+	# to "pool of everything"
+	touch $file || {
+		error_noexit "failed to use fallback striping for missing pool"
+		return 1
+	}
+	# setstripe on an empty pool should fail
+	$SETSTRIPE -p $pool $file 2>/dev/null && {
+		error_noexit "expected failure when creating file" \
+							"with missing pool"
+		return 2
+	}
+
+	# get param should return err once pool is gone
+	if wait_update $HOSTNAME "lctl get_param -n \
+		lov.$FSNAME-*.pools.$pool 2>/dev/null || echo foo" "foo"
+	then
+		remove_pool_from_list $FSNAME.$pool
+		return 0
+	fi
+	error_noexit "Pool $FSNAME.$pool is not destroyed"
+	return 3
+}
+
+# Get and check the actual stripe count of one file.
+# Usage: check_stripe_count <file> <expected_stripe_count>
+check_stripe_count() {
+	local file=$1
+	local expected=$2
+	local actual
+
+	[[ -z "$file" || -z "$expected" ]] &&
+		error "check_stripe_count: invalid argument"
+
+	local cmd="$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"
+	fi
+}
+
+# Get and check the actual list of OST indices on one file.
+# Usage: check_obdidx <file> <expected_comma_separated_list_of_ost_indices>
+check_obdidx() {
+	local file=$1
+	local expected=$2
+	local obdidx
+
+	[[ -z "$file" || -z "$expected" ]] &&
+		error "check_obdidx: invalid argument!"
+
+	obdidx=$(comma_list $($GETSTRIPE $file | grep -A $OSTCOUNT obdidx |
+			      grep -v obdidx | awk '{print $1}' | xargs))
+
+	[[ $obdidx = $expected ]] ||
+		error "list of OST indices on $file is $obdidx," \
+		      "should be $expected"
+}
+
+# Get and check the actual OST index of the first stripe on one file.
+# Usage: check_start_ost_idx <file> <expected_start_ost_idx>
+check_start_ost_idx() {
+	local file=$1
+	local expected=$2
+	local start_ost_idx
+
+	[[ -z "$file" || -z "$expected" ]] &&
+		error "check_start_ost_idx: invalid argument!"
+
+	start_ost_idx=$($GETSTRIPE $file | grep -A 1 obdidx | grep -v obdidx |
+			awk '{print $1}')
+
+	[[ $start_ost_idx = $expected ]] ||
+		error "OST index of the first stripe on $file is" \
+		      "$start_ost_idx, should be $expected"
+}
+
+killall_process () {
+	local clients=${1:-$(hostname)}
+	local name=$2
+	local signal=$3
+	local rc=0
+
+	do_nodes $clients "killall $signal $name"
+}