X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Ftest-framework.sh;h=95ddf56ba0c3517068de209df404c978e7302bd5;hp=43b17e91cb47e313e5ce8a69d2139590ee25331f;hb=15dd813536ad06a119dfb2358f00281eed22a98b;hpb=85cbe1a3ee6940f7468884bca43cd736a5365694

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 43b17e9..95ddf56 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -54,16 +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 $path | sed -r 's/\/+$//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() {
@@ -133,6 +142,7 @@ init_test_env() {
 	export TEST_FAILED=false
 	export FAIL_ON_SKIP_ENV=${FAIL_ON_SKIP_ENV:-false}
 	export RPC_MODE=${RPC_MODE:-false}
+	export DO_CLEANUP=${DO_CLEANUP:-true}
 
     export MKE2FS=$MKE2FS
     if [ -z "$MKE2FS" ]; then
@@ -284,6 +294,8 @@ init_test_env() {
     export DIR2
     export SAVE_PWD=${SAVE_PWD:-$LUSTRE/tests}
     export AT_MAX_PATH
+    export LDEV=${LDEV:-"$LUSTRE/scripts/ldev"}
+    [ ! -f "$LDEV" ] && export LDEV=$(which ldev 2> /dev/null)
 
     if [ "$ACCEPTOR_PORT" ]; then
         export PORT_OPT="--port $ACCEPTOR_PORT"
@@ -324,20 +336,20 @@ init_test_env() {
 	export SHUTDOWN_ATTEMPTS=${SHUTDOWN_ATTEMPTS:-3}
 	export OSD_TRACK_DECLARES_LBUG=${OSD_TRACK_DECLARES_LBUG:-"yes"}
 
-    # command line
+	# command line
 
-    while getopts "rvwf:" opt $*; do
-        case $opt in
-            f) CONFIG=$OPTARG;;
-            r) REFORMAT=--reformat;;
-            v) VERBOSE=true;;
-            w) WRITECONF=writeconf;;
-            \?) usage;;
-        esac
-    done
+	while getopts "rvwf:" opt $*; do
+		case $opt in
+			f) CONFIG=$OPTARG;;
+			r) REFORMAT=yes;;
+			v) VERBOSE=true;;
+			w) WRITECONF=writeconf;;
+			\?) usage;;
+		esac
+	done
 
-    shift $((OPTIND - 1))
-    ONLY=${ONLY:-$*}
+	shift $((OPTIND - 1))
+	ONLY=${ONLY:-$*}
 
 	# print the durations of each test if "true"
 	DDETAILS=${DDETAILS:-false}
@@ -392,9 +404,7 @@ export LINUX_VERSION_CODE=$(version_code ${LINUX_VERSION//\./ })
 lustre_build_version() {
 	local facet=${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
+	# lustre: 2.8.52
 	local VER=$(do_facet $facet $LCTL get_param -n version 2> /dev/null |
 		    awk '/lustre: / { print $2 }')
 	# lctl 2.6.50
@@ -560,13 +570,15 @@ 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
@@ -574,6 +586,8 @@ 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
@@ -585,7 +599,7 @@ 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
@@ -686,6 +700,18 @@ fs_log_size() {
 	echo -n $size
 }
 
+fs_inode_ksize() {
+	local facet=${1:-$SINGLEMDS}
+	local fstype=$(facet_fstype $facet)
+	local size=0
+	case $fstype in
+		ldiskfs) size=4;;  # ~4KB per inode
+		zfs)     size=11;; # 10 to 11KB per inode
+	esac
+
+	echo -n $size
+}
+
 check_gss_daemon_nodes() {
     local list=$1
     dname=$2
@@ -817,7 +843,7 @@ facet_type() {
 facet_number() {
 	local facet=$1
 
-	if [ $facet == mgs ]; then
+	if [ $facet == mgs ] || [ $facet == client ]; then
 		return 1
 	fi
 
@@ -964,6 +990,19 @@ facet_vdevice() {
 	echo -n $device
 }
 
+running_in_vm() {
+	local virt=$(virt-what 2> /dev/null)
+
+	[ $? -eq 0 ] && [ -n "$virt" ] && { echo $virt; return; }
+
+	virt=$(dmidecode -s system-product-name | awk '{print $1}')
+
+	case $virt in
+		VMware|KVM|VirtualBox|Parallels) echo ${virt,,} ;;
+		*) ;;
+	esac
+}
+
 #
 # Re-read the partition table on failover partner host.
 # After a ZFS storage pool is created on a shared device, the partition table
@@ -1135,7 +1174,7 @@ set_debug_size () {
     if [ -f /sys/devices/system/cpu/possible ]; then
         local cpus=$(($(cut -d "-" -f 2 /sys/devices/system/cpu/possible)+1))
     else
-        local cpus=$(getconf _NPROCESSORS_CONF)
+        local cpus=$(getconf _NPROCESSORS_CONF 2>/dev/null)
     fi
 
     # bug 19944, adjust size to be -gt num_possible_cpus()
@@ -1177,6 +1216,15 @@ set_default_debug_facet () {
     set_default_debug_nodes $node
 }
 
+set_hostid () {
+    local hostid=${1:-$(hostid)}
+
+    if [ ! -s /etc/hostid ]; then
+	printf $(echo -n $hostid |
+	    sed 's/\(..\)\(..\)\(..\)\(..\)/\\x\4\\x\3\\x\2\\x\1/') >/etc/hostid
+    fi
+}
+
 # Facet functions
 mount_facets () {
 	local facets=${1:-$(get_facets)}
@@ -1223,6 +1271,8 @@ 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
 
@@ -1236,6 +1286,16 @@ 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
@@ -1251,6 +1311,11 @@ mount_facet() {
 		return $RC
 	fi
 
+	health=$(do_facet ${facet} "$LCTL get_param -n health_check")
+	if [[ "$health" != "healthy" ]]; then
+		error "$facet is in a unhealthy state"
+	fi
+
 	set_default_debug_facet $facet
 
 	if [[ $facet == mds* ]]; then
@@ -1261,7 +1326,6 @@ mount_facet() {
 	if [[ $opts =~ .*nosvc.* ]]; then
 		echo "Start ${!dev} without service"
 	else
-		local fstype=$(facet_fstype $facet)
 
 		case $fstype in
 		ldiskfs)
@@ -1279,6 +1343,12 @@ mount_facet() {
 		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}
@@ -1482,32 +1552,44 @@ setup_quota(){
 }
 
 zconf_mount() {
-    local client=$1
-    local mnt=$2
-    local opts=${3:-$MOUNT_OPTS}
-    opts=${opts:+-o $opts}
-    local flags=${4:-$MOUNT_FLAGS}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
-        exit 1
-    fi
-
-    echo "Starting client: $client: $flags $opts $device $mnt"
-    do_node $client mkdir -p $mnt
-    do_node $client $MOUNT_CMD $flags $opts $device $mnt || return 1
-
-    set_default_debug_nodes $client
+	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: $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
 
-    return 0
+	return 0
 }
 
 zconf_umount() {
     local client=$1
     local mnt=$2
     local force
-    local busy 
+    local busy
     local need_kill
 
     [ "$3" ] && force=-f
@@ -1548,13 +1630,13 @@ if [ \\\$running -ne \\\$mpts ]; then
     echo \\\$(hostname) env are INSANE!;
     exit 1;
 fi"
-    [ $? -eq 0 ] || rc=1 
+    [ $? -eq 0 ] || rc=1
     done
     return $rc
 }
 
 sanity_mount_check_servers () {
-    [ "$CLIENTONLY" ] && 
+    [ -n "$CLIENTONLY" ] &&
         { echo "CLIENTONLY mode, skip mount_check_servers"; return 0; } || true
     echo Checking servers environments
 
@@ -1590,21 +1672,35 @@ sanity_mount_check () {
 
 # mount clients if not mouted
 zconf_mount_clients() {
-    local clients=$1
-    local mnt=$2
-    local opts=${3:-$MOUNT_OPTS}
-    opts=${opts:+-o $opts}
-    local flags=${4:-$MOUNT_FLAGS}
-
-    local device=$MGSNID:/$FSNAME
-    if [ -z "$mnt" -o -z "$FSNAME" ]; then
-        echo Bad zconf mount command: opt=$flags $opts dev=$device mnt=$mnt
-        exit 1
-    fi
-
-    echo "Starting client $clients: $flags $opts $device $mnt"
-
-    do_nodes $clients "
+	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: $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 "
 running=\\\$(mount | grep -c $mnt' ');
 rc=0;
 if [ \\\$running -eq 0 ] ; then
@@ -1614,12 +1710,12 @@ if [ \\\$running -eq 0 ] ; then
 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() {
@@ -1662,7 +1758,7 @@ shutdown_node_hard () {
         wait_for_function --quiet "! ping -w 3 -c 1 $host" 5 1 && return 0
         echo "waiting for $host to fail attempts=$attempts"
         [ $i -lt $attempts ] || \
-            { echo "$host still pingable after power down! attempts=$attempts" && return 1; } 
+            { echo "$host still pingable after power down! attempts=$attempts" && return 1; }
     done
 }
 
@@ -2356,16 +2452,17 @@ client_evicted() {
 }
 
 client_reconnect_try() {
-    uname -n >> $MOUNT/recon
-    if [ -z "$CLIENTS" ]; then
-        df $MOUNT; uname -n >> $MOUNT/recon
-    else
-        do_nodes $CLIENTS "df $MOUNT; uname -n >> $MOUNT/recon" > /dev/null
-    fi
-    echo Connected clients:
-    cat $MOUNT/recon
-    ls -l $MOUNT/recon > /dev/null
-    rm $MOUNT/recon
+	local f=$MOUNT/recon
+
+	uname -n >> $f
+	if [ -z "$CLIENTS" ]; then
+		$LFS df $MOUNT; uname -n >> $f
+	else
+		do_nodes $CLIENTS "$LFS df $MOUNT; uname -n >> $f" > /dev/null
+	fi
+	echo "Connected clients: $(cat $f)"
+	ls -l $f > /dev/null
+	rm $f
 }
 
 client_reconnect() {
@@ -2460,7 +2557,7 @@ obd_name() {
 replay_barrier() {
 	local facet=$1
 	do_facet $facet "sync; sync; sync"
-	df $MOUNT
+	$LFS df $MOUNT
 
 	# make sure there will be no seq change
 	local clients=${CLIENTS:-$HOSTNAME}
@@ -2545,7 +2642,7 @@ fail() {
 
 	facet_failover $* || error "failover: $?"
 	wait_clients_import_state "$clients" "$facets" FULL
-	clients_up || error "post-failover df: $?"
+	clients_up || error "post-failover stat: $?"
 }
 
 fail_nodf() {
@@ -2559,8 +2656,8 @@ fail_abort() {
 	change_active $facet
 	wait_for_facet $facet
 	mount_facet $facet -o abort_recovery
-	clients_up || echo "first df failed: $?"
-	clients_up || error "post-failover df: $?"
+	clients_up || echo "first stat failed: $?"
+	clients_up || error "post-failover stat: $?"
 }
 
 do_lmc() {
@@ -3167,7 +3264,7 @@ facet_mntpt () {
     [[ $facet = mgs ]] && combined_mgs_mds && facet="mds1"
 
     local var=${facet}_MOUNT
-    eval mntpt=${!var:-${MOUNT%/*}/$facet}
+    eval mntpt=${!var:-${MOUNT}-$facet}
 
     echo -n $mntpt
 }
@@ -3175,7 +3272,7 @@ facet_mntpt () {
 mount_ldiskfs() {
 	local facet=$1
 	local dev=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local opts
 
 	if ! do_facet $facet test -b $dev; then
@@ -3187,7 +3284,7 @@ mount_ldiskfs() {
 unmount_ldiskfs() {
 	local facet=$1
 	local dev=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 
 	do_facet $facet $UMOUNT $mnt
 }
@@ -3199,7 +3296,7 @@ var_name() {
 mount_zfs() {
 	local facet=$1
 	local ds=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local canmnt
 	local mntpt
 
@@ -3222,7 +3319,7 @@ mount_zfs() {
 unmount_zfs() {
 	local facet=$1
 	local ds=$(facet_device $facet)
-	local mnt=$(facet_mntpt $facet)
+	local mnt=${2:-$(facet_mntpt $facet)}
 	local var_mntpt=mz_$(var_name ${facet}_$ds)_mountpoint
 	local var_canmnt=mz_$(var_name ${facet}_$ds)_canmount
 	local mntpt=${!var_mntpt}
@@ -3238,16 +3335,18 @@ unmount_zfs() {
 
 mount_fstype() {
 	local facet=$1
+	local mnt=$2
 	local fstype=$(facet_fstype $facet)
 
-	mount_$fstype $facet
+	mount_$fstype $facet $mnt
 }
 
 unmount_fstype() {
 	local facet=$1
+	local mnt=$2
 	local fstype=$(facet_fstype $facet)
 
-	unmount_$fstype $facet
+	unmount_$fstype $facet $mnt
 }
 
 ########
@@ -3267,7 +3366,8 @@ stopall() {
     zconf_umount_clients $clients $MOUNT "$*" || true
     [ -n "$MOUNT2" ] && zconf_umount_clients $clients $MOUNT2 "$*" || true
 
-    [ "$CLIENTONLY" ] && return
+    [ -n "$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
@@ -3290,12 +3390,14 @@ stopall() {
 }
 
 cleanup_echo_devs () {
-    local devs=$($LCTL dl | grep echo | awk '{print $4}')
+	trap 0
+	local dev
+	local devs=$($LCTL dl | grep echo | awk '{print $4}')
 
-    for dev in $devs; do
-        $LCTL --device $dev cleanup
-        $LCTL --device $dev detach
-    done
+	for dev in $devs; do
+		$LCTL --device $dev cleanup
+		$LCTL --device $dev detach
+	done
 }
 
 cleanupall() {
@@ -3356,7 +3458,6 @@ mkfs_opts() {
 	opts+=${LDLM_TIMEOUT:+" --param=sys.ldlm_timeout=$LDLM_TIMEOUT"}
 
 	if [ $type == MDS ]; then
-		opts+=${SECLEVEL:+" --param=mdt.sec_level"}
 		opts+=${MDSCAPA:+" --param-mdt.capa=$MDSCAPA"}
 		opts+=${STRIPE_BYTES:+" --param=lov.stripesize=$STRIPE_BYTES"}
 		opts+=${STRIPES_PER_OBJ:+" --param=lov.stripecount=$STRIPES_PER_OBJ"}
@@ -3380,7 +3481,6 @@ mkfs_opts() {
 	fi
 
 	if [ $type == OST ]; then
-		opts+=${SECLEVEL:+" --param=ost.sec_level"}
 		opts+=${OSSCAPA:+" --param=ost.capa=$OSSCAPA"}
 
 		if [ $fstype == ldiskfs ]; then
@@ -3475,9 +3575,16 @@ format_ost() {
 
 formatall() {
 	stopall
+	# Set hostid for ZFS/SPL zpool import protection
+	# (Assumes MDS version is also OSS version)
+	if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.8.54) ];
+	then
+	    do_rpc_nodes "$(comma_list $(remote_nodes_list))" set_hostid
+	fi
+
 	# We need ldiskfs here, may as well load them all
 	load_modules
-	[ "$CLIENTONLY" ] && return
+	[ -n "$CLIENTONLY" ] && return
 	echo Formatting mgs, mds, osts
 	if ! combined_mgs_mds ; then
 		format_mgs
@@ -3567,21 +3674,19 @@ writeconf_all () {
 }
 
 setupall() {
-    nfs_client_mode && return
+	nfs_client_mode && return
 	cifs_client_mode && return
 
-    sanity_mount_check ||
-        error "environments are insane!"
+	sanity_mount_check || error "environments are insane!"
 
-    load_modules
+	load_modules
 
-    if [ -z "$CLIENTONLY" ]; then
-        echo Setup mgs, mdt, osts
-        echo $WRITECONF | grep -q "writeconf" && \
-            writeconf_all
-        if ! combined_mgs_mds ; then
+	if [ -z "$CLIENTONLY" ]; then
+		echo Setup mgs, mdt, osts
+		echo $WRITECONF | grep -q "writeconf" && writeconf_all
+		if ! combined_mgs_mds ; then
 			start mgs $(mgsdevname) $MGS_MOUNT_OPTS
-        fi
+		fi
 
         for num in `seq $MDSCOUNT`; do
             DEVNAME=$(mdsdevname $num)
@@ -3650,7 +3755,7 @@ mounted_lustre_filesystems() {
 }
 
 init_facet_vars () {
-	[ "$CLIENTONLY" ] && return 0
+	[ -n "$CLIENTONLY" ] && return 0
 	local facet=$1
 	shift
 	local device=$1
@@ -3775,14 +3880,17 @@ set_conf_param_and_check() {
 	do_facet mgs "$LCTL conf_param $PARAM='$FINAL'" ||
 		error "conf_param $PARAM failed"
 
-	wait_update $(facet_host $myfacet) "$TEST" "$FINAL" ||
+	wait_update_facet $myfacet "$TEST" "$FINAL" ||
 		error "check $PARAM failed!"
 }
 
 init_param_vars () {
-	remote_mds_nodsh ||
-		TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
+	TIMEOUT=$(lctl get_param -n timeout)
+	TIMEOUT=${TIMEOUT:-20}
+
+	remote_mds_nodsh && log "Using TIMEOUT=$TIMEOUT" && return 0
 
+	TIMEOUT=$(do_facet $SINGLEMDS "lctl get_param -n timeout")
 	log "Using TIMEOUT=$TIMEOUT"
 
 	osc_ensure_active $SINGLEMDS $TIMEOUT
@@ -3843,14 +3951,14 @@ check_config_client () {
     local mntpt=$1
 
     local mounted=$(mount | grep " $mntpt ")
-    if [ "$CLIENTONLY" ]; then
+    if [ -n "$CLIENTONLY" ]; then
         # bug 18021
         # CLIENTONLY should not depend on *_HOST settings
         local mgc=$($LCTL device_list | awk '/MGC/ {print $4}')
         # in theory someone could create a new,
         # client-only config file that assumed lustre was already
         # configured and didn't set the MGSNID. If MGSNID is not set,
-        # then we should use the mgs nid currently being used 
+        # then we should use the mgs nid currently being used
         # as the default value. bug 18021
         [[ x$MGSNID = x ]] &&
             MGSNID=${mgc//MGC/}
@@ -3865,9 +3973,9 @@ check_config_client () {
         return 0
     fi
 
-    local myMGS_host=$mgs_HOST   
+    local myMGS_host=$mgs_HOST
     if [ "$NETTYPE" = "ptl" ]; then
-        myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//) 
+        myMGS_host=$(h2ptl $mgs_HOST | sed -e s/@ptl//)
     fi
 
     echo Checking config lustre mounted on $mntpt
@@ -3932,21 +4040,22 @@ 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 do_check=true
-    # 1.
-    # both MOUNT and MOUNT2 are not mounted
-    if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
-        [ "$REFORMAT" ] && formatall
-        # setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
-        setupall
-        is_mounted $MOUNT || error "NAME=$NAME not mounted"
-        export I_MOUNTED=yes
-        do_check=false
+	local MOUNTED=$(mounted_lustre_filesystems)
+
+	local do_check=true
+	# 1.
+	# both MOUNT and MOUNT2 are not mounted
+	if ! is_mounted $MOUNT && ! is_mounted $MOUNT2; then
+		[ "$REFORMAT" = "yes" ] && formatall
+		# setupall mounts both MOUNT and MOUNT2 (if MOUNT_2 is set)
+		setupall
+		is_mounted $MOUNT || error "NAME=$NAME not mounted"
+		export I_MOUNTED=yes
+		do_check=false
     # 2.
     # MOUNT2 is mounted
     elif is_mounted $MOUNT2; then
@@ -3969,7 +4078,7 @@ check_and_setup_lustre() {
                     restore_mount $MOUNT2
                     export I_MOUNTED2=yes
                 fi
-            fi 
+            fi
 
     # 5.
     # MOUNT is mounted MOUNT2 is not mounted
@@ -3990,7 +4099,7 @@ check_and_setup_lustre() {
         set_default_debug_nodes $(comma_list $(nodes_list))
     fi
 
-	if [ $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ] ; then
+	if [ -z "$CLIENTONLY" -a $(lower $OSD_TRACK_DECLARES_LBUG) == 'yes' ]; then
 		local facets=""
 		[ "$(facet_fstype ost1)" = "ldiskfs" ] &&
 			facets="$(get_facets OST)"
@@ -4010,12 +4119,14 @@ check_and_setup_lustre() {
 		set_flavor_all $SEC
 	fi
 
-	#Enable remote MDT create for testing
-	for num in $(seq $MDSCOUNT); do
-		do_facet mds$num \
-			lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
-				2>/dev/null
-	done
+	if [ -z "$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
@@ -4234,8 +4345,12 @@ check_and_cleanup_lustre() {
 	fi
 
 	if is_mounted $MOUNT; then
-		[ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
-			error "remove sub-test dirs failed"
+		if $DO_CLEANUP; then
+			[ -n "$DIR" ] && rm -rf $DIR/[Rdfs][0-9]* ||
+				error "remove sub-test dirs failed"
+		else
+			echo "skip cleanup"
+		fi
 		[ "$ENABLE_QUOTA" ] && restore_quota || true
 	fi
 
@@ -4558,7 +4673,8 @@ clear_failloc() {
 }
 
 set_nodes_failloc () {
-	do_nodes $(comma_list $1)  lctl set_param fail_val=0 fail_loc=$2
+	local fv=${3:-0}
+	do_nodes $(comma_list $1)  lctl set_param fail_val=$fv fail_loc=$2
 }
 
 cancel_lru_locks() {
@@ -4605,13 +4721,21 @@ pgcache_empty() {
 }
 
 debugsave() {
-    DEBUGSAVE="$(lctl get_param -n debug)"
+	DEBUGSAVE="$(lctl get_param -n debug)"
+	DEBUGSAVE_SERVER=$(do_facet $SINGLEMDS "$LCTL get_param -n debug")
 }
 
 debugrestore() {
-    [ -n "$DEBUGSAVE" ] && \
-        do_nodes $(comma_list $(nodes_list)) "$LCTL set_param debug=\\\"${DEBUGSAVE}\\\";"
-    DEBUGSAVE=""
+	[ -n "$DEBUGSAVE" ] &&
+		do_nodes $CLIENTS "$LCTL set_param debug=\\\"${DEBUGSAVE}\\\""||
+		true
+	DEBUGSAVE=""
+
+	[ -n "DEBUGSAVE_SERVER" ] &&
+		do_nodes $(comma_list $(all_server_nodes)) \
+			 "$LCTL set_param debug=\\\"${DEBUGSAVE_SERVER}\\\"" ||
+			 true
+	DEBUGSAVE_SERVER=""
 }
 
 debug_size_save() {
@@ -4642,20 +4766,17 @@ stop_full_debug_logging() {
 
 # prints bash call stack
 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
@@ -4665,10 +4786,8 @@ error_noexit() {
 		dump=false
 	fi
 
-
 	log " ${TESTSUITE} ${TESTNAME}: @@@@@@ ${TYPE}: $@ "
-	print_stack_trace >&2
-
+	(print_stack_trace 2) >&2
 	mkdir -p $LOGDIR
 	# We need to dump the logs on all nodes
 	if $dump; then
@@ -4687,6 +4806,17 @@ error_noexit() {
 			echo "$@" > $LOGDIR/err
 		fi
 	fi
+
+	# cleanup the env for failed tests
+	reset_fail_loc
+}
+
+##################################
+# Test interface
+##################################
+
+error_noexit() {
+	report_error "$@"
 }
 
 exit_status () {
@@ -4698,12 +4828,13 @@ 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.
@@ -4713,15 +4844,27 @@ 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
 }
 
+# Throw an error if it's not running in vm - usually for performance
+# verification
+error_not_in_vm() {
+	local virt=$(running_in_vm)
+	if [[ -n "$virt" ]]; then
+		echo "running in VM '$virt', ignore error"
+		error_ignore env=$virt "$@"
+	else
+		error "$@"
+	fi
+}
+
 skip_env () {
 	$FAIL_ON_SKIP_ENV && error false $@ || skip $@
 }
@@ -4907,17 +5050,17 @@ 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 \
-	    fail_val=0 2>/dev/null || true"
-    echo done.
+	echo -n "Resetting fail_loc on all nodes..."
+	do_nodes $(comma_list $(nodes_list)) "lctl set_param -n fail_loc=0 \
+	    fail_val=0 2>/dev/null" || true
+	echo done.
 }
 
 
 #
-# Log a message (on all nodes) padded with "=" before and after. 
+# Log a message (on all nodes) padded with "=" before and after.
 # Also appends a timestamp and prepends the testsuite name.
-# 
+#
 
 EQUALS="===================================================================================================="
 banner() {
@@ -4929,6 +5072,18 @@ 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.
 #
@@ -4944,12 +5099,17 @@ 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"
@@ -4958,6 +5118,7 @@ run_one() {
 	unset tdir
 	unset tfile
 	umask $SAVE_UMASK
+	$CLEANUP
 	return 0
 }
 
@@ -5150,15 +5311,15 @@ remote_mds ()
 
 remote_mds_nodsh()
 {
-    [ "$CLIENTONLY" ] && return 0 || true
-    remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	remote_mds && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$mds_HOST" ]
 }
 
 require_dsh_mds()
 {
-        remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" && \
-            MSKIPPED=1 && return 1
-        return 0
+	remote_mds_nodsh && echo "SKIP: $TESTSUITE: remote MDS with nodsh" &&
+		MSKIPPED=1 && return 1
+	return 0
 }
 
 remote_ost ()
@@ -5172,8 +5333,8 @@ remote_ost ()
 
 remote_ost_nodsh()
 {
-    [ "$CLIENTONLY" ] && return 0 || true 
-    remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	remote_ost && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
 require_dsh_ost()
@@ -5185,10 +5346,10 @@ 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" ]
+	[ -n "$CLIENTONLY" ] && return 0 || true
+	local MGS
+	MGS=$(facet_host mgs)
+	remote_node $MGS && [ "$PDSH" = "no_dsh" -o -z "$PDSH" -o -z "$ost_HOST" ]
 }
 
 local_mode ()
@@ -5349,19 +5510,14 @@ get_random_entry () {
     rnodes=${rnodes//,/ }
 
     local -a nodes=($rnodes)
-    local num=${#nodes[@]} 
+    local num=${#nodes[@]}
     local i=$((RANDOM * num * 2 / 65536))
 
     echo ${nodes[i]}
 }
 
 client_only () {
-    [ "$CLIENTONLY" ] || [ "$CLIENTMODSONLY" = yes ]
-}
-
-is_patchless ()
-{
-    lctl get_param version | grep -q patchless
+	[ -n "$CLIENTONLY" ] || [ "x$CLIENTMODSONLY" = "xyes" ]
 }
 
 check_versions () {
@@ -6070,19 +6226,21 @@ 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() {
@@ -6162,7 +6320,7 @@ gather_logs () {
     suffix="$ts.log"
     echo "Dumping lctl log to ${prefix}.*.${suffix}"
 
-    if [ "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
+    if [ -n "$CLIENTONLY" -o "$PDSH" == "no_dsh" ]; then
         echo "Dumping logs only on local client."
         $LCTL dk > ${prefix}.debug_log.$(hostname -s).${suffix}
         dmesg > ${prefix}.dmesg.$(hostname -s).${suffix}
@@ -6568,29 +6726,36 @@ check_write_access() {
 }
 
 init_logging() {
-    if [[ -n $YAML_LOG ]]; then
-        return
-    fi
-    local SAVE_UMASK=`umask`
-    umask 0000
-
-    export YAML_LOG=${LOGDIR}/results.yml
-    mkdir -p $LOGDIR
-    init_clients_lists
-
-    if [ ! -f $YAML_LOG ]; then       # If the yaml log already exists then we will just append to it
-      if check_shared_dir $LOGDIR; then
-          touch $LOGDIR/shared
-          echo "Logging to shared log directory: $LOGDIR"
-      else
-          echo "Logging to local directory: $LOGDIR"
-      fi
+	[[ -n $YAML_LOG ]] && return
+	local save_umask=$(umask)
+	umask 0000
 
-      yml_nodes_file $LOGDIR >> $YAML_LOG
-      yml_results_file >> $YAML_LOG
-    fi
+	export YAML_LOG=${LOGDIR}/results.yml
+	mkdir -p $LOGDIR
+	init_clients_lists
+
+	# If the yaml log already exists then we will just append to it
+	if [ ! -f $YAML_LOG ]; then
+		if check_shared_dir $LOGDIR; then
+			touch $LOGDIR/shared
+			echo "Logging to shared log directory: $LOGDIR"
+		else
+			echo "Logging to local directory: $LOGDIR"
+		fi
+
+		yml_nodes_file $LOGDIR >> $YAML_LOG
+		yml_results_file >> $YAML_LOG
+	fi
 
-    umask $SAVE_UMASK
+	umask $save_umask
+
+	# If modules are not yet loaded then older "lctl lustre_build_version"
+	# will fail.  Use lctl build version instead.
+	log "Client: $($LCTL lustre_build_version)"
+	log "MDS: $(do_facet $SINGLEMDS $LCTL lustre_build_version 2>/dev/null||
+		    do_facet $SINGLEMDS $LCTL --version)"
+	log "OSS: $(do_facet ost1 $LCTL lustre_build_version 2> /dev/null ||
+		    do_facet ost1 $LCTL --version)"
 }
 
 log_test() {
@@ -6792,11 +6957,10 @@ get_obd_size() {
 #
 get_page_size() {
 	local facet=$1
-	local size
+	local size=$(getconf PAGE_SIZE 2>/dev/null)
 
-	size=$(do_facet $facet getconf PAGE_SIZE)
-	[[ ${PIPESTATUS[0]} = 0 && -n "$size" ]] || size=4096
-	echo -n $size
+	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet getconf PAGE_SIZE)
+	echo -n ${size:-4096}
 }
 
 #
@@ -6807,20 +6971,20 @@ get_block_count() {
 	local device=$2
 	local count
 
-	count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+	[ -z "$CLIENTONLY" ] && count=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
 		awk '/^Block count:/ {print $3}')
-	echo -n $count
+	echo -n ${count:-0}
 }
 
 # Get the block size of the filesystem.
 get_block_size() {
-    local facet=$1
-    local device=$2
-    local size
+	local facet=$1
+	local device=$2
+	local size
 
-    size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
-           awk '/^Block size:/ {print $3}')
-    echo $size
+	[ -z "$CLIENTONLY" ] && size=$(do_facet $facet "$DUMPE2FS -h $device 2>&1" |
+		awk '/^Block size:/ {print $3}')
+	echo -n ${size:-0}
 }
 
 # Check whether the "large_xattr" feature is enabled or not.