X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=c352d8978649462917dba2ce02b6d0ca6b99fb8f;hp=c0aac29605ad3b06da4166abf5f38f6068d1627c;hb=a4d0b9612568d20baec2b53a25119bf1e697234e;hpb=72c1f7095203cc1badadf581c66f9546476438ab

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
index c0aac29..c352d89 100755
--- a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -77,8 +77,8 @@ if (( $LINUX_VERSION_CODE >= $(version_code 4.18.0) &&
 	ALWAYS_EXCEPT+=" 411"
 fi
 
-#                                  5              12     8   12  (min)"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 300o"
+#                                  5              12     8   12  15   (min)"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 60i 64b 68 71 115 135 136 230d 300o"
 
 if [ "$mds1_FSTYPE" = "zfs" ]; then
 	# bug number for skipped test:
@@ -5682,38 +5682,51 @@ run_test 51b "exceed 64k subdirectory nlink limit on create, verify unlink"
 test_51d() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 	[[ $OSTCOUNT -lt 3 ]] && skip_env "needs >= 3 OSTs"
+	local qos_old
 
 	test_mkdir $DIR/$tdir
+	$LFS setstripe -c $OSTCOUNT $DIR/$tdir
+
+	qos_old=$(do_facet mds1 \
+		"$LCTL get_param -n lod.$FSNAME-*.qos_threshold_rr" | head -n 1)
+	do_nodes $(comma_list $(mdts_nodes)) \
+		"$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=100"
+	stack_trap "do_nodes $(comma_list $(mdts_nodes)) \
+		'$LCTL set_param lod.$FSNAME-*.qos_threshold_rr=${qos_old%%%}'"
+
 	createmany -o $DIR/$tdir/t- 1000
 	$LFS getstripe $DIR/$tdir > $TMP/$tfile
-	for N in $(seq 0 $((OSTCOUNT - 1))); do
-		OBJS[$N]=$(awk -vobjs=0 '($1 == '$N') { objs += 1 } \
-			END { printf("%0.0f", objs) }' $TMP/$tfile)
-		OBJS0[$N]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
-			'($1 == '$N') { objs += 1 } \
-			END { printf("%0.0f", objs) }')
-		log "OST$N has ${OBJS[$N]} objects, ${OBJS0[$N]} are index 0"
+	for ((n = 0; n < $OSTCOUNT; n++)); do
+		objs[$n]=$(awk -vobjs=0 '($1 == '$n') { objs += 1 } \
+			   END { printf("%0.0f", objs) }' $TMP/$tfile)
+		objs0[$n]=$(grep -A 1 idx $TMP/$tfile | awk -vobjs=0 \
+			    '($1 == '$n') { objs += 1 } \
+			    END { printf("%0.0f", objs) }')
+		log "OST$n has ${objs[$n]} objects, ${objs0[$n]} are index 0"
 	done
 	unlinkmany $DIR/$tdir/t- 1000
 
-	NLAST=0
-	for N in $(seq 1 $((OSTCOUNT - 1))); do
-		[[ ${OBJS[$N]} -lt $((${OBJS[$NLAST]} - 20)) ]] &&
-			error "OST $N has less objects vs OST $NLAST" \
-			      " (${OBJS[$N]} < ${OBJS[$NLAST]}"
-		[[ ${OBJS[$N]} -gt $((${OBJS[$NLAST]} + 20)) ]] &&
-			error "OST $N has less objects vs OST $NLAST" \
-			      " (${OBJS[$N]} < ${OBJS[$NLAST]}"
-
-		[[ ${OBJS0[$N]} -lt $((${OBJS0[$NLAST]} - 20)) ]] &&
-			error "OST $N has less #0 objects vs OST $NLAST" \
-			      " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
-		[[ ${OBJS0[$N]} -gt $((${OBJS0[$NLAST]} + 20)) ]] &&
-			error "OST $N has less #0 objects vs OST $NLAST" \
-			      " (${OBJS0[$N]} < ${OBJS0[$NLAST]}"
-		NLAST=$N
+	nlast=0
+	for ((n = 0; n < $OSTCOUNT; n++)); do
+		(( ${objs[$n]} > ${objs[$nlast]} * 4 / 5 )) ||
+			{ $LFS df && $LFS df -i &&
+			error "OST $n has fewer objects vs. OST $nlast" \
+			      " (${objs[$n]} < ${objs[$nlast]}"; }
+		(( ${objs[$n]} < ${objs[$nlast]} * 5 / 4 )) ||
+			{ $LFS df && $LFS df -i &&
+			error "OST $n has fewer objects vs. OST $nlast" \
+			      " (${objs[$n]} < ${objs[$nlast]}"; }
+
+		(( ${objs0[$n]} > ${objs0[$nlast]} * 4 / 5 )) ||
+			{ $LFS df && $LFS df -i &&
+			error "OST $n has fewer #0 objects vs. OST $nlast" \
+			      " (${objs0[$n]} < ${objs0[$nlast]}"; }
+		(( ${objs0[$n]} < ${objs0[$nlast]} * 5 / 4 )) ||
+			{ $LFS df && $LFS df -i &&
+			error "OST $n has fewer #0 objects vs. OST $nlast" \
+			      " (${objs0[$n]} < ${objs0[$nlast]}"; }
+		nlast=$n
 	done
-	rm -f $TMP/$tfile
 }
 run_test 51d "check object distribution"
 
@@ -6131,6 +6144,47 @@ test_56d() {
 }
 run_test 56d "'lfs df -v' prints only configured devices"
 
+test_56e() {
+	err_enoent=2 # No such file or directory
+	err_eopnotsupp=95 # Operation not supported
+
+	enoent_mnt=/pmt1 # Invalid dentry. Path not present
+	notsup_mnt=/tmp  # Valid dentry, but Not a lustreFS
+
+	# Check for handling of path not exists
+	output=$($LFS df $enoent_mnt 2>&1)
+	ret=$?
+
+	fs=$(echo $output | awk -F: '{print $2}' | awk '{print $3}' | tr -d \')
+	[[ $fs = $enoent_mnt && $ret -eq $err_enoent ]] ||
+		error "expect failure $err_enoent, not $ret"
+
+	# Check for handling of non-Lustre FS
+	output=$($LFS df $notsup_mnt)
+	ret=$?
+
+	fs=$(echo $output | awk '{print $1}' | awk -F: '{print $2}')
+	[[ $fs = $notsup_mnt && $ret -eq $err_eopnotsupp ]] ||
+		error "expect success $err_eopnotsupp, not $ret"
+
+	# Check for multiple LustreFS argument
+	output=$($LFS df $MOUNT $MOUNT $MOUNT | grep -c "filesystem_summary:")
+	ret=$?
+
+	[[ $output -eq 3 && $ret -eq 0 ]] ||
+		error "expect success 3, not $output, rc = $ret"
+
+	# Check for correct non-Lustre FS handling among multiple
+	# LustreFS argument
+	output=$($LFS df $MOUNT $notsup_mnt $MOUNT |
+		grep -c "filesystem_summary:"; exit ${PIPESTATUS[0]})
+	ret=$?
+
+	[[ $output -eq 2 && $ret -eq $err_eopnotsupp ]] ||
+		error "expect success 2, not $output, rc = $ret"
+}
+run_test 56e "'lfs df' Handle non LustreFS & multiple LustreFS"
+
 NUMFILES=3
 NUMDIRS=3
 setup_56() {
@@ -7644,6 +7698,85 @@ test_56xf() {
 }
 run_test 56xf "FID is not lost during migration of a composite layout file"
 
+check_file_ost_range() {
+	local file="$1"
+	shift
+	local range="$*"
+	local -a file_range
+	local idx
+
+	file_range=($($LFS getstripe -y "$file" |
+		awk '/l_ost_idx:/ { print $NF }'))
+
+	if [[ "${#file_range[@]}" = 0 ]]; then
+		echo "No osts found for $file"
+		return 1
+	fi
+
+	for idx in "${file_range[@]}"; do
+		[[ " $range " =~ " $idx " ]] ||
+			return 1
+	done
+
+	return 0
+}
+
+sub_test_56xg() {
+	local stripe_opt="$1"
+	local pool="$2"
+	shift 2
+	local pool_ostidx="$(seq $* | tr '\n' ' ')"
+
+	$LFS migrate $stripe_opt -p $pool $DIR/$tfile ||
+		error "Fail to migrate $tfile on $pool"
+	[[ "$($LFS getstripe -p $DIR/$tfile)" = "$pool" ]] ||
+		error "$tfile is not in pool $pool"
+	check_file_ost_range "$DIR/$tfile" $pool_ostidx ||
+		error "$tfile osts mismatch with pool $pool (osts $pool_ostidx)"
+}
+
+test_56xg() {
+	[[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+	[[ $OSTCOUNT -ge 2 ]] || skip "needs >= 2 OSTs"
+	[[ $MDS1_VERSION -gt $(version_code 2.14.52) ]] ||
+		skip "Need MDS version newer than 2.14.52"
+
+	local -a pool_names=("${TESTNAME}_0" "${TESTNAME}_1" "${TESTNAME}_2")
+	local -a pool_ranges=("0 0" "1 1" "0 1")
+
+	# init pools
+	for i in "${!pool_names[@]}"; do
+		pool_add ${pool_names[$i]} ||
+			error "pool_add failed (pool: ${pool_names[$i]})"
+		pool_add_targets ${pool_names[$i]} ${pool_ranges[$i]} ||
+			error "pool_add_targets failed (pool: ${pool_names[$i]})"
+	done
+
+	# init the file to migrate
+	$LFS setstripe -c1 -i1 $DIR/$tfile ||
+		error "Unable to create $tfile on OST1"
+	dd if=/dev/urandom of=$DIR/$tfile bs=1M count=4 status=none ||
+		error "Unable to write on $tfile"
+
+	echo "1. migrate $tfile on pool ${pool_names[0]}"
+	sub_test_56xg "-c-1" "${pool_names[0]}" ${pool_ranges[0]}
+
+	echo "2. migrate $tfile on pool ${pool_names[2]}"
+	sub_test_56xg "-c-1 -S2M" "${pool_names[2]}" ${pool_ranges[2]}
+
+	echo "3. migrate $tfile on pool ${pool_names[1]}"
+	sub_test_56xg "-n -c-1" "${pool_names[1]}" ${pool_ranges[1]}
+
+	echo "4. migrate $tfile on pool ${pool_names[2]} with default stripe parameters"
+	sub_test_56xg "" "${pool_names[2]}" ${pool_ranges[2]}
+	echo
+
+	# Clean pools
+	destroy_test_pools ||
+		error "pool_destroy failed"
+}
+run_test 56xg "lfs migrate pool support"
+
 test_56y() {
 	[ $MDS1_VERSION -lt $(version_code 2.4.53) ] &&
 		skip "No HSM $(lustre_build_version $SINGLEMDS) MDS < 2.4.53"
@@ -8696,6 +8829,104 @@ test_64f() {
 }
 run_test 64f "check grant consumption (with grant allocation)"
 
+test_64g() {
+	#[ $MDS1_VERSION -lt $(version_code 2.14.54) ] &&
+	#	skip "Need MDS version at least 2.14.54"
+
+	local mdts=$(comma_list $(mdts_nodes))
+
+	local old=$($LCTL get_param mdc.$FSNAME-*.grant_shrink_interval |
+			tr '\n' ' ')
+	stack_trap "$LCTL set_param $old"
+
+	# generate dirty pages and increase dirty granted on MDT
+	stack_trap "rm -f $DIR/$tfile-*"
+	for (( i = 0; i < 10; i++)); do
+		$LFS setstripe -E 1M -L mdt $DIR/$tfile-$i ||
+			error "can't set stripe"
+		dd if=/dev/zero of=$DIR/$tfile-$i bs=128k count=1 ||
+			error "can't dd"
+		$LFS getstripe $DIR/$tfile-$i | grep -q pattern.*mdt || {
+			$LFS getstripe $DIR/$tfile-$i
+			error "not DoM file"
+		}
+	done
+
+	# flush dirty pages
+	sync
+
+	# wait until grant shrink reset grant dirty on MDTs
+	for ((i = 0; i < 120; i++)); do
+		grant_dirty=$(do_nodes $mdts $LCTL get_param -n  mdt.*.tot_dirty |
+			awk '{sum=sum+$1} END {print sum}')
+		vm_dirty=$(awk '/Dirty:/{print $2}' /proc/meminfo)
+		echo "$grant_dirty grants, $vm_dirty pages"
+		(( grant_dirty + vm_dirty == 0 )) && break
+		(( i == 3 )) && sync &&
+			$LCTL set_param mdc.$FSNAME-*.grant_shrink_interval=5
+		sleep 1
+	done
+
+	grant_dirty=$(do_nodes $mdts $LCTL get_param -n  mdt.*.tot_dirty |
+		awk '{sum=sum+$1} END {print sum}')
+	(( grant_dirty == 0 )) || error "$grant_dirty on MDT"
+}
+run_test 64g "grant shrink on MDT"
+
+test_64h() {
+	local instance=$($LFS getname -i $DIR)
+	local osc_tgt="$FSNAME-OST0000-osc-$instance"
+	local num_exps=$(do_facet ost1 \
+	    $LCTL get_param -n obdfilter.*OST0000*.num_exports)
+	local max_brw_size=$(import_param $osc_tgt max_brw_size)
+	local avail=$($LCTL get_param -n osc.*OST0000-osc-$instance.kbytesavail)
+	local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+
+	# 10MiB is for file to be written, max_brw_size * 16 *
+	# num_exps is space reserve so that tgt_grant_shrink() decided
+	# to not shrink
+	local expect=$((max_brw_size * 16 * num_exps + 10 * 1048576))
+	(( avail * 1024 < expect )) &&
+		skip "need $expect bytes on ost1, have $(( avail * 1024 )) only"
+
+	save_lustre_params client "osc.*OST0000*.grant_shrink" > $p
+	save_lustre_params client "osc.*OST0000*.grant_shrink_interval" >> $p
+	stack_trap "restore_lustre_params < $p; rm -f $save" EXIT
+	$LCTL set_param osc.*OST0000*.grant_shrink=1
+	$LCTL set_param osc.*OST0000*.grant_shrink_interval=10
+
+	$LFS setstripe -c 1 -i 0 $DIR/$tfile
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=10 oflag=sync
+
+	# drop cache so that coming read would do rpc
+	cancel_lru_locks osc
+
+	# shrink interval is set to 10, pause for 7 seconds so that
+	# grant thread did not wake up yet but coming read entered
+	# shrink mode for rpc (osc_should_shrink_grant())
+	sleep 7
+
+	declare -a cur_grant_bytes
+	declare -a tot_granted
+	cur_grant_bytes[0]=$($LCTL get_param -n osc.*OST0000*.cur_grant_bytes)
+	tot_granted[0]=$(do_facet ost1 \
+	    $LCTL get_param -n obdfilter.*OST0000*.tot_granted)
+
+	dd if=$DIR/$tfile bs=4K count=1 of=/dev/null
+
+	cur_grant_bytes[1]=$($LCTL get_param -n osc.*OST0000*.cur_grant_bytes)
+	tot_granted[1]=$(do_facet ost1 \
+	    $LCTL get_param -n obdfilter.*OST0000*.tot_granted)
+
+	# grant change should be equal on both sides
+	(( cur_grant_bytes[0] - cur_grant_bytes[1] ==
+		tot_granted[0] - tot_granted[1])) ||
+		error "grant change mismatch, "				       \
+			"server: ${tot_granted[0]} to ${tot_granted[1]}, "     \
+			"client: ${cur_grant_bytes[0]} to ${cur_grant_bytes[1]}"
+}
+run_test 64h "grant shrink on read"
+
 # bug 1414 - set/get directories' stripe info
 test_65a() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
@@ -9720,6 +9951,36 @@ test_77n() {
 }
 run_test 77n "Verify read from a hole inside contiguous blocks with T10PI"
 
+test_77o() {
+	(( $CLIENT_VERSION >= $(version_code 2.14.54) )) ||
+		skip "Need at least version 2.14.54"
+	local ofd=obdfilter
+	local mdt=mdt
+
+	# print OST checksum_type
+	echo "$ofd.$FSNAME-*.checksum_type:"
+	do_nodes $(comma_list $(osts_nodes)) \
+		$LCTL get_param -n $ofd.$FSNAME-*.checksum_type
+
+	# print MDT checksum_type
+	echo "$mdt.$FSNAME-*.checksum_type:"
+	do_nodes $(comma_list $(mdts_nodes)) \
+		$LCTL get_param -n $mdt.$FSNAME-*.checksum_type
+
+	local o_count=$(do_nodes $(comma_list $(osts_nodes)) \
+		   $LCTL get_param -n $ofd.$FSNAME-*.checksum_type | wc -l)
+
+	(( $o_count == $OSTCOUNT )) ||
+		error "found $o_count checksums, not \$MDSCOUNT=$OSTCOUNT"
+
+	local m_count=$(do_nodes $(comma_list $(mdts_nodes)) \
+		   $LCTL get_param -n $mdt.$FSNAME-*.checksum_type | wc -l)
+
+	(( $m_count == $MDSCOUNT )) ||
+		error "found $m_count checksums, not \$MDSCOUNT=$MDSCOUNT"
+}
+run_test 77o "Verify checksum_type for server (mdt and ofd(obdfilter))"
+
 cleanup_test_78() {
 	trap 0
 	rm -f $DIR/$tfile
@@ -16352,7 +16613,7 @@ test_160o() {
 
 	changelog_register --user test_160o -m unlnk+close+open ||
 		error "changelog_register failed"
-	# drop server mask so it doesn't interfere
+
 	do_facet $SINGLEMDS $LCTL --device $mdt \
 				changelog_register -u "Tt3_-#" &&
 		error "bad symbols in name should fail"
@@ -16443,6 +16704,28 @@ test_160p() {
 }
 run_test 160p "Changelog orphan cleanup with no users"
 
+test_160q() {
+	local mdt="$(facet_svc $SINGLEMDS)"
+	local clu
+
+	[[ $PARALLEL != "yes" ]] || skip "skip parallel run"
+	remote_mds_nodsh && skip "remote MDS with nodsh"
+	[ $MDS1_VERSION -ge $(version_code 2.14.54) ] ||
+		skip "Need MDS version at least 2.14.54"
+
+	# set server mask to minimal value like server init does
+	changelog_chmask "MARK"
+	clu=$(do_facet $SINGLEMDS $LCTL --device $mdt changelog_register -n) ||
+		error "changelog_register failed"
+	# check effective mask again, should be treated as DEFMASK now
+	mask=$(do_facet $SINGLEMDS $LCTL get_param \
+				mdd.$mdt.changelog_current_mask -n)
+	do_facet $SINGLEMDS $LCTL --device $mdt changelog_deregister $clu ||
+		error "changelog_deregister failed"
+	[[ $mask == *"HLINK"* ]] || error "mask is not DEFMASK as expected"
+}
+run_test 160q "changelog effective mask is DEFMASK if not set"
+
 test_161a() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 
@@ -18178,23 +18461,23 @@ test_208() {
 	$MULTIOP $DIR/$tfile oO_CREAT:O_RDWR:eRE+eU || error "get lease error"
 
 	echo "==== test 2: verify lease can be broken by upcoming open"
-	$MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+	$MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
 	local PID=$!
 	sleep 1
 
-	$MULTIOP $DIR/$tfile oO_RDONLY:c
+	$MULTIOP $DIR/$tfile oO_RDWR:c
 	kill -USR1 $PID && wait $PID || error "break lease error"
 
 	echo "==== test 3: verify lease can't be granted if an open already exists"
-	$MULTIOP $DIR/$tfile oO_RDONLY:_c &
+	$MULTIOP $DIR/$tfile oO_RDWR:_c &
 	local PID=$!
 	sleep 1
 
-	$MULTIOP $DIR/$tfile oO_RDONLY:eReUc && error "apply lease should fail"
+	$MULTIOP $DIR/$tfile oO_RDWR:eReUc && error "apply lease should fail"
 	kill -USR1 $PID && wait $PID || error "open file error"
 
 	echo "==== test 4: lease can sustain over recovery"
-	$MULTIOP $DIR/$tfile oO_RDONLY:eR_E+eUc &
+	$MULTIOP $DIR/$tfile oO_RDWR:eR_E+eUc &
 	PID=$!
 	sleep 1
 
@@ -18203,7 +18486,7 @@ test_208() {
 	kill -USR1 $PID && wait $PID || error "lease broken over recovery"
 
 	echo "==== test 5: lease broken can't be regained by replay"
-	$MULTIOP $DIR/$tfile oO_RDONLY:eR_E-eUc &
+	$MULTIOP $DIR/$tfile oO_RDWR:eR_E-eUc &
 	PID=$!
 	sleep 1
 
@@ -18649,26 +18932,48 @@ run_test 223 "osc reenqueue if without AGL lock granted ======================="
 
 test_224a() { # LU-1039, MRP-303
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
-
 	#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB   0x508
 	$LCTL set_param fail_loc=0x508
-	dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=fsync
 	$LCTL set_param fail_loc=0
 	df $DIR
 }
 run_test 224a "Don't panic on bulk IO failure"
 
-test_224b() { # LU-1039, MRP-303
+test_224bd_sub() { # LU-1039, MRP-303
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
+	local timeout=$1
 
-	dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
+	shift
+	dd if=/dev/urandom of=$TMP/$tfile bs=1M count=1
+
+	$LFS setstripe -c 1 -i 0 $DIR/$tfile
+
+	dd if=$TMP/$tfile of=$DIR/$tfile bs=1M count=1
 	cancel_lru_locks osc
+	set_checksums 0
+	stack_trap "set_checksums $ORIG_CSUM" EXIT
+	local at_max_saved=0
+
+	# adaptive timeouts may prevent seeing the issue
+	if at_is_enabled; then
+		at_max_saved=$(at_max_get mds)
+		at_max_set 0 mds client
+		stack_trap "at_max_set $at_max_saved mds client" EXIT
+	fi
+
 	#define OBD_FAIL_PTLRPC_CLIENT_BULK_CB2   0x515
-	$LCTL set_param fail_loc=0x515
-	dd of=/dev/null if=$DIR/$tfile bs=4096 count=1
-	$LCTL set_param fail_loc=0
+	do_facet ost1 $LCTL set_param fail_val=$timeout fail_loc=0x80000515
+	dd of=$TMP/$tfile.new if=$DIR/$tfile bs=1M count=1 || "$@"
+
+	do_facet ost1 $LCTL set_param fail_loc=0
+	cmp $TMP/$tfile $TMP/$tfile.new || error "file contents wrong"
 	df $DIR
 }
+
+test_224b() {
+	test_224bd_sub 3 error "dd failed"
+}
 run_test 224b "Don't panic on bulk IO failure"
 
 test_224c() { # LU-6441
@@ -18709,6 +19014,11 @@ test_224c() { # LU-6441
 }
 run_test 224c "Don't hang if one of md lost during large bulk RPC"
 
+test_224d() { # LU-11169
+	test_224bd_sub $((TIMEOUT + 2)) error "dd failed"
+}
+run_test 224d "Don't corrupt data on bulk IO timeout"
+
 MDSSURVEY=${MDSSURVEY:-$(which mds-survey 2>/dev/null || true)}
 test_225a () {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
@@ -19388,7 +19698,7 @@ test_230d() {
 		error "migrate remote dir error"
 
 	echo "Finish migration, then checking.."
-	for file in $(find $migrate_dir); do
+	for file in $(find $migrate_dir -maxdepth 1); do
 		mdt_index=$($LFS getstripe -m $file)
 		if [ $mdt_index -lt $new_index ] ||
 		   [ $mdt_index -gt $((new_index + new_count - 1)) ]; then
@@ -19945,6 +20255,48 @@ test_230t()
 }
 run_test 230t "migrate directory with project ID set"
 
+test_230u()
+{
+	(( MDSCOUNT > 3 )) || skip_env "needs >= 4 MDTs"
+	(( MDS1_VERSION >= $(version_code 2.14.53) )) ||
+		skip "Need MDS version at least 2.14.53"
+
+	local count
+
+	mkdir_on_mdt0 $DIR/$tdir || error "mkdir $tdir failed"
+	mkdir $DIR/$tdir/sub{0..99} || error "mkdir sub failed"
+	$LFS migrate -m -1 $DIR/$tdir/sub{0..99} || error "migrate sub failed"
+	for i in $(seq 0 $((MDSCOUNT - 1))); do
+		count=$($LFS getstripe -m $DIR/$tdir/sub* | grep -c ^$i)
+		echo "$count dirs migrated to MDT$i"
+	done
+	count=$($LFS getstripe -m $DIR/$tdir/sub* | sort -u | wc -l)
+	(( count >= MDSCOUNT - 1 )) || error "dirs migrated to $count MDTs"
+}
+run_test 230u "migrate directory by QOS"
+
+test_230v()
+{
+	(( MDSCOUNT > 3 )) || skip_env "needs >= 4 MDTs"
+	(( MDS1_VERSION >= $(version_code 2.14.53) )) ||
+		skip "Need MDS version at least 2.14.53"
+
+	local count
+
+	mkdir $DIR/$tdir || error "mkdir $tdir failed"
+	mkdir $DIR/$tdir/sub{0..99} || error "mkdir sub failed"
+	$LFS migrate -m 0,2,1 $DIR/$tdir || error "migrate $tdir failed"
+	for i in $(seq 0 $((MDSCOUNT - 1))); do
+		count=$($LFS getstripe -m $DIR/$tdir/sub* | grep -c ^$i)
+		echo "$count subdirs migrated to MDT$i"
+		(( i == 3 )) && (( count > 0 )) &&
+			error "subdir shouldn't be migrated to MDT3"
+	done
+	count=$($LFS getstripe -m $DIR/$tdir/sub* | sort -u | wc -l)
+	(( count == 3 )) || error "dirs migrated to $count MDTs"
+}
+run_test 230v "subdir migrated to the MDT where its parent is located"
+
 test_231a()
 {
 	# For simplicity this test assumes that max_pages_per_rpc
@@ -20844,18 +21196,20 @@ run_test 253 "Check object allocation limit"
 test_254() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 	remote_mds_nodsh && skip "remote MDS with nodsh"
-	do_facet $SINGLEMDS $LCTL get_param -n mdd.$MDT0.changelog_size ||
+
+	local mdt=$(facet_svc $SINGLEMDS)
+
+	do_facet $SINGLEMDS $LCTL get_param -n mdd.$mdt.changelog_size ||
 		skip "MDS does not support changelog_size"
 
 	local cl_user
-	local MDT0=$(facet_svc $SINGLEMDS)
 
 	changelog_register || error "changelog_register failed"
 
 	changelog_clear 0 || error "changelog_clear failed"
 
 	local size1=$(do_facet $SINGLEMDS \
-		      $LCTL get_param -n mdd.$MDT0.changelog_size)
+		      $LCTL get_param -n mdd.$mdt.changelog_size)
 	echo "Changelog size $size1"
 
 	rm -rf $DIR/$tdir
@@ -20870,7 +21224,7 @@ test_254() {
 	rm $DIR/$tdir/pics/desktop.jpg
 
 	local size2=$(do_facet $SINGLEMDS \
-		      $LCTL get_param -n mdd.$MDT0.changelog_size)
+		      $LCTL get_param -n mdd.$mdt.changelog_size)
 	echo "Changelog size after work $size2"
 
 	(( $size2 > $size1 )) ||
@@ -21799,6 +22153,16 @@ test_270h() {
 }
 run_test 270h "DoM: DoM stripe removal when disabled on server"
 
+test_270i() {
+	(( $MDS1_VERSION >= $(version_code 2.14.54) )) ||
+		skip "Need MDS version at least 2.14.54"
+
+	mkdir $DIR/$tdir
+	$LFS setstripe -L mdt -S 128k -c -1 $DIR/$tdir &&
+		error "setstripe should fail" || true
+}
+run_test 270i "DoM: setting invalid DoM striping should fail"
+
 test_271a() {
 	[ $MDS1_VERSION -lt $(version_code 2.10.55) ] &&
 		skip "Need MDS version at least 2.10.55"
@@ -24627,6 +24991,8 @@ run_test 412 "mkdir on specific MDTs"
 
 generate_uneven_mdts() {
 	local threshold=$1
+	local lmv_qos_maxage
+	local lod_qos_maxage
 	local ffree
 	local bavail
 	local max
@@ -24636,6 +25002,17 @@ generate_uneven_mdts() {
 	local tmp
 	local i
 
+	lmv_qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage)
+	$LCTL set_param lmv.*.qos_maxage=1
+	stack_trap "$LCTL set_param \
+		lmv.*.qos_maxage=$lmv_qos_maxage > /dev/null" RETURN
+	lod_qos_maxage=$(do_facet mds1 $LCTL get_param -n \
+		lod.$FSNAME-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }')
+	do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param \
+		lod.*.mdt_qos_maxage=1
+	stack_trap "do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param \
+		lod.*.mdt_qos_maxage=$lod_qos_maxage > /dev/null" RETURN
+
 	echo
 	echo "Check for uneven MDTs: "
 
@@ -24659,9 +25036,15 @@ generate_uneven_mdts() {
 		fi
 	done
 
+	(( ${ffree[min_index]} > 0 )) ||
+		skip "no free files in MDT$min_index"
+	(( ${ffree[min_index]} < 10000000 )) ||
+		skip "too many free files in MDT$min_index"
+
 	# Check if we need to generate uneven MDTs
 	local diff=$(((max - min) * 100 / min))
 	local testdir=$DIR/$tdir-fillmdt
+	local start
 
 	mkdir -p $testdir
 
@@ -24669,17 +25052,21 @@ generate_uneven_mdts() {
 	while (( diff < threshold )); do
 		# generate uneven MDTs, create till $threshold% diff
 		echo -n "weight diff=$diff% must be > $threshold% ..."
-		echo "Fill MDT$min_index with 100 files: loop $i"
+		echo "Fill MDT$min_index with 1000 files: loop $i"
 		testdir=$DIR/$tdir-fillmdt/$i
 		[ -d $testdir ] || $LFS mkdir -i $min_index $testdir ||
 			error "mkdir $testdir failed"
 		$LFS setstripe -E 1M -L mdt $testdir ||
 			error "setstripe $testdir failed"
-		for F in f.{0..99}; do
-			dd if=/dev/zero of=$testdir/$F bs=1M count=1 > \
+		start=$SECONDS
+		for F in f.{0..999}; do
+			dd if=/dev/zero of=$testdir/$F bs=64K count=1 > \
 				/dev/null 2>&1 || error "dd $F failed"
 		done
 
+		# wait for QOS to update
+		(( SECONDS < start + 1 )) && sleep $((start + 1 - SECONDS))
+
 		ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-*.filesfree))
 		bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-*.kbytesavail))
 		max=$(((${ffree[max_index]} >> 8) * \
@@ -24746,7 +25133,6 @@ test_qos_mkdir() {
 	local stripe_index=$($LFS getstripe -m $testdir)
 	local test_mkdir_rr=true
 
-	echo "dirstripe: '$($LFS getdirstripe $testdir)'"
 	getfattr -d -m dmv -e hex $testdir | grep dmv
 	if (( $? == 0 && $MDS1_VERSION >= $(version_code 2.14.51) )); then
 		echo "defstripe: '$($LFS getdirstripe -D $testdir)'"
@@ -24823,7 +25209,7 @@ test_qos_mkdir() {
 
 	(( ${ffree[min_index]} > 0 )) ||
 		skip "no free files in MDT$min_index"
-	(( ${ffree[min_index]} < 100000000 )) ||
+	(( ${ffree[min_index]} < 10000000 )) ||
 		skip "too many free files in MDT$min_index"
 
 	echo "MDT filesfree available: ${ffree[@]}"
@@ -24851,32 +25237,33 @@ test_qos_mkdir() {
 			error "$mkdir_cmd subdir$i failed"
 	done
 
+	max=0
 	for (( i = 0; i < $MDSCOUNT; i++ )); do
 		count=$($LFS getdirstripe -i $testdir/* | grep -c "^$i$")
+		(( count > max )) && max=$count
 		echo "$count directories created on MDT$i"
-
-		if [ $stripe_count -gt 1 ]; then
-			count=$($LFS getdirstripe $testdir/* |
-				grep -c -P "^\s+$i\t")
-			echo "$count stripes created on MDT$i"
-		fi
 	done
 
-	max=$($LFS getdirstripe -i $testdir/* | grep -c "^$max_index$")
 	min=$($LFS getdirstripe -i $testdir/* | grep -c "^$min_index$")
 
 	# D-value should > 10% of averge
-	(( max - min >= num / 10 )) ||
+	(( max - min > num / 10 )) ||
 		error "subdirs shouldn't be evenly distributed: $max - $min < $((num / 10))"
 
-	# 5% for stripes
+	# ditto for stripes
 	if (( stripe_count > 1 )); then
-		max=$($LFS getdirstripe $testdir/* |
-		      grep -c -P "^\s+$max_index\t")
+		max=0
+		for (( i = 0; i < $MDSCOUNT; i++ )); do
+			count=$($LFS getdirstripe $testdir/* |
+				grep -c -P "^\s+$i\t")
+			(( count > max )) && max=$count
+			echo "$count stripes created on MDT$i"
+		done
+
 		min=$($LFS getdirstripe $testdir/* |
 			grep -c -P "^\s+$min_index\t")
-		(( max - min >= num * stripe_count / 20 )) ||
-			error "stripes shouldn't be evenly distributed: $max - $min < $((num / 20)) * $stripe_count"
+		(( max - min > num * stripe_count / 10 )) ||
+			error "stripes shouldn't be evenly distributed: $max - $min < $((num / 10)) * $stripe_count"
 	fi
 }
 
@@ -25020,7 +25407,7 @@ test_413z() {
 	local pid
 
 	for subdir in $(\ls -1 -d $DIR/d413*-fillmdt/*); do
-		unlinkmany $subdir/f. 100 &
+		unlinkmany $subdir/f. 1000 &
 		pids="$pids $!"
 	done
 
@@ -25145,7 +25532,8 @@ check_lfs_df() {
 	[ "$1" == "blocks" ] && inodes= || inodes="-i"
 
 	for count in {1..100}; do
-		cancel_lru_locks
+		do_nodes "$CLIENTS" \
+			$LCTL set_param ldlm.namespaces.*.lru_size=clear
 		sync; sleep 0.2
 
 		# read the lines of interest
@@ -25158,7 +25546,9 @@ check_lfs_df() {
 		# "<NID>:/<fsname>" for df, "filesystem_summary:" for lfs df
 		# compare the two outputs
 		passed=true
-		for i in {1..5}; do
+		#  skip "available" on MDT until LU-13997 is fixed.
+		#for i in {1..5}; do
+		for i in 1 2 4 5; do
 			[ "${df_out[i]}" != "${lfs_df_out[i]}" ] && passed=false
 		done
 		$passed && break
@@ -27183,6 +27573,36 @@ test_822() {
 }
 run_test 822 "test precreate failure"
 
+test_823() {
+	local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+	local OST_MAX_PRECREATE=20000
+
+	save_lustre_params mds1 \
+		"osp.$FSNAME-OST*-osc-MDT0000.max_create_count" > $p
+	do_facet $SINGLEMDS "$LCTL set_param -n \
+		osp.$FSNAME-OST*MDT0000.max_create_count=0"
+	do_facet $SINGLEMDS "$LCTL set_param -n \
+		osp.$FSNAME-OST0000*MDT0000.max_create_count=$OST_MAX_PRECREATE"
+
+	stack_trap "restore_lustre_params < $p; rm $p"
+
+	do_facet $SINGLEMDS "$LCTL set_param -n \
+		osp.$FSNAME-OST*-osc-MDT*.create_count=100200"
+
+	local count=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+		      osp.$FSNAME-OST0000*MDT0000.create_count")
+	local max=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+		    osp.$FSNAME-OST0000*MDT0000.max_create_count")
+	local expect_count=$(((($max/2)/256) * 256))
+
+	log "setting create_count to 100200:"
+	log " -result- count: $count with max: $max, expecting: $expect_count"
+
+	[[ $count -eq expect_count ]] ||
+		error "Create count not set to max precreate."
+}
+run_test 823 "Setting create_count > OST_MAX_PRECREATE is lowered to maximum"
+
 #
 # tests that do cleanup/setup should be run at the end
 #
@@ -27238,6 +27658,21 @@ test_902() {
 }
 run_test 902 "test short write doesn't hang lustre"
 
+# LU-14711
+test_903() {
+	$LFS setstripe -i 0 -c 1 $DIR/$tfile $DIR/${tfile}-2
+	echo "blah" > $DIR/${tfile}-2
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=6 conv=fsync
+	#define OBD_FAIL_OSC_SLOW_PAGE_EVICT 0x417
+	$LCTL set_param fail_loc=0x417 fail_val=20
+
+	mv $DIR/${tfile}-2 $DIR/$tfile # Destroys the big object
+	sleep 1 # To start the destroy
+	wait_destroy_complete 150 || error "Destroy taking too long"
+	cat $DIR/$tfile > /dev/null || error "Evicted"
+}
+run_test 903 "Test long page discard does not cause evictions"
+
 complete $SECONDS
 [ -f $EXT2_DEV ] && rm $EXT2_DEV || true
 check_and_cleanup_lustre