X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=541fae3408cd5d0df3cbb6182f65d33b0cc00080;hp=027271c65a37e1486cc3c277a93db1e857ce1aaf;hb=59a3d7237df82ac2ce14b6779577584163f9122b;hpb=0ad54d59777366fba8ee61eaaa27b3060c91782f

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
old mode 100644
new mode 100755
index 027271c..541fae3
--- a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -45,8 +45,8 @@ ALWAYS_EXCEPT+="               42a     42b     42c "
 ALWAYS_EXCEPT+=" 407     312 "
 
 if $SHARED_KEY; then
-	# bug number:    LU-9795 LU-9795 LU-9795 LU-9795 LU-12781
-	ALWAYS_EXCEPT+=" 17n     60a     133g    300f    272a"
+	# bug number:    LU-9795 LU-9795 LU-9795 LU-9795
+	ALWAYS_EXCEPT+=" 17n     60a     133g    300f"
 fi
 
 selinux_status=$(getenforce)
@@ -59,12 +59,24 @@ fi
 if [[ $(uname -m) = aarch64 ]]; then
 	# bug number:	 LU-11596
 	ALWAYS_EXCEPT+=" $GRANT_CHECK_LIST"
-	# bug number:	 LU-11671 LU-11667 LU-4398
-	ALWAYS_EXCEPT+=" 45	  317      817"
+	# bug number:	 LU-11671 LU-11667
+	ALWAYS_EXCEPT+=" 45	  317"
+fi
+
+# skip nfs tests on kernels >= 4.14.0 until they are fixed
+if [ $LINUX_VERSION_CODE -ge $(version_code 4.14.0) ]; then
+	# bug number:	LU-12661
+	ALWAYS_EXCEPT+=" 817"
+fi
+# skip cgroup tests on RHEL8.1 kernels until they are fixed
+if (( $LINUX_VERSION_CODE >= $(version_code 4.18.0) &&
+      $LINUX_VERSION_CODE <  $(version_code 5.4.0) )); then
+	# bug number:	LU-13063
+	ALWAYS_EXCEPT+=" 411"
 fi
 
 #                                  5          12          (min)"
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 135 136 300o"
 
 if [ "$mds1_FSTYPE" = "zfs" ]; then
 	# bug number for skipped test:
@@ -544,11 +556,6 @@ test_17g() {
 	[ $MDS1_VERSION -le $(version_code 2.3.55) ] &&
 		TESTS="4094 4095"
 
-	# skip long symlink name for rhel6.5.
-	# rhel6.5 has a limit (PATH_MAX - sizeof(struct filename))
-	grep -q '6.5' /etc/redhat-release &>/dev/null &&
-		TESTS="59 60 61 4062 4063"
-
 	for i in $TESTS; do
 		local SYMNAME=$(str_repeat 'x' $i)
 		ln -s $SYMNAME $DIR/$tdir/f$i || error "failed $i-char symlink"
@@ -2635,7 +2642,7 @@ test_27I() {
 run_test 27I "check that root dir striping does not break parent dir one"
 
 test_27J() {
-	[[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.51) ]] &&
+	[[ $MDS1_VERSION -le $(version_code 2.12.51) ]] &&
 		skip "Need MDS version newer than 2.12.51"
 
 	test_mkdir $DIR/$tdir
@@ -2730,7 +2737,7 @@ test_27J() {
 run_test 27J "basic ops on file with foreign LOV"
 
 test_27K() {
-	[[ $(lustre_version_code $SINGLEMDS) -le $(version_code 2.12.49) ]] &&
+	[[ $MDS1_VERSION -le $(version_code 2.12.49) ]] &&
 		skip "Need MDS version newer than 2.12.49"
 
 	test_mkdir $DIR/$tdir
@@ -4981,7 +4988,7 @@ test_49() { # LU-1030
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 	remote_ost_nodsh && skip "remote OST with nodsh"
 
-	# get ost1 size - lustre-OST0000
+	# get ost1 size - $FSNAME-OST0000
 	ost1_size=$(do_facet ost1 $LFS df | grep ${ost1_svc} |
 		awk '{ print $4 }')
 	# write 800M at maximum
@@ -5744,6 +5751,65 @@ test_56ob() {
 }
 run_test 56ob "check lfs find -atime -mtime -ctime with units"
 
+test_newerXY_base() {
+	local x=$1
+	local y=$2
+	local dir=$DIR/$tdir
+	local ref
+	local negref
+
+	if [ $y == "t" ]; then
+		ref="\"$(date +"%Y-%m-%d %H:%M:%S")\""
+	else
+		ref=$DIR/$tfile.newer
+		touch $ref || error "touch $ref failed"
+	fi
+	sleep 2
+	setup_56 $dir $NUMFILES $NUMDIRS "-i0 -c1" "-i0 -c1"
+	sleep 2
+	if [ $y == "t" ]; then
+		negref="\"$(date +"%Y-%m-%d %H:%M:%S")\""
+	else
+		negref=$DIR/$tfile.newerneg
+		touch $negref || error "touch $negref failed"
+	fi
+
+	local cmd="$LFS find $dir -newer$x$y $ref"
+	local nums=$(eval $cmd | wc -l)
+	local expected=$(((NUMFILES + 2) * NUMDIRS + 1))
+
+	[ $nums -eq $expected ] ||
+		error "'$cmd' wrong: found $nums, expected $expected"
+
+	cmd="$LFS find $dir ! -newer$x$y $negref"
+	nums=$(eval $cmd | wc -l)
+	[ $nums -eq $expected ] ||
+		error "'$cmd' wrong: found $nums, expected $expected"
+
+	cmd="$LFS find $dir -newer$x$y $ref ! -newer$x$y $negref"
+	nums=$(eval $cmd | wc -l)
+	[ $nums -eq $expected ] ||
+		error "'$cmd' wrong: found $nums, expected $expected"
+
+	rm -rf $DIR/*
+}
+
+test_56oc() {
+	test_newerXY_base "a" "a"
+	test_newerXY_base "a" "m"
+	test_newerXY_base "a" "c"
+	test_newerXY_base "m" "a"
+	test_newerXY_base "m" "m"
+	test_newerXY_base "m" "c"
+	test_newerXY_base "c" "a"
+	test_newerXY_base "c" "m"
+	test_newerXY_base "c" "c"
+	test_newerXY_base "a" "t"
+	test_newerXY_base "m" "t"
+	test_newerXY_base "c" "t"
+}
+run_test 56oc "check lfs find -newerXY work"
+
 test_56p() {
 	[ $RUNAS_ID -eq $UID ] &&
 		skip_env "RUNAS_ID = UID = $UID -- skipping"
@@ -7773,7 +7839,7 @@ run_test 65m "normal user can't set filesystem default stripe"
 
 test_65n() {
 	[ -n "$FILESET" ] && skip "Not functional for FILESET set"
-	[[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.12.50) ]] ||
+	[[ $MDS1_VERSION -ge $(version_code 2.12.50) ]] ||
 		skip "Need MDS version at least 2.12.50"
 	[[ $PARALLEL != "yes" ]] || skip "skip parallel run"
 
@@ -7816,8 +7882,13 @@ test_65n() {
 
 	local dir3=$MOUNT/$tdir-3
 	mkdir $dir3 || error "mkdir $dir3 failed"
-	! getfattr -n trusted.lov $dir3 &> /dev/null ||
-		error "$dir3 shouldn't have LOV EA"
+	# $dir3 shouldn't have LOV EA, but "lfs getstripe -d $dir3" should show
+	# the root layout, which is the actual default layout that will be used
+	# when new files are created in $dir3.
+	local dir3_layout=$(get_layout_param $dir3)
+	local root_dir_layout=$(get_layout_param $MOUNT)
+	[[ "$dir3_layout" = "$root_dir_layout" ]] ||
+		error "$dir3 should show the default layout from $MOUNT"
 
 	# set OST pool on root directory
 	local pool=$TESTNAME
@@ -7839,8 +7910,14 @@ test_65n() {
 
 	local dir4=$MOUNT/$tdir-4
 	mkdir $dir4 || error "mkdir $dir4 failed"
-	! getfattr -n trusted.lov $dir4 &> /dev/null ||
-		error "$dir4 shouldn't have LOV EA"
+	local dir4_layout=$(get_layout_param $dir4)
+	root_dir_layout=$(get_layout_param $MOUNT)
+	echo "$LFS getstripe -d $dir4"
+	$LFS getstripe -d $dir4
+	echo "$LFS getstripe -d $MOUNT"
+	$LFS getstripe -d $MOUNT
+	[[ "$dir4_layout" = "$root_dir_layout" ]] ||
+		error "$dir4 should show the default layout from $MOUNT"
 
 	# new file created in $dir4 should inherit the pool from
 	# the filesystem default
@@ -7859,7 +7936,7 @@ test_65n() {
 	local dir5=$dir4/$tdir-5
 	mkdir $dir5 || error "mkdir $dir5 failed"
 
-	local dir4_layout=$(get_layout_param $dir4)
+	dir4_layout=$(get_layout_param $dir4)
 	local dir5_layout=$(get_layout_param $dir5)
 	[[ "$dir4_layout" = "$dir5_layout" ]] ||
 		error "$dir5 should inherit the default layout from $dir4"
@@ -8521,31 +8598,35 @@ test_80() { # bug 10718
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 
 	# relax strong synchronous semantics for slow backends like ZFS
-	local soc="obdfilter.*.sync_on_lock_cancel"
-	local soc_old=$(do_facet ost1 lctl get_param -n $soc | head -n1)
-	local hosts=
-	if [ "$soc_old" != "never" ] &&
-		[ "$ost1_FSTYPE" != "ldiskfs" ]; then
-			hosts=$(for host in $(seq -f "ost%g" 1 $OSTCOUNT); do
-				facet_active_host $host; done | sort -u)
-			do_nodes $hosts lctl set_param $soc=never
-	fi
+	if [ "$ost1_FSTYPE" != "ldiskfs" ]; then
+		local soc="obdfilter.*.sync_lock_cancel"
+		local save=$(do_facet ost1 $LCTL get_param -n $soc | head -n1)
 
-        dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M
-        sync; sleep 1; sync
-        local BEFORE=`date +%s`
-        cancel_lru_locks osc
-        local AFTER=`date +%s`
-        local DIFF=$((AFTER-BEFORE))
-        if [ $DIFF -gt 1 ] ; then
-                error "elapsed for 1M@1T = $DIFF"
-        fi
+		# "sync_on_lock_cancel" was broken by v2_11_55_0-26-g7059644e9a
+		if [ -z "$save" ]; then
+			soc="obdfilter.*.sync_on_lock_cancel"
+			save=$(do_facet ost1 $LCTL get_param -n $soc | head -n1)
+		fi
 
-        [ -n "$hosts" ] && do_nodes $hosts lctl set_param $soc=$soc_old
+		if [ "$save" != "never" ]; then
+			local hosts=$(comma_list $(osts_nodes))
 
-        rm -f $DIR/$tfile
+			do_nodes $hosts $LCTL set_param $soc=never
+			stack_trap "do_nodes $hosts $LCTL set_param $soc=$save"
+		fi
+	fi
+
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 seek=1M
+	sync; sleep 1; sync
+	local before=$(date +%s)
+	cancel_lru_locks osc
+	local after=$(date +%s)
+	local diff=$((after - before))
+	[ $diff -le 1 ] || error "elapsed for 1M@1T = $diff"
+
+	rm -f $DIR/$tfile
 }
-run_test 80 "Page eviction is equally fast at high offsets too  ===="
+run_test 80 "Page eviction is equally fast at high offsets too"
 
 test_81a() { # LU-456
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
@@ -8851,11 +8932,6 @@ test_101c() {
 }
 run_test 101c "check stripe_size aligned read-ahead ================="
 
-set_read_ahead() {
-	$LCTL get_param -n llite.*.max_read_ahead_mb | head -n 1
-	$LCTL set_param -n llite.*.max_read_ahead_mb $1 > /dev/null 2>&1
-}
-
 test_101d() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 
@@ -8875,7 +8951,10 @@ test_101d() {
 	cancel_lru_locks osc
 
 	echo Disable read-ahead
-	local old_READAHEAD=$(set_read_ahead 0)
+	local old_RA=$($LCTL get_param -n llite.*.max_read_ahead_mb | head -n 1)
+	$LCTL set_param -n llite.*.max_read_ahead_mb=0
+	stack_trap "$LCTL set_param -n llite.*.max_read_ahead_mb $old_RA" EXIT
+	$LCTL get_param -n llite.*.max_read_ahead_mb
 
 	echo Reading the test file $file with read-ahead disabled
 	local raOFF=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB")
@@ -8883,7 +8962,7 @@ test_101d() {
 	echo Cancel LRU locks on lustre client to flush the client cache
 	cancel_lru_locks osc
 	echo Enable read-ahead with ${ra_MB}MB
-	set_read_ahead $ra_MB
+	$LCTL set_param -n llite.*.max_read_ahead_mb=$ra_MB
 
 	echo Reading the test file $file with read-ahead enabled
 	local raON=$(do_and_time "dd if=$file of=/dev/null bs=1M count=$sz_MB")
@@ -8891,7 +8970,6 @@ test_101d() {
 	echo "read-ahead disabled time read $raOFF"
 	echo "read-ahead enabled  time read $raON"
 
-	set_read_ahead $old_READAHEAD
 	rm -f $file
 	wait_delete_completed
 
@@ -11082,6 +11160,27 @@ test_123b () { # statahead(bug 15027)
 }
 run_test 123b "not panic with network error in statahead enqueue (bug 15027)"
 
+test_123c() {
+	[[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs"
+
+	test_mkdir -i 0 -c 1 $DIR/$tdir.0
+	test_mkdir -i 1 -c 1 $DIR/$tdir.1
+	touch $DIR/$tdir.1/{1..3}
+	mv $DIR/$tdir.1/{1..3} $DIR/$tdir.0
+
+	remount_client $MOUNT
+
+	$MULTIOP $DIR/$tdir.0 Q
+
+	# let statahead to complete
+	ls -l $DIR/$tdir.0 > /dev/null
+
+	testid=$(echo $TESTNAME | tr '_' ' ')
+	dmesg | tac | sed "/$testid/,$ d" | grep "Can not initialize inode" &&
+		error "statahead warning" || true
+}
+run_test 123c "Can not initialize inode warning on DNE statahead"
+
 test_124a() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 	$LCTL get_param -n mdc.*.connect_flags | grep -q lru_resize ||
@@ -12392,33 +12491,22 @@ test_133g() {
 	remote_mds_nodsh && skip "remote MDS with nodsh"
 	remote_ost_nodsh && skip "remote OST with nodsh"
 
-	# eventually, this can also be replaced with "lctl get_param -R",
-	# but not until that option is always available on the server
 	local facet
 	for facet in mds1 ost1; do
-		[ $(lustre_version_code $facet) -le $(version_code 2.5.54) ] &&
-			skip_noexit "Too old lustre on $facet"
-		local facet_proc_dirs=$(do_facet $facet \
-					\\\ls -d $proc_regexp 2>/dev/null)
-		echo "${facet}_proc_dirs='$facet_proc_dirs'"
-		[ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet"
-		do_facet $facet find $facet_proc_dirs \
-			! -name req_history \
-			-exec cat '{}' \\\; &> /dev/null
-
-		do_facet $facet find $facet_proc_dirs \
-			! -name req_history \
-			-type f \
-			-exec cat '{}' \\\; &> /dev/null ||
-				error "proc file read failed"
-
-		do_facet $facet find $facet_proc_dirs \
-			-ignore_readdir_race \
-			-type f \
-			-not -name force_lbug \
-			-not -name changelog_mask \
-			-exec badarea_io '{}' \\\; ||
-				error_133 "$facet find $facet_proc_dirs failed"
+		local facet_ver=$(lustre_version_code $facet)
+		if [ $facet_ver -ge $(version_code 2.7.65) ]; then
+			do_facet $facet "$LCTL get_param -R '*'" &> /dev/null
+		else
+			log "$facet: too old lustre for get_param -R"
+		fi
+		if [ $facet_ver -ge $(version_code 2.5.54) ]; then
+			do_facet $facet "$LCTL list_param -R '*' | grep '=' |
+				tr -d= | egrep -v 'force_lbug|changelog_mask' |
+				xargs badarea_io" ||
+					error_133 "$facet badarea_io failed"
+		else
+			skip_noexit "$facet: too old lustre for get_param -R"
+		fi
 	done
 
 	# remount the FS in case writes/reads /proc break the FS
@@ -12534,6 +12622,73 @@ test_134b() {
 }
 run_test 134b "Server rejects lock request when reaching lock_limit_mb"
 
+test_135() {
+	remote_mds_nodsh && skip "remote MDS with nodsh"
+	[[ $MDS1_VERSION -lt $(version_code 2.13.50) ]] &&
+		skip "Need MDS version at least 2.13.50"
+	local fname
+
+	mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir"
+
+#define OBD_FAIL_PLAIN_RECORDS 0x1319
+	#set only one record at plain llog
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0x1319 fail_val=1
+
+	#fill already existed plain llog each 64767
+	#wrapping whole catalog
+	createmany -o -u $DIR/$tdir/$tfile- $((64767 * 1))
+
+	createmany -o $DIR/$tdir/$tfile_ 64700
+	for (( i = 0; i < 64700; i = i + 2 ))
+	do
+		rm $DIR/$tdir/$tfile_$i &
+		rm $DIR/$tdir/$tfile_$((i + 1)) &
+		local pid=$!
+		wait $pid
+	done
+
+	#waiting osp synchronization
+	wait_delete_completed
+}
+run_test 135 "Race catalog processing"
+
+test_136() {
+	remote_mds_nodsh && skip "remote MDS with nodsh"
+	[[ $MDS1_VERSION -lt $(version_code 2.13.50) ]] &&
+		skip "Need MDS version at least 2.13.50"
+	local fname
+
+	mkdir -p $DIR/$tdir || error "failed to create $DIR/$tdir"
+	$SETSTRIPE -c 1 -i 0 $DIR/$tdir || error "failed to set striping"
+	#set only one record at plain llog
+#define OBD_FAIL_CATALOG_FULL_CHECK                0x131a
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0x131a fail_val=1
+
+	#fill already existed 2 plain llogs each 64767
+	#wrapping whole catalog
+	createmany -o -u $DIR/$tdir/$tfile- $((64767 * 1))
+	createmany -o -u $DIR/$tdir/$tfile- $((64767 * 3 / 2))
+	wait_delete_completed
+
+	createmany -o $DIR/$tdir/$tfile_ 10
+	sleep 25
+
+	do_facet $SINGLEMDS $LCTL set_param fail_val=3
+	for (( i = 0; i < 10; i = i + 3 ))
+	do
+		rm $DIR/$tdir/$tfile_$i &
+		rm $DIR/$tdir/$tfile_$((i + 1)) &
+		local pid=$!
+		wait $pid
+		sleep 7
+		rm $DIR/$tdir/$tfile_$((i + 2)) &
+	done
+
+	#waiting osp synchronization
+	wait_delete_completed
+}
+run_test 136 "Race catalog processing 2"
+
 test_140() { #bug-17379
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 
@@ -13718,14 +13873,17 @@ test_160f() {
 
 	# generate some changelog records to accumulate on each MDT
 	test_mkdir -c $MDSCOUNT $DIR/$tdir || error "test_mkdir $tdir failed"
+	log "$(date +%s): creating first files"
 	createmany -m $DIR/$tdir/$tfile $((MDSCOUNT * 2)) ||
 		error "create $DIR/$tdir/$tfile failed"
 
 	# check changelogs have been generated
+	local start=$SECONDS
+	local idle_time=$((MDSCOUNT * 5 + 5))
 	local nbcl=$(changelog_dump | wc -l)
 	[[ $nbcl -eq 0 ]] && error "no changelogs found"
 
-	for param in "changelog_max_idle_time=10" \
+	for param in "changelog_max_idle_time=$idle_time" \
 		     "changelog_gc=1" \
 		     "changelog_min_gc_interval=2" \
 		     "changelog_min_free_cat_entries=3"; do
@@ -13737,8 +13895,11 @@ test_160f() {
 		do_nodes $mdts $LCTL set_param mdd.*.$param
 	done
 
-	# force cl_user2 to be idle (1st part)
-	sleep 9
+	# force cl_user2 to be idle (1st part), but also cancel the
+	# cl_user1 records so that it is not evicted later in the test.
+	local sleep1=$((idle_time / 2))
+	echo "$(date +%s): sleep1 $sleep1/${idle_time}s"
+	sleep $sleep1
 
 	# simulate changelog catalog almost full
 	#define OBD_FAIL_CAT_FREE_RECORDS	0x1313
@@ -13774,13 +13935,16 @@ test_160f() {
 			      "$user_rec1, but is $user_rec2"
 	done
 
-	# force cl_user2 to be idle (2nd part) and to reach
-	# changelog_max_idle_time
-	sleep 2
+	# force cl_user2 idle (2nd part) to just exceed changelog_max_idle_time
+	local sleep2=$((idle_time - (SECONDS - start) + 1))
+	echo "$(date +%s): sleep2 $sleep2/${idle_time}s"
+	sleep $sleep2
 
-	# generate one more changelog to trigger fail_loc
-	createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) ||
-		error "create $DIR/$tdir/${tfile}bis failed"
+	# Generate one more changelog to trigger GC at fail_loc for cl_user2.
+	# cl_user1 should be OK because it recently processed records.
+	echo "$(date +%s): creating $((MDSCOUNT * 2)) files"
+	createmany -m $DIR/$tdir/${tfile}b $((MDSCOUNT * 2)) ||
+		error "create $DIR/$tdir/${tfile}b failed"
 
 	# ensure gc thread is done
 	for i in $(mdts_nodes); do
@@ -14156,8 +14320,10 @@ test_160j() {
 		skip "Need MDS version at least 2.12.56"
 
 	mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed"
+	stack_trap "umount $MOUNT2" EXIT
 
 	changelog_register || error "first changelog_register failed"
+	stack_trap "changelog_deregister" EXIT
 
 	# generate some changelog
 	test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed"
@@ -14166,10 +14332,13 @@ test_160j() {
 
 	# open the changelog device
 	exec 3>/dev/changelog-$FSNAME-MDT0000
+	stack_trap "exec 3>&-" EXIT
 	exec 4</dev/changelog-$FSNAME-MDT0000
+	stack_trap "exec 4<&-" EXIT
 
 	# umount the first lustre mount
 	umount $MOUNT
+	stack_trap "mount_client $MOUNT" EXIT
 
 	# read changelog
 	cat <&4 >/dev/null || error "read changelog failed"
@@ -14181,15 +14350,6 @@ test_160j() {
 
 	printf 'clear:'$cl_user':0' >&3
 
-	# close
-	exec 3>&-
-	exec 4<&-
-
-	# cleanup
-	changelog_deregister || error "changelog_deregister failed"
-
-	umount $MOUNT2
-	mount_client $MOUNT || error "mount_client on $MOUNT failed"
 }
 run_test 160j "client can be umounted  while its chanangelog is being used"
 
@@ -15383,7 +15543,7 @@ jobstats_set() {
 		"$FSNAME.sys.jobid_var" $new_jobenv
 }
 
-test_205() { # Job stats
+test_205a() { # Job stats
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
 	[[ $MDS1_VERSION -ge $(version_code 2.7.1) ]] ||
 		skip "Need MDS version with at least 2.7.1"
@@ -15485,7 +15645,18 @@ test_205() { # Job stats
 
 	verify_jobstats "touch $DIR/$tfile" $SINGLEMDS
 }
-run_test 205 "Verify job stats"
+run_test 205a "Verify job stats"
+
+# LU-13117
+test_205b() {
+	$LCTL set_param jobid_var=USER jobid_name="%e.%u"
+	env -i USERTESTJOBSTATS=foolish touch $DIR/$tfile.1
+	do_facet $SINGLEMDS $LCTL get_param mdt.*.job_stats |
+		grep job_id: | grep foolish &&
+			error "Unexpected jobid found"
+	true
+}
+run_test 205b "Verify job stats jobid parsing"
 
 # LU-1480, LU-1773 and LU-1657
 test_206() {
@@ -16022,8 +16193,7 @@ test_224c() { # LU-6441
 	save_writethrough $p
 	set_cache writethrough on
 
-	local pages_per_rpc=$($LCTL get_param \
-				osc.*.max_pages_per_rpc)
+	local pages_per_rpc=$($LCTL get_param osc.*.max_pages_per_rpc)
 	local at_max=$($LCTL get_param -n at_max)
 	local timeout=$($LCTL get_param -n timeout)
 	local test_at="at_max"
@@ -18908,7 +19078,7 @@ test_272a() {
 		error "md5sum differ: $old_md5, $new_md5"
 
 	[ $($LFS getstripe -c $dom) -eq 2 ] ||
-		error "migrate stripe count bad: $(LFS getstripe -c $dom) != 2"
+		error "bad final stripe count: $($LFS getstripe -c $dom) != 2"
 }
 run_test 272a "DoM migration: new layout with the same DOM component"
 
@@ -19698,8 +19868,9 @@ test_300k() {
 
 	# this test needs a huge transaction
 	local kb
-	kb=$(do_facet $SINGLEMDS lctl get_param -n osd*.lustre-MDT0000.kbytestotal)
-	[ $kb -lt $((1024*1024)) ] && skip "too small mds: $kb"
+	kb=$(do_facet $SINGLEMDS "$LCTL get_param -n \
+	     osd*.$FSNAME-MDT0000.kbytestotal")
+	[ $kb -lt $((1024*1024)) ] && skip "MDT0 too small: $kb"
 
 	local stripe_count
 	local file
@@ -19907,7 +20078,7 @@ test_300q() {
 run_test 300q "create remote directory under orphan directory"
 
 test_300r() {
-	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.55) ] &&
+	[ $MDS1_VERSION -lt $(version_code 2.7.55) ] &&
 		skip "Need MDS version at least 2.7.55" && return
 	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
 
@@ -20365,6 +20536,105 @@ test_319() {
 }
 run_test 319 "lost lease lock on migrate error"
 
+test_398a() { # LU-4198
+	$LFS setstripe -c 1 -i 0 $DIR/$tfile
+	$LCTL set_param ldlm.namespaces.*.lru_size=clear
+
+	# request a new lock on client
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1
+
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc
+	local lock_count=$($LCTL get_param -n \
+			   ldlm.namespaces.*-OST0000-osc-ffff*.lru_size)
+	[[ $lock_count -eq 0 ]] || error "lock should be cancelled by direct IO"
+
+	$LCTL set_param ldlm.namespaces.*-OST0000-osc-ffff*.lru_size=clear
+
+	# no lock cached, should use lockless IO and not enqueue new lock
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 oflag=direct conv=notrunc
+	lock_count=$($LCTL get_param -n \
+		     ldlm.namespaces.*-OST0000-osc-ffff*.lru_size)
+	[[ $lock_count -eq 0 ]] || error "no lock should be held by direct IO"
+}
+run_test 398a "direct IO should cancel lock otherwise lockless"
+
+test_398b() { # LU-4198
+	which fio || skip_env "no fio installed"
+	$LFS setstripe -c -1 $DIR/$tfile
+
+	local size=12
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=$size
+
+	local njobs=4
+	echo "mix direct rw ${size}M to OST0 by fio with $njobs jobs..."
+	fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE --direct=1 \
+		--numjobs=$njobs --fallocate=none \
+		--iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+		--filename=$DIR/$tfile &
+	bg_pid=$!
+
+	echo "mix buffer rw ${size}M to OST0 by fio with $njobs jobs..."
+	fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE \
+		--numjobs=$njobs --fallocate=none \
+		--iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+		--filename=$DIR/$tfile || true
+	wait $bg_pid
+
+	rm -rf $DIR/$tfile
+}
+run_test 398b "DIO and buffer IO race"
+
+test_398c() { # LU-4198
+	which fio || skip_env "no fio installed"
+
+	saved_debug=$($LCTL get_param -n debug)
+	$LCTL set_param debug=0
+
+	local size=$(lctl get_param -n osc.$FSNAME-OST0000*.kbytesavail | head -1)
+	((size /= 1024)) # by megabytes
+	((size /= 2)) # write half of the OST at most
+	[ $size -gt 40 ] && size=40 #reduce test time anyway
+
+	$LFS setstripe -c 1 $DIR/$tfile
+
+	# it seems like ldiskfs reserves more space than necessary if the
+	# writing blocks are not mapped, so it extends the file firstly
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=$size && sync
+	cancel_lru_locks osc
+
+	# clear and verify rpc_stats later
+	$LCTL set_param osc.${FSNAME}-OST0000-osc-ffff*.rpc_stats=clear
+
+	local njobs=4
+	echo "writing ${size}M to OST0 by fio with $njobs jobs..."
+	fio --name=rand-write --rw=randwrite --bs=$PAGE_SIZE --direct=1 \
+		--numjobs=$njobs --fallocate=none --ioengine=libaio \
+		--iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+		--filename=$DIR/$tfile
+	[ $? -eq 0 ] || error "fio write error"
+
+	[ $($LCTL get_param -n \
+	 ldlm.namespaces.${FSNAME}-OST0000-osc-ffff*.lock_count) -eq 0 ] ||
+		error "Locks were requested while doing AIO"
+
+	# get the percentage of 1-page I/O
+	pct=$($LCTL get_param osc.${FSNAME}-OST0000-osc-ffff*.rpc_stats |
+		grep -A 1 'pages per rpc' | grep -v 'pages per rpc' |
+		awk '{print $7}')
+	[ $pct -le 50 ] || error "$pct% of I/O are 1-page"
+
+	echo "mix rw ${size}M to OST0 by fio with $njobs jobs..."
+	fio --name=rand-rw --rw=randrw --bs=$PAGE_SIZE --direct=1 \
+		--numjobs=$njobs --fallocate=none --ioengine=libaio \
+		--iodepth=16 --allow_file_create=0 --size=$((size/njobs))M \
+		--filename=$DIR/$tfile
+	[ $? -eq 0 ] || error "fio mixed read write error"
+
+	rm -rf $DIR/$tfile
+	$LCTL set_param debug="$saved_debug"
+}
+run_test 398c "run fio to test AIO"
+
 test_fake_rw() {
 	local read_write=$1
 	if [ "$read_write" = "write" ]; then
@@ -20381,7 +20651,7 @@ test_fake_rw() {
 
 	$LFS setstripe -c 1 -i 0 $DIR/$tfile
 
-	# get ost1 size - lustre-OST0000
+	# get ost1 size - $FSNAME-OST0000
 	local ost1_avail_size=$($LFS df | awk /${ost1_svc}/'{ print $4 }')
 	local blocks=$((ost1_avail_size/2/1024)) # half avail space by megabytes
 	[ $blocks -gt 1000 ] && blocks=1000 # 1G in maximum
@@ -20878,7 +21148,7 @@ run_test 411 "Slab allocation error with cgroup does not LBUG"
 
 test_412() {
 	[ $MDSCOUNT -lt 2 ] && skip_env "needs >= 2 MDTs"
-	if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then
+	if [ $MDS1_VERSION -lt $(version_code 2.10.55) ]; then
 		skip "Need server version at least 2.10.55"
 	fi
 
@@ -20923,13 +21193,13 @@ test_qos_mkdir() {
 		lmv.*.qos_maxage=$lmv_qos_maxage > /dev/null" EXIT
 
 	lod_qos_prio_free=$(do_facet mds1 $LCTL get_param -n \
-		lod.lustre-MDT0000-mdtlov.mdt_qos_prio_free | head -n1)
+		lod.$FSNAME-MDT0000-mdtlov.mdt_qos_prio_free | head -n1)
 	lod_qos_prio_free=${lod_qos_prio_free%%%}
 	lod_qos_threshold_rr=$(do_facet mds1 $LCTL get_param -n \
-		lod.lustre-MDT0000-mdtlov.mdt_qos_threshold_rr | head -n1)
+		lod.$FSNAME-MDT0000-mdtlov.mdt_qos_threshold_rr | head -n1)
 	lod_qos_threshold_rr=${lod_qos_threshold_rr%%%}
 	lod_qos_maxage=$(do_facet mds1 $LCTL get_param -n \
-		lod.lustre-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }')
+		lod.$FSNAME-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }')
 	stack_trap "do_nodes $mdts $LCTL set_param \
 		lod.*.mdt_qos_prio_free=$lod_qos_prio_free > /dev/null" EXIT
 	stack_trap "do_nodes $mdts $LCTL set_param \
@@ -21146,7 +21416,7 @@ run_test 414 "simulate ENOMEM in ptlrpc_register_bulk()"
 
 test_415() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run"
-	[ $(lustre_version_code mds1) -lt $(version_code 2.11.52) ] &&
+	[ $MDS1_VERSION -lt $(version_code 2.11.52) ] &&
 		skip "Need server version at least 2.11.52"
 
 	# LU-11102
@@ -21188,7 +21458,7 @@ test_415() {
 run_test 415 "lock revoke is not missing"
 
 test_416() {
-	[ $(lustre_version_code mds1) -lt $(version_code 2.11.55) ] &&
+	[ $MDS1_VERSION -lt $(version_code 2.11.55) ] &&
 		skip "Need server version at least 2.11.55"
 
 	# define OBD_FAIL_OSD_TXN_START    0x19a
@@ -21643,7 +21913,7 @@ test_422() {
 run_test 422 "kill a process with RPC in progress"
 
 prep_801() {
-	[[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] ||
+	[[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
 	[[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
 		skip "Need server version at least 2.9.55"
 
@@ -21874,7 +22144,7 @@ cleanup_802a() {
 
 test_802a() {
 	[[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test"
-	[[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] ||
+	[[ $MDS1_VERSION -lt $(version_code 2.9.55) ]] ||
 	[[ $OST1_VERSION -lt $(version_code 2.9.55) ]] &&
 		skip "Need server version at least 2.9.55"
 
@@ -22101,7 +22371,7 @@ test_805() {
 	local usedkb
 	local old
 	local quota
-	local pref="osd-zfs.lustre-MDT0000."
+	local pref="osd-zfs.$FSNAME-MDT0000."
 
 	# limit available space on MDS dataset to meet nospace issue
 	# quickly. then ZFS 0.7.2 can use reserved space if asked
@@ -22393,7 +22663,7 @@ test_810() {
 run_test 810 "partial page writes on ZFS (LU-11663)"
 
 test_811() {
-	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.11.56) ] &&
+	[ $MDS1_VERSION -lt $(version_code 2.11.56) ] &&
 		skip "Need MDS version at least 2.11.56"
 
 	#define OBD_FAIL_MDS_ORPHAN_DELETE	0x165
@@ -22403,13 +22673,12 @@ test_811() {
 	stop mds1
 	start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
 
-	sleep 5
-	[[ $(do_facet mds1 pgrep orph_.*-MDD | wc -l) -eq 0 ]] ||
+	wait_update_facet mds1 "pgrep orph_.*-MDD | wc -l" "0" ||
 		error "MDD orphan cleanup thread not quit"
 }
 run_test 811 "orphan name stub can be cleaned up in startup"
 
-test_812() {
+test_812a() {
 	[ $OST1_VERSION -lt $(version_code 2.12.51) ] &&
 		skip "OST < 2.12.51 doesn't support this fail_loc"
 	[ "$SHARED_KEY" = true ] &&
@@ -22430,7 +22699,31 @@ test_812() {
 
 	stat $DIR/$tfile >/dev/null || error "can't stat file"
 }
-run_test 812 "do not drop reqs generated when imp is going to idle (LU-11951)"
+run_test 812a "do not drop reqs generated when imp is going to idle (LU-11951)"
+
+test_812b() { # LU-12378
+	[ $OST1_VERSION -lt $(version_code 2.12.51) ] &&
+		skip "OST < 2.12.51 doesn't support this fail_loc"
+	[ "$SHARED_KEY" = true ] &&
+		skip "OSC connections never go IDLE with Shared-Keys enabled"
+
+	$LFS setstripe -c 1 -i 0 $DIR/$tfile || error "setstripe failed"
+	# ensure ost1 is connected
+	stat $DIR/$tfile >/dev/null || error "can't stat"
+	wait_osc_import_state client ost1 FULL
+	# no locks, no reqs to let the connection idle
+	cancel_lru_locks osc
+
+	# delay OST_DISCONNECT on OST1 to put OSC into intermediate state
+#define OBD_FAIL_OST_DISCONNECT_DELAY	 0x245
+	do_facet ost1 "$LCTL set_param fail_loc=0x245 fail_val=8"
+	wait_osc_import_state client ost1 CONNECTING
+	do_facet ost1 "$LCTL set_param fail_loc=0 fail_val=0"
+
+	$LFS quota -u 0 $DIR/ || error "lfs quota should succeed"
+	wait_osc_import_state client ost1 IDLE
+}
+run_test 812b "do not drop no resend request for idle connect"
 
 test_813() {
 	local file_heat_sav=$($LCTL get_param -n llite.*.file_heat 2>/dev/null)
@@ -22700,6 +22993,42 @@ test_819b() {
 }
 run_test 819b "too big niobuf in write"
 
+
+function test_820_start_ost() {
+	sleep 5
+
+	for num in $(seq $OSTCOUNT); do
+		start ost$num $(ostdevname $num) $OST_MOUNT_OPTS
+	done
+}
+
+test_820() {
+	[[ $MDSCOUNT -lt 2 ]] && skip_env "needs >= 2 MDTs"
+
+	mkdir $DIR/$tdir
+	umount_client $MOUNT || error "umount failed"
+	for num in $(seq $OSTCOUNT); do
+		stop ost$num
+	done
+
+	# mount client with no active OSTs
+	# so that the client can't initialize max LOV EA size
+	# from OSC notifications
+	mount_client $MOUNT || error "mount failed"
+	# delay OST starting to keep this 0 max EA size for a while
+	test_820_start_ost &
+
+	# create a directory on MDS2
+	test_mkdir -i 1 -c1 $DIR/$tdir/mds2 ||
+		error "Failed to create directory"
+	# open intent should update default EA size
+	# see mdc_update_max_ea_from_body()
+	# notice this is the very first RPC to MDS2
+	cp /etc/services $DIR/$tdir/mds2 ||
+		error "Failed to copy files to mds$n"
+}
+run_test 820 "update max EA from open intent"
+
 #
 # tests that do cleanup/setup should be run at the end
 #