X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=fb48a6727af56420ace035e778558210cc447293;hp=9f333841dab621f02243a375a902d375953600bb;hb=4a8e06fb5af88f93a936ed8ba0718ff5d3554c9f;hpb=4fed33473ca2964ff19f61fdb8501b2210f923de;ds=sidebyside

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
old mode 100644
new mode 100755
index 9f33384..fb48a67
--- 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)
@@ -64,13 +64,19 @@ if [[ $(uname -m) = aarch64 ]]; then
 fi
 
 # skip nfs tests on kernels >= 4.14.0 until they are fixed
-if [ $LINUX_VERSION_CODE -ge $(version_code 4.14.0) ];then
+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:
@@ -550,11 +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"
@@ -7881,8 +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
@@ -7904,8 +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 $dir4 should inherit the pool from
 	# the filesystem default
@@ -7924,7 +7936,7 @@ test_65n() {
 	local dir5=$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"
@@ -8586,31 +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"
@@ -11147,6 +11163,27 @@ 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 ||
@@ -12457,33 +12494,22 @@ 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
@@ -12599,6 +12625,73 @@ 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"
 
@@ -13783,14 +13876,17 @@ 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
@@ -13802,8 +13898,11 @@ 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
@@ -13839,13 +13938,16 @@ 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 $(mdts_nodes); do
@@ -15448,7 +15550,7 @@ jobstats_set() {
 		"$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"
@@ -15550,7 +15652,18 @@ 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() {
@@ -18973,7 +19086,7 @@ 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"