X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=8cdf3fe2a26a3a9f7b8ad50c38e47392c61a1088;hb=82e494a36e9ea4f51ec163ab15beb9fdda7fa8d6;hp=46f9e8c4eb6f8cd7528e97d0df570154bc8efdb5;hpb=3442db6faf685fbdbd092bdfdc8d273e4990a141;p=fs%2Flustre-release.git

diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh
index 46f9e8c..8cdf3fe 100755
--- a/lustre/tests/sanity.sh
+++ b/lustre/tests/sanity.sh
@@ -5791,6 +5791,92 @@ test_64c() {
 }
 run_test 64c "verify grant shrink"
 
+# this does exactly what osc_request.c:osc_announce_cached() does in
+# order to calculate max amount of grants to ask from server
+want_grant() {
+	local tgt=$1
+
+	local page_size=$(get_page_size client)
+
+	local nrpages=$($LCTL get_param -n osc.${tgt}.max_pages_per_rpc)
+	local rpc_in_flight=$($LCTL get_param -n osc.${tgt}.max_rpcs_in_flight)
+
+	((rpc_in_flight ++));
+	nrpages=$((nrpages * rpc_in_flight))
+
+	local dirty_max_pages=$($LCTL get_param -n osc.${tgt}.max_dirty_mb)
+
+	dirty_max_pages=$((dirty_max_pages * 1024 * 1024 / page_size))
+
+	[[ $dirty_max_pages -gt $nrpages ]] && nrpages=$dirty_max_pages
+	local undirty=$((nrpages * page_size))
+
+	local max_extent_pages
+	max_extent_pages=$($LCTL get_param osc.${tgt}.import |
+	    grep grant_max_extent_size | awk '{print $2}')
+	max_extent_pages=$((max_extent_pages / page_size))
+	local nrextents=$(((nrpages + max_extent_pages - 1) / max_extent_pages))
+	local grant_extent_tax
+	grant_extent_tax=$($LCTL get_param osc.${tgt}.import |
+	    grep grant_extent_tax | awk '{print $2}')
+
+	undirty=$((undirty + nrextents * grant_extent_tax))
+
+	echo $undirty
+}
+
+# this is size of unit for grant allocation. It should be equal to
+# what tgt_grant.c:tgt_grant_chunk() calculates
+grant_chunk() {
+	local tgt=$1
+	local max_brw_size
+	local grant_extent_tax
+
+	max_brw_size=$($LCTL get_param osc.${tgt}.import |
+	    grep max_brw_size | awk '{print $2}')
+
+	grant_extent_tax=$($LCTL get_param osc.${tgt}.import |
+	    grep grant_extent_tax | awk '{print $2}')
+
+	echo $(((max_brw_size + grant_extent_tax) * 2))
+}
+
+test_64d() {
+	[ $(lustre_version_code ost1) -lt $(version_code 2.10.56) ] &&
+		skip "OST < 2.10.55 doesn't limit grants enough" && return 0
+
+	local tgt=$($LCTL dl | grep "0000-osc-[^mM]" | awk '{print $4}')
+
+	[[ $($LCTL get_param osc.${tgt}.import |
+		    grep "connect_flags:.*grant_param") ]] || \
+			{ skip "no grant_param connect flag"; return; }
+
+	local olddebug=$($LCTL get_param -n debug 2> /dev/null)
+
+	$LCTL set_param debug="$OLDDEBUG" 2> /dev/null || true
+
+	local max_cur_granted=$(($(want_grant $tgt) + $(grant_chunk $tgt)))
+
+	$SETSTRIPE $DIR/$tfile -i 0 -c 1
+	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1000 &
+	ddpid=$!
+
+	while true
+	do
+		local cur_grant=$($LCTL get_param -n osc.${tgt}.cur_grant_bytes)
+		if [[ $cur_grant -gt $max_cur_granted ]]
+		then
+			kill $ddpid
+			error "cur_grant $cur_grant > $max_cur_granted"
+		fi
+		kill -0 $ddpid
+		[[ $? -ne 0 ]] && break;
+		sleep 2
+	done
+	$LCTL set_param debug="$olddebug" 2> /dev/null || true
+}
+run_test 64d "check grant limit exceed"
+
 # bug 1414 - set/get directories' stripe info
 test_65a() {
 	[ $PARALLEL == "yes" ] && skip "skip parallel run" && return
@@ -7044,14 +7130,19 @@ test_101g() {
 	local list=$(comma_list $(osts_nodes))
 	local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
 	local brw_size="obdfilter.*.brw_size"
+	local ostver=$(lustre_version_code ost1)
+	local cliver=$(lustre_version_code client)
 
 	$LFS setstripe -i 0 -c 1 $DIR/$tfile
 
 	local orig_mb=$(do_facet ost1 $LCTL get_param -n $brw_size | head -n 1)
-	if [ $(lustre_version_code ost1) -ge $(version_code 2.8.52) -a \
-	     $(lustre_version_code client) -ge $(version_code 2.8.52) ]; then
-		[ $(lustre_version_code ost1) -ge $(version_code 2.9.52) ] &&
-			suffix="M"
+	if [ $ostver -ge $(version_code 2.8.52) -o \
+	     \( $ostver -ge $(version_code 2.7.17) -a \
+		$ostver -lt $(version_code 2.7.50) \) ] &&
+	   [ $cliver -ge $(version_code 2.8.52) -o \
+	     \( $cliver -ge $(version_code 2.7.17) -a \
+		$cliver -lt $(version_code 2.7.50) \) ]; then
+		[ $ostver -ge $(version_code 2.9.52) ] && suffix="M"
 		if [[ $orig_mb -lt 16 ]]; then
 			save_lustre_params $osts "$brw_size" > $p
 			do_nodes $list $LCTL set_param -n $brw_size=16$suffix ||
@@ -8507,6 +8598,34 @@ run_test 118m "fdatasync dir ========="
 
 [ "$SLOW" = "no" ] && [ -n "$OLD_RESENDCOUNT" ] && set_resend_count $OLD_RESENDCOUNT
 
+test_118n()
+{
+	local begin
+	local end
+
+	[ $PARALLEL == "yes" ] && skip "skip parallel run" && return
+	remote_ost_nodsh && skip "remote OSTs with nodsh" && return
+
+	# Sleep to avoid a cached response.
+	#define OBD_STATFS_CACHE_SECONDS 1
+	sleep 2
+
+	# Inject a 10 second delay in the OST_STATFS handler.
+	#define OBD_FAIL_OST_STATFS_DELAY 0x242
+	set_nodes_failloc "$(osts_nodes)" 0x242
+
+	begin=$SECONDS
+	stat --file-system $MOUNT > /dev/null
+	end=$SECONDS
+
+	set_nodes_failloc "$(osts_nodes)" 0
+
+	if ((end - begin > 20)); then
+	    error "statfs took $((end - begin)) seconds, expected 10"
+	fi
+}
+run_test 118n "statfs() sends OST_STATFS requests in parallel"
+
 test_119a() # bug 11737
 {
         BSIZE=$((512 * 1024))
@@ -9382,7 +9501,7 @@ test_130a() {
 		skip "ORI-366/LU-1941: FIEMAP unimplemented on ZFS" && return
 	[ $RC != 0 ] && error "filefrag $fm_file failed"
 
-	filefrag_op=$(filefrag -ve $fm_file |
+	filefrag_op=$(filefrag -ve -k $fm_file |
 			sed -n '/ext:/,/found/{/ext:/d; /found/d; p}')
 	lun=$($GETSTRIPE -i $fm_file)
 
@@ -9432,7 +9551,7 @@ test_130b() {
 		error "dd failed on $fm_file"
 
 	filefrag -ves $fm_file || error "filefrag $fm_file failed"
-	filefrag_op=$(filefrag -ve $fm_file |
+	filefrag_op=$(filefrag -ve -k $fm_file |
 			sed -n '/ext:/,/found/{/ext:/d; /found/d; p}')
 
 	last_lun=$(echo $filefrag_op | cut -d: -f5 |
@@ -9491,7 +9610,7 @@ test_130c() {
 			error "dd failed on $fm_file"
 
 	filefrag -ves $fm_file || error "filefrag $fm_file failed"
-	filefrag_op=$(filefrag -ve $fm_file |
+	filefrag_op=$(filefrag -ve -k $fm_file |
 		sed -n '/ext:/,/found/{/ext:/d; /found/d; p}')
 
 	last_lun=$(echo $filefrag_op | cut -d: -f5 |
@@ -9559,7 +9678,7 @@ test_130d() {
 		error "dd failed on $fm_file"
 
 	filefrag -ves $fm_file || error "filefrag $fm_file failed"
-	filefrag_op=$(filefrag -ve $fm_file |
+	filefrag_op=$(filefrag -ve -k $fm_file |
 			sed -n '/ext:/,/found/{/ext:/d; /found/d; p}')
 
 	last_lun=$(echo $filefrag_op | cut -d: -f5 |
@@ -9621,7 +9740,7 @@ test_130e() {
 	done
 
 	filefrag -ves $fm_file || error "filefrag $fm_file failed"
-	filefrag_op=$(filefrag -ve $fm_file |
+	filefrag_op=$(filefrag -ve -k $fm_file |
 			sed -n '/ext:/,/found/{/ext:/d; /found/d; p}')
 
 	last_lun=$(echo $filefrag_op | cut -d: -f5 |
@@ -10063,6 +10182,7 @@ test_133g() {
 	echo "proc_dirs='$proc_dirs'"
 	[ -n "$proc_dirs" ] || error "no proc_dirs on $HOSTNAME"
 	find $proc_dirs \
+		-ignore_readdir_race \
 		-type f \
 		-not -name force_lbug \
 		-not -name changelog_mask \
@@ -10078,6 +10198,7 @@ test_133g() {
 		echo "${facet}_proc_dirs='$facet_proc_dirs'"
 		[ -z "$facet_proc_dirs" ] && error "no proc_dirs on $facet"
 		do_facet $facet find $facet_proc_dirs \
+			-ignore_readdir_race \
 			-type f \
 			-not -name force_lbug \
 			-not -name changelog_mask \
@@ -12238,6 +12359,8 @@ test_184a() {
 
 	cmp $ref1 $file2 || error "content compare failed ($ref1 != $file2)"
 	cmp $ref2 $file1 || error "content compare failed ($ref2 != $file1)"
+
+	lfsck_verify_pfid $file1 $file2 || error "PFID are not transferred"
 }
 run_test 184a "Basic layout swap"
 
@@ -14165,19 +14288,44 @@ test_231b() {
 }
 run_test 231b "must not assert on fully utilized OST request buffer"
 
-test_232() {
+test_232a() {
 	mkdir -p $DIR/$tdir
+	$LFS setstripe -c1 -i0 $DIR/$tdir/$tfile
+
 	#define OBD_FAIL_LDLM_OST_LVB		 0x31c
-	$LCTL set_param fail_loc=0x31c
+	do_facet ost1 $LCTL set_param fail_loc=0x31c
 
 	# ignore dd failure
 	dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=1 || true
 
-	$LCTL set_param fail_loc=0
+	do_facet ost1 $LCTL set_param fail_loc=0
 	umount_client $MOUNT || error "umount failed"
 	mount_client $MOUNT || error "mount failed"
+	stop ost1 || error "cannot stop ost1"
+	start ost1 $(ostdevname 1) $OST_MOUNT_OPTS || error "cannot start ost1"
 }
-run_test 232 "failed lock should not block umount"
+run_test 232a "failed lock should not block umount"
+
+test_232b() {
+	mkdir -p $DIR/$tdir
+	$LFS setstripe -c1 -i0 $DIR/$tdir/$tfile
+	dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=1
+	sync
+	cancel_lru_locks osc
+
+	#define OBD_FAIL_LDLM_OST_LVB		 0x31c
+	do_facet ost1 $LCTL set_param fail_loc=0x31c
+
+	# ignore failure
+	$LFS data_version $DIR/$tdir/$tfile || true
+
+	do_facet ost1 $LCTL set_param fail_loc=0
+	umount_client $MOUNT || error "umount failed"
+	mount_client $MOUNT || error "mount failed"
+	stop ost1 || error "cannot stop ost1"
+	start ost1 $(ostdevname 1) $OST_MOUNT_OPTS || error "cannot start ost1"
+}
+run_test 232b "failed data version lock should not block umount"
 
 test_233a() {
 	[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.64) ] ||
@@ -16701,6 +16849,24 @@ test_314() {
 }
 run_test 314 "OSP shouldn't fail after last_rcvd update failure"
 
+test_315() { # LU-618
+	local file=$DIR/$tfile
+	rm -f $file
+
+	$MULTIOP $file oO_CREAT:O_DIRECT:O_RDWR:w4096000c
+	$MULTIOP $file oO_RDONLY:r4096000_c &
+	PID=$!
+
+	sleep 2
+
+	local rbytes=$(awk '/read_bytes/ { print $2 }' /proc/$PID/io)
+	kill -USR1 $PID
+
+	[ $rbytes -gt 4000000 ] || error "read is not accounted ($rbytes)"
+	rm -f $file
+}
+run_test 315 "read should be accounted"
+
 test_fake_rw() {
 	local read_write=$1
 	if [ "$read_write" = "write" ]; then
@@ -17504,6 +17670,137 @@ test_802() {
 }
 run_test 802 "simulate readonly device"
 
+test_803() {
+	[[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return
+	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.54) ] &&
+		skip "MDS needs to be newer than 2.10.54" && return
+
+	mkdir -p $DIR/$tdir
+	# Create some objects on all MDTs to trigger related logs objects
+	for idx in $(seq $MDSCOUNT); do
+		$LFS mkdir -c $MDSCOUNT -i $((idx % $MDSCOUNT)) \
+			$DIR/$tdir/dir${idx} ||
+			error "Fail to create $DIR/$tdir/dir${idx}"
+	done
+
+	sync; sleep 5
+	echo "before create:"
+	$LFS df -i $MOUNT
+	local before_used=$($LFS df -i | grep MDT0000_UUID | awk '{print $3}')
+
+	for ((i=0; i<10; i++)); do
+		$LFS mkdir -c 1 -i 1 $DIR/$tdir/foo$i ||
+			error "Fail to create $DIR/$tdir/foo$i"
+	done
+
+	sync; sleep 5
+	echo "after create:"
+	$LFS df -i $MOUNT
+	local after_used=$($LFS df -i | grep MDT0000_UUID | awk '{print $3}')
+
+	[ $after_used -ge $((before_used + 10)) ] ||
+		error "before ($before_used) + 10 > after ($after_used)"
+
+	for ((i=0; i<10; i++)); do
+		rm -rf $DIR/$tdir/foo$i ||
+			error "Fail to remove $DIR/$tdir/foo$i"
+	done
+
+	wait_delete_completed
+	echo "after unlink:"
+	$LFS df -i $MOUNT
+	before_used=$after_used
+	after_used=$($LFS df -i | grep MDT0000_UUID | awk '{print $3}')
+
+	[ $after_used -le $((before_used - 8)) ] ||
+		error "before ($before_used) - 8 < after ($after_used)"
+}
+run_test 803 "verify agent object for remote object"
+
+test_804() {
+	[[ $MDSCOUNT -lt 2 ]] && skip "needs >= 2 MDTs" && return
+	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.10.54) ] &&
+		skip "MDS needs to be newer than 2.10.54" && return
+
+	[ "$(facet_fstype $SINGLEMDS)" != "ldiskfs" ] &&
+		skip "ldiskfs only test" && return 0
+
+	mkdir -p $DIR/$tdir
+	$LFS mkdir -c 1 -i 1 $DIR/$tdir/dir0 ||
+		error "Fail to create $DIR/$tdir/dir0"
+
+	local fid=$($LFS path2fid $DIR/$tdir/dir0)
+	local dev=$(mdsdevname 2)
+
+	do_facet mds2 "$DEBUGFS -c -R 'ls /REMOTE_PARENT_DIR' $dev" |
+		grep ${fid} || error "NOT found agent entry for dir0"
+
+	$LFS mkdir -c $MDSCOUNT -i 0 $DIR/$tdir/dir1 ||
+		error "Fail to create $DIR/$tdir/dir1"
+
+	touch $DIR/$tdir/dir1/foo0 ||
+		error "Fail to create $DIR/$tdir/dir1/foo0"
+	fid=$($LFS path2fid $DIR/$tdir/dir1/foo0)
+	local rc=0
+
+	for idx in $(seq $MDSCOUNT); do
+		dev=$(mdsdevname $idx)
+		do_facet mds${idx} \
+			"$DEBUGFS -c -R 'ls /REMOTE_PARENT_DIR' $dev" |
+			grep ${fid} && rc=$idx
+	done
+
+	mv $DIR/$tdir/dir1/foo0 $DIR/$tdir/dir1/foo1 ||
+		error "Fail to rename foo0 to foo1"
+	if [ $rc -eq 0 ]; then
+		for idx in $(seq $MDSCOUNT); do
+			dev=$(mdsdevname $idx)
+			do_facet mds${idx} \
+			"$DEBUGFS -c -R 'ls /REMOTE_PARENT_DIR' $dev" |
+			grep ${fid} && rc=$idx
+		done
+	fi
+
+	mv $DIR/$tdir/dir1/foo1 $DIR/$tdir/dir1/foo2 ||
+		error "Fail to rename foo1 to foo2"
+	if [ $rc -eq 0 ]; then
+		for idx in $(seq $MDSCOUNT); do
+			dev=$(mdsdevname $idx)
+			do_facet mds${idx} \
+			"$DEBUGFS -c -R 'ls /REMOTE_PARENT_DIR' $dev" |
+			grep ${fid} && rc=$idx
+		done
+	fi
+
+	[ $rc -ne 0 ] || error "NOT found agent entry for foo"
+
+	ln $DIR/$tdir/dir1/foo2 $DIR/$tdir/dir0/guard ||
+		error "Fail to link to $DIR/$tdir/dir1/foo2"
+	mv $DIR/$tdir/dir1/foo2 $DIR/$tdir/dir1/foo0 ||
+		error "Fail to rename foo2 to foo0"
+	unlink $DIR/$tdir/dir1/foo0 ||
+		error "Fail to unlink $DIR/$tdir/dir1/foo0"
+	rm -rf $DIR/$tdir/dir0 ||
+		error "Fail to rm $DIR/$tdir/dir0"
+
+	for idx in $(seq $MDSCOUNT); do
+		dev=$(mdsdevname $idx)
+		rc=0
+
+		stop mds${idx}
+		run_e2fsck $(facet_active_host mds$idx) $dev -n ||
+			rc=$?
+		start mds${idx} $dev $MDS_MOUNT_OPTS ||
+			error "mount mds$idx failed"
+		df $MOUNT > /dev/null 2>&1
+
+		# e2fsck should not return error
+		[ $rc -eq 0 ] ||
+			error "e2fsck detected error on MDT${idx}: rc=$rc"
+	done
+}
+run_test 804 "verify agent entry for remote entry"
+
 #
 # tests that do cleanup/setup should be run at the end
 #