selftests: mptcp: run mptcp_sockopt from a new netns

Not running it from a new netns causes issues if some MPTCP settings are
modified, e.g. if MPTCP is disabled from the sysctl knob, if multiple
addresses are available and added to the MPTCP path-manager, etc.

In these cases, the created connection will not behave as expected, e.g.
unable to create an MPTCP socket, more than one subflow is seen, etc.

A new "sandbox" net namespace is now created and used to run
mptcp_sockopt from this controlled environment.

Fixes: ce99791 ("selftests: mptcp: add mptcp getsockopt test cases")
Reviewed-by: Mat Martineau <[email protected]>
Signed-off-by: Matthieu Baerts <[email protected]>
Signed-off-by: Mat Martineau <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

Author: matttbe

tools/testing/selftests/net/mptcp/mptcp_sockopt.sh

@@ -35,8 +35,9 @@ init()
 
 	ns1="ns1-$rndh"
 	ns2="ns2-$rndh"
+	ns_sbox="ns_sbox-$rndh"
 
-	for netns in "$ns1" "$ns2";do
+	for netns in "$ns1" "$ns2" "$ns_sbox";do
 		ip netns add $netns || exit $ksft_skip
 		ip -net $netns link set lo up
 		ip netns exec $netns sysctl -q net.mptcp.enabled=1
@@ -73,7 +74,7 @@ init()
 
 cleanup()
 {
-	for netns in "$ns1" "$ns2"; do
+	for netns in "$ns1" "$ns2" "$ns_sbox"; do
 		ip netns del $netns
 	done
 	rm -f "$cin" "$cout"
@@ -243,7 +244,7 @@ do_mptcp_sockopt_tests()
 {
 	local lret=0
 
-	./mptcp_sockopt
+	ip netns exec "$ns_sbox" ./mptcp_sockopt
 	lret=$?
 
 	if [ $lret -ne 0 ]; then
@@ -252,7 +253,7 @@ do_mptcp_sockopt_tests()
 		return
 	fi
 
-	./mptcp_sockopt -6
+	ip netns exec "$ns_sbox" ./mptcp_sockopt -6
 	lret=$?
 
 	if [ $lret -ne 0 ]; then