selftests/bpf: Prevent client connect before server bind in test_tc_t…

…unnel.sh

[ Upstream commit f803bcf ]

In some systems, the netcat server can incur in delay to start listening.
When this happens, the test can randomly fail in various points.
This is an example error message:

   # ip gre none gso
   # encap 192.168.1.1 to 192.168.1.2, type gre, mac none len 2000
   # test basic connectivity
   # Ncat: Connection refused.

The issue stems from a race condition between the netcat client and server.
The test author had addressed this problem by implementing a sleep, which
I have removed in this patch.
This patch introduces a function capable of sleeping for up to two seconds.
However, it can terminate the waiting period early if the port is reported
to be listening.

Signed-off-by: Alessandro Carminati (Red Hat) <[email protected]>
Signed-off-by: Andrii Nakryiko <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]
Signed-off-by: Sasha Levin <[email protected]>

tools/testing/selftests/bpf/test_tc_tunnel.sh

@@ -72,7 +72,6 @@ cleanup() {
 server_listen() {
 	ip netns exec "${ns2}" nc "${netcat_opt}" -l "${port}" > "${outfile}" &
 	server_pid=$!
-	sleep 0.2
 }
 
 client_connect() {
@@ -93,6 +92,16 @@ verify_data() {
 	fi
 }
 
+wait_for_port() {
+	for i in $(seq 20); do
+		if ip netns exec "${ns2}" ss ${2:--4}OHntl | grep -q "$1"; then
+			return 0
+		fi
+		sleep 0.1
+	done
+	return 1
+}
+
 set -e
 
 # no arguments: automated test, run all
@@ -193,6 +202,7 @@ setup
 # basic communication works
 echo "test basic connectivity"
 server_listen
+wait_for_port ${port} ${netcat_opt}
 client_connect
 verify_data
 
@@ -204,6 +214,7 @@ ip netns exec "${ns1}" tc filter add dev veth1 egress \
 	section "encap_${tuntype}_${mac}"
 echo "test bpf encap without decap (expect failure)"
 server_listen
+wait_for_port ${port} ${netcat_opt}
 ! client_connect
 
 if [[ "$tuntype" =~ "udp" ]]; then