add regression integration test for #1754

Validate that the RX queue won't freeze under high stateless ingress
traffic when encountering bottlenecks on the tap.

Signed-off-by: Serban Iorga <[email protected]>

Author: Serban Iorga

tests/host_tools/network.py

@@ -320,6 +320,16 @@ def netns(self):
         """Return the network namespace of this tap."""
         return self._netns
 
+    def set_tx_queue_len(self, tx_queue_len):
+        """Set the length of the tap's TX queue."""
+        utils.run_cmd(
+            'ip netns exec {} ip link set {} txqueuelen {}'.format(
+                self.netns,
+                self.name,
+                tx_queue_len
+            )
+        )
+
     def __del__(self):
         """Destructor doing tap interface clean up."""
         # pylint: disable=subprocess-run-check

tests/integration_tests/functional/test_net.py

@@ -0,0 +1,51 @@
+# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+# SPDX-License-Identifier: Apache-2.0
+"""Tests for the net device."""
+from subprocess import run, PIPE
+import time
+
+import host_tools.network as net_tools
+
+# The iperf version to run this tests with
+IPERF_BINARY = 'iperf3'
+
+
+def test_high_ingress_traffic(test_microvm_with_ssh, network_config):
+    """Run iperf rx with high UDP traffic."""
+    test_microvm = test_microvm_with_ssh
+    test_microvm.spawn()
+
+    test_microvm.basic_config()
+
+    # Create tap before configuring interface.
+    tap, _host_ip, guest_ip = test_microvm.ssh_network_config(
+        network_config,
+        '1'
+    )
+    # Set the tap's tx queue len to 5. This increases the probability
+    # of filling the tap under high ingress traffic.
+    tap.set_tx_queue_len(5)
+
+    # Start the microvm.
+    test_microvm.start()
+
+    # Start iperf3 server on the guest.
+    ssh_connection = net_tools.SSHConnection(test_microvm.ssh_config)
+    ssh_connection.execute_command('{} -sD\n'.format(IPERF_BINARY))
+    time.sleep(1)
+
+    # Start iperf3 client on the host. Send 1Gbps UDP traffic.
+    # If the net device breaks, iperf will freeze. We have to use a timeout.
+    run(
+        'timeout 30 {} {} -c {} -u -V -b 1000000000 -t 30'.format(
+            test_microvm.jailer.netns_cmd_prefix(),
+            IPERF_BINARY,
+            guest_ip,
+        ), stdout=PIPE, shell=True, check=False
+    )
+
+    # Check if the high ingress traffic broke the net interface.
+    # If the net interface still works we should be able to execute
+    # ssh commands.
+    exit_code, _, _ = ssh_connection.execute_command('echo success\n')
+    assert exit_code == 0