VPP/STN Testing

From fd.io
< VPP
Revision as of 22:04, 4 October 2017 by Hagbard (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Setup

# In this simulation, we use 192.168.1.2/24 as the 'shared address' between a vpp interface, and one end of a veth pair connected to vpp

# We need to setup a testing environment that uses the stn path in an analogous manner to what we need it to do.
# To do this, we create a 'pretendroot' namespace, and connect vpp to it with an eth0 interface in the pretendroot namespace
sudo ip netns add pretendroot
sudo ip link add name vpp2eth0 type veth peer name eth0
# Don't forget to disable the tcp offload, or we will get bitten by the Linux kernels misbehavior around checksums over veth pairs
sudo ethtool --offload  eth0  rx off  tx off
sudo ip link set eth0 netns pretendroot
sudo ip netns exec pretendroot ip addr add 192.168.1.2/24 dev eth0
# Set the mac address of the eth0 interface.  VPP will use 00:00:00:00:00:02 to address it, so we need to make sure the kernel accepts that mac for the eth0 interface
sudo ip netns exec pretendroot ip link set dev eth0 address 00:00:00:00:00:02
sudo ip netns exec pretendroot ip link set dev eth0 up

# Playing the part of our 'realnic' in this simulation, we create a 'realeth0' veth to connect to vpp.
# So that we can use it to originate traffic towards vpp, we give it IP 192.168.1.1/24
sudo ip link add name vpp2realeth0 type veth peer name realeth0
sudo ethtool --offload  realeth0  rx off  tx off
sudo ip addr add 192.168.1.1/24 dev realeth0
sudo ip link set dev realeth0 up

# Connect the vpp side of realeth0 (vpp2realeth0), addres it, and make sure its working
sudo vppctl create host-interface name vpp2realeth0
sudo vppctl set interface ip address host-vpp2realeth0 192.168.1.2/24
sudo vppctl set int state host-vpp2realeth0 up
ping 192.168.1.2

# Connect the vpp side of eth0 (vpp2eth0), set it up as an unnumbered slave of vpp2realeth0
sudo vppctl create host-interface name vpp2eth0
REALETH0_INT_IDX=$(sudo vppctl show int | grep host-vpp2realeth0 | awk '{print $2}')
echo ${REALETH0_INT_IDX}
ETH0_INT_IDX=$(sudo vppctl show int | grep host-vpp2eth0 | awk '{print $2}')
echo ${ETH0_INT_IDX}
echo sw_interface_set_unnumbered sw_if_index ${REALETH0_INT_IDX} unnum_if_index ${ETH0_INT_IDX} | sudo vpp_api_test | grep -v "#vat"
sudo vppctl set int state host-vpp2eth0 up

# Setup proxy arp for vpp2eth0
sudo vppctl set ip arp proxy 192.168.1.1-192.168.1.255
sudo vppctl set interface proxy-arp host-vpp2eth0 enable

# Setup the stn rule for 192.168.1.2 to host-vpp2eth0
sudo vppctl stn rule address 192.168.1.2 interface host-vpp2eth0
sudo vppctl show stn rule

Test 1: Ping from pretendroot (192.168.1.2) to realeth0 (192.168.1.1) via vpp

# Don't forget to setup vpp trace
sudo vppctl trace add af-packet-input 100
sudo ip netns exec pretendroot ping 192.168.1.1

Test1 Results

Ping works in this situation.

Test 2

Set up an echo server in the pretendroot namespace, and telnet to it.

In one shell:

sudo ip netns exec pretendroot /bin/bash
ncat -l 2000 -k -c 'xargs -n1 echo'

In a second shell:

telnet 192.168.1.2 2000

And try the echo server.

Test 2 Results

TCP works in this senario.

Test 1: Analysis

If you look at Packet 4, you will see the icp echo response is being addressed to mac address 00:00:00:00:00:02 (and bogus source 00:00:00:00:00:01) which causes the eth0 interface to reject the response. Tests for TCP and UDP showed the same issue.