test-ct-nat-tcp.sh

#!/bin/bash
#
# Test CT ipv4 NAT with tcp traffic
#

my_dir="$(dirname "$0")"
. $my_dir/common.sh

require_module act_ct
echo 1 > /proc/sys/net/netfilter/nf_conntrack_tcp_be_liberal

IP1="7.7.7.1"
IP2="7.7.7.2"
IP3="7.7.7.3"

enable_switchdev
require_interfaces REP REP2
unbind_vfs
bind_vfs
reset_tc $REP
reset_tc $REP2

mac1=`cat /sys/class/net/$VF/address`
mac2=`cat /sys/class/net/$VF2/address`

test "$mac1" || fail "no mac1"
test "$mac2" || fail "no mac2"

function cleanup() {
    ip netns del ns0 2> /dev/null
    ip netns del ns1 2> /dev/null
    reset_tc $REP
    reset_tc $REP2
}
trap cleanup EXIT

function get_pkts() {
    # single table tc show doesn't have nested keys attribute
    s1=`tc -j -p -s  filter show dev $REP protocol ip ingress | jq '.[] | select(.options.ct_state == "+trk+est") | .options.actions[0].stats.packets' || 0`
    # upstream tc dump
    s2=`tc -j -p -s  filter show dev $REP protocol ip ingress | jq '.[] | select(.options.keys.ct_state == "+trk+est") | .options.actions[0].stats.packets' || 0`

    echo $(( s1 > s2 ? s1 : s2 ))
}

function run() {
    title "Test CT nat tcp"
    tc_test_verbose
    config_vf ns0 $VF $REP $IP1
    config_vf ns1 $VF2 $REP2 $IP2
    ip -netns ns0 neigh replace $IP3 dev $VF lladdr $mac2
    ip -netns ns1 neigh replace $IP1 dev $VF2 lladdr $mac1

    flag=""
    # use this flag to test miss handling
    #flag2="skip_hw"

    echo "add ct rules"
    tc_filter add dev $REP ingress protocol ip prio 2 flower $flag $tc_verbose \
        dst_mac $mac2 ct_state -trk \
        action ct nat action goto chain 1

    tc_filter add dev $REP ingress protocol ip chain 1 prio 2 flower $flag $tc_verbose \
        dst_mac $mac2 ct_state +trk+new \
        action ct commit nat dst addr $IP2 \
        action mirred egress redirect dev $REP2

    tc_filter add dev $REP ingress protocol ip chain 1 prio 2 flower $flag $flag2 $tc_verbose \
        dst_mac $mac2 ct_state +trk+est \
        action mirred egress redirect dev $REP2

    # chain0,ct -> chain1,fwd
    tc_filter add dev $REP2 ingress protocol ip prio 2 flower $flag $tc_verbose \
        dst_mac $mac1 ct_state -trk \
        action ct nat action goto chain 1

    tc_filter add dev $REP2 ingress protocol ip prio 2 chain 1 flower $flag $flag2 $tc_verbose \
        dst_mac $mac1 ct_state +trk+est \
        action mirred egress redirect dev $REP

    fail_if_err

    echo $REP
    tc filter show dev $REP ingress
    echo $REP2
    tc filter show dev $REP2 ingress

    t=15
    echo "run traffic for $t seconds"
    ip netns exec ns1 timeout $((t+2)) iperf3 -s -D
    sleep 1

    ip netns exec ns0 timeout $t iperf3 -t $t -c $IP3 -P 1 -i 1 &
    sleep 2
    pidof iperf3 &>/dev/null || err "iperf failed"

    echo "sniff packets on $REP"
    # first 4 packets not offloaded until conn is in established state.
    timeout $((t-4)) tcpdump -qnnei $REP -c 10 'tcp' &
    pid=$!

    sleep 4
    pkts1=`get_pkts`

    sleep $((t-6))
    killall -q -9 iperf3
    wait $! 2>/dev/null

    title "verify tc stats"
    pkts2=`get_pkts`
    let a=pkts2-pkts1
    if (( a < 100 )); then
        err "TC stats are not updated"
    fi

    title "verify traffic offloaded"
    verify_no_traffic $pid

    reset_tc $REP
    reset_tc $REP2
    # wait for traces as merging & offloading is done in workqueue.
    sleep 3
}


run
test_done