test-ct-tcp-commit.sh

#!/bin/bash
#
# Test CT where all tc rules with CT action also do commit.
#
# Bug SW #2632768: [MT CT] Connection not offloaded when all CT action used have commit in them

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

require_module act_ct

IP1="7.7.7.1"
IP2="7.7.7.2"

enable_switchdev
require_interfaces REP REP2
unbind_vfs
bind_vfs
reset_tc $REP $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 $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 TCP"
    tc_test_verbose
    config_vf ns0 $VF $REP $IP1
    config_vf ns1 $VF2 $REP2 $IP2

    echo "add arp rules"
    tc_filter add dev $REP ingress protocol arp prio 1 flower $tc_verbose \
        action mirred egress redirect dev $REP2

    tc_filter add dev $REP2 ingress protocol arp prio 1 flower $tc_verbose \
        action mirred egress redirect dev $REP

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

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

    tc_filter add dev $REP ingress protocol ip chain 1 prio 2 flower $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 $tc_verbose \
        dst_mac $mac1 \
        action ct commit \
        action goto chain 1

    tc_filter add dev $REP2 ingress protocol ip prio 2 chain 1 flower $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=12
    echo "run traffic for $t seconds"
    ip netns exec ns1 timeout $((t+1)) iperf -s &
    sleep 0.5
    ip netns exec ns0 timeout $((t+1)) iperf -t $t -c $IP2 -P 3 &

    sleep 2
    pidof iperf &>/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 $REP -c 10 'tcp' &
    pid=$!

    pkts1=`get_pkts`

    sleep $t
    killall -9 iperf &>/dev/null
    wait $! 2>/dev/null

    title "verify tc stats"
    pkts2=`get_pkts`
    let a=pkts2-pkts1
    if (( a < 5 )); then
        err "TC stats are not updated"
    fi

    title "Verify no traffic on $REP"
    verify_no_traffic $pid

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


run
test_done