qdisc-hfsc-fq_codel.sh

#!/bin/sh
# Calvin's simple traffic shaper script
# Currently set up to handle traffic shaping on mb gateway
# Note that the ingress and egress rates are hardcoded in the commands below.

set -e

IFACE=$1
if [ -z "$IFACE" ]; then
	echo "Usage: $0 <iface>"
	exit 1
fi

if [ ! -d /sys/class/net/"$IFACE" ]; then
	echo "'$IFACE' does not appear to be a network interface."
	exit 1
fi

##############################################################################
# EGRESS

tc qdisc delete dev "$IFACE" root >/dev/null 2>&1 || true
tc qdisc add dev "$IFACE" root handle 1: hfsc default 20

# Overall egress limit, all other classes are children of this
tc class add dev "$IFACE" parent 1: classid 1:1 hfsc sc rate 20000kbit ul rate 20000kbit

# The "sc" rate is the min realtime bandwidth, the "ul" is the upper limit
# The "sc" rates for all of the traffic classes should sum to the overall
# limit
# fq_codel is used to manage packet queues, and makes good decisions on which
# packets to drop

# The default traffic class
# When link is at capacity, runs at 19mbit.
# When link not at capacity, allowed to use idle capacity.
tc class add dev "$IFACE" parent 1:1 classid 1:20 hfsc sc rate 19000kbit
tc qdisc add dev "$IFACE" parent 1:20 handle 200: fq_codel

# Traffic rules for rika
# When link is at capacity, runs at 1mbit.
# When link not at capacity, allowed to use idle capacity
tc class add dev "$IFACE" parent 1:1 classid 1:30 hfsc sc rate 1000kbit
tc qdisc add dev "$IFACE" parent 1:30 handle 300: fq_codel
# Match based on src ip of rika
tc filter add dev "$IFACE" parent 1: protocol ip u32 match ip src 72.29.167.158/32 flowid 1:30

##############################################################################
# INGRESS

# Set up the "ifb" device used to apply queue disciplines on the device
IFB_IFACE="${IFACE}_ingress"
tc qdisc delete dev "$IFACE" ingress >/dev/null 2>&1 || true
if [ -d "/sys/class/net/$IFB_IFACE" ]
then
	ip link delete "$IFB_IFACE"
fi
ip link add "$IFB_IFACE" type ifb
ip link set "$IFB_IFACE" up

# Forward ingress traffic to the ifb device
tc qdisc add dev "$IFACE" handle ffff: ingress
tc filter add dev "$IFACE" parent ffff: protocol all u32 match u32 0 0 action mirred egress redirect dev "$IFB_IFACE"

# Now the ingress ifb device can be set up in the same way as the egress device.

tc qdisc delete dev "$IFB_IFACE" root >/dev/null 2>&1 || true
tc qdisc add dev "$IFB_IFACE" root handle 1: hfsc default 20

tc class add dev "$IFB_IFACE" parent 1: classid 1:1 hfsc sc rate 5000kbit ul rate 5000kbit

# default traffic class
tc class add dev "$IFB_IFACE" parent 1:1 classid 1:20 hfsc sc rate 5000kbit
tc qdisc add dev "$IFB_IFACE" parent 1:20 handle 200: fq_codel