The OpenWrt firewall implementation is the mechanism by which network traffic is filtered coming through the router. At a high level, one of three outcomes will occur: either the packet is discarded (dropped) without any further action, rejected (with an appropriate response to the source), or accepted (routed to the destination). Note that the router itself is a destination for management and monitoring.

The OpenWrt firewall revolves around the Linux netfilter project. There are the following main components to the OpenWrt firewall:

1. the firewall3 application
2. a set of netfilter hooks in the kernel networking stacks
3. a set of linux kernel modules that handle the inspection of network packets
4. a set of kernel tuning parameters to configure the network stacks and firewall modules

This documentation is based on OpenWrt 18.06.0. Many of the configurations have been tested against this release using the test network

The fw3 application package is the main application used to provision the firewall. It was developed by the OpenWrt team specifically for the project.

Each of the network stacks have netfilter functions call hooks embedded at specific places in the code. As a network packet moves through the stack, each hook is called to check the packet against possible netfilter rules bound to the hook.

The netfilter hook code uses the NF_HOOK set of macros. Each hook takes the following arguments:

• network protocol: unspec (all), ipv4, ipv6, arp, bridge, decnet
• hook num: PRE_ROUTING, LOCAL_IN, FORWARD, LOCAL_OUT, POST_ROUTING
• net structure: context for the network stack
• socket: BSD socket used for packet
• network packet: a socket buffer containing the network packet
• incoming device (interface): the source of the packet
• outgoing device (interface): the destination of the packet after routing
• a function callback if the packet passes the filter

The netfilter kernel modules are loaded at boot depend on the configured. There are roughly 35 kernel modules to support the standard netfilter capabilities but there are many more depending on the requirements of the router. For example, many routers use the ipset feature. This adds ~16 additional kernel modules.

Most of the netfilter modules are small, providing a single specific capability. For example:

• ipt_REJECT performs REJECT (target),
• xt_multiport performs match of the IP port (match)
• xt_TCPMSS performs Maximum Segment Size adjustment in the TCP header (target in mangle table)

Several of the netfilter modules are larger. For example:

• nf_conntrack performs connection tracking for masquerading (NAT) and packet de-fragmentation.

/etc/init.d/sysctl is executed at boot time. This is a shell script that loads /etc/sysctl.conf and all files under /etc/sysctl.d/. These set/tune kernel parameters to provide OpenWrt features. See sysctl.conf

All are parameters documented under the Documentation/networking directory of kernel source tree so the specifics will not be repeated here. See ip-sysctl.txt and nf_conntrack-sysctl.txt for reference.

Since the OpenWrt feature set is fairly static, the kernel parameters almost certainly do not need to tuned beyond the defaults provided in the build.

Notice that netfilter bridging support in the kernel is disabled! See ip-sysctl.txt:

bridge-nf-call-iptables - BOOLEAN
	1 : pass bridged IPv4 traffic to iptables' chains.
	0 : disable this.
	Default: 1

: the sysctl directives in /etc/init.d/sysctl are a superset of those in the /etc/sysctl.d files. The files in /etc/sysctl.d are entirely redundant. This may be a product of the build process.