OpenBSD manual page server

Manual Page Search Parameters

DIVERT(4) Device Drivers Manual DIVERT(4)

divertkernel packet diversion mechanism

#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>



Divert sockets are part of a mechanism completely integrated with pf(4) that queues raw packets from the kernel stack to userspace applications, and vice versa.

A divert socket must be bound to a divert port through bind(2), which only the superuser can do. Divert ports have their own number space, completely separated from tcp(4) and udp(4). When pf(4) processes a packet that matches a rule with the divert-packet parameter (see pf.conf(5) for details) it is sent to the divert socket listening on the divert port specified in the rule. Note that divert-packet should not be confused with divert-to or divert-reply, which do not use divert sockets. If there are no divert sockets listening, the packets are dropped.

Packets can be read via read(2), recv(2), or recvfrom(2) from the divert socket. The application that is processing the packets can then reinject them into the kernel. With recvfrom(2), an interface IP address is passed if it is an inbound packet. Outbound packets provide the unspecified address. When reinjecting, use this address as argument to sendto(2). This allows the kernel to guess the original incoming interface and and process it as an incoming packet. If no interface IP address is given, the reinjected packet is treated as an outgoing packet. Since the userspace application could have modified the packets, upon reinjection basic sanity checks are done to ensure that the packets are still valid. The packets' IPv4 and protocol checksums (TCP, UDP, ICMP, and ICMPv6) are also recalculated.

Writing to a divert socket can be achieved using sendto(2) and it will skip pf(4) filters to avoid loops. Note that this means that a reinjected inbound packet will also not run through the pf out rules after being forwarded. A diverted packet that is not reinjected into the kernel stack is lost.

Receive and send divert socket buffer space can be tuned through sysctl(8). netstat(1) shows information relevant to divert sockets. Note that the default is 64k and too short to handle full sized UDP packets.

The IP_DIVERTFL socket option on the IPPROTO_IP level controls whether both inbound and outbound packets are diverted (the default) or only packets travelling in one direction. It cannot be reset once set. Valid values are IPPROTO_DIVERT_INIT for the direction of the initial packet of a flow, and IPPROTO_DIVERT_RESP for the direction of the response packets. The direction is relative to the packet direction. So for pf out rules, it is the other way around. If one filter is active, it specifies which packets should not be diverted. Both directions can be combined as bit fields, but then the traffic is not filtered; not using the P_DIVERTFL option has the same effect.

The following PF rule queues outbound IPv4 packets to TCP port 80, as well as the return traffic, on the em0 interface to divert port 700:

pass out on em0 inet proto tcp to port 80 divert-packet port 700

The following program reads packets on divert port 700 and reinjects them back into the kernel. This program does not perform any processing of the packets, apart from discarding invalid IP packets.

#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <netinet/tcp.h>
#include <arpa/inet.h>
#include <stdio.h>
#include <string.h>
#include <err.h>

#define DIVERT_PORT 700

main(int argc, char *argv[])
	int fd, s;
	struct sockaddr_in sin;
	socklen_t sin_len;

	if (fd == -1)
		err(1, "socket");

	memset(&sin, 0, sizeof(sin));
	sin.sin_family = AF_INET;
	sin.sin_port = htons(DIVERT_PORT);
	sin.sin_addr.s_addr = 0;

	sin_len = sizeof(struct sockaddr_in);

	s = bind(fd, (struct sockaddr *) &sin, sin_len);
	if (s == -1)
		err(1, "bind");

	for (;;) {
		ssize_t n;
		char packet[IP_MAXPACKET];
		struct ip *ip;
		struct tcphdr *th;
		int hlen;
		char src[48], dst[48];

		memset(packet, 0, sizeof(packet));
		n = recvfrom(fd, packet, sizeof(packet), 0,
		    (struct sockaddr *) &sin, &sin_len);
		if (n == -1) {
		if (n < sizeof(struct ip)) {
			warnx("packet is too short");

		ip = (struct ip *) packet;
		hlen = ip->ip_hl << 2;
		if (hlen < sizeof(struct ip) || ntohs(ip->ip_len) < hlen ||
		    n < ntohs(ip->ip_len)) {
			warnx("invalid IPv4 packet");

		th = (struct tcphdr *) (packet + hlen);

		if (inet_ntop(AF_INET, &ip->ip_src, src,
		    sizeof(src)) == NULL)
			(void)strlcpy(src, "?", sizeof(src));

		if (inet_ntop(AF_INET, &ip->ip_dst, dst,
		    sizeof(dst)) == NULL)
			(void)strlcpy(dst, "?", sizeof(dst));

		printf("%s:%u -> %s:%u\n",

		n = sendto(fd, packet, n, 0, (struct sockaddr *) &sin,
		if (n == -1)

	return 0;

setsockopt(2), socket(2), ip(4), pf.conf(5)

The divert protocol first appeared in OpenBSD 4.7.

August 25, 2017 OpenBSD-6.2