/* sysdep_netlink.c - Linux netlink glue
 *
 * Copyright (C) 2007-2009 Timo Teräs <timo.teras@iki.fi>
 * All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 or later as
 * published by the Free Software Foundation.
 *
 * See http://www.gnu.org/ for details.
 */

#include <time.h>
#include <stdio.h>
#include <fcntl.h>
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <malloc.h>
#include <string.h>
#include <sys/uio.h>
#include <sys/wait.h>
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <asm/types.h>
#include <arpa/inet.h>
#include <linux/netlink.h>
#include <linux/rtnetlink.h>
#include <linux/ip.h>
#include <linux/if_arp.h>
#include <linux/if_tunnel.h>

#include "libev.h"
#include "nhrp_common.h"
#include "nhrp_interface.h"
#include "nhrp_peer.h"

#define NETLINK_KERNEL_BUFFER	(256 * 1024)
#define NETLINK_RECV_BUFFER	(8 * 1024)

#define NLMSG_TAIL(nmsg) \
	((struct rtattr *) (((void *) (nmsg)) + NLMSG_ALIGN((nmsg)->nlmsg_len)))

#define NDA_RTA(r)  ((struct rtattr*)(((char*)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
#define NDA_PAYLOAD(n) NLMSG_PAYLOAD(n,sizeof(struct ndmsg))

typedef void (*netlink_dispatch_f)(struct nlmsghdr *msg);

struct netlink_fd {
	int fd;
	__u32 seq;
	struct ev_io io;

	int dispatch_size;
	const netlink_dispatch_f *dispatch;
};

static const int netlink_groups[] = {
	0,
	RTMGRP_NEIGH,
	RTMGRP_LINK,
	RTMGRP_IPV4_IFADDR,
	RTMGRP_IPV4_ROUTE,
};
static struct netlink_fd netlink_fds[ARRAY_SIZE(netlink_groups)];
#define talk_fd netlink_fds[0]

static struct ev_io packet_io;

static u_int16_t translate_mtu(u_int16_t mtu)
{
	/* if mtu is ethernet standard, do not advertise it
	 * pmtu should be working */
	if (mtu == 1500)
		return 0;
	return mtu;
}

static void netlink_parse_rtattr(struct rtattr *tb[], int max, struct rtattr *rta, int len)
{
	memset(tb, 0, sizeof(struct rtattr *) * (max + 1));
	while (RTA_OK(rta, len)) {
		if (rta->rta_type <= max)
			tb[rta->rta_type] = rta;
		rta = RTA_NEXT(rta,len);
	}
}

static int netlink_add_rtattr_l(struct nlmsghdr *n, int maxlen, int type,
				const void *data, int alen)
{
	int len = RTA_LENGTH(alen);
	struct rtattr *rta;

	if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len) > maxlen)
		return FALSE;

	rta = NLMSG_TAIL(n);
	rta->rta_type = type;
	rta->rta_len = len;
	memcpy(RTA_DATA(rta), data, alen);
#ifdef VALGRIND
	/* Clear the padding area to avoid spurious warnings */
	memset(RTA_DATA(rta) + alen, 0, RTA_ALIGN(len) - alen);
#endif
	n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len);
	return TRUE;
}

static int netlink_receive(struct netlink_fd *fd, struct nlmsghdr *reply)
{
	struct sockaddr_nl nladdr;
	struct iovec iov;
	struct msghdr msg = {
		.msg_name = &nladdr,
		.msg_namelen = sizeof(nladdr),
		.msg_iov = &iov,
		.msg_iovlen = 1,
	};
	int got_reply = FALSE, len;
	char buf[NETLINK_RECV_BUFFER];

	iov.iov_base = buf;
	while (!got_reply) {
		int status;
		struct nlmsghdr *h;

		iov.iov_len = sizeof(buf);
		status = recvmsg(fd->fd, &msg, MSG_DONTWAIT);
		if (status < 0) {
			if (errno == EINTR)
				continue;
			if (errno == EAGAIN)
				return reply == NULL;
			nhrp_perror("Netlink overrun");
			continue;
		}

		if (status == 0) {
			nhrp_error("Netlink returned EOF");
			return FALSE;
		}

		h = (struct nlmsghdr *) buf;
		while (NLMSG_OK(h, status)) {
			if (reply != NULL &&
			    h->nlmsg_seq == reply->nlmsg_seq) {
				len = h->nlmsg_len;
				if (len > reply->nlmsg_len) {
					nhrp_error("Netlink message truncated");
					len = reply->nlmsg_len;
				}
				memcpy(reply, h, len);
				got_reply = TRUE;
			} else if (h->nlmsg_type <= fd->dispatch_size &&
				fd->dispatch[h->nlmsg_type] != NULL) {
				fd->dispatch[h->nlmsg_type](h);
			} else if (h->nlmsg_type != NLMSG_DONE) {
				nhrp_info("Unknown NLmsg: 0x%08x, len %d",
					  h->nlmsg_type, h->nlmsg_len);
			}
			h = NLMSG_NEXT(h, status);
		}
	}

	return TRUE;
}

static int netlink_send(struct netlink_fd *fd, struct nlmsghdr *req)
{
	struct sockaddr_nl nladdr;
	struct iovec iov = {
		.iov_base = (void*) req,
		.iov_len = req->nlmsg_len
	};
	struct msghdr msg = {
		.msg_name = &nladdr,
		.msg_namelen = sizeof(nladdr),
		.msg_iov = &iov,
		.msg_iovlen = 1,
	};
	int status;

	memset(&nladdr, 0, sizeof(nladdr));
	nladdr.nl_family = AF_NETLINK;

	req->nlmsg_seq = ++fd->seq;

	status = sendmsg(fd->fd, &msg, 0);
	if (status < 0) {
		nhrp_perror("Cannot talk to rtnetlink");
		return FALSE;
	}
	return TRUE;
}

static int netlink_talk(struct netlink_fd *fd, struct nlmsghdr *req,
		 size_t replysize, struct nlmsghdr *reply)
{
	if (reply == NULL)
		req->nlmsg_flags |= NLM_F_ACK;

	if (!netlink_send(fd, req))
		return FALSE;

	if (reply == NULL)
		return TRUE;

	reply->nlmsg_len = replysize;
	return netlink_receive(fd, reply);
}

static int netlink_enumerate(struct netlink_fd *fd, int family, int type)
{
	struct {
		struct nlmsghdr nlh;
		struct rtgenmsg g;
	} req;
	struct sockaddr_nl addr;

	memset(&addr, 0, sizeof(addr));
	addr.nl_family = AF_NETLINK;

	memset(&req, 0, sizeof(req));
	req.nlh.nlmsg_len = sizeof(req);
	req.nlh.nlmsg_type = type;
	req.nlh.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
	req.nlh.nlmsg_pid = 0;
	req.nlh.nlmsg_seq = ++fd->seq;
	req.g.rtgen_family = family;

	return sendto(fd->fd, (void *) &req, sizeof(req), 0,
		      (struct sockaddr *) &addr, sizeof(addr)) >= 0;
}

static void netlink_read_cb(struct ev_io *w, int revents)
{
	struct netlink_fd *nfd = container_of(w, struct netlink_fd, io);

	if (revents & EV_READ)
		netlink_receive(nfd, NULL);
}

static int do_get_ioctl(const char *basedev, struct ip_tunnel_parm *p)
{
	struct ifreq ifr;

#ifdef VALGRIND
	/* Valgrind does not have SIOCGETTUNNEL description, so clear
	 * the memory structs to avoid spurious warnings */
	memset(&ifr, 0, sizeof(ifr));
	memset(p, 0, sizeof(*p));
#endif

	strncpy(ifr.ifr_name, basedev, IFNAMSIZ);
	ifr.ifr_ifru.ifru_data = (void *) p;
	if (ioctl(packet_io.fd, SIOCGETTUNNEL, &ifr)) {
		nhrp_perror("ioctl(SIOCGETTUNNEL)");
		return FALSE;
	}
	return TRUE;
}

#ifndef NHRP_NO_NBMA_GRE

static int netlink_add_nested_rtattr_u32(struct rtattr *rta, int maxlen,
					 int type, uint32_t value)
{
	int len = RTA_LENGTH(4);
	struct rtattr *subrta;

	if (RTA_ALIGN(rta->rta_len) + len > maxlen)
		return FALSE;

	subrta = (struct rtattr*)(((char*)rta) + RTA_ALIGN(rta->rta_len));
	subrta->rta_type = type;
	subrta->rta_len = len;
	memcpy(RTA_DATA(subrta), &value, 4);
	rta->rta_len = NLMSG_ALIGN(rta->rta_len) + len;
	return TRUE;
}

static int netlink_configure_arp(struct nhrp_interface *iface, int pf)
{
	struct {
		struct nlmsghdr n;
		struct ndtmsg ndtm;
		char buf[256];
	} req;
	struct {
		struct rtattr rta;
		char buf[256];
	} parms;

	memset(&req.n, 0, sizeof(req.n));
	memset(&req.ndtm, 0, sizeof(req.ndtm));

	req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndtmsg));
	req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_REPLACE;
	req.n.nlmsg_type = RTM_SETNEIGHTBL;

	req.ndtm.ndtm_family = pf;

	netlink_add_rtattr_l(&req.n, sizeof(req), NDTA_NAME,
			     "arp_cache", 10);

	parms.rta.rta_type = NDTA_PARMS;
	parms.rta.rta_len = RTA_LENGTH(0);
	netlink_add_nested_rtattr_u32(&parms.rta, sizeof(parms),
				      NDTPA_IFINDEX, iface->index);
	netlink_add_nested_rtattr_u32(&parms.rta, sizeof(parms),
				      NDTPA_APP_PROBES, 1);
	netlink_add_nested_rtattr_u32(&parms.rta, sizeof(parms),
				      NDTPA_MCAST_PROBES, 0);
	netlink_add_nested_rtattr_u32(&parms.rta, sizeof(parms),
				      NDTPA_UCAST_PROBES, 0);

	netlink_add_rtattr_l(&req.n, sizeof(req), NDTA_PARMS,
			     parms.buf, parms.rta.rta_len - RTA_LENGTH(0));

	return netlink_send(&talk_fd, &req.n);
}

static int netlink_link_arp_on(struct nhrp_interface *iface)
{
	struct ifreq ifr;

	strncpy(ifr.ifr_name, iface->name, IFNAMSIZ);
	if (ioctl(packet_io.fd, SIOCGIFFLAGS, &ifr)) {
		nhrp_perror("ioctl(SIOCGIFFLAGS)");
		return FALSE;
	}
	if (ifr.ifr_flags & IFF_NOARP) {
		ifr.ifr_flags &= ~IFF_NOARP;
		if (ioctl(packet_io.fd, SIOCSIFFLAGS, &ifr)) {
			nhrp_perror("ioctl(SIOCSIFFLAGS)");
			return FALSE;
		}
	}
	return TRUE;
}

#else

static int netlink_configure_arp(struct nhrp_interface *iface, int pf)
{
	return TRUE;
}

static int netlink_link_arp_on(struct nhrp_interface *iface)
{
	return TRUE;
}

#endif

static int proc_icmp_redirect_off(const char *interface)
{
	char fname[256];
	int fd, ret = FALSE;

	sprintf(fname, "/proc/sys/net/ipv4/conf/%s/send_redirects", interface);
	fd = open(fname, O_WRONLY);
	if (fd < 0)
		return FALSE;
	if (write(fd, "0\n", 2) == 2)
		ret = TRUE;
	close(fd);

	return ret;
}

static void netlink_neigh_request(struct nlmsghdr *msg)
{
	struct ndmsg *ndm = NLMSG_DATA(msg);
	struct rtattr *rta[NDA_MAX+1];
	struct nhrp_peer *peer;
	struct nhrp_address addr;
	struct nhrp_interface *iface;
	char tmp[64];

	netlink_parse_rtattr(rta, NDA_MAX, NDA_RTA(ndm), NDA_PAYLOAD(msg));
	if (rta[NDA_DST] == NULL)
		return;

	iface = nhrp_interface_get_by_index(ndm->ndm_ifindex, 0);
	if (iface == NULL)
		return;

	nhrp_address_set(&addr, ndm->ndm_family,
			 RTA_PAYLOAD(rta[NDA_DST]),
			 RTA_DATA(rta[NDA_DST]));

	nhrp_debug("NL-ARP(%s) who-has %s",
		   iface->name, nhrp_address_format(&addr, sizeof(tmp), tmp));

	peer = nhrp_peer_route(iface, &addr, 0, ~BIT(NHRP_PEER_TYPE_LOCAL_ROUTE));
	if (peer == NULL)
		return;

	if (peer->flags & NHRP_PEER_FLAG_UP)
		kernel_inject_neighbor(&addr, &peer->next_hop_address, iface);

	if (peer->next_hop_address.type != PF_UNSPEC &&
	    nhrp_address_cmp(&addr, &peer->protocol_address) != 0)
		nhrp_peer_traffic_indication(iface, peer->afnum, &addr);
}

static void netlink_neigh_update(struct nlmsghdr *msg)
{
	struct ndmsg *ndm = NLMSG_DATA(msg);
	struct rtattr *rta[NDA_MAX+1];
	struct nhrp_interface *iface;
	struct nhrp_peer_selector sel;
	int used = FALSE;

	netlink_parse_rtattr(rta, NDA_MAX, NDA_RTA(ndm), NDA_PAYLOAD(msg));
	if (rta[NDA_DST] == NULL)
		return;

	if (!(ndm->ndm_state & (NUD_STALE | NUD_FAILED | NUD_REACHABLE)))
		return;

	iface = nhrp_interface_get_by_index(ndm->ndm_ifindex, 0);
	if (iface == NULL)
		return;

	memset(&sel, 0, sizeof(sel));
	sel.flags = NHRP_PEER_FIND_EXACT;
	sel.interface = iface;
	nhrp_address_set(&sel.protocol_address, ndm->ndm_family,
			 RTA_PAYLOAD(rta[NDA_DST]),
			 RTA_DATA(rta[NDA_DST]));

	if (msg->nlmsg_type == RTM_NEWNEIGH && (ndm->ndm_state & NUD_REACHABLE))
		used = TRUE;

	nhrp_peer_foreach(nhrp_peer_set_used_matching,
			  (void*) (intptr_t) used, &sel);
}

static void netlink_link_new(struct nlmsghdr *msg)
{
	struct nhrp_interface *iface;
	struct ifinfomsg *ifi = NLMSG_DATA(msg);
	struct rtattr *rta[IFLA_MAX+1];
	const char *ifname;
	struct ip_tunnel_parm cfg;
	int configuration_changed = FALSE;

	netlink_parse_rtattr(rta, IFLA_MAX, IFLA_RTA(ifi), IFLA_PAYLOAD(msg));
	if (rta[IFLA_IFNAME] == NULL)
		return;

	ifname = RTA_DATA(rta[IFLA_IFNAME]);
	iface = nhrp_interface_get_by_name(ifname, TRUE);
	if (iface == NULL)
		return;

	if (rta[IFLA_MTU])
		iface->mtu = *((unsigned*)RTA_DATA(rta[IFLA_MTU]));

	if (((ifi->ifi_change & IFF_UP) || (iface->index == 0)) &&
	    (ifi->ifi_flags & IFF_UP)) {
		nhrp_info("Interface %s: configured UP, mtu=%d",
			  ifname, iface->mtu);
		nhrp_interface_run_script(iface, "interface-up");
	} else {
		nhrp_info("Interface %s: config change, mtu=%d",
			  ifname, iface->mtu);
	}

	iface->index = ifi->ifi_index;
	nhrp_interface_hash(iface);

	if (!(iface->flags & NHRP_INTERFACE_FLAG_CONFIGURED))
		return;

	switch (ifi->ifi_type) {
	case ARPHRD_IPGRE:
		iface->afnum = AFNUM_INET;
		/* try hard to get the interface nbma address */
		do_get_ioctl(ifname, &cfg);
		if (iface->gre_key != ntohl(cfg.i_key)) {
			configuration_changed = TRUE;
			iface->gre_key = ntohl(cfg.i_key);
		}
		if (cfg.iph.saddr) {
			struct nhrp_address saddr;
			nhrp_address_set(&saddr, PF_INET, 4, (uint8_t *) &cfg.iph.saddr);
			if (nhrp_address_cmp(&iface->nbma_address, &saddr) || iface->link_index) {
				configuration_changed = TRUE;
				iface->nbma_address = saddr;
				iface->link_index = 0;
			}
		} else if (cfg.link) {
			if (cfg.link != iface->link_index) {
				configuration_changed = TRUE;
				nhrp_address_set_type(&iface->nbma_address, PF_UNSPEC);
				iface->link_index = cfg.link;
			}
		} else {
			if (iface->link_index || iface->nbma_address.type != PF_UNSPEC) {
				configuration_changed = TRUE;
				/* Mark the interface as owning all NBMA addresses
				 * this works when there's only one GRE interface */
				iface->link_index = 0;
				nhrp_address_set_type(&iface->nbma_address, PF_UNSPEC);
				nhrp_info("WARNING: Cannot figure out NBMA address for "
					  "interface '%s'. Using route hints.", ifname);
			}
		}
		break;
	}

	if (!(iface->flags & NHRP_INTERFACE_FLAG_SHORTCUT_DEST)) {
		netlink_configure_arp(iface, PF_INET);
		netlink_link_arp_on(iface);
		proc_icmp_redirect_off(iface->name);
	}

	if (configuration_changed) {
		struct nhrp_peer_selector sel;
		int count = 0;

		/* Reset the interface values we detect later */
		memset(&iface->nat_cie, 0, sizeof(iface->nat_cie));
		iface->nbma_mtu = 0;
		if (iface->link_index) {
			/* Reenumerate addresses if needed */
			netlink_enumerate(&talk_fd, PF_UNSPEC, RTM_GETADDR);
			netlink_read_cb(&talk_fd.io, EV_READ);
		}

		/* Purge all NHRP entries for this interface */
		memset(&sel, 0, sizeof(sel));
		sel.type_mask = NHRP_PEER_TYPEMASK_PURGEABLE;
		sel.interface = iface;
		nhrp_peer_foreach(nhrp_peer_purge_matching, &count, &sel);
		nhrp_info("Interface %s: GRE configuration changed. Purged %d peers.",
			  ifname, count);
	}
}

static void netlink_link_del(struct nlmsghdr *msg)
{
	struct nhrp_interface *iface;
	struct ifinfomsg *ifi = NLMSG_DATA(msg);
	struct rtattr *rta[IFLA_MAX+1];
	const char *ifname;

	netlink_parse_rtattr(rta, IFLA_MAX, IFLA_RTA(ifi), IFLA_PAYLOAD(msg));
	if (rta[IFLA_IFNAME] == NULL)
		return;

	ifname = RTA_DATA(rta[IFLA_IFNAME]);
	iface = nhrp_interface_get_by_name(ifname, FALSE);
	if (iface == NULL)
		return;

	nhrp_info("Interface '%s' deleted", ifname);
	iface->index = 0;
	iface->link_index = 0;
	nhrp_interface_hash(iface);

	nhrp_address_set_type(&iface->nbma_address, PF_UNSPEC);
	nhrp_address_set_type(&iface->protocol_address, PF_UNSPEC);
}

static int netlink_addr_new_nbma(void *ctx, struct nhrp_interface *iface)
{
	struct nlmsghdr *msg = (struct nlmsghdr *) ctx;
	struct ifaddrmsg *ifa = NLMSG_DATA(msg);
	struct rtattr *rta[IFA_MAX+1];
	struct nhrp_interface *nbma_iface;

	if (iface->link_index == ifa->ifa_index) {
		netlink_parse_rtattr(rta, IFA_MAX, IFA_RTA(ifa),
				     IFA_PAYLOAD(msg));

		if (rta[IFA_LOCAL] == NULL)
			return 0;

		nhrp_address_set(&iface->nbma_address, ifa->ifa_family,
				 RTA_PAYLOAD(rta[IFA_LOCAL]),
				 RTA_DATA(rta[IFA_LOCAL]));

		nbma_iface = nhrp_interface_get_by_index(ifa->ifa_index, FALSE);
		if (nbma_iface != NULL) {
			iface->nbma_mtu = translate_mtu(nbma_iface->mtu);
		}
	}

	return 0;
}

static void netlink_addr_new(struct nlmsghdr *msg)
{
	struct nhrp_interface *iface;
	struct nhrp_peer *peer, *bcast;
	struct ifaddrmsg *ifa = NLMSG_DATA(msg);
	struct rtattr *rta[IFA_MAX+1];

	if (!(ifa->ifa_flags & IFA_F_SECONDARY))
		nhrp_interface_foreach(netlink_addr_new_nbma, msg);

	netlink_parse_rtattr(rta, IFA_MAX, IFA_RTA(ifa), IFA_PAYLOAD(msg));
	iface = nhrp_interface_get_by_index(ifa->ifa_index, FALSE);
	if (iface == NULL || rta[IFA_LOCAL] == NULL)
		return;

	/* Shortcut destination stuff is extracted from routes;
	 * not from local address information. */
	if (iface->flags & NHRP_INTERFACE_FLAG_SHORTCUT_DEST)
		return;
	if (!(iface->flags & NHRP_INTERFACE_FLAG_CONFIGURED))
		return;

	nhrp_address_set(&iface->protocol_address, ifa->ifa_family,
			 RTA_PAYLOAD(rta[IFA_LOCAL]),
			 RTA_DATA(rta[IFA_LOCAL]));
	iface->protocol_address_prefix = ifa->ifa_prefixlen;

	peer = nhrp_peer_alloc(iface);
	peer->type = NHRP_PEER_TYPE_LOCAL_ADDR;
	peer->afnum = AFNUM_RESERVED;
	nhrp_address_set(&peer->protocol_address, ifa->ifa_family,
			 RTA_PAYLOAD(rta[IFA_LOCAL]),
			 RTA_DATA(rta[IFA_LOCAL]));
	switch (ifa->ifa_family) {
	case PF_INET:
		peer->protocol_type = ETHPROTO_IP;
		peer->prefix_length = peer->protocol_address.addr_len * 8;
		nhrp_peer_insert(peer);
		break;
	default:
		nhrp_peer_put(peer);
		return;
	}

	bcast = nhrp_peer_alloc(iface);
	bcast->type = peer->type;
	bcast->afnum = peer->afnum;
	bcast->protocol_type = peer->protocol_type;
	bcast->prefix_length = peer->prefix_length;
	bcast->protocol_address = peer->protocol_address;
	nhrp_address_set_broadcast(&bcast->protocol_address,
				   ifa->ifa_prefixlen);
	bcast->next_hop_address = peer->protocol_address;
	nhrp_peer_insert(bcast);
	nhrp_peer_put(bcast);

	nhrp_peer_put(peer);
}

struct netlink_del_addr_msg {
	int interface_index;
	struct nhrp_address address;
};

static int netlink_addr_del_nbma(void *ctx, struct nhrp_interface *iface)
{
	struct netlink_del_addr_msg *msg = (struct netlink_del_addr_msg *) ctx;

	if (iface->link_index == msg->interface_index &&
	    nhrp_address_cmp(&msg->address, &iface->nbma_address) == 0)
		nhrp_address_set_type(&iface->nbma_address, PF_UNSPEC);

	return 0;
}

static int netlink_addr_purge_nbma(void *ctx, struct nhrp_peer *peer)
{
	struct netlink_del_addr_msg *msg = (struct netlink_del_addr_msg *) ctx;

	if (nhrp_address_cmp(&peer->my_nbma_address, &msg->address) == 0)
		nhrp_peer_purge(peer, "address-removed");

	return 0;
}

static void netlink_addr_del(struct nlmsghdr *nlmsg)
{
	struct netlink_del_addr_msg msg;
	struct nhrp_interface *iface;
	struct ifaddrmsg *ifa = NLMSG_DATA(nlmsg);
	struct rtattr *rta[IFA_MAX+1];
	struct nhrp_peer_selector sel;

	netlink_parse_rtattr(rta, IFA_MAX, IFA_RTA(ifa), IFA_PAYLOAD(nlmsg));
	if (rta[IFA_LOCAL] == NULL)
		return;

	msg.interface_index = ifa->ifa_index;
	nhrp_address_set(&msg.address, ifa->ifa_family,
			 RTA_PAYLOAD(rta[IFA_LOCAL]),
			 RTA_DATA(rta[IFA_LOCAL]));

	if (!(ifa->ifa_flags & IFA_F_SECONDARY))
		nhrp_interface_foreach(netlink_addr_del_nbma, &msg);
	nhrp_peer_foreach(netlink_addr_purge_nbma, &msg, NULL);

	iface = nhrp_interface_get_by_index(ifa->ifa_index, FALSE);
	if (iface == NULL)
		return;

	memset(&sel, 0, sizeof(sel));
	sel.flags = NHRP_PEER_FIND_EXACT;
	sel.type_mask = BIT(NHRP_PEER_TYPE_LOCAL_ADDR);
	sel.interface = iface;
	sel.protocol_address = msg.address;
	sel.prefix_length = sel.protocol_address.addr_len * 8;

	if (nhrp_address_cmp(&sel.protocol_address, &iface->protocol_address) == 0)
		nhrp_address_set_type(&iface->protocol_address, PF_UNSPEC);
	nhrp_peer_foreach(nhrp_peer_remove_matching, NULL, &sel);

	nhrp_address_set_broadcast(&sel.protocol_address, ifa->ifa_prefixlen);
	sel.next_hop_address = msg.address;
	nhrp_peer_foreach(nhrp_peer_remove_matching, NULL, &sel);
}

static void netlink_route_new(struct nlmsghdr *msg)
{
	struct nhrp_interface *iface;
	struct nhrp_peer *peer;
	struct rtmsg *rtm = NLMSG_DATA(msg);
	struct rtattr *rta[RTA_MAX+1];
	int type = 0;

	netlink_parse_rtattr(rta, RTA_MAX, RTM_RTA(rtm), RTM_PAYLOAD(msg));
	if (rta[RTA_OIF] == NULL || rta[RTA_DST] == NULL)
		return;

	if (rtm->rtm_family != PF_INET)
		return;

	iface = nhrp_interface_get_by_index(*(int*)RTA_DATA(rta[RTA_OIF]),
					    FALSE);
	if (iface == NULL)
		return;

	if (iface->flags & NHRP_INTERFACE_FLAG_SHORTCUT_DEST) {
		/* Local shortcut target routes */
		if (rtm->rtm_table != RT_TABLE_MAIN)
			return;
		type = NHRP_PEER_TYPE_LOCAL_ADDR;
	} else if (iface->flags & NHRP_INTERFACE_FLAG_CONFIGURED) {
		/* Routes which might get additional outbound
		 * shortcuts */
		if (rtm->rtm_table != iface->route_table ||
		    rtm->rtm_protocol == RTPROT_KERNEL)
			return;
		type = NHRP_PEER_TYPE_LOCAL_ROUTE;
	}
	if (type == 0)
		return;

	peer = nhrp_peer_alloc(iface);
	peer->type = type;
	peer->afnum = AFNUM_RESERVED;
	nhrp_address_set(&peer->protocol_address, rtm->rtm_family,
			 RTA_PAYLOAD(rta[RTA_DST]),
			 RTA_DATA(rta[RTA_DST]));
	if (rta[RTA_GATEWAY] != NULL) {
		nhrp_address_set(&peer->next_hop_address,
				 rtm->rtm_family,
				 RTA_PAYLOAD(rta[RTA_GATEWAY]),
				 RTA_DATA(rta[RTA_GATEWAY]));
	}
	peer->protocol_type = nhrp_protocol_from_pf(rtm->rtm_family);
	peer->prefix_length = rtm->rtm_dst_len;
	nhrp_peer_insert(peer);
	nhrp_peer_put(peer);
}

static void netlink_route_del(struct nlmsghdr *msg)
{
	struct nhrp_interface *iface;
	struct rtmsg *rtm = NLMSG_DATA(msg);
	struct rtattr *rta[RTA_MAX+1];
	struct nhrp_peer_selector sel;
	int type = 0;

	netlink_parse_rtattr(rta, RTA_MAX, RTM_RTA(rtm), RTM_PAYLOAD(msg));
	if (rta[RTA_OIF] == NULL || rta[RTA_DST] == NULL)
		return;

	if (rtm->rtm_family != PF_INET)
		return;

	iface = nhrp_interface_get_by_index(*(int*)RTA_DATA(rta[RTA_OIF]),
					    FALSE);
	if (iface == NULL)
		return;

	if (iface->flags & NHRP_INTERFACE_FLAG_SHORTCUT_DEST) {
		/* Local shortcut target routes */
		if (rtm->rtm_table != RT_TABLE_MAIN)
			return;
		type = NHRP_PEER_TYPE_LOCAL_ADDR;
	} else if (iface->flags & NHRP_INTERFACE_FLAG_CONFIGURED) {
		/* Routes which might get additional outbound
		 * shortcuts */
		if (rtm->rtm_table != iface->route_table ||
		    rtm->rtm_protocol == RTPROT_KERNEL)
			return;
		type = NHRP_PEER_TYPE_LOCAL_ROUTE;
	}
	if (type == 0)
		return;

	memset(&sel, 0, sizeof(sel));
	sel.flags = NHRP_PEER_FIND_EXACT;
	sel.type_mask = BIT(type);
	sel.interface = iface;
	nhrp_address_set(&sel.protocol_address, rtm->rtm_family,
			 RTA_PAYLOAD(rta[RTA_DST]),
			 RTA_DATA(rta[RTA_DST]));
	if (rta[RTA_GATEWAY] != NULL) {
		nhrp_address_set(&sel.next_hop_address,
				 rtm->rtm_family,
				 RTA_PAYLOAD(rta[RTA_GATEWAY]),
				 RTA_DATA(rta[RTA_GATEWAY]));
	}
	sel.prefix_length = rtm->rtm_dst_len;
	nhrp_peer_foreach(nhrp_peer_remove_matching, NULL, &sel);
}

static const netlink_dispatch_f route_dispatch[RTM_MAX] = {
	[RTM_GETNEIGH] = netlink_neigh_request,
	[RTM_NEWNEIGH] = netlink_neigh_update,
	[RTM_DELNEIGH] = netlink_neigh_update,
	[RTM_NEWLINK] = netlink_link_new,
	[RTM_DELLINK] = netlink_link_del,
	[RTM_NEWADDR] = netlink_addr_new,
	[RTM_DELADDR] = netlink_addr_del,
	[RTM_NEWROUTE] = netlink_route_new,
	[RTM_DELROUTE] = netlink_route_del,
};

static void netlink_stop_listening(struct netlink_fd *fd)
{
	ev_io_stop(&fd->io);
}

static void netlink_close(struct netlink_fd *fd)
{
	if (fd->fd >= 0) {
		netlink_stop_listening(fd);
		close(fd->fd);
		fd->fd = 0;
	}
}

static int netlink_open(struct netlink_fd *fd, int protocol, int groups)
{
	struct sockaddr_nl addr;
	int buf = NETLINK_KERNEL_BUFFER;

	fd->fd = socket(AF_NETLINK, SOCK_RAW, protocol);
	fd->seq = time(NULL);
	if (fd->fd < 0) {
		nhrp_perror("Cannot open netlink socket");
		return FALSE;
	}

	fcntl(fd->fd, F_SETFD, FD_CLOEXEC);
	if (setsockopt(fd->fd, SOL_SOCKET, SO_SNDBUF, &buf, sizeof(buf)) < 0) {
		nhrp_perror("SO_SNDBUF");
		goto error;
	}

	if (setsockopt(fd->fd, SOL_SOCKET, SO_RCVBUF, &buf, sizeof(buf)) < 0) {
		nhrp_perror("SO_RCVBUF");
		goto error;
	}

	memset(&addr, 0, sizeof(addr));
	addr.nl_family = AF_NETLINK;
	addr.nl_groups = groups;
	if (bind(fd->fd, (struct sockaddr *) &addr, sizeof(addr)) < 0) {
		nhrp_perror("Cannot bind netlink socket");
		goto error;
	}

	ev_io_init(&fd->io, netlink_read_cb, fd->fd, EV_READ);
	ev_io_start(&fd->io);

	return TRUE;

error:
	netlink_close(fd);
	return FALSE;
}

static void pfpacket_read_cb(struct ev_io *w, int revents)
{
	struct sockaddr_ll lladdr;
	struct nhrp_interface *iface;
	struct iovec iov;
	struct msghdr msg = {
		.msg_name = &lladdr,
		.msg_namelen = sizeof(lladdr),
		.msg_iov = &iov,
		.msg_iovlen = 1,
	};
	uint8_t buf[1500];
	struct nhrp_address from;
	int fd = w->fd;
	int i;

	iov.iov_base = buf;
	for (i = 0; i < 2; i++) {
		int status;

		iov.iov_len = sizeof(buf);
		status = recvmsg(fd, &msg, MSG_DONTWAIT);
		if (status < 0) {
			if (errno == EINTR)
				continue;
			if (errno == EAGAIN)
				return;
			nhrp_perror("PF_PACKET overrun");
			continue;
		}

		if (status == 0) {
			nhrp_error("PF_PACKET returned EOF");
			return;
		}

		iface = nhrp_interface_get_by_index(lladdr.sll_ifindex, FALSE);
		if (iface == NULL)
			continue;

		nhrp_address_set(&from, PF_INET, lladdr.sll_halen, lladdr.sll_addr);
		if (memcmp(lladdr.sll_addr, "\x00\x00\x00\x00", 4) == 0)
			nhrp_address_set_type(&from, PF_UNSPEC);
		nhrp_packet_receive(buf, status, iface, &from);
	}
}

int kernel_init(void)
{
	int fd, i;

	proc_icmp_redirect_off("all");

	fd = socket(PF_PACKET, SOCK_DGRAM, ETHPROTO_NHRP);
	if (fd < 0) {
		nhrp_error("Unable to create PF_PACKET socket");
		return FALSE;
	}

	fcntl(fd, F_SETFD, FD_CLOEXEC);
	ev_io_init(&packet_io, pfpacket_read_cb, fd, EV_READ);
	ev_io_start(&packet_io);

	for (i = 0; i < ARRAY_SIZE(netlink_groups); i++) {
		netlink_fds[i].dispatch_size = sizeof(route_dispatch) / sizeof(route_dispatch[0]);
		netlink_fds[i].dispatch = route_dispatch;
		if (!netlink_open(&netlink_fds[i], NETLINK_ROUTE,
				  netlink_groups[i]))
			goto err_close_all;
	}

	netlink_enumerate(&talk_fd, PF_UNSPEC, RTM_GETLINK);
	netlink_read_cb(&talk_fd.io, EV_READ);

	netlink_enumerate(&talk_fd, PF_UNSPEC, RTM_GETADDR);
	netlink_read_cb(&talk_fd.io, EV_READ);

	netlink_enumerate(&talk_fd, PF_UNSPEC, RTM_GETROUTE);
	netlink_read_cb(&talk_fd.io, EV_READ);

	return TRUE;

err_close_all:
	kernel_cleanup();
	return FALSE;
}

void kernel_stop_listening(void)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(netlink_groups); i++)
		netlink_stop_listening(&netlink_fds[i]);
	ev_io_stop(&packet_io);
}

void kernel_cleanup(void)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(netlink_groups); i++)
		netlink_close(&netlink_fds[i]);
	ev_io_stop(&packet_io);
	close(packet_io.fd);
}

int kernel_route(struct nhrp_interface *out_iface,
		 struct nhrp_address *dest,
		 struct nhrp_address *default_source,
		 struct nhrp_address *next_hop,
		 u_int16_t *mtu)
{
	struct {
		struct nlmsghdr 	n;
		struct rtmsg 		r;
		char   			buf[1024];
	} req;
	struct rtmsg *r = NLMSG_DATA(&req.n);
	struct rtattr *rta[RTA_MAX+1];

	memset(&req, 0, sizeof(req));
	req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
	req.n.nlmsg_flags = NLM_F_REQUEST;
	req.n.nlmsg_type = RTM_GETROUTE;
	req.r.rtm_family = dest->type;

	netlink_add_rtattr_l(&req.n, sizeof(req), RTA_DST,
			     dest->addr, dest->addr_len);
	req.r.rtm_dst_len = dest->addr_len * 8;

	if (default_source != NULL && default_source->type != PF_UNSPEC)
		netlink_add_rtattr_l(&req.n, sizeof(req), RTA_SRC,
				     default_source->addr,
				     default_source->addr_len);
	if (out_iface != NULL)
		netlink_add_rtattr_l(&req.n, sizeof(req), RTA_OIF,
				     &out_iface->index, sizeof(int));

	if (!netlink_talk(&talk_fd, &req.n, sizeof(req), &req.n))
		return FALSE;

	netlink_parse_rtattr(rta, RTA_MAX, RTM_RTA(r), RTM_PAYLOAD(&req.n));

	if (default_source != NULL && default_source->type == PF_UNSPEC &&
	    rta[RTA_PREFSRC] != NULL) {
		nhrp_address_set(default_source, dest->type,
				 RTA_PAYLOAD(rta[RTA_PREFSRC]),
				 RTA_DATA(rta[RTA_PREFSRC]));
	}

	if (next_hop != NULL) {
		if (rta[RTA_GATEWAY] != NULL) {
			nhrp_address_set(next_hop, dest->type,
					 RTA_PAYLOAD(rta[RTA_GATEWAY]),
					 RTA_DATA(rta[RTA_GATEWAY]));
		} else {
			*next_hop = *dest;
		}
	}

	if (mtu != NULL) {
		*mtu = 0;

		if (rta[RTA_OIF] != NULL) {
			struct nhrp_interface *nbma_iface;

			/* We use interface MTU here instead of the route
			 * cache MTU from RTA_METRICS/RTAX_MTU since we
			 * don't want to announce mtu if PMTU works */
			nbma_iface = nhrp_interface_get_by_index(
				*(int*)RTA_DATA(rta[RTA_OIF]),
				FALSE);
			if (nbma_iface != NULL)
				*mtu = translate_mtu(nbma_iface->mtu);
		}
	}

	return TRUE;
}

int kernel_send(uint8_t *packet, size_t bytes, struct nhrp_interface *out,
		struct nhrp_address *to)
{
	struct sockaddr_ll lladdr;
	struct iovec iov = {
		.iov_base = (void*) packet,
		.iov_len = bytes
	};
	struct msghdr msg = {
		.msg_name = &lladdr,
		.msg_namelen = sizeof(lladdr),
		.msg_iov = &iov,
		.msg_iovlen = 1,
	};
	int status;

	if (to->addr_len > sizeof(lladdr.sll_addr)) {
		nhrp_error("Destination NBMA address too long");
		return FALSE;
	}

	memset(&lladdr, 0, sizeof(lladdr));
	lladdr.sll_family = AF_PACKET;
	lladdr.sll_protocol = ETHPROTO_NHRP;
	lladdr.sll_ifindex = out->index;
	lladdr.sll_halen = to->addr_len;
	memcpy(lladdr.sll_addr, to->addr, to->addr_len);

	status = sendmsg(packet_io.fd, &msg, 0);
	if (status < 0) {
		nhrp_error("Cannot send packet to %s(%d): %s",
			   out->name, out->index, strerror(errno));
		return FALSE;
	}

	return TRUE;
}

int kernel_inject_neighbor(struct nhrp_address *neighbor,
			   struct nhrp_address *hwaddr,
			   struct nhrp_interface *dev)
{
	struct {
		struct nlmsghdr 	n;
		struct ndmsg 		ndm;
		char   			buf[256];
	} req;
	char neigh[64], nbma[64];

	memset(&req.n, 0, sizeof(req.n));
	memset(&req.ndm, 0, sizeof(req.ndm));

	req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
	req.n.nlmsg_flags = NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE;
	req.n.nlmsg_type = RTM_NEWNEIGH;
	req.ndm.ndm_family = neighbor->type;
	req.ndm.ndm_ifindex = dev->index;
	req.ndm.ndm_type = RTN_UNICAST;

	netlink_add_rtattr_l(&req.n, sizeof(req), NDA_DST,
			     neighbor->addr, neighbor->addr_len);

	if (hwaddr != NULL && hwaddr->type != PF_UNSPEC) {
		req.ndm.ndm_state = NUD_REACHABLE;

		netlink_add_rtattr_l(&req.n, sizeof(req), NDA_LLADDR,
				     hwaddr->addr, hwaddr->addr_len);

		nhrp_debug("NL-ARP(%s) %s is-at %s",
			   dev->name,
			   nhrp_address_format(neighbor, sizeof(neigh), neigh),
			   nhrp_address_format(hwaddr, sizeof(nbma), nbma));
	} else {
		req.ndm.ndm_state = NUD_FAILED;

		nhrp_debug("NL-ARP(%s) %s not-reachable",
			   dev->name,
			   nhrp_address_format(neighbor, sizeof(neigh), neigh));
	}

	return netlink_send(&talk_fd, &req.n);
}