/*
 * route-bsd.c
 *
 * Copyright (c) 2001 Dug Song <dugsong@monkey.org>
 * Copyright (c) 1999 Masaki Hirabaru <masaki@merit.edu>
 * 
 * $Id: route-bsd.c 555 2005-02-10 05:18:38Z dugsong $
 */

#include "config.h"

#include <sys/param.h>
#include <sys/types.h>
#include <sys/socket.h>
#ifdef HAVE_SYS_SYSCTL_H
#include <sys/sysctl.h>
#endif
#ifdef HAVE_STREAMS_MIB2
#include <sys/stream.h>
#include <sys/tihdr.h>
#include <sys/tiuser.h>
#include <inet/common.h>
#include <inet/mib2.h>
#include <inet/ip.h>
#undef IP_ADDR_LEN
#include <stropts.h>
#elif defined(HAVE_STREAMS_ROUTE)
#include <sys/stream.h>
#include <sys/stropts.h>
#endif
#ifdef HAVE_GETKERNINFO
#include <sys/kinfo.h>
#endif

#define route_t	oroute_t	/* XXX - unixware */
#include <net/route.h>
#undef route_t
#include <net/if.h>
#include <netinet/in.h>

#include <errno.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

#include "dnet.h"

#if defined(RT_ROUNDUP) && defined(__NetBSD__)
/* NetBSD defines this macro rounding to 64-bit boundaries.
   http://fxr.watson.org/fxr/ident?v=NETBSD;i=RT_ROUNDUP */
#define ROUNDUP(a) RT_ROUNDUP(a)
#else
/* Unix Network Programming, 3rd edition says that sockaddr structures in
   rt_msghdr should be padded so their addresses start on a multiple of
   sizeof(u_long). But on 64-bit Mac OS X 10.6 at least, this is false. Apple's
   netstat code uses 4-byte padding, not 8-byte. This is relevant for IPv6
   addresses, for which sa_len == 28.
   http://www.opensource.apple.com/source/network_cmds/network_cmds-329.2.2/netstat.tproj/route.c */
#ifdef __APPLE__
#define RT_MSGHDR_ALIGNMENT sizeof(uint32_t)
#else
#define RT_MSGHDR_ALIGNMENT sizeof(unsigned long)
#endif
#define ROUNDUP(a) \
	((a) > 0 ? (1 + (((a) - 1) | (RT_MSGHDR_ALIGNMENT - 1))) : RT_MSGHDR_ALIGNMENT)
#endif

#ifdef HAVE_SOCKADDR_SA_LEN
#define NEXTSA(s) \
	((struct sockaddr *)((u_char *)(s) + ROUNDUP((s)->sa_len)))
#else
#define NEXTSA(s) \
	((struct sockaddr *)((u_char *)(s) + ROUNDUP(sizeof(*(s)))))
#endif

struct route_handle {
	int	fd;
	int	seq;
#ifdef HAVE_STREAMS_MIB2
	int	ip_fd;
#endif
};

#ifdef DEBUG
static void
route_msg_print(struct rt_msghdr *rtm)
{
	printf("v: %d type: 0x%x flags: 0x%x addrs: 0x%x pid: %d seq: %d\n",
	    rtm->rtm_version, rtm->rtm_type, rtm->rtm_flags,
	    rtm->rtm_addrs, rtm->rtm_pid, rtm->rtm_seq);
}
#endif

static int
route_msg(route_t *r, int type, char intf_name[INTF_NAME_LEN], struct addr *dst, struct addr *gw)
{
	struct addr net;
	struct rt_msghdr *rtm;
	struct sockaddr *sa;
	u_char buf[BUFSIZ];
	pid_t pid;
	int len;

	memset(buf, 0, sizeof(buf));

	rtm = (struct rt_msghdr *)buf;
	rtm->rtm_version = RTM_VERSION;
	if ((rtm->rtm_type = type) != RTM_DELETE)
		rtm->rtm_flags = RTF_UP;
	rtm->rtm_addrs = RTA_DST;
	rtm->rtm_seq = ++r->seq;

	/* Destination */
	sa = (struct sockaddr *)(rtm + 1);
	if (addr_net(dst, &net) < 0 || addr_ntos(&net, sa) < 0)
		return (-1);
	sa = NEXTSA(sa);

	/* Gateway */
	if (gw != NULL && type != RTM_GET) {
		rtm->rtm_flags |= RTF_GATEWAY;
		rtm->rtm_addrs |= RTA_GATEWAY;
		if (addr_ntos(gw, sa) < 0)
			return (-1);
		sa = NEXTSA(sa);
	}
	/* Netmask */
	if (dst->addr_ip == IP_ADDR_ANY || dst->addr_bits < IP_ADDR_BITS) {
		rtm->rtm_addrs |= RTA_NETMASK;
		if (addr_btos(dst->addr_bits, sa) < 0)
			return (-1);
		sa = NEXTSA(sa);
	} else
		rtm->rtm_flags |= RTF_HOST;
	
	rtm->rtm_msglen = (u_char *)sa - buf;
#ifdef DEBUG
	route_msg_print(rtm);
#endif
#ifdef HAVE_STREAMS_ROUTE
	if (ioctl(r->fd, RTSTR_SEND, rtm) < 0)
		return (-1);
#else
	if (write(r->fd, buf, rtm->rtm_msglen) < 0)
		return (-1);

	pid = getpid();
	
	while (type == RTM_GET && (len = read(r->fd, buf, sizeof(buf))) > 0) {
		if (len < (int)sizeof(*rtm)) {
			return (-1);
		}
		if (rtm->rtm_type == type && rtm->rtm_pid == pid &&
		    rtm->rtm_seq == r->seq) {
			if (rtm->rtm_errno) {
				errno = rtm->rtm_errno;
				return (-1);
			}
			break;
		}
	}
#endif
	if (type == RTM_GET && (rtm->rtm_addrs & (RTA_DST|RTA_GATEWAY)) ==
	    (RTA_DST|RTA_GATEWAY)) {
		sa = (struct sockaddr *)(rtm + 1);
		sa = NEXTSA(sa);
		
		if (addr_ston(sa, gw) < 0 || gw->addr_type != ADDR_TYPE_IP) {
			errno = ESRCH;
			return (-1);
		}

		if (intf_name != NULL) {
			char namebuf[IF_NAMESIZE];

			if (if_indextoname(rtm->rtm_index, namebuf) == NULL) {
				errno = ESRCH;
				return (-1);
			}
			strlcpy(intf_name, namebuf, INTF_NAME_LEN);
		}
	}
	return (0);
}

route_t *
route_open(void)
{
	route_t *r;
	
	if ((r = calloc(1, sizeof(*r))) != NULL) {
		r->fd = -1;
#ifdef HAVE_STREAMS_MIB2
		if ((r->ip_fd = open(IP_DEV_NAME, O_RDWR)) < 0)
			return (route_close(r));
#endif
#ifdef HAVE_STREAMS_ROUTE
		if ((r->fd = open("/dev/route", O_RDWR, 0)) < 0)
#else
		if ((r->fd = socket(PF_ROUTE, SOCK_RAW, AF_INET)) < 0)
#endif
			return (route_close(r));
	}
	return (r);
}

int
route_add(route_t *r, const struct route_entry *entry)
{
	struct route_entry rtent;
	
	memcpy(&rtent, entry, sizeof(rtent));
	
	if (route_msg(r, RTM_ADD, NULL, &rtent.route_dst, &rtent.route_gw) < 0)
		return (-1);
	
	return (0);
}

int
route_delete(route_t *r, const struct route_entry *entry)
{
	struct route_entry rtent;
	
	memcpy(&rtent, entry, sizeof(rtent));
	
	if (route_get(r, &rtent) < 0)
		return (-1);
	
	if (route_msg(r, RTM_DELETE, NULL, &rtent.route_dst, &rtent.route_gw) < 0)
		return (-1);
	
	return (0);
}

int
route_get(route_t *r, struct route_entry *entry)
{
	if (route_msg(r, RTM_GET, entry->intf_name, &entry->route_dst, &entry->route_gw) < 0)
		return (-1);
	entry->intf_name[0] = '\0';
	entry->metric = 0;
	
	return (0);
}

#if defined(HAVE_SYS_SYSCTL_H) || defined(HAVE_STREAMS_ROUTE) || defined(HAVE_GETKERNINFO)
/* This wrapper around addr_ston, on failure, checks for a gateway address
 * family of AF_LINK, and if it finds one, stores an all-zero address of the
 * same type as dst. The all-zero address is a convention for same-subnet
 * routing table entries. */
static int
addr_ston_gateway(const struct addr *dst,
	const struct sockaddr *sa, struct addr *a)
{
	int rc;

	rc = addr_ston(sa, a);
	if (rc == 0)
		return rc;

#ifdef HAVE_NET_IF_DL_H
# ifdef AF_LINK
	if (sa->sa_family == AF_LINK) {
		memset(a, 0, sizeof(*a));
		a->addr_type = dst->addr_type;
		return (0);
	}
# endif
#endif

	return (-1);
}

int
route_loop(route_t *r, route_handler callback, void *arg)
{
	struct rt_msghdr *rtm;
	struct route_entry entry;
	struct sockaddr *sa;
	char *buf, *lim, *next;
	int ret;
#ifdef HAVE_SYS_SYSCTL_H
	int mib[6] = { CTL_NET, PF_ROUTE, 0, 0 /* XXX */, NET_RT_DUMP, 0 };
	size_t len;
	
	if (sysctl(mib, 6, NULL, &len, NULL, 0) < 0)
		return (-1);

	if (len == 0)
		return (0);
	
	if ((buf = malloc(len)) == NULL)
		return (-1);
	
	if (sysctl(mib, 6, buf, &len, NULL, 0) < 0) {
		free(buf);
		return (-1);
	}
	lim = buf + len;
	next = buf;
#elif defined(HAVE_GETKERNINFO)
	int len = getkerninfo(KINFO_RT_DUMP,0,0,0);

	if (len == 0)
		return (0);

	if ((buf = malloc(len)) == NULL)
		return (-1);

	if (getkerninfo(KINFO_RT_DUMP,buf,&len,0) < 0) {
		free(buf);
		return (-1);
	}
	lim = buf + len;
	next = buf;
#else /* HAVE_STREAMS_ROUTE */
	struct rt_giarg giarg, *gp;

	memset(&giarg, 0, sizeof(giarg));
	giarg.gi_op = KINFO_RT_DUMP;

	if (ioctl(r->fd, RTSTR_GETROUTE, &giarg) < 0)
		return (-1);

	if ((buf = malloc(giarg.gi_size)) == NULL)
		return (-1);

	gp = (struct rt_giarg *)buf;
	gp->gi_size = giarg.gi_size;
	gp->gi_op = KINFO_RT_DUMP;
	gp->gi_where = buf;
	gp->gi_arg = RTF_UP | RTF_GATEWAY;

	if (ioctl(r->fd, RTSTR_GETROUTE, buf) < 0) {
		free(buf);
		return (-1);
	}
	lim = buf + gp->gi_size;
	next = buf + sizeof(giarg);
#endif
	/* This loop assumes that RTA_DST, RTA_GATEWAY, and RTA_NETMASK have the
	 * values, 1, 2, and 4 respectively. Cf. Unix Network Programming,
	 * p. 494, function get_rtaddrs. */
	for (ret = 0; next < lim; next += rtm->rtm_msglen) {
		char namebuf[IF_NAMESIZE];
		sa_family_t sfam;
		rtm = (struct rt_msghdr *)next;
		sa = (struct sockaddr *)(rtm + 1);
		/* peek at address family */
		sfam = sa->sa_family;

		if (if_indextoname(rtm->rtm_index, namebuf) == NULL)
			continue;
		strlcpy(entry.intf_name, namebuf, sizeof(entry.intf_name));

		if ((rtm->rtm_addrs & RTA_DST) == 0)
			/* Need a destination. */
			continue;
		if (addr_ston(sa, &entry.route_dst) < 0)
			continue;

		if ((rtm->rtm_addrs & RTA_GATEWAY) == 0)
			/* Need a gateway. */
			continue;
		sa = NEXTSA(sa);
		if (addr_ston_gateway(&entry.route_dst, sa, &entry.route_gw) < 0)
			continue;
		
		if (entry.route_dst.addr_type != entry.route_gw.addr_type ||
		    (entry.route_dst.addr_type != ADDR_TYPE_IP &&
			entry.route_dst.addr_type != ADDR_TYPE_IP6))
			continue;

		if (rtm->rtm_addrs & RTA_NETMASK) {
			sa = NEXTSA(sa);
			/* FreeBSD for IPv6 uses a different AF for netmasks. Force the same one. */
			sa->sa_family = sfam;
			if (addr_stob(sa, &entry.route_dst.addr_bits) < 0)
				continue;
		}

		entry.metric = 0;

		if ((ret = callback(&entry, arg)) != 0)
			break;
	}
	free(buf);
	
	return (ret);
}
#elif defined(HAVE_STREAMS_MIB2)

#ifdef IRE_DEFAULT		/* This means Solaris 5.6 */
/* I'm not sure if they are compatible, though -- masaki */
#define IRE_ROUTE IRE_CACHE
#define IRE_ROUTE_REDIRECT IRE_HOST_REDIRECT
#endif /* IRE_DEFAULT */

int
route_loop(route_t *r, route_handler callback, void *arg)
{
	struct route_entry entry;
	struct strbuf msg;
	struct T_optmgmt_req *tor;
	struct T_optmgmt_ack *toa;
	struct T_error_ack *tea;
	struct opthdr *opt;
	u_char buf[8192];
	int flags, rc, rtable, ret;

	tor = (struct T_optmgmt_req *)buf;
	toa = (struct T_optmgmt_ack *)buf;
	tea = (struct T_error_ack *)buf;

	tor->PRIM_type = T_OPTMGMT_REQ;
	tor->OPT_offset = sizeof(*tor);
	tor->OPT_length = sizeof(*opt);
	tor->MGMT_flags = T_CURRENT;
	
	opt = (struct opthdr *)(tor + 1);
	opt->level = MIB2_IP;
	opt->name = opt->len = 0;
	
	msg.maxlen = sizeof(buf);
	msg.len = sizeof(*tor) + sizeof(*opt);
	msg.buf = buf;
	
	if (putmsg(r->ip_fd, &msg, NULL, 0) < 0)
		return (-1);
	
	opt = (struct opthdr *)(toa + 1);
	msg.maxlen = sizeof(buf);
	
	for (;;) {
		mib2_ipRouteEntry_t *rt, *rtend;

		flags = 0;
		if ((rc = getmsg(r->ip_fd, &msg, NULL, &flags)) < 0)
			return (-1);

		/* See if we're finished. */
		if (rc == 0 &&
		    msg.len >= sizeof(*toa) &&
		    toa->PRIM_type == T_OPTMGMT_ACK &&
		    toa->MGMT_flags == T_SUCCESS && opt->len == 0)
			break;

		if (msg.len >= sizeof(*tea) && tea->PRIM_type == T_ERROR_ACK)
			return (-1);
		
		if (rc != MOREDATA || msg.len < (int)sizeof(*toa) ||
		    toa->PRIM_type != T_OPTMGMT_ACK ||
		    toa->MGMT_flags != T_SUCCESS)
			return (-1);
		
		rtable = (opt->level == MIB2_IP && opt->name == MIB2_IP_21);
		
		msg.maxlen = sizeof(buf) - (sizeof(buf) % sizeof(*rt));
		msg.len = 0;
		flags = 0;
		
		do {
			struct sockaddr_in sin;

			rc = getmsg(r->ip_fd, NULL, &msg, &flags);
			
			if (rc != 0 && rc != MOREDATA)
				return (-1);
			
			if (!rtable)
				continue;
			
			rt = (mib2_ipRouteEntry_t *)msg.buf;
			rtend = (mib2_ipRouteEntry_t *)(msg.buf + msg.len);

			sin.sin_family = AF_INET;

			for ( ; rt < rtend; rt++) {
				if ((rt->ipRouteInfo.re_ire_type &
				    (IRE_BROADCAST|IRE_ROUTE_REDIRECT|
					IRE_LOCAL|IRE_ROUTE)) != 0 ||
				    rt->ipRouteNextHop == IP_ADDR_ANY)
					continue;
				
				entry.intf_name[0] = '\0';

				sin.sin_addr.s_addr = rt->ipRouteNextHop;
				addr_ston((struct sockaddr *)&sin,
				    &entry.route_gw);
				
				sin.sin_addr.s_addr = rt->ipRouteDest;
				addr_ston((struct sockaddr *)&sin,
				    &entry.route_dst);
				
				sin.sin_addr.s_addr = rt->ipRouteMask;
				addr_stob((struct sockaddr *)&sin,
				    &entry.route_dst.addr_bits);

				entry.metric = 0;
				
				if ((ret = callback(&entry, arg)) != 0)
					return (ret);
			}
		} while (rc == MOREDATA);
	}

	tor = (struct T_optmgmt_req *)buf;
	toa = (struct T_optmgmt_ack *)buf;
	tea = (struct T_error_ack *)buf;

	tor->PRIM_type = T_OPTMGMT_REQ;
	tor->OPT_offset = sizeof(*tor);
	tor->OPT_length = sizeof(*opt);
	tor->MGMT_flags = T_CURRENT;
	
	opt = (struct opthdr *)(tor + 1);
	opt->level = MIB2_IP6;
	opt->name = opt->len = 0;
	
	msg.maxlen = sizeof(buf);
	msg.len = sizeof(*tor) + sizeof(*opt);
	msg.buf = buf;
	
	if (putmsg(r->ip_fd, &msg, NULL, 0) < 0)
		return (-1);
	
	opt = (struct opthdr *)(toa + 1);
	msg.maxlen = sizeof(buf);
	
	for (;;) {
		mib2_ipv6RouteEntry_t *rt, *rtend;

		flags = 0;
		if ((rc = getmsg(r->ip_fd, &msg, NULL, &flags)) < 0)
			return (-1);

		/* See if we're finished. */
		if (rc == 0 &&
		    msg.len >= sizeof(*toa) &&
		    toa->PRIM_type == T_OPTMGMT_ACK &&
		    toa->MGMT_flags == T_SUCCESS && opt->len == 0)
			break;

		if (msg.len >= sizeof(*tea) && tea->PRIM_type == T_ERROR_ACK)
			return (-1);
		
		if (rc != MOREDATA || msg.len < (int)sizeof(*toa) ||
		    toa->PRIM_type != T_OPTMGMT_ACK ||
		    toa->MGMT_flags != T_SUCCESS)
			return (-1);
		
		rtable = (opt->level == MIB2_IP6 && opt->name == MIB2_IP6_ROUTE);
		
		msg.maxlen = sizeof(buf) - (sizeof(buf) % sizeof(*rt));
		msg.len = 0;
		flags = 0;
		
		do {
			struct sockaddr_in6 sin6;

			rc = getmsg(r->ip_fd, NULL, &msg, &flags);
			
			if (rc != 0 && rc != MOREDATA)
				return (-1);
			
			if (!rtable)
				continue;
			
			rt = (mib2_ipv6RouteEntry_t *)msg.buf;
			rtend = (mib2_ipv6RouteEntry_t *)(msg.buf + msg.len);

			sin6.sin6_family = AF_INET6;

			for ( ; rt < rtend; rt++) {
				if ((rt->ipv6RouteInfo.re_ire_type &
				    (IRE_BROADCAST|IRE_ROUTE_REDIRECT|
					IRE_LOCAL|IRE_ROUTE)) != 0 ||
				    memcmp(&rt->ipv6RouteNextHop, IP6_ADDR_UNSPEC, IP6_ADDR_LEN) == 0)
					continue;
				
				entry.intf_name[0] = '\0';

				sin6.sin6_addr = rt->ipv6RouteNextHop;
				addr_ston((struct sockaddr *)&sin6,
				    &entry.route_gw);
				
				sin6.sin6_addr = rt->ipv6RouteDest;
				addr_ston((struct sockaddr *)&sin6,
				    &entry.route_dst);
				
				entry.route_dst.addr_bits = rt->ipv6RoutePfxLength;
				
				if ((ret = callback(&entry, arg)) != 0)
					return (ret);
			}
		} while (rc == MOREDATA);
	}
	return (0);
}
#elif defined(HAVE_NET_RADIX_H)
/* XXX - Tru64, others? */
#include <nlist.h>

static int
_kread(int fd, void *addr, void *buf, int len)
{
	if (lseek(fd, (off_t)addr, SEEK_SET) == (off_t)-1L)
		return (-1);
	return (read(fd, buf, len) == len ? 0 : -1);
}

static int
_radix_walk(int fd, struct radix_node *rn, route_handler callback, void *arg)
{
	struct radix_node rnode;
	struct rtentry rt;
	struct sockaddr_in sin;
	struct route_entry entry;
	int ret = 0;
 again:
	_kread(fd, rn, &rnode, sizeof(rnode));
	if (rnode.rn_b < 0) {
		if (!(rnode.rn_flags & RNF_ROOT)) {
			entry.intf_name[0] = '\0';
			_kread(fd, rn, &rt, sizeof(rt));
			_kread(fd, rt_key(&rt), &sin, sizeof(sin));
			addr_ston((struct sockaddr *)&sin, &entry.route_dst);
			if (!(rt.rt_flags & RTF_HOST)) {
				_kread(fd, rt_mask(&rt), &sin, sizeof(sin));
				addr_stob((struct sockaddr *)&sin,
				    &entry.route_dst.addr_bits);
			}
			_kread(fd, rt.rt_gateway, &sin, sizeof(sin));
			addr_ston((struct sockaddr *)&sin, &entry.route_gw);
			entry.metric = 0;
			if ((ret = callback(&entry, arg)) != 0)
				return (ret);
		}
		if ((rn = rnode.rn_dupedkey))
			goto again;
	} else {
		rn = rnode.rn_r;
		if ((ret = _radix_walk(fd, rnode.rn_l, callback, arg)) != 0)
			return (ret);
		if ((ret = _radix_walk(fd, rn, callback, arg)) != 0)
			return (ret);
	}
	return (ret);
}

int
route_loop(route_t *r, route_handler callback, void *arg)
{
	struct radix_node_head *rnh, head;
	struct nlist nl[2];
	int fd, ret = 0;

	memset(nl, 0, sizeof(nl));
	nl[0].n_name = "radix_node_head";
	
	if (knlist(nl) < 0 || nl[0].n_type == 0 ||
	    (fd = open("/dev/kmem", O_RDONLY, 0)) < 0)
		return (-1);
	
	for (_kread(fd, (void *)nl[0].n_value, &rnh, sizeof(rnh));
	    rnh != NULL; rnh = head.rnh_next) {
		_kread(fd, rnh, &head, sizeof(head));
		/* XXX - only IPv4 for now... */
		if (head.rnh_af == AF_INET) {
			if ((ret = _radix_walk(fd, head.rnh_treetop,
				 callback, arg)) != 0)
				break;
		}
	}
	close(fd);
	return (ret);
}
#else
int
route_loop(route_t *r, route_handler callback, void *arg)
{
	errno = ENOSYS;
	return (-1);
}
#endif

route_t *
route_close(route_t *r)
{
	if (r != NULL) {
#ifdef HAVE_STREAMS_MIB2
		if (r->ip_fd >= 0)
			close(r->ip_fd);
#endif
		if (r->fd >= 0)
			close(r->fd);
		free(r);
	}
	return (NULL);
}