xref: /onnv-gate/usr/src/cmd/cmd-inet/usr.sbin/in.routed/output.c (revision 5381:6bff17151099)

/*
 * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 1983, 1988, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgment:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 * $FreeBSD: src/sbin/routed/output.c,v 1.7 2000/08/11 08:24:38 sheldonh Exp $
 */

#pragma ident	"%Z%%M%	%I%	%E% SMI"

#include "defs.h"
#include <md5.h>
#include <alloca.h>

uint_t update_seqno;


/*
 * walk the tree of routes with this for output
 */
static	struct {
	struct sockaddr_in to;
	in_addr_t	to_mask;
	in_addr_t	to_net;
	in_addr_t	to_std_mask;
	in_addr_t	to_std_net;
	struct interface *ifp;		/* usually output interface */
	struct auth	*a;
	uint8_t		metric;		/* adjust metrics by interface */
	uint32_t	npackets;
	uint32_t	gen_limit;
#define	WS_GEN_LIMIT_MAX	1024
	uint16_t	state;
#define	WS_ST_FLASH	0x001	/* send only changed routes */
#define	WS_ST_RIP2_ALL	0x002	/* send full featured RIPv2 */
#define	WS_ST_AG	0x004	/* ok to aggregate subnets */
#define	WS_ST_SUPER_AG	0x008	/* ok to aggregate networks */
#define	WS_ST_QUERY	0x010	/* responding to a query */
#define	WS_ST_TO_ON_NET	0x020	/* sending onto one of our nets */
#define	WS_ST_DEFAULT	0x040	/* faking a default */
} ws;

/* A buffer for what can be heard by both RIPv1 and RIPv2 listeners */
struct ws_buf v12buf;
static	union pkt_buf ripv12_buf;

/* Another for only RIPv2 listeners */
static	struct ws_buf v2buf;
static	union pkt_buf rip_v2_buf;



void
bufinit(void)
{
	ripv12_buf.rip.rip_cmd = RIPCMD_RESPONSE;
	v12buf.buf = &ripv12_buf.rip;
	v12buf.base = &v12buf.buf->rip_nets[0];

	rip_v2_buf.rip.rip_cmd = RIPCMD_RESPONSE;
	rip_v2_buf.rip.rip_vers = RIPv2;
	v2buf.buf = &rip_v2_buf.rip;
	v2buf.base = &v2buf.buf->rip_nets[0];
}


/*
 * Send the contents of the global buffer via the non-multicast socket
 */
int					/* <0 on failure */
output(enum output_type type,
    struct sockaddr_in *dst,		/* send to here */
    struct interface *ifp,
    struct rip *buf,
    int size)			/* this many bytes */
{
	struct sockaddr_in sin;
	int flags;
	const char *msg;
	int res;
	int ifindex;
	struct in_addr addr;

	sin = *dst;
	if (sin.sin_port == 0)
		sin.sin_port = htons(RIP_PORT);

	flags = 0;

	if (ifp == NULL && type == OUT_MULTICAST) {
		msglog("Cannot send RIP message to %s",
		    inet_ntoa(sin.sin_addr));
		return (-1);
	}

	switch (type) {
	case OUT_QUERY:
		msg = "Answer Query";
		break;
	case OUT_UNICAST:
		msg = "Send";
		flags = MSG_DONTROUTE;
		break;
	case OUT_BROADCAST:
		msg = "Send bcast";
		break;
	case OUT_MULTICAST:
		msg = "Send mcast";
		break;

	case NO_OUT_MULTICAST:
	case NO_OUT_RIPV2:
	default:
#ifdef DEBUG
		abort();
#endif
		return (-1);
	}

	/*
	 * IP_PKTINFO overrides IP_MULTICAST_IF, so we don't set ifindex
	 * for multicast traffic.
	 */
	ifindex = (type != OUT_MULTICAST && type != OUT_QUERY &&
	    ifp != NULL && ifp->int_phys != NULL) ?
	    ifp->int_phys->phyi_index : 0;

	if (rip_sock_interface != ifp) {
		/*
		 * For multicast, we have to choose the source
		 * address.  This is either the local address
		 * (non-point-to-point) or the remote address.
		 */
		if (ifp != NULL) {
			addr.s_addr = (ifp->int_if_flags & IFF_POINTOPOINT) ?
			    ifp->int_dstaddr : ifp->int_addr;
			if (type == OUT_MULTICAST &&
			    setsockopt(rip_sock, IPPROTO_IP,
			    IP_MULTICAST_IF, &addr, sizeof (addr)) == -1) {
				LOGERR("setsockopt(rip_sock, IP_MULTICAST_IF)");
				return (-1);
			}
		}
		rip_sock_interface = ifp;
	}

	trace_rip(msg, "to", &sin, ifp, buf, size);

	res = sendtoif(rip_sock, buf, size, flags, &sin, ifindex);
	if (res < 0 && (ifp == NULL || !(ifp->int_state & IS_BROKE))) {
		writelog(LOG_WARNING, "%s sendto(%s%s%s.%d): %s", msg,
		    ifp != NULL ? ifp->int_name : "",
		    ifp != NULL ? ", " : "",
		    inet_ntoa(sin.sin_addr),
		    ntohs(sin.sin_port),
		    rip_strerror(errno));
	}

	return (res);
}

/*
 * Semantically identical to sendto(), but sends the message through a
 * specific interface (if ifindex is non-zero) using IP_PKTINFO.
 */
int
sendtoif(int fd, const void *buf, uint_t bufsize, uint_t flags,
    struct sockaddr_in *sinp, uint_t ifindex)
{
	struct iovec iov;
	struct msghdr msg;
	struct cmsghdr *cmsgp;
	struct in_pktinfo *ipip;

	iov.iov_base = (void *)buf;
	iov.iov_len = bufsize;

	(void) memset(&msg, 0, sizeof (struct msghdr));
	msg.msg_name = (struct sockaddr *)sinp;
	msg.msg_namelen = sizeof (struct sockaddr_in);
	msg.msg_iov = &iov;
	msg.msg_iovlen = 1;

	if (ifindex != 0) {
		/*
		 * We can't precisely predict the alignment padding we'll
		 * need, so we allocate the maximum alignment and then
		 * use CMSG_NXTHDR() to fix it up at the end.
		 */
		msg.msg_controllen = sizeof (*cmsgp) + _MAX_ALIGNMENT +
		    sizeof (*ipip) + _MAX_ALIGNMENT + sizeof (*cmsgp);
		msg.msg_control = alloca(msg.msg_controllen);

		cmsgp = CMSG_FIRSTHDR(&msg);
		ipip = (void *)CMSG_DATA(cmsgp);
		(void) memset(ipip, 0, sizeof (struct in_pktinfo));
		ipip->ipi_ifindex = ifindex;
		cmsgp->cmsg_len = (caddr_t)(ipip + 1) - (caddr_t)cmsgp;
		cmsgp->cmsg_type = IP_PKTINFO;
		cmsgp->cmsg_level = IPPROTO_IP;

		/*
		 * Correct the control message length.
		 */
		cmsgp = CMSG_NXTHDR(&msg, cmsgp);
		msg.msg_controllen = (caddr_t)cmsgp - (caddr_t)msg.msg_control;
	}

	return (sendmsg(fd, &msg, flags));
}

/*
 * Find the first key for a packet to send.
 * Try for a key that is eligible and has not expired, but settle for
 * the last key if they have all expired.
 * If no key is ready yet, give up.
 */
struct auth *
find_auth(struct interface *ifp)
{
	struct auth *ap, *res = NULL;
	int i;


	if (ifp == NULL)
		return (NULL);

	if ((ap = ifp->int_auth) == NULL)
		return (NULL);

	for (i = 0; i < MAX_AUTH_KEYS; i++, ap++) {
		/* stop looking after the last key */
		if (ap->type == RIP_AUTH_NONE)
			break;

		/* ignore keys that are not ready yet */
		if ((ulong_t)ap->start > (ulong_t)clk.tv_sec)
			continue;

		if ((ulong_t)ap->end < (ulong_t)clk.tv_sec) {
			/* note best expired password as a fall-back */
			if (res == NULL ||
			    (((ulong_t)ap->end > (ulong_t)res->end)) &&
			    ((ulong_t)res->end < (ulong_t)clk.tv_sec))
				res = ap;
			continue;
		}

		/* note key with the best future */
		if (res == NULL || (ulong_t)res->end < (ulong_t)ap->end)
			res = ap;
	}
	return (res);
}


void
clr_ws_buf(struct ws_buf *wb, struct auth *ap)
{
	struct netauth *na;

	wb->lim = wb->base + NETS_LEN;
	wb->n = wb->base;
	(void) memset(wb->n, 0, NETS_LEN*sizeof (*wb->n));

	/*
	 * (start to) install authentication if appropriate
	 */
	if (ap == NULL)
		return;

	na = (struct netauth *)wb->n;
	if (ap->type == RIP_AUTH_PW) {
		na->a_family = RIP_AF_AUTH;
		na->a_type = RIP_AUTH_PW;
		(void) memcpy(na->au.au_pw, ap->key, sizeof (na->au.au_pw));
		wb->n++;

	} else if (ap->type ==  RIP_AUTH_MD5) {
		na->a_family = RIP_AF_AUTH;
		na->a_type = RIP_AUTH_MD5;
		na->au.a_md5.md5_keyid = ap->keyid;
		na->au.a_md5.md5_auth_len = RIP_AUTH_MD5_LEN;
		na->au.a_md5.md5_seqno = htonl(clk.tv_sec);
		wb->n++;
		wb->lim--;		/* make room for trailer */
	}
}


void
end_md5_auth(struct ws_buf *wb, struct auth *ap)
{
	struct netauth *na, *na2;
	MD5_CTX md5_ctx;
	int len;

	na = (struct netauth *)wb->base;
	na2 = (struct netauth *)wb->n;
	len = (char *)na2-(char *)wb->buf;
	na2->a_family = RIP_AF_AUTH;
	na2->a_type = RIP_AUTH_TRAILER;
	na->au.a_md5.md5_pkt_len = htons(len);
	MD5Init(&md5_ctx);
	/* len+4 to include auth trailer's family/type in MD5 sum */
	MD5Update(&md5_ctx, (uchar_t *)wb->buf, len + 4);
	MD5Update(&md5_ctx, ap->key, RIP_AUTH_MD5_LEN);
	MD5Final(na2->au.au_pw, &md5_ctx);
	wb->n++;
}


/*
 * Send the buffer
 */
static void
supply_write(struct ws_buf *wb)
{
	/*
	 * Output multicast only if legal.
	 * If we would multicast and it would be illegal, then discard the
	 * packet.
	 */
	switch (wb->type) {
	case NO_OUT_MULTICAST:
		trace_pkt("skip multicast to %s because impossible",
		    naddr_ntoa(ws.to.sin_addr.s_addr));
		break;
	case NO_OUT_RIPV2:
		break;
	default:
		if (ws.a != NULL && ws.a->type == RIP_AUTH_MD5)
			end_md5_auth(wb, ws.a);
		if (output(wb->type, &ws.to, ws.ifp, wb->buf,
		    ((char *)wb->n - (char *)wb->buf)) < 0 && ws.ifp != NULL)
			if_sick(ws.ifp, _B_FALSE);
		ws.npackets++;
		break;
	}

	clr_ws_buf(wb, ws.a);
}


/*
 * Put an entry into the packet
 */
static void
supply_out(struct ag_info *ag)
{
	uint32_t dstcount;
	in_addr_t mask, v1_mask, dst_h, ddst_h = 0;
	struct ws_buf *wb;


	/*
	 * Skip this route if doing a flash update and it and the routes
	 * it aggregates have not changed recently.
	 */
	if (ag->ag_seqno < update_seqno && (ws.state & WS_ST_FLASH))
		return;

	dst_h = ag->ag_dst_h;
	mask = ag->ag_mask;
	v1_mask = ripv1_mask_host(htonl(dst_h),
	    (ws.state & WS_ST_TO_ON_NET) ? ws.ifp : NULL);
	dstcount = 0;

	/*
	 * If we are sending RIPv2 packets that cannot (or must not) be
	 * heard by RIPv1 listeners, do not worry about sub- or supernets.
	 * Subnets (from other networks) can only be sent via multicast.
	 * A pair of subnet routes might have been promoted so that they
	 * are legal to send by RIPv1.
	 * If RIPv1 is off, use the multicast buffer.
	 */
	if ((ws.state & WS_ST_RIP2_ALL) ||
	    ((ag->ag_state & AGS_RIPV2) && v1_mask != mask)) {
		/* use the RIPv2-only buffer */
		wb = &v2buf;

	} else {
		/*
		 * use the RIPv1-or-RIPv2 buffer
		 */
		wb = &v12buf;

		/*
		 * Convert supernet route into corresponding set of network
		 * routes for RIPv1, but leave non-contiguous netmasks
		 * to ag_check().
		 */
		if (v1_mask > mask &&
		    mask + (mask & -mask) == 0) {
			ddst_h = v1_mask & -v1_mask;
			dstcount = (v1_mask & ~mask)/ddst_h;

			if (dstcount > ws.gen_limit) {
				/*
				 * Punt if we would have to generate an
				 * unreasonable number of routes.
				 */
				if (TRACECONTENTS)
					trace_misc("sending %s-->%s as 1"
					    " instead of %d routes",
					    addrname(htonl(dst_h), mask, 1),
					    naddr_ntoa(ws.to.sin_addr.s_addr),
					    dstcount + 1);
				dstcount = 0;

			} else {
				mask = v1_mask;
				ws.gen_limit -= dstcount;
			}
		}
	}

	do {
		wb->n->n_family = RIP_AF_INET;
		wb->n->n_dst = htonl(dst_h);
		/*
		 * If the route is from router-discovery or we are
		 * shutting down, or this is a broken/sick interface,
		 * admit only a bad metric.
		 */
		wb->n->n_metric = ((stopint || ag->ag_metric < 1 ||
		    (ag->ag_ifp && (ag->ag_ifp->int_state &
		    (IS_BROKE|IS_SICK)))) ? HOPCNT_INFINITY : ag->ag_metric);
		wb->n->n_metric = htonl(wb->n->n_metric);
		/*
		 * Any non-zero bits in the supposedly unused RIPv1 fields
		 * cause the old `routed` to ignore the route.
		 * That means the mask and so forth cannot be sent
		 * in the hybrid RIPv1/RIPv2 mode.
		 */
		if (ws.state & WS_ST_RIP2_ALL) {
			if (ag->ag_nhop != 0 &&
			    ((ws.state & WS_ST_QUERY) ||
			    (ag->ag_nhop != ws.ifp->int_addr &&
			    on_net(ag->ag_nhop, ws.ifp->int_net,
			    ws.ifp->int_mask)) &&
			    ifwithaddr(ag->ag_nhop, _B_FALSE, _B_FALSE) ==
			    NULL))
				wb->n->n_nhop = ag->ag_nhop;
			wb->n->n_mask = htonl(mask);
			wb->n->n_tag = ag->ag_tag;
		}
		dst_h += ddst_h;

		if (++wb->n >= wb->lim)
			supply_write(wb);
	} while (dstcount-- > 0);
}


/*
 * Supply one route from the table
 */
/* ARGSUSED */
static int
walk_supply(struct radix_node *rn, void *argp)
{
#define	RT ((struct rt_entry *)rn)
	ushort_t ags;
	uint8_t metric, pref;
	in_addr_t dst, nhop;
	struct rt_spare *rts;
	uint_t sparecount;


	/*
	 * Do not advertise external remote interfaces or passive interfaces.
	 */
	if ((RT->rt_state & RS_IF) && RT->rt_ifp != NULL &&
	    (RT->rt_ifp->int_state & IS_PASSIVE) &&
	    !(RT->rt_state & RS_MHOME))
		return (0);
	/*
	 * Do not advertise routes learnt from /etc/gateways.
	 */
	if (RT->rt_spares[0].rts_origin == RO_FILE)
		return (0);

	/*
	 * Do not advertise routes which would lead to forwarding on a
	 * non-forwarding interface.
	 */
	if (RT->rt_state & RS_NOPROPAGATE)
		return (0);

	/*
	 * If being quiet about our ability to forward, then
	 * do not say anything unless responding to a query,
	 * except about our main interface.
	 */
	if (!should_supply(NULL) && !(ws.state & WS_ST_QUERY) &&
	    !(RT->rt_state & RS_MHOME))
		return (0);

	dst = RT->rt_dst;

	/*
	 * do not collide with the fake default route
	 */
	if (dst == RIP_DEFAULT && (ws.state & WS_ST_DEFAULT))
		return (0);

	if (RT->rt_state & RS_NET_SYN) {
		if (RT->rt_state & RS_NET_INT) {
			/*
			 * Do not send manual synthetic network routes
			 * into the subnet.
			 */
			if (on_net(ws.to.sin_addr.s_addr,
			    ntohl(dst), RT->rt_mask))
				return (0);

		} else {
			/*
			 * Do not send automatic synthetic network routes
			 * if they are not needed because no RIPv1 listeners
			 * can hear them.
			 */
			if (ws.state & WS_ST_RIP2_ALL)
				return (0);

			/*
			 * Do not send automatic synthetic network routes to
			 * the real subnet.
			 */
			if (on_net(ws.to.sin_addr.s_addr,
			    ntohl(dst), RT->rt_mask))
				return (0);
		}
		nhop = 0;

	} else {
		/*
		 * Advertise the next hop if this is not a route for one
		 * of our interfaces and the next hop is on the same
		 * network as the target.
		 * The final determination is made by supply_out().
		 */
		if (!(RT->rt_state & RS_IF) && !(RT->rt_state & RS_MHOME) &&
		    RT->rt_gate != loopaddr)
			nhop = RT->rt_gate;
		else
			nhop = 0;
	}

	metric = RT->rt_metric;
	ags = 0;

	if (!RT_ISHOST(RT)) {
		/*
		 * Always suppress network routes into other, existing
		 * network routes
		 */
		ags |= AGS_SUPPRESS;

		/*
		 * Generate supernets if allowed.
		 * If we can be heard by RIPv1 systems, we will
		 * later convert back to ordinary nets.
		 * This unifies dealing with received supernets.
		 */
		if ((ws.state & WS_ST_AG) && ((RT->rt_state & RS_SUBNET) ||
		    (ws.state & WS_ST_SUPER_AG)))
			ags |= AGS_AGGREGATE;
	} else if (!(RT->rt_state & RS_MHOME)) {
		/*
		 * We should always suppress (into existing network routes)
		 * the host routes for the local end of our point-to-point
		 * links.
		 * If we are suppressing host routes in general, then do so.
		 * Avoid advertising host routes onto their own network,
		 * where they should be handled by proxy-ARP.
		 */
		if ((RT->rt_state & RS_LOCAL) || ridhosts ||
		    on_net(dst, ws.to_net, ws.to_mask))
			ags |= AGS_SUPPRESS;

		/*
		 * Aggregate stray host routes into network routes if allowed.
		 * We cannot aggregate host routes into small network routes
		 * without confusing RIPv1 listeners into thinking the
		 * network routes are host routes.
		 */
		if ((ws.state & WS_ST_AG) && (ws.state & WS_ST_RIP2_ALL))
			ags |= AGS_AGGREGATE;
	}

	/*
	 * Do not send RIPv1 advertisements of subnets to other
	 * networks. If possible, multicast them by RIPv2.
	 */
	if ((RT->rt_state & RS_SUBNET) && !(ws.state & WS_ST_RIP2_ALL) &&
	    !on_net(dst, ws.to_std_net, ws.to_std_mask))
		ags |= AGS_RIPV2 | AGS_AGGREGATE;


	/*
	 * Do not send a route back to where it came from, except in
	 * response to a query.  This is "split-horizon".  That means not
	 * advertising back to the same network	and so via the same interface.
	 *
	 * We want to suppress routes that might have been fragmented
	 * from this route by a RIPv1 router and sent back to us, and so we
	 * cannot forget this route here.  Let the split-horizon route
	 * suppress the fragmented routes and then itself be forgotten.
	 *
	 * Include the routes for both ends of point-to-point interfaces
	 * among those suppressed by split-horizon, since the other side
	 * should knows them as well as we do.
	 *
	 * Notice spare routes with the same metric that we are about to
	 * advertise, to split the horizon on redundant, inactive paths.
	 */
	if (ws.ifp != NULL && !(ws.state & WS_ST_QUERY) &&
	    (ws.state & WS_ST_TO_ON_NET) && (!(RT->rt_state & RS_IF) ||
	    (ws.ifp->int_if_flags & IFF_POINTOPOINT))) {
		for (rts = RT->rt_spares, sparecount = 0;
		    sparecount < RT->rt_num_spares; sparecount++, rts++) {
			if (rts->rts_metric > metric || rts->rts_ifp != ws.ifp)
				continue;

			/*
			 * If we do not mark the route with AGS_SPLIT_HZ here,
			 * it will be poisoned-reverse, or advertised back
			 * toward its source with an infinite metric.
			 * If we have recently advertised the route with a
			 * better metric than we now have, then we should
			 * poison-reverse the route before suppressing it for
			 * split-horizon.
			 *
			 * In almost all cases, if there is no spare for the
			 * route then it is either old and dead or a brand
			 * new route. If it is brand new, there is no need
			 * for poison-reverse. If it is old and dead, it
			 * is already poisoned.
			 */
			if (RT->rt_poison_time < now_expire ||
			    RT->rt_poison_metric >= metric ||
			    RT->rt_spares[1].rts_gate == 0) {
				ags |= AGS_SPLIT_HZ;
				ags &= ~AGS_SUPPRESS;
			}
			metric = HOPCNT_INFINITY;
			break;
		}
	}

	/*
	 * Keep track of the best metric with which the
	 * route has been advertised recently.
	 */
	if (RT->rt_poison_metric >= metric ||
	    RT->rt_poison_time < now_expire) {
		RT->rt_poison_time = now.tv_sec;
		RT->rt_poison_metric = metric;
	}

	/*
	 * Adjust the outgoing metric by the cost of the link.
	 * Avoid aggregation when a route is counting to infinity.
	 */
	pref = RT->rt_poison_metric + ws.metric;
	metric += ws.metric;

	/*
	 * If this is a static route pointing to the same interface
	 * upon which we are sending out the RIP RESPONSE
	 * adjust the preference so that we don't aggregate into this
	 * route. Note that the maximum possible hop count on a route
	 * per RFC 2453 is 16 (HOPCNT_INFINITY)
	 */
	if ((RT->rt_state & RS_STATIC) && (ws.ifp == RT->rt_ifp))
		pref = (HOPCNT_INFINITY+1);

	/*
	 * Do not advertise stable routes that will be ignored,
	 * unless we are answering a query.
	 * If the route recently was advertised with a metric that
	 * would have been less than infinity through this interface,
	 * we need to continue to advertise it in order to poison it.
	 */
	if (metric >= HOPCNT_INFINITY) {
		if (!(ws.state & WS_ST_QUERY) && (pref >= HOPCNT_INFINITY ||
		    RT->rt_poison_time < now_garbage))
			return (0);

		metric = HOPCNT_INFINITY;
	}

	/*
	 * supply this route out on the wire- we only care about dest/mask
	 * and so can ignore all rt_spares[i] with i > 0
	 */
	ag_check(dst, RT->rt_mask, 0, RT->rt_ifp, nhop, metric, pref,
	    RT->rt_seqno, RT->rt_tag, ags, supply_out);
	return (0);
#undef RT
}


/*
 * Supply dst with the contents of the routing tables.
 * If this won't fit in one packet, chop it up into several.
 */
void
supply(struct sockaddr_in *dst,
    struct interface *ifp,	/* output interface */
    enum output_type type,
    int flash,			/* 1=flash update */
    int vers,			/* RIP version */
    boolean_t passwd_ok)	/* OK to include cleartext password */
{
	struct rt_entry *rt;
	uint8_t def_metric;


	ws.state = 0;
	ws.gen_limit = WS_GEN_LIMIT_MAX;

	ws.to = *dst;
	ws.to_std_mask = std_mask(ws.to.sin_addr.s_addr);
	ws.to_std_net = ntohl(ws.to.sin_addr.s_addr) & ws.to_std_mask;

	if (ifp != NULL) {
		ws.to_mask = ifp->int_mask;
		ws.to_net = ifp->int_net;
		if (on_net(ws.to.sin_addr.s_addr, ws.to_net, ws.to_mask) ||
		    type == OUT_MULTICAST)
			ws.state |= WS_ST_TO_ON_NET;

	} else {
		ws.to_mask = ripv1_mask_net(ws.to.sin_addr.s_addr, NULL);
		ws.to_net = ntohl(ws.to.sin_addr.s_addr) & ws.to_mask;
		rt = rtfind(dst->sin_addr.s_addr);
		if (rt != NULL)
			ifp = rt->rt_ifp;
		else
			return;
	}

	ws.npackets = 0;
	if (flash)
		ws.state |= WS_ST_FLASH;

	ws.ifp = ifp;

	/*
	 * Routes in the table were already adjusted by their respective
	 * destination interface costs (which are zero by default) on
	 * input.  The following is the value by which each route's metric
	 * will be bumped up on output.
	 */
	ws.metric = 1;

	ripv12_buf.rip.rip_vers = vers;

	switch (type) {
	case OUT_MULTICAST:
		if (ifp->int_if_flags & IFF_MULTICAST)
			v2buf.type = OUT_MULTICAST;
		else
			v2buf.type = NO_OUT_MULTICAST;
		v12buf.type = OUT_BROADCAST;
		break;

	case OUT_QUERY:
		ws.state |= WS_ST_QUERY;
		/* FALLTHROUGH */
	case OUT_BROADCAST:
	case OUT_UNICAST:
		v2buf.type = (vers == RIPv2) ? type : NO_OUT_RIPV2;
		v12buf.type = type;
		break;

	case NO_OUT_MULTICAST:
	case NO_OUT_RIPV2:
		return;			/* no output */
	}

	if (vers == RIPv2) {
		/* full RIPv2 only if cannot be heard by RIPv1 listeners */
		if (type != OUT_BROADCAST)
			ws.state |= WS_ST_RIP2_ALL;
		if ((ws.state & WS_ST_QUERY) || !(ws.state & WS_ST_TO_ON_NET)) {
			ws.state |= (WS_ST_AG | WS_ST_SUPER_AG);
		} else if (ifp == NULL || !(ifp->int_state & IS_NO_AG)) {
			ws.state |= WS_ST_AG;
			if (type != OUT_BROADCAST && (ifp == NULL ||
			    !(ifp->int_state & IS_NO_SUPER_AG)))
				ws.state |= WS_ST_SUPER_AG;
		}

		/* See if this packet needs authenticating */
		ws.a = find_auth(ifp);
		if (!passwd_ok && ws.a != NULL && ws.a->type == RIP_AUTH_PW)
			ws.a = NULL;
		if (ws.a != NULL && (ulong_t)ws.a->end < (ulong_t)clk.tv_sec &&
		    !ws.a->warnedflag) {
			/*
			 * If the best key is an expired one, we may as
			 * well use it.  Log this event.
			 */
			writelog(LOG_WARNING,
			    "Using expired auth while transmitting to %s",
			    naddr_ntoa(ws.to.sin_addr.s_addr));
			ws.a->warnedflag = 1;
		}
	} else {
		ws.a = NULL;
	}

	clr_ws_buf(&v12buf, ws.a);
	clr_ws_buf(&v2buf, ws.a);

	/*
	 * Fake a default route if asked and if there is not already
	 * a better, real default route.
	 */
	if (should_supply(NULL) && (def_metric = ifp->int_d_metric) != 0) {
		if (NULL == (rt = rtget(RIP_DEFAULT, 0)) ||
		    rt->rt_metric+ws.metric >= def_metric) {
			ws.state |= WS_ST_DEFAULT;
			ag_check(0, 0, 0, NULL, 0, def_metric, def_metric,
			    0, 0, 0, supply_out);
		} else {
			def_metric = rt->rt_metric+ws.metric;
		}

		/*
		 * If both RIPv2 and the poor-man's router discovery
		 * kludge are on, arrange to advertise an extra
		 * default route via RIPv1.
		 */
		if ((ws.state & WS_ST_RIP2_ALL) &&
		    (ifp->int_state & IS_PM_RDISC)) {
			ripv12_buf.rip.rip_vers = RIPv1;
			v12buf.n->n_family = RIP_AF_INET;
			v12buf.n->n_dst = htonl(RIP_DEFAULT);
			v12buf.n->n_metric = htonl(def_metric);
			v12buf.n++;
		}
	}

	(void) rn_walktree(rhead, walk_supply, NULL);
	ag_flush(0, 0, supply_out);

	/*
	 * Flush the packet buffers, provided they are not empty and
	 * do not contain only the password.
	 */
	if (v12buf.n != v12buf.base &&
	    (v12buf.n > v12buf.base+1 ||
	    v12buf.base->n_family != RIP_AF_AUTH))
		supply_write(&v12buf);
	if (v2buf.n != v2buf.base && (v2buf.n > v2buf.base+1 ||
	    v2buf.base->n_family != RIP_AF_AUTH))
		supply_write(&v2buf);

	/*
	 * If we sent nothing and this is an answer to a query, send
	 * an empty buffer.
	 */
	if (ws.npackets == 0 && (ws.state & WS_ST_QUERY)) {
		supply_write(&v2buf);
		if (ws.npackets == 0)
			supply_write(&v12buf);
	}
}


/*
 * send all of the routing table or just do a flash update
 */
void
rip_bcast(int flash)
{
	static struct sockaddr_in dst = {AF_INET};
	struct interface *ifp;
	enum output_type type;
	int vers;
	struct timeval rtime;


	need_flash = _B_FALSE;
	intvl_random(&rtime, MIN_WAITTIME, MAX_WAITTIME);
	no_flash = rtime;
	timevaladd(&no_flash, &now);

	if (!rip_enabled)
		return;

	trace_act("send %s and inhibit dynamic updates for %.3f sec",
	    flash ? "dynamic update" : "all routes",
	    rtime.tv_sec + ((double)rtime.tv_usec)/1000000.0);

	for (ifp = ifnet; ifp != NULL; ifp = ifp->int_next) {
		/*
		 * Skip interfaces not doing RIP or for which IP
		 * forwarding isn't turned on.  Skip duplicate
		 * interfaces, we don't want to generate duplicate
		 * packets.  Do try broken interfaces to see if they
		 * have healed.
		 */
		if (IS_RIP_OUT_OFF(ifp->int_state) ||
		    (ifp->int_state & IS_DUP) ||
		    !IS_IFF_ROUTING(ifp->int_if_flags))
			continue;

		/* skip turned off interfaces */
		if (!IS_IFF_UP(ifp->int_if_flags))
			continue;

		/* skip interfaces we shouldn't use */
		if (IS_IFF_QUIET(ifp->int_if_flags))
			continue;

		vers = (ifp->int_state & IS_NO_RIPV1_OUT) ? RIPv2 : RIPv1;
		dst.sin_addr.s_addr = ifp->int_ripout_addr;

		/*
		 * Ignore the interface if it's not broadcast,
		 * point-to-point, or remote.  It must be non-broadcast
		 * multiaccess, and therefore unsupported.
		 */
		if (!(ifp->int_if_flags & (IFF_BROADCAST | IFF_POINTOPOINT)) &&
		    !(ifp->int_state & IS_REMOTE))
			continue;

		type = (ifp->int_if_flags & IFF_BROADCAST) ?
		    OUT_BROADCAST : OUT_UNICAST;
		if (vers == RIPv2 && (ifp->int_if_flags & IFF_MULTICAST) &&
		    !(ifp->int_state & IS_NO_RIP_MCAST))
			type = OUT_MULTICAST;

		supply(&dst, ifp, type, flash, vers, _B_TRUE);
	}

	update_seqno++;			/* all routes are up to date */
}


/*
 * Ask for routes
 * Do it only once to an interface, and not even after the interface
 * was broken and recovered.
 */
void
rip_query(void)
{
	static struct sockaddr_in dst = {AF_INET};
	struct interface *ifp;
	struct rip buf;
	enum output_type type;


	if (!rip_enabled)
		return;

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

	for (ifp = ifnet; ifp; ifp = ifp->int_next) {
		/*
		 * Skip interfaces those already queried.  Do not ask
		 * via interfaces through which we don't accept input.
		 * Do not ask via interfaces that cannot send RIP
		 * packets.  Don't send queries on duplicate
		 * interfaces, that would generate duplicate packets
		 * on link.  Do try broken interfaces to see if they
		 * have healed.
		 */
		if (IS_RIP_IN_OFF(ifp->int_state) ||
		    (ifp->int_state & IS_DUP) ||
		    ifp->int_query_time != NEVER)
			continue;

		/* skip turned off interfaces */
		if (!IS_IFF_UP(ifp->int_if_flags))
			continue;

		/* skip interfaces we shouldn't use */
		if (IS_IFF_QUIET(ifp->int_if_flags))
			continue;

		/*
		 * Ignore the interface if it's not broadcast,
		 * point-to-point, or remote.  It must be non-broadcast
		 * multiaccess, and therefore unsupported.
		 */
		if (!(ifp->int_if_flags & (IFF_BROADCAST | IFF_POINTOPOINT)) &&
		    !(ifp->int_state & IS_REMOTE))
			continue;

		buf.rip_cmd = RIPCMD_REQUEST;
		buf.rip_nets[0].n_family = RIP_AF_UNSPEC;
		buf.rip_nets[0].n_metric = htonl(HOPCNT_INFINITY);

		/*
		 * Send a RIPv1 query only if allowed and if we will
		 * listen to RIPv1 routers.
		 */
		if ((ifp->int_state & IS_NO_RIPV1_OUT) ||
		    (ifp->int_state & IS_NO_RIPV1_IN)) {
			buf.rip_vers = RIPv2;
		} else {
			buf.rip_vers = RIPv1;
		}

		dst.sin_addr.s_addr = ifp->int_ripout_addr;

		type = (ifp->int_if_flags & IFF_BROADCAST) ?
		    OUT_BROADCAST : OUT_UNICAST;
		if (buf.rip_vers == RIPv2 &&
		    (ifp->int_if_flags & IFF_MULTICAST) &&
		    !(ifp->int_state & IS_NO_RIP_MCAST))
			type = OUT_MULTICAST;

		ifp->int_query_time = now.tv_sec+SUPPLY_INTERVAL;
		if (output(type, &dst, ifp, &buf, sizeof (buf)) < 0)
			if_sick(ifp, _B_FALSE);
	}
}