Home | History | Annotate | Line # | Download | only in netinet
ip_output.c revision 1.75
      1 /*	$NetBSD: ip_output.c,v 1.75 2000/06/28 02:59:32 mrg Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*-
     33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
     34  * All rights reserved.
     35  *
     36  * This code is derived from software contributed to The NetBSD Foundation
     37  * by Public Access Networks Corporation ("Panix").  It was developed under
     38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     39  *
     40  * Redistribution and use in source and binary forms, with or without
     41  * modification, are permitted provided that the following conditions
     42  * are met:
     43  * 1. Redistributions of source code must retain the above copyright
     44  *    notice, this list of conditions and the following disclaimer.
     45  * 2. Redistributions in binary form must reproduce the above copyright
     46  *    notice, this list of conditions and the following disclaimer in the
     47  *    documentation and/or other materials provided with the distribution.
     48  * 3. All advertising materials mentioning features or use of this software
     49  *    must display the following acknowledgement:
     50  *	This product includes software developed by the NetBSD
     51  *	Foundation, Inc. and its contributors.
     52  * 4. Neither the name of The NetBSD Foundation nor the names of its
     53  *    contributors may be used to endorse or promote products derived
     54  *    from this software without specific prior written permission.
     55  *
     56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66  * POSSIBILITY OF SUCH DAMAGE.
     67  */
     68 
     69 /*
     70  * Copyright (c) 1982, 1986, 1988, 1990, 1993
     71  *	The Regents of the University of California.  All rights reserved.
     72  *
     73  * Redistribution and use in source and binary forms, with or without
     74  * modification, are permitted provided that the following conditions
     75  * are met:
     76  * 1. Redistributions of source code must retain the above copyright
     77  *    notice, this list of conditions and the following disclaimer.
     78  * 2. Redistributions in binary form must reproduce the above copyright
     79  *    notice, this list of conditions and the following disclaimer in the
     80  *    documentation and/or other materials provided with the distribution.
     81  * 3. All advertising materials mentioning features or use of this software
     82  *    must display the following acknowledgement:
     83  *	This product includes software developed by the University of
     84  *	California, Berkeley and its contributors.
     85  * 4. Neither the name of the University nor the names of its contributors
     86  *    may be used to endorse or promote products derived from this software
     87  *    without specific prior written permission.
     88  *
     89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     99  * SUCH DAMAGE.
    100  *
    101  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
    102  */
    103 
    104 #include "opt_pfil_hooks.h"
    105 #include "opt_ipsec.h"
    106 #include "opt_mrouting.h"
    107 
    108 #include <sys/param.h>
    109 #include <sys/malloc.h>
    110 #include <sys/mbuf.h>
    111 #include <sys/errno.h>
    112 #include <sys/protosw.h>
    113 #include <sys/socket.h>
    114 #include <sys/socketvar.h>
    115 #include <sys/systm.h>
    116 #include <sys/proc.h>
    117 
    118 #include <net/if.h>
    119 #include <net/route.h>
    120 #include <net/pfil.h>
    121 
    122 #include <netinet/in.h>
    123 #include <netinet/in_systm.h>
    124 #include <netinet/ip.h>
    125 #include <netinet/in_pcb.h>
    126 #include <netinet/in_var.h>
    127 #include <netinet/ip_var.h>
    128 
    129 #ifdef MROUTING
    130 #include <netinet/ip_mroute.h>
    131 #endif
    132 
    133 #ifdef __vax__
    134 #include <machine/mtpr.h>
    135 #endif
    136 
    137 #include <machine/stdarg.h>
    138 
    139 #ifdef IPSEC
    140 #include <netinet6/ipsec.h>
    141 #include <netkey/key.h>
    142 #include <netkey/key_debug.h>
    143 #endif /*IPSEC*/
    144 
    145 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
    146 static void ip_mloopback
    147 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
    148 
    149 /*
    150  * IP output.  The packet in mbuf chain m contains a skeletal IP
    151  * header (with len, off, ttl, proto, tos, src, dst).
    152  * The mbuf chain containing the packet will be freed.
    153  * The mbuf opt, if present, will not be freed.
    154  */
    155 int
    156 #if __STDC__
    157 ip_output(struct mbuf *m0, ...)
    158 #else
    159 ip_output(m0, va_alist)
    160 	struct mbuf *m0;
    161 	va_dcl
    162 #endif
    163 {
    164 	struct ip *ip, *mhip;
    165 	struct ifnet *ifp;
    166 	struct mbuf *m = m0;
    167 	int hlen = sizeof (struct ip);
    168 	int len, off, error = 0;
    169 	struct route iproute;
    170 	struct sockaddr_in *dst;
    171 #if IFA_STATS
    172 	struct sockaddr_in src;
    173 #endif
    174 	struct in_ifaddr *ia;
    175 	struct mbuf *opt;
    176 	struct route *ro;
    177 	int flags;
    178 	int *mtu_p;
    179 	int mtu;
    180 	struct ip_moptions *imo;
    181 	va_list ap;
    182 #ifdef PFIL_HOOKS
    183 	struct packet_filter_hook *pfh;
    184 	struct mbuf *m1;
    185 	int rv;
    186 #endif /* PFIL_HOOKS */
    187 #ifdef IPSEC
    188 	struct socket *so;
    189 	struct secpolicy *sp = NULL;
    190 #endif /*IPSEC*/
    191 
    192 	va_start(ap, m0);
    193 	opt = va_arg(ap, struct mbuf *);
    194 	ro = va_arg(ap, struct route *);
    195 	flags = va_arg(ap, int);
    196 	imo = va_arg(ap, struct ip_moptions *);
    197 	if (flags & IP_RETURNMTU)
    198 		mtu_p = va_arg(ap, int *);
    199 	else
    200 		mtu_p = NULL;
    201 	va_end(ap);
    202 
    203 #ifdef IPSEC
    204 	so = ipsec_getsocket(m);
    205 	ipsec_setsocket(m, NULL);
    206 #endif /*IPSEC*/
    207 
    208 #ifdef	DIAGNOSTIC
    209 	if ((m->m_flags & M_PKTHDR) == 0)
    210 		panic("ip_output no HDR");
    211 #endif
    212 	if (opt) {
    213 		m = ip_insertoptions(m, opt, &len);
    214 		hlen = len;
    215 	}
    216 	ip = mtod(m, struct ip *);
    217 	/*
    218 	 * Fill in IP header.
    219 	 */
    220 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
    221 		ip->ip_v = IPVERSION;
    222 		ip->ip_off &= IP_DF;
    223 		ip->ip_id = htons(ip_id++);
    224 		ip->ip_hl = hlen >> 2;
    225 		ipstat.ips_localout++;
    226 	} else {
    227 		hlen = ip->ip_hl << 2;
    228 	}
    229 	/*
    230 	 * Route packet.
    231 	 */
    232 	if (ro == 0) {
    233 		ro = &iproute;
    234 		bzero((caddr_t)ro, sizeof (*ro));
    235 	}
    236 	dst = satosin(&ro->ro_dst);
    237 	/*
    238 	 * If there is a cached route,
    239 	 * check that it is to the same destination
    240 	 * and is still up.  If not, free it and try again.
    241 	 */
    242 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    243 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
    244 		RTFREE(ro->ro_rt);
    245 		ro->ro_rt = (struct rtentry *)0;
    246 	}
    247 	if (ro->ro_rt == 0) {
    248 		dst->sin_family = AF_INET;
    249 		dst->sin_len = sizeof(*dst);
    250 		dst->sin_addr = ip->ip_dst;
    251 	}
    252 	/*
    253 	 * If routing to interface only,
    254 	 * short circuit routing lookup.
    255 	 */
    256 	if (flags & IP_ROUTETOIF) {
    257 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
    258 			ipstat.ips_noroute++;
    259 			error = ENETUNREACH;
    260 			goto bad;
    261 		}
    262 		ifp = ia->ia_ifp;
    263 		mtu = ifp->if_mtu;
    264 		ip->ip_ttl = 1;
    265 	} else {
    266 		if (ro->ro_rt == 0)
    267 			rtalloc(ro);
    268 		if (ro->ro_rt == 0) {
    269 			ipstat.ips_noroute++;
    270 			error = EHOSTUNREACH;
    271 			goto bad;
    272 		}
    273 		ia = ifatoia(ro->ro_rt->rt_ifa);
    274 		ifp = ro->ro_rt->rt_ifp;
    275 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
    276 			mtu = ifp->if_mtu;
    277 		ro->ro_rt->rt_use++;
    278 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    279 			dst = satosin(ro->ro_rt->rt_gateway);
    280 	}
    281 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
    282 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
    283 		struct in_multi *inm;
    284 
    285 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
    286 			M_BCAST : M_MCAST;
    287 		/*
    288 		 * IP destination address is multicast.  Make sure "dst"
    289 		 * still points to the address in "ro".  (It may have been
    290 		 * changed to point to a gateway address, above.)
    291 		 */
    292 		dst = satosin(&ro->ro_dst);
    293 		/*
    294 		 * See if the caller provided any multicast options
    295 		 */
    296 		if (imo != NULL) {
    297 			ip->ip_ttl = imo->imo_multicast_ttl;
    298 			if (imo->imo_multicast_ifp != NULL) {
    299 				ifp = imo->imo_multicast_ifp;
    300 				mtu = ifp->if_mtu;
    301 			}
    302 		} else
    303 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
    304 		/*
    305 		 * Confirm that the outgoing interface supports multicast.
    306 		 */
    307 		if (((m->m_flags & M_MCAST) &&
    308 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
    309 		    ((m->m_flags & M_BCAST) &&
    310 		     (ifp->if_flags & IFF_BROADCAST) == 0))  {
    311 			ipstat.ips_noroute++;
    312 			error = ENETUNREACH;
    313 			goto bad;
    314 		}
    315 		/*
    316 		 * If source address not specified yet, use an address
    317 		 * of outgoing interface.
    318 		 */
    319 		if (in_nullhost(ip->ip_src)) {
    320 			struct in_ifaddr *ia;
    321 
    322 			IFP_TO_IA(ifp, ia);
    323 			ip->ip_src = ia->ia_addr.sin_addr;
    324 		}
    325 
    326 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
    327 		if (inm != NULL &&
    328 		   (imo == NULL || imo->imo_multicast_loop)) {
    329 			/*
    330 			 * If we belong to the destination multicast group
    331 			 * on the outgoing interface, and the caller did not
    332 			 * forbid loopback, loop back a copy.
    333 			 */
    334 			ip_mloopback(ifp, m, dst);
    335 		}
    336 #ifdef MROUTING
    337 		else {
    338 			/*
    339 			 * If we are acting as a multicast router, perform
    340 			 * multicast forwarding as if the packet had just
    341 			 * arrived on the interface to which we are about
    342 			 * to send.  The multicast forwarding function
    343 			 * recursively calls this function, using the
    344 			 * IP_FORWARDING flag to prevent infinite recursion.
    345 			 *
    346 			 * Multicasts that are looped back by ip_mloopback(),
    347 			 * above, will be forwarded by the ip_input() routine,
    348 			 * if necessary.
    349 			 */
    350 			extern struct socket *ip_mrouter;
    351 
    352 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
    353 				if (ip_mforward(m, ifp) != 0) {
    354 					m_freem(m);
    355 					goto done;
    356 				}
    357 			}
    358 		}
    359 #endif
    360 		/*
    361 		 * Multicasts with a time-to-live of zero may be looped-
    362 		 * back, above, but must not be transmitted on a network.
    363 		 * Also, multicasts addressed to the loopback interface
    364 		 * are not sent -- the above call to ip_mloopback() will
    365 		 * loop back a copy if this host actually belongs to the
    366 		 * destination group on the loopback interface.
    367 		 */
    368 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
    369 			m_freem(m);
    370 			goto done;
    371 		}
    372 
    373 		goto sendit;
    374 	}
    375 #ifndef notdef
    376 	/*
    377 	 * If source address not specified yet, use address
    378 	 * of outgoing interface.
    379 	 */
    380 	if (in_nullhost(ip->ip_src))
    381 		ip->ip_src = ia->ia_addr.sin_addr;
    382 #endif
    383 
    384 	/*
    385 	 * packets with Class-D address as source are not valid per
    386 	 * RFC 1112
    387 	 */
    388 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    389 		ipstat.ips_odropped++;
    390 		error = EADDRNOTAVAIL;
    391 		goto bad;
    392 	}
    393 
    394 	/*
    395 	 * Look for broadcast address and
    396 	 * and verify user is allowed to send
    397 	 * such a packet.
    398 	 */
    399 	if (in_broadcast(dst->sin_addr, ifp)) {
    400 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
    401 			error = EADDRNOTAVAIL;
    402 			goto bad;
    403 		}
    404 		if ((flags & IP_ALLOWBROADCAST) == 0) {
    405 			error = EACCES;
    406 			goto bad;
    407 		}
    408 		/* don't allow broadcast messages to be fragmented */
    409 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
    410 			error = EMSGSIZE;
    411 			goto bad;
    412 		}
    413 		m->m_flags |= M_BCAST;
    414 	} else
    415 		m->m_flags &= ~M_BCAST;
    416 
    417 sendit:
    418 #ifdef PFIL_HOOKS
    419 	/*
    420 	 * Run through list of hooks for output packets.
    421 	 */
    422 	m1 = m;
    423 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
    424 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
    425 		if (pfh->pfil_func) {
    426 		    	rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
    427 			if (rv) {
    428 				error = EHOSTUNREACH;
    429 				goto done;
    430 			}
    431 			m = m1;
    432 			if (m == NULL)
    433 				goto done;
    434 			ip = mtod(m, struct ip *);
    435 		}
    436 #endif /* PFIL_HOOKS */
    437 
    438 #ifdef IPSEC
    439 	/* get SP for this packet */
    440 	if (so == NULL)
    441 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
    442 	else
    443 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    444 
    445 	if (sp == NULL) {
    446 		ipsecstat.out_inval++;
    447 		goto bad;
    448 	}
    449 
    450 	error = 0;
    451 
    452 	/* check policy */
    453 	switch (sp->policy) {
    454 	case IPSEC_POLICY_DISCARD:
    455 		/*
    456 		 * This packet is just discarded.
    457 		 */
    458 		ipsecstat.out_polvio++;
    459 		goto bad;
    460 
    461 	case IPSEC_POLICY_BYPASS:
    462 	case IPSEC_POLICY_NONE:
    463 		/* no need to do IPsec. */
    464 		goto skip_ipsec;
    465 
    466 	case IPSEC_POLICY_IPSEC:
    467 		if (sp->req == NULL) {
    468 			/* XXX should be panic ? */
    469 			printf("ip_output: No IPsec request specified.\n");
    470 			error = EINVAL;
    471 			goto bad;
    472 		}
    473 		break;
    474 
    475 	case IPSEC_POLICY_ENTRUST:
    476 	default:
    477 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
    478 	}
    479 
    480 	ip->ip_len = htons((u_short)ip->ip_len);
    481 	ip->ip_off = htons((u_short)ip->ip_off);
    482 	ip->ip_sum = 0;
    483 
    484     {
    485 	struct ipsec_output_state state;
    486 	bzero(&state, sizeof(state));
    487 	state.m = m;
    488 	if (flags & IP_ROUTETOIF) {
    489 		state.ro = &iproute;
    490 		bzero(&iproute, sizeof(iproute));
    491 	} else
    492 		state.ro = ro;
    493 	state.dst = (struct sockaddr *)dst;
    494 
    495 	error = ipsec4_output(&state, sp, flags);
    496 
    497 	m = state.m;
    498 	if (flags & IP_ROUTETOIF) {
    499 		/*
    500 		 * if we have tunnel mode SA, we may need to ignore
    501 		 * IP_ROUTETOIF.
    502 		 */
    503 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
    504 			flags &= ~IP_ROUTETOIF;
    505 			ro = state.ro;
    506 		}
    507 	} else
    508 		ro = state.ro;
    509 	dst = (struct sockaddr_in *)state.dst;
    510 	if (error) {
    511 		/* mbuf is already reclaimed in ipsec4_output. */
    512 		m0 = NULL;
    513 		switch (error) {
    514 		case EHOSTUNREACH:
    515 		case ENETUNREACH:
    516 		case EMSGSIZE:
    517 		case ENOBUFS:
    518 		case ENOMEM:
    519 			break;
    520 		default:
    521 			printf("ip4_output (ipsec): error code %d\n", error);
    522 			/*fall through*/
    523 		case ENOENT:
    524 			/* don't show these error codes to the user */
    525 			error = 0;
    526 			break;
    527 		}
    528 		goto bad;
    529 	}
    530     }
    531 
    532 	/* be sure to update variables that are affected by ipsec4_output() */
    533 	ip = mtod(m, struct ip *);
    534 #ifdef _IP_VHL
    535 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
    536 #else
    537 	hlen = ip->ip_hl << 2;
    538 #endif
    539 	if (ro->ro_rt == NULL) {
    540 		if ((flags & IP_ROUTETOIF) == 0) {
    541 			printf("ip_output: "
    542 				"can't update route after IPsec processing\n");
    543 			error = EHOSTUNREACH;	/*XXX*/
    544 			goto bad;
    545 		}
    546 	} else {
    547 		/* nobody uses ia beyond here */
    548 		ifp = ro->ro_rt->rt_ifp;
    549 	}
    550 
    551 	/* make it flipped, again. */
    552 	ip->ip_len = ntohs((u_short)ip->ip_len);
    553 	ip->ip_off = ntohs((u_short)ip->ip_off);
    554 skip_ipsec:
    555 #endif /*IPSEC*/
    556 
    557 	/*
    558 	 * If small enough for mtu of path, can just send directly.
    559 	 */
    560 	if ((u_int16_t)ip->ip_len <= mtu) {
    561 #if IFA_STATS
    562 		/*
    563 		 * search for the source address structure to
    564 		 * maintain output statistics.
    565 		 */
    566 		bzero((caddr_t*) &src, sizeof(src));
    567 		src.sin_family = AF_INET;
    568 		src.sin_addr.s_addr = ip->ip_src.s_addr;
    569 		src.sin_len = sizeof(src);
    570 		ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
    571 		if (ia)
    572 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
    573 #endif
    574 		HTONS(ip->ip_len);
    575 		HTONS(ip->ip_off);
    576 		ip->ip_sum = 0;
    577 		ip->ip_sum = in_cksum(m, hlen);
    578 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
    579 		goto done;
    580 	}
    581 
    582 	/*
    583 	 * Too large for interface; fragment if possible.
    584 	 * Must be able to put at least 8 bytes per fragment.
    585 	 */
    586 #if 0
    587 	/*
    588 	 * If IPsec packet is too big for the interface, try fragment it.
    589 	 * XXX This really is a quickhack.  May be inappropriate.
    590 	 * XXX fails if somebody is sending AH'ed packet, with:
    591 	 *	sizeof(packet without AH) < mtu < sizeof(packet with AH)
    592 	 */
    593 	if (sab && ip->ip_p != IPPROTO_AH && (flags & IP_FORWARDING) == 0)
    594 		ip->ip_off &= ~IP_DF;
    595 #endif /*IPSEC*/
    596 	if (ip->ip_off & IP_DF) {
    597 		if (flags & IP_RETURNMTU)
    598 			*mtu_p = mtu;
    599 		error = EMSGSIZE;
    600 		ipstat.ips_cantfrag++;
    601 		goto bad;
    602 	}
    603 	len = (mtu - hlen) &~ 7;
    604 	if (len < 8) {
    605 		error = EMSGSIZE;
    606 		goto bad;
    607 	}
    608 
    609     {
    610 	int mhlen, firstlen = len;
    611 	struct mbuf **mnext = &m->m_nextpkt;
    612 	int fragments = 0;
    613 	int s;
    614 
    615 	/*
    616 	 * Loop through length of segment after first fragment,
    617 	 * make new header and copy data of each part and link onto chain.
    618 	 */
    619 	m0 = m;
    620 	mhlen = sizeof (struct ip);
    621 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
    622 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
    623 		if (m == 0) {
    624 			error = ENOBUFS;
    625 			ipstat.ips_odropped++;
    626 			goto sendorfree;
    627 		}
    628 		*mnext = m;
    629 		mnext = &m->m_nextpkt;
    630 		m->m_data += max_linkhdr;
    631 		mhip = mtod(m, struct ip *);
    632 		*mhip = *ip;
    633 		/* we must inherit MCAST and BCAST flags */
    634 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
    635 		if (hlen > sizeof (struct ip)) {
    636 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
    637 			mhip->ip_hl = mhlen >> 2;
    638 		}
    639 		m->m_len = mhlen;
    640 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
    641 		if (ip->ip_off & IP_MF)
    642 			mhip->ip_off |= IP_MF;
    643 		if (off + len >= (u_int16_t)ip->ip_len)
    644 			len = (u_int16_t)ip->ip_len - off;
    645 		else
    646 			mhip->ip_off |= IP_MF;
    647 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
    648 		m->m_next = m_copy(m0, off, len);
    649 		if (m->m_next == 0) {
    650 			error = ENOBUFS;	/* ??? */
    651 			ipstat.ips_odropped++;
    652 			goto sendorfree;
    653 		}
    654 		m->m_pkthdr.len = mhlen + len;
    655 		m->m_pkthdr.rcvif = (struct ifnet *)0;
    656 		HTONS(mhip->ip_off);
    657 		mhip->ip_sum = 0;
    658 		mhip->ip_sum = in_cksum(m, mhlen);
    659 		ipstat.ips_ofragments++;
    660 		fragments++;
    661 	}
    662 	/*
    663 	 * Update first fragment by trimming what's been copied out
    664 	 * and updating header, then send each fragment (in order).
    665 	 */
    666 	m = m0;
    667 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
    668 	m->m_pkthdr.len = hlen + firstlen;
    669 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
    670 	ip->ip_off |= IP_MF;
    671 	HTONS(ip->ip_off);
    672 	ip->ip_sum = 0;
    673 	ip->ip_sum = in_cksum(m, hlen);
    674 sendorfree:
    675 	/*
    676 	 * If there is no room for all the fragments, don't queue
    677 	 * any of them.
    678 	 */
    679 	s = splimp();
    680 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
    681 		error = ENOBUFS;
    682 	splx(s);
    683 	for (m = m0; m; m = m0) {
    684 		m0 = m->m_nextpkt;
    685 		m->m_nextpkt = 0;
    686 		if (error == 0) {
    687 #if IFA_STATS
    688 			/*
    689 			 * search for the source address structure to
    690 			 * maintain output statistics.
    691 			 */
    692 			bzero((caddr_t*) &src, sizeof(src));
    693 			src.sin_family = AF_INET;
    694 			src.sin_addr.s_addr = ip->ip_src.s_addr;
    695 			src.sin_len = sizeof(src);
    696 			ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
    697 			if (ia) {
    698 				ia->ia_ifa.ifa_data.ifad_outbytes +=
    699 					ntohs(ip->ip_len);
    700 			}
    701 #endif
    702 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
    703 			    ro->ro_rt);
    704 		} else
    705 			m_freem(m);
    706 	}
    707 
    708 	if (error == 0)
    709 		ipstat.ips_fragmented++;
    710     }
    711 done:
    712 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
    713 		RTFREE(ro->ro_rt);
    714 		ro->ro_rt = 0;
    715 	}
    716 
    717 #ifdef IPSEC
    718 	if (sp != NULL) {
    719 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
    720 			printf("DP ip_output call free SP:%p\n", sp));
    721 		key_freesp(sp);
    722 	}
    723 #endif /* IPSEC */
    724 
    725 	return (error);
    726 bad:
    727 	m_freem(m);
    728 	goto done;
    729 }
    730 
    731 /*
    732  * Determine the maximum length of the options to be inserted;
    733  * we would far rather allocate too much space rather than too little.
    734  */
    735 
    736 u_int
    737 ip_optlen(inp)
    738 	struct inpcb *inp;
    739 {
    740 	struct mbuf *m = inp->inp_options;
    741 
    742 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
    743 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
    744 	else
    745 		return 0;
    746 }
    747 
    748 
    749 /*
    750  * Insert IP options into preformed packet.
    751  * Adjust IP destination as required for IP source routing,
    752  * as indicated by a non-zero in_addr at the start of the options.
    753  */
    754 static struct mbuf *
    755 ip_insertoptions(m, opt, phlen)
    756 	struct mbuf *m;
    757 	struct mbuf *opt;
    758 	int *phlen;
    759 {
    760 	struct ipoption *p = mtod(opt, struct ipoption *);
    761 	struct mbuf *n;
    762 	struct ip *ip = mtod(m, struct ip *);
    763 	unsigned optlen;
    764 
    765 	optlen = opt->m_len - sizeof(p->ipopt_dst);
    766 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
    767 		return (m);		/* XXX should fail */
    768 	if (!in_nullhost(p->ipopt_dst))
    769 		ip->ip_dst = p->ipopt_dst;
    770 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
    771 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
    772 		if (n == 0)
    773 			return (m);
    774 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
    775 		m->m_len -= sizeof(struct ip);
    776 		m->m_data += sizeof(struct ip);
    777 		n->m_next = m;
    778 		m = n;
    779 		m->m_len = optlen + sizeof(struct ip);
    780 		m->m_data += max_linkhdr;
    781 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
    782 	} else {
    783 		m->m_data -= optlen;
    784 		m->m_len += optlen;
    785 		m->m_pkthdr.len += optlen;
    786 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
    787 	}
    788 	ip = mtod(m, struct ip *);
    789 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
    790 	*phlen = sizeof(struct ip) + optlen;
    791 	ip->ip_len += optlen;
    792 	return (m);
    793 }
    794 
    795 /*
    796  * Copy options from ip to jp,
    797  * omitting those not copied during fragmentation.
    798  */
    799 int
    800 ip_optcopy(ip, jp)
    801 	struct ip *ip, *jp;
    802 {
    803 	u_char *cp, *dp;
    804 	int opt, optlen, cnt;
    805 
    806 	cp = (u_char *)(ip + 1);
    807 	dp = (u_char *)(jp + 1);
    808 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
    809 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
    810 		opt = cp[0];
    811 		if (opt == IPOPT_EOL)
    812 			break;
    813 		if (opt == IPOPT_NOP) {
    814 			/* Preserve for IP mcast tunnel's LSRR alignment. */
    815 			*dp++ = IPOPT_NOP;
    816 			optlen = 1;
    817 			continue;
    818 		}
    819 #ifdef DIAGNOSTIC
    820 		if (cnt < IPOPT_OLEN + sizeof(*cp))
    821 			panic("malformed IPv4 option passed to ip_optcopy");
    822 #endif
    823 		optlen = cp[IPOPT_OLEN];
    824 #ifdef DIAGNOSTIC
    825 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
    826 			panic("malformed IPv4 option passed to ip_optcopy");
    827 #endif
    828 		/* bogus lengths should have been caught by ip_dooptions */
    829 		if (optlen > cnt)
    830 			optlen = cnt;
    831 		if (IPOPT_COPIED(opt)) {
    832 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
    833 			dp += optlen;
    834 		}
    835 	}
    836 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
    837 		*dp++ = IPOPT_EOL;
    838 	return (optlen);
    839 }
    840 
    841 /*
    842  * IP socket option processing.
    843  */
    844 int
    845 ip_ctloutput(op, so, level, optname, mp)
    846 	int op;
    847 	struct socket *so;
    848 	int level, optname;
    849 	struct mbuf **mp;
    850 {
    851 	struct inpcb *inp = sotoinpcb(so);
    852 	struct mbuf *m = *mp;
    853 	int optval = 0;
    854 	int error = 0;
    855 #ifdef IPSEC
    856 #ifdef __NetBSD__
    857 	struct proc *p = curproc;	/*XXX*/
    858 #endif
    859 #endif
    860 
    861 	if (level != IPPROTO_IP) {
    862 		error = EINVAL;
    863 		if (op == PRCO_SETOPT && *mp)
    864 			(void) m_free(*mp);
    865 	} else switch (op) {
    866 
    867 	case PRCO_SETOPT:
    868 		switch (optname) {
    869 		case IP_OPTIONS:
    870 #ifdef notyet
    871 		case IP_RETOPTS:
    872 			return (ip_pcbopts(optname, &inp->inp_options, m));
    873 #else
    874 			return (ip_pcbopts(&inp->inp_options, m));
    875 #endif
    876 
    877 		case IP_TOS:
    878 		case IP_TTL:
    879 		case IP_RECVOPTS:
    880 		case IP_RECVRETOPTS:
    881 		case IP_RECVDSTADDR:
    882 		case IP_RECVIF:
    883 			if (m == NULL || m->m_len != sizeof(int))
    884 				error = EINVAL;
    885 			else {
    886 				optval = *mtod(m, int *);
    887 				switch (optname) {
    888 
    889 				case IP_TOS:
    890 					inp->inp_ip.ip_tos = optval;
    891 					break;
    892 
    893 				case IP_TTL:
    894 					inp->inp_ip.ip_ttl = optval;
    895 					break;
    896 #define	OPTSET(bit) \
    897 	if (optval) \
    898 		inp->inp_flags |= bit; \
    899 	else \
    900 		inp->inp_flags &= ~bit;
    901 
    902 				case IP_RECVOPTS:
    903 					OPTSET(INP_RECVOPTS);
    904 					break;
    905 
    906 				case IP_RECVRETOPTS:
    907 					OPTSET(INP_RECVRETOPTS);
    908 					break;
    909 
    910 				case IP_RECVDSTADDR:
    911 					OPTSET(INP_RECVDSTADDR);
    912 					break;
    913 
    914 				case IP_RECVIF:
    915 					OPTSET(INP_RECVIF);
    916 					break;
    917 				}
    918 			}
    919 			break;
    920 #undef OPTSET
    921 
    922 		case IP_MULTICAST_IF:
    923 		case IP_MULTICAST_TTL:
    924 		case IP_MULTICAST_LOOP:
    925 		case IP_ADD_MEMBERSHIP:
    926 		case IP_DROP_MEMBERSHIP:
    927 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
    928 			break;
    929 
    930 		case IP_PORTRANGE:
    931 			if (m == 0 || m->m_len != sizeof(int))
    932 				error = EINVAL;
    933 			else {
    934 				optval = *mtod(m, int *);
    935 
    936 				switch (optval) {
    937 
    938 				case IP_PORTRANGE_DEFAULT:
    939 				case IP_PORTRANGE_HIGH:
    940 					inp->inp_flags &= ~(INP_LOWPORT);
    941 					break;
    942 
    943 				case IP_PORTRANGE_LOW:
    944 					inp->inp_flags |= INP_LOWPORT;
    945 					break;
    946 
    947 				default:
    948 					error = EINVAL;
    949 					break;
    950 				}
    951 			}
    952 			break;
    953 
    954 #ifdef IPSEC
    955 		case IP_IPSEC_POLICY:
    956 		{
    957 			caddr_t req = NULL;
    958 			size_t len = 0;
    959 			int priv = 0;
    960 
    961 #ifdef __NetBSD__
    962 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
    963 				priv = 0;
    964 			else
    965 				priv = 1;
    966 #else
    967 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
    968 #endif
    969 			if (m) {
    970 				req = mtod(m, caddr_t);
    971 				len = m->m_len;
    972 			}
    973 			error = ipsec4_set_policy(inp, optname, req, len, priv);
    974 			break;
    975 		    }
    976 #endif /*IPSEC*/
    977 
    978 		default:
    979 			error = ENOPROTOOPT;
    980 			break;
    981 		}
    982 		if (m)
    983 			(void)m_free(m);
    984 		break;
    985 
    986 	case PRCO_GETOPT:
    987 		switch (optname) {
    988 		case IP_OPTIONS:
    989 		case IP_RETOPTS:
    990 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
    991 			if (inp->inp_options) {
    992 				m->m_len = inp->inp_options->m_len;
    993 				bcopy(mtod(inp->inp_options, caddr_t),
    994 				    mtod(m, caddr_t), (unsigned)m->m_len);
    995 			} else
    996 				m->m_len = 0;
    997 			break;
    998 
    999 		case IP_TOS:
   1000 		case IP_TTL:
   1001 		case IP_RECVOPTS:
   1002 		case IP_RECVRETOPTS:
   1003 		case IP_RECVDSTADDR:
   1004 		case IP_RECVIF:
   1005 		case IP_ERRORMTU:
   1006 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1007 			m->m_len = sizeof(int);
   1008 			switch (optname) {
   1009 
   1010 			case IP_TOS:
   1011 				optval = inp->inp_ip.ip_tos;
   1012 				break;
   1013 
   1014 			case IP_TTL:
   1015 				optval = inp->inp_ip.ip_ttl;
   1016 				break;
   1017 
   1018 			case IP_ERRORMTU:
   1019 				optval = inp->inp_errormtu;
   1020 				break;
   1021 
   1022 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
   1023 
   1024 			case IP_RECVOPTS:
   1025 				optval = OPTBIT(INP_RECVOPTS);
   1026 				break;
   1027 
   1028 			case IP_RECVRETOPTS:
   1029 				optval = OPTBIT(INP_RECVRETOPTS);
   1030 				break;
   1031 
   1032 			case IP_RECVDSTADDR:
   1033 				optval = OPTBIT(INP_RECVDSTADDR);
   1034 				break;
   1035 
   1036 			case IP_RECVIF:
   1037 				optval = OPTBIT(INP_RECVIF);
   1038 				break;
   1039 			}
   1040 			*mtod(m, int *) = optval;
   1041 			break;
   1042 
   1043 #ifdef IPSEC
   1044 		case IP_IPSEC_POLICY:
   1045 		{
   1046 			caddr_t req = NULL;
   1047 			size_t len;
   1048 
   1049 			if (m) {
   1050 				req = mtod(m, caddr_t);
   1051 				len = m->m_len;
   1052 			}
   1053 			error = ipsec4_get_policy(inp, req, len, mp);
   1054 			break;
   1055 		}
   1056 #endif /*IPSEC*/
   1057 
   1058 		case IP_MULTICAST_IF:
   1059 		case IP_MULTICAST_TTL:
   1060 		case IP_MULTICAST_LOOP:
   1061 		case IP_ADD_MEMBERSHIP:
   1062 		case IP_DROP_MEMBERSHIP:
   1063 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
   1064 			break;
   1065 
   1066 		case IP_PORTRANGE:
   1067 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1068 			m->m_len = sizeof(int);
   1069 
   1070 			if (inp->inp_flags & INP_LOWPORT)
   1071 				optval = IP_PORTRANGE_LOW;
   1072 			else
   1073 				optval = IP_PORTRANGE_DEFAULT;
   1074 
   1075 			*mtod(m, int *) = optval;
   1076 			break;
   1077 
   1078 		default:
   1079 			error = ENOPROTOOPT;
   1080 			break;
   1081 		}
   1082 		break;
   1083 	}
   1084 	return (error);
   1085 }
   1086 
   1087 /*
   1088  * Set up IP options in pcb for insertion in output packets.
   1089  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1090  * with destination address if source routed.
   1091  */
   1092 int
   1093 #ifdef notyet
   1094 ip_pcbopts(optname, pcbopt, m)
   1095 	int optname;
   1096 #else
   1097 ip_pcbopts(pcbopt, m)
   1098 #endif
   1099 	struct mbuf **pcbopt;
   1100 	struct mbuf *m;
   1101 {
   1102 	int cnt, optlen;
   1103 	u_char *cp;
   1104 	u_char opt;
   1105 
   1106 	/* turn off any old options */
   1107 	if (*pcbopt)
   1108 		(void)m_free(*pcbopt);
   1109 	*pcbopt = 0;
   1110 	if (m == (struct mbuf *)0 || m->m_len == 0) {
   1111 		/*
   1112 		 * Only turning off any previous options.
   1113 		 */
   1114 		if (m)
   1115 			(void)m_free(m);
   1116 		return (0);
   1117 	}
   1118 
   1119 #ifndef	vax
   1120 	if (m->m_len % sizeof(int32_t))
   1121 		goto bad;
   1122 #endif
   1123 	/*
   1124 	 * IP first-hop destination address will be stored before
   1125 	 * actual options; move other options back
   1126 	 * and clear it when none present.
   1127 	 */
   1128 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
   1129 		goto bad;
   1130 	cnt = m->m_len;
   1131 	m->m_len += sizeof(struct in_addr);
   1132 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
   1133 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
   1134 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
   1135 
   1136 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
   1137 		opt = cp[IPOPT_OPTVAL];
   1138 		if (opt == IPOPT_EOL)
   1139 			break;
   1140 		if (opt == IPOPT_NOP)
   1141 			optlen = 1;
   1142 		else {
   1143 			if (cnt < IPOPT_OLEN + sizeof(*cp))
   1144 				goto bad;
   1145 			optlen = cp[IPOPT_OLEN];
   1146 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
   1147 				goto bad;
   1148 		}
   1149 		switch (opt) {
   1150 
   1151 		default:
   1152 			break;
   1153 
   1154 		case IPOPT_LSRR:
   1155 		case IPOPT_SSRR:
   1156 			/*
   1157 			 * user process specifies route as:
   1158 			 *	->A->B->C->D
   1159 			 * D must be our final destination (but we can't
   1160 			 * check that since we may not have connected yet).
   1161 			 * A is first hop destination, which doesn't appear in
   1162 			 * actual IP option, but is stored before the options.
   1163 			 */
   1164 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
   1165 				goto bad;
   1166 			m->m_len -= sizeof(struct in_addr);
   1167 			cnt -= sizeof(struct in_addr);
   1168 			optlen -= sizeof(struct in_addr);
   1169 			cp[IPOPT_OLEN] = optlen;
   1170 			/*
   1171 			 * Move first hop before start of options.
   1172 			 */
   1173 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
   1174 			    sizeof(struct in_addr));
   1175 			/*
   1176 			 * Then copy rest of options back
   1177 			 * to close up the deleted entry.
   1178 			 */
   1179 			memmove(&cp[IPOPT_OFFSET+1],
   1180                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
   1181 			    (unsigned)cnt + sizeof(struct in_addr));
   1182 			break;
   1183 		}
   1184 	}
   1185 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
   1186 		goto bad;
   1187 	*pcbopt = m;
   1188 	return (0);
   1189 
   1190 bad:
   1191 	(void)m_free(m);
   1192 	return (EINVAL);
   1193 }
   1194 
   1195 /*
   1196  * Set the IP multicast options in response to user setsockopt().
   1197  */
   1198 int
   1199 ip_setmoptions(optname, imop, m)
   1200 	int optname;
   1201 	struct ip_moptions **imop;
   1202 	struct mbuf *m;
   1203 {
   1204 	int error = 0;
   1205 	u_char loop;
   1206 	int i;
   1207 	struct in_addr addr;
   1208 	struct ip_mreq *mreq;
   1209 	struct ifnet *ifp;
   1210 	struct ip_moptions *imo = *imop;
   1211 	struct route ro;
   1212 	struct sockaddr_in *dst;
   1213 
   1214 	if (imo == NULL) {
   1215 		/*
   1216 		 * No multicast option buffer attached to the pcb;
   1217 		 * allocate one and initialize to default values.
   1218 		 */
   1219 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
   1220 		    M_WAITOK);
   1221 
   1222 		if (imo == NULL)
   1223 			return (ENOBUFS);
   1224 		*imop = imo;
   1225 		imo->imo_multicast_ifp = NULL;
   1226 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
   1227 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
   1228 		imo->imo_num_memberships = 0;
   1229 	}
   1230 
   1231 	switch (optname) {
   1232 
   1233 	case IP_MULTICAST_IF:
   1234 		/*
   1235 		 * Select the interface for outgoing multicast packets.
   1236 		 */
   1237 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
   1238 			error = EINVAL;
   1239 			break;
   1240 		}
   1241 		addr = *(mtod(m, struct in_addr *));
   1242 		/*
   1243 		 * INADDR_ANY is used to remove a previous selection.
   1244 		 * When no interface is selected, a default one is
   1245 		 * chosen every time a multicast packet is sent.
   1246 		 */
   1247 		if (in_nullhost(addr)) {
   1248 			imo->imo_multicast_ifp = NULL;
   1249 			break;
   1250 		}
   1251 		/*
   1252 		 * The selected interface is identified by its local
   1253 		 * IP address.  Find the interface and confirm that
   1254 		 * it supports multicasting.
   1255 		 */
   1256 		INADDR_TO_IFP(addr, ifp);
   1257 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1258 			error = EADDRNOTAVAIL;
   1259 			break;
   1260 		}
   1261 		imo->imo_multicast_ifp = ifp;
   1262 		break;
   1263 
   1264 	case IP_MULTICAST_TTL:
   1265 		/*
   1266 		 * Set the IP time-to-live for outgoing multicast packets.
   1267 		 */
   1268 		if (m == NULL || m->m_len != 1) {
   1269 			error = EINVAL;
   1270 			break;
   1271 		}
   1272 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
   1273 		break;
   1274 
   1275 	case IP_MULTICAST_LOOP:
   1276 		/*
   1277 		 * Set the loopback flag for outgoing multicast packets.
   1278 		 * Must be zero or one.
   1279 		 */
   1280 		if (m == NULL || m->m_len != 1 ||
   1281 		   (loop = *(mtod(m, u_char *))) > 1) {
   1282 			error = EINVAL;
   1283 			break;
   1284 		}
   1285 		imo->imo_multicast_loop = loop;
   1286 		break;
   1287 
   1288 	case IP_ADD_MEMBERSHIP:
   1289 		/*
   1290 		 * Add a multicast group membership.
   1291 		 * Group must be a valid IP multicast address.
   1292 		 */
   1293 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
   1294 			error = EINVAL;
   1295 			break;
   1296 		}
   1297 		mreq = mtod(m, struct ip_mreq *);
   1298 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
   1299 			error = EINVAL;
   1300 			break;
   1301 		}
   1302 		/*
   1303 		 * If no interface address was provided, use the interface of
   1304 		 * the route to the given multicast address.
   1305 		 */
   1306 		if (in_nullhost(mreq->imr_interface)) {
   1307 			bzero((caddr_t)&ro, sizeof(ro));
   1308 			ro.ro_rt = NULL;
   1309 			dst = satosin(&ro.ro_dst);
   1310 			dst->sin_len = sizeof(*dst);
   1311 			dst->sin_family = AF_INET;
   1312 			dst->sin_addr = mreq->imr_multiaddr;
   1313 			rtalloc(&ro);
   1314 			if (ro.ro_rt == NULL) {
   1315 				error = EADDRNOTAVAIL;
   1316 				break;
   1317 			}
   1318 			ifp = ro.ro_rt->rt_ifp;
   1319 			rtfree(ro.ro_rt);
   1320 		} else {
   1321 			INADDR_TO_IFP(mreq->imr_interface, ifp);
   1322 		}
   1323 		/*
   1324 		 * See if we found an interface, and confirm that it
   1325 		 * supports multicast.
   1326 		 */
   1327 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1328 			error = EADDRNOTAVAIL;
   1329 			break;
   1330 		}
   1331 		/*
   1332 		 * See if the membership already exists or if all the
   1333 		 * membership slots are full.
   1334 		 */
   1335 		for (i = 0; i < imo->imo_num_memberships; ++i) {
   1336 			if (imo->imo_membership[i]->inm_ifp == ifp &&
   1337 			    in_hosteq(imo->imo_membership[i]->inm_addr,
   1338 				      mreq->imr_multiaddr))
   1339 				break;
   1340 		}
   1341 		if (i < imo->imo_num_memberships) {
   1342 			error = EADDRINUSE;
   1343 			break;
   1344 		}
   1345 		if (i == IP_MAX_MEMBERSHIPS) {
   1346 			error = ETOOMANYREFS;
   1347 			break;
   1348 		}
   1349 		/*
   1350 		 * Everything looks good; add a new record to the multicast
   1351 		 * address list for the given interface.
   1352 		 */
   1353 		if ((imo->imo_membership[i] =
   1354 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
   1355 			error = ENOBUFS;
   1356 			break;
   1357 		}
   1358 		++imo->imo_num_memberships;
   1359 		break;
   1360 
   1361 	case IP_DROP_MEMBERSHIP:
   1362 		/*
   1363 		 * Drop a multicast group membership.
   1364 		 * Group must be a valid IP multicast address.
   1365 		 */
   1366 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
   1367 			error = EINVAL;
   1368 			break;
   1369 		}
   1370 		mreq = mtod(m, struct ip_mreq *);
   1371 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
   1372 			error = EINVAL;
   1373 			break;
   1374 		}
   1375 		/*
   1376 		 * If an interface address was specified, get a pointer
   1377 		 * to its ifnet structure.
   1378 		 */
   1379 		if (in_nullhost(mreq->imr_interface))
   1380 			ifp = NULL;
   1381 		else {
   1382 			INADDR_TO_IFP(mreq->imr_interface, ifp);
   1383 			if (ifp == NULL) {
   1384 				error = EADDRNOTAVAIL;
   1385 				break;
   1386 			}
   1387 		}
   1388 		/*
   1389 		 * Find the membership in the membership array.
   1390 		 */
   1391 		for (i = 0; i < imo->imo_num_memberships; ++i) {
   1392 			if ((ifp == NULL ||
   1393 			     imo->imo_membership[i]->inm_ifp == ifp) &&
   1394 			     in_hosteq(imo->imo_membership[i]->inm_addr,
   1395 				       mreq->imr_multiaddr))
   1396 				break;
   1397 		}
   1398 		if (i == imo->imo_num_memberships) {
   1399 			error = EADDRNOTAVAIL;
   1400 			break;
   1401 		}
   1402 		/*
   1403 		 * Give up the multicast address record to which the
   1404 		 * membership points.
   1405 		 */
   1406 		in_delmulti(imo->imo_membership[i]);
   1407 		/*
   1408 		 * Remove the gap in the membership array.
   1409 		 */
   1410 		for (++i; i < imo->imo_num_memberships; ++i)
   1411 			imo->imo_membership[i-1] = imo->imo_membership[i];
   1412 		--imo->imo_num_memberships;
   1413 		break;
   1414 
   1415 	default:
   1416 		error = EOPNOTSUPP;
   1417 		break;
   1418 	}
   1419 
   1420 	/*
   1421 	 * If all options have default values, no need to keep the mbuf.
   1422 	 */
   1423 	if (imo->imo_multicast_ifp == NULL &&
   1424 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
   1425 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
   1426 	    imo->imo_num_memberships == 0) {
   1427 		free(*imop, M_IPMOPTS);
   1428 		*imop = NULL;
   1429 	}
   1430 
   1431 	return (error);
   1432 }
   1433 
   1434 /*
   1435  * Return the IP multicast options in response to user getsockopt().
   1436  */
   1437 int
   1438 ip_getmoptions(optname, imo, mp)
   1439 	int optname;
   1440 	struct ip_moptions *imo;
   1441 	struct mbuf **mp;
   1442 {
   1443 	u_char *ttl;
   1444 	u_char *loop;
   1445 	struct in_addr *addr;
   1446 	struct in_ifaddr *ia;
   1447 
   1448 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1449 
   1450 	switch (optname) {
   1451 
   1452 	case IP_MULTICAST_IF:
   1453 		addr = mtod(*mp, struct in_addr *);
   1454 		(*mp)->m_len = sizeof(struct in_addr);
   1455 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
   1456 			*addr = zeroin_addr;
   1457 		else {
   1458 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
   1459 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
   1460 		}
   1461 		return (0);
   1462 
   1463 	case IP_MULTICAST_TTL:
   1464 		ttl = mtod(*mp, u_char *);
   1465 		(*mp)->m_len = 1;
   1466 		*ttl = imo ? imo->imo_multicast_ttl
   1467 			   : IP_DEFAULT_MULTICAST_TTL;
   1468 		return (0);
   1469 
   1470 	case IP_MULTICAST_LOOP:
   1471 		loop = mtod(*mp, u_char *);
   1472 		(*mp)->m_len = 1;
   1473 		*loop = imo ? imo->imo_multicast_loop
   1474 			    : IP_DEFAULT_MULTICAST_LOOP;
   1475 		return (0);
   1476 
   1477 	default:
   1478 		return (EOPNOTSUPP);
   1479 	}
   1480 }
   1481 
   1482 /*
   1483  * Discard the IP multicast options.
   1484  */
   1485 void
   1486 ip_freemoptions(imo)
   1487 	struct ip_moptions *imo;
   1488 {
   1489 	int i;
   1490 
   1491 	if (imo != NULL) {
   1492 		for (i = 0; i < imo->imo_num_memberships; ++i)
   1493 			in_delmulti(imo->imo_membership[i]);
   1494 		free(imo, M_IPMOPTS);
   1495 	}
   1496 }
   1497 
   1498 /*
   1499  * Routine called from ip_output() to loop back a copy of an IP multicast
   1500  * packet to the input queue of a specified interface.  Note that this
   1501  * calls the output routine of the loopback "driver", but with an interface
   1502  * pointer that might NOT be &loif -- easier than replicating that code here.
   1503  */
   1504 static void
   1505 ip_mloopback(ifp, m, dst)
   1506 	struct ifnet *ifp;
   1507 	struct mbuf *m;
   1508 	struct sockaddr_in *dst;
   1509 {
   1510 	struct ip *ip;
   1511 	struct mbuf *copym;
   1512 
   1513 	copym = m_copy(m, 0, M_COPYALL);
   1514 	if (copym != NULL
   1515 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
   1516 		copym = m_pullup(copym, sizeof(struct ip));
   1517 	if (copym != NULL) {
   1518 		/*
   1519 		 * We don't bother to fragment if the IP length is greater
   1520 		 * than the interface's MTU.  Can this possibly matter?
   1521 		 */
   1522 		ip = mtod(copym, struct ip *);
   1523 		HTONS(ip->ip_len);
   1524 		HTONS(ip->ip_off);
   1525 		ip->ip_sum = 0;
   1526 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
   1527 		(void) looutput(ifp, copym, sintosa(dst), NULL);
   1528 	}
   1529 }
   1530