xref: /netbsd-src/sys/netinet/ip_output.c (revision 89c5a767f8fc7a4633b2d409966e2becbb98ff92)
1 /*	$NetBSD: ip_output.c,v 1.69 2000/03/01 12:49:35 itojun 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 
117 #include <vm/vm.h>
118 #include <sys/proc.h>
119 
120 #include <net/if.h>
121 #include <net/route.h>
122 #include <net/pfil.h>
123 
124 #include <netinet/in.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/in_pcb.h>
128 #include <netinet/in_var.h>
129 #include <netinet/ip_var.h>
130 
131 #ifdef __vax__
132 #include <machine/mtpr.h>
133 #endif
134 
135 #include <machine/stdarg.h>
136 
137 #ifdef IPSEC
138 #include <netinet6/ipsec.h>
139 #include <netkey/key.h>
140 #include <netkey/key_debug.h>
141 #endif /*IPSEC*/
142 
143 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
144 static void ip_mloopback
145 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
146 
147 /*
148  * IP output.  The packet in mbuf chain m contains a skeletal IP
149  * header (with len, off, ttl, proto, tos, src, dst).
150  * The mbuf chain containing the packet will be freed.
151  * The mbuf opt, if present, will not be freed.
152  */
153 int
154 #if __STDC__
155 ip_output(struct mbuf *m0, ...)
156 #else
157 ip_output(m0, va_alist)
158 	struct mbuf *m0;
159 	va_dcl
160 #endif
161 {
162 	register struct ip *ip, *mhip;
163 	register struct ifnet *ifp;
164 	register struct mbuf *m = m0;
165 	register int hlen = sizeof (struct ip);
166 	int len, off, error = 0;
167 	struct route iproute;
168 	struct sockaddr_in *dst;
169 #if IFA_STATS
170 	struct sockaddr_in src;
171 #endif
172 	struct in_ifaddr *ia;
173 	struct mbuf *opt;
174 	struct route *ro;
175 	int flags;
176 	int *mtu_p;
177 	int mtu;
178 	struct ip_moptions *imo;
179 	va_list ap;
180 #ifdef PFIL_HOOKS
181 	struct packet_filter_hook *pfh;
182 	struct mbuf *m1;
183 	int rv;
184 #endif /* PFIL_HOOKS */
185 #ifdef IPSEC
186 	struct socket *so;
187 	struct secpolicy *sp = NULL;
188 #endif /*IPSEC*/
189 
190 	va_start(ap, m0);
191 	opt = va_arg(ap, struct mbuf *);
192 	ro = va_arg(ap, struct route *);
193 	flags = va_arg(ap, int);
194 	imo = va_arg(ap, struct ip_moptions *);
195 	if (flags & IP_RETURNMTU)
196 		mtu_p = va_arg(ap, int *);
197 	else
198 		mtu_p = NULL;
199 	va_end(ap);
200 
201 #ifdef IPSEC
202 	so = ipsec_getsocket(m);
203 	ipsec_setsocket(m, NULL);
204 #endif /*IPSEC*/
205 
206 #ifdef	DIAGNOSTIC
207 	if ((m->m_flags & M_PKTHDR) == 0)
208 		panic("ip_output no HDR");
209 #endif
210 	if (opt) {
211 		m = ip_insertoptions(m, opt, &len);
212 		hlen = len;
213 	}
214 	ip = mtod(m, struct ip *);
215 	/*
216 	 * Fill in IP header.
217 	 */
218 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
219 		ip->ip_v = IPVERSION;
220 		ip->ip_off &= IP_DF;
221 		ip->ip_id = htons(ip_id++);
222 		ip->ip_hl = hlen >> 2;
223 		ipstat.ips_localout++;
224 	} else {
225 		hlen = ip->ip_hl << 2;
226 	}
227 	/*
228 	 * Route packet.
229 	 */
230 	if (ro == 0) {
231 		ro = &iproute;
232 		bzero((caddr_t)ro, sizeof (*ro));
233 	}
234 	dst = satosin(&ro->ro_dst);
235 	/*
236 	 * If there is a cached route,
237 	 * check that it is to the same destination
238 	 * and is still up.  If not, free it and try again.
239 	 */
240 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
241 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
242 		RTFREE(ro->ro_rt);
243 		ro->ro_rt = (struct rtentry *)0;
244 	}
245 	if (ro->ro_rt == 0) {
246 		dst->sin_family = AF_INET;
247 		dst->sin_len = sizeof(*dst);
248 		dst->sin_addr = ip->ip_dst;
249 	}
250 	/*
251 	 * If routing to interface only,
252 	 * short circuit routing lookup.
253 	 */
254 	if (flags & IP_ROUTETOIF) {
255 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
256 			ipstat.ips_noroute++;
257 			error = ENETUNREACH;
258 			goto bad;
259 		}
260 		ifp = ia->ia_ifp;
261 		mtu = ifp->if_mtu;
262 		ip->ip_ttl = 1;
263 	} else {
264 		if (ro->ro_rt == 0)
265 			rtalloc(ro);
266 		if (ro->ro_rt == 0) {
267 			ipstat.ips_noroute++;
268 			error = EHOSTUNREACH;
269 			goto bad;
270 		}
271 		ia = ifatoia(ro->ro_rt->rt_ifa);
272 		ifp = ro->ro_rt->rt_ifp;
273 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
274 			mtu = ifp->if_mtu;
275 		ro->ro_rt->rt_use++;
276 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
277 			dst = satosin(ro->ro_rt->rt_gateway);
278 	}
279 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
280 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
281 		struct in_multi *inm;
282 
283 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
284 			M_BCAST : M_MCAST;
285 		/*
286 		 * IP destination address is multicast.  Make sure "dst"
287 		 * still points to the address in "ro".  (It may have been
288 		 * changed to point to a gateway address, above.)
289 		 */
290 		dst = satosin(&ro->ro_dst);
291 		/*
292 		 * See if the caller provided any multicast options
293 		 */
294 		if (imo != NULL) {
295 			ip->ip_ttl = imo->imo_multicast_ttl;
296 			if (imo->imo_multicast_ifp != NULL) {
297 				ifp = imo->imo_multicast_ifp;
298 				mtu = ifp->if_mtu;
299 			}
300 		} else
301 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
302 		/*
303 		 * Confirm that the outgoing interface supports multicast.
304 		 */
305 		if (((m->m_flags & M_MCAST) &&
306 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
307 		    ((m->m_flags & M_BCAST) &&
308 		     (ifp->if_flags & IFF_BROADCAST) == 0))  {
309 			ipstat.ips_noroute++;
310 			error = ENETUNREACH;
311 			goto bad;
312 		}
313 		/*
314 		 * If source address not specified yet, use an address
315 		 * of outgoing interface.
316 		 */
317 		if (in_nullhost(ip->ip_src)) {
318 			register struct in_ifaddr *ia;
319 
320 			IFP_TO_IA(ifp, ia);
321 			ip->ip_src = ia->ia_addr.sin_addr;
322 		}
323 
324 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
325 		if (inm != NULL &&
326 		   (imo == NULL || imo->imo_multicast_loop)) {
327 			/*
328 			 * If we belong to the destination multicast group
329 			 * on the outgoing interface, and the caller did not
330 			 * forbid loopback, loop back a copy.
331 			 */
332 			ip_mloopback(ifp, m, dst);
333 		}
334 #ifdef MROUTING
335 		else {
336 			/*
337 			 * If we are acting as a multicast router, perform
338 			 * multicast forwarding as if the packet had just
339 			 * arrived on the interface to which we are about
340 			 * to send.  The multicast forwarding function
341 			 * recursively calls this function, using the
342 			 * IP_FORWARDING flag to prevent infinite recursion.
343 			 *
344 			 * Multicasts that are looped back by ip_mloopback(),
345 			 * above, will be forwarded by the ip_input() routine,
346 			 * if necessary.
347 			 */
348 			extern struct socket *ip_mrouter;
349 
350 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
351 				if (ip_mforward(m, ifp) != 0) {
352 					m_freem(m);
353 					goto done;
354 				}
355 			}
356 		}
357 #endif
358 		/*
359 		 * Multicasts with a time-to-live of zero may be looped-
360 		 * back, above, but must not be transmitted on a network.
361 		 * Also, multicasts addressed to the loopback interface
362 		 * are not sent -- the above call to ip_mloopback() will
363 		 * loop back a copy if this host actually belongs to the
364 		 * destination group on the loopback interface.
365 		 */
366 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
367 			m_freem(m);
368 			goto done;
369 		}
370 
371 		goto sendit;
372 	}
373 #ifndef notdef
374 	/*
375 	 * If source address not specified yet, use address
376 	 * of outgoing interface.
377 	 */
378 	if (in_nullhost(ip->ip_src))
379 		ip->ip_src = ia->ia_addr.sin_addr;
380 #endif
381 
382 	/*
383 	 * packets with Class-D address as source are not valid per
384 	 * RFC 1112
385 	 */
386 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
387 		ipstat.ips_odropped++;
388 		error = EADDRNOTAVAIL;
389 		goto bad;
390 	}
391 
392 	/*
393 	 * Look for broadcast address and
394 	 * and verify user is allowed to send
395 	 * such a packet.
396 	 */
397 	if (in_broadcast(dst->sin_addr, ifp)) {
398 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
399 			error = EADDRNOTAVAIL;
400 			goto bad;
401 		}
402 		if ((flags & IP_ALLOWBROADCAST) == 0) {
403 			error = EACCES;
404 			goto bad;
405 		}
406 		/* don't allow broadcast messages to be fragmented */
407 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
408 			error = EMSGSIZE;
409 			goto bad;
410 		}
411 		m->m_flags |= M_BCAST;
412 	} else
413 		m->m_flags &= ~M_BCAST;
414 
415 sendit:
416 #ifdef PFIL_HOOKS
417 	/*
418 	 * Run through list of hooks for output packets.
419 	 */
420 	m1 = m;
421 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
422 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
423 		if (pfh->pfil_func) {
424 		    	rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
425 			if (rv) {
426 				error = EHOSTUNREACH;
427 				goto done;
428 			}
429 			m = m1;
430 			if (m == NULL)
431 				goto done;
432 			ip = mtod(m, struct ip *);
433 		}
434 #endif /* PFIL_HOOKS */
435 
436 #ifdef IPSEC
437 	/* get SP for this packet */
438 	if (so == NULL)
439 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
440 	else
441 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
442 
443 	if (sp == NULL) {
444 		ipsecstat.out_inval++;
445 		goto bad;
446 	}
447 
448 	error = 0;
449 
450 	/* check policy */
451 	switch (sp->policy) {
452 	case IPSEC_POLICY_DISCARD:
453 		/*
454 		 * This packet is just discarded.
455 		 */
456 		ipsecstat.out_polvio++;
457 		goto bad;
458 
459 	case IPSEC_POLICY_BYPASS:
460 	case IPSEC_POLICY_NONE:
461 		/* no need to do IPsec. */
462 		goto skip_ipsec;
463 
464 	case IPSEC_POLICY_IPSEC:
465 		if (sp->req == NULL) {
466 			/* XXX should be panic ? */
467 			printf("ip_output: No IPsec request specified.\n");
468 			error = EINVAL;
469 			goto bad;
470 		}
471 		break;
472 
473 	case IPSEC_POLICY_ENTRUST:
474 	default:
475 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
476 	}
477 
478 	ip->ip_len = htons((u_short)ip->ip_len);
479 	ip->ip_off = htons((u_short)ip->ip_off);
480 	ip->ip_sum = 0;
481 
482     {
483 	struct ipsec_output_state state;
484 	bzero(&state, sizeof(state));
485 	state.m = m;
486 	if (flags & IP_ROUTETOIF) {
487 		state.ro = &iproute;
488 		bzero(&iproute, sizeof(iproute));
489 	} else
490 		state.ro = ro;
491 	state.dst = (struct sockaddr *)dst;
492 
493 	error = ipsec4_output(&state, sp, flags);
494 
495 	m = state.m;
496 	if (flags & IP_ROUTETOIF) {
497 		/*
498 		 * if we have tunnel mode SA, we may need to ignore
499 		 * IP_ROUTETOIF.
500 		 */
501 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
502 			flags &= ~IP_ROUTETOIF;
503 			ro = state.ro;
504 		}
505 	} else
506 		ro = state.ro;
507 	dst = (struct sockaddr_in *)state.dst;
508 	if (error) {
509 		/* mbuf is already reclaimed in ipsec4_output. */
510 		m0 = NULL;
511 		switch (error) {
512 		case EHOSTUNREACH:
513 		case ENETUNREACH:
514 		case EMSGSIZE:
515 		case ENOBUFS:
516 		case ENOMEM:
517 			break;
518 		default:
519 			printf("ip4_output (ipsec): error code %d\n", error);
520 			/*fall through*/
521 		case ENOENT:
522 			/* don't show these error codes to the user */
523 			error = 0;
524 			break;
525 		}
526 		goto bad;
527 	}
528     }
529 
530 	/* be sure to update variables that are affected by ipsec4_output() */
531 	ip = mtod(m, struct ip *);
532 #ifdef _IP_VHL
533 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
534 #else
535 	hlen = ip->ip_hl << 2;
536 #endif
537 	if (ro->ro_rt == NULL) {
538 		if ((flags & IP_ROUTETOIF) == 0) {
539 			printf("ip_output: "
540 				"can't update route after IPsec processing\n");
541 			error = EHOSTUNREACH;	/*XXX*/
542 			goto bad;
543 		}
544 	} else {
545 		/* nobody uses ia beyond here */
546 		ifp = ro->ro_rt->rt_ifp;
547 	}
548 
549 	/* make it flipped, again. */
550 	ip->ip_len = ntohs((u_short)ip->ip_len);
551 	ip->ip_off = ntohs((u_short)ip->ip_off);
552 skip_ipsec:
553 #endif /*IPSEC*/
554 
555 	/*
556 	 * If small enough for mtu of path, can just send directly.
557 	 */
558 	if ((u_int16_t)ip->ip_len <= mtu) {
559 #if IFA_STATS
560 		/*
561 		 * search for the source address structure to
562 		 * maintain output statistics.
563 		 */
564 		bzero((caddr_t*) &src, sizeof(src));
565 		src.sin_family = AF_INET;
566 		src.sin_addr.s_addr = ip->ip_src.s_addr;
567 		src.sin_len = sizeof(src);
568 		ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
569 		if (ia)
570 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
571 #endif
572 		HTONS(ip->ip_len);
573 		HTONS(ip->ip_off);
574 		ip->ip_sum = 0;
575 		ip->ip_sum = in_cksum(m, hlen);
576 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
577 		goto done;
578 	}
579 
580 	/*
581 	 * Too large for interface; fragment if possible.
582 	 * Must be able to put at least 8 bytes per fragment.
583 	 */
584 #if 0
585 	/*
586 	 * If IPsec packet is too big for the interface, try fragment it.
587 	 * XXX This really is a quickhack.  May be inappropriate.
588 	 * XXX fails if somebody is sending AH'ed packet, with:
589 	 *	sizeof(packet without AH) < mtu < sizeof(packet with AH)
590 	 */
591 	if (sab && ip->ip_p != IPPROTO_AH && (flags & IP_FORWARDING) == 0)
592 		ip->ip_off &= ~IP_DF;
593 #endif /*IPSEC*/
594 	if (ip->ip_off & IP_DF) {
595 		if (flags & IP_RETURNMTU)
596 			*mtu_p = mtu;
597 		error = EMSGSIZE;
598 		ipstat.ips_cantfrag++;
599 		goto bad;
600 	}
601 	len = (mtu - hlen) &~ 7;
602 	if (len < 8) {
603 		error = EMSGSIZE;
604 		goto bad;
605 	}
606 
607     {
608 	int mhlen, firstlen = len;
609 	struct mbuf **mnext = &m->m_nextpkt;
610 	int fragments = 0;
611 	int s;
612 
613 	/*
614 	 * Loop through length of segment after first fragment,
615 	 * make new header and copy data of each part and link onto chain.
616 	 */
617 	m0 = m;
618 	mhlen = sizeof (struct ip);
619 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
620 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
621 		if (m == 0) {
622 			error = ENOBUFS;
623 			ipstat.ips_odropped++;
624 			goto sendorfree;
625 		}
626 		*mnext = m;
627 		mnext = &m->m_nextpkt;
628 		m->m_data += max_linkhdr;
629 		mhip = mtod(m, struct ip *);
630 		*mhip = *ip;
631 		if (hlen > sizeof (struct ip)) {
632 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
633 			mhip->ip_hl = mhlen >> 2;
634 		}
635 		m->m_len = mhlen;
636 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
637 		if (ip->ip_off & IP_MF)
638 			mhip->ip_off |= IP_MF;
639 		if (off + len >= (u_int16_t)ip->ip_len)
640 			len = (u_int16_t)ip->ip_len - off;
641 		else
642 			mhip->ip_off |= IP_MF;
643 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
644 		m->m_next = m_copy(m0, off, len);
645 		if (m->m_next == 0) {
646 			error = ENOBUFS;	/* ??? */
647 			ipstat.ips_odropped++;
648 			goto sendorfree;
649 		}
650 		m->m_pkthdr.len = mhlen + len;
651 		m->m_pkthdr.rcvif = (struct ifnet *)0;
652 		HTONS(mhip->ip_off);
653 		mhip->ip_sum = 0;
654 		mhip->ip_sum = in_cksum(m, mhlen);
655 		ipstat.ips_ofragments++;
656 		fragments++;
657 	}
658 	/*
659 	 * Update first fragment by trimming what's been copied out
660 	 * and updating header, then send each fragment (in order).
661 	 */
662 	m = m0;
663 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
664 	m->m_pkthdr.len = hlen + firstlen;
665 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
666 	ip->ip_off |= IP_MF;
667 	HTONS(ip->ip_off);
668 	ip->ip_sum = 0;
669 	ip->ip_sum = in_cksum(m, hlen);
670 sendorfree:
671 	/*
672 	 * If there is no room for all the fragments, don't queue
673 	 * any of them.
674 	 */
675 	s = splimp();
676 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
677 		error = ENOBUFS;
678 	splx(s);
679 	for (m = m0; m; m = m0) {
680 		m0 = m->m_nextpkt;
681 		m->m_nextpkt = 0;
682 		if (error == 0) {
683 #if IFA_STATS
684 			/*
685 			 * search for the source address structure to
686 			 * maintain output statistics.
687 			 */
688 			bzero((caddr_t*) &src, sizeof(src));
689 			src.sin_family = AF_INET;
690 			src.sin_addr.s_addr = ip->ip_src.s_addr;
691 			src.sin_len = sizeof(src);
692 			ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
693 			if (ia) {
694 				ia->ia_ifa.ifa_data.ifad_outbytes +=
695 					ntohs(ip->ip_len);
696 			}
697 #endif
698 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
699 			    ro->ro_rt);
700 		} else
701 			m_freem(m);
702 	}
703 
704 	if (error == 0)
705 		ipstat.ips_fragmented++;
706     }
707 done:
708 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
709 		RTFREE(ro->ro_rt);
710 		ro->ro_rt = 0;
711 	}
712 
713 #ifdef IPSEC
714 	if (sp != NULL) {
715 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
716 			printf("DP ip_output call free SP:%p\n", sp));
717 		key_freesp(sp);
718 	}
719 #endif /* IPSEC */
720 
721 	return (error);
722 bad:
723 	m_freem(m);
724 	goto done;
725 }
726 
727 /*
728  * Determine the maximum length of the options to be inserted;
729  * we would far rather allocate too much space rather than too little.
730  */
731 
732 u_int
733 ip_optlen(inp)
734 	struct inpcb *inp;
735 {
736 	struct mbuf *m = inp->inp_options;
737 
738 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
739 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
740 	else
741 		return 0;
742 }
743 
744 
745 /*
746  * Insert IP options into preformed packet.
747  * Adjust IP destination as required for IP source routing,
748  * as indicated by a non-zero in_addr at the start of the options.
749  */
750 static struct mbuf *
751 ip_insertoptions(m, opt, phlen)
752 	register struct mbuf *m;
753 	struct mbuf *opt;
754 	int *phlen;
755 {
756 	register struct ipoption *p = mtod(opt, struct ipoption *);
757 	struct mbuf *n;
758 	register struct ip *ip = mtod(m, struct ip *);
759 	unsigned optlen;
760 
761 	optlen = opt->m_len - sizeof(p->ipopt_dst);
762 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
763 		return (m);		/* XXX should fail */
764 	if (!in_nullhost(p->ipopt_dst))
765 		ip->ip_dst = p->ipopt_dst;
766 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
767 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
768 		if (n == 0)
769 			return (m);
770 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
771 		m->m_len -= sizeof(struct ip);
772 		m->m_data += sizeof(struct ip);
773 		n->m_next = m;
774 		m = n;
775 		m->m_len = optlen + sizeof(struct ip);
776 		m->m_data += max_linkhdr;
777 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
778 	} else {
779 		m->m_data -= optlen;
780 		m->m_len += optlen;
781 		m->m_pkthdr.len += optlen;
782 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
783 	}
784 	ip = mtod(m, struct ip *);
785 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
786 	*phlen = sizeof(struct ip) + optlen;
787 	ip->ip_len += optlen;
788 	return (m);
789 }
790 
791 /*
792  * Copy options from ip to jp,
793  * omitting those not copied during fragmentation.
794  */
795 int
796 ip_optcopy(ip, jp)
797 	struct ip *ip, *jp;
798 {
799 	register u_char *cp, *dp;
800 	int opt, optlen, cnt;
801 
802 	cp = (u_char *)(ip + 1);
803 	dp = (u_char *)(jp + 1);
804 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
805 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
806 		opt = cp[0];
807 		if (opt == IPOPT_EOL)
808 			break;
809 		if (opt == IPOPT_NOP) {
810 			/* Preserve for IP mcast tunnel's LSRR alignment. */
811 			*dp++ = IPOPT_NOP;
812 			optlen = 1;
813 			continue;
814 		} else
815 			optlen = cp[IPOPT_OLEN];
816 		/* bogus lengths should have been caught by ip_dooptions */
817 		if (optlen > cnt)
818 			optlen = cnt;
819 		if (IPOPT_COPIED(opt)) {
820 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
821 			dp += optlen;
822 		}
823 	}
824 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
825 		*dp++ = IPOPT_EOL;
826 	return (optlen);
827 }
828 
829 /*
830  * IP socket option processing.
831  */
832 int
833 ip_ctloutput(op, so, level, optname, mp)
834 	int op;
835 	struct socket *so;
836 	int level, optname;
837 	struct mbuf **mp;
838 {
839 	register struct inpcb *inp = sotoinpcb(so);
840 	register struct mbuf *m = *mp;
841 	register int optval = 0;
842 	int error = 0;
843 #ifdef IPSEC
844 #ifdef __NetBSD__
845 	struct proc *p = curproc;	/*XXX*/
846 #endif
847 #endif
848 
849 	if (level != IPPROTO_IP) {
850 		error = EINVAL;
851 		if (op == PRCO_SETOPT && *mp)
852 			(void) m_free(*mp);
853 	} else switch (op) {
854 
855 	case PRCO_SETOPT:
856 		switch (optname) {
857 		case IP_OPTIONS:
858 #ifdef notyet
859 		case IP_RETOPTS:
860 			return (ip_pcbopts(optname, &inp->inp_options, m));
861 #else
862 			return (ip_pcbopts(&inp->inp_options, m));
863 #endif
864 
865 		case IP_TOS:
866 		case IP_TTL:
867 		case IP_RECVOPTS:
868 		case IP_RECVRETOPTS:
869 		case IP_RECVDSTADDR:
870 		case IP_RECVIF:
871 			if (m == NULL || m->m_len != sizeof(int))
872 				error = EINVAL;
873 			else {
874 				optval = *mtod(m, int *);
875 				switch (optname) {
876 
877 				case IP_TOS:
878 					inp->inp_ip.ip_tos = optval;
879 					break;
880 
881 				case IP_TTL:
882 					inp->inp_ip.ip_ttl = optval;
883 					break;
884 #define	OPTSET(bit) \
885 	if (optval) \
886 		inp->inp_flags |= bit; \
887 	else \
888 		inp->inp_flags &= ~bit;
889 
890 				case IP_RECVOPTS:
891 					OPTSET(INP_RECVOPTS);
892 					break;
893 
894 				case IP_RECVRETOPTS:
895 					OPTSET(INP_RECVRETOPTS);
896 					break;
897 
898 				case IP_RECVDSTADDR:
899 					OPTSET(INP_RECVDSTADDR);
900 					break;
901 
902 				case IP_RECVIF:
903 					OPTSET(INP_RECVIF);
904 					break;
905 				}
906 			}
907 			break;
908 #undef OPTSET
909 
910 		case IP_MULTICAST_IF:
911 		case IP_MULTICAST_TTL:
912 		case IP_MULTICAST_LOOP:
913 		case IP_ADD_MEMBERSHIP:
914 		case IP_DROP_MEMBERSHIP:
915 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
916 			break;
917 
918 		case IP_PORTRANGE:
919 			if (m == 0 || m->m_len != sizeof(int))
920 				error = EINVAL;
921 			else {
922 				optval = *mtod(m, int *);
923 
924 				switch (optval) {
925 
926 				case IP_PORTRANGE_DEFAULT:
927 				case IP_PORTRANGE_HIGH:
928 					inp->inp_flags &= ~(INP_LOWPORT);
929 					break;
930 
931 				case IP_PORTRANGE_LOW:
932 					inp->inp_flags |= INP_LOWPORT;
933 					break;
934 
935 				default:
936 					error = EINVAL;
937 					break;
938 				}
939 			}
940 			break;
941 
942 #ifdef IPSEC
943 		case IP_IPSEC_POLICY:
944 		{
945 			caddr_t req = NULL;
946 			size_t len = 0;
947 			int priv = 0;
948 
949 #ifdef __NetBSD__
950 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
951 				priv = 0;
952 			else
953 				priv = 1;
954 #else
955 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
956 #endif
957 			if (m) {
958 				req = mtod(m, caddr_t);
959 				len = m->m_len;
960 			}
961 			error = ipsec4_set_policy(inp, optname, req, len, priv);
962 			break;
963 		    }
964 #endif /*IPSEC*/
965 
966 		default:
967 			error = ENOPROTOOPT;
968 			break;
969 		}
970 		if (m)
971 			(void)m_free(m);
972 		break;
973 
974 	case PRCO_GETOPT:
975 		switch (optname) {
976 		case IP_OPTIONS:
977 		case IP_RETOPTS:
978 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
979 			if (inp->inp_options) {
980 				m->m_len = inp->inp_options->m_len;
981 				bcopy(mtod(inp->inp_options, caddr_t),
982 				    mtod(m, caddr_t), (unsigned)m->m_len);
983 			} else
984 				m->m_len = 0;
985 			break;
986 
987 		case IP_TOS:
988 		case IP_TTL:
989 		case IP_RECVOPTS:
990 		case IP_RECVRETOPTS:
991 		case IP_RECVDSTADDR:
992 		case IP_RECVIF:
993 		case IP_ERRORMTU:
994 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
995 			m->m_len = sizeof(int);
996 			switch (optname) {
997 
998 			case IP_TOS:
999 				optval = inp->inp_ip.ip_tos;
1000 				break;
1001 
1002 			case IP_TTL:
1003 				optval = inp->inp_ip.ip_ttl;
1004 				break;
1005 
1006 			case IP_ERRORMTU:
1007 				optval = inp->inp_errormtu;
1008 				break;
1009 
1010 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1011 
1012 			case IP_RECVOPTS:
1013 				optval = OPTBIT(INP_RECVOPTS);
1014 				break;
1015 
1016 			case IP_RECVRETOPTS:
1017 				optval = OPTBIT(INP_RECVRETOPTS);
1018 				break;
1019 
1020 			case IP_RECVDSTADDR:
1021 				optval = OPTBIT(INP_RECVDSTADDR);
1022 				break;
1023 
1024 			case IP_RECVIF:
1025 				optval = OPTBIT(INP_RECVIF);
1026 				break;
1027 			}
1028 			*mtod(m, int *) = optval;
1029 			break;
1030 
1031 #ifdef IPSEC
1032 		case IP_IPSEC_POLICY:
1033 		{
1034 			caddr_t req = NULL;
1035 			size_t len;
1036 
1037 			if (m) {
1038 				req = mtod(m, caddr_t);
1039 				len = m->m_len;
1040 			}
1041 			error = ipsec4_get_policy(inp, req, len, mp);
1042 			break;
1043 		}
1044 #endif /*IPSEC*/
1045 
1046 		case IP_MULTICAST_IF:
1047 		case IP_MULTICAST_TTL:
1048 		case IP_MULTICAST_LOOP:
1049 		case IP_ADD_MEMBERSHIP:
1050 		case IP_DROP_MEMBERSHIP:
1051 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1052 			break;
1053 
1054 		case IP_PORTRANGE:
1055 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1056 			m->m_len = sizeof(int);
1057 
1058 			if (inp->inp_flags & INP_LOWPORT)
1059 				optval = IP_PORTRANGE_LOW;
1060 			else
1061 				optval = IP_PORTRANGE_DEFAULT;
1062 
1063 			*mtod(m, int *) = optval;
1064 			break;
1065 
1066 		default:
1067 			error = ENOPROTOOPT;
1068 			break;
1069 		}
1070 		break;
1071 	}
1072 	return (error);
1073 }
1074 
1075 /*
1076  * Set up IP options in pcb for insertion in output packets.
1077  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1078  * with destination address if source routed.
1079  */
1080 int
1081 #ifdef notyet
1082 ip_pcbopts(optname, pcbopt, m)
1083 	int optname;
1084 #else
1085 ip_pcbopts(pcbopt, m)
1086 #endif
1087 	struct mbuf **pcbopt;
1088 	register struct mbuf *m;
1089 {
1090 	register int cnt, optlen;
1091 	register u_char *cp;
1092 	u_char opt;
1093 
1094 	/* turn off any old options */
1095 	if (*pcbopt)
1096 		(void)m_free(*pcbopt);
1097 	*pcbopt = 0;
1098 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1099 		/*
1100 		 * Only turning off any previous options.
1101 		 */
1102 		if (m)
1103 			(void)m_free(m);
1104 		return (0);
1105 	}
1106 
1107 #ifndef	vax
1108 	if (m->m_len % sizeof(int32_t))
1109 		goto bad;
1110 #endif
1111 	/*
1112 	 * IP first-hop destination address will be stored before
1113 	 * actual options; move other options back
1114 	 * and clear it when none present.
1115 	 */
1116 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1117 		goto bad;
1118 	cnt = m->m_len;
1119 	m->m_len += sizeof(struct in_addr);
1120 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1121 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1122 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1123 
1124 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1125 		opt = cp[IPOPT_OPTVAL];
1126 		if (opt == IPOPT_EOL)
1127 			break;
1128 		if (opt == IPOPT_NOP)
1129 			optlen = 1;
1130 		else {
1131 			optlen = cp[IPOPT_OLEN];
1132 			if (optlen <= IPOPT_OLEN || optlen > cnt)
1133 				goto bad;
1134 		}
1135 		switch (opt) {
1136 
1137 		default:
1138 			break;
1139 
1140 		case IPOPT_LSRR:
1141 		case IPOPT_SSRR:
1142 			/*
1143 			 * user process specifies route as:
1144 			 *	->A->B->C->D
1145 			 * D must be our final destination (but we can't
1146 			 * check that since we may not have connected yet).
1147 			 * A is first hop destination, which doesn't appear in
1148 			 * actual IP option, but is stored before the options.
1149 			 */
1150 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1151 				goto bad;
1152 			m->m_len -= sizeof(struct in_addr);
1153 			cnt -= sizeof(struct in_addr);
1154 			optlen -= sizeof(struct in_addr);
1155 			cp[IPOPT_OLEN] = optlen;
1156 			/*
1157 			 * Move first hop before start of options.
1158 			 */
1159 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1160 			    sizeof(struct in_addr));
1161 			/*
1162 			 * Then copy rest of options back
1163 			 * to close up the deleted entry.
1164 			 */
1165 			memmove(&cp[IPOPT_OFFSET+1],
1166                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1167 			    (unsigned)cnt + sizeof(struct in_addr));
1168 			break;
1169 		}
1170 	}
1171 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1172 		goto bad;
1173 	*pcbopt = m;
1174 	return (0);
1175 
1176 bad:
1177 	(void)m_free(m);
1178 	return (EINVAL);
1179 }
1180 
1181 /*
1182  * Set the IP multicast options in response to user setsockopt().
1183  */
1184 int
1185 ip_setmoptions(optname, imop, m)
1186 	int optname;
1187 	struct ip_moptions **imop;
1188 	struct mbuf *m;
1189 {
1190 	register int error = 0;
1191 	u_char loop;
1192 	register int i;
1193 	struct in_addr addr;
1194 	register struct ip_mreq *mreq;
1195 	register struct ifnet *ifp;
1196 	register struct ip_moptions *imo = *imop;
1197 	struct route ro;
1198 	register struct sockaddr_in *dst;
1199 
1200 	if (imo == NULL) {
1201 		/*
1202 		 * No multicast option buffer attached to the pcb;
1203 		 * allocate one and initialize to default values.
1204 		 */
1205 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1206 		    M_WAITOK);
1207 
1208 		if (imo == NULL)
1209 			return (ENOBUFS);
1210 		*imop = imo;
1211 		imo->imo_multicast_ifp = NULL;
1212 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1213 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1214 		imo->imo_num_memberships = 0;
1215 	}
1216 
1217 	switch (optname) {
1218 
1219 	case IP_MULTICAST_IF:
1220 		/*
1221 		 * Select the interface for outgoing multicast packets.
1222 		 */
1223 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1224 			error = EINVAL;
1225 			break;
1226 		}
1227 		addr = *(mtod(m, struct in_addr *));
1228 		/*
1229 		 * INADDR_ANY is used to remove a previous selection.
1230 		 * When no interface is selected, a default one is
1231 		 * chosen every time a multicast packet is sent.
1232 		 */
1233 		if (in_nullhost(addr)) {
1234 			imo->imo_multicast_ifp = NULL;
1235 			break;
1236 		}
1237 		/*
1238 		 * The selected interface is identified by its local
1239 		 * IP address.  Find the interface and confirm that
1240 		 * it supports multicasting.
1241 		 */
1242 		INADDR_TO_IFP(addr, ifp);
1243 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1244 			error = EADDRNOTAVAIL;
1245 			break;
1246 		}
1247 		imo->imo_multicast_ifp = ifp;
1248 		break;
1249 
1250 	case IP_MULTICAST_TTL:
1251 		/*
1252 		 * Set the IP time-to-live for outgoing multicast packets.
1253 		 */
1254 		if (m == NULL || m->m_len != 1) {
1255 			error = EINVAL;
1256 			break;
1257 		}
1258 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1259 		break;
1260 
1261 	case IP_MULTICAST_LOOP:
1262 		/*
1263 		 * Set the loopback flag for outgoing multicast packets.
1264 		 * Must be zero or one.
1265 		 */
1266 		if (m == NULL || m->m_len != 1 ||
1267 		   (loop = *(mtod(m, u_char *))) > 1) {
1268 			error = EINVAL;
1269 			break;
1270 		}
1271 		imo->imo_multicast_loop = loop;
1272 		break;
1273 
1274 	case IP_ADD_MEMBERSHIP:
1275 		/*
1276 		 * Add a multicast group membership.
1277 		 * Group must be a valid IP multicast address.
1278 		 */
1279 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1280 			error = EINVAL;
1281 			break;
1282 		}
1283 		mreq = mtod(m, struct ip_mreq *);
1284 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1285 			error = EINVAL;
1286 			break;
1287 		}
1288 		/*
1289 		 * If no interface address was provided, use the interface of
1290 		 * the route to the given multicast address.
1291 		 */
1292 		if (in_nullhost(mreq->imr_interface)) {
1293 			bzero((caddr_t)&ro, sizeof(ro));
1294 			ro.ro_rt = NULL;
1295 			dst = satosin(&ro.ro_dst);
1296 			dst->sin_len = sizeof(*dst);
1297 			dst->sin_family = AF_INET;
1298 			dst->sin_addr = mreq->imr_multiaddr;
1299 			rtalloc(&ro);
1300 			if (ro.ro_rt == NULL) {
1301 				error = EADDRNOTAVAIL;
1302 				break;
1303 			}
1304 			ifp = ro.ro_rt->rt_ifp;
1305 			rtfree(ro.ro_rt);
1306 		} else {
1307 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1308 		}
1309 		/*
1310 		 * See if we found an interface, and confirm that it
1311 		 * supports multicast.
1312 		 */
1313 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1314 			error = EADDRNOTAVAIL;
1315 			break;
1316 		}
1317 		/*
1318 		 * See if the membership already exists or if all the
1319 		 * membership slots are full.
1320 		 */
1321 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1322 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1323 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1324 				      mreq->imr_multiaddr))
1325 				break;
1326 		}
1327 		if (i < imo->imo_num_memberships) {
1328 			error = EADDRINUSE;
1329 			break;
1330 		}
1331 		if (i == IP_MAX_MEMBERSHIPS) {
1332 			error = ETOOMANYREFS;
1333 			break;
1334 		}
1335 		/*
1336 		 * Everything looks good; add a new record to the multicast
1337 		 * address list for the given interface.
1338 		 */
1339 		if ((imo->imo_membership[i] =
1340 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1341 			error = ENOBUFS;
1342 			break;
1343 		}
1344 		++imo->imo_num_memberships;
1345 		break;
1346 
1347 	case IP_DROP_MEMBERSHIP:
1348 		/*
1349 		 * Drop a multicast group membership.
1350 		 * Group must be a valid IP multicast address.
1351 		 */
1352 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1353 			error = EINVAL;
1354 			break;
1355 		}
1356 		mreq = mtod(m, struct ip_mreq *);
1357 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1358 			error = EINVAL;
1359 			break;
1360 		}
1361 		/*
1362 		 * If an interface address was specified, get a pointer
1363 		 * to its ifnet structure.
1364 		 */
1365 		if (in_nullhost(mreq->imr_interface))
1366 			ifp = NULL;
1367 		else {
1368 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1369 			if (ifp == NULL) {
1370 				error = EADDRNOTAVAIL;
1371 				break;
1372 			}
1373 		}
1374 		/*
1375 		 * Find the membership in the membership array.
1376 		 */
1377 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1378 			if ((ifp == NULL ||
1379 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1380 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1381 				       mreq->imr_multiaddr))
1382 				break;
1383 		}
1384 		if (i == imo->imo_num_memberships) {
1385 			error = EADDRNOTAVAIL;
1386 			break;
1387 		}
1388 		/*
1389 		 * Give up the multicast address record to which the
1390 		 * membership points.
1391 		 */
1392 		in_delmulti(imo->imo_membership[i]);
1393 		/*
1394 		 * Remove the gap in the membership array.
1395 		 */
1396 		for (++i; i < imo->imo_num_memberships; ++i)
1397 			imo->imo_membership[i-1] = imo->imo_membership[i];
1398 		--imo->imo_num_memberships;
1399 		break;
1400 
1401 	default:
1402 		error = EOPNOTSUPP;
1403 		break;
1404 	}
1405 
1406 	/*
1407 	 * If all options have default values, no need to keep the mbuf.
1408 	 */
1409 	if (imo->imo_multicast_ifp == NULL &&
1410 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1411 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1412 	    imo->imo_num_memberships == 0) {
1413 		free(*imop, M_IPMOPTS);
1414 		*imop = NULL;
1415 	}
1416 
1417 	return (error);
1418 }
1419 
1420 /*
1421  * Return the IP multicast options in response to user getsockopt().
1422  */
1423 int
1424 ip_getmoptions(optname, imo, mp)
1425 	int optname;
1426 	register struct ip_moptions *imo;
1427 	register struct mbuf **mp;
1428 {
1429 	u_char *ttl;
1430 	u_char *loop;
1431 	struct in_addr *addr;
1432 	struct in_ifaddr *ia;
1433 
1434 	*mp = m_get(M_WAIT, MT_SOOPTS);
1435 
1436 	switch (optname) {
1437 
1438 	case IP_MULTICAST_IF:
1439 		addr = mtod(*mp, struct in_addr *);
1440 		(*mp)->m_len = sizeof(struct in_addr);
1441 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1442 			*addr = zeroin_addr;
1443 		else {
1444 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1445 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1446 		}
1447 		return (0);
1448 
1449 	case IP_MULTICAST_TTL:
1450 		ttl = mtod(*mp, u_char *);
1451 		(*mp)->m_len = 1;
1452 		*ttl = imo ? imo->imo_multicast_ttl
1453 			   : IP_DEFAULT_MULTICAST_TTL;
1454 		return (0);
1455 
1456 	case IP_MULTICAST_LOOP:
1457 		loop = mtod(*mp, u_char *);
1458 		(*mp)->m_len = 1;
1459 		*loop = imo ? imo->imo_multicast_loop
1460 			    : IP_DEFAULT_MULTICAST_LOOP;
1461 		return (0);
1462 
1463 	default:
1464 		return (EOPNOTSUPP);
1465 	}
1466 }
1467 
1468 /*
1469  * Discard the IP multicast options.
1470  */
1471 void
1472 ip_freemoptions(imo)
1473 	register struct ip_moptions *imo;
1474 {
1475 	register int i;
1476 
1477 	if (imo != NULL) {
1478 		for (i = 0; i < imo->imo_num_memberships; ++i)
1479 			in_delmulti(imo->imo_membership[i]);
1480 		free(imo, M_IPMOPTS);
1481 	}
1482 }
1483 
1484 /*
1485  * Routine called from ip_output() to loop back a copy of an IP multicast
1486  * packet to the input queue of a specified interface.  Note that this
1487  * calls the output routine of the loopback "driver", but with an interface
1488  * pointer that might NOT be &loif -- easier than replicating that code here.
1489  */
1490 static void
1491 ip_mloopback(ifp, m, dst)
1492 	struct ifnet *ifp;
1493 	register struct mbuf *m;
1494 	register struct sockaddr_in *dst;
1495 {
1496 	register struct ip *ip;
1497 	struct mbuf *copym;
1498 
1499 	copym = m_copy(m, 0, M_COPYALL);
1500         if (copym != NULL
1501 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1502 		copym = m_pullup(copym, sizeof(struct ip));
1503 	if (copym != NULL) {
1504 		/*
1505 		 * We don't bother to fragment if the IP length is greater
1506 		 * than the interface's MTU.  Can this possibly matter?
1507 		 */
1508 		ip = mtod(copym, struct ip *);
1509 		HTONS(ip->ip_len);
1510 		HTONS(ip->ip_off);
1511 		ip->ip_sum = 0;
1512 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1513 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1514 	}
1515 }
1516