xref: /netbsd-src/sys/netinet/ip_output.c (revision bcc8ec9959e7b01e313d813067bfb43a3ad70551)
1 /*	$NetBSD: ip_output.c,v 1.81 2001/01/13 07:19:33 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 #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 struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
147 static void ip_mloopback
148 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
149 
150 #ifdef PFIL_HOOKS
151 extern struct pfil_head inet_pfil_hook;			/* XXX */
152 #endif
153 
154 /*
155  * IP output.  The packet in mbuf chain m contains a skeletal IP
156  * header (with len, off, ttl, proto, tos, src, dst).
157  * The mbuf chain containing the packet will be freed.
158  * The mbuf opt, if present, will not be freed.
159  */
160 int
161 #if __STDC__
162 ip_output(struct mbuf *m0, ...)
163 #else
164 ip_output(m0, va_alist)
165 	struct mbuf *m0;
166 	va_dcl
167 #endif
168 {
169 	struct ip *ip, *mhip;
170 	struct ifnet *ifp;
171 	struct mbuf *m = m0;
172 	int hlen = sizeof (struct ip);
173 	int len, off, error = 0;
174 	struct route iproute;
175 	struct sockaddr_in *dst;
176 	struct in_ifaddr *ia;
177 	struct mbuf *opt;
178 	struct route *ro;
179 	int flags;
180 	int *mtu_p;
181 	int mtu;
182 	struct ip_moptions *imo;
183 	va_list ap;
184 #ifdef IPSEC
185 	struct socket *so;
186 	struct secpolicy *sp = NULL;
187 #endif /*IPSEC*/
188 	u_int16_t ip_len;
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 			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 	/*
417 	 * If we're doing Path MTU Discovery, we need to set DF unless
418 	 * the route's MTU is locked.
419 	 */
420 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
421 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
422 		ip->ip_off |= IP_DF;
423 
424 	/*
425 	 * Remember the current ip_len and ip_off, and swap them into
426 	 * network order.
427 	 */
428 	ip_len = ip->ip_len;
429 
430 	HTONS(ip->ip_len);
431 	HTONS(ip->ip_off);
432 
433 #ifdef PFIL_HOOKS
434 	/*
435 	 * Run through list of hooks for output packets.
436 	 */
437 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp,
438 				    PFIL_OUT)) != 0)
439 		goto done;
440 	if (m == NULL)
441 		goto done;
442 
443 	ip = mtod(m, struct ip *);
444 #endif /* PFIL_HOOKS */
445 
446 #ifdef IPSEC
447 	/* get SP for this packet */
448 	if (so == NULL)
449 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
450 	else
451 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
452 
453 	if (sp == NULL) {
454 		ipsecstat.out_inval++;
455 		goto bad;
456 	}
457 
458 	error = 0;
459 
460 	/* check policy */
461 	switch (sp->policy) {
462 	case IPSEC_POLICY_DISCARD:
463 		/*
464 		 * This packet is just discarded.
465 		 */
466 		ipsecstat.out_polvio++;
467 		goto bad;
468 
469 	case IPSEC_POLICY_BYPASS:
470 	case IPSEC_POLICY_NONE:
471 		/* no need to do IPsec. */
472 		goto skip_ipsec;
473 
474 	case IPSEC_POLICY_IPSEC:
475 		if (sp->req == NULL) {
476 			/* XXX should be panic ? */
477 			printf("ip_output: No IPsec request specified.\n");
478 			error = EINVAL;
479 			goto bad;
480 		}
481 		break;
482 
483 	case IPSEC_POLICY_ENTRUST:
484 	default:
485 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
486 	}
487 
488 	/*
489 	 * ipsec4_output() expects ip_len and ip_off in network
490 	 * order.  They have been set to network order above.
491 	 */
492 
493     {
494 	struct ipsec_output_state state;
495 	bzero(&state, sizeof(state));
496 	state.m = m;
497 	if (flags & IP_ROUTETOIF) {
498 		state.ro = &iproute;
499 		bzero(&iproute, sizeof(iproute));
500 	} else
501 		state.ro = ro;
502 	state.dst = (struct sockaddr *)dst;
503 
504 	error = ipsec4_output(&state, sp, flags);
505 
506 	m = state.m;
507 	if (flags & IP_ROUTETOIF) {
508 		/*
509 		 * if we have tunnel mode SA, we may need to ignore
510 		 * IP_ROUTETOIF.
511 		 */
512 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
513 			flags &= ~IP_ROUTETOIF;
514 			ro = state.ro;
515 		}
516 	} else
517 		ro = state.ro;
518 	dst = (struct sockaddr_in *)state.dst;
519 	if (error) {
520 		/* mbuf is already reclaimed in ipsec4_output. */
521 		m0 = NULL;
522 		switch (error) {
523 		case EHOSTUNREACH:
524 		case ENETUNREACH:
525 		case EMSGSIZE:
526 		case ENOBUFS:
527 		case ENOMEM:
528 			break;
529 		default:
530 			printf("ip4_output (ipsec): error code %d\n", error);
531 			/*fall through*/
532 		case ENOENT:
533 			/* don't show these error codes to the user */
534 			error = 0;
535 			break;
536 		}
537 		goto bad;
538 	}
539     }
540 
541 	/* be sure to update variables that are affected by ipsec4_output() */
542 	ip = mtod(m, struct ip *);
543 #ifdef _IP_VHL
544 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
545 #else
546 	hlen = ip->ip_hl << 2;
547 #endif
548 	ip_len = ntohs(ip->ip_len);
549 
550 	if (ro->ro_rt == NULL) {
551 		if ((flags & IP_ROUTETOIF) == 0) {
552 			printf("ip_output: "
553 				"can't update route after IPsec processing\n");
554 			error = EHOSTUNREACH;	/*XXX*/
555 			goto bad;
556 		}
557 	} else {
558 		/* nobody uses ia beyond here */
559 		ifp = ro->ro_rt->rt_ifp;
560 	}
561 
562 skip_ipsec:
563 #endif /*IPSEC*/
564 
565 	/*
566 	 * If small enough for mtu of path, can just send directly.
567 	 */
568 	if (ip_len <= mtu) {
569 #if IFA_STATS
570 		/*
571 		 * search for the source address structure to
572 		 * maintain output statistics.
573 		 */
574 		INADDR_TO_IA(ip->ip_src, ia);
575 		if (ia)
576 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
577 #endif
578 		ip->ip_sum = 0;
579 		ip->ip_sum = in_cksum(m, hlen);
580 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
581 		goto done;
582 	}
583 
584 	/*
585 	 * Too large for interface; fragment if possible.
586 	 * Must be able to put at least 8 bytes per fragment.
587 	 *
588 	 * Note we swap ip_len and ip_off into host order to make
589 	 * the logic below a little simpler.
590 	 */
591 
592 	NTOHS(ip->ip_len);
593 	NTOHS(ip->ip_off);
594 
595 #if 0
596 	/*
597 	 * If IPsec packet is too big for the interface, try fragment it.
598 	 * XXX This really is a quickhack.  May be inappropriate.
599 	 * XXX fails if somebody is sending AH'ed packet, with:
600 	 *	sizeof(packet without AH) < mtu < sizeof(packet with AH)
601 	 */
602 	if (sab && ip->ip_p != IPPROTO_AH && (flags & IP_FORWARDING) == 0)
603 		ip->ip_off &= ~IP_DF;
604 #endif /*IPSEC*/
605 	if (ip->ip_off & IP_DF) {
606 		if (flags & IP_RETURNMTU)
607 			*mtu_p = mtu;
608 		error = EMSGSIZE;
609 		ipstat.ips_cantfrag++;
610 		goto bad;
611 	}
612 	len = (mtu - hlen) &~ 7;
613 	if (len < 8) {
614 		error = EMSGSIZE;
615 		goto bad;
616 	}
617 
618     {
619 	int mhlen, firstlen = len;
620 	struct mbuf **mnext = &m->m_nextpkt;
621 	int fragments = 0;
622 	int s;
623 
624 	/*
625 	 * Loop through length of segment after first fragment,
626 	 * make new header and copy data of each part and link onto chain.
627 	 */
628 	m0 = m;
629 	mhlen = sizeof (struct ip);
630 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
631 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
632 		if (m == 0) {
633 			error = ENOBUFS;
634 			ipstat.ips_odropped++;
635 			goto sendorfree;
636 		}
637 		*mnext = m;
638 		mnext = &m->m_nextpkt;
639 		m->m_data += max_linkhdr;
640 		mhip = mtod(m, struct ip *);
641 		*mhip = *ip;
642 		/* we must inherit MCAST and BCAST flags */
643 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
644 		if (hlen > sizeof (struct ip)) {
645 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
646 			mhip->ip_hl = mhlen >> 2;
647 		}
648 		m->m_len = mhlen;
649 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
650 		if (ip->ip_off & IP_MF)
651 			mhip->ip_off |= IP_MF;
652 		if (off + len >= (u_int16_t)ip->ip_len)
653 			len = (u_int16_t)ip->ip_len - off;
654 		else
655 			mhip->ip_off |= IP_MF;
656 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
657 		m->m_next = m_copy(m0, off, len);
658 		if (m->m_next == 0) {
659 			error = ENOBUFS;	/* ??? */
660 			ipstat.ips_odropped++;
661 			goto sendorfree;
662 		}
663 		m->m_pkthdr.len = mhlen + len;
664 		m->m_pkthdr.rcvif = (struct ifnet *)0;
665 		HTONS(mhip->ip_off);
666 		mhip->ip_sum = 0;
667 		mhip->ip_sum = in_cksum(m, mhlen);
668 		ipstat.ips_ofragments++;
669 		fragments++;
670 	}
671 	/*
672 	 * Update first fragment by trimming what's been copied out
673 	 * and updating header, then send each fragment (in order).
674 	 */
675 	m = m0;
676 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
677 	m->m_pkthdr.len = hlen + firstlen;
678 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
679 	ip->ip_off |= IP_MF;
680 	HTONS(ip->ip_off);
681 	ip->ip_sum = 0;
682 	ip->ip_sum = in_cksum(m, hlen);
683 sendorfree:
684 	/*
685 	 * If there is no room for all the fragments, don't queue
686 	 * any of them.
687 	 */
688 	s = splimp();
689 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
690 		error = ENOBUFS;
691 	splx(s);
692 	for (m = m0; m; m = m0) {
693 		m0 = m->m_nextpkt;
694 		m->m_nextpkt = 0;
695 		if (error == 0) {
696 #if IFA_STATS
697 			/*
698 			 * search for the source address structure to
699 			 * maintain output statistics.
700 			 */
701 			INADDR_TO_IA(ip->ip_src, ia);
702 			if (ia) {
703 				ia->ia_ifa.ifa_data.ifad_outbytes +=
704 					ntohs(ip->ip_len);
705 			}
706 #endif
707 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
708 			    ro->ro_rt);
709 		} else
710 			m_freem(m);
711 	}
712 
713 	if (error == 0)
714 		ipstat.ips_fragmented++;
715     }
716 done:
717 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
718 		RTFREE(ro->ro_rt);
719 		ro->ro_rt = 0;
720 	}
721 
722 #ifdef IPSEC
723 	if (sp != NULL) {
724 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
725 			printf("DP ip_output call free SP:%p\n", sp));
726 		key_freesp(sp);
727 	}
728 #endif /* IPSEC */
729 
730 	return (error);
731 bad:
732 	m_freem(m);
733 	goto done;
734 }
735 
736 /*
737  * Determine the maximum length of the options to be inserted;
738  * we would far rather allocate too much space rather than too little.
739  */
740 
741 u_int
742 ip_optlen(inp)
743 	struct inpcb *inp;
744 {
745 	struct mbuf *m = inp->inp_options;
746 
747 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
748 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
749 	else
750 		return 0;
751 }
752 
753 
754 /*
755  * Insert IP options into preformed packet.
756  * Adjust IP destination as required for IP source routing,
757  * as indicated by a non-zero in_addr at the start of the options.
758  */
759 static struct mbuf *
760 ip_insertoptions(m, opt, phlen)
761 	struct mbuf *m;
762 	struct mbuf *opt;
763 	int *phlen;
764 {
765 	struct ipoption *p = mtod(opt, struct ipoption *);
766 	struct mbuf *n;
767 	struct ip *ip = mtod(m, struct ip *);
768 	unsigned optlen;
769 
770 	optlen = opt->m_len - sizeof(p->ipopt_dst);
771 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
772 		return (m);		/* XXX should fail */
773 	if (!in_nullhost(p->ipopt_dst))
774 		ip->ip_dst = p->ipopt_dst;
775 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
776 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
777 		if (n == 0)
778 			return (m);
779 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
780 		m->m_len -= sizeof(struct ip);
781 		m->m_data += sizeof(struct ip);
782 		n->m_next = m;
783 		m = n;
784 		m->m_len = optlen + sizeof(struct ip);
785 		m->m_data += max_linkhdr;
786 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
787 	} else {
788 		m->m_data -= optlen;
789 		m->m_len += optlen;
790 		m->m_pkthdr.len += optlen;
791 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
792 	}
793 	ip = mtod(m, struct ip *);
794 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
795 	*phlen = sizeof(struct ip) + optlen;
796 	ip->ip_len += optlen;
797 	return (m);
798 }
799 
800 /*
801  * Copy options from ip to jp,
802  * omitting those not copied during fragmentation.
803  */
804 int
805 ip_optcopy(ip, jp)
806 	struct ip *ip, *jp;
807 {
808 	u_char *cp, *dp;
809 	int opt, optlen, cnt;
810 
811 	cp = (u_char *)(ip + 1);
812 	dp = (u_char *)(jp + 1);
813 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
814 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
815 		opt = cp[0];
816 		if (opt == IPOPT_EOL)
817 			break;
818 		if (opt == IPOPT_NOP) {
819 			/* Preserve for IP mcast tunnel's LSRR alignment. */
820 			*dp++ = IPOPT_NOP;
821 			optlen = 1;
822 			continue;
823 		}
824 #ifdef DIAGNOSTIC
825 		if (cnt < IPOPT_OLEN + sizeof(*cp))
826 			panic("malformed IPv4 option passed to ip_optcopy");
827 #endif
828 		optlen = cp[IPOPT_OLEN];
829 #ifdef DIAGNOSTIC
830 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
831 			panic("malformed IPv4 option passed to ip_optcopy");
832 #endif
833 		/* bogus lengths should have been caught by ip_dooptions */
834 		if (optlen > cnt)
835 			optlen = cnt;
836 		if (IPOPT_COPIED(opt)) {
837 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
838 			dp += optlen;
839 		}
840 	}
841 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
842 		*dp++ = IPOPT_EOL;
843 	return (optlen);
844 }
845 
846 /*
847  * IP socket option processing.
848  */
849 int
850 ip_ctloutput(op, so, level, optname, mp)
851 	int op;
852 	struct socket *so;
853 	int level, optname;
854 	struct mbuf **mp;
855 {
856 	struct inpcb *inp = sotoinpcb(so);
857 	struct mbuf *m = *mp;
858 	int optval = 0;
859 	int error = 0;
860 #ifdef IPSEC
861 #ifdef __NetBSD__
862 	struct proc *p = curproc;	/*XXX*/
863 #endif
864 #endif
865 
866 	if (level != IPPROTO_IP) {
867 		error = EINVAL;
868 		if (op == PRCO_SETOPT && *mp)
869 			(void) m_free(*mp);
870 	} else switch (op) {
871 
872 	case PRCO_SETOPT:
873 		switch (optname) {
874 		case IP_OPTIONS:
875 #ifdef notyet
876 		case IP_RETOPTS:
877 			return (ip_pcbopts(optname, &inp->inp_options, m));
878 #else
879 			return (ip_pcbopts(&inp->inp_options, m));
880 #endif
881 
882 		case IP_TOS:
883 		case IP_TTL:
884 		case IP_RECVOPTS:
885 		case IP_RECVRETOPTS:
886 		case IP_RECVDSTADDR:
887 		case IP_RECVIF:
888 			if (m == NULL || m->m_len != sizeof(int))
889 				error = EINVAL;
890 			else {
891 				optval = *mtod(m, int *);
892 				switch (optname) {
893 
894 				case IP_TOS:
895 					inp->inp_ip.ip_tos = optval;
896 					break;
897 
898 				case IP_TTL:
899 					inp->inp_ip.ip_ttl = optval;
900 					break;
901 #define	OPTSET(bit) \
902 	if (optval) \
903 		inp->inp_flags |= bit; \
904 	else \
905 		inp->inp_flags &= ~bit;
906 
907 				case IP_RECVOPTS:
908 					OPTSET(INP_RECVOPTS);
909 					break;
910 
911 				case IP_RECVRETOPTS:
912 					OPTSET(INP_RECVRETOPTS);
913 					break;
914 
915 				case IP_RECVDSTADDR:
916 					OPTSET(INP_RECVDSTADDR);
917 					break;
918 
919 				case IP_RECVIF:
920 					OPTSET(INP_RECVIF);
921 					break;
922 				}
923 			}
924 			break;
925 #undef OPTSET
926 
927 		case IP_MULTICAST_IF:
928 		case IP_MULTICAST_TTL:
929 		case IP_MULTICAST_LOOP:
930 		case IP_ADD_MEMBERSHIP:
931 		case IP_DROP_MEMBERSHIP:
932 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
933 			break;
934 
935 		case IP_PORTRANGE:
936 			if (m == 0 || m->m_len != sizeof(int))
937 				error = EINVAL;
938 			else {
939 				optval = *mtod(m, int *);
940 
941 				switch (optval) {
942 
943 				case IP_PORTRANGE_DEFAULT:
944 				case IP_PORTRANGE_HIGH:
945 					inp->inp_flags &= ~(INP_LOWPORT);
946 					break;
947 
948 				case IP_PORTRANGE_LOW:
949 					inp->inp_flags |= INP_LOWPORT;
950 					break;
951 
952 				default:
953 					error = EINVAL;
954 					break;
955 				}
956 			}
957 			break;
958 
959 #ifdef IPSEC
960 		case IP_IPSEC_POLICY:
961 		{
962 			caddr_t req = NULL;
963 			size_t len = 0;
964 			int priv = 0;
965 
966 #ifdef __NetBSD__
967 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
968 				priv = 0;
969 			else
970 				priv = 1;
971 #else
972 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
973 #endif
974 			if (m) {
975 				req = mtod(m, caddr_t);
976 				len = m->m_len;
977 			}
978 			error = ipsec4_set_policy(inp, optname, req, len, priv);
979 			break;
980 		    }
981 #endif /*IPSEC*/
982 
983 		default:
984 			error = ENOPROTOOPT;
985 			break;
986 		}
987 		if (m)
988 			(void)m_free(m);
989 		break;
990 
991 	case PRCO_GETOPT:
992 		switch (optname) {
993 		case IP_OPTIONS:
994 		case IP_RETOPTS:
995 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
996 			if (inp->inp_options) {
997 				m->m_len = inp->inp_options->m_len;
998 				bcopy(mtod(inp->inp_options, caddr_t),
999 				    mtod(m, caddr_t), (unsigned)m->m_len);
1000 			} else
1001 				m->m_len = 0;
1002 			break;
1003 
1004 		case IP_TOS:
1005 		case IP_TTL:
1006 		case IP_RECVOPTS:
1007 		case IP_RECVRETOPTS:
1008 		case IP_RECVDSTADDR:
1009 		case IP_RECVIF:
1010 		case IP_ERRORMTU:
1011 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1012 			m->m_len = sizeof(int);
1013 			switch (optname) {
1014 
1015 			case IP_TOS:
1016 				optval = inp->inp_ip.ip_tos;
1017 				break;
1018 
1019 			case IP_TTL:
1020 				optval = inp->inp_ip.ip_ttl;
1021 				break;
1022 
1023 			case IP_ERRORMTU:
1024 				optval = inp->inp_errormtu;
1025 				break;
1026 
1027 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1028 
1029 			case IP_RECVOPTS:
1030 				optval = OPTBIT(INP_RECVOPTS);
1031 				break;
1032 
1033 			case IP_RECVRETOPTS:
1034 				optval = OPTBIT(INP_RECVRETOPTS);
1035 				break;
1036 
1037 			case IP_RECVDSTADDR:
1038 				optval = OPTBIT(INP_RECVDSTADDR);
1039 				break;
1040 
1041 			case IP_RECVIF:
1042 				optval = OPTBIT(INP_RECVIF);
1043 				break;
1044 			}
1045 			*mtod(m, int *) = optval;
1046 			break;
1047 
1048 #ifdef IPSEC
1049 		case IP_IPSEC_POLICY:
1050 		{
1051 			caddr_t req = NULL;
1052 			size_t len = 0;
1053 
1054 			if (m) {
1055 				req = mtod(m, caddr_t);
1056 				len = m->m_len;
1057 			}
1058 			error = ipsec4_get_policy(inp, req, len, mp);
1059 			break;
1060 		}
1061 #endif /*IPSEC*/
1062 
1063 		case IP_MULTICAST_IF:
1064 		case IP_MULTICAST_TTL:
1065 		case IP_MULTICAST_LOOP:
1066 		case IP_ADD_MEMBERSHIP:
1067 		case IP_DROP_MEMBERSHIP:
1068 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1069 			break;
1070 
1071 		case IP_PORTRANGE:
1072 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1073 			m->m_len = sizeof(int);
1074 
1075 			if (inp->inp_flags & INP_LOWPORT)
1076 				optval = IP_PORTRANGE_LOW;
1077 			else
1078 				optval = IP_PORTRANGE_DEFAULT;
1079 
1080 			*mtod(m, int *) = optval;
1081 			break;
1082 
1083 		default:
1084 			error = ENOPROTOOPT;
1085 			break;
1086 		}
1087 		break;
1088 	}
1089 	return (error);
1090 }
1091 
1092 /*
1093  * Set up IP options in pcb for insertion in output packets.
1094  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1095  * with destination address if source routed.
1096  */
1097 int
1098 #ifdef notyet
1099 ip_pcbopts(optname, pcbopt, m)
1100 	int optname;
1101 #else
1102 ip_pcbopts(pcbopt, m)
1103 #endif
1104 	struct mbuf **pcbopt;
1105 	struct mbuf *m;
1106 {
1107 	int cnt, optlen;
1108 	u_char *cp;
1109 	u_char opt;
1110 
1111 	/* turn off any old options */
1112 	if (*pcbopt)
1113 		(void)m_free(*pcbopt);
1114 	*pcbopt = 0;
1115 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1116 		/*
1117 		 * Only turning off any previous options.
1118 		 */
1119 		if (m)
1120 			(void)m_free(m);
1121 		return (0);
1122 	}
1123 
1124 #ifndef	vax
1125 	if (m->m_len % sizeof(int32_t))
1126 		goto bad;
1127 #endif
1128 	/*
1129 	 * IP first-hop destination address will be stored before
1130 	 * actual options; move other options back
1131 	 * and clear it when none present.
1132 	 */
1133 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1134 		goto bad;
1135 	cnt = m->m_len;
1136 	m->m_len += sizeof(struct in_addr);
1137 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1138 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1139 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1140 
1141 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1142 		opt = cp[IPOPT_OPTVAL];
1143 		if (opt == IPOPT_EOL)
1144 			break;
1145 		if (opt == IPOPT_NOP)
1146 			optlen = 1;
1147 		else {
1148 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1149 				goto bad;
1150 			optlen = cp[IPOPT_OLEN];
1151 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1152 				goto bad;
1153 		}
1154 		switch (opt) {
1155 
1156 		default:
1157 			break;
1158 
1159 		case IPOPT_LSRR:
1160 		case IPOPT_SSRR:
1161 			/*
1162 			 * user process specifies route as:
1163 			 *	->A->B->C->D
1164 			 * D must be our final destination (but we can't
1165 			 * check that since we may not have connected yet).
1166 			 * A is first hop destination, which doesn't appear in
1167 			 * actual IP option, but is stored before the options.
1168 			 */
1169 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1170 				goto bad;
1171 			m->m_len -= sizeof(struct in_addr);
1172 			cnt -= sizeof(struct in_addr);
1173 			optlen -= sizeof(struct in_addr);
1174 			cp[IPOPT_OLEN] = optlen;
1175 			/*
1176 			 * Move first hop before start of options.
1177 			 */
1178 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1179 			    sizeof(struct in_addr));
1180 			/*
1181 			 * Then copy rest of options back
1182 			 * to close up the deleted entry.
1183 			 */
1184 			memmove(&cp[IPOPT_OFFSET+1],
1185                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1186 			    (unsigned)cnt + sizeof(struct in_addr));
1187 			break;
1188 		}
1189 	}
1190 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1191 		goto bad;
1192 	*pcbopt = m;
1193 	return (0);
1194 
1195 bad:
1196 	(void)m_free(m);
1197 	return (EINVAL);
1198 }
1199 
1200 /*
1201  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1202  */
1203 static struct ifnet *
1204 ip_multicast_if(a, ifindexp)
1205 	struct in_addr *a;
1206 	int *ifindexp;
1207 {
1208 	int ifindex;
1209 	struct ifnet *ifp;
1210 
1211 	if (ifindexp)
1212 		*ifindexp = 0;
1213 	if (ntohl(a->s_addr) >> 24 == 0) {
1214 		ifindex = ntohl(a->s_addr) & 0xffffff;
1215 		if (ifindex < 0 || if_index < ifindex)
1216 			return NULL;
1217 		ifp = ifindex2ifnet[ifindex];
1218 		if (ifindexp)
1219 			*ifindexp = ifindex;
1220 	} else {
1221 		INADDR_TO_IFP(*a, ifp);
1222 	}
1223 	return ifp;
1224 }
1225 
1226 /*
1227  * Set the IP multicast options in response to user setsockopt().
1228  */
1229 int
1230 ip_setmoptions(optname, imop, m)
1231 	int optname;
1232 	struct ip_moptions **imop;
1233 	struct mbuf *m;
1234 {
1235 	int error = 0;
1236 	u_char loop;
1237 	int i;
1238 	struct in_addr addr;
1239 	struct ip_mreq *mreq;
1240 	struct ifnet *ifp;
1241 	struct ip_moptions *imo = *imop;
1242 	struct route ro;
1243 	struct sockaddr_in *dst;
1244 	int ifindex;
1245 
1246 	if (imo == NULL) {
1247 		/*
1248 		 * No multicast option buffer attached to the pcb;
1249 		 * allocate one and initialize to default values.
1250 		 */
1251 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1252 		    M_WAITOK);
1253 
1254 		if (imo == NULL)
1255 			return (ENOBUFS);
1256 		*imop = imo;
1257 		imo->imo_multicast_ifp = NULL;
1258 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1259 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1260 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1261 		imo->imo_num_memberships = 0;
1262 	}
1263 
1264 	switch (optname) {
1265 
1266 	case IP_MULTICAST_IF:
1267 		/*
1268 		 * Select the interface for outgoing multicast packets.
1269 		 */
1270 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1271 			error = EINVAL;
1272 			break;
1273 		}
1274 		addr = *(mtod(m, struct in_addr *));
1275 		/*
1276 		 * INADDR_ANY is used to remove a previous selection.
1277 		 * When no interface is selected, a default one is
1278 		 * chosen every time a multicast packet is sent.
1279 		 */
1280 		if (in_nullhost(addr)) {
1281 			imo->imo_multicast_ifp = NULL;
1282 			break;
1283 		}
1284 		/*
1285 		 * The selected interface is identified by its local
1286 		 * IP address.  Find the interface and confirm that
1287 		 * it supports multicasting.
1288 		 */
1289 		ifp = ip_multicast_if(&addr, &ifindex);
1290 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1291 			error = EADDRNOTAVAIL;
1292 			break;
1293 		}
1294 		imo->imo_multicast_ifp = ifp;
1295 		if (ifindex)
1296 			imo->imo_multicast_addr = addr;
1297 		else
1298 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1299 		break;
1300 
1301 	case IP_MULTICAST_TTL:
1302 		/*
1303 		 * Set the IP time-to-live for outgoing multicast packets.
1304 		 */
1305 		if (m == NULL || m->m_len != 1) {
1306 			error = EINVAL;
1307 			break;
1308 		}
1309 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1310 		break;
1311 
1312 	case IP_MULTICAST_LOOP:
1313 		/*
1314 		 * Set the loopback flag for outgoing multicast packets.
1315 		 * Must be zero or one.
1316 		 */
1317 		if (m == NULL || m->m_len != 1 ||
1318 		   (loop = *(mtod(m, u_char *))) > 1) {
1319 			error = EINVAL;
1320 			break;
1321 		}
1322 		imo->imo_multicast_loop = loop;
1323 		break;
1324 
1325 	case IP_ADD_MEMBERSHIP:
1326 		/*
1327 		 * Add a multicast group membership.
1328 		 * Group must be a valid IP multicast address.
1329 		 */
1330 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1331 			error = EINVAL;
1332 			break;
1333 		}
1334 		mreq = mtod(m, struct ip_mreq *);
1335 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1336 			error = EINVAL;
1337 			break;
1338 		}
1339 		/*
1340 		 * If no interface address was provided, use the interface of
1341 		 * the route to the given multicast address.
1342 		 */
1343 		if (in_nullhost(mreq->imr_interface)) {
1344 			bzero((caddr_t)&ro, sizeof(ro));
1345 			ro.ro_rt = NULL;
1346 			dst = satosin(&ro.ro_dst);
1347 			dst->sin_len = sizeof(*dst);
1348 			dst->sin_family = AF_INET;
1349 			dst->sin_addr = mreq->imr_multiaddr;
1350 			rtalloc(&ro);
1351 			if (ro.ro_rt == NULL) {
1352 				error = EADDRNOTAVAIL;
1353 				break;
1354 			}
1355 			ifp = ro.ro_rt->rt_ifp;
1356 			rtfree(ro.ro_rt);
1357 		} else {
1358 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1359 		}
1360 		/*
1361 		 * See if we found an interface, and confirm that it
1362 		 * supports multicast.
1363 		 */
1364 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1365 			error = EADDRNOTAVAIL;
1366 			break;
1367 		}
1368 		/*
1369 		 * See if the membership already exists or if all the
1370 		 * membership slots are full.
1371 		 */
1372 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1373 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1374 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1375 				      mreq->imr_multiaddr))
1376 				break;
1377 		}
1378 		if (i < imo->imo_num_memberships) {
1379 			error = EADDRINUSE;
1380 			break;
1381 		}
1382 		if (i == IP_MAX_MEMBERSHIPS) {
1383 			error = ETOOMANYREFS;
1384 			break;
1385 		}
1386 		/*
1387 		 * Everything looks good; add a new record to the multicast
1388 		 * address list for the given interface.
1389 		 */
1390 		if ((imo->imo_membership[i] =
1391 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1392 			error = ENOBUFS;
1393 			break;
1394 		}
1395 		++imo->imo_num_memberships;
1396 		break;
1397 
1398 	case IP_DROP_MEMBERSHIP:
1399 		/*
1400 		 * Drop a multicast group membership.
1401 		 * Group must be a valid IP multicast address.
1402 		 */
1403 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1404 			error = EINVAL;
1405 			break;
1406 		}
1407 		mreq = mtod(m, struct ip_mreq *);
1408 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1409 			error = EINVAL;
1410 			break;
1411 		}
1412 		/*
1413 		 * If an interface address was specified, get a pointer
1414 		 * to its ifnet structure.
1415 		 */
1416 		if (in_nullhost(mreq->imr_interface))
1417 			ifp = NULL;
1418 		else {
1419 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1420 			if (ifp == NULL) {
1421 				error = EADDRNOTAVAIL;
1422 				break;
1423 			}
1424 		}
1425 		/*
1426 		 * Find the membership in the membership array.
1427 		 */
1428 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1429 			if ((ifp == NULL ||
1430 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1431 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1432 				       mreq->imr_multiaddr))
1433 				break;
1434 		}
1435 		if (i == imo->imo_num_memberships) {
1436 			error = EADDRNOTAVAIL;
1437 			break;
1438 		}
1439 		/*
1440 		 * Give up the multicast address record to which the
1441 		 * membership points.
1442 		 */
1443 		in_delmulti(imo->imo_membership[i]);
1444 		/*
1445 		 * Remove the gap in the membership array.
1446 		 */
1447 		for (++i; i < imo->imo_num_memberships; ++i)
1448 			imo->imo_membership[i-1] = imo->imo_membership[i];
1449 		--imo->imo_num_memberships;
1450 		break;
1451 
1452 	default:
1453 		error = EOPNOTSUPP;
1454 		break;
1455 	}
1456 
1457 	/*
1458 	 * If all options have default values, no need to keep the mbuf.
1459 	 */
1460 	if (imo->imo_multicast_ifp == NULL &&
1461 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1462 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1463 	    imo->imo_num_memberships == 0) {
1464 		free(*imop, M_IPMOPTS);
1465 		*imop = NULL;
1466 	}
1467 
1468 	return (error);
1469 }
1470 
1471 /*
1472  * Return the IP multicast options in response to user getsockopt().
1473  */
1474 int
1475 ip_getmoptions(optname, imo, mp)
1476 	int optname;
1477 	struct ip_moptions *imo;
1478 	struct mbuf **mp;
1479 {
1480 	u_char *ttl;
1481 	u_char *loop;
1482 	struct in_addr *addr;
1483 	struct in_ifaddr *ia;
1484 
1485 	*mp = m_get(M_WAIT, MT_SOOPTS);
1486 
1487 	switch (optname) {
1488 
1489 	case IP_MULTICAST_IF:
1490 		addr = mtod(*mp, struct in_addr *);
1491 		(*mp)->m_len = sizeof(struct in_addr);
1492 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1493 			*addr = zeroin_addr;
1494 		else if (imo->imo_multicast_addr.s_addr) {
1495 			/* return the value user has set */
1496 			*addr = imo->imo_multicast_addr;
1497 		} else {
1498 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1499 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1500 		}
1501 		return (0);
1502 
1503 	case IP_MULTICAST_TTL:
1504 		ttl = mtod(*mp, u_char *);
1505 		(*mp)->m_len = 1;
1506 		*ttl = imo ? imo->imo_multicast_ttl
1507 			   : IP_DEFAULT_MULTICAST_TTL;
1508 		return (0);
1509 
1510 	case IP_MULTICAST_LOOP:
1511 		loop = mtod(*mp, u_char *);
1512 		(*mp)->m_len = 1;
1513 		*loop = imo ? imo->imo_multicast_loop
1514 			    : IP_DEFAULT_MULTICAST_LOOP;
1515 		return (0);
1516 
1517 	default:
1518 		return (EOPNOTSUPP);
1519 	}
1520 }
1521 
1522 /*
1523  * Discard the IP multicast options.
1524  */
1525 void
1526 ip_freemoptions(imo)
1527 	struct ip_moptions *imo;
1528 {
1529 	int i;
1530 
1531 	if (imo != NULL) {
1532 		for (i = 0; i < imo->imo_num_memberships; ++i)
1533 			in_delmulti(imo->imo_membership[i]);
1534 		free(imo, M_IPMOPTS);
1535 	}
1536 }
1537 
1538 /*
1539  * Routine called from ip_output() to loop back a copy of an IP multicast
1540  * packet to the input queue of a specified interface.  Note that this
1541  * calls the output routine of the loopback "driver", but with an interface
1542  * pointer that might NOT be &loif -- easier than replicating that code here.
1543  */
1544 static void
1545 ip_mloopback(ifp, m, dst)
1546 	struct ifnet *ifp;
1547 	struct mbuf *m;
1548 	struct sockaddr_in *dst;
1549 {
1550 	struct ip *ip;
1551 	struct mbuf *copym;
1552 
1553 	copym = m_copy(m, 0, M_COPYALL);
1554 	if (copym != NULL
1555 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1556 		copym = m_pullup(copym, sizeof(struct ip));
1557 	if (copym != NULL) {
1558 		/*
1559 		 * We don't bother to fragment if the IP length is greater
1560 		 * than the interface's MTU.  Can this possibly matter?
1561 		 */
1562 		ip = mtod(copym, struct ip *);
1563 		HTONS(ip->ip_len);
1564 		HTONS(ip->ip_off);
1565 		ip->ip_sum = 0;
1566 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1567 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1568 	}
1569 }
1570