xref: /netbsd-src/sys/netinet/ip_output.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: ip_output.c,v 1.74 2000/05/10 03:31:30 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 MROUTING
132 #include <netinet/ip_mroute.h>
133 #endif
134 
135 #ifdef __vax__
136 #include <machine/mtpr.h>
137 #endif
138 
139 #include <machine/stdarg.h>
140 
141 #ifdef IPSEC
142 #include <netinet6/ipsec.h>
143 #include <netkey/key.h>
144 #include <netkey/key_debug.h>
145 #endif /*IPSEC*/
146 
147 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
148 static void ip_mloopback
149 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
150 
151 /*
152  * IP output.  The packet in mbuf chain m contains a skeletal IP
153  * header (with len, off, ttl, proto, tos, src, dst).
154  * The mbuf chain containing the packet will be freed.
155  * The mbuf opt, if present, will not be freed.
156  */
157 int
158 #if __STDC__
159 ip_output(struct mbuf *m0, ...)
160 #else
161 ip_output(m0, va_alist)
162 	struct mbuf *m0;
163 	va_dcl
164 #endif
165 {
166 	struct ip *ip, *mhip;
167 	struct ifnet *ifp;
168 	struct mbuf *m = m0;
169 	int hlen = sizeof (struct ip);
170 	int len, off, error = 0;
171 	struct route iproute;
172 	struct sockaddr_in *dst;
173 #if IFA_STATS
174 	struct sockaddr_in src;
175 #endif
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 PFIL_HOOKS
185 	struct packet_filter_hook *pfh;
186 	struct mbuf *m1;
187 	int rv;
188 #endif /* PFIL_HOOKS */
189 #ifdef IPSEC
190 	struct socket *so;
191 	struct secpolicy *sp = NULL;
192 #endif /*IPSEC*/
193 
194 	va_start(ap, m0);
195 	opt = va_arg(ap, struct mbuf *);
196 	ro = va_arg(ap, struct route *);
197 	flags = va_arg(ap, int);
198 	imo = va_arg(ap, struct ip_moptions *);
199 	if (flags & IP_RETURNMTU)
200 		mtu_p = va_arg(ap, int *);
201 	else
202 		mtu_p = NULL;
203 	va_end(ap);
204 
205 #ifdef IPSEC
206 	so = ipsec_getsocket(m);
207 	ipsec_setsocket(m, NULL);
208 #endif /*IPSEC*/
209 
210 #ifdef	DIAGNOSTIC
211 	if ((m->m_flags & M_PKTHDR) == 0)
212 		panic("ip_output no HDR");
213 #endif
214 	if (opt) {
215 		m = ip_insertoptions(m, opt, &len);
216 		hlen = len;
217 	}
218 	ip = mtod(m, struct ip *);
219 	/*
220 	 * Fill in IP header.
221 	 */
222 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
223 		ip->ip_v = IPVERSION;
224 		ip->ip_off &= IP_DF;
225 		ip->ip_id = htons(ip_id++);
226 		ip->ip_hl = hlen >> 2;
227 		ipstat.ips_localout++;
228 	} else {
229 		hlen = ip->ip_hl << 2;
230 	}
231 	/*
232 	 * Route packet.
233 	 */
234 	if (ro == 0) {
235 		ro = &iproute;
236 		bzero((caddr_t)ro, sizeof (*ro));
237 	}
238 	dst = satosin(&ro->ro_dst);
239 	/*
240 	 * If there is a cached route,
241 	 * check that it is to the same destination
242 	 * and is still up.  If not, free it and try again.
243 	 */
244 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
245 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
246 		RTFREE(ro->ro_rt);
247 		ro->ro_rt = (struct rtentry *)0;
248 	}
249 	if (ro->ro_rt == 0) {
250 		dst->sin_family = AF_INET;
251 		dst->sin_len = sizeof(*dst);
252 		dst->sin_addr = ip->ip_dst;
253 	}
254 	/*
255 	 * If routing to interface only,
256 	 * short circuit routing lookup.
257 	 */
258 	if (flags & IP_ROUTETOIF) {
259 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
260 			ipstat.ips_noroute++;
261 			error = ENETUNREACH;
262 			goto bad;
263 		}
264 		ifp = ia->ia_ifp;
265 		mtu = ifp->if_mtu;
266 		ip->ip_ttl = 1;
267 	} else {
268 		if (ro->ro_rt == 0)
269 			rtalloc(ro);
270 		if (ro->ro_rt == 0) {
271 			ipstat.ips_noroute++;
272 			error = EHOSTUNREACH;
273 			goto bad;
274 		}
275 		ia = ifatoia(ro->ro_rt->rt_ifa);
276 		ifp = ro->ro_rt->rt_ifp;
277 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
278 			mtu = ifp->if_mtu;
279 		ro->ro_rt->rt_use++;
280 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
281 			dst = satosin(ro->ro_rt->rt_gateway);
282 	}
283 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
284 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
285 		struct in_multi *inm;
286 
287 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
288 			M_BCAST : M_MCAST;
289 		/*
290 		 * IP destination address is multicast.  Make sure "dst"
291 		 * still points to the address in "ro".  (It may have been
292 		 * changed to point to a gateway address, above.)
293 		 */
294 		dst = satosin(&ro->ro_dst);
295 		/*
296 		 * See if the caller provided any multicast options
297 		 */
298 		if (imo != NULL) {
299 			ip->ip_ttl = imo->imo_multicast_ttl;
300 			if (imo->imo_multicast_ifp != NULL) {
301 				ifp = imo->imo_multicast_ifp;
302 				mtu = ifp->if_mtu;
303 			}
304 		} else
305 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
306 		/*
307 		 * Confirm that the outgoing interface supports multicast.
308 		 */
309 		if (((m->m_flags & M_MCAST) &&
310 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
311 		    ((m->m_flags & M_BCAST) &&
312 		     (ifp->if_flags & IFF_BROADCAST) == 0))  {
313 			ipstat.ips_noroute++;
314 			error = ENETUNREACH;
315 			goto bad;
316 		}
317 		/*
318 		 * If source address not specified yet, use an address
319 		 * of outgoing interface.
320 		 */
321 		if (in_nullhost(ip->ip_src)) {
322 			struct in_ifaddr *ia;
323 
324 			IFP_TO_IA(ifp, ia);
325 			ip->ip_src = ia->ia_addr.sin_addr;
326 		}
327 
328 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
329 		if (inm != NULL &&
330 		   (imo == NULL || imo->imo_multicast_loop)) {
331 			/*
332 			 * If we belong to the destination multicast group
333 			 * on the outgoing interface, and the caller did not
334 			 * forbid loopback, loop back a copy.
335 			 */
336 			ip_mloopback(ifp, m, dst);
337 		}
338 #ifdef MROUTING
339 		else {
340 			/*
341 			 * If we are acting as a multicast router, perform
342 			 * multicast forwarding as if the packet had just
343 			 * arrived on the interface to which we are about
344 			 * to send.  The multicast forwarding function
345 			 * recursively calls this function, using the
346 			 * IP_FORWARDING flag to prevent infinite recursion.
347 			 *
348 			 * Multicasts that are looped back by ip_mloopback(),
349 			 * above, will be forwarded by the ip_input() routine,
350 			 * if necessary.
351 			 */
352 			extern struct socket *ip_mrouter;
353 
354 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
355 				if (ip_mforward(m, ifp) != 0) {
356 					m_freem(m);
357 					goto done;
358 				}
359 			}
360 		}
361 #endif
362 		/*
363 		 * Multicasts with a time-to-live of zero may be looped-
364 		 * back, above, but must not be transmitted on a network.
365 		 * Also, multicasts addressed to the loopback interface
366 		 * are not sent -- the above call to ip_mloopback() will
367 		 * loop back a copy if this host actually belongs to the
368 		 * destination group on the loopback interface.
369 		 */
370 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
371 			m_freem(m);
372 			goto done;
373 		}
374 
375 		goto sendit;
376 	}
377 #ifndef notdef
378 	/*
379 	 * If source address not specified yet, use address
380 	 * of outgoing interface.
381 	 */
382 	if (in_nullhost(ip->ip_src))
383 		ip->ip_src = ia->ia_addr.sin_addr;
384 #endif
385 
386 	/*
387 	 * packets with Class-D address as source are not valid per
388 	 * RFC 1112
389 	 */
390 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
391 		ipstat.ips_odropped++;
392 		error = EADDRNOTAVAIL;
393 		goto bad;
394 	}
395 
396 	/*
397 	 * Look for broadcast address and
398 	 * and verify user is allowed to send
399 	 * such a packet.
400 	 */
401 	if (in_broadcast(dst->sin_addr, ifp)) {
402 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
403 			error = EADDRNOTAVAIL;
404 			goto bad;
405 		}
406 		if ((flags & IP_ALLOWBROADCAST) == 0) {
407 			error = EACCES;
408 			goto bad;
409 		}
410 		/* don't allow broadcast messages to be fragmented */
411 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
412 			error = EMSGSIZE;
413 			goto bad;
414 		}
415 		m->m_flags |= M_BCAST;
416 	} else
417 		m->m_flags &= ~M_BCAST;
418 
419 sendit:
420 #ifdef PFIL_HOOKS
421 	/*
422 	 * Run through list of hooks for output packets.
423 	 */
424 	m1 = m;
425 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
426 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
427 		if (pfh->pfil_func) {
428 		    	rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
429 			if (rv) {
430 				error = EHOSTUNREACH;
431 				goto done;
432 			}
433 			m = m1;
434 			if (m == NULL)
435 				goto done;
436 			ip = mtod(m, struct ip *);
437 		}
438 #endif /* PFIL_HOOKS */
439 
440 #ifdef IPSEC
441 	/* get SP for this packet */
442 	if (so == NULL)
443 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
444 	else
445 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
446 
447 	if (sp == NULL) {
448 		ipsecstat.out_inval++;
449 		goto bad;
450 	}
451 
452 	error = 0;
453 
454 	/* check policy */
455 	switch (sp->policy) {
456 	case IPSEC_POLICY_DISCARD:
457 		/*
458 		 * This packet is just discarded.
459 		 */
460 		ipsecstat.out_polvio++;
461 		goto bad;
462 
463 	case IPSEC_POLICY_BYPASS:
464 	case IPSEC_POLICY_NONE:
465 		/* no need to do IPsec. */
466 		goto skip_ipsec;
467 
468 	case IPSEC_POLICY_IPSEC:
469 		if (sp->req == NULL) {
470 			/* XXX should be panic ? */
471 			printf("ip_output: No IPsec request specified.\n");
472 			error = EINVAL;
473 			goto bad;
474 		}
475 		break;
476 
477 	case IPSEC_POLICY_ENTRUST:
478 	default:
479 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
480 	}
481 
482 	ip->ip_len = htons((u_short)ip->ip_len);
483 	ip->ip_off = htons((u_short)ip->ip_off);
484 	ip->ip_sum = 0;
485 
486     {
487 	struct ipsec_output_state state;
488 	bzero(&state, sizeof(state));
489 	state.m = m;
490 	if (flags & IP_ROUTETOIF) {
491 		state.ro = &iproute;
492 		bzero(&iproute, sizeof(iproute));
493 	} else
494 		state.ro = ro;
495 	state.dst = (struct sockaddr *)dst;
496 
497 	error = ipsec4_output(&state, sp, flags);
498 
499 	m = state.m;
500 	if (flags & IP_ROUTETOIF) {
501 		/*
502 		 * if we have tunnel mode SA, we may need to ignore
503 		 * IP_ROUTETOIF.
504 		 */
505 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
506 			flags &= ~IP_ROUTETOIF;
507 			ro = state.ro;
508 		}
509 	} else
510 		ro = state.ro;
511 	dst = (struct sockaddr_in *)state.dst;
512 	if (error) {
513 		/* mbuf is already reclaimed in ipsec4_output. */
514 		m0 = NULL;
515 		switch (error) {
516 		case EHOSTUNREACH:
517 		case ENETUNREACH:
518 		case EMSGSIZE:
519 		case ENOBUFS:
520 		case ENOMEM:
521 			break;
522 		default:
523 			printf("ip4_output (ipsec): error code %d\n", error);
524 			/*fall through*/
525 		case ENOENT:
526 			/* don't show these error codes to the user */
527 			error = 0;
528 			break;
529 		}
530 		goto bad;
531 	}
532     }
533 
534 	/* be sure to update variables that are affected by ipsec4_output() */
535 	ip = mtod(m, struct ip *);
536 #ifdef _IP_VHL
537 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
538 #else
539 	hlen = ip->ip_hl << 2;
540 #endif
541 	if (ro->ro_rt == NULL) {
542 		if ((flags & IP_ROUTETOIF) == 0) {
543 			printf("ip_output: "
544 				"can't update route after IPsec processing\n");
545 			error = EHOSTUNREACH;	/*XXX*/
546 			goto bad;
547 		}
548 	} else {
549 		/* nobody uses ia beyond here */
550 		ifp = ro->ro_rt->rt_ifp;
551 	}
552 
553 	/* make it flipped, again. */
554 	ip->ip_len = ntohs((u_short)ip->ip_len);
555 	ip->ip_off = ntohs((u_short)ip->ip_off);
556 skip_ipsec:
557 #endif /*IPSEC*/
558 
559 	/*
560 	 * If small enough for mtu of path, can just send directly.
561 	 */
562 	if ((u_int16_t)ip->ip_len <= mtu) {
563 #if IFA_STATS
564 		/*
565 		 * search for the source address structure to
566 		 * maintain output statistics.
567 		 */
568 		bzero((caddr_t*) &src, sizeof(src));
569 		src.sin_family = AF_INET;
570 		src.sin_addr.s_addr = ip->ip_src.s_addr;
571 		src.sin_len = sizeof(src);
572 		ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
573 		if (ia)
574 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
575 #endif
576 		HTONS(ip->ip_len);
577 		HTONS(ip->ip_off);
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 #if 0
589 	/*
590 	 * If IPsec packet is too big for the interface, try fragment it.
591 	 * XXX This really is a quickhack.  May be inappropriate.
592 	 * XXX fails if somebody is sending AH'ed packet, with:
593 	 *	sizeof(packet without AH) < mtu < sizeof(packet with AH)
594 	 */
595 	if (sab && ip->ip_p != IPPROTO_AH && (flags & IP_FORWARDING) == 0)
596 		ip->ip_off &= ~IP_DF;
597 #endif /*IPSEC*/
598 	if (ip->ip_off & IP_DF) {
599 		if (flags & IP_RETURNMTU)
600 			*mtu_p = mtu;
601 		error = EMSGSIZE;
602 		ipstat.ips_cantfrag++;
603 		goto bad;
604 	}
605 	len = (mtu - hlen) &~ 7;
606 	if (len < 8) {
607 		error = EMSGSIZE;
608 		goto bad;
609 	}
610 
611     {
612 	int mhlen, firstlen = len;
613 	struct mbuf **mnext = &m->m_nextpkt;
614 	int fragments = 0;
615 	int s;
616 
617 	/*
618 	 * Loop through length of segment after first fragment,
619 	 * make new header and copy data of each part and link onto chain.
620 	 */
621 	m0 = m;
622 	mhlen = sizeof (struct ip);
623 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
624 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
625 		if (m == 0) {
626 			error = ENOBUFS;
627 			ipstat.ips_odropped++;
628 			goto sendorfree;
629 		}
630 		*mnext = m;
631 		mnext = &m->m_nextpkt;
632 		m->m_data += max_linkhdr;
633 		mhip = mtod(m, struct ip *);
634 		*mhip = *ip;
635 		/* we must inherit MCAST and BCAST flags */
636 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
637 		if (hlen > sizeof (struct ip)) {
638 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
639 			mhip->ip_hl = mhlen >> 2;
640 		}
641 		m->m_len = mhlen;
642 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
643 		if (ip->ip_off & IP_MF)
644 			mhip->ip_off |= IP_MF;
645 		if (off + len >= (u_int16_t)ip->ip_len)
646 			len = (u_int16_t)ip->ip_len - off;
647 		else
648 			mhip->ip_off |= IP_MF;
649 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
650 		m->m_next = m_copy(m0, off, len);
651 		if (m->m_next == 0) {
652 			error = ENOBUFS;	/* ??? */
653 			ipstat.ips_odropped++;
654 			goto sendorfree;
655 		}
656 		m->m_pkthdr.len = mhlen + len;
657 		m->m_pkthdr.rcvif = (struct ifnet *)0;
658 		HTONS(mhip->ip_off);
659 		mhip->ip_sum = 0;
660 		mhip->ip_sum = in_cksum(m, mhlen);
661 		ipstat.ips_ofragments++;
662 		fragments++;
663 	}
664 	/*
665 	 * Update first fragment by trimming what's been copied out
666 	 * and updating header, then send each fragment (in order).
667 	 */
668 	m = m0;
669 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
670 	m->m_pkthdr.len = hlen + firstlen;
671 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
672 	ip->ip_off |= IP_MF;
673 	HTONS(ip->ip_off);
674 	ip->ip_sum = 0;
675 	ip->ip_sum = in_cksum(m, hlen);
676 sendorfree:
677 	/*
678 	 * If there is no room for all the fragments, don't queue
679 	 * any of them.
680 	 */
681 	s = splimp();
682 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
683 		error = ENOBUFS;
684 	splx(s);
685 	for (m = m0; m; m = m0) {
686 		m0 = m->m_nextpkt;
687 		m->m_nextpkt = 0;
688 		if (error == 0) {
689 #if IFA_STATS
690 			/*
691 			 * search for the source address structure to
692 			 * maintain output statistics.
693 			 */
694 			bzero((caddr_t*) &src, sizeof(src));
695 			src.sin_family = AF_INET;
696 			src.sin_addr.s_addr = ip->ip_src.s_addr;
697 			src.sin_len = sizeof(src);
698 			ia = ifatoia(ifa_ifwithladdr(sintosa(&src)));
699 			if (ia) {
700 				ia->ia_ifa.ifa_data.ifad_outbytes +=
701 					ntohs(ip->ip_len);
702 			}
703 #endif
704 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
705 			    ro->ro_rt);
706 		} else
707 			m_freem(m);
708 	}
709 
710 	if (error == 0)
711 		ipstat.ips_fragmented++;
712     }
713 done:
714 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
715 		RTFREE(ro->ro_rt);
716 		ro->ro_rt = 0;
717 	}
718 
719 #ifdef IPSEC
720 	if (sp != NULL) {
721 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
722 			printf("DP ip_output call free SP:%p\n", sp));
723 		key_freesp(sp);
724 	}
725 #endif /* IPSEC */
726 
727 	return (error);
728 bad:
729 	m_freem(m);
730 	goto done;
731 }
732 
733 /*
734  * Determine the maximum length of the options to be inserted;
735  * we would far rather allocate too much space rather than too little.
736  */
737 
738 u_int
739 ip_optlen(inp)
740 	struct inpcb *inp;
741 {
742 	struct mbuf *m = inp->inp_options;
743 
744 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
745 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
746 	else
747 		return 0;
748 }
749 
750 
751 /*
752  * Insert IP options into preformed packet.
753  * Adjust IP destination as required for IP source routing,
754  * as indicated by a non-zero in_addr at the start of the options.
755  */
756 static struct mbuf *
757 ip_insertoptions(m, opt, phlen)
758 	struct mbuf *m;
759 	struct mbuf *opt;
760 	int *phlen;
761 {
762 	struct ipoption *p = mtod(opt, struct ipoption *);
763 	struct mbuf *n;
764 	struct ip *ip = mtod(m, struct ip *);
765 	unsigned optlen;
766 
767 	optlen = opt->m_len - sizeof(p->ipopt_dst);
768 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
769 		return (m);		/* XXX should fail */
770 	if (!in_nullhost(p->ipopt_dst))
771 		ip->ip_dst = p->ipopt_dst;
772 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
773 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
774 		if (n == 0)
775 			return (m);
776 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
777 		m->m_len -= sizeof(struct ip);
778 		m->m_data += sizeof(struct ip);
779 		n->m_next = m;
780 		m = n;
781 		m->m_len = optlen + sizeof(struct ip);
782 		m->m_data += max_linkhdr;
783 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
784 	} else {
785 		m->m_data -= optlen;
786 		m->m_len += optlen;
787 		m->m_pkthdr.len += optlen;
788 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
789 	}
790 	ip = mtod(m, struct ip *);
791 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
792 	*phlen = sizeof(struct ip) + optlen;
793 	ip->ip_len += optlen;
794 	return (m);
795 }
796 
797 /*
798  * Copy options from ip to jp,
799  * omitting those not copied during fragmentation.
800  */
801 int
802 ip_optcopy(ip, jp)
803 	struct ip *ip, *jp;
804 {
805 	u_char *cp, *dp;
806 	int opt, optlen, cnt;
807 
808 	cp = (u_char *)(ip + 1);
809 	dp = (u_char *)(jp + 1);
810 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
811 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
812 		opt = cp[0];
813 		if (opt == IPOPT_EOL)
814 			break;
815 		if (opt == IPOPT_NOP) {
816 			/* Preserve for IP mcast tunnel's LSRR alignment. */
817 			*dp++ = IPOPT_NOP;
818 			optlen = 1;
819 			continue;
820 		}
821 #ifdef DIAGNOSTIC
822 		if (cnt < IPOPT_OLEN + sizeof(*cp))
823 			panic("malformed IPv4 option passed to ip_optcopy");
824 #endif
825 		optlen = cp[IPOPT_OLEN];
826 #ifdef DIAGNOSTIC
827 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
828 			panic("malformed IPv4 option passed to ip_optcopy");
829 #endif
830 		/* bogus lengths should have been caught by ip_dooptions */
831 		if (optlen > cnt)
832 			optlen = cnt;
833 		if (IPOPT_COPIED(opt)) {
834 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
835 			dp += optlen;
836 		}
837 	}
838 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
839 		*dp++ = IPOPT_EOL;
840 	return (optlen);
841 }
842 
843 /*
844  * IP socket option processing.
845  */
846 int
847 ip_ctloutput(op, so, level, optname, mp)
848 	int op;
849 	struct socket *so;
850 	int level, optname;
851 	struct mbuf **mp;
852 {
853 	struct inpcb *inp = sotoinpcb(so);
854 	struct mbuf *m = *mp;
855 	int optval = 0;
856 	int error = 0;
857 #ifdef IPSEC
858 #ifdef __NetBSD__
859 	struct proc *p = curproc;	/*XXX*/
860 #endif
861 #endif
862 
863 	if (level != IPPROTO_IP) {
864 		error = EINVAL;
865 		if (op == PRCO_SETOPT && *mp)
866 			(void) m_free(*mp);
867 	} else switch (op) {
868 
869 	case PRCO_SETOPT:
870 		switch (optname) {
871 		case IP_OPTIONS:
872 #ifdef notyet
873 		case IP_RETOPTS:
874 			return (ip_pcbopts(optname, &inp->inp_options, m));
875 #else
876 			return (ip_pcbopts(&inp->inp_options, m));
877 #endif
878 
879 		case IP_TOS:
880 		case IP_TTL:
881 		case IP_RECVOPTS:
882 		case IP_RECVRETOPTS:
883 		case IP_RECVDSTADDR:
884 		case IP_RECVIF:
885 			if (m == NULL || m->m_len != sizeof(int))
886 				error = EINVAL;
887 			else {
888 				optval = *mtod(m, int *);
889 				switch (optname) {
890 
891 				case IP_TOS:
892 					inp->inp_ip.ip_tos = optval;
893 					break;
894 
895 				case IP_TTL:
896 					inp->inp_ip.ip_ttl = optval;
897 					break;
898 #define	OPTSET(bit) \
899 	if (optval) \
900 		inp->inp_flags |= bit; \
901 	else \
902 		inp->inp_flags &= ~bit;
903 
904 				case IP_RECVOPTS:
905 					OPTSET(INP_RECVOPTS);
906 					break;
907 
908 				case IP_RECVRETOPTS:
909 					OPTSET(INP_RECVRETOPTS);
910 					break;
911 
912 				case IP_RECVDSTADDR:
913 					OPTSET(INP_RECVDSTADDR);
914 					break;
915 
916 				case IP_RECVIF:
917 					OPTSET(INP_RECVIF);
918 					break;
919 				}
920 			}
921 			break;
922 #undef OPTSET
923 
924 		case IP_MULTICAST_IF:
925 		case IP_MULTICAST_TTL:
926 		case IP_MULTICAST_LOOP:
927 		case IP_ADD_MEMBERSHIP:
928 		case IP_DROP_MEMBERSHIP:
929 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
930 			break;
931 
932 		case IP_PORTRANGE:
933 			if (m == 0 || m->m_len != sizeof(int))
934 				error = EINVAL;
935 			else {
936 				optval = *mtod(m, int *);
937 
938 				switch (optval) {
939 
940 				case IP_PORTRANGE_DEFAULT:
941 				case IP_PORTRANGE_HIGH:
942 					inp->inp_flags &= ~(INP_LOWPORT);
943 					break;
944 
945 				case IP_PORTRANGE_LOW:
946 					inp->inp_flags |= INP_LOWPORT;
947 					break;
948 
949 				default:
950 					error = EINVAL;
951 					break;
952 				}
953 			}
954 			break;
955 
956 #ifdef IPSEC
957 		case IP_IPSEC_POLICY:
958 		{
959 			caddr_t req = NULL;
960 			size_t len = 0;
961 			int priv = 0;
962 
963 #ifdef __NetBSD__
964 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
965 				priv = 0;
966 			else
967 				priv = 1;
968 #else
969 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
970 #endif
971 			if (m) {
972 				req = mtod(m, caddr_t);
973 				len = m->m_len;
974 			}
975 			error = ipsec4_set_policy(inp, optname, req, len, priv);
976 			break;
977 		    }
978 #endif /*IPSEC*/
979 
980 		default:
981 			error = ENOPROTOOPT;
982 			break;
983 		}
984 		if (m)
985 			(void)m_free(m);
986 		break;
987 
988 	case PRCO_GETOPT:
989 		switch (optname) {
990 		case IP_OPTIONS:
991 		case IP_RETOPTS:
992 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
993 			if (inp->inp_options) {
994 				m->m_len = inp->inp_options->m_len;
995 				bcopy(mtod(inp->inp_options, caddr_t),
996 				    mtod(m, caddr_t), (unsigned)m->m_len);
997 			} else
998 				m->m_len = 0;
999 			break;
1000 
1001 		case IP_TOS:
1002 		case IP_TTL:
1003 		case IP_RECVOPTS:
1004 		case IP_RECVRETOPTS:
1005 		case IP_RECVDSTADDR:
1006 		case IP_RECVIF:
1007 		case IP_ERRORMTU:
1008 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1009 			m->m_len = sizeof(int);
1010 			switch (optname) {
1011 
1012 			case IP_TOS:
1013 				optval = inp->inp_ip.ip_tos;
1014 				break;
1015 
1016 			case IP_TTL:
1017 				optval = inp->inp_ip.ip_ttl;
1018 				break;
1019 
1020 			case IP_ERRORMTU:
1021 				optval = inp->inp_errormtu;
1022 				break;
1023 
1024 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1025 
1026 			case IP_RECVOPTS:
1027 				optval = OPTBIT(INP_RECVOPTS);
1028 				break;
1029 
1030 			case IP_RECVRETOPTS:
1031 				optval = OPTBIT(INP_RECVRETOPTS);
1032 				break;
1033 
1034 			case IP_RECVDSTADDR:
1035 				optval = OPTBIT(INP_RECVDSTADDR);
1036 				break;
1037 
1038 			case IP_RECVIF:
1039 				optval = OPTBIT(INP_RECVIF);
1040 				break;
1041 			}
1042 			*mtod(m, int *) = optval;
1043 			break;
1044 
1045 #ifdef IPSEC
1046 		case IP_IPSEC_POLICY:
1047 		{
1048 			caddr_t req = NULL;
1049 			size_t len;
1050 
1051 			if (m) {
1052 				req = mtod(m, caddr_t);
1053 				len = m->m_len;
1054 			}
1055 			error = ipsec4_get_policy(inp, req, len, mp);
1056 			break;
1057 		}
1058 #endif /*IPSEC*/
1059 
1060 		case IP_MULTICAST_IF:
1061 		case IP_MULTICAST_TTL:
1062 		case IP_MULTICAST_LOOP:
1063 		case IP_ADD_MEMBERSHIP:
1064 		case IP_DROP_MEMBERSHIP:
1065 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1066 			break;
1067 
1068 		case IP_PORTRANGE:
1069 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1070 			m->m_len = sizeof(int);
1071 
1072 			if (inp->inp_flags & INP_LOWPORT)
1073 				optval = IP_PORTRANGE_LOW;
1074 			else
1075 				optval = IP_PORTRANGE_DEFAULT;
1076 
1077 			*mtod(m, int *) = optval;
1078 			break;
1079 
1080 		default:
1081 			error = ENOPROTOOPT;
1082 			break;
1083 		}
1084 		break;
1085 	}
1086 	return (error);
1087 }
1088 
1089 /*
1090  * Set up IP options in pcb for insertion in output packets.
1091  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1092  * with destination address if source routed.
1093  */
1094 int
1095 #ifdef notyet
1096 ip_pcbopts(optname, pcbopt, m)
1097 	int optname;
1098 #else
1099 ip_pcbopts(pcbopt, m)
1100 #endif
1101 	struct mbuf **pcbopt;
1102 	struct mbuf *m;
1103 {
1104 	int cnt, optlen;
1105 	u_char *cp;
1106 	u_char opt;
1107 
1108 	/* turn off any old options */
1109 	if (*pcbopt)
1110 		(void)m_free(*pcbopt);
1111 	*pcbopt = 0;
1112 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1113 		/*
1114 		 * Only turning off any previous options.
1115 		 */
1116 		if (m)
1117 			(void)m_free(m);
1118 		return (0);
1119 	}
1120 
1121 #ifndef	vax
1122 	if (m->m_len % sizeof(int32_t))
1123 		goto bad;
1124 #endif
1125 	/*
1126 	 * IP first-hop destination address will be stored before
1127 	 * actual options; move other options back
1128 	 * and clear it when none present.
1129 	 */
1130 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1131 		goto bad;
1132 	cnt = m->m_len;
1133 	m->m_len += sizeof(struct in_addr);
1134 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1135 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1136 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1137 
1138 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1139 		opt = cp[IPOPT_OPTVAL];
1140 		if (opt == IPOPT_EOL)
1141 			break;
1142 		if (opt == IPOPT_NOP)
1143 			optlen = 1;
1144 		else {
1145 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1146 				goto bad;
1147 			optlen = cp[IPOPT_OLEN];
1148 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1149 				goto bad;
1150 		}
1151 		switch (opt) {
1152 
1153 		default:
1154 			break;
1155 
1156 		case IPOPT_LSRR:
1157 		case IPOPT_SSRR:
1158 			/*
1159 			 * user process specifies route as:
1160 			 *	->A->B->C->D
1161 			 * D must be our final destination (but we can't
1162 			 * check that since we may not have connected yet).
1163 			 * A is first hop destination, which doesn't appear in
1164 			 * actual IP option, but is stored before the options.
1165 			 */
1166 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1167 				goto bad;
1168 			m->m_len -= sizeof(struct in_addr);
1169 			cnt -= sizeof(struct in_addr);
1170 			optlen -= sizeof(struct in_addr);
1171 			cp[IPOPT_OLEN] = optlen;
1172 			/*
1173 			 * Move first hop before start of options.
1174 			 */
1175 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1176 			    sizeof(struct in_addr));
1177 			/*
1178 			 * Then copy rest of options back
1179 			 * to close up the deleted entry.
1180 			 */
1181 			memmove(&cp[IPOPT_OFFSET+1],
1182                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1183 			    (unsigned)cnt + sizeof(struct in_addr));
1184 			break;
1185 		}
1186 	}
1187 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1188 		goto bad;
1189 	*pcbopt = m;
1190 	return (0);
1191 
1192 bad:
1193 	(void)m_free(m);
1194 	return (EINVAL);
1195 }
1196 
1197 /*
1198  * Set the IP multicast options in response to user setsockopt().
1199  */
1200 int
1201 ip_setmoptions(optname, imop, m)
1202 	int optname;
1203 	struct ip_moptions **imop;
1204 	struct mbuf *m;
1205 {
1206 	int error = 0;
1207 	u_char loop;
1208 	int i;
1209 	struct in_addr addr;
1210 	struct ip_mreq *mreq;
1211 	struct ifnet *ifp;
1212 	struct ip_moptions *imo = *imop;
1213 	struct route ro;
1214 	struct sockaddr_in *dst;
1215 
1216 	if (imo == NULL) {
1217 		/*
1218 		 * No multicast option buffer attached to the pcb;
1219 		 * allocate one and initialize to default values.
1220 		 */
1221 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1222 		    M_WAITOK);
1223 
1224 		if (imo == NULL)
1225 			return (ENOBUFS);
1226 		*imop = imo;
1227 		imo->imo_multicast_ifp = NULL;
1228 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1229 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1230 		imo->imo_num_memberships = 0;
1231 	}
1232 
1233 	switch (optname) {
1234 
1235 	case IP_MULTICAST_IF:
1236 		/*
1237 		 * Select the interface for outgoing multicast packets.
1238 		 */
1239 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1240 			error = EINVAL;
1241 			break;
1242 		}
1243 		addr = *(mtod(m, struct in_addr *));
1244 		/*
1245 		 * INADDR_ANY is used to remove a previous selection.
1246 		 * When no interface is selected, a default one is
1247 		 * chosen every time a multicast packet is sent.
1248 		 */
1249 		if (in_nullhost(addr)) {
1250 			imo->imo_multicast_ifp = NULL;
1251 			break;
1252 		}
1253 		/*
1254 		 * The selected interface is identified by its local
1255 		 * IP address.  Find the interface and confirm that
1256 		 * it supports multicasting.
1257 		 */
1258 		INADDR_TO_IFP(addr, ifp);
1259 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1260 			error = EADDRNOTAVAIL;
1261 			break;
1262 		}
1263 		imo->imo_multicast_ifp = ifp;
1264 		break;
1265 
1266 	case IP_MULTICAST_TTL:
1267 		/*
1268 		 * Set the IP time-to-live for outgoing multicast packets.
1269 		 */
1270 		if (m == NULL || m->m_len != 1) {
1271 			error = EINVAL;
1272 			break;
1273 		}
1274 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1275 		break;
1276 
1277 	case IP_MULTICAST_LOOP:
1278 		/*
1279 		 * Set the loopback flag for outgoing multicast packets.
1280 		 * Must be zero or one.
1281 		 */
1282 		if (m == NULL || m->m_len != 1 ||
1283 		   (loop = *(mtod(m, u_char *))) > 1) {
1284 			error = EINVAL;
1285 			break;
1286 		}
1287 		imo->imo_multicast_loop = loop;
1288 		break;
1289 
1290 	case IP_ADD_MEMBERSHIP:
1291 		/*
1292 		 * Add a multicast group membership.
1293 		 * Group must be a valid IP multicast address.
1294 		 */
1295 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1296 			error = EINVAL;
1297 			break;
1298 		}
1299 		mreq = mtod(m, struct ip_mreq *);
1300 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1301 			error = EINVAL;
1302 			break;
1303 		}
1304 		/*
1305 		 * If no interface address was provided, use the interface of
1306 		 * the route to the given multicast address.
1307 		 */
1308 		if (in_nullhost(mreq->imr_interface)) {
1309 			bzero((caddr_t)&ro, sizeof(ro));
1310 			ro.ro_rt = NULL;
1311 			dst = satosin(&ro.ro_dst);
1312 			dst->sin_len = sizeof(*dst);
1313 			dst->sin_family = AF_INET;
1314 			dst->sin_addr = mreq->imr_multiaddr;
1315 			rtalloc(&ro);
1316 			if (ro.ro_rt == NULL) {
1317 				error = EADDRNOTAVAIL;
1318 				break;
1319 			}
1320 			ifp = ro.ro_rt->rt_ifp;
1321 			rtfree(ro.ro_rt);
1322 		} else {
1323 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1324 		}
1325 		/*
1326 		 * See if we found an interface, and confirm that it
1327 		 * supports multicast.
1328 		 */
1329 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1330 			error = EADDRNOTAVAIL;
1331 			break;
1332 		}
1333 		/*
1334 		 * See if the membership already exists or if all the
1335 		 * membership slots are full.
1336 		 */
1337 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1338 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1339 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1340 				      mreq->imr_multiaddr))
1341 				break;
1342 		}
1343 		if (i < imo->imo_num_memberships) {
1344 			error = EADDRINUSE;
1345 			break;
1346 		}
1347 		if (i == IP_MAX_MEMBERSHIPS) {
1348 			error = ETOOMANYREFS;
1349 			break;
1350 		}
1351 		/*
1352 		 * Everything looks good; add a new record to the multicast
1353 		 * address list for the given interface.
1354 		 */
1355 		if ((imo->imo_membership[i] =
1356 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1357 			error = ENOBUFS;
1358 			break;
1359 		}
1360 		++imo->imo_num_memberships;
1361 		break;
1362 
1363 	case IP_DROP_MEMBERSHIP:
1364 		/*
1365 		 * Drop a multicast group membership.
1366 		 * Group must be a valid IP multicast address.
1367 		 */
1368 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1369 			error = EINVAL;
1370 			break;
1371 		}
1372 		mreq = mtod(m, struct ip_mreq *);
1373 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1374 			error = EINVAL;
1375 			break;
1376 		}
1377 		/*
1378 		 * If an interface address was specified, get a pointer
1379 		 * to its ifnet structure.
1380 		 */
1381 		if (in_nullhost(mreq->imr_interface))
1382 			ifp = NULL;
1383 		else {
1384 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1385 			if (ifp == NULL) {
1386 				error = EADDRNOTAVAIL;
1387 				break;
1388 			}
1389 		}
1390 		/*
1391 		 * Find the membership in the membership array.
1392 		 */
1393 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1394 			if ((ifp == NULL ||
1395 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1396 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1397 				       mreq->imr_multiaddr))
1398 				break;
1399 		}
1400 		if (i == imo->imo_num_memberships) {
1401 			error = EADDRNOTAVAIL;
1402 			break;
1403 		}
1404 		/*
1405 		 * Give up the multicast address record to which the
1406 		 * membership points.
1407 		 */
1408 		in_delmulti(imo->imo_membership[i]);
1409 		/*
1410 		 * Remove the gap in the membership array.
1411 		 */
1412 		for (++i; i < imo->imo_num_memberships; ++i)
1413 			imo->imo_membership[i-1] = imo->imo_membership[i];
1414 		--imo->imo_num_memberships;
1415 		break;
1416 
1417 	default:
1418 		error = EOPNOTSUPP;
1419 		break;
1420 	}
1421 
1422 	/*
1423 	 * If all options have default values, no need to keep the mbuf.
1424 	 */
1425 	if (imo->imo_multicast_ifp == NULL &&
1426 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1427 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1428 	    imo->imo_num_memberships == 0) {
1429 		free(*imop, M_IPMOPTS);
1430 		*imop = NULL;
1431 	}
1432 
1433 	return (error);
1434 }
1435 
1436 /*
1437  * Return the IP multicast options in response to user getsockopt().
1438  */
1439 int
1440 ip_getmoptions(optname, imo, mp)
1441 	int optname;
1442 	struct ip_moptions *imo;
1443 	struct mbuf **mp;
1444 {
1445 	u_char *ttl;
1446 	u_char *loop;
1447 	struct in_addr *addr;
1448 	struct in_ifaddr *ia;
1449 
1450 	*mp = m_get(M_WAIT, MT_SOOPTS);
1451 
1452 	switch (optname) {
1453 
1454 	case IP_MULTICAST_IF:
1455 		addr = mtod(*mp, struct in_addr *);
1456 		(*mp)->m_len = sizeof(struct in_addr);
1457 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1458 			*addr = zeroin_addr;
1459 		else {
1460 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1461 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1462 		}
1463 		return (0);
1464 
1465 	case IP_MULTICAST_TTL:
1466 		ttl = mtod(*mp, u_char *);
1467 		(*mp)->m_len = 1;
1468 		*ttl = imo ? imo->imo_multicast_ttl
1469 			   : IP_DEFAULT_MULTICAST_TTL;
1470 		return (0);
1471 
1472 	case IP_MULTICAST_LOOP:
1473 		loop = mtod(*mp, u_char *);
1474 		(*mp)->m_len = 1;
1475 		*loop = imo ? imo->imo_multicast_loop
1476 			    : IP_DEFAULT_MULTICAST_LOOP;
1477 		return (0);
1478 
1479 	default:
1480 		return (EOPNOTSUPP);
1481 	}
1482 }
1483 
1484 /*
1485  * Discard the IP multicast options.
1486  */
1487 void
1488 ip_freemoptions(imo)
1489 	struct ip_moptions *imo;
1490 {
1491 	int i;
1492 
1493 	if (imo != NULL) {
1494 		for (i = 0; i < imo->imo_num_memberships; ++i)
1495 			in_delmulti(imo->imo_membership[i]);
1496 		free(imo, M_IPMOPTS);
1497 	}
1498 }
1499 
1500 /*
1501  * Routine called from ip_output() to loop back a copy of an IP multicast
1502  * packet to the input queue of a specified interface.  Note that this
1503  * calls the output routine of the loopback "driver", but with an interface
1504  * pointer that might NOT be &loif -- easier than replicating that code here.
1505  */
1506 static void
1507 ip_mloopback(ifp, m, dst)
1508 	struct ifnet *ifp;
1509 	struct mbuf *m;
1510 	struct sockaddr_in *dst;
1511 {
1512 	struct ip *ip;
1513 	struct mbuf *copym;
1514 
1515 	copym = m_copy(m, 0, M_COPYALL);
1516 	if (copym != NULL
1517 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1518 		copym = m_pullup(copym, sizeof(struct ip));
1519 	if (copym != NULL) {
1520 		/*
1521 		 * We don't bother to fragment if the IP length is greater
1522 		 * than the interface's MTU.  Can this possibly matter?
1523 		 */
1524 		ip = mtod(copym, struct ip *);
1525 		HTONS(ip->ip_len);
1526 		HTONS(ip->ip_off);
1527 		ip->ip_sum = 0;
1528 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1529 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1530 	}
1531 }
1532