xref: /original-bsd/sys/netinet/ip_output.c (revision 6471873a)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	Regents of the University of California.  All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)ip_output.c	8.1 (Berkeley) 06/10/93
8  */
9 
10 #include <sys/param.h>
11 #include <sys/malloc.h>
12 #include <sys/mbuf.h>
13 #include <sys/errno.h>
14 #include <sys/protosw.h>
15 #include <sys/socket.h>
16 #include <sys/socketvar.h>
17 
18 #include <net/if.h>
19 #include <net/route.h>
20 
21 #include <netinet/in.h>
22 #include <netinet/in_systm.h>
23 #include <netinet/ip.h>
24 #include <netinet/in_pcb.h>
25 #include <netinet/in_var.h>
26 #include <netinet/ip_var.h>
27 
28 #ifdef vax
29 #include <machine/mtpr.h>
30 #endif
31 
32 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
33 static void ip_mloopback
34 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
35 
36 /*
37  * IP output.  The packet in mbuf chain m contains a skeletal IP
38  * header (with len, off, ttl, proto, tos, src, dst).
39  * The mbuf chain containing the packet will be freed.
40  * The mbuf opt, if present, will not be freed.
41  */
42 int
43 ip_output(m0, opt, ro, flags, imo)
44 	struct mbuf *m0;
45 	struct mbuf *opt;
46 	struct route *ro;
47 	int flags;
48 	struct ip_moptions *imo;
49 {
50 	register struct ip *ip, *mhip;
51 	register struct ifnet *ifp;
52 	register struct mbuf *m = m0;
53 	register int hlen = sizeof (struct ip);
54 	int len, off, error = 0;
55 	struct route iproute;
56 	struct sockaddr_in *dst;
57 	struct in_ifaddr *ia;
58 
59 #ifdef	DIAGNOSTIC
60 	if ((m->m_flags & M_PKTHDR) == 0)
61 		panic("ip_output no HDR");
62 #endif
63 	if (opt) {
64 		m = ip_insertoptions(m, opt, &len);
65 		hlen = len;
66 	}
67 	ip = mtod(m, struct ip *);
68 	/*
69 	 * Fill in IP header.
70 	 */
71 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
72 		ip->ip_v = IPVERSION;
73 		ip->ip_off &= IP_DF;
74 		ip->ip_id = htons(ip_id++);
75 		ip->ip_hl = hlen >> 2;
76 		ipstat.ips_localout++;
77 	} else {
78 		hlen = ip->ip_hl << 2;
79 	}
80 	/*
81 	 * Route packet.
82 	 */
83 	if (ro == 0) {
84 		ro = &iproute;
85 		bzero((caddr_t)ro, sizeof (*ro));
86 	}
87 	dst = (struct sockaddr_in *)&ro->ro_dst;
88 	/*
89 	 * If there is a cached route,
90 	 * check that it is to the same destination
91 	 * and is still up.  If not, free it and try again.
92 	 */
93 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
94 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
95 		RTFREE(ro->ro_rt);
96 		ro->ro_rt = (struct rtentry *)0;
97 	}
98 	if (ro->ro_rt == 0) {
99 		dst->sin_family = AF_INET;
100 		dst->sin_len = sizeof(*dst);
101 		dst->sin_addr = ip->ip_dst;
102 	}
103 	/*
104 	 * If routing to interface only,
105 	 * short circuit routing lookup.
106 	 */
107 #define ifatoia(ifa)	((struct in_ifaddr *)(ifa))
108 #define sintosa(sin)	((struct sockaddr *)(sin))
109 	if (flags & IP_ROUTETOIF) {
110 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
111 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
112 			ipstat.ips_noroute++;
113 			error = ENETUNREACH;
114 			goto bad;
115 		}
116 		ifp = ia->ia_ifp;
117 		ip->ip_ttl = 1;
118 	} else {
119 		if (ro->ro_rt == 0)
120 			rtalloc(ro);
121 		if (ro->ro_rt == 0) {
122 			ipstat.ips_noroute++;
123 			error = EHOSTUNREACH;
124 			goto bad;
125 		}
126 		ia = ifatoia(ro->ro_rt->rt_ifa);
127 		ifp = ro->ro_rt->rt_ifp;
128 		ro->ro_rt->rt_use++;
129 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
130 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
131 	}
132 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
133 		struct in_multi *inm;
134 		extern struct ifnet loif;
135 
136 		m->m_flags |= M_MCAST;
137 		/*
138 		 * IP destination address is multicast.  Make sure "dst"
139 		 * still points to the address in "ro".  (It may have been
140 		 * changed to point to a gateway address, above.)
141 		 */
142 		dst = (struct sockaddr_in *)&ro->ro_dst;
143 		/*
144 		 * See if the caller provided any multicast options
145 		 */
146 		if (imo != NULL) {
147 			ip->ip_ttl = imo->imo_multicast_ttl;
148 			if (imo->imo_multicast_ifp != NULL)
149 				ifp = imo->imo_multicast_ifp;
150 		} else
151 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
152 		/*
153 		 * Confirm that the outgoing interface supports multicast.
154 		 */
155 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
156 			ipstat.ips_noroute++;
157 			error = ENETUNREACH;
158 			goto bad;
159 		}
160 		/*
161 		 * If source address not specified yet, use address
162 		 * of outgoing interface.
163 		 */
164 		if (ip->ip_src.s_addr == INADDR_ANY) {
165 			register struct in_ifaddr *ia;
166 
167 			for (ia = in_ifaddr; ia; ia = ia->ia_next)
168 				if (ia->ia_ifp == ifp) {
169 					ip->ip_src = IA_SIN(ia)->sin_addr;
170 					break;
171 				}
172 		}
173 
174 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
175 		if (inm != NULL &&
176 		   (imo == NULL || imo->imo_multicast_loop)) {
177 			/*
178 			 * If we belong to the destination multicast group
179 			 * on the outgoing interface, and the caller did not
180 			 * forbid loopback, loop back a copy.
181 			 */
182 			ip_mloopback(ifp, m, dst);
183 		}
184 #ifdef MROUTING
185 		else {
186 			/*
187 			 * If we are acting as a multicast router, perform
188 			 * multicast forwarding as if the packet had just
189 			 * arrived on the interface to which we are about
190 			 * to send.  The multicast forwarding function
191 			 * recursively calls this function, using the
192 			 * IP_FORWARDING flag to prevent infinite recursion.
193 			 *
194 			 * Multicasts that are looped back by ip_mloopback(),
195 			 * above, will be forwarded by the ip_input() routine,
196 			 * if necessary.
197 			 */
198 			extern struct socket *ip_mrouter;
199 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
200 				if (ip_mforward(m, ifp) != 0) {
201 					m_freem(m);
202 					goto done;
203 				}
204 			}
205 		}
206 #endif
207 		/*
208 		 * Multicasts with a time-to-live of zero may be looped-
209 		 * back, above, but must not be transmitted on a network.
210 		 * Also, multicasts addressed to the loopback interface
211 		 * are not sent -- the above call to ip_mloopback() will
212 		 * loop back a copy if this host actually belongs to the
213 		 * destination group on the loopback interface.
214 		 */
215 		if (ip->ip_ttl == 0 || ifp == &loif) {
216 			m_freem(m);
217 			goto done;
218 		}
219 
220 		goto sendit;
221 	}
222 #ifndef notdef
223 	/*
224 	 * If source address not specified yet, use address
225 	 * of outgoing interface.
226 	 */
227 	if (ip->ip_src.s_addr == INADDR_ANY)
228 		ip->ip_src = IA_SIN(ia)->sin_addr;
229 #endif
230 	/*
231 	 * Look for broadcast address and
232 	 * and verify user is allowed to send
233 	 * such a packet.
234 	 */
235 	if (in_broadcast(dst->sin_addr, ifp)) {
236 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
237 			error = EADDRNOTAVAIL;
238 			goto bad;
239 		}
240 		if ((flags & IP_ALLOWBROADCAST) == 0) {
241 			error = EACCES;
242 			goto bad;
243 		}
244 		/* don't allow broadcast messages to be fragmented */
245 		if ((u_short)ip->ip_len > ifp->if_mtu) {
246 			error = EMSGSIZE;
247 			goto bad;
248 		}
249 		m->m_flags |= M_BCAST;
250 	} else
251 		m->m_flags &= ~M_BCAST;
252 
253 sendit:
254 	/*
255 	 * If small enough for interface, can just send directly.
256 	 */
257 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
258 		ip->ip_len = htons((u_short)ip->ip_len);
259 		ip->ip_off = htons((u_short)ip->ip_off);
260 		ip->ip_sum = 0;
261 		ip->ip_sum = in_cksum(m, hlen);
262 		error = (*ifp->if_output)(ifp, m,
263 				(struct sockaddr *)dst, ro->ro_rt);
264 		goto done;
265 	}
266 	/*
267 	 * Too large for interface; fragment if possible.
268 	 * Must be able to put at least 8 bytes per fragment.
269 	 */
270 	if (ip->ip_off & IP_DF) {
271 		error = EMSGSIZE;
272 		ipstat.ips_cantfrag++;
273 		goto bad;
274 	}
275 	len = (ifp->if_mtu - hlen) &~ 7;
276 	if (len < 8) {
277 		error = EMSGSIZE;
278 		goto bad;
279 	}
280 
281     {
282 	int mhlen, firstlen = len;
283 	struct mbuf **mnext = &m->m_nextpkt;
284 
285 	/*
286 	 * Loop through length of segment after first fragment,
287 	 * make new header and copy data of each part and link onto chain.
288 	 */
289 	m0 = m;
290 	mhlen = sizeof (struct ip);
291 	for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
292 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
293 		if (m == 0) {
294 			error = ENOBUFS;
295 			ipstat.ips_odropped++;
296 			goto sendorfree;
297 		}
298 		m->m_data += max_linkhdr;
299 		mhip = mtod(m, struct ip *);
300 		*mhip = *ip;
301 		if (hlen > sizeof (struct ip)) {
302 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
303 			mhip->ip_hl = mhlen >> 2;
304 		}
305 		m->m_len = mhlen;
306 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
307 		if (ip->ip_off & IP_MF)
308 			mhip->ip_off |= IP_MF;
309 		if (off + len >= (u_short)ip->ip_len)
310 			len = (u_short)ip->ip_len - off;
311 		else
312 			mhip->ip_off |= IP_MF;
313 		mhip->ip_len = htons((u_short)(len + mhlen));
314 		m->m_next = m_copy(m0, off, len);
315 		if (m->m_next == 0) {
316 			(void) m_free(m);
317 			error = ENOBUFS;	/* ??? */
318 			ipstat.ips_odropped++;
319 			goto sendorfree;
320 		}
321 		m->m_pkthdr.len = mhlen + len;
322 		m->m_pkthdr.rcvif = (struct ifnet *)0;
323 		mhip->ip_off = htons((u_short)mhip->ip_off);
324 		mhip->ip_sum = 0;
325 		mhip->ip_sum = in_cksum(m, mhlen);
326 		*mnext = m;
327 		mnext = &m->m_nextpkt;
328 		ipstat.ips_ofragments++;
329 	}
330 	/*
331 	 * Update first fragment by trimming what's been copied out
332 	 * and updating header, then send each fragment (in order).
333 	 */
334 	m = m0;
335 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
336 	m->m_pkthdr.len = hlen + firstlen;
337 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
338 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
339 	ip->ip_sum = 0;
340 	ip->ip_sum = in_cksum(m, hlen);
341 sendorfree:
342 	for (m = m0; m; m = m0) {
343 		m0 = m->m_nextpkt;
344 		m->m_nextpkt = 0;
345 		if (error == 0)
346 			error = (*ifp->if_output)(ifp, m,
347 			    (struct sockaddr *)dst, ro->ro_rt);
348 		else
349 			m_freem(m);
350 	}
351 
352 	if (error == 0)
353 		ipstat.ips_fragmented++;
354     }
355 done:
356 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
357 		RTFREE(ro->ro_rt);
358 	return (error);
359 bad:
360 	m_freem(m0);
361 	goto done;
362 }
363 
364 /*
365  * Insert IP options into preformed packet.
366  * Adjust IP destination as required for IP source routing,
367  * as indicated by a non-zero in_addr at the start of the options.
368  */
369 static struct mbuf *
370 ip_insertoptions(m, opt, phlen)
371 	register struct mbuf *m;
372 	struct mbuf *opt;
373 	int *phlen;
374 {
375 	register struct ipoption *p = mtod(opt, struct ipoption *);
376 	struct mbuf *n;
377 	register struct ip *ip = mtod(m, struct ip *);
378 	unsigned optlen;
379 
380 	optlen = opt->m_len - sizeof(p->ipopt_dst);
381 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
382 		return (m);		/* XXX should fail */
383 	if (p->ipopt_dst.s_addr)
384 		ip->ip_dst = p->ipopt_dst;
385 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
386 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
387 		if (n == 0)
388 			return (m);
389 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
390 		m->m_len -= sizeof(struct ip);
391 		m->m_data += sizeof(struct ip);
392 		n->m_next = m;
393 		m = n;
394 		m->m_len = optlen + sizeof(struct ip);
395 		m->m_data += max_linkhdr;
396 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
397 	} else {
398 		m->m_data -= optlen;
399 		m->m_len += optlen;
400 		m->m_pkthdr.len += optlen;
401 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
402 	}
403 	ip = mtod(m, struct ip *);
404 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
405 	*phlen = sizeof(struct ip) + optlen;
406 	ip->ip_len += optlen;
407 	return (m);
408 }
409 
410 /*
411  * Copy options from ip to jp,
412  * omitting those not copied during fragmentation.
413  */
414 int
415 ip_optcopy(ip, jp)
416 	struct ip *ip, *jp;
417 {
418 	register u_char *cp, *dp;
419 	int opt, optlen, cnt;
420 
421 	cp = (u_char *)(ip + 1);
422 	dp = (u_char *)(jp + 1);
423 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
424 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
425 		opt = cp[0];
426 		if (opt == IPOPT_EOL)
427 			break;
428 		if (opt == IPOPT_NOP) {
429 			/* Preserve for IP mcast tunnel's LSRR alignment. */
430 			*dp++ = IPOPT_NOP;
431 			optlen = 1;
432 			continue;
433 		} else
434 			optlen = cp[IPOPT_OLEN];
435 		/* bogus lengths should have been caught by ip_dooptions */
436 		if (optlen > cnt)
437 			optlen = cnt;
438 		if (IPOPT_COPIED(opt)) {
439 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
440 			dp += optlen;
441 		}
442 	}
443 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
444 		*dp++ = IPOPT_EOL;
445 	return (optlen);
446 }
447 
448 /*
449  * IP socket option processing.
450  */
451 int
452 ip_ctloutput(op, so, level, optname, mp)
453 	int op;
454 	struct socket *so;
455 	int level, optname;
456 	struct mbuf **mp;
457 {
458 	register struct inpcb *inp = sotoinpcb(so);
459 	register struct mbuf *m = *mp;
460 	register int optval;
461 	int error = 0;
462 
463 	if (level != IPPROTO_IP)
464 		goto freeit;
465 	else switch (op) {
466 
467 	case PRCO_SETOPT:
468 		switch (optname) {
469 		case IP_OPTIONS:
470 #ifdef notyet
471 		case IP_RETOPTS:
472 			return (ip_pcbopts(optname, &inp->inp_options, m));
473 #else
474 			return (ip_pcbopts(&inp->inp_options, m));
475 #endif
476 
477 		case IP_TOS:
478 		case IP_TTL:
479 		case IP_RECVOPTS:
480 		case IP_RECVRETOPTS:
481 		case IP_RECVDSTADDR:
482 			if (m->m_len != sizeof(int))
483 				error = EINVAL;
484 			else {
485 				optval = *mtod(m, int *);
486 				switch (optname) {
487 
488 				case IP_TOS:
489 					inp->inp_ip.ip_tos = optval;
490 					break;
491 
492 				case IP_TTL:
493 					inp->inp_ip.ip_ttl = optval;
494 					break;
495 #define	OPTSET(bit) \
496 	if (optval) \
497 		inp->inp_flags |= bit; \
498 	else \
499 		inp->inp_flags &= ~bit;
500 
501 				case IP_RECVOPTS:
502 					OPTSET(INP_RECVOPTS);
503 					break;
504 
505 				case IP_RECVRETOPTS:
506 					OPTSET(INP_RECVRETOPTS);
507 					break;
508 
509 				case IP_RECVDSTADDR:
510 					OPTSET(INP_RECVDSTADDR);
511 					break;
512 				}
513 			}
514 			break;
515 #undef OPTSET
516 
517 		case IP_MULTICAST_IF:
518 		case IP_MULTICAST_TTL:
519 		case IP_MULTICAST_LOOP:
520 		case IP_ADD_MEMBERSHIP:
521 		case IP_DROP_MEMBERSHIP:
522 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
523 			break;
524 
525 		freeit:
526 		default:
527 			error = EINVAL;
528 			break;
529 		}
530 		if (m)
531 			(void)m_free(m);
532 		break;
533 
534 	case PRCO_GETOPT:
535 		switch (optname) {
536 		case IP_OPTIONS:
537 		case IP_RETOPTS:
538 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
539 			if (inp->inp_options) {
540 				m->m_len = inp->inp_options->m_len;
541 				bcopy(mtod(inp->inp_options, caddr_t),
542 				    mtod(m, caddr_t), (unsigned)m->m_len);
543 			} else
544 				m->m_len = 0;
545 			break;
546 
547 		case IP_TOS:
548 		case IP_TTL:
549 		case IP_RECVOPTS:
550 		case IP_RECVRETOPTS:
551 		case IP_RECVDSTADDR:
552 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
553 			m->m_len = sizeof(int);
554 			switch (optname) {
555 
556 			case IP_TOS:
557 				optval = inp->inp_ip.ip_tos;
558 				break;
559 
560 			case IP_TTL:
561 				optval = inp->inp_ip.ip_ttl;
562 				break;
563 
564 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
565 
566 			case IP_RECVOPTS:
567 				optval = OPTBIT(INP_RECVOPTS);
568 				break;
569 
570 			case IP_RECVRETOPTS:
571 				optval = OPTBIT(INP_RECVRETOPTS);
572 				break;
573 
574 			case IP_RECVDSTADDR:
575 				optval = OPTBIT(INP_RECVDSTADDR);
576 				break;
577 			}
578 			*mtod(m, int *) = optval;
579 			break;
580 
581 		case IP_MULTICAST_IF:
582 		case IP_MULTICAST_TTL:
583 		case IP_MULTICAST_LOOP:
584 		case IP_ADD_MEMBERSHIP:
585 		case IP_DROP_MEMBERSHIP:
586 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
587 			break;
588 
589 		default:
590 			error = EINVAL;
591 			break;
592 		}
593 		break;
594 	}
595 	return (error);
596 }
597 
598 /*
599  * Set up IP options in pcb for insertion in output packets.
600  * Store in mbuf with pointer in pcbopt, adding pseudo-option
601  * with destination address if source routed.
602  */
603 int
604 #ifdef notyet
605 ip_pcbopts(optname, pcbopt, m)
606 	int optname;
607 #else
608 ip_pcbopts(pcbopt, m)
609 #endif
610 	struct mbuf **pcbopt;
611 	register struct mbuf *m;
612 {
613 	register cnt, optlen;
614 	register u_char *cp;
615 	u_char opt;
616 
617 	/* turn off any old options */
618 	if (*pcbopt)
619 		(void)m_free(*pcbopt);
620 	*pcbopt = 0;
621 	if (m == (struct mbuf *)0 || m->m_len == 0) {
622 		/*
623 		 * Only turning off any previous options.
624 		 */
625 		if (m)
626 			(void)m_free(m);
627 		return (0);
628 	}
629 
630 #ifndef	vax
631 	if (m->m_len % sizeof(long))
632 		goto bad;
633 #endif
634 	/*
635 	 * IP first-hop destination address will be stored before
636 	 * actual options; move other options back
637 	 * and clear it when none present.
638 	 */
639 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
640 		goto bad;
641 	cnt = m->m_len;
642 	m->m_len += sizeof(struct in_addr);
643 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
644 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
645 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
646 
647 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
648 		opt = cp[IPOPT_OPTVAL];
649 		if (opt == IPOPT_EOL)
650 			break;
651 		if (opt == IPOPT_NOP)
652 			optlen = 1;
653 		else {
654 			optlen = cp[IPOPT_OLEN];
655 			if (optlen <= IPOPT_OLEN || optlen > cnt)
656 				goto bad;
657 		}
658 		switch (opt) {
659 
660 		default:
661 			break;
662 
663 		case IPOPT_LSRR:
664 		case IPOPT_SSRR:
665 			/*
666 			 * user process specifies route as:
667 			 *	->A->B->C->D
668 			 * D must be our final destination (but we can't
669 			 * check that since we may not have connected yet).
670 			 * A is first hop destination, which doesn't appear in
671 			 * actual IP option, but is stored before the options.
672 			 */
673 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
674 				goto bad;
675 			m->m_len -= sizeof(struct in_addr);
676 			cnt -= sizeof(struct in_addr);
677 			optlen -= sizeof(struct in_addr);
678 			cp[IPOPT_OLEN] = optlen;
679 			/*
680 			 * Move first hop before start of options.
681 			 */
682 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
683 			    sizeof(struct in_addr));
684 			/*
685 			 * Then copy rest of options back
686 			 * to close up the deleted entry.
687 			 */
688 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
689 			    sizeof(struct in_addr)),
690 			    (caddr_t)&cp[IPOPT_OFFSET+1],
691 			    (unsigned)cnt + sizeof(struct in_addr));
692 			break;
693 		}
694 	}
695 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
696 		goto bad;
697 	*pcbopt = m;
698 	return (0);
699 
700 bad:
701 	(void)m_free(m);
702 	return (EINVAL);
703 }
704 
705 /*
706  * Set the IP multicast options in response to user setsockopt().
707  */
708 int
709 ip_setmoptions(optname, imop, m)
710 	int optname;
711 	struct ip_moptions **imop;
712 	struct mbuf *m;
713 {
714 	register int error = 0;
715 	u_char loop;
716 	register int i;
717 	struct in_addr addr;
718 	register struct ip_mreq *mreq;
719 	register struct ifnet *ifp;
720 	register struct ip_moptions *imo = *imop;
721 	struct route ro;
722 	register struct sockaddr_in *dst;
723 
724 	if (imo == NULL) {
725 		/*
726 		 * No multicast option buffer attached to the pcb;
727 		 * allocate one and initialize to default values.
728 		 */
729 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
730 		    M_WAITOK);
731 
732 		if (imo == NULL)
733 			return (ENOBUFS);
734 		*imop = imo;
735 		imo->imo_multicast_ifp = NULL;
736 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
737 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
738 		imo->imo_num_memberships = 0;
739 	}
740 
741 	switch (optname) {
742 
743 	case IP_MULTICAST_IF:
744 		/*
745 		 * Select the interface for outgoing multicast packets.
746 		 */
747 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
748 			error = EINVAL;
749 			break;
750 		}
751 		addr = *(mtod(m, struct in_addr *));
752 		/*
753 		 * INADDR_ANY is used to remove a previous selection.
754 		 * When no interface is selected, a default one is
755 		 * chosen every time a multicast packet is sent.
756 		 */
757 		if (addr.s_addr == INADDR_ANY) {
758 			imo->imo_multicast_ifp = NULL;
759 			break;
760 		}
761 		/*
762 		 * The selected interface is identified by its local
763 		 * IP address.  Find the interface and confirm that
764 		 * it supports multicasting.
765 		 */
766 		INADDR_TO_IFP(addr, ifp);
767 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
768 			error = EADDRNOTAVAIL;
769 			break;
770 		}
771 		imo->imo_multicast_ifp = ifp;
772 		break;
773 
774 	case IP_MULTICAST_TTL:
775 		/*
776 		 * Set the IP time-to-live for outgoing multicast packets.
777 		 */
778 		if (m == NULL || m->m_len != 1) {
779 			error = EINVAL;
780 			break;
781 		}
782 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
783 		break;
784 
785 	case IP_MULTICAST_LOOP:
786 		/*
787 		 * Set the loopback flag for outgoing multicast packets.
788 		 * Must be zero or one.
789 		 */
790 		if (m == NULL || m->m_len != 1 ||
791 		   (loop = *(mtod(m, u_char *))) > 1) {
792 			error = EINVAL;
793 			break;
794 		}
795 		imo->imo_multicast_loop = loop;
796 		break;
797 
798 	case IP_ADD_MEMBERSHIP:
799 		/*
800 		 * Add a multicast group membership.
801 		 * Group must be a valid IP multicast address.
802 		 */
803 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
804 			error = EINVAL;
805 			break;
806 		}
807 		mreq = mtod(m, struct ip_mreq *);
808 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
809 			error = EINVAL;
810 			break;
811 		}
812 		/*
813 		 * If no interface address was provided, use the interface of
814 		 * the route to the given multicast address.
815 		 */
816 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
817 			ro.ro_rt = NULL;
818 			dst = (struct sockaddr_in *)&ro.ro_dst;
819 			dst->sin_len = sizeof(*dst);
820 			dst->sin_family = AF_INET;
821 			dst->sin_addr = mreq->imr_multiaddr;
822 			rtalloc(&ro);
823 			if (ro.ro_rt == NULL) {
824 				error = EADDRNOTAVAIL;
825 				break;
826 			}
827 			ifp = ro.ro_rt->rt_ifp;
828 			rtfree(ro.ro_rt);
829 		}
830 		else {
831 			INADDR_TO_IFP(mreq->imr_interface, ifp);
832 		}
833 		/*
834 		 * See if we found an interface, and confirm that it
835 		 * supports multicast.
836 		 */
837 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
838 			error = EADDRNOTAVAIL;
839 			break;
840 		}
841 		/*
842 		 * See if the membership already exists or if all the
843 		 * membership slots are full.
844 		 */
845 		for (i = 0; i < imo->imo_num_memberships; ++i) {
846 			if (imo->imo_membership[i]->inm_ifp == ifp &&
847 			    imo->imo_membership[i]->inm_addr.s_addr
848 						== mreq->imr_multiaddr.s_addr)
849 				break;
850 		}
851 		if (i < imo->imo_num_memberships) {
852 			error = EADDRINUSE;
853 			break;
854 		}
855 		if (i == IP_MAX_MEMBERSHIPS) {
856 			error = ETOOMANYREFS;
857 			break;
858 		}
859 		/*
860 		 * Everything looks good; add a new record to the multicast
861 		 * address list for the given interface.
862 		 */
863 		if ((imo->imo_membership[i] =
864 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
865 			error = ENOBUFS;
866 			break;
867 		}
868 		++imo->imo_num_memberships;
869 		break;
870 
871 	case IP_DROP_MEMBERSHIP:
872 		/*
873 		 * Drop a multicast group membership.
874 		 * Group must be a valid IP multicast address.
875 		 */
876 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
877 			error = EINVAL;
878 			break;
879 		}
880 		mreq = mtod(m, struct ip_mreq *);
881 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
882 			error = EINVAL;
883 			break;
884 		}
885 		/*
886 		 * If an interface address was specified, get a pointer
887 		 * to its ifnet structure.
888 		 */
889 		if (mreq->imr_interface.s_addr == INADDR_ANY)
890 			ifp = NULL;
891 		else {
892 			INADDR_TO_IFP(mreq->imr_interface, ifp);
893 			if (ifp == NULL) {
894 				error = EADDRNOTAVAIL;
895 				break;
896 			}
897 		}
898 		/*
899 		 * Find the membership in the membership array.
900 		 */
901 		for (i = 0; i < imo->imo_num_memberships; ++i) {
902 			if ((ifp == NULL ||
903 			     imo->imo_membership[i]->inm_ifp == ifp) &&
904 			     imo->imo_membership[i]->inm_addr.s_addr ==
905 			     mreq->imr_multiaddr.s_addr)
906 				break;
907 		}
908 		if (i == imo->imo_num_memberships) {
909 			error = EADDRNOTAVAIL;
910 			break;
911 		}
912 		/*
913 		 * Give up the multicast address record to which the
914 		 * membership points.
915 		 */
916 		in_delmulti(imo->imo_membership[i]);
917 		/*
918 		 * Remove the gap in the membership array.
919 		 */
920 		for (++i; i < imo->imo_num_memberships; ++i)
921 			imo->imo_membership[i-1] = imo->imo_membership[i];
922 		--imo->imo_num_memberships;
923 		break;
924 
925 	default:
926 		error = EOPNOTSUPP;
927 		break;
928 	}
929 
930 	/*
931 	 * If all options have default values, no need to keep the mbuf.
932 	 */
933 	if (imo->imo_multicast_ifp == NULL &&
934 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
935 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
936 	    imo->imo_num_memberships == 0) {
937 		free(*imop, M_IPMOPTS);
938 		*imop = NULL;
939 	}
940 
941 	return (error);
942 }
943 
944 /*
945  * Return the IP multicast options in response to user getsockopt().
946  */
947 int
948 ip_getmoptions(optname, imo, mp)
949 	int optname;
950 	register struct ip_moptions *imo;
951 	register struct mbuf **mp;
952 {
953 	u_char *ttl;
954 	u_char *loop;
955 	struct in_addr *addr;
956 	struct in_ifaddr *ia;
957 
958 	*mp = m_get(M_WAIT, MT_SOOPTS);
959 
960 	switch (optname) {
961 
962 	case IP_MULTICAST_IF:
963 		addr = mtod(*mp, struct in_addr *);
964 		(*mp)->m_len = sizeof(struct in_addr);
965 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
966 			addr->s_addr = INADDR_ANY;
967 		else {
968 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
969 			addr->s_addr = (ia == NULL) ? INADDR_ANY
970 					: IA_SIN(ia)->sin_addr.s_addr;
971 		}
972 		return (0);
973 
974 	case IP_MULTICAST_TTL:
975 		ttl = mtod(*mp, u_char *);
976 		(*mp)->m_len = 1;
977 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
978 				     : imo->imo_multicast_ttl;
979 		return (0);
980 
981 	case IP_MULTICAST_LOOP:
982 		loop = mtod(*mp, u_char *);
983 		(*mp)->m_len = 1;
984 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
985 				      : imo->imo_multicast_loop;
986 		return (0);
987 
988 	default:
989 		return (EOPNOTSUPP);
990 	}
991 }
992 
993 /*
994  * Discard the IP multicast options.
995  */
996 void
997 ip_freemoptions(imo)
998 	register struct ip_moptions *imo;
999 {
1000 	register int i;
1001 
1002 	if (imo != NULL) {
1003 		for (i = 0; i < imo->imo_num_memberships; ++i)
1004 			in_delmulti(imo->imo_membership[i]);
1005 		free(imo, M_IPMOPTS);
1006 	}
1007 }
1008 
1009 /*
1010  * Routine called from ip_output() to loop back a copy of an IP multicast
1011  * packet to the input queue of a specified interface.  Note that this
1012  * calls the output routine of the loopback "driver", but with an interface
1013  * pointer that might NOT be &loif -- easier than replicating that code here.
1014  */
1015 static void
1016 ip_mloopback(ifp, m, dst)
1017 	struct ifnet *ifp;
1018 	register struct mbuf *m;
1019 	register struct sockaddr_in *dst;
1020 {
1021 	register struct ip *ip;
1022 	struct mbuf *copym;
1023 
1024 	copym = m_copy(m, 0, M_COPYALL);
1025 	if (copym != NULL) {
1026 		/*
1027 		 * We don't bother to fragment if the IP length is greater
1028 		 * than the interface's MTU.  Can this possibly matter?
1029 		 */
1030 		ip = mtod(copym, struct ip *);
1031 		ip->ip_len = htons((u_short)ip->ip_len);
1032 		ip->ip_off = htons((u_short)ip->ip_off);
1033 		ip->ip_sum = 0;
1034 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1035 		(void) looutput(ifp, copym, (struct sockaddr *)dst, NULL);
1036 	}
1037 }
1038