xref: /netbsd-src/sys/netinet/ip_output.c (revision d0fed6c87ddc40a8bffa6f99e7433ddfc864dd83)
1 /*	$NetBSD: ip_output.c,v 1.39 1997/04/15 00:41:53 christos Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1988, 1990, 1993
5  *	The Regents of the University of California.  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. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
36  */
37 
38 #include <sys/param.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/errno.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/systm.h>
46 
47 #include <net/if.h>
48 #include <net/route.h>
49 #include <net/pfil.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 
58 #ifdef vax
59 #include <machine/mtpr.h>
60 #endif
61 
62 #include <machine/stdarg.h>
63 
64 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
65 static void ip_mloopback
66 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
67 
68 /*
69  * IP output.  The packet in mbuf chain m contains a skeletal IP
70  * header (with len, off, ttl, proto, tos, src, dst).
71  * The mbuf chain containing the packet will be freed.
72  * The mbuf opt, if present, will not be freed.
73  */
74 int
75 #if __STDC__
76 ip_output(struct mbuf *m0, ...)
77 #else
78 ip_output(m0, va_alist)
79 	struct mbuf *m0;
80 	va_dcl
81 #endif
82 {
83 	register struct ip *ip, *mhip;
84 	register struct ifnet *ifp;
85 	register struct mbuf *m = m0;
86 	register int hlen = sizeof (struct ip);
87 	int len, off, error = 0;
88 	struct route iproute;
89 	struct sockaddr_in *dst;
90 	struct in_ifaddr *ia;
91 	struct mbuf *opt;
92 	struct route *ro;
93 	int flags;
94 	struct ip_moptions *imo;
95 	va_list ap;
96 #ifdef PFIL_HOOKS
97 	struct packet_filter_hook *pfh;
98 	struct mbuf *m1;
99 	int rv;
100 #endif /* PFIL_HOOKS */
101 
102 	va_start(ap, m0);
103 	opt = va_arg(ap, struct mbuf *);
104 	ro = va_arg(ap, struct route *);
105 	flags = va_arg(ap, int);
106 	imo = va_arg(ap, struct ip_moptions *);
107 	va_end(ap);
108 
109 #ifdef	DIAGNOSTIC
110 	if ((m->m_flags & M_PKTHDR) == 0)
111 		panic("ip_output no HDR");
112 #endif
113 	if (opt) {
114 		m = ip_insertoptions(m, opt, &len);
115 		hlen = len;
116 	}
117 	ip = mtod(m, struct ip *);
118 	/*
119 	 * Fill in IP header.
120 	 */
121 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
122 		ip->ip_v = IPVERSION;
123 		ip->ip_off &= IP_DF;
124 		ip->ip_id = htons(ip_id++);
125 		ip->ip_hl = hlen >> 2;
126 		ipstat.ips_localout++;
127 	} else {
128 		hlen = ip->ip_hl << 2;
129 	}
130 	/*
131 	 * Route packet.
132 	 */
133 	if (ro == 0) {
134 		ro = &iproute;
135 		bzero((caddr_t)ro, sizeof (*ro));
136 	}
137 	dst = satosin(&ro->ro_dst);
138 	/*
139 	 * If there is a cached route,
140 	 * check that it is to the same destination
141 	 * and is still up.  If not, free it and try again.
142 	 */
143 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
144 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
145 		RTFREE(ro->ro_rt);
146 		ro->ro_rt = (struct rtentry *)0;
147 	}
148 	if (ro->ro_rt == 0) {
149 		dst->sin_family = AF_INET;
150 		dst->sin_len = sizeof(*dst);
151 		dst->sin_addr = ip->ip_dst;
152 	}
153 	/*
154 	 * If routing to interface only,
155 	 * short circuit routing lookup.
156 	 */
157 	if (flags & IP_ROUTETOIF) {
158 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
159 			ipstat.ips_noroute++;
160 			error = ENETUNREACH;
161 			goto bad;
162 		}
163 		ifp = ia->ia_ifp;
164 		ip->ip_ttl = 1;
165 	} else {
166 		if (ro->ro_rt == 0)
167 			rtalloc(ro);
168 		if (ro->ro_rt == 0) {
169 			ipstat.ips_noroute++;
170 			error = EHOSTUNREACH;
171 			goto bad;
172 		}
173 		ia = ifatoia(ro->ro_rt->rt_ifa);
174 		ifp = ro->ro_rt->rt_ifp;
175 		ro->ro_rt->rt_use++;
176 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
177 			dst = satosin(ro->ro_rt->rt_gateway);
178 	}
179 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
180 		struct in_multi *inm;
181 
182 		m->m_flags |= M_MCAST;
183 		/*
184 		 * IP destination address is multicast.  Make sure "dst"
185 		 * still points to the address in "ro".  (It may have been
186 		 * changed to point to a gateway address, above.)
187 		 */
188 		dst = satosin(&ro->ro_dst);
189 		/*
190 		 * See if the caller provided any multicast options
191 		 */
192 		if (imo != NULL) {
193 			ip->ip_ttl = imo->imo_multicast_ttl;
194 			if (imo->imo_multicast_ifp != NULL)
195 				ifp = imo->imo_multicast_ifp;
196 		} else
197 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
198 		/*
199 		 * Confirm that the outgoing interface supports multicast.
200 		 */
201 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
202 			ipstat.ips_noroute++;
203 			error = ENETUNREACH;
204 			goto bad;
205 		}
206 		/*
207 		 * If source address not specified yet, use address
208 		 * of outgoing interface.
209 		 */
210 		if (in_nullhost(ip->ip_src)) {
211 			register struct in_ifaddr *ia;
212 
213 			for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next)
214 				if (ia->ia_ifp == ifp) {
215 					ip->ip_src = ia->ia_addr.sin_addr;
216 					break;
217 				}
218 		}
219 
220 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
221 		if (inm != NULL &&
222 		   (imo == NULL || imo->imo_multicast_loop)) {
223 			/*
224 			 * If we belong to the destination multicast group
225 			 * on the outgoing interface, and the caller did not
226 			 * forbid loopback, loop back a copy.
227 			 */
228 			ip_mloopback(ifp, m, dst);
229 		}
230 #ifdef MROUTING
231 		else {
232 			/*
233 			 * If we are acting as a multicast router, perform
234 			 * multicast forwarding as if the packet had just
235 			 * arrived on the interface to which we are about
236 			 * to send.  The multicast forwarding function
237 			 * recursively calls this function, using the
238 			 * IP_FORWARDING flag to prevent infinite recursion.
239 			 *
240 			 * Multicasts that are looped back by ip_mloopback(),
241 			 * above, will be forwarded by the ip_input() routine,
242 			 * if necessary.
243 			 */
244 			extern struct socket *ip_mrouter;
245 
246 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
247 				if (ip_mforward(m, ifp) != 0) {
248 					m_freem(m);
249 					goto done;
250 				}
251 			}
252 		}
253 #endif
254 		/*
255 		 * Multicasts with a time-to-live of zero may be looped-
256 		 * back, above, but must not be transmitted on a network.
257 		 * Also, multicasts addressed to the loopback interface
258 		 * are not sent -- the above call to ip_mloopback() will
259 		 * loop back a copy if this host actually belongs to the
260 		 * destination group on the loopback interface.
261 		 */
262 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
263 			m_freem(m);
264 			goto done;
265 		}
266 
267 		goto sendit;
268 	}
269 #ifndef notdef
270 	/*
271 	 * If source address not specified yet, use address
272 	 * of outgoing interface.
273 	 */
274 	if (in_nullhost(ip->ip_src))
275 		ip->ip_src = ia->ia_addr.sin_addr;
276 #endif
277 	/*
278 	 * Look for broadcast address and
279 	 * and verify user is allowed to send
280 	 * such a packet.
281 	 */
282 	if (in_broadcast(dst->sin_addr, ifp)) {
283 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
284 			error = EADDRNOTAVAIL;
285 			goto bad;
286 		}
287 		if ((flags & IP_ALLOWBROADCAST) == 0) {
288 			error = EACCES;
289 			goto bad;
290 		}
291 		/* don't allow broadcast messages to be fragmented */
292 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
293 			error = EMSGSIZE;
294 			goto bad;
295 		}
296 		m->m_flags |= M_BCAST;
297 	} else
298 		m->m_flags &= ~M_BCAST;
299 
300 #ifdef PFIL_HOOKS
301 	/*
302 	 * Run through list of hooks for output packets.
303 	 */
304 	m1 = m;
305 	for (pfh = pfil_hook_get(PFIL_OUT); pfh; pfh = pfh->pfil_link.le_next)
306 		if (pfh->pfil_func) {
307 		    	rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
308 			if (rv) {
309 				error = EHOSTUNREACH;
310 				goto done;
311 			}
312 			ip = mtod(m = m1, struct ip *);
313 		}
314 #endif /* PFIL_HOOKS */
315 sendit:
316 	/*
317 	 * If small enough for interface, can just send directly.
318 	 */
319 	if ((u_int16_t)ip->ip_len <= ifp->if_mtu) {
320 		ip->ip_len = htons((u_int16_t)ip->ip_len);
321 		ip->ip_off = htons((u_int16_t)ip->ip_off);
322 		ip->ip_sum = 0;
323 		ip->ip_sum = in_cksum(m, hlen);
324 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
325 		goto done;
326 	}
327 	/*
328 	 * Too large for interface; fragment if possible.
329 	 * Must be able to put at least 8 bytes per fragment.
330 	 */
331 	if (ip->ip_off & IP_DF) {
332 		error = EMSGSIZE;
333 		ipstat.ips_cantfrag++;
334 		goto bad;
335 	}
336 	len = (ifp->if_mtu - hlen) &~ 7;
337 	if (len < 8) {
338 		error = EMSGSIZE;
339 		goto bad;
340 	}
341 
342     {
343 	int mhlen, firstlen = len;
344 	struct mbuf **mnext = &m->m_nextpkt;
345 
346 	/*
347 	 * Loop through length of segment after first fragment,
348 	 * make new header and copy data of each part and link onto chain.
349 	 */
350 	m0 = m;
351 	mhlen = sizeof (struct ip);
352 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
353 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
354 		if (m == 0) {
355 			error = ENOBUFS;
356 			ipstat.ips_odropped++;
357 			goto sendorfree;
358 		}
359 		*mnext = m;
360 		mnext = &m->m_nextpkt;
361 		m->m_data += max_linkhdr;
362 		mhip = mtod(m, struct ip *);
363 		*mhip = *ip;
364 		if (hlen > sizeof (struct ip)) {
365 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
366 			mhip->ip_hl = mhlen >> 2;
367 		}
368 		m->m_len = mhlen;
369 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
370 		if (ip->ip_off & IP_MF)
371 			mhip->ip_off |= IP_MF;
372 		if (off + len >= (u_int16_t)ip->ip_len)
373 			len = (u_int16_t)ip->ip_len - off;
374 		else
375 			mhip->ip_off |= IP_MF;
376 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
377 		m->m_next = m_copy(m0, off, len);
378 		if (m->m_next == 0) {
379 			error = ENOBUFS;	/* ??? */
380 			ipstat.ips_odropped++;
381 			goto sendorfree;
382 		}
383 		m->m_pkthdr.len = mhlen + len;
384 		m->m_pkthdr.rcvif = (struct ifnet *)0;
385 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
386 		mhip->ip_sum = 0;
387 		mhip->ip_sum = in_cksum(m, mhlen);
388 		ipstat.ips_ofragments++;
389 	}
390 	/*
391 	 * Update first fragment by trimming what's been copied out
392 	 * and updating header, then send each fragment (in order).
393 	 */
394 	m = m0;
395 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
396 	m->m_pkthdr.len = hlen + firstlen;
397 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
398 	ip->ip_off = htons((u_int16_t)(ip->ip_off | IP_MF));
399 	ip->ip_sum = 0;
400 	ip->ip_sum = in_cksum(m, hlen);
401 sendorfree:
402 	for (m = m0; m; m = m0) {
403 		m0 = m->m_nextpkt;
404 		m->m_nextpkt = 0;
405 		if (error == 0)
406 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
407 			    ro->ro_rt);
408 		else
409 			m_freem(m);
410 	}
411 
412 	if (error == 0)
413 		ipstat.ips_fragmented++;
414     }
415 done:
416 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
417 		RTFREE(ro->ro_rt);
418 		ro->ro_rt = 0;
419 	}
420 	return (error);
421 bad:
422 	m_freem(m);
423 	goto done;
424 }
425 
426 /*
427  * Insert IP options into preformed packet.
428  * Adjust IP destination as required for IP source routing,
429  * as indicated by a non-zero in_addr at the start of the options.
430  */
431 static struct mbuf *
432 ip_insertoptions(m, opt, phlen)
433 	register struct mbuf *m;
434 	struct mbuf *opt;
435 	int *phlen;
436 {
437 	register struct ipoption *p = mtod(opt, struct ipoption *);
438 	struct mbuf *n;
439 	register struct ip *ip = mtod(m, struct ip *);
440 	unsigned optlen;
441 
442 	optlen = opt->m_len - sizeof(p->ipopt_dst);
443 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
444 		return (m);		/* XXX should fail */
445 	if (!in_nullhost(p->ipopt_dst))
446 		ip->ip_dst = p->ipopt_dst;
447 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
448 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
449 		if (n == 0)
450 			return (m);
451 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
452 		m->m_len -= sizeof(struct ip);
453 		m->m_data += sizeof(struct ip);
454 		n->m_next = m;
455 		m = n;
456 		m->m_len = optlen + sizeof(struct ip);
457 		m->m_data += max_linkhdr;
458 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
459 	} else {
460 		m->m_data -= optlen;
461 		m->m_len += optlen;
462 		m->m_pkthdr.len += optlen;
463 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
464 	}
465 	ip = mtod(m, struct ip *);
466 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
467 	*phlen = sizeof(struct ip) + optlen;
468 	ip->ip_len += optlen;
469 	return (m);
470 }
471 
472 /*
473  * Copy options from ip to jp,
474  * omitting those not copied during fragmentation.
475  */
476 int
477 ip_optcopy(ip, jp)
478 	struct ip *ip, *jp;
479 {
480 	register u_char *cp, *dp;
481 	int opt, optlen, cnt;
482 
483 	cp = (u_char *)(ip + 1);
484 	dp = (u_char *)(jp + 1);
485 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
486 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
487 		opt = cp[0];
488 		if (opt == IPOPT_EOL)
489 			break;
490 		if (opt == IPOPT_NOP) {
491 			/* Preserve for IP mcast tunnel's LSRR alignment. */
492 			*dp++ = IPOPT_NOP;
493 			optlen = 1;
494 			continue;
495 		} else
496 			optlen = cp[IPOPT_OLEN];
497 		/* bogus lengths should have been caught by ip_dooptions */
498 		if (optlen > cnt)
499 			optlen = cnt;
500 		if (IPOPT_COPIED(opt)) {
501 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
502 			dp += optlen;
503 		}
504 	}
505 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
506 		*dp++ = IPOPT_EOL;
507 	return (optlen);
508 }
509 
510 /*
511  * IP socket option processing.
512  */
513 int
514 ip_ctloutput(op, so, level, optname, mp)
515 	int op;
516 	struct socket *so;
517 	int level, optname;
518 	struct mbuf **mp;
519 {
520 	register struct inpcb *inp = sotoinpcb(so);
521 	register struct mbuf *m = *mp;
522 	register int optval = 0;
523 	int error = 0;
524 
525 	if (level != IPPROTO_IP) {
526 		error = EINVAL;
527 		if (op == PRCO_SETOPT && *mp)
528 			(void) m_free(*mp);
529 	} else switch (op) {
530 
531 	case PRCO_SETOPT:
532 		switch (optname) {
533 		case IP_OPTIONS:
534 #ifdef notyet
535 		case IP_RETOPTS:
536 			return (ip_pcbopts(optname, &inp->inp_options, m));
537 #else
538 			return (ip_pcbopts(&inp->inp_options, m));
539 #endif
540 
541 		case IP_TOS:
542 		case IP_TTL:
543 		case IP_RECVOPTS:
544 		case IP_RECVRETOPTS:
545 		case IP_RECVDSTADDR:
546 		case IP_RECVIF:
547 			if (m == NULL || m->m_len != sizeof(int))
548 				error = EINVAL;
549 			else {
550 				optval = *mtod(m, int *);
551 				switch (optname) {
552 
553 				case IP_TOS:
554 					inp->inp_ip.ip_tos = optval;
555 					break;
556 
557 				case IP_TTL:
558 					inp->inp_ip.ip_ttl = optval;
559 					break;
560 #define	OPTSET(bit) \
561 	if (optval) \
562 		inp->inp_flags |= bit; \
563 	else \
564 		inp->inp_flags &= ~bit;
565 
566 				case IP_RECVOPTS:
567 					OPTSET(INP_RECVOPTS);
568 					break;
569 
570 				case IP_RECVRETOPTS:
571 					OPTSET(INP_RECVRETOPTS);
572 					break;
573 
574 				case IP_RECVDSTADDR:
575 					OPTSET(INP_RECVDSTADDR);
576 					break;
577 
578 				case IP_RECVIF:
579 					OPTSET(INP_RECVIF);
580 					break;
581 				}
582 			}
583 			break;
584 #undef OPTSET
585 
586 		case IP_MULTICAST_IF:
587 		case IP_MULTICAST_TTL:
588 		case IP_MULTICAST_LOOP:
589 		case IP_ADD_MEMBERSHIP:
590 		case IP_DROP_MEMBERSHIP:
591 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
592 			break;
593 
594 		default:
595 			error = ENOPROTOOPT;
596 			break;
597 		}
598 		if (m)
599 			(void)m_free(m);
600 		break;
601 
602 	case PRCO_GETOPT:
603 		switch (optname) {
604 		case IP_OPTIONS:
605 		case IP_RETOPTS:
606 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
607 			if (inp->inp_options) {
608 				m->m_len = inp->inp_options->m_len;
609 				bcopy(mtod(inp->inp_options, caddr_t),
610 				    mtod(m, caddr_t), (unsigned)m->m_len);
611 			} else
612 				m->m_len = 0;
613 			break;
614 
615 		case IP_TOS:
616 		case IP_TTL:
617 		case IP_RECVOPTS:
618 		case IP_RECVRETOPTS:
619 		case IP_RECVDSTADDR:
620 		case IP_RECVIF:
621 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
622 			m->m_len = sizeof(int);
623 			switch (optname) {
624 
625 			case IP_TOS:
626 				optval = inp->inp_ip.ip_tos;
627 				break;
628 
629 			case IP_TTL:
630 				optval = inp->inp_ip.ip_ttl;
631 				break;
632 
633 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
634 
635 			case IP_RECVOPTS:
636 				optval = OPTBIT(INP_RECVOPTS);
637 				break;
638 
639 			case IP_RECVRETOPTS:
640 				optval = OPTBIT(INP_RECVRETOPTS);
641 				break;
642 
643 			case IP_RECVDSTADDR:
644 				optval = OPTBIT(INP_RECVDSTADDR);
645 				break;
646 
647 			case IP_RECVIF:
648 				optval = OPTBIT(INP_RECVIF);
649 				break;
650 			}
651 			*mtod(m, int *) = optval;
652 			break;
653 
654 		case IP_MULTICAST_IF:
655 		case IP_MULTICAST_TTL:
656 		case IP_MULTICAST_LOOP:
657 		case IP_ADD_MEMBERSHIP:
658 		case IP_DROP_MEMBERSHIP:
659 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
660 			break;
661 
662 		default:
663 			error = ENOPROTOOPT;
664 			break;
665 		}
666 		break;
667 	}
668 	return (error);
669 }
670 
671 /*
672  * Set up IP options in pcb for insertion in output packets.
673  * Store in mbuf with pointer in pcbopt, adding pseudo-option
674  * with destination address if source routed.
675  */
676 int
677 #ifdef notyet
678 ip_pcbopts(optname, pcbopt, m)
679 	int optname;
680 #else
681 ip_pcbopts(pcbopt, m)
682 #endif
683 	struct mbuf **pcbopt;
684 	register struct mbuf *m;
685 {
686 	register cnt, optlen;
687 	register u_char *cp;
688 	u_char opt;
689 
690 	/* turn off any old options */
691 	if (*pcbopt)
692 		(void)m_free(*pcbopt);
693 	*pcbopt = 0;
694 	if (m == (struct mbuf *)0 || m->m_len == 0) {
695 		/*
696 		 * Only turning off any previous options.
697 		 */
698 		if (m)
699 			(void)m_free(m);
700 		return (0);
701 	}
702 
703 #ifndef	vax
704 	if (m->m_len % sizeof(int32_t))
705 		goto bad;
706 #endif
707 	/*
708 	 * IP first-hop destination address will be stored before
709 	 * actual options; move other options back
710 	 * and clear it when none present.
711 	 */
712 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
713 		goto bad;
714 	cnt = m->m_len;
715 	m->m_len += sizeof(struct in_addr);
716 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
717 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
718 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
719 
720 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
721 		opt = cp[IPOPT_OPTVAL];
722 		if (opt == IPOPT_EOL)
723 			break;
724 		if (opt == IPOPT_NOP)
725 			optlen = 1;
726 		else {
727 			optlen = cp[IPOPT_OLEN];
728 			if (optlen <= IPOPT_OLEN || optlen > cnt)
729 				goto bad;
730 		}
731 		switch (opt) {
732 
733 		default:
734 			break;
735 
736 		case IPOPT_LSRR:
737 		case IPOPT_SSRR:
738 			/*
739 			 * user process specifies route as:
740 			 *	->A->B->C->D
741 			 * D must be our final destination (but we can't
742 			 * check that since we may not have connected yet).
743 			 * A is first hop destination, which doesn't appear in
744 			 * actual IP option, but is stored before the options.
745 			 */
746 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
747 				goto bad;
748 			m->m_len -= sizeof(struct in_addr);
749 			cnt -= sizeof(struct in_addr);
750 			optlen -= sizeof(struct in_addr);
751 			cp[IPOPT_OLEN] = optlen;
752 			/*
753 			 * Move first hop before start of options.
754 			 */
755 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
756 			    sizeof(struct in_addr));
757 			/*
758 			 * Then copy rest of options back
759 			 * to close up the deleted entry.
760 			 */
761 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
762 			    sizeof(struct in_addr)),
763 			    (caddr_t)&cp[IPOPT_OFFSET+1],
764 			    (unsigned)cnt + sizeof(struct in_addr));
765 			break;
766 		}
767 	}
768 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
769 		goto bad;
770 	*pcbopt = m;
771 	return (0);
772 
773 bad:
774 	(void)m_free(m);
775 	return (EINVAL);
776 }
777 
778 /*
779  * Set the IP multicast options in response to user setsockopt().
780  */
781 int
782 ip_setmoptions(optname, imop, m)
783 	int optname;
784 	struct ip_moptions **imop;
785 	struct mbuf *m;
786 {
787 	register int error = 0;
788 	u_char loop;
789 	register int i;
790 	struct in_addr addr;
791 	register struct ip_mreq *mreq;
792 	register struct ifnet *ifp;
793 	register struct ip_moptions *imo = *imop;
794 	struct route ro;
795 	register struct sockaddr_in *dst;
796 
797 	if (imo == NULL) {
798 		/*
799 		 * No multicast option buffer attached to the pcb;
800 		 * allocate one and initialize to default values.
801 		 */
802 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
803 		    M_WAITOK);
804 
805 		if (imo == NULL)
806 			return (ENOBUFS);
807 		*imop = imo;
808 		imo->imo_multicast_ifp = NULL;
809 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
810 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
811 		imo->imo_num_memberships = 0;
812 	}
813 
814 	switch (optname) {
815 
816 	case IP_MULTICAST_IF:
817 		/*
818 		 * Select the interface for outgoing multicast packets.
819 		 */
820 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
821 			error = EINVAL;
822 			break;
823 		}
824 		addr = *(mtod(m, struct in_addr *));
825 		/*
826 		 * INADDR_ANY is used to remove a previous selection.
827 		 * When no interface is selected, a default one is
828 		 * chosen every time a multicast packet is sent.
829 		 */
830 		if (in_nullhost(addr)) {
831 			imo->imo_multicast_ifp = NULL;
832 			break;
833 		}
834 		/*
835 		 * The selected interface is identified by its local
836 		 * IP address.  Find the interface and confirm that
837 		 * it supports multicasting.
838 		 */
839 		INADDR_TO_IFP(addr, ifp);
840 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
841 			error = EADDRNOTAVAIL;
842 			break;
843 		}
844 		imo->imo_multicast_ifp = ifp;
845 		break;
846 
847 	case IP_MULTICAST_TTL:
848 		/*
849 		 * Set the IP time-to-live for outgoing multicast packets.
850 		 */
851 		if (m == NULL || m->m_len != 1) {
852 			error = EINVAL;
853 			break;
854 		}
855 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
856 		break;
857 
858 	case IP_MULTICAST_LOOP:
859 		/*
860 		 * Set the loopback flag for outgoing multicast packets.
861 		 * Must be zero or one.
862 		 */
863 		if (m == NULL || m->m_len != 1 ||
864 		   (loop = *(mtod(m, u_char *))) > 1) {
865 			error = EINVAL;
866 			break;
867 		}
868 		imo->imo_multicast_loop = loop;
869 		break;
870 
871 	case IP_ADD_MEMBERSHIP:
872 		/*
873 		 * Add 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(mreq->imr_multiaddr.s_addr)) {
882 			error = EINVAL;
883 			break;
884 		}
885 		/*
886 		 * If no interface address was provided, use the interface of
887 		 * the route to the given multicast address.
888 		 */
889 		if (in_nullhost(mreq->imr_interface)) {
890 			ro.ro_rt = NULL;
891 			dst = satosin(&ro.ro_dst);
892 			dst->sin_len = sizeof(*dst);
893 			dst->sin_family = AF_INET;
894 			dst->sin_addr = mreq->imr_multiaddr;
895 			rtalloc(&ro);
896 			if (ro.ro_rt == NULL) {
897 				error = EADDRNOTAVAIL;
898 				break;
899 			}
900 			ifp = ro.ro_rt->rt_ifp;
901 			rtfree(ro.ro_rt);
902 		} else {
903 			INADDR_TO_IFP(mreq->imr_interface, ifp);
904 		}
905 		/*
906 		 * See if we found an interface, and confirm that it
907 		 * supports multicast.
908 		 */
909 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
910 			error = EADDRNOTAVAIL;
911 			break;
912 		}
913 		/*
914 		 * See if the membership already exists or if all the
915 		 * membership slots are full.
916 		 */
917 		for (i = 0; i < imo->imo_num_memberships; ++i) {
918 			if (imo->imo_membership[i]->inm_ifp == ifp &&
919 			    in_hosteq(imo->imo_membership[i]->inm_addr,
920 				      mreq->imr_multiaddr))
921 				break;
922 		}
923 		if (i < imo->imo_num_memberships) {
924 			error = EADDRINUSE;
925 			break;
926 		}
927 		if (i == IP_MAX_MEMBERSHIPS) {
928 			error = ETOOMANYREFS;
929 			break;
930 		}
931 		/*
932 		 * Everything looks good; add a new record to the multicast
933 		 * address list for the given interface.
934 		 */
935 		if ((imo->imo_membership[i] =
936 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
937 			error = ENOBUFS;
938 			break;
939 		}
940 		++imo->imo_num_memberships;
941 		break;
942 
943 	case IP_DROP_MEMBERSHIP:
944 		/*
945 		 * Drop a multicast group membership.
946 		 * Group must be a valid IP multicast address.
947 		 */
948 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
949 			error = EINVAL;
950 			break;
951 		}
952 		mreq = mtod(m, struct ip_mreq *);
953 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
954 			error = EINVAL;
955 			break;
956 		}
957 		/*
958 		 * If an interface address was specified, get a pointer
959 		 * to its ifnet structure.
960 		 */
961 		if (in_nullhost(mreq->imr_interface))
962 			ifp = NULL;
963 		else {
964 			INADDR_TO_IFP(mreq->imr_interface, ifp);
965 			if (ifp == NULL) {
966 				error = EADDRNOTAVAIL;
967 				break;
968 			}
969 		}
970 		/*
971 		 * Find the membership in the membership array.
972 		 */
973 		for (i = 0; i < imo->imo_num_memberships; ++i) {
974 			if ((ifp == NULL ||
975 			     imo->imo_membership[i]->inm_ifp == ifp) &&
976 			     in_hosteq(imo->imo_membership[i]->inm_addr,
977 				       mreq->imr_multiaddr))
978 				break;
979 		}
980 		if (i == imo->imo_num_memberships) {
981 			error = EADDRNOTAVAIL;
982 			break;
983 		}
984 		/*
985 		 * Give up the multicast address record to which the
986 		 * membership points.
987 		 */
988 		in_delmulti(imo->imo_membership[i]);
989 		/*
990 		 * Remove the gap in the membership array.
991 		 */
992 		for (++i; i < imo->imo_num_memberships; ++i)
993 			imo->imo_membership[i-1] = imo->imo_membership[i];
994 		--imo->imo_num_memberships;
995 		break;
996 
997 	default:
998 		error = EOPNOTSUPP;
999 		break;
1000 	}
1001 
1002 	/*
1003 	 * If all options have default values, no need to keep the mbuf.
1004 	 */
1005 	if (imo->imo_multicast_ifp == NULL &&
1006 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1007 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1008 	    imo->imo_num_memberships == 0) {
1009 		free(*imop, M_IPMOPTS);
1010 		*imop = NULL;
1011 	}
1012 
1013 	return (error);
1014 }
1015 
1016 /*
1017  * Return the IP multicast options in response to user getsockopt().
1018  */
1019 int
1020 ip_getmoptions(optname, imo, mp)
1021 	int optname;
1022 	register struct ip_moptions *imo;
1023 	register struct mbuf **mp;
1024 {
1025 	u_char *ttl;
1026 	u_char *loop;
1027 	struct in_addr *addr;
1028 	struct in_ifaddr *ia;
1029 
1030 	*mp = m_get(M_WAIT, MT_SOOPTS);
1031 
1032 	switch (optname) {
1033 
1034 	case IP_MULTICAST_IF:
1035 		addr = mtod(*mp, struct in_addr *);
1036 		(*mp)->m_len = sizeof(struct in_addr);
1037 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1038 			*addr = zeroin_addr;
1039 		else {
1040 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1041 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1042 		}
1043 		return (0);
1044 
1045 	case IP_MULTICAST_TTL:
1046 		ttl = mtod(*mp, u_char *);
1047 		(*mp)->m_len = 1;
1048 		*ttl = imo ? imo->imo_multicast_ttl
1049 			   : IP_DEFAULT_MULTICAST_TTL;
1050 		return (0);
1051 
1052 	case IP_MULTICAST_LOOP:
1053 		loop = mtod(*mp, u_char *);
1054 		(*mp)->m_len = 1;
1055 		*loop = imo ? imo->imo_multicast_loop
1056 			    : IP_DEFAULT_MULTICAST_LOOP;
1057 		return (0);
1058 
1059 	default:
1060 		return (EOPNOTSUPP);
1061 	}
1062 }
1063 
1064 /*
1065  * Discard the IP multicast options.
1066  */
1067 void
1068 ip_freemoptions(imo)
1069 	register struct ip_moptions *imo;
1070 {
1071 	register int i;
1072 
1073 	if (imo != NULL) {
1074 		for (i = 0; i < imo->imo_num_memberships; ++i)
1075 			in_delmulti(imo->imo_membership[i]);
1076 		free(imo, M_IPMOPTS);
1077 	}
1078 }
1079 
1080 /*
1081  * Routine called from ip_output() to loop back a copy of an IP multicast
1082  * packet to the input queue of a specified interface.  Note that this
1083  * calls the output routine of the loopback "driver", but with an interface
1084  * pointer that might NOT be &loif -- easier than replicating that code here.
1085  */
1086 static void
1087 ip_mloopback(ifp, m, dst)
1088 	struct ifnet *ifp;
1089 	register struct mbuf *m;
1090 	register struct sockaddr_in *dst;
1091 {
1092 	register struct ip *ip;
1093 	struct mbuf *copym;
1094 
1095 	copym = m_copy(m, 0, M_COPYALL);
1096 	if (copym != NULL) {
1097 		/*
1098 		 * We don't bother to fragment if the IP length is greater
1099 		 * than the interface's MTU.  Can this possibly matter?
1100 		 */
1101 		ip = mtod(copym, struct ip *);
1102 		ip->ip_len = htons((u_int16_t)ip->ip_len);
1103 		ip->ip_off = htons((u_int16_t)ip->ip_off);
1104 		ip->ip_sum = 0;
1105 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1106 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1107 	}
1108 }
1109