xref: /netbsd-src/sys/netinet/ip_output.c (revision 81b108b45f75f89f1e3ffad9fb6f074e771c0935)
1 /*	$NetBSD: ip_output.c,v 1.32 1996/09/14 14:40:27 mrg 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 
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_var.h>
56 
57 #ifdef PFIL_HOOKS
58 #include <net/pfil.h>
59 #endif /* PFIL_HOOKS */
60 
61 #ifdef vax
62 #include <machine/mtpr.h>
63 #endif
64 
65 #include <machine/stdarg.h>
66 
67 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
68 static void ip_mloopback
69 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
70 
71 /*
72  * IP output.  The packet in mbuf chain m contains a skeletal IP
73  * header (with len, off, ttl, proto, tos, src, dst).
74  * The mbuf chain containing the packet will be freed.
75  * The mbuf opt, if present, will not be freed.
76  */
77 int
78 #if __STDC__
79 ip_output(struct mbuf *m0, ...)
80 #else
81 ip_output(m0, va_alist)
82 	struct mbuf *m0;
83 	va_dcl
84 #endif
85 {
86 	register struct ip *ip, *mhip;
87 	register struct ifnet *ifp;
88 	register struct mbuf *m = m0;
89 	register int hlen = sizeof (struct ip);
90 	int len, off, error = 0;
91 	struct route iproute;
92 	struct sockaddr_in *dst;
93 	struct in_ifaddr *ia;
94 	struct mbuf *opt;
95 	struct route *ro;
96 	int flags;
97 	struct ip_moptions *imo;
98 	va_list ap;
99 #ifdef PFIL_HOOKS
100 	struct packet_filter_hook *pfh;
101 	struct mbuf *m1;
102 #endif /* PFIL_HOOKS */
103 
104 	va_start(ap, m0);
105 	opt = va_arg(ap, struct mbuf *);
106 	ro = va_arg(ap, struct route *);
107 	flags = va_arg(ap, int);
108 	imo = va_arg(ap, struct ip_moptions *);
109 	va_end(ap);
110 
111 #ifdef	DIAGNOSTIC
112 	if ((m->m_flags & M_PKTHDR) == 0)
113 		panic("ip_output no HDR");
114 #endif
115 	if (opt) {
116 		m = ip_insertoptions(m, opt, &len);
117 		hlen = len;
118 	}
119 	ip = mtod(m, struct ip *);
120 	/*
121 	 * Fill in IP header.
122 	 */
123 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
124 		ip->ip_v = IPVERSION;
125 		ip->ip_off &= IP_DF;
126 		ip->ip_id = htons(ip_id++);
127 		ip->ip_hl = hlen >> 2;
128 		ipstat.ips_localout++;
129 	} else {
130 		hlen = ip->ip_hl << 2;
131 	}
132 	/*
133 	 * Route packet.
134 	 */
135 	if (ro == 0) {
136 		ro = &iproute;
137 		bzero((caddr_t)ro, sizeof (*ro));
138 	}
139 	dst = satosin(&ro->ro_dst);
140 	/*
141 	 * If there is a cached route,
142 	 * check that it is to the same destination
143 	 * and is still up.  If not, free it and try again.
144 	 */
145 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
146 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
147 		RTFREE(ro->ro_rt);
148 		ro->ro_rt = (struct rtentry *)0;
149 	}
150 	if (ro->ro_rt == 0) {
151 		dst->sin_family = AF_INET;
152 		dst->sin_len = sizeof(*dst);
153 		dst->sin_addr = ip->ip_dst;
154 	}
155 	/*
156 	 * If routing to interface only,
157 	 * short circuit routing lookup.
158 	 */
159 	if (flags & IP_ROUTETOIF) {
160 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
161 			ipstat.ips_noroute++;
162 			error = ENETUNREACH;
163 			goto bad;
164 		}
165 		ifp = ia->ia_ifp;
166 		ip->ip_ttl = 1;
167 	} else {
168 		if (ro->ro_rt == 0)
169 			rtalloc(ro);
170 		if (ro->ro_rt == 0) {
171 			ipstat.ips_noroute++;
172 			error = EHOSTUNREACH;
173 			goto bad;
174 		}
175 		ia = ifatoia(ro->ro_rt->rt_ifa);
176 		ifp = ro->ro_rt->rt_ifp;
177 		ro->ro_rt->rt_use++;
178 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
179 			dst = satosin(ro->ro_rt->rt_gateway);
180 	}
181 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
182 		struct in_multi *inm;
183 
184 		m->m_flags |= M_MCAST;
185 		/*
186 		 * IP destination address is multicast.  Make sure "dst"
187 		 * still points to the address in "ro".  (It may have been
188 		 * changed to point to a gateway address, above.)
189 		 */
190 		dst = satosin(&ro->ro_dst);
191 		/*
192 		 * See if the caller provided any multicast options
193 		 */
194 		if (imo != NULL) {
195 			ip->ip_ttl = imo->imo_multicast_ttl;
196 			if (imo->imo_multicast_ifp != NULL)
197 				ifp = imo->imo_multicast_ifp;
198 		} else
199 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
200 		/*
201 		 * Confirm that the outgoing interface supports multicast.
202 		 */
203 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
204 			ipstat.ips_noroute++;
205 			error = ENETUNREACH;
206 			goto bad;
207 		}
208 		/*
209 		 * If source address not specified yet, use address
210 		 * of outgoing interface.
211 		 */
212 		if (in_nullhost(ip->ip_src)) {
213 			register struct in_ifaddr *ia;
214 
215 			for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next)
216 				if (ia->ia_ifp == ifp) {
217 					ip->ip_src = ia->ia_addr.sin_addr;
218 					break;
219 				}
220 		}
221 
222 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
223 		if (inm != NULL &&
224 		   (imo == NULL || imo->imo_multicast_loop)) {
225 			/*
226 			 * If we belong to the destination multicast group
227 			 * on the outgoing interface, and the caller did not
228 			 * forbid loopback, loop back a copy.
229 			 */
230 			ip_mloopback(ifp, m, dst);
231 		}
232 #ifdef MROUTING
233 		else {
234 			/*
235 			 * If we are acting as a multicast router, perform
236 			 * multicast forwarding as if the packet had just
237 			 * arrived on the interface to which we are about
238 			 * to send.  The multicast forwarding function
239 			 * recursively calls this function, using the
240 			 * IP_FORWARDING flag to prevent infinite recursion.
241 			 *
242 			 * Multicasts that are looped back by ip_mloopback(),
243 			 * above, will be forwarded by the ip_input() routine,
244 			 * if necessary.
245 			 */
246 			extern struct socket *ip_mrouter;
247 
248 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
249 				if (ip_mforward(m, ifp) != 0) {
250 					m_freem(m);
251 					goto done;
252 				}
253 			}
254 		}
255 #endif
256 		/*
257 		 * Multicasts with a time-to-live of zero may be looped-
258 		 * back, above, but must not be transmitted on a network.
259 		 * Also, multicasts addressed to the loopback interface
260 		 * are not sent -- the above call to ip_mloopback() will
261 		 * loop back a copy if this host actually belongs to the
262 		 * destination group on the loopback interface.
263 		 */
264 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
265 			m_freem(m);
266 			goto done;
267 		}
268 
269 		goto sendit;
270 	}
271 #ifndef notdef
272 	/*
273 	 * If source address not specified yet, use address
274 	 * of outgoing interface.
275 	 */
276 	if (in_nullhost(ip->ip_src))
277 		ip->ip_src = ia->ia_addr.sin_addr;
278 #endif
279 	/*
280 	 * Look for broadcast address and
281 	 * and verify user is allowed to send
282 	 * such a packet.
283 	 */
284 	if (in_broadcast(dst->sin_addr, ifp)) {
285 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
286 			error = EADDRNOTAVAIL;
287 			goto bad;
288 		}
289 		if ((flags & IP_ALLOWBROADCAST) == 0) {
290 			error = EACCES;
291 			goto bad;
292 		}
293 		/* don't allow broadcast messages to be fragmented */
294 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
295 			error = EMSGSIZE;
296 			goto bad;
297 		}
298 		m->m_flags |= M_BCAST;
299 	} else
300 		m->m_flags &= ~M_BCAST;
301 
302 #ifdef PFIL_HOOKS
303 	/*
304 	 * Run through list of hooks for output packets.
305 	 */
306 	m1 = m;
307 	for (pfh = pfil_hook_get(PFIL_OUT); pfh; pfh = pfh->pfil_link.le_next)
308 		if (pfh->pfil_func) {
309 		    	if (pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 1, &m1)) {
310 				error = EHOSTUNREACH;
311 				goto bad;
312 			}
313 			ip = mtod(m = m1, struct ip *);
314 		}
315 #endif /* PFIL_HOOKS */
316 sendit:
317 	/*
318 	 * If small enough for interface, can just send directly.
319 	 */
320 	if ((u_int16_t)ip->ip_len <= ifp->if_mtu) {
321 		ip->ip_len = htons((u_int16_t)ip->ip_len);
322 		ip->ip_off = htons((u_int16_t)ip->ip_off);
323 		ip->ip_sum = 0;
324 		ip->ip_sum = in_cksum(m, hlen);
325 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
326 		goto done;
327 	}
328 	/*
329 	 * Too large for interface; fragment if possible.
330 	 * Must be able to put at least 8 bytes per fragment.
331 	 */
332 	if (ip->ip_off & IP_DF) {
333 		error = EMSGSIZE;
334 		ipstat.ips_cantfrag++;
335 		goto bad;
336 	}
337 	len = (ifp->if_mtu - hlen) &~ 7;
338 	if (len < 8) {
339 		error = EMSGSIZE;
340 		goto bad;
341 	}
342 
343     {
344 	int mhlen, firstlen = len;
345 	struct mbuf **mnext = &m->m_nextpkt;
346 
347 	/*
348 	 * Loop through length of segment after first fragment,
349 	 * make new header and copy data of each part and link onto chain.
350 	 */
351 	m0 = m;
352 	mhlen = sizeof (struct ip);
353 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
354 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
355 		if (m == 0) {
356 			error = ENOBUFS;
357 			ipstat.ips_odropped++;
358 			goto sendorfree;
359 		}
360 		*mnext = m;
361 		mnext = &m->m_nextpkt;
362 		m->m_data += max_linkhdr;
363 		mhip = mtod(m, struct ip *);
364 		*mhip = *ip;
365 		if (hlen > sizeof (struct ip)) {
366 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
367 			mhip->ip_hl = mhlen >> 2;
368 		}
369 		m->m_len = mhlen;
370 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
371 		if (ip->ip_off & IP_MF)
372 			mhip->ip_off |= IP_MF;
373 		if (off + len >= (u_int16_t)ip->ip_len)
374 			len = (u_int16_t)ip->ip_len - off;
375 		else
376 			mhip->ip_off |= IP_MF;
377 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
378 		m->m_next = m_copy(m0, off, len);
379 		if (m->m_next == 0) {
380 			error = ENOBUFS;	/* ??? */
381 			ipstat.ips_odropped++;
382 			goto sendorfree;
383 		}
384 		m->m_pkthdr.len = mhlen + len;
385 		m->m_pkthdr.rcvif = (struct ifnet *)0;
386 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
387 		mhip->ip_sum = 0;
388 		mhip->ip_sum = in_cksum(m, mhlen);
389 		ipstat.ips_ofragments++;
390 	}
391 	/*
392 	 * Update first fragment by trimming what's been copied out
393 	 * and updating header, then send each fragment (in order).
394 	 */
395 	m = m0;
396 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
397 	m->m_pkthdr.len = hlen + firstlen;
398 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
399 	ip->ip_off = htons((u_int16_t)(ip->ip_off | IP_MF));
400 	ip->ip_sum = 0;
401 	ip->ip_sum = in_cksum(m, hlen);
402 sendorfree:
403 	for (m = m0; m; m = m0) {
404 		m0 = m->m_nextpkt;
405 		m->m_nextpkt = 0;
406 		if (error == 0)
407 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
408 			    ro->ro_rt);
409 		else
410 			m_freem(m);
411 	}
412 
413 	if (error == 0)
414 		ipstat.ips_fragmented++;
415     }
416 done:
417 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
418 		RTFREE(ro->ro_rt);
419 		ro->ro_rt = 0;
420 	}
421 	return (error);
422 bad:
423 #ifdef PFIL_HOOKS
424 	m1 = m;
425 	for (pfh = pfil_hook_get(PFIL_BAD); pfh; pfh = pfh->pfil_link.le_next)
426 		if (pfh->pfil_func) {
427 			(void)pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 2, &m1);
428 			ip = mtod(m = m1, struct ip *);
429 		}
430 #endif /* PFIL_HOOKS */
431 	m_freem(m0);
432 	goto done;
433 }
434 
435 /*
436  * Insert IP options into preformed packet.
437  * Adjust IP destination as required for IP source routing,
438  * as indicated by a non-zero in_addr at the start of the options.
439  */
440 static struct mbuf *
441 ip_insertoptions(m, opt, phlen)
442 	register struct mbuf *m;
443 	struct mbuf *opt;
444 	int *phlen;
445 {
446 	register struct ipoption *p = mtod(opt, struct ipoption *);
447 	struct mbuf *n;
448 	register struct ip *ip = mtod(m, struct ip *);
449 	unsigned optlen;
450 
451 	optlen = opt->m_len - sizeof(p->ipopt_dst);
452 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
453 		return (m);		/* XXX should fail */
454 	if (!in_nullhost(p->ipopt_dst))
455 		ip->ip_dst = p->ipopt_dst;
456 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
457 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
458 		if (n == 0)
459 			return (m);
460 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
461 		m->m_len -= sizeof(struct ip);
462 		m->m_data += sizeof(struct ip);
463 		n->m_next = m;
464 		m = n;
465 		m->m_len = optlen + sizeof(struct ip);
466 		m->m_data += max_linkhdr;
467 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
468 	} else {
469 		m->m_data -= optlen;
470 		m->m_len += optlen;
471 		m->m_pkthdr.len += optlen;
472 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
473 	}
474 	ip = mtod(m, struct ip *);
475 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
476 	*phlen = sizeof(struct ip) + optlen;
477 	ip->ip_len += optlen;
478 	return (m);
479 }
480 
481 /*
482  * Copy options from ip to jp,
483  * omitting those not copied during fragmentation.
484  */
485 int
486 ip_optcopy(ip, jp)
487 	struct ip *ip, *jp;
488 {
489 	register u_char *cp, *dp;
490 	int opt, optlen, cnt;
491 
492 	cp = (u_char *)(ip + 1);
493 	dp = (u_char *)(jp + 1);
494 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
495 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
496 		opt = cp[0];
497 		if (opt == IPOPT_EOL)
498 			break;
499 		if (opt == IPOPT_NOP) {
500 			/* Preserve for IP mcast tunnel's LSRR alignment. */
501 			*dp++ = IPOPT_NOP;
502 			optlen = 1;
503 			continue;
504 		} else
505 			optlen = cp[IPOPT_OLEN];
506 		/* bogus lengths should have been caught by ip_dooptions */
507 		if (optlen > cnt)
508 			optlen = cnt;
509 		if (IPOPT_COPIED(opt)) {
510 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
511 			dp += optlen;
512 		}
513 	}
514 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
515 		*dp++ = IPOPT_EOL;
516 	return (optlen);
517 }
518 
519 /*
520  * IP socket option processing.
521  */
522 int
523 ip_ctloutput(op, so, level, optname, mp)
524 	int op;
525 	struct socket *so;
526 	int level, optname;
527 	struct mbuf **mp;
528 {
529 	register struct inpcb *inp = sotoinpcb(so);
530 	register struct mbuf *m = *mp;
531 	register int optval = 0;
532 	int error = 0;
533 
534 	if (level != IPPROTO_IP) {
535 		error = EINVAL;
536 		if (op == PRCO_SETOPT && *mp)
537 			(void) m_free(*mp);
538 	} else switch (op) {
539 
540 	case PRCO_SETOPT:
541 		switch (optname) {
542 		case IP_OPTIONS:
543 #ifdef notyet
544 		case IP_RETOPTS:
545 			return (ip_pcbopts(optname, &inp->inp_options, m));
546 #else
547 			return (ip_pcbopts(&inp->inp_options, m));
548 #endif
549 
550 		case IP_TOS:
551 		case IP_TTL:
552 		case IP_RECVOPTS:
553 		case IP_RECVRETOPTS:
554 		case IP_RECVDSTADDR:
555 			if (m == NULL || m->m_len != sizeof(int))
556 				error = EINVAL;
557 			else {
558 				optval = *mtod(m, int *);
559 				switch (optname) {
560 
561 				case IP_TOS:
562 					inp->inp_ip.ip_tos = optval;
563 					break;
564 
565 				case IP_TTL:
566 					inp->inp_ip.ip_ttl = optval;
567 					break;
568 #define	OPTSET(bit) \
569 	if (optval) \
570 		inp->inp_flags |= bit; \
571 	else \
572 		inp->inp_flags &= ~bit;
573 
574 				case IP_RECVOPTS:
575 					OPTSET(INP_RECVOPTS);
576 					break;
577 
578 				case IP_RECVRETOPTS:
579 					OPTSET(INP_RECVRETOPTS);
580 					break;
581 
582 				case IP_RECVDSTADDR:
583 					OPTSET(INP_RECVDSTADDR);
584 					break;
585 				}
586 			}
587 			break;
588 #undef OPTSET
589 
590 		case IP_MULTICAST_IF:
591 		case IP_MULTICAST_TTL:
592 		case IP_MULTICAST_LOOP:
593 		case IP_ADD_MEMBERSHIP:
594 		case IP_DROP_MEMBERSHIP:
595 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
596 			break;
597 
598 		default:
599 			error = ENOPROTOOPT;
600 			break;
601 		}
602 		if (m)
603 			(void)m_free(m);
604 		break;
605 
606 	case PRCO_GETOPT:
607 		switch (optname) {
608 		case IP_OPTIONS:
609 		case IP_RETOPTS:
610 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
611 			if (inp->inp_options) {
612 				m->m_len = inp->inp_options->m_len;
613 				bcopy(mtod(inp->inp_options, caddr_t),
614 				    mtod(m, caddr_t), (unsigned)m->m_len);
615 			} else
616 				m->m_len = 0;
617 			break;
618 
619 		case IP_TOS:
620 		case IP_TTL:
621 		case IP_RECVOPTS:
622 		case IP_RECVRETOPTS:
623 		case IP_RECVDSTADDR:
624 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
625 			m->m_len = sizeof(int);
626 			switch (optname) {
627 
628 			case IP_TOS:
629 				optval = inp->inp_ip.ip_tos;
630 				break;
631 
632 			case IP_TTL:
633 				optval = inp->inp_ip.ip_ttl;
634 				break;
635 
636 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
637 
638 			case IP_RECVOPTS:
639 				optval = OPTBIT(INP_RECVOPTS);
640 				break;
641 
642 			case IP_RECVRETOPTS:
643 				optval = OPTBIT(INP_RECVRETOPTS);
644 				break;
645 
646 			case IP_RECVDSTADDR:
647 				optval = OPTBIT(INP_RECVDSTADDR);
648 				break;
649 			}
650 			*mtod(m, int *) = optval;
651 			break;
652 
653 		case IP_MULTICAST_IF:
654 		case IP_MULTICAST_TTL:
655 		case IP_MULTICAST_LOOP:
656 		case IP_ADD_MEMBERSHIP:
657 		case IP_DROP_MEMBERSHIP:
658 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
659 			break;
660 
661 		default:
662 			error = ENOPROTOOPT;
663 			break;
664 		}
665 		break;
666 	}
667 	return (error);
668 }
669 
670 /*
671  * Set up IP options in pcb for insertion in output packets.
672  * Store in mbuf with pointer in pcbopt, adding pseudo-option
673  * with destination address if source routed.
674  */
675 int
676 #ifdef notyet
677 ip_pcbopts(optname, pcbopt, m)
678 	int optname;
679 #else
680 ip_pcbopts(pcbopt, m)
681 #endif
682 	struct mbuf **pcbopt;
683 	register struct mbuf *m;
684 {
685 	register cnt, optlen;
686 	register u_char *cp;
687 	u_char opt;
688 
689 	/* turn off any old options */
690 	if (*pcbopt)
691 		(void)m_free(*pcbopt);
692 	*pcbopt = 0;
693 	if (m == (struct mbuf *)0 || m->m_len == 0) {
694 		/*
695 		 * Only turning off any previous options.
696 		 */
697 		if (m)
698 			(void)m_free(m);
699 		return (0);
700 	}
701 
702 #ifndef	vax
703 	if (m->m_len % sizeof(int32_t))
704 		goto bad;
705 #endif
706 	/*
707 	 * IP first-hop destination address will be stored before
708 	 * actual options; move other options back
709 	 * and clear it when none present.
710 	 */
711 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
712 		goto bad;
713 	cnt = m->m_len;
714 	m->m_len += sizeof(struct in_addr);
715 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
716 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
717 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
718 
719 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
720 		opt = cp[IPOPT_OPTVAL];
721 		if (opt == IPOPT_EOL)
722 			break;
723 		if (opt == IPOPT_NOP)
724 			optlen = 1;
725 		else {
726 			optlen = cp[IPOPT_OLEN];
727 			if (optlen <= IPOPT_OLEN || optlen > cnt)
728 				goto bad;
729 		}
730 		switch (opt) {
731 
732 		default:
733 			break;
734 
735 		case IPOPT_LSRR:
736 		case IPOPT_SSRR:
737 			/*
738 			 * user process specifies route as:
739 			 *	->A->B->C->D
740 			 * D must be our final destination (but we can't
741 			 * check that since we may not have connected yet).
742 			 * A is first hop destination, which doesn't appear in
743 			 * actual IP option, but is stored before the options.
744 			 */
745 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
746 				goto bad;
747 			m->m_len -= sizeof(struct in_addr);
748 			cnt -= sizeof(struct in_addr);
749 			optlen -= sizeof(struct in_addr);
750 			cp[IPOPT_OLEN] = optlen;
751 			/*
752 			 * Move first hop before start of options.
753 			 */
754 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
755 			    sizeof(struct in_addr));
756 			/*
757 			 * Then copy rest of options back
758 			 * to close up the deleted entry.
759 			 */
760 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
761 			    sizeof(struct in_addr)),
762 			    (caddr_t)&cp[IPOPT_OFFSET+1],
763 			    (unsigned)cnt + sizeof(struct in_addr));
764 			break;
765 		}
766 	}
767 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
768 		goto bad;
769 	*pcbopt = m;
770 	return (0);
771 
772 bad:
773 	(void)m_free(m);
774 	return (EINVAL);
775 }
776 
777 /*
778  * Set the IP multicast options in response to user setsockopt().
779  */
780 int
781 ip_setmoptions(optname, imop, m)
782 	int optname;
783 	struct ip_moptions **imop;
784 	struct mbuf *m;
785 {
786 	register int error = 0;
787 	u_char loop;
788 	register int i;
789 	struct in_addr addr;
790 	register struct ip_mreq *mreq;
791 	register struct ifnet *ifp;
792 	register struct ip_moptions *imo = *imop;
793 	struct route ro;
794 	register struct sockaddr_in *dst;
795 
796 	if (imo == NULL) {
797 		/*
798 		 * No multicast option buffer attached to the pcb;
799 		 * allocate one and initialize to default values.
800 		 */
801 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
802 		    M_WAITOK);
803 
804 		if (imo == NULL)
805 			return (ENOBUFS);
806 		*imop = imo;
807 		imo->imo_multicast_ifp = NULL;
808 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
809 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
810 		imo->imo_num_memberships = 0;
811 	}
812 
813 	switch (optname) {
814 
815 	case IP_MULTICAST_IF:
816 		/*
817 		 * Select the interface for outgoing multicast packets.
818 		 */
819 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
820 			error = EINVAL;
821 			break;
822 		}
823 		addr = *(mtod(m, struct in_addr *));
824 		/*
825 		 * INADDR_ANY is used to remove a previous selection.
826 		 * When no interface is selected, a default one is
827 		 * chosen every time a multicast packet is sent.
828 		 */
829 		if (in_nullhost(addr)) {
830 			imo->imo_multicast_ifp = NULL;
831 			break;
832 		}
833 		/*
834 		 * The selected interface is identified by its local
835 		 * IP address.  Find the interface and confirm that
836 		 * it supports multicasting.
837 		 */
838 		INADDR_TO_IFP(addr, ifp);
839 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
840 			error = EADDRNOTAVAIL;
841 			break;
842 		}
843 		imo->imo_multicast_ifp = ifp;
844 		break;
845 
846 	case IP_MULTICAST_TTL:
847 		/*
848 		 * Set the IP time-to-live for outgoing multicast packets.
849 		 */
850 		if (m == NULL || m->m_len != 1) {
851 			error = EINVAL;
852 			break;
853 		}
854 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
855 		break;
856 
857 	case IP_MULTICAST_LOOP:
858 		/*
859 		 * Set the loopback flag for outgoing multicast packets.
860 		 * Must be zero or one.
861 		 */
862 		if (m == NULL || m->m_len != 1 ||
863 		   (loop = *(mtod(m, u_char *))) > 1) {
864 			error = EINVAL;
865 			break;
866 		}
867 		imo->imo_multicast_loop = loop;
868 		break;
869 
870 	case IP_ADD_MEMBERSHIP:
871 		/*
872 		 * Add a multicast group membership.
873 		 * Group must be a valid IP multicast address.
874 		 */
875 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
876 			error = EINVAL;
877 			break;
878 		}
879 		mreq = mtod(m, struct ip_mreq *);
880 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
881 			error = EINVAL;
882 			break;
883 		}
884 		/*
885 		 * If no interface address was provided, use the interface of
886 		 * the route to the given multicast address.
887 		 */
888 		if (in_nullhost(mreq->imr_interface)) {
889 			ro.ro_rt = NULL;
890 			dst = satosin(&ro.ro_dst);
891 			dst->sin_len = sizeof(*dst);
892 			dst->sin_family = AF_INET;
893 			dst->sin_addr = mreq->imr_multiaddr;
894 			rtalloc(&ro);
895 			if (ro.ro_rt == NULL) {
896 				error = EADDRNOTAVAIL;
897 				break;
898 			}
899 			ifp = ro.ro_rt->rt_ifp;
900 			rtfree(ro.ro_rt);
901 		} else {
902 			INADDR_TO_IFP(mreq->imr_interface, ifp);
903 		}
904 		/*
905 		 * See if we found an interface, and confirm that it
906 		 * supports multicast.
907 		 */
908 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
909 			error = EADDRNOTAVAIL;
910 			break;
911 		}
912 		/*
913 		 * See if the membership already exists or if all the
914 		 * membership slots are full.
915 		 */
916 		for (i = 0; i < imo->imo_num_memberships; ++i) {
917 			if (imo->imo_membership[i]->inm_ifp == ifp &&
918 			    in_hosteq(imo->imo_membership[i]->inm_addr,
919 				      mreq->imr_multiaddr))
920 				break;
921 		}
922 		if (i < imo->imo_num_memberships) {
923 			error = EADDRINUSE;
924 			break;
925 		}
926 		if (i == IP_MAX_MEMBERSHIPS) {
927 			error = ETOOMANYREFS;
928 			break;
929 		}
930 		/*
931 		 * Everything looks good; add a new record to the multicast
932 		 * address list for the given interface.
933 		 */
934 		if ((imo->imo_membership[i] =
935 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
936 			error = ENOBUFS;
937 			break;
938 		}
939 		++imo->imo_num_memberships;
940 		break;
941 
942 	case IP_DROP_MEMBERSHIP:
943 		/*
944 		 * Drop a multicast group membership.
945 		 * Group must be a valid IP multicast address.
946 		 */
947 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
948 			error = EINVAL;
949 			break;
950 		}
951 		mreq = mtod(m, struct ip_mreq *);
952 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
953 			error = EINVAL;
954 			break;
955 		}
956 		/*
957 		 * If an interface address was specified, get a pointer
958 		 * to its ifnet structure.
959 		 */
960 		if (in_nullhost(mreq->imr_interface))
961 			ifp = NULL;
962 		else {
963 			INADDR_TO_IFP(mreq->imr_interface, ifp);
964 			if (ifp == NULL) {
965 				error = EADDRNOTAVAIL;
966 				break;
967 			}
968 		}
969 		/*
970 		 * Find the membership in the membership array.
971 		 */
972 		for (i = 0; i < imo->imo_num_memberships; ++i) {
973 			if ((ifp == NULL ||
974 			     imo->imo_membership[i]->inm_ifp == ifp) &&
975 			     in_hosteq(imo->imo_membership[i]->inm_addr,
976 				       mreq->imr_multiaddr))
977 				break;
978 		}
979 		if (i == imo->imo_num_memberships) {
980 			error = EADDRNOTAVAIL;
981 			break;
982 		}
983 		/*
984 		 * Give up the multicast address record to which the
985 		 * membership points.
986 		 */
987 		in_delmulti(imo->imo_membership[i]);
988 		/*
989 		 * Remove the gap in the membership array.
990 		 */
991 		for (++i; i < imo->imo_num_memberships; ++i)
992 			imo->imo_membership[i-1] = imo->imo_membership[i];
993 		--imo->imo_num_memberships;
994 		break;
995 
996 	default:
997 		error = EOPNOTSUPP;
998 		break;
999 	}
1000 
1001 	/*
1002 	 * If all options have default values, no need to keep the mbuf.
1003 	 */
1004 	if (imo->imo_multicast_ifp == NULL &&
1005 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1006 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1007 	    imo->imo_num_memberships == 0) {
1008 		free(*imop, M_IPMOPTS);
1009 		*imop = NULL;
1010 	}
1011 
1012 	return (error);
1013 }
1014 
1015 /*
1016  * Return the IP multicast options in response to user getsockopt().
1017  */
1018 int
1019 ip_getmoptions(optname, imo, mp)
1020 	int optname;
1021 	register struct ip_moptions *imo;
1022 	register struct mbuf **mp;
1023 {
1024 	u_char *ttl;
1025 	u_char *loop;
1026 	struct in_addr *addr;
1027 	struct in_ifaddr *ia;
1028 
1029 	*mp = m_get(M_WAIT, MT_SOOPTS);
1030 
1031 	switch (optname) {
1032 
1033 	case IP_MULTICAST_IF:
1034 		addr = mtod(*mp, struct in_addr *);
1035 		(*mp)->m_len = sizeof(struct in_addr);
1036 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1037 			*addr = zeroin_addr;
1038 		else {
1039 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1040 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1041 		}
1042 		return (0);
1043 
1044 	case IP_MULTICAST_TTL:
1045 		ttl = mtod(*mp, u_char *);
1046 		(*mp)->m_len = 1;
1047 		*ttl = imo ? imo->imo_multicast_ttl
1048 			   : IP_DEFAULT_MULTICAST_TTL;
1049 		return (0);
1050 
1051 	case IP_MULTICAST_LOOP:
1052 		loop = mtod(*mp, u_char *);
1053 		(*mp)->m_len = 1;
1054 		*loop = imo ? imo->imo_multicast_loop
1055 			    : IP_DEFAULT_MULTICAST_LOOP;
1056 		return (0);
1057 
1058 	default:
1059 		return (EOPNOTSUPP);
1060 	}
1061 }
1062 
1063 /*
1064  * Discard the IP multicast options.
1065  */
1066 void
1067 ip_freemoptions(imo)
1068 	register struct ip_moptions *imo;
1069 {
1070 	register int i;
1071 
1072 	if (imo != NULL) {
1073 		for (i = 0; i < imo->imo_num_memberships; ++i)
1074 			in_delmulti(imo->imo_membership[i]);
1075 		free(imo, M_IPMOPTS);
1076 	}
1077 }
1078 
1079 /*
1080  * Routine called from ip_output() to loop back a copy of an IP multicast
1081  * packet to the input queue of a specified interface.  Note that this
1082  * calls the output routine of the loopback "driver", but with an interface
1083  * pointer that might NOT be &loif -- easier than replicating that code here.
1084  */
1085 static void
1086 ip_mloopback(ifp, m, dst)
1087 	struct ifnet *ifp;
1088 	register struct mbuf *m;
1089 	register struct sockaddr_in *dst;
1090 {
1091 	register struct ip *ip;
1092 	struct mbuf *copym;
1093 
1094 	copym = m_copy(m, 0, M_COPYALL);
1095 	if (copym != NULL) {
1096 		/*
1097 		 * We don't bother to fragment if the IP length is greater
1098 		 * than the interface's MTU.  Can this possibly matter?
1099 		 */
1100 		ip = mtod(copym, struct ip *);
1101 		ip->ip_len = htons((u_int16_t)ip->ip_len);
1102 		ip->ip_off = htons((u_int16_t)ip->ip_off);
1103 		ip->ip_sum = 0;
1104 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1105 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1106 	}
1107 }
1108