xref: /netbsd-src/sys/netinet/ip_output.c (revision 76dfffe33547c37f8bdd446e3e4ab0f3c16cea4b)
1 /*	$NetBSD: ip_output.c,v 1.34 1996/10/22 11:27:07 veego 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 done;
312 			} else {
313 				ip = mtod(m = m1, struct ip *);
314 			}
315 		}
316 #endif /* PFIL_HOOKS */
317 sendit:
318 	/*
319 	 * If small enough for interface, can just send directly.
320 	 */
321 	if ((u_int16_t)ip->ip_len <= ifp->if_mtu) {
322 		ip->ip_len = htons((u_int16_t)ip->ip_len);
323 		ip->ip_off = htons((u_int16_t)ip->ip_off);
324 		ip->ip_sum = 0;
325 		ip->ip_sum = in_cksum(m, hlen);
326 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
327 		goto done;
328 	}
329 	/*
330 	 * Too large for interface; fragment if possible.
331 	 * Must be able to put at least 8 bytes per fragment.
332 	 */
333 	if (ip->ip_off & IP_DF) {
334 		error = EMSGSIZE;
335 		ipstat.ips_cantfrag++;
336 		goto bad;
337 	}
338 	len = (ifp->if_mtu - hlen) &~ 7;
339 	if (len < 8) {
340 		error = EMSGSIZE;
341 		goto bad;
342 	}
343 
344     {
345 	int mhlen, firstlen = len;
346 	struct mbuf **mnext = &m->m_nextpkt;
347 
348 	/*
349 	 * Loop through length of segment after first fragment,
350 	 * make new header and copy data of each part and link onto chain.
351 	 */
352 	m0 = m;
353 	mhlen = sizeof (struct ip);
354 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
355 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
356 		if (m == 0) {
357 			error = ENOBUFS;
358 			ipstat.ips_odropped++;
359 			goto sendorfree;
360 		}
361 		*mnext = m;
362 		mnext = &m->m_nextpkt;
363 		m->m_data += max_linkhdr;
364 		mhip = mtod(m, struct ip *);
365 		*mhip = *ip;
366 		if (hlen > sizeof (struct ip)) {
367 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
368 			mhip->ip_hl = mhlen >> 2;
369 		}
370 		m->m_len = mhlen;
371 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
372 		if (ip->ip_off & IP_MF)
373 			mhip->ip_off |= IP_MF;
374 		if (off + len >= (u_int16_t)ip->ip_len)
375 			len = (u_int16_t)ip->ip_len - off;
376 		else
377 			mhip->ip_off |= IP_MF;
378 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
379 		m->m_next = m_copy(m0, off, len);
380 		if (m->m_next == 0) {
381 			error = ENOBUFS;	/* ??? */
382 			ipstat.ips_odropped++;
383 			goto sendorfree;
384 		}
385 		m->m_pkthdr.len = mhlen + len;
386 		m->m_pkthdr.rcvif = (struct ifnet *)0;
387 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
388 		mhip->ip_sum = 0;
389 		mhip->ip_sum = in_cksum(m, mhlen);
390 		ipstat.ips_ofragments++;
391 	}
392 	/*
393 	 * Update first fragment by trimming what's been copied out
394 	 * and updating header, then send each fragment (in order).
395 	 */
396 	m = m0;
397 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
398 	m->m_pkthdr.len = hlen + firstlen;
399 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
400 	ip->ip_off = htons((u_int16_t)(ip->ip_off | IP_MF));
401 	ip->ip_sum = 0;
402 	ip->ip_sum = in_cksum(m, hlen);
403 sendorfree:
404 	for (m = m0; m; m = m0) {
405 		m0 = m->m_nextpkt;
406 		m->m_nextpkt = 0;
407 		if (error == 0)
408 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
409 			    ro->ro_rt);
410 		else
411 			m_freem(m);
412 	}
413 
414 	if (error == 0)
415 		ipstat.ips_fragmented++;
416     }
417 done:
418 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
419 		RTFREE(ro->ro_rt);
420 		ro->ro_rt = 0;
421 	}
422 	return (error);
423 bad:
424 #ifdef PFIL_HOOKS
425 	m1 = m;
426 	for (pfh = pfil_hook_get(PFIL_BAD); pfh; pfh = pfh->pfil_link.le_next)
427 		if (pfh->pfil_func) {
428 			(void)pfh->pfil_func(ip, hlen, m->m_pkthdr.rcvif, 2, &m1);
429 			ip = mtod(m = m1, struct ip *);
430 		}
431 #endif /* PFIL_HOOKS */
432 	m_freem(m);
433 	goto done;
434 }
435 
436 /*
437  * Insert IP options into preformed packet.
438  * Adjust IP destination as required for IP source routing,
439  * as indicated by a non-zero in_addr at the start of the options.
440  */
441 static struct mbuf *
442 ip_insertoptions(m, opt, phlen)
443 	register struct mbuf *m;
444 	struct mbuf *opt;
445 	int *phlen;
446 {
447 	register struct ipoption *p = mtod(opt, struct ipoption *);
448 	struct mbuf *n;
449 	register struct ip *ip = mtod(m, struct ip *);
450 	unsigned optlen;
451 
452 	optlen = opt->m_len - sizeof(p->ipopt_dst);
453 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
454 		return (m);		/* XXX should fail */
455 	if (!in_nullhost(p->ipopt_dst))
456 		ip->ip_dst = p->ipopt_dst;
457 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
458 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
459 		if (n == 0)
460 			return (m);
461 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
462 		m->m_len -= sizeof(struct ip);
463 		m->m_data += sizeof(struct ip);
464 		n->m_next = m;
465 		m = n;
466 		m->m_len = optlen + sizeof(struct ip);
467 		m->m_data += max_linkhdr;
468 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
469 	} else {
470 		m->m_data -= optlen;
471 		m->m_len += optlen;
472 		m->m_pkthdr.len += optlen;
473 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
474 	}
475 	ip = mtod(m, struct ip *);
476 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
477 	*phlen = sizeof(struct ip) + optlen;
478 	ip->ip_len += optlen;
479 	return (m);
480 }
481 
482 /*
483  * Copy options from ip to jp,
484  * omitting those not copied during fragmentation.
485  */
486 int
487 ip_optcopy(ip, jp)
488 	struct ip *ip, *jp;
489 {
490 	register u_char *cp, *dp;
491 	int opt, optlen, cnt;
492 
493 	cp = (u_char *)(ip + 1);
494 	dp = (u_char *)(jp + 1);
495 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
496 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
497 		opt = cp[0];
498 		if (opt == IPOPT_EOL)
499 			break;
500 		if (opt == IPOPT_NOP) {
501 			/* Preserve for IP mcast tunnel's LSRR alignment. */
502 			*dp++ = IPOPT_NOP;
503 			optlen = 1;
504 			continue;
505 		} else
506 			optlen = cp[IPOPT_OLEN];
507 		/* bogus lengths should have been caught by ip_dooptions */
508 		if (optlen > cnt)
509 			optlen = cnt;
510 		if (IPOPT_COPIED(opt)) {
511 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
512 			dp += optlen;
513 		}
514 	}
515 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
516 		*dp++ = IPOPT_EOL;
517 	return (optlen);
518 }
519 
520 /*
521  * IP socket option processing.
522  */
523 int
524 ip_ctloutput(op, so, level, optname, mp)
525 	int op;
526 	struct socket *so;
527 	int level, optname;
528 	struct mbuf **mp;
529 {
530 	register struct inpcb *inp = sotoinpcb(so);
531 	register struct mbuf *m = *mp;
532 	register int optval = 0;
533 	int error = 0;
534 
535 	if (level != IPPROTO_IP) {
536 		error = EINVAL;
537 		if (op == PRCO_SETOPT && *mp)
538 			(void) m_free(*mp);
539 	} else switch (op) {
540 
541 	case PRCO_SETOPT:
542 		switch (optname) {
543 		case IP_OPTIONS:
544 #ifdef notyet
545 		case IP_RETOPTS:
546 			return (ip_pcbopts(optname, &inp->inp_options, m));
547 #else
548 			return (ip_pcbopts(&inp->inp_options, m));
549 #endif
550 
551 		case IP_TOS:
552 		case IP_TTL:
553 		case IP_RECVOPTS:
554 		case IP_RECVRETOPTS:
555 		case IP_RECVDSTADDR:
556 			if (m == NULL || m->m_len != sizeof(int))
557 				error = EINVAL;
558 			else {
559 				optval = *mtod(m, int *);
560 				switch (optname) {
561 
562 				case IP_TOS:
563 					inp->inp_ip.ip_tos = optval;
564 					break;
565 
566 				case IP_TTL:
567 					inp->inp_ip.ip_ttl = optval;
568 					break;
569 #define	OPTSET(bit) \
570 	if (optval) \
571 		inp->inp_flags |= bit; \
572 	else \
573 		inp->inp_flags &= ~bit;
574 
575 				case IP_RECVOPTS:
576 					OPTSET(INP_RECVOPTS);
577 					break;
578 
579 				case IP_RECVRETOPTS:
580 					OPTSET(INP_RECVRETOPTS);
581 					break;
582 
583 				case IP_RECVDSTADDR:
584 					OPTSET(INP_RECVDSTADDR);
585 					break;
586 				}
587 			}
588 			break;
589 #undef OPTSET
590 
591 		case IP_MULTICAST_IF:
592 		case IP_MULTICAST_TTL:
593 		case IP_MULTICAST_LOOP:
594 		case IP_ADD_MEMBERSHIP:
595 		case IP_DROP_MEMBERSHIP:
596 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
597 			break;
598 
599 		default:
600 			error = ENOPROTOOPT;
601 			break;
602 		}
603 		if (m)
604 			(void)m_free(m);
605 		break;
606 
607 	case PRCO_GETOPT:
608 		switch (optname) {
609 		case IP_OPTIONS:
610 		case IP_RETOPTS:
611 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
612 			if (inp->inp_options) {
613 				m->m_len = inp->inp_options->m_len;
614 				bcopy(mtod(inp->inp_options, caddr_t),
615 				    mtod(m, caddr_t), (unsigned)m->m_len);
616 			} else
617 				m->m_len = 0;
618 			break;
619 
620 		case IP_TOS:
621 		case IP_TTL:
622 		case IP_RECVOPTS:
623 		case IP_RECVRETOPTS:
624 		case IP_RECVDSTADDR:
625 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
626 			m->m_len = sizeof(int);
627 			switch (optname) {
628 
629 			case IP_TOS:
630 				optval = inp->inp_ip.ip_tos;
631 				break;
632 
633 			case IP_TTL:
634 				optval = inp->inp_ip.ip_ttl;
635 				break;
636 
637 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
638 
639 			case IP_RECVOPTS:
640 				optval = OPTBIT(INP_RECVOPTS);
641 				break;
642 
643 			case IP_RECVRETOPTS:
644 				optval = OPTBIT(INP_RECVRETOPTS);
645 				break;
646 
647 			case IP_RECVDSTADDR:
648 				optval = OPTBIT(INP_RECVDSTADDR);
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