xref: /freebsd/sys/netinet/ip_output.c (revision b7a1a0a5ff8895e6a5c9404d02357ac9e2eb8091)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
34  * $Id: ip_output.c,v 1.10 1994/12/12 17:20:54 ugen Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.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 
46 #include <net/if.h>
47 #include <net/route.h>
48 
49 #include <netinet/in.h>
50 #include <netinet/in_systm.h>
51 #include <netinet/ip.h>
52 #include <netinet/in_pcb.h>
53 #include <netinet/in_var.h>
54 #include <netinet/ip_var.h>
55 
56 #ifdef IPFIREWALL
57 #include <netinet/ip_fw.h>
58 #endif
59 #ifdef IPACCT
60 #include <netinet/ip_fw.h>
61 #endif
62 
63 #ifdef vax
64 #include <machine/mtpr.h>
65 #endif
66 
67 u_short ip_id;
68 
69 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
70 static void ip_mloopback
71 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
72 
73 /*
74  * IP output.  The packet in mbuf chain m contains a skeletal IP
75  * header (with len, off, ttl, proto, tos, src, dst).
76  * The mbuf chain containing the packet will be freed.
77  * The mbuf opt, if present, will not be freed.
78  */
79 int
80 ip_output(m0, opt, ro, flags, imo)
81 	struct mbuf *m0;
82 	struct mbuf *opt;
83 	struct route *ro;
84 	int flags;
85 	struct ip_moptions *imo;
86 {
87 	register struct ip *ip, *mhip;
88 	register struct ifnet *ifp;
89 	register struct mbuf *m = m0;
90 	register int hlen = sizeof (struct ip);
91 	int len, off, error = 0;
92 	struct route iproute;
93 	struct sockaddr_in *dst;
94 	struct in_ifaddr *ia;
95 
96 #ifdef	DIAGNOSTIC
97 	if ((m->m_flags & M_PKTHDR) == 0)
98 		panic("ip_output no HDR");
99 #endif
100 	if (opt) {
101 		m = ip_insertoptions(m, opt, &len);
102 		hlen = len;
103 	}
104 	ip = mtod(m, struct ip *);
105 	/*
106 	 * Fill in IP header.
107 	 */
108 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
109 		ip->ip_v = IPVERSION;
110 		ip->ip_off &= IP_DF;
111 		ip->ip_id = htons(ip_id++);
112 		ip->ip_hl = hlen >> 2;
113 		ipstat.ips_localout++;
114 	} else {
115 		hlen = ip->ip_hl << 2;
116 	}
117 	/*
118 	 * Route packet.
119 	 */
120 	if (ro == 0) {
121 		ro = &iproute;
122 		bzero((caddr_t)ro, sizeof (*ro));
123 	}
124 	dst = (struct sockaddr_in *)&ro->ro_dst;
125 	/*
126 	 * If there is a cached route,
127 	 * check that it is to the same destination
128 	 * and is still up.  If not, free it and try again.
129 	 */
130 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
131 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
132 		RTFREE(ro->ro_rt);
133 		ro->ro_rt = (struct rtentry *)0;
134 	}
135 	if (ro->ro_rt == 0) {
136 		dst->sin_family = AF_INET;
137 		dst->sin_len = sizeof(*dst);
138 		dst->sin_addr = ip->ip_dst;
139 	}
140 	/*
141 	 * If routing to interface only,
142 	 * short circuit routing lookup.
143 	 */
144 #define ifatoia(ifa)	((struct in_ifaddr *)(ifa))
145 #define sintosa(sin)	((struct sockaddr *)(sin))
146 	if (flags & IP_ROUTETOIF) {
147 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
148 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
149 			ipstat.ips_noroute++;
150 			error = ENETUNREACH;
151 			goto bad;
152 		}
153 		ifp = ia->ia_ifp;
154 		ip->ip_ttl = 1;
155 	} else {
156 		/*
157 		 * If this is the case, we probably don't want to allocate
158 		 * a protocol-cloned route since we didn't get one from the
159 		 * ULP.  This lets TCP do its thing, while not burdening
160 		 * forwarding or ICMP with the overhead of cloning a route.
161 		 * Of course, we still want to do any cloning requested by
162 		 * the link layer, as this is probably required in all cases
163 		 * for correct operation (as it is for ARP).
164 		 */
165 		if (ro->ro_rt == 0)
166 			rtalloc_ign(ro, RTF_PRCLONING);
167 		if (ro->ro_rt == 0) {
168 			ipstat.ips_noroute++;
169 			error = EHOSTUNREACH;
170 			goto bad;
171 		}
172 		ia = ifatoia(ro->ro_rt->rt_ifa);
173 		ifp = ro->ro_rt->rt_ifp;
174 		ro->ro_rt->rt_use++;
175 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
176 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
177 	}
178 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
179 		struct in_multi *inm;
180 		extern struct ifnet loif;
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 = (struct sockaddr_in *)&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 (ip->ip_src.s_addr == INADDR_ANY) {
211 			register struct in_ifaddr *ia;
212 
213 			for (ia = in_ifaddr; ia; ia = ia->ia_next)
214 				if (ia->ia_ifp == ifp) {
215 					ip->ip_src = IA_SIN(ia)->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 		else {
231 			/*
232 			 * If we are acting as a multicast router, perform
233 			 * multicast forwarding as if the packet had just
234 			 * arrived on the interface to which we are about
235 			 * to send.  The multicast forwarding function
236 			 * recursively calls this function, using the
237 			 * IP_FORWARDING flag to prevent infinite recursion.
238 			 *
239 			 * Multicasts that are looped back by ip_mloopback(),
240 			 * above, will be forwarded by the ip_input() routine,
241 			 * if necessary.
242 			 */
243 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
244 				/*
245 				 * Check if rsvp daemon is running. If not, don't
246 				 * set ip_moptions. This ensures that the packet
247 				 * is multicast and not just sent down one link
248 				 * as prescribed by rsvpd.
249 				 */
250 				if (ip_rsvpd == NULL)
251 				  imo = NULL;
252 				if (ip_mforward(ip, ifp, m, imo) != 0) {
253 					m_freem(m);
254 					goto done;
255 				}
256 			}
257 		}
258 
259 		/*
260 		 * Multicasts with a time-to-live of zero may be looped-
261 		 * back, above, but must not be transmitted on a network.
262 		 * Also, multicasts addressed to the loopback interface
263 		 * are not sent -- the above call to ip_mloopback() will
264 		 * loop back a copy if this host actually belongs to the
265 		 * destination group on the loopback interface.
266 		 */
267 		if (ip->ip_ttl == 0 || ifp == &loif) {
268 			m_freem(m);
269 			goto done;
270 		}
271 
272 		goto sendit;
273 	}
274 #ifndef notdef
275 	/*
276 	 * If source address not specified yet, use address
277 	 * of outgoing interface.
278 	 */
279 	if (ip->ip_src.s_addr == INADDR_ANY)
280 		ip->ip_src = IA_SIN(ia)->sin_addr;
281 #endif
282 	/*
283 	 * Verify that we have any chance at all of being able to queue
284 	 *      the packet or packet fragments
285 	 */
286 	if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
287 		ifp->if_snd.ifq_maxlen) {
288 			error = ENOBUFS;
289 			goto bad;
290 	}
291 
292 	/*
293 	 * Look for broadcast address and
294 	 * and verify user is allowed to send
295 	 * such a packet.
296 	 */
297 	if (in_broadcast(dst->sin_addr, ifp)) {
298 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
299 			error = EADDRNOTAVAIL;
300 			goto bad;
301 		}
302 		if ((flags & IP_ALLOWBROADCAST) == 0) {
303 			error = EACCES;
304 			goto bad;
305 		}
306 		/* don't allow broadcast messages to be fragmented */
307 		if ((u_short)ip->ip_len > ifp->if_mtu) {
308 			error = EMSGSIZE;
309 			goto bad;
310 		}
311 		m->m_flags |= M_BCAST;
312 	} else
313 		m->m_flags &= ~M_BCAST;
314 
315 sendit:
316 	/*
317 	 * If small enough for interface, can just send directly.
318 	 */
319 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
320 		ip->ip_len = htons((u_short)ip->ip_len);
321 		ip->ip_off = htons((u_short)ip->ip_off);
322 		ip->ip_sum = 0;
323 		ip->ip_sum = in_cksum(m, hlen);
324 		error = (*ifp->if_output)(ifp, m,
325 				(struct sockaddr *)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_short)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 		m->m_data += max_linkhdr;
361 		mhip = mtod(m, struct ip *);
362 		*mhip = *ip;
363 		if (hlen > sizeof (struct ip)) {
364 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
365 			mhip->ip_hl = mhlen >> 2;
366 		}
367 		m->m_len = mhlen;
368 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
369 		if (ip->ip_off & IP_MF)
370 			mhip->ip_off |= IP_MF;
371 		if (off + len >= (u_short)ip->ip_len)
372 			len = (u_short)ip->ip_len - off;
373 		else
374 			mhip->ip_off |= IP_MF;
375 		mhip->ip_len = htons((u_short)(len + mhlen));
376 		m->m_next = m_copy(m0, off, len);
377 		if (m->m_next == 0) {
378 			(void) m_free(m);
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_short)mhip->ip_off);
386 		mhip->ip_sum = 0;
387 		mhip->ip_sum = in_cksum(m, mhlen);
388 		*mnext = m;
389 		mnext = &m->m_nextpkt;
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_short)ip->ip_len);
398 	m->m_pkthdr.len = hlen + firstlen;
399 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
400 	ip->ip_off = htons((u_short)(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,
409 			    (struct sockaddr *)dst, 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 #ifdef IPACCT
421 	/*
422 	 * Count outgoing packet,here we count both our packets and
423 	 * those we forward.
424 	 * Here we want to convert ip_len to host byte order when counting
425 	 * so we set 3rd arg to 1.
426 	 * This is locally generated packet so it has not
427 	 * incoming interface.
428 	 */
429 	ip_acct_cnt(ip,NULL,ip_acct_chain,1);
430 #endif
431 	return (error);
432 bad:
433 	m_freem(m0);
434 	goto done;
435 }
436 
437 /*
438  * Insert IP options into preformed packet.
439  * Adjust IP destination as required for IP source routing,
440  * as indicated by a non-zero in_addr at the start of the options.
441  */
442 static struct mbuf *
443 ip_insertoptions(m, opt, phlen)
444 	register struct mbuf *m;
445 	struct mbuf *opt;
446 	int *phlen;
447 {
448 	register struct ipoption *p = mtod(opt, struct ipoption *);
449 	struct mbuf *n;
450 	register struct ip *ip = mtod(m, struct ip *);
451 	unsigned optlen;
452 
453 	optlen = opt->m_len - sizeof(p->ipopt_dst);
454 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
455 		return (m);		/* XXX should fail */
456 	if (p->ipopt_dst.s_addr)
457 		ip->ip_dst = p->ipopt_dst;
458 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
459 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
460 		if (n == 0)
461 			return (m);
462 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
463 		m->m_len -= sizeof(struct ip);
464 		m->m_data += sizeof(struct ip);
465 		n->m_next = m;
466 		m = n;
467 		m->m_len = optlen + sizeof(struct ip);
468 		m->m_data += max_linkhdr;
469 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
470 	} else {
471 		m->m_data -= optlen;
472 		m->m_len += optlen;
473 		m->m_pkthdr.len += optlen;
474 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
475 	}
476 	ip = mtod(m, struct ip *);
477 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
478 	*phlen = sizeof(struct ip) + optlen;
479 	ip->ip_len += optlen;
480 	return (m);
481 }
482 
483 /*
484  * Copy options from ip to jp,
485  * omitting those not copied during fragmentation.
486  */
487 int
488 ip_optcopy(ip, jp)
489 	struct ip *ip, *jp;
490 {
491 	register u_char *cp, *dp;
492 	int opt, optlen, cnt;
493 
494 	cp = (u_char *)(ip + 1);
495 	dp = (u_char *)(jp + 1);
496 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
497 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
498 		opt = cp[0];
499 		if (opt == IPOPT_EOL)
500 			break;
501 		if (opt == IPOPT_NOP) {
502 			/* Preserve for IP mcast tunnel's LSRR alignment. */
503 			*dp++ = IPOPT_NOP;
504 			optlen = 1;
505 			continue;
506 		} else
507 			optlen = cp[IPOPT_OLEN];
508 		/* bogus lengths should have been caught by ip_dooptions */
509 		if (optlen > cnt)
510 			optlen = cnt;
511 		if (IPOPT_COPIED(opt)) {
512 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
513 			dp += optlen;
514 		}
515 	}
516 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
517 		*dp++ = IPOPT_EOL;
518 	return (optlen);
519 }
520 
521 /*
522  * IP socket option processing.
523  */
524 int
525 ip_ctloutput(op, so, level, optname, mp)
526 	int op;
527 	struct socket *so;
528 	int level, optname;
529 	struct mbuf **mp;
530 {
531 	register struct inpcb *inp = sotoinpcb(so);
532 	register struct mbuf *m = *mp;
533 	register int optval = 0;
534 	int error = 0;
535 
536 	if (level != IPPROTO_IP) {
537 		error = EINVAL;
538 		if (op == PRCO_SETOPT && *mp)
539 			(void) m_free(*mp);
540 	} else switch (op) {
541 
542 	case PRCO_SETOPT:
543 		switch (optname) {
544 		case IP_OPTIONS:
545 #ifdef notyet
546 		case IP_RETOPTS:
547 			return (ip_pcbopts(optname, &inp->inp_options, m));
548 #else
549 			return (ip_pcbopts(&inp->inp_options, m));
550 #endif
551 
552 		case IP_TOS:
553 		case IP_TTL:
554 		case IP_RECVOPTS:
555 		case IP_RECVRETOPTS:
556 		case IP_RECVDSTADDR:
557 			if (m->m_len != sizeof(int))
558 				error = EINVAL;
559 			else {
560 				optval = *mtod(m, int *);
561 				switch (optname) {
562 
563 				case IP_TOS:
564 					inp->inp_ip.ip_tos = optval;
565 					break;
566 
567 				case IP_TTL:
568 					inp->inp_ip.ip_ttl = optval;
569 					break;
570 #define	OPTSET(bit) \
571 	if (optval) \
572 		inp->inp_flags |= bit; \
573 	else \
574 		inp->inp_flags &= ~bit;
575 
576 				case IP_RECVOPTS:
577 					OPTSET(INP_RECVOPTS);
578 					break;
579 
580 				case IP_RECVRETOPTS:
581 					OPTSET(INP_RECVRETOPTS);
582 					break;
583 
584 				case IP_RECVDSTADDR:
585 					OPTSET(INP_RECVDSTADDR);
586 					break;
587 				}
588 			}
589 			break;
590 #undef OPTSET
591 
592 		case IP_MULTICAST_IF:
593 		case IP_MULTICAST_VIF:
594 		case IP_MULTICAST_TTL:
595 		case IP_MULTICAST_LOOP:
596 		case IP_ADD_MEMBERSHIP:
597 		case IP_DROP_MEMBERSHIP:
598 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
599 			break;
600 
601 		default:
602 			error = ENOPROTOOPT;
603 			break;
604 		}
605 		if (m)
606 			(void)m_free(m);
607 		break;
608 
609 	case PRCO_GETOPT:
610 		switch (optname) {
611 		case IP_OPTIONS:
612 		case IP_RETOPTS:
613 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
614 			if (inp->inp_options) {
615 				m->m_len = inp->inp_options->m_len;
616 				bcopy(mtod(inp->inp_options, caddr_t),
617 				    mtod(m, caddr_t), (unsigned)m->m_len);
618 			} else
619 				m->m_len = 0;
620 			break;
621 
622 		case IP_TOS:
623 		case IP_TTL:
624 		case IP_RECVOPTS:
625 		case IP_RECVRETOPTS:
626 		case IP_RECVDSTADDR:
627 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
628 			m->m_len = sizeof(int);
629 			switch (optname) {
630 
631 			case IP_TOS:
632 				optval = inp->inp_ip.ip_tos;
633 				break;
634 
635 			case IP_TTL:
636 				optval = inp->inp_ip.ip_ttl;
637 				break;
638 
639 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
640 
641 			case IP_RECVOPTS:
642 				optval = OPTBIT(INP_RECVOPTS);
643 				break;
644 
645 			case IP_RECVRETOPTS:
646 				optval = OPTBIT(INP_RECVRETOPTS);
647 				break;
648 
649 			case IP_RECVDSTADDR:
650 				optval = OPTBIT(INP_RECVDSTADDR);
651 				break;
652 			}
653 			*mtod(m, int *) = optval;
654 			break;
655 
656 		case IP_MULTICAST_IF:
657 		case IP_MULTICAST_VIF:
658 		case IP_MULTICAST_TTL:
659 		case IP_MULTICAST_LOOP:
660 		case IP_ADD_MEMBERSHIP:
661 		case IP_DROP_MEMBERSHIP:
662 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
663 			break;
664 
665 		default:
666 			error = ENOPROTOOPT;
667 			break;
668 		}
669 		break;
670 	}
671 	return (error);
672 }
673 
674 /*
675  * Set up IP options in pcb for insertion in output packets.
676  * Store in mbuf with pointer in pcbopt, adding pseudo-option
677  * with destination address if source routed.
678  */
679 int
680 #ifdef notyet
681 ip_pcbopts(optname, pcbopt, m)
682 	int optname;
683 #else
684 ip_pcbopts(pcbopt, m)
685 #endif
686 	struct mbuf **pcbopt;
687 	register struct mbuf *m;
688 {
689 	register cnt, optlen;
690 	register u_char *cp;
691 	u_char opt;
692 
693 	/* turn off any old options */
694 	if (*pcbopt)
695 		(void)m_free(*pcbopt);
696 	*pcbopt = 0;
697 	if (m == (struct mbuf *)0 || m->m_len == 0) {
698 		/*
699 		 * Only turning off any previous options.
700 		 */
701 		if (m)
702 			(void)m_free(m);
703 		return (0);
704 	}
705 
706 #ifndef	vax
707 	if (m->m_len % sizeof(long))
708 		goto bad;
709 #endif
710 	/*
711 	 * IP first-hop destination address will be stored before
712 	 * actual options; move other options back
713 	 * and clear it when none present.
714 	 */
715 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
716 		goto bad;
717 	cnt = m->m_len;
718 	m->m_len += sizeof(struct in_addr);
719 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
720 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
721 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
722 
723 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
724 		opt = cp[IPOPT_OPTVAL];
725 		if (opt == IPOPT_EOL)
726 			break;
727 		if (opt == IPOPT_NOP)
728 			optlen = 1;
729 		else {
730 			optlen = cp[IPOPT_OLEN];
731 			if (optlen <= IPOPT_OLEN || optlen > cnt)
732 				goto bad;
733 		}
734 		switch (opt) {
735 
736 		default:
737 			break;
738 
739 		case IPOPT_LSRR:
740 		case IPOPT_SSRR:
741 			/*
742 			 * user process specifies route as:
743 			 *	->A->B->C->D
744 			 * D must be our final destination (but we can't
745 			 * check that since we may not have connected yet).
746 			 * A is first hop destination, which doesn't appear in
747 			 * actual IP option, but is stored before the options.
748 			 */
749 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
750 				goto bad;
751 			m->m_len -= sizeof(struct in_addr);
752 			cnt -= sizeof(struct in_addr);
753 			optlen -= sizeof(struct in_addr);
754 			cp[IPOPT_OLEN] = optlen;
755 			/*
756 			 * Move first hop before start of options.
757 			 */
758 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
759 			    sizeof(struct in_addr));
760 			/*
761 			 * Then copy rest of options back
762 			 * to close up the deleted entry.
763 			 */
764 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
765 			    sizeof(struct in_addr)),
766 			    (caddr_t)&cp[IPOPT_OFFSET+1],
767 			    (unsigned)cnt + sizeof(struct in_addr));
768 			break;
769 		}
770 	}
771 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
772 		goto bad;
773 	*pcbopt = m;
774 	return (0);
775 
776 bad:
777 	(void)m_free(m);
778 	return (EINVAL);
779 }
780 
781 /*
782  * Set the IP multicast options in response to user setsockopt().
783  */
784 int
785 ip_setmoptions(optname, imop, m)
786 	int optname;
787 	struct ip_moptions **imop;
788 	struct mbuf *m;
789 {
790 	register int error = 0;
791 	u_char loop;
792 	register int i;
793 	struct in_addr addr;
794 	register struct ip_mreq *mreq;
795 	register struct ifnet *ifp;
796 	register struct ip_moptions *imo = *imop;
797 	struct route ro;
798 	register struct sockaddr_in *dst;
799 
800 	if (imo == NULL) {
801 		/*
802 		 * No multicast option buffer attached to the pcb;
803 		 * allocate one and initialize to default values.
804 		 */
805 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
806 		    M_WAITOK);
807 
808 		if (imo == NULL)
809 			return (ENOBUFS);
810 		*imop = imo;
811 		imo->imo_multicast_ifp = NULL;
812 		imo->imo_multicast_vif = 0;
813 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
814 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
815 		imo->imo_num_memberships = 0;
816 	}
817 
818 	switch (optname) {
819 		extern int (*legal_vif_num)(int);
820 	/* store an index number for the vif you wanna use in the send */
821 	case IP_MULTICAST_VIF:
822 		if (!legal_vif_num) {
823 			error = EOPNOTSUPP;
824 			break;
825 		}
826 		if (m == NULL || m->m_len != sizeof(int)) {
827 			error = EINVAL;
828 			break;
829 		}
830 		i = *(mtod(m, int *));
831 		if (!legal_vif_num(i)) {
832 			error = EINVAL;
833 			break;
834 		}
835 		imo->imo_multicast_vif = i;
836 		break;
837 
838 	case IP_MULTICAST_IF:
839 		/*
840 		 * Select the interface for outgoing multicast packets.
841 		 */
842 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
843 			error = EINVAL;
844 			break;
845 		}
846 		addr = *(mtod(m, struct in_addr *));
847 		/*
848 		 * INADDR_ANY is used to remove a previous selection.
849 		 * When no interface is selected, a default one is
850 		 * chosen every time a multicast packet is sent.
851 		 */
852 		if (addr.s_addr == INADDR_ANY) {
853 			imo->imo_multicast_ifp = NULL;
854 			break;
855 		}
856 		/*
857 		 * The selected interface is identified by its local
858 		 * IP address.  Find the interface and confirm that
859 		 * it supports multicasting.
860 		 */
861 		INADDR_TO_IFP(addr, ifp);
862 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
863 			error = EADDRNOTAVAIL;
864 			break;
865 		}
866 		imo->imo_multicast_ifp = ifp;
867 		break;
868 
869 	case IP_MULTICAST_TTL:
870 		/*
871 		 * Set the IP time-to-live for outgoing multicast packets.
872 		 */
873 		if (m == NULL || m->m_len != 1) {
874 			error = EINVAL;
875 			break;
876 		}
877 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
878 		break;
879 
880 	case IP_MULTICAST_LOOP:
881 		/*
882 		 * Set the loopback flag for outgoing multicast packets.
883 		 * Must be zero or one.
884 		 */
885 		if (m == NULL || m->m_len != 1 ||
886 		   (loop = *(mtod(m, u_char *))) > 1) {
887 			error = EINVAL;
888 			break;
889 		}
890 		imo->imo_multicast_loop = loop;
891 		break;
892 
893 	case IP_ADD_MEMBERSHIP:
894 		/*
895 		 * Add a multicast group membership.
896 		 * Group must be a valid IP multicast address.
897 		 */
898 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
899 			error = EINVAL;
900 			break;
901 		}
902 		mreq = mtod(m, struct ip_mreq *);
903 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
904 			error = EINVAL;
905 			break;
906 		}
907 		/*
908 		 * If no interface address was provided, use the interface of
909 		 * the route to the given multicast address.
910 		 */
911 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
912 			ro.ro_rt = NULL;
913 			dst = (struct sockaddr_in *)&ro.ro_dst;
914 			dst->sin_len = sizeof(*dst);
915 			dst->sin_family = AF_INET;
916 			dst->sin_addr = mreq->imr_multiaddr;
917 			rtalloc(&ro);
918 			if (ro.ro_rt == NULL) {
919 				error = EADDRNOTAVAIL;
920 				break;
921 			}
922 			ifp = ro.ro_rt->rt_ifp;
923 			rtfree(ro.ro_rt);
924 		}
925 		else {
926 			INADDR_TO_IFP(mreq->imr_interface, ifp);
927 		}
928 		/*
929 		 * See if we found an interface, and confirm that it
930 		 * supports multicast.
931 		 */
932 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
933 			error = EADDRNOTAVAIL;
934 			break;
935 		}
936 		/*
937 		 * See if the membership already exists or if all the
938 		 * membership slots are full.
939 		 */
940 		for (i = 0; i < imo->imo_num_memberships; ++i) {
941 			if (imo->imo_membership[i]->inm_ifp == ifp &&
942 			    imo->imo_membership[i]->inm_addr.s_addr
943 						== mreq->imr_multiaddr.s_addr)
944 				break;
945 		}
946 		if (i < imo->imo_num_memberships) {
947 			error = EADDRINUSE;
948 			break;
949 		}
950 		if (i == IP_MAX_MEMBERSHIPS) {
951 			error = ETOOMANYREFS;
952 			break;
953 		}
954 		/*
955 		 * Everything looks good; add a new record to the multicast
956 		 * address list for the given interface.
957 		 */
958 		if ((imo->imo_membership[i] =
959 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
960 			error = ENOBUFS;
961 			break;
962 		}
963 		++imo->imo_num_memberships;
964 		break;
965 
966 	case IP_DROP_MEMBERSHIP:
967 		/*
968 		 * Drop a multicast group membership.
969 		 * Group must be a valid IP multicast address.
970 		 */
971 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
972 			error = EINVAL;
973 			break;
974 		}
975 		mreq = mtod(m, struct ip_mreq *);
976 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
977 			error = EINVAL;
978 			break;
979 		}
980 		/*
981 		 * If an interface address was specified, get a pointer
982 		 * to its ifnet structure.
983 		 */
984 		if (mreq->imr_interface.s_addr == INADDR_ANY)
985 			ifp = NULL;
986 		else {
987 			INADDR_TO_IFP(mreq->imr_interface, ifp);
988 			if (ifp == NULL) {
989 				error = EADDRNOTAVAIL;
990 				break;
991 			}
992 		}
993 		/*
994 		 * Find the membership in the membership array.
995 		 */
996 		for (i = 0; i < imo->imo_num_memberships; ++i) {
997 			if ((ifp == NULL ||
998 			     imo->imo_membership[i]->inm_ifp == ifp) &&
999 			     imo->imo_membership[i]->inm_addr.s_addr ==
1000 			     mreq->imr_multiaddr.s_addr)
1001 				break;
1002 		}
1003 		if (i == imo->imo_num_memberships) {
1004 			error = EADDRNOTAVAIL;
1005 			break;
1006 		}
1007 		/*
1008 		 * Give up the multicast address record to which the
1009 		 * membership points.
1010 		 */
1011 		in_delmulti(imo->imo_membership[i]);
1012 		/*
1013 		 * Remove the gap in the membership array.
1014 		 */
1015 		for (++i; i < imo->imo_num_memberships; ++i)
1016 			imo->imo_membership[i-1] = imo->imo_membership[i];
1017 		--imo->imo_num_memberships;
1018 		break;
1019 
1020 	default:
1021 		error = EOPNOTSUPP;
1022 		break;
1023 	}
1024 
1025 	/*
1026 	 * If all options have default values, no need to keep the mbuf.
1027 	 */
1028 	if (imo->imo_multicast_ifp == NULL &&
1029 	    imo->imo_multicast_vif == 0 &&
1030 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1031 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1032 	    imo->imo_num_memberships == 0) {
1033 		free(*imop, M_IPMOPTS);
1034 		*imop = NULL;
1035 	}
1036 
1037 	return (error);
1038 }
1039 
1040 /*
1041  * Return the IP multicast options in response to user getsockopt().
1042  */
1043 int
1044 ip_getmoptions(optname, imo, mp)
1045 	int optname;
1046 	register struct ip_moptions *imo;
1047 	register struct mbuf **mp;
1048 {
1049 	u_char *ttl;
1050 	u_char *loop;
1051 	struct in_addr *addr;
1052 	struct in_ifaddr *ia;
1053 
1054 	*mp = m_get(M_WAIT, MT_SOOPTS);
1055 
1056 	switch (optname) {
1057 
1058 	case IP_MULTICAST_VIF:
1059 		if (imo != NULL)
1060 			*(mtod(*mp, int *)) = imo->imo_multicast_vif;
1061 		else
1062 			*(mtod(*mp, int *)) = 7890;
1063 		(*mp)->m_len = sizeof(int);
1064 		return(0);
1065 
1066 	case IP_MULTICAST_IF:
1067 		addr = mtod(*mp, struct in_addr *);
1068 		(*mp)->m_len = sizeof(struct in_addr);
1069 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1070 			addr->s_addr = INADDR_ANY;
1071 		else {
1072 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1073 			addr->s_addr = (ia == NULL) ? INADDR_ANY
1074 					: IA_SIN(ia)->sin_addr.s_addr;
1075 		}
1076 		return (0);
1077 
1078 	case IP_MULTICAST_TTL:
1079 		ttl = mtod(*mp, u_char *);
1080 		(*mp)->m_len = 1;
1081 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1082 				     : imo->imo_multicast_ttl;
1083 		return (0);
1084 
1085 	case IP_MULTICAST_LOOP:
1086 		loop = mtod(*mp, u_char *);
1087 		(*mp)->m_len = 1;
1088 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
1089 				      : imo->imo_multicast_loop;
1090 		return (0);
1091 
1092 	default:
1093 		return (EOPNOTSUPP);
1094 	}
1095 }
1096 
1097 /*
1098  * Discard the IP multicast options.
1099  */
1100 void
1101 ip_freemoptions(imo)
1102 	register struct ip_moptions *imo;
1103 {
1104 	register int i;
1105 
1106 	if (imo != NULL) {
1107 		for (i = 0; i < imo->imo_num_memberships; ++i)
1108 			in_delmulti(imo->imo_membership[i]);
1109 		free(imo, M_IPMOPTS);
1110 	}
1111 }
1112 
1113 /*
1114  * Routine called from ip_output() to loop back a copy of an IP multicast
1115  * packet to the input queue of a specified interface.  Note that this
1116  * calls the output routine of the loopback "driver", but with an interface
1117  * pointer that might NOT be &loif -- easier than replicating that code here.
1118  */
1119 static void
1120 ip_mloopback(ifp, m, dst)
1121 	struct ifnet *ifp;
1122 	register struct mbuf *m;
1123 	register struct sockaddr_in *dst;
1124 {
1125 	register struct ip *ip;
1126 	struct mbuf *copym;
1127 
1128 	copym = m_copy(m, 0, M_COPYALL);
1129 	if (copym != NULL) {
1130 		/*
1131 		 * We don't bother to fragment if the IP length is greater
1132 		 * than the interface's MTU.  Can this possibly matter?
1133 		 */
1134 		ip = mtod(copym, struct ip *);
1135 		ip->ip_len = htons((u_short)ip->ip_len);
1136 		ip->ip_off = htons((u_short)ip->ip_off);
1137 		ip->ip_sum = 0;
1138 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1139 		(void) looutput(ifp, copym, (struct sockaddr *)dst, NULL);
1140 	}
1141 }
1142