xref: /original-bsd/sys/netinet/ip_output.c (revision 5447a84a)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)ip_output.c	8.3 (Berkeley) 01/21/94
8  */
9 
10 #include <sys/param.h>
11 #include <sys/malloc.h>
12 #include <sys/mbuf.h>
13 #include <sys/errno.h>
14 #include <sys/protosw.h>
15 #include <sys/socket.h>
16 #include <sys/socketvar.h>
17 
18 #include <net/if.h>
19 #include <net/route.h>
20 
21 #include <netinet/in.h>
22 #include <netinet/in_systm.h>
23 #include <netinet/ip.h>
24 #include <netinet/in_pcb.h>
25 #include <netinet/in_var.h>
26 #include <netinet/ip_var.h>
27 
28 #ifdef vax
29 #include <machine/mtpr.h>
30 #endif
31 
32 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
33 static void ip_mloopback
34 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
35 
36 /*
37  * IP output.  The packet in mbuf chain m contains a skeletal IP
38  * header (with len, off, ttl, proto, tos, src, dst).
39  * The mbuf chain containing the packet will be freed.
40  * The mbuf opt, if present, will not be freed.
41  */
42 int
ip_output(m0,opt,ro,flags,imo)43 ip_output(m0, opt, ro, flags, imo)
44 	struct mbuf *m0;
45 	struct mbuf *opt;
46 	struct route *ro;
47 	int flags;
48 	struct ip_moptions *imo;
49 {
50 	register struct ip *ip, *mhip;
51 	register struct ifnet *ifp;
52 	register struct mbuf *m = m0;
53 	register int hlen = sizeof (struct ip);
54 	int len, off, error = 0;
55 	struct route iproute;
56 	struct sockaddr_in *dst;
57 	struct in_ifaddr *ia;
58 
59 #ifdef	DIAGNOSTIC
60 	if ((m->m_flags & M_PKTHDR) == 0)
61 		panic("ip_output no HDR");
62 #endif
63 	if (opt) {
64 		m = ip_insertoptions(m, opt, &len);
65 		hlen = len;
66 	}
67 	ip = mtod(m, struct ip *);
68 	/*
69 	 * Fill in IP header.
70 	 */
71 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
72 		ip->ip_v = IPVERSION;
73 		ip->ip_off &= IP_DF;
74 		ip->ip_id = htons(ip_id++);
75 		ip->ip_hl = hlen >> 2;
76 		ipstat.ips_localout++;
77 	} else {
78 		hlen = ip->ip_hl << 2;
79 	}
80 	/*
81 	 * Route packet.
82 	 */
83 	if (ro == 0) {
84 		ro = &iproute;
85 		bzero((caddr_t)ro, sizeof (*ro));
86 	}
87 	dst = (struct sockaddr_in *)&ro->ro_dst;
88 	/*
89 	 * If there is a cached route,
90 	 * check that it is to the same destination
91 	 * and is still up.  If not, free it and try again.
92 	 */
93 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
94 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
95 		RTFREE(ro->ro_rt);
96 		ro->ro_rt = (struct rtentry *)0;
97 	}
98 	if (ro->ro_rt == 0) {
99 		dst->sin_family = AF_INET;
100 		dst->sin_len = sizeof(*dst);
101 		dst->sin_addr = ip->ip_dst;
102 	}
103 	/*
104 	 * If routing to interface only,
105 	 * short circuit routing lookup.
106 	 */
107 #define ifatoia(ifa)	((struct in_ifaddr *)(ifa))
108 #define sintosa(sin)	((struct sockaddr *)(sin))
109 	if (flags & IP_ROUTETOIF) {
110 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
111 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
112 			ipstat.ips_noroute++;
113 			error = ENETUNREACH;
114 			goto bad;
115 		}
116 		ifp = ia->ia_ifp;
117 		ip->ip_ttl = 1;
118 	} else {
119 		if (ro->ro_rt == 0)
120 			rtalloc(ro);
121 		if (ro->ro_rt == 0) {
122 			ipstat.ips_noroute++;
123 			error = EHOSTUNREACH;
124 			goto bad;
125 		}
126 		ia = ifatoia(ro->ro_rt->rt_ifa);
127 		ifp = ro->ro_rt->rt_ifp;
128 		ro->ro_rt->rt_use++;
129 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
130 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
131 	}
132 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
133 		struct in_multi *inm;
134 		extern struct ifnet loif;
135 
136 		m->m_flags |= M_MCAST;
137 		/*
138 		 * IP destination address is multicast.  Make sure "dst"
139 		 * still points to the address in "ro".  (It may have been
140 		 * changed to point to a gateway address, above.)
141 		 */
142 		dst = (struct sockaddr_in *)&ro->ro_dst;
143 		/*
144 		 * See if the caller provided any multicast options
145 		 */
146 		if (imo != NULL) {
147 			ip->ip_ttl = imo->imo_multicast_ttl;
148 			if (imo->imo_multicast_ifp != NULL)
149 				ifp = imo->imo_multicast_ifp;
150 		} else
151 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
152 		/*
153 		 * Confirm that the outgoing interface supports multicast.
154 		 */
155 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
156 			ipstat.ips_noroute++;
157 			error = ENETUNREACH;
158 			goto bad;
159 		}
160 		/*
161 		 * If source address not specified yet, use address
162 		 * of outgoing interface.
163 		 */
164 		if (ip->ip_src.s_addr == INADDR_ANY) {
165 			register struct in_ifaddr *ia;
166 
167 			for (ia = in_ifaddr; ia; ia = ia->ia_next)
168 				if (ia->ia_ifp == ifp) {
169 					ip->ip_src = IA_SIN(ia)->sin_addr;
170 					break;
171 				}
172 		}
173 
174 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
175 		if (inm != NULL &&
176 		   (imo == NULL || imo->imo_multicast_loop)) {
177 			/*
178 			 * If we belong to the destination multicast group
179 			 * on the outgoing interface, and the caller did not
180 			 * forbid loopback, loop back a copy.
181 			 */
182 			ip_mloopback(ifp, m, dst);
183 		}
184 #ifdef MROUTING
185 		else {
186 			/*
187 			 * If we are acting as a multicast router, perform
188 			 * multicast forwarding as if the packet had just
189 			 * arrived on the interface to which we are about
190 			 * to send.  The multicast forwarding function
191 			 * recursively calls this function, using the
192 			 * IP_FORWARDING flag to prevent infinite recursion.
193 			 *
194 			 * Multicasts that are looped back by ip_mloopback(),
195 			 * above, will be forwarded by the ip_input() routine,
196 			 * if necessary.
197 			 */
198 			extern struct socket *ip_mrouter;
199 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
200 				if (ip_mforward(m, ifp) != 0) {
201 					m_freem(m);
202 					goto done;
203 				}
204 			}
205 		}
206 #endif
207 		/*
208 		 * Multicasts with a time-to-live of zero may be looped-
209 		 * back, above, but must not be transmitted on a network.
210 		 * Also, multicasts addressed to the loopback interface
211 		 * are not sent -- the above call to ip_mloopback() will
212 		 * loop back a copy if this host actually belongs to the
213 		 * destination group on the loopback interface.
214 		 */
215 		if (ip->ip_ttl == 0 || ifp == &loif) {
216 			m_freem(m);
217 			goto done;
218 		}
219 
220 		goto sendit;
221 	}
222 #ifndef notdef
223 	/*
224 	 * If source address not specified yet, use address
225 	 * of outgoing interface.
226 	 */
227 	if (ip->ip_src.s_addr == INADDR_ANY)
228 		ip->ip_src = IA_SIN(ia)->sin_addr;
229 #endif
230 	/*
231 	 * Look for broadcast address and
232 	 * and verify user is allowed to send
233 	 * such a packet.
234 	 */
235 	if (in_broadcast(dst->sin_addr, ifp)) {
236 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
237 			error = EADDRNOTAVAIL;
238 			goto bad;
239 		}
240 		if ((flags & IP_ALLOWBROADCAST) == 0) {
241 			error = EACCES;
242 			goto bad;
243 		}
244 		/* don't allow broadcast messages to be fragmented */
245 		if ((u_short)ip->ip_len > ifp->if_mtu) {
246 			error = EMSGSIZE;
247 			goto bad;
248 		}
249 		m->m_flags |= M_BCAST;
250 	} else
251 		m->m_flags &= ~M_BCAST;
252 
253 sendit:
254 	/*
255 	 * If small enough for interface, can just send directly.
256 	 */
257 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
258 		ip->ip_len = htons((u_short)ip->ip_len);
259 		ip->ip_off = htons((u_short)ip->ip_off);
260 		ip->ip_sum = 0;
261 		ip->ip_sum = in_cksum(m, hlen);
262 		error = (*ifp->if_output)(ifp, m,
263 				(struct sockaddr *)dst, ro->ro_rt);
264 		goto done;
265 	}
266 	/*
267 	 * Too large for interface; fragment if possible.
268 	 * Must be able to put at least 8 bytes per fragment.
269 	 */
270 	if (ip->ip_off & IP_DF) {
271 		error = EMSGSIZE;
272 		ipstat.ips_cantfrag++;
273 		goto bad;
274 	}
275 	len = (ifp->if_mtu - hlen) &~ 7;
276 	if (len < 8) {
277 		error = EMSGSIZE;
278 		goto bad;
279 	}
280 
281     {
282 	int mhlen, firstlen = len;
283 	struct mbuf **mnext = &m->m_nextpkt;
284 
285 	/*
286 	 * Loop through length of segment after first fragment,
287 	 * make new header and copy data of each part and link onto chain.
288 	 */
289 	m0 = m;
290 	mhlen = sizeof (struct ip);
291 	for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
292 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
293 		if (m == 0) {
294 			error = ENOBUFS;
295 			ipstat.ips_odropped++;
296 			goto sendorfree;
297 		}
298 		m->m_data += max_linkhdr;
299 		mhip = mtod(m, struct ip *);
300 		*mhip = *ip;
301 		if (hlen > sizeof (struct ip)) {
302 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
303 			mhip->ip_hl = mhlen >> 2;
304 		}
305 		m->m_len = mhlen;
306 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
307 		if (ip->ip_off & IP_MF)
308 			mhip->ip_off |= IP_MF;
309 		if (off + len >= (u_short)ip->ip_len)
310 			len = (u_short)ip->ip_len - off;
311 		else
312 			mhip->ip_off |= IP_MF;
313 		mhip->ip_len = htons((u_short)(len + mhlen));
314 		m->m_next = m_copy(m0, off, len);
315 		if (m->m_next == 0) {
316 			(void) m_free(m);
317 			error = ENOBUFS;	/* ??? */
318 			ipstat.ips_odropped++;
319 			goto sendorfree;
320 		}
321 		m->m_pkthdr.len = mhlen + len;
322 		m->m_pkthdr.rcvif = (struct ifnet *)0;
323 		mhip->ip_off = htons((u_short)mhip->ip_off);
324 		mhip->ip_sum = 0;
325 		mhip->ip_sum = in_cksum(m, mhlen);
326 		*mnext = m;
327 		mnext = &m->m_nextpkt;
328 		ipstat.ips_ofragments++;
329 	}
330 	/*
331 	 * Update first fragment by trimming what's been copied out
332 	 * and updating header, then send each fragment (in order).
333 	 */
334 	m = m0;
335 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
336 	m->m_pkthdr.len = hlen + firstlen;
337 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
338 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
339 	ip->ip_sum = 0;
340 	ip->ip_sum = in_cksum(m, hlen);
341 sendorfree:
342 	for (m = m0; m; m = m0) {
343 		m0 = m->m_nextpkt;
344 		m->m_nextpkt = 0;
345 		if (error == 0)
346 			error = (*ifp->if_output)(ifp, m,
347 			    (struct sockaddr *)dst, ro->ro_rt);
348 		else
349 			m_freem(m);
350 	}
351 
352 	if (error == 0)
353 		ipstat.ips_fragmented++;
354     }
355 done:
356 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
357 		RTFREE(ro->ro_rt);
358 	return (error);
359 bad:
360 	m_freem(m0);
361 	goto done;
362 }
363 
364 /*
365  * Insert IP options into preformed packet.
366  * Adjust IP destination as required for IP source routing,
367  * as indicated by a non-zero in_addr at the start of the options.
368  */
369 static struct mbuf *
ip_insertoptions(m,opt,phlen)370 ip_insertoptions(m, opt, phlen)
371 	register struct mbuf *m;
372 	struct mbuf *opt;
373 	int *phlen;
374 {
375 	register struct ipoption *p = mtod(opt, struct ipoption *);
376 	struct mbuf *n;
377 	register struct ip *ip = mtod(m, struct ip *);
378 	unsigned optlen;
379 
380 	optlen = opt->m_len - sizeof(p->ipopt_dst);
381 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
382 		return (m);		/* XXX should fail */
383 	if (p->ipopt_dst.s_addr)
384 		ip->ip_dst = p->ipopt_dst;
385 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
386 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
387 		if (n == 0)
388 			return (m);
389 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
390 		m->m_len -= sizeof(struct ip);
391 		m->m_data += sizeof(struct ip);
392 		n->m_next = m;
393 		m = n;
394 		m->m_len = optlen + sizeof(struct ip);
395 		m->m_data += max_linkhdr;
396 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
397 	} else {
398 		m->m_data -= optlen;
399 		m->m_len += optlen;
400 		m->m_pkthdr.len += optlen;
401 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
402 	}
403 	ip = mtod(m, struct ip *);
404 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
405 	*phlen = sizeof(struct ip) + optlen;
406 	ip->ip_len += optlen;
407 	return (m);
408 }
409 
410 /*
411  * Copy options from ip to jp,
412  * omitting those not copied during fragmentation.
413  */
414 int
ip_optcopy(ip,jp)415 ip_optcopy(ip, jp)
416 	struct ip *ip, *jp;
417 {
418 	register u_char *cp, *dp;
419 	int opt, optlen, cnt;
420 
421 	cp = (u_char *)(ip + 1);
422 	dp = (u_char *)(jp + 1);
423 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
424 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
425 		opt = cp[0];
426 		if (opt == IPOPT_EOL)
427 			break;
428 		if (opt == IPOPT_NOP) {
429 			/* Preserve for IP mcast tunnel's LSRR alignment. */
430 			*dp++ = IPOPT_NOP;
431 			optlen = 1;
432 			continue;
433 		} else
434 			optlen = cp[IPOPT_OLEN];
435 		/* bogus lengths should have been caught by ip_dooptions */
436 		if (optlen > cnt)
437 			optlen = cnt;
438 		if (IPOPT_COPIED(opt)) {
439 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
440 			dp += optlen;
441 		}
442 	}
443 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
444 		*dp++ = IPOPT_EOL;
445 	return (optlen);
446 }
447 
448 /*
449  * IP socket option processing.
450  */
451 int
ip_ctloutput(op,so,level,optname,mp)452 ip_ctloutput(op, so, level, optname, mp)
453 	int op;
454 	struct socket *so;
455 	int level, optname;
456 	struct mbuf **mp;
457 {
458 	register struct inpcb *inp = sotoinpcb(so);
459 	register struct mbuf *m = *mp;
460 	register int optval;
461 	int error = 0;
462 
463 	if (level != IPPROTO_IP) {
464 		error = EINVAL;
465 		if (op == PRCO_SETOPT && *mp)
466 			(void) m_free(*mp);
467 	} else switch (op) {
468 
469 	case PRCO_SETOPT:
470 		switch (optname) {
471 		case IP_OPTIONS:
472 #ifdef notyet
473 		case IP_RETOPTS:
474 			return (ip_pcbopts(optname, &inp->inp_options, m));
475 #else
476 			return (ip_pcbopts(&inp->inp_options, m));
477 #endif
478 
479 		case IP_TOS:
480 		case IP_TTL:
481 		case IP_RECVOPTS:
482 		case IP_RECVRETOPTS:
483 		case IP_RECVDSTADDR:
484 			if (m->m_len != sizeof(int))
485 				error = EINVAL;
486 			else {
487 				optval = *mtod(m, int *);
488 				switch (optname) {
489 
490 				case IP_TOS:
491 					inp->inp_ip.ip_tos = optval;
492 					break;
493 
494 				case IP_TTL:
495 					inp->inp_ip.ip_ttl = optval;
496 					break;
497 #define	OPTSET(bit) \
498 	if (optval) \
499 		inp->inp_flags |= bit; \
500 	else \
501 		inp->inp_flags &= ~bit;
502 
503 				case IP_RECVOPTS:
504 					OPTSET(INP_RECVOPTS);
505 					break;
506 
507 				case IP_RECVRETOPTS:
508 					OPTSET(INP_RECVRETOPTS);
509 					break;
510 
511 				case IP_RECVDSTADDR:
512 					OPTSET(INP_RECVDSTADDR);
513 					break;
514 				}
515 			}
516 			break;
517 #undef OPTSET
518 
519 		case IP_MULTICAST_IF:
520 		case IP_MULTICAST_TTL:
521 		case IP_MULTICAST_LOOP:
522 		case IP_ADD_MEMBERSHIP:
523 		case IP_DROP_MEMBERSHIP:
524 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
525 			break;
526 
527 		default:
528 			error = ENOPROTOOPT;
529 			break;
530 		}
531 		if (m)
532 			(void)m_free(m);
533 		break;
534 
535 	case PRCO_GETOPT:
536 		switch (optname) {
537 		case IP_OPTIONS:
538 		case IP_RETOPTS:
539 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
540 			if (inp->inp_options) {
541 				m->m_len = inp->inp_options->m_len;
542 				bcopy(mtod(inp->inp_options, caddr_t),
543 				    mtod(m, caddr_t), (unsigned)m->m_len);
544 			} else
545 				m->m_len = 0;
546 			break;
547 
548 		case IP_TOS:
549 		case IP_TTL:
550 		case IP_RECVOPTS:
551 		case IP_RECVRETOPTS:
552 		case IP_RECVDSTADDR:
553 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
554 			m->m_len = sizeof(int);
555 			switch (optname) {
556 
557 			case IP_TOS:
558 				optval = inp->inp_ip.ip_tos;
559 				break;
560 
561 			case IP_TTL:
562 				optval = inp->inp_ip.ip_ttl;
563 				break;
564 
565 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
566 
567 			case IP_RECVOPTS:
568 				optval = OPTBIT(INP_RECVOPTS);
569 				break;
570 
571 			case IP_RECVRETOPTS:
572 				optval = OPTBIT(INP_RECVRETOPTS);
573 				break;
574 
575 			case IP_RECVDSTADDR:
576 				optval = OPTBIT(INP_RECVDSTADDR);
577 				break;
578 			}
579 			*mtod(m, int *) = optval;
580 			break;
581 
582 		case IP_MULTICAST_IF:
583 		case IP_MULTICAST_TTL:
584 		case IP_MULTICAST_LOOP:
585 		case IP_ADD_MEMBERSHIP:
586 		case IP_DROP_MEMBERSHIP:
587 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
588 			break;
589 
590 		default:
591 			error = ENOPROTOOPT;
592 			break;
593 		}
594 		break;
595 	}
596 	return (error);
597 }
598 
599 /*
600  * Set up IP options in pcb for insertion in output packets.
601  * Store in mbuf with pointer in pcbopt, adding pseudo-option
602  * with destination address if source routed.
603  */
604 int
605 #ifdef notyet
ip_pcbopts(optname,pcbopt,m)606 ip_pcbopts(optname, pcbopt, m)
607 	int optname;
608 #else
609 ip_pcbopts(pcbopt, m)
610 #endif
611 	struct mbuf **pcbopt;
612 	register struct mbuf *m;
613 {
614 	register cnt, optlen;
615 	register u_char *cp;
616 	u_char opt;
617 
618 	/* turn off any old options */
619 	if (*pcbopt)
620 		(void)m_free(*pcbopt);
621 	*pcbopt = 0;
622 	if (m == (struct mbuf *)0 || m->m_len == 0) {
623 		/*
624 		 * Only turning off any previous options.
625 		 */
626 		if (m)
627 			(void)m_free(m);
628 		return (0);
629 	}
630 
631 #ifndef	vax
632 	if (m->m_len % sizeof(long))
633 		goto bad;
634 #endif
635 	/*
636 	 * IP first-hop destination address will be stored before
637 	 * actual options; move other options back
638 	 * and clear it when none present.
639 	 */
640 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
641 		goto bad;
642 	cnt = m->m_len;
643 	m->m_len += sizeof(struct in_addr);
644 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
645 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
646 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
647 
648 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
649 		opt = cp[IPOPT_OPTVAL];
650 		if (opt == IPOPT_EOL)
651 			break;
652 		if (opt == IPOPT_NOP)
653 			optlen = 1;
654 		else {
655 			optlen = cp[IPOPT_OLEN];
656 			if (optlen <= IPOPT_OLEN || optlen > cnt)
657 				goto bad;
658 		}
659 		switch (opt) {
660 
661 		default:
662 			break;
663 
664 		case IPOPT_LSRR:
665 		case IPOPT_SSRR:
666 			/*
667 			 * user process specifies route as:
668 			 *	->A->B->C->D
669 			 * D must be our final destination (but we can't
670 			 * check that since we may not have connected yet).
671 			 * A is first hop destination, which doesn't appear in
672 			 * actual IP option, but is stored before the options.
673 			 */
674 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
675 				goto bad;
676 			m->m_len -= sizeof(struct in_addr);
677 			cnt -= sizeof(struct in_addr);
678 			optlen -= sizeof(struct in_addr);
679 			cp[IPOPT_OLEN] = optlen;
680 			/*
681 			 * Move first hop before start of options.
682 			 */
683 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
684 			    sizeof(struct in_addr));
685 			/*
686 			 * Then copy rest of options back
687 			 * to close up the deleted entry.
688 			 */
689 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
690 			    sizeof(struct in_addr)),
691 			    (caddr_t)&cp[IPOPT_OFFSET+1],
692 			    (unsigned)cnt + sizeof(struct in_addr));
693 			break;
694 		}
695 	}
696 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
697 		goto bad;
698 	*pcbopt = m;
699 	return (0);
700 
701 bad:
702 	(void)m_free(m);
703 	return (EINVAL);
704 }
705 
706 /*
707  * Set the IP multicast options in response to user setsockopt().
708  */
709 int
ip_setmoptions(optname,imop,m)710 ip_setmoptions(optname, imop, m)
711 	int optname;
712 	struct ip_moptions **imop;
713 	struct mbuf *m;
714 {
715 	register int error = 0;
716 	u_char loop;
717 	register int i;
718 	struct in_addr addr;
719 	register struct ip_mreq *mreq;
720 	register struct ifnet *ifp;
721 	register struct ip_moptions *imo = *imop;
722 	struct route ro;
723 	register struct sockaddr_in *dst;
724 
725 	if (imo == NULL) {
726 		/*
727 		 * No multicast option buffer attached to the pcb;
728 		 * allocate one and initialize to default values.
729 		 */
730 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
731 		    M_WAITOK);
732 
733 		if (imo == NULL)
734 			return (ENOBUFS);
735 		*imop = imo;
736 		imo->imo_multicast_ifp = NULL;
737 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
738 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
739 		imo->imo_num_memberships = 0;
740 	}
741 
742 	switch (optname) {
743 
744 	case IP_MULTICAST_IF:
745 		/*
746 		 * Select the interface for outgoing multicast packets.
747 		 */
748 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
749 			error = EINVAL;
750 			break;
751 		}
752 		addr = *(mtod(m, struct in_addr *));
753 		/*
754 		 * INADDR_ANY is used to remove a previous selection.
755 		 * When no interface is selected, a default one is
756 		 * chosen every time a multicast packet is sent.
757 		 */
758 		if (addr.s_addr == INADDR_ANY) {
759 			imo->imo_multicast_ifp = NULL;
760 			break;
761 		}
762 		/*
763 		 * The selected interface is identified by its local
764 		 * IP address.  Find the interface and confirm that
765 		 * it supports multicasting.
766 		 */
767 		INADDR_TO_IFP(addr, ifp);
768 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
769 			error = EADDRNOTAVAIL;
770 			break;
771 		}
772 		imo->imo_multicast_ifp = ifp;
773 		break;
774 
775 	case IP_MULTICAST_TTL:
776 		/*
777 		 * Set the IP time-to-live for outgoing multicast packets.
778 		 */
779 		if (m == NULL || m->m_len != 1) {
780 			error = EINVAL;
781 			break;
782 		}
783 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
784 		break;
785 
786 	case IP_MULTICAST_LOOP:
787 		/*
788 		 * Set the loopback flag for outgoing multicast packets.
789 		 * Must be zero or one.
790 		 */
791 		if (m == NULL || m->m_len != 1 ||
792 		   (loop = *(mtod(m, u_char *))) > 1) {
793 			error = EINVAL;
794 			break;
795 		}
796 		imo->imo_multicast_loop = loop;
797 		break;
798 
799 	case IP_ADD_MEMBERSHIP:
800 		/*
801 		 * Add a multicast group membership.
802 		 * Group must be a valid IP multicast address.
803 		 */
804 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
805 			error = EINVAL;
806 			break;
807 		}
808 		mreq = mtod(m, struct ip_mreq *);
809 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
810 			error = EINVAL;
811 			break;
812 		}
813 		/*
814 		 * If no interface address was provided, use the interface of
815 		 * the route to the given multicast address.
816 		 */
817 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
818 			ro.ro_rt = NULL;
819 			dst = (struct sockaddr_in *)&ro.ro_dst;
820 			dst->sin_len = sizeof(*dst);
821 			dst->sin_family = AF_INET;
822 			dst->sin_addr = mreq->imr_multiaddr;
823 			rtalloc(&ro);
824 			if (ro.ro_rt == NULL) {
825 				error = EADDRNOTAVAIL;
826 				break;
827 			}
828 			ifp = ro.ro_rt->rt_ifp;
829 			rtfree(ro.ro_rt);
830 		}
831 		else {
832 			INADDR_TO_IFP(mreq->imr_interface, ifp);
833 		}
834 		/*
835 		 * See if we found an interface, and confirm that it
836 		 * supports multicast.
837 		 */
838 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
839 			error = EADDRNOTAVAIL;
840 			break;
841 		}
842 		/*
843 		 * See if the membership already exists or if all the
844 		 * membership slots are full.
845 		 */
846 		for (i = 0; i < imo->imo_num_memberships; ++i) {
847 			if (imo->imo_membership[i]->inm_ifp == ifp &&
848 			    imo->imo_membership[i]->inm_addr.s_addr
849 						== mreq->imr_multiaddr.s_addr)
850 				break;
851 		}
852 		if (i < imo->imo_num_memberships) {
853 			error = EADDRINUSE;
854 			break;
855 		}
856 		if (i == IP_MAX_MEMBERSHIPS) {
857 			error = ETOOMANYREFS;
858 			break;
859 		}
860 		/*
861 		 * Everything looks good; add a new record to the multicast
862 		 * address list for the given interface.
863 		 */
864 		if ((imo->imo_membership[i] =
865 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
866 			error = ENOBUFS;
867 			break;
868 		}
869 		++imo->imo_num_memberships;
870 		break;
871 
872 	case IP_DROP_MEMBERSHIP:
873 		/*
874 		 * Drop a multicast group membership.
875 		 * Group must be a valid IP multicast address.
876 		 */
877 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
878 			error = EINVAL;
879 			break;
880 		}
881 		mreq = mtod(m, struct ip_mreq *);
882 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
883 			error = EINVAL;
884 			break;
885 		}
886 		/*
887 		 * If an interface address was specified, get a pointer
888 		 * to its ifnet structure.
889 		 */
890 		if (mreq->imr_interface.s_addr == INADDR_ANY)
891 			ifp = NULL;
892 		else {
893 			INADDR_TO_IFP(mreq->imr_interface, ifp);
894 			if (ifp == NULL) {
895 				error = EADDRNOTAVAIL;
896 				break;
897 			}
898 		}
899 		/*
900 		 * Find the membership in the membership array.
901 		 */
902 		for (i = 0; i < imo->imo_num_memberships; ++i) {
903 			if ((ifp == NULL ||
904 			     imo->imo_membership[i]->inm_ifp == ifp) &&
905 			     imo->imo_membership[i]->inm_addr.s_addr ==
906 			     mreq->imr_multiaddr.s_addr)
907 				break;
908 		}
909 		if (i == imo->imo_num_memberships) {
910 			error = EADDRNOTAVAIL;
911 			break;
912 		}
913 		/*
914 		 * Give up the multicast address record to which the
915 		 * membership points.
916 		 */
917 		in_delmulti(imo->imo_membership[i]);
918 		/*
919 		 * Remove the gap in the membership array.
920 		 */
921 		for (++i; i < imo->imo_num_memberships; ++i)
922 			imo->imo_membership[i-1] = imo->imo_membership[i];
923 		--imo->imo_num_memberships;
924 		break;
925 
926 	default:
927 		error = EOPNOTSUPP;
928 		break;
929 	}
930 
931 	/*
932 	 * If all options have default values, no need to keep the mbuf.
933 	 */
934 	if (imo->imo_multicast_ifp == NULL &&
935 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
936 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
937 	    imo->imo_num_memberships == 0) {
938 		free(*imop, M_IPMOPTS);
939 		*imop = NULL;
940 	}
941 
942 	return (error);
943 }
944 
945 /*
946  * Return the IP multicast options in response to user getsockopt().
947  */
948 int
ip_getmoptions(optname,imo,mp)949 ip_getmoptions(optname, imo, mp)
950 	int optname;
951 	register struct ip_moptions *imo;
952 	register struct mbuf **mp;
953 {
954 	u_char *ttl;
955 	u_char *loop;
956 	struct in_addr *addr;
957 	struct in_ifaddr *ia;
958 
959 	*mp = m_get(M_WAIT, MT_SOOPTS);
960 
961 	switch (optname) {
962 
963 	case IP_MULTICAST_IF:
964 		addr = mtod(*mp, struct in_addr *);
965 		(*mp)->m_len = sizeof(struct in_addr);
966 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
967 			addr->s_addr = INADDR_ANY;
968 		else {
969 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
970 			addr->s_addr = (ia == NULL) ? INADDR_ANY
971 					: IA_SIN(ia)->sin_addr.s_addr;
972 		}
973 		return (0);
974 
975 	case IP_MULTICAST_TTL:
976 		ttl = mtod(*mp, u_char *);
977 		(*mp)->m_len = 1;
978 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
979 				     : imo->imo_multicast_ttl;
980 		return (0);
981 
982 	case IP_MULTICAST_LOOP:
983 		loop = mtod(*mp, u_char *);
984 		(*mp)->m_len = 1;
985 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
986 				      : imo->imo_multicast_loop;
987 		return (0);
988 
989 	default:
990 		return (EOPNOTSUPP);
991 	}
992 }
993 
994 /*
995  * Discard the IP multicast options.
996  */
997 void
ip_freemoptions(imo)998 ip_freemoptions(imo)
999 	register struct ip_moptions *imo;
1000 {
1001 	register int i;
1002 
1003 	if (imo != NULL) {
1004 		for (i = 0; i < imo->imo_num_memberships; ++i)
1005 			in_delmulti(imo->imo_membership[i]);
1006 		free(imo, M_IPMOPTS);
1007 	}
1008 }
1009 
1010 /*
1011  * Routine called from ip_output() to loop back a copy of an IP multicast
1012  * packet to the input queue of a specified interface.  Note that this
1013  * calls the output routine of the loopback "driver", but with an interface
1014  * pointer that might NOT be &loif -- easier than replicating that code here.
1015  */
1016 static void
ip_mloopback(ifp,m,dst)1017 ip_mloopback(ifp, m, dst)
1018 	struct ifnet *ifp;
1019 	register struct mbuf *m;
1020 	register struct sockaddr_in *dst;
1021 {
1022 	register struct ip *ip;
1023 	struct mbuf *copym;
1024 
1025 	copym = m_copy(m, 0, M_COPYALL);
1026 	if (copym != NULL) {
1027 		/*
1028 		 * We don't bother to fragment if the IP length is greater
1029 		 * than the interface's MTU.  Can this possibly matter?
1030 		 */
1031 		ip = mtod(copym, struct ip *);
1032 		ip->ip_len = htons((u_short)ip->ip_len);
1033 		ip->ip_off = htons((u_short)ip->ip_off);
1034 		ip->ip_sum = 0;
1035 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1036 		(void) looutput(ifp, copym, (struct sockaddr *)dst, NULL);
1037 	}
1038 }
1039