xref: /csrg-svn/sys/netinet/ip_output.c (revision 57433)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California.
3  * All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)ip_output.c	7.27 (Berkeley) 01/08/93
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 struct	mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
33 static	void ip_mloopback __P((struct ifnet *, struct mbuf *,
34 	    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
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) == 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 	if (flags & IP_ROUTETOIF) {
108 
109 		ia = (struct in_ifaddr *)ifa_ifwithdstaddr((struct sockaddr *)dst);
110 		if (ia == 0)
111 			ia = in_iaonnetof(in_netof(ip->ip_dst));
112 		if (ia == 0) {
113 			ipstat.ips_noroute++;
114 			error = ENETUNREACH;
115 			goto bad;
116 		}
117 		ifp = ia->ia_ifp;
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 = (struct in_ifaddr *)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 		extern struct socket *ip_mrouter;
136 
137 		m->m_flags |= M_MCAST;
138 		/*
139 		 * IP destination address is multicast.  Make sure "dst"
140 		 * still points to the address in "ro".  (It may have been
141 		 * changed to point to a gateway address, above.)
142 		 */
143 		dst = (struct sockaddr_in *)&ro->ro_dst;
144 		/*
145 		 * See if the caller provided any multicast options
146 		 */
147 		if (imo != NULL) {
148 			ip->ip_ttl = imo->imo_multicast_ttl;
149 			if (imo->imo_multicast_ifp != NULL)
150 				ifp = imo->imo_multicast_ifp;
151 		} else
152 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
153 		/*
154 		 * Confirm that the outgoing interface supports multicast.
155 		 */
156 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
157 			ipstat.ips_noroute++;
158 			error = ENETUNREACH;
159 			goto bad;
160 		}
161 		/*
162 		 * If source address not specified yet, use address
163 		 * of outgoing interface.
164 		 */
165 		if (ip->ip_src.s_addr == INADDR_ANY) {
166 			register struct in_ifaddr *ia;
167 
168 			for (ia = in_ifaddr; ia; ia = ia->ia_next)
169 				if (ia->ia_ifp == ifp) {
170 					ip->ip_src = IA_SIN(ia)->sin_addr;
171 					break;
172 				}
173 		}
174 
175 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
176 		if (inm != NULL &&
177 		   (imo == NULL || imo->imo_multicast_loop)) {
178 			/*
179 			 * If we belong to the destination multicast group
180 			 * on the outgoing interface, and the caller did not
181 			 * forbid loopback, loop back a copy.
182 			 */
183 			ip_mloopback(ifp, m, dst);
184 		}
185 #ifdef MROUTING
186 		else if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
187 			/*
188 			 * If we are acting as a multicast router, perform
189 			 * multicast forwarding as if the packet had just
190 			 * arrived on the interface to which we are about
191 			 * to send.  The multicast forwarding function
192 			 * recursively calls this function, using the
193 			 * IP_FORWARDING flag to prevent infinite recursion.
194 			 *
195 			 * Multicasts that are looped back by ip_mloopback(),
196 			 * above, will be forwarded by the ip_input() routine,
197 			 * if necessary.
198 			 */
199 			if (ip_mforward(m, ifp) != 0) {
200 				m_freem(m);
201 				goto done;
202 			}
203 		}
204 #endif
205 		/*
206 		 * Multicasts with a time-to-live of zero may be looped-
207 		 * back, above, but must not be transmitted on a network.
208 		 * Also, multicasts addressed to the loopback interface
209 		 * are not sent -- the above call to ip_mloopback() will
210 		 * loop back a copy if this host actually belongs to the
211 		 * destination group on the loopback interface.
212 		 */
213 		if (ip->ip_ttl == 0 || ifp == &loif) {
214 			m_freem(m);
215 			goto done;
216 		}
217 
218 		goto sendit;
219 	}
220 #ifndef notdef
221 	/*
222 	 * If source address not specified yet, use address
223 	 * of outgoing interface.
224 	 */
225 	if (ip->ip_src.s_addr == INADDR_ANY)
226 		ip->ip_src = IA_SIN(ia)->sin_addr;
227 #endif
228 	/*
229 	 * Look for broadcast address and
230 	 * and verify user is allowed to send
231 	 * such a packet.
232 	 */
233 	if (in_broadcast(dst->sin_addr)) {
234 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
235 			error = EADDRNOTAVAIL;
236 			goto bad;
237 		}
238 		if ((flags & IP_ALLOWBROADCAST) == 0) {
239 			error = EACCES;
240 			goto bad;
241 		}
242 		/* don't allow broadcast messages to be fragmented */
243 		if ((u_short)ip->ip_len > ifp->if_mtu) {
244 			error = EMSGSIZE;
245 			goto bad;
246 		}
247 		m->m_flags |= M_BCAST;
248 	} else
249 		m->m_flags &= ~M_BCAST;
250 
251 sendit:
252 	/*
253 	 * If small enough for interface, can just send directly.
254 	 */
255 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
256 		ip->ip_len = htons((u_short)ip->ip_len);
257 		ip->ip_off = htons((u_short)ip->ip_off);
258 		ip->ip_sum = 0;
259 		ip->ip_sum = in_cksum(m, hlen);
260 		error = (*ifp->if_output)(ifp, m,
261 				(struct sockaddr *)dst, ro->ro_rt);
262 		goto done;
263 	}
264 	/*
265 	 * Too large for interface; fragment if possible.
266 	 * Must be able to put at least 8 bytes per fragment.
267 	 */
268 	if (ip->ip_off & IP_DF) {
269 		error = EMSGSIZE;
270 		ipstat.ips_cantfrag++;
271 		goto bad;
272 	}
273 	len = (ifp->if_mtu - hlen) &~ 7;
274 	if (len < 8) {
275 		error = EMSGSIZE;
276 		goto bad;
277 	}
278 
279     {
280 	int mhlen, firstlen = len;
281 	struct mbuf **mnext = &m->m_nextpkt;
282 
283 	/*
284 	 * Loop through length of segment after first fragment,
285 	 * make new header and copy data of each part and link onto chain.
286 	 */
287 	m0 = m;
288 	mhlen = sizeof (struct ip);
289 	for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
290 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
291 		if (m == 0) {
292 			error = ENOBUFS;
293 			ipstat.ips_odropped++;
294 			goto sendorfree;
295 		}
296 		m->m_data += max_linkhdr;
297 		mhip = mtod(m, struct ip *);
298 		*mhip = *ip;
299 		if (hlen > sizeof (struct ip)) {
300 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
301 			mhip->ip_hl = mhlen >> 2;
302 		}
303 		m->m_len = mhlen;
304 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
305 		if (ip->ip_off & IP_MF)
306 			mhip->ip_off |= IP_MF;
307 		if (off + len >= (u_short)ip->ip_len)
308 			len = (u_short)ip->ip_len - off;
309 		else
310 			mhip->ip_off |= IP_MF;
311 		mhip->ip_len = htons((u_short)(len + mhlen));
312 		m->m_next = m_copy(m0, off, len);
313 		if (m->m_next == 0) {
314 			error = ENOBUFS;	/* ??? */
315 			ipstat.ips_odropped++;
316 			goto sendorfree;
317 		}
318 		m->m_pkthdr.len = mhlen + len;
319 		m->m_pkthdr.rcvif = (struct ifnet *)0;
320 		mhip->ip_off = htons((u_short)mhip->ip_off);
321 		mhip->ip_sum = 0;
322 		mhip->ip_sum = in_cksum(m, mhlen);
323 		*mnext = m;
324 		mnext = &m->m_nextpkt;
325 		ipstat.ips_ofragments++;
326 	}
327 	/*
328 	 * Update first fragment by trimming what's been copied out
329 	 * and updating header, then send each fragment (in order).
330 	 */
331 	m = m0;
332 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
333 	m->m_pkthdr.len = hlen + firstlen;
334 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
335 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
336 	ip->ip_sum = 0;
337 	ip->ip_sum = in_cksum(m, hlen);
338 sendorfree:
339 	for (m = m0; m; m = m0) {
340 		m0 = m->m_nextpkt;
341 		m->m_nextpkt = 0;
342 		if (error == 0)
343 			error = (*ifp->if_output)(ifp, m,
344 			    (struct sockaddr *)dst, ro->ro_rt);
345 		else
346 			m_freem(m);
347 	}
348 
349 	if (error == 0)
350 		ipstat.ips_fragmented++;
351     }
352 done:
353 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
354 		RTFREE(ro->ro_rt);
355 	return (error);
356 bad:
357 	m_freem(m0);
358 	goto done;
359 }
360 
361 /*
362  * Insert IP options into preformed packet.
363  * Adjust IP destination as required for IP source routing,
364  * as indicated by a non-zero in_addr at the start of the options.
365  */
366 struct mbuf *
367 ip_insertoptions(m, opt, phlen)
368 	register struct mbuf *m;
369 	struct mbuf *opt;
370 	int *phlen;
371 {
372 	register struct ipoption *p = mtod(opt, struct ipoption *);
373 	struct mbuf *n;
374 	register struct ip *ip = mtod(m, struct ip *);
375 	unsigned optlen;
376 
377 	optlen = opt->m_len - sizeof(p->ipopt_dst);
378 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
379 		return (m);		/* XXX should fail */
380 	if (p->ipopt_dst.s_addr)
381 		ip->ip_dst = p->ipopt_dst;
382 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
383 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
384 		if (n == 0)
385 			return (m);
386 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
387 		m->m_len -= sizeof(struct ip);
388 		m->m_data += sizeof(struct ip);
389 		n->m_next = m;
390 		m = n;
391 		m->m_len = optlen + sizeof(struct ip);
392 		m->m_data += max_linkhdr;
393 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
394 	} else {
395 		m->m_data -= optlen;
396 		m->m_len += optlen;
397 		m->m_pkthdr.len += optlen;
398 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
399 	}
400 	ip = mtod(m, struct ip *);
401 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
402 	*phlen = sizeof(struct ip) + optlen;
403 	ip->ip_len += optlen;
404 	return (m);
405 }
406 
407 /*
408  * Copy options from ip to jp,
409  * omitting those not copied during fragmentation.
410  */
411 int
412 ip_optcopy(ip, jp)
413 	struct ip *ip, *jp;
414 {
415 	register u_char *cp, *dp;
416 	int opt, optlen, cnt;
417 
418 	cp = (u_char *)(ip + 1);
419 	dp = (u_char *)(jp + 1);
420 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
421 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
422 		opt = cp[0];
423 		if (opt == IPOPT_EOL)
424 			break;
425 		if (opt == IPOPT_NOP) {
426 			/* Preserve for IP mcast tunnel's LSRR alignment. */
427 			*dp++ = IPOPT_NOP;
428 			optlen = 1;
429 			continue;
430 		} else
431 			optlen = cp[IPOPT_OLEN];
432 		/* bogus lengths should have been caught by ip_dooptions */
433 		if (optlen > cnt)
434 			optlen = cnt;
435 		if (IPOPT_COPIED(opt)) {
436 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
437 			dp += optlen;
438 		}
439 	}
440 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
441 		*dp++ = IPOPT_EOL;
442 	return (optlen);
443 }
444 
445 /*
446  * IP socket option processing.
447  */
448 int
449 ip_ctloutput(op, so, level, optname, mp)
450 	int op;
451 	struct socket *so;
452 	int level, optname;
453 	struct mbuf **mp;
454 {
455 	register struct inpcb *inp = sotoinpcb(so);
456 	register struct mbuf *m = *mp;
457 	register int optval;
458 	int error = 0;
459 
460 	if (level != IPPROTO_IP)
461 		goto freeit;
462 	else switch (op) {
463 
464 	case PRCO_SETOPT:
465 		switch (optname) {
466 		case IP_OPTIONS:
467 #ifdef notyet
468 		case IP_RETOPTS:
469 			return (ip_pcbopts(optname, &inp->inp_options, m));
470 #else
471 			return (ip_pcbopts(&inp->inp_options, m));
472 #endif
473 
474 		case IP_TOS:
475 		case IP_TTL:
476 		case IP_RECVOPTS:
477 		case IP_RECVRETOPTS:
478 		case IP_RECVDSTADDR:
479 			if (m->m_len != sizeof(int))
480 				error = EINVAL;
481 			else {
482 				optval = *mtod(m, int *);
483 				switch (optname) {
484 
485 				case IP_TOS:
486 					inp->inp_ip.ip_tos = optval;
487 					break;
488 
489 				case IP_TTL:
490 					inp->inp_ip.ip_ttl = optval;
491 					break;
492 #define	OPTSET(bit) \
493 	if (optval) \
494 		inp->inp_flags |= bit; \
495 	else \
496 		inp->inp_flags &= ~bit;
497 
498 				case IP_RECVOPTS:
499 					OPTSET(INP_RECVOPTS);
500 					break;
501 
502 				case IP_RECVRETOPTS:
503 					OPTSET(INP_RECVRETOPTS);
504 					break;
505 
506 				case IP_RECVDSTADDR:
507 					OPTSET(INP_RECVDSTADDR);
508 					break;
509 				}
510 			}
511 			break;
512 #undef OPTSET
513 
514 		case IP_MULTICAST_IF:
515 		case IP_MULTICAST_TTL:
516 		case IP_MULTICAST_LOOP:
517 		case IP_ADD_MEMBERSHIP:
518 		case IP_DROP_MEMBERSHIP:
519 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
520 			break;
521 
522 		freeit:
523 		default:
524 			error = EINVAL;
525 			break;
526 		}
527 		if (m)
528 			(void)m_free(m);
529 		break;
530 
531 	case PRCO_GETOPT:
532 		switch (optname) {
533 		case IP_OPTIONS:
534 		case IP_RETOPTS:
535 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
536 			if (inp->inp_options) {
537 				m->m_len = inp->inp_options->m_len;
538 				bcopy(mtod(inp->inp_options, caddr_t),
539 				    mtod(m, caddr_t), (unsigned)m->m_len);
540 			} else
541 				m->m_len = 0;
542 			break;
543 
544 		case IP_TOS:
545 		case IP_TTL:
546 		case IP_RECVOPTS:
547 		case IP_RECVRETOPTS:
548 		case IP_RECVDSTADDR:
549 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
550 			m->m_len = sizeof(int);
551 			switch (optname) {
552 
553 			case IP_TOS:
554 				optval = inp->inp_ip.ip_tos;
555 				break;
556 
557 			case IP_TTL:
558 				optval = inp->inp_ip.ip_ttl;
559 				break;
560 
561 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
562 
563 			case IP_RECVOPTS:
564 				optval = OPTBIT(INP_RECVOPTS);
565 				break;
566 
567 			case IP_RECVRETOPTS:
568 				optval = OPTBIT(INP_RECVRETOPTS);
569 				break;
570 
571 			case IP_RECVDSTADDR:
572 				optval = OPTBIT(INP_RECVDSTADDR);
573 				break;
574 			}
575 			*mtod(m, int *) = optval;
576 			break;
577 
578 		case IP_MULTICAST_IF:
579 		case IP_MULTICAST_TTL:
580 		case IP_MULTICAST_LOOP:
581 		case IP_ADD_MEMBERSHIP:
582 		case IP_DROP_MEMBERSHIP:
583 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
584 			break;
585 
586 		default:
587 			error = EINVAL;
588 			break;
589 		}
590 		break;
591 	}
592 	return (error);
593 }
594 
595 /*
596  * Set up IP options in pcb for insertion in output packets.
597  * Store in mbuf with pointer in pcbopt, adding pseudo-option
598  * with destination address if source routed.
599  */
600 int
601 #ifdef notyet
602 ip_pcbopts(optname, pcbopt, m)
603 	int optname;
604 #else
605 ip_pcbopts(pcbopt, m)
606 #endif
607 	struct mbuf **pcbopt;
608 	register struct mbuf *m;
609 {
610 	register cnt, optlen;
611 	register u_char *cp;
612 	u_char opt;
613 
614 	/* turn off any old options */
615 	if (*pcbopt)
616 		(void)m_free(*pcbopt);
617 	*pcbopt = 0;
618 	if (m == (struct mbuf *)0 || m->m_len == 0) {
619 		/*
620 		 * Only turning off any previous options.
621 		 */
622 		if (m)
623 			(void)m_free(m);
624 		return (0);
625 	}
626 
627 #ifndef	vax
628 	if (m->m_len % sizeof(long))
629 		goto bad;
630 #endif
631 	/*
632 	 * IP first-hop destination address will be stored before
633 	 * actual options; move other options back
634 	 * and clear it when none present.
635 	 */
636 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
637 		goto bad;
638 	cnt = m->m_len;
639 	m->m_len += sizeof(struct in_addr);
640 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
641 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
642 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
643 
644 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
645 		opt = cp[IPOPT_OPTVAL];
646 		if (opt == IPOPT_EOL)
647 			break;
648 		if (opt == IPOPT_NOP)
649 			optlen = 1;
650 		else {
651 			optlen = cp[IPOPT_OLEN];
652 			if (optlen <= IPOPT_OLEN || optlen > cnt)
653 				goto bad;
654 		}
655 		switch (opt) {
656 
657 		default:
658 			break;
659 
660 		case IPOPT_LSRR:
661 		case IPOPT_SSRR:
662 			/*
663 			 * user process specifies route as:
664 			 *	->A->B->C->D
665 			 * D must be our final destination (but we can't
666 			 * check that since we may not have connected yet).
667 			 * A is first hop destination, which doesn't appear in
668 			 * actual IP option, but is stored before the options.
669 			 */
670 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
671 				goto bad;
672 			m->m_len -= sizeof(struct in_addr);
673 			cnt -= sizeof(struct in_addr);
674 			optlen -= sizeof(struct in_addr);
675 			cp[IPOPT_OLEN] = optlen;
676 			/*
677 			 * Move first hop before start of options.
678 			 */
679 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
680 			    sizeof(struct in_addr));
681 			/*
682 			 * Then copy rest of options back
683 			 * to close up the deleted entry.
684 			 */
685 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
686 			    sizeof(struct in_addr)),
687 			    (caddr_t)&cp[IPOPT_OFFSET+1],
688 			    (unsigned)cnt + sizeof(struct in_addr));
689 			break;
690 		}
691 	}
692 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
693 		goto bad;
694 	*pcbopt = m;
695 	return (0);
696 
697 bad:
698 	(void)m_free(m);
699 	return (EINVAL);
700 }
701 
702 /*
703  * Set the IP multicast options in response to user setsockopt().
704  */
705 int
706 ip_setmoptions(optname, imop, m)
707 	int optname;
708 	struct ip_moptions **imop;
709 	struct mbuf *m;
710 {
711 	register int error = 0;
712 	u_char loop;
713 	register int i;
714 	struct in_addr addr;
715 	register struct ip_mreq *mreq;
716 	register struct ifnet *ifp;
717 	register struct ip_moptions *imo = *imop;
718 	struct route ro;
719 	register struct sockaddr_in *dst;
720 
721 	if (imo == NULL) {
722 		/*
723 		 * No multicast option buffer attached to the pcb;
724 		 * allocate one and initialize to default values.
725 		 */
726 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
727 		    M_WAITOK);
728 
729 		if (imo == NULL)
730 			return (ENOBUFS);
731 		*imop = imo;
732 		imo->imo_multicast_ifp = NULL;
733 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
734 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
735 		imo->imo_num_memberships = 0;
736 	}
737 
738 	switch (optname) {
739 
740 	case IP_MULTICAST_IF:
741 		/*
742 		 * Select the interface for outgoing multicast packets.
743 		 */
744 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
745 			error = EINVAL;
746 			break;
747 		}
748 		addr = *(mtod(m, struct in_addr *));
749 		/*
750 		 * INADDR_ANY is used to remove a previous selection.
751 		 * When no interface is selected, a default one is
752 		 * chosen every time a multicast packet is sent.
753 		 */
754 		if (addr.s_addr == INADDR_ANY) {
755 			imo->imo_multicast_ifp = NULL;
756 			break;
757 		}
758 		/*
759 		 * The selected interface is identified by its local
760 		 * IP address.  Find the interface and confirm that
761 		 * it supports multicasting.
762 		 */
763 		INADDR_TO_IFP(addr, ifp);
764 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
765 			error = EADDRNOTAVAIL;
766 			break;
767 		}
768 		imo->imo_multicast_ifp = ifp;
769 		break;
770 
771 	case IP_MULTICAST_TTL:
772 		/*
773 		 * Set the IP time-to-live for outgoing multicast packets.
774 		 */
775 		if (m == NULL || m->m_len != 1) {
776 			error = EINVAL;
777 			break;
778 		}
779 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
780 		break;
781 
782 	case IP_MULTICAST_LOOP:
783 		/*
784 		 * Set the loopback flag for outgoing multicast packets.
785 		 * Must be zero or one.
786 		 */
787 		if (m == NULL || m->m_len != 1 ||
788 		   (loop = *(mtod(m, u_char *))) > 1) {
789 			error = EINVAL;
790 			break;
791 		}
792 		imo->imo_multicast_loop = loop;
793 		break;
794 
795 	case IP_ADD_MEMBERSHIP:
796 		/*
797 		 * Add a multicast group membership.
798 		 * Group must be a valid IP multicast address.
799 		 */
800 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
801 			error = EINVAL;
802 			break;
803 		}
804 		mreq = mtod(m, struct ip_mreq *);
805 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
806 			error = EINVAL;
807 			break;
808 		}
809 		/*
810 		 * If no interface address was provided, use the interface of
811 		 * the route to the given multicast address.
812 		 */
813 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
814 			ro.ro_rt = NULL;
815 			dst = (struct sockaddr_in *)&ro.ro_dst;
816 			dst->sin_len = sizeof(*dst);
817 			dst->sin_family = AF_INET;
818 			dst->sin_addr = mreq->imr_multiaddr;
819 			rtalloc(&ro);
820 			if (ro.ro_rt == NULL) {
821 				error = EADDRNOTAVAIL;
822 				break;
823 			}
824 			ifp = ro.ro_rt->rt_ifp;
825 			rtfree(ro.ro_rt);
826 		}
827 		else {
828 			INADDR_TO_IFP(mreq->imr_interface, ifp);
829 		}
830 		/*
831 		 * See if we found an interface, and confirm that it
832 		 * supports multicast.
833 		 */
834 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
835 			error = EADDRNOTAVAIL;
836 			break;
837 		}
838 		/*
839 		 * See if the membership already exists or if all the
840 		 * membership slots are full.
841 		 */
842 		for (i = 0; i < imo->imo_num_memberships; ++i) {
843 			if (imo->imo_membership[i]->inm_ifp == ifp &&
844 			    imo->imo_membership[i]->inm_addr.s_addr
845 						== mreq->imr_multiaddr.s_addr)
846 				break;
847 		}
848 		if (i < imo->imo_num_memberships) {
849 			error = EADDRINUSE;
850 			break;
851 		}
852 		if (i == IP_MAX_MEMBERSHIPS) {
853 			error = ETOOMANYREFS;
854 			break;
855 		}
856 		/*
857 		 * Everything looks good; add a new record to the multicast
858 		 * address list for the given interface.
859 		 */
860 		if ((imo->imo_membership[i] =
861 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
862 			error = ENOBUFS;
863 			break;
864 		}
865 		++imo->imo_num_memberships;
866 		break;
867 
868 	case IP_DROP_MEMBERSHIP:
869 		/*
870 		 * Drop a multicast group membership.
871 		 * Group must be a valid IP multicast address.
872 		 */
873 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
874 			error = EINVAL;
875 			break;
876 		}
877 		mreq = mtod(m, struct ip_mreq *);
878 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
879 			error = EINVAL;
880 			break;
881 		}
882 		/*
883 		 * If an interface address was specified, get a pointer
884 		 * to its ifnet structure.
885 		 */
886 		if (mreq->imr_interface.s_addr == INADDR_ANY)
887 			ifp = NULL;
888 		else {
889 			INADDR_TO_IFP(mreq->imr_interface, ifp);
890 			if (ifp == NULL) {
891 				error = EADDRNOTAVAIL;
892 				break;
893 			}
894 		}
895 		/*
896 		 * Find the membership in the membership array.
897 		 */
898 		for (i = 0; i < imo->imo_num_memberships; ++i) {
899 			if ((ifp == NULL ||
900 			     imo->imo_membership[i]->inm_ifp == ifp) &&
901 			     imo->imo_membership[i]->inm_addr.s_addr ==
902 			     mreq->imr_multiaddr.s_addr)
903 				break;
904 		}
905 		if (i == imo->imo_num_memberships) {
906 			error = EADDRNOTAVAIL;
907 			break;
908 		}
909 		/*
910 		 * Give up the multicast address record to which the
911 		 * membership points.
912 		 */
913 		in_delmulti(imo->imo_membership[i]);
914 		/*
915 		 * Remove the gap in the membership array.
916 		 */
917 		for (++i; i < imo->imo_num_memberships; ++i)
918 			imo->imo_membership[i-1] = imo->imo_membership[i];
919 		--imo->imo_num_memberships;
920 		break;
921 
922 	default:
923 		error = EOPNOTSUPP;
924 		break;
925 	}
926 
927 	/*
928 	 * If all options have default values, no need to keep the mbuf.
929 	 */
930 	if (imo->imo_multicast_ifp == NULL &&
931 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
932 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
933 	    imo->imo_num_memberships == 0) {
934 		free(*imop, M_IPMOPTS);
935 		*imop = NULL;
936 	}
937 
938 	return (error);
939 }
940 
941 /*
942  * Return the IP multicast options in response to user getsockopt().
943  */
944 int
945 ip_getmoptions(optname, imo, mp)
946 	int optname;
947 	register struct ip_moptions *imo;
948 	register struct mbuf **mp;
949 {
950 	u_char *ttl;
951 	u_char *loop;
952 	struct in_addr *addr;
953 	struct in_ifaddr *ia;
954 
955 	*mp = m_get(M_WAIT, MT_SOOPTS);
956 
957 	switch (optname) {
958 
959 	case IP_MULTICAST_IF:
960 		addr = mtod(*mp, struct in_addr *);
961 		(*mp)->m_len = sizeof(struct in_addr);
962 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
963 			addr->s_addr = INADDR_ANY;
964 		else {
965 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
966 			addr->s_addr = (ia == NULL) ? INADDR_ANY
967 					: IA_SIN(ia)->sin_addr.s_addr;
968 		}
969 		return (0);
970 
971 	case IP_MULTICAST_TTL:
972 		ttl = mtod(*mp, u_char *);
973 		(*mp)->m_len = 1;
974 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
975 				     : imo->imo_multicast_ttl;
976 		return (0);
977 
978 	case IP_MULTICAST_LOOP:
979 		loop = mtod(*mp, u_char *);
980 		(*mp)->m_len = 1;
981 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
982 				      : imo->imo_multicast_loop;
983 		return (0);
984 
985 	default:
986 		return (EOPNOTSUPP);
987 	}
988 }
989 
990 /*
991  * Discard the IP multicast options.
992  */
993 void
994 ip_freemoptions(imo)
995 	register struct ip_moptions *imo;
996 {
997 	register int i;
998 
999 	if (imo != NULL) {
1000 		for (i = 0; i < imo->imo_num_memberships; ++i)
1001 			in_delmulti(imo->imo_membership[i]);
1002 		free(imo, M_IPMOPTS);
1003 	}
1004 }
1005 
1006 /*
1007  * Routine called from ip_output() to loop back a copy of an IP multicast
1008  * packet to the input queue of a specified interface.  Note that this
1009  * calls the output routine of the loopback "driver", but with an interface
1010  * pointer that might NOT be &loif -- easier than replicating that code here.
1011  */
1012 static void
1013 ip_mloopback(ifp, m, dst)
1014 	struct ifnet *ifp;
1015 	register struct mbuf *m;
1016 	register struct sockaddr_in *dst;
1017 {
1018 	register struct ip *ip;
1019 	struct mbuf *copym;
1020 
1021 	copym = m_copy(m, 0, M_COPYALL);
1022 	if (copym != NULL) {
1023 		/*
1024 		 * We don't bother to fragment if the IP length is greater
1025 		 * than the interface's MTU.  Can this possibly matter?
1026 		 */
1027 		ip = mtod(copym, struct ip *);
1028 		ip->ip_len = htons((u_short)ip->ip_len);
1029 		ip->ip_off = htons((u_short)ip->ip_off);
1030 		ip->ip_sum = 0;
1031 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1032 		(void) looutput(ifp, copym, (struct sockaddr *)dst);
1033 	}
1034 }
1035