xref: /netbsd-src/sys/netinet/ip_output.c (revision c41a4eebefede43f6950f838a387dc18c6a431bf)
1 /*	$NetBSD: ip_output.c,v 1.41 1998/01/07 22:51:25 lukem 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 #include <net/pfil.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 
58 #ifdef vax
59 #include <machine/mtpr.h>
60 #endif
61 
62 #include <machine/stdarg.h>
63 
64 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
65 static void ip_mloopback
66 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
67 
68 /*
69  * IP output.  The packet in mbuf chain m contains a skeletal IP
70  * header (with len, off, ttl, proto, tos, src, dst).
71  * The mbuf chain containing the packet will be freed.
72  * The mbuf opt, if present, will not be freed.
73  */
74 int
75 #if __STDC__
76 ip_output(struct mbuf *m0, ...)
77 #else
78 ip_output(m0, va_alist)
79 	struct mbuf *m0;
80 	va_dcl
81 #endif
82 {
83 	register struct ip *ip, *mhip;
84 	register struct ifnet *ifp;
85 	register struct mbuf *m = m0;
86 	register int hlen = sizeof (struct ip);
87 	int len, off, error = 0;
88 	struct route iproute;
89 	struct sockaddr_in *dst;
90 	struct in_ifaddr *ia;
91 	struct mbuf *opt;
92 	struct route *ro;
93 	int flags;
94 	int *mtu_p;
95 	struct ip_moptions *imo;
96 	va_list ap;
97 #ifdef PFIL_HOOKS
98 	struct packet_filter_hook *pfh;
99 	struct mbuf *m1;
100 	int rv;
101 #endif /* PFIL_HOOKS */
102 
103 	va_start(ap, m0);
104 	opt = va_arg(ap, struct mbuf *);
105 	ro = va_arg(ap, struct route *);
106 	flags = va_arg(ap, int);
107 	imo = va_arg(ap, struct ip_moptions *);
108 	if (flags & IP_RETURNMTU)
109 		mtu_p = va_arg(ap, int *);
110 	else
111 		mtu_p = NULL;
112 	va_end(ap);
113 
114 #ifdef	DIAGNOSTIC
115 	if ((m->m_flags & M_PKTHDR) == 0)
116 		panic("ip_output no HDR");
117 #endif
118 	if (opt) {
119 		m = ip_insertoptions(m, opt, &len);
120 		hlen = len;
121 	}
122 	ip = mtod(m, struct ip *);
123 	/*
124 	 * Fill in IP header.
125 	 */
126 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
127 		ip->ip_v = IPVERSION;
128 		ip->ip_off &= IP_DF;
129 		ip->ip_id = htons(ip_id++);
130 		ip->ip_hl = hlen >> 2;
131 		ipstat.ips_localout++;
132 	} else {
133 		hlen = ip->ip_hl << 2;
134 	}
135 	/*
136 	 * Route packet.
137 	 */
138 	if (ro == 0) {
139 		ro = &iproute;
140 		bzero((caddr_t)ro, sizeof (*ro));
141 	}
142 	dst = satosin(&ro->ro_dst);
143 	/*
144 	 * If there is a cached route,
145 	 * check that it is to the same destination
146 	 * and is still up.  If not, free it and try again.
147 	 */
148 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
149 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
150 		RTFREE(ro->ro_rt);
151 		ro->ro_rt = (struct rtentry *)0;
152 	}
153 	if (ro->ro_rt == 0) {
154 		dst->sin_family = AF_INET;
155 		dst->sin_len = sizeof(*dst);
156 		dst->sin_addr = ip->ip_dst;
157 	}
158 	/*
159 	 * If routing to interface only,
160 	 * short circuit routing lookup.
161 	 */
162 	if (flags & IP_ROUTETOIF) {
163 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
164 			ipstat.ips_noroute++;
165 			error = ENETUNREACH;
166 			goto bad;
167 		}
168 		ifp = ia->ia_ifp;
169 		ip->ip_ttl = 1;
170 	} else {
171 		if (ro->ro_rt == 0)
172 			rtalloc(ro);
173 		if (ro->ro_rt == 0) {
174 			ipstat.ips_noroute++;
175 			error = EHOSTUNREACH;
176 			goto bad;
177 		}
178 		ia = ifatoia(ro->ro_rt->rt_ifa);
179 		ifp = ro->ro_rt->rt_ifp;
180 		ro->ro_rt->rt_use++;
181 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
182 			dst = satosin(ro->ro_rt->rt_gateway);
183 	}
184 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
185 		struct in_multi *inm;
186 
187 		m->m_flags |= M_MCAST;
188 		/*
189 		 * IP destination address is multicast.  Make sure "dst"
190 		 * still points to the address in "ro".  (It may have been
191 		 * changed to point to a gateway address, above.)
192 		 */
193 		dst = satosin(&ro->ro_dst);
194 		/*
195 		 * See if the caller provided any multicast options
196 		 */
197 		if (imo != NULL) {
198 			ip->ip_ttl = imo->imo_multicast_ttl;
199 			if (imo->imo_multicast_ifp != NULL)
200 				ifp = imo->imo_multicast_ifp;
201 		} else
202 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
203 		/*
204 		 * Confirm that the outgoing interface supports multicast.
205 		 */
206 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
207 			ipstat.ips_noroute++;
208 			error = ENETUNREACH;
209 			goto bad;
210 		}
211 		/*
212 		 * If source address not specified yet, use address
213 		 * of outgoing interface.
214 		 */
215 		if (in_nullhost(ip->ip_src)) {
216 			register struct in_ifaddr *ia;
217 
218 			for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next)
219 				if (ia->ia_ifp == ifp) {
220 					ip->ip_src = ia->ia_addr.sin_addr;
221 					break;
222 				}
223 		}
224 
225 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
226 		if (inm != NULL &&
227 		   (imo == NULL || imo->imo_multicast_loop)) {
228 			/*
229 			 * If we belong to the destination multicast group
230 			 * on the outgoing interface, and the caller did not
231 			 * forbid loopback, loop back a copy.
232 			 */
233 			ip_mloopback(ifp, m, dst);
234 		}
235 #ifdef MROUTING
236 		else {
237 			/*
238 			 * If we are acting as a multicast router, perform
239 			 * multicast forwarding as if the packet had just
240 			 * arrived on the interface to which we are about
241 			 * to send.  The multicast forwarding function
242 			 * recursively calls this function, using the
243 			 * IP_FORWARDING flag to prevent infinite recursion.
244 			 *
245 			 * Multicasts that are looped back by ip_mloopback(),
246 			 * above, will be forwarded by the ip_input() routine,
247 			 * if necessary.
248 			 */
249 			extern struct socket *ip_mrouter;
250 
251 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
252 				if (ip_mforward(m, ifp) != 0) {
253 					m_freem(m);
254 					goto done;
255 				}
256 			}
257 		}
258 #endif
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->if_flags & IFF_LOOPBACK) != 0) {
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 (in_nullhost(ip->ip_src))
280 		ip->ip_src = ia->ia_addr.sin_addr;
281 #endif
282 	/*
283 	 * Look for broadcast address and
284 	 * and verify user is allowed to send
285 	 * such a packet.
286 	 */
287 	if (in_broadcast(dst->sin_addr, ifp)) {
288 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
289 			error = EADDRNOTAVAIL;
290 			goto bad;
291 		}
292 		if ((flags & IP_ALLOWBROADCAST) == 0) {
293 			error = EACCES;
294 			goto bad;
295 		}
296 		/* don't allow broadcast messages to be fragmented */
297 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
298 			error = EMSGSIZE;
299 			goto bad;
300 		}
301 		m->m_flags |= M_BCAST;
302 	} else
303 		m->m_flags &= ~M_BCAST;
304 
305 #ifdef PFIL_HOOKS
306 	/*
307 	 * Run through list of hooks for output packets.
308 	 */
309 	m1 = m;
310 	for (pfh = pfil_hook_get(PFIL_OUT); pfh; pfh = pfh->pfil_link.le_next)
311 		if (pfh->pfil_func) {
312 		    	rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
313 			if (rv) {
314 				error = EHOSTUNREACH;
315 				goto done;
316 			}
317 			ip = mtod(m = m1, struct ip *);
318 		}
319 #endif /* PFIL_HOOKS */
320 sendit:
321 	/*
322 	 * If small enough for interface, can just send directly.
323 	 */
324 	if ((u_int16_t)ip->ip_len <= ifp->if_mtu) {
325 		ip->ip_len = htons((u_int16_t)ip->ip_len);
326 		ip->ip_off = htons((u_int16_t)ip->ip_off);
327 		ip->ip_sum = 0;
328 		ip->ip_sum = in_cksum(m, hlen);
329 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
330 		goto done;
331 	}
332 	/*
333 	 * Too large for interface; fragment if possible.
334 	 * Must be able to put at least 8 bytes per fragment.
335 	 */
336 	if (ip->ip_off & IP_DF) {
337 		if (flags & IP_RETURNMTU)
338 			*mtu_p = ifp->if_mtu;
339 		error = EMSGSIZE;
340 		ipstat.ips_cantfrag++;
341 		goto bad;
342 	}
343 	len = (ifp->if_mtu - hlen) &~ 7;
344 	if (len < 8) {
345 		error = EMSGSIZE;
346 		goto bad;
347 	}
348 
349     {
350 	int mhlen, firstlen = len;
351 	struct mbuf **mnext = &m->m_nextpkt;
352 
353 	/*
354 	 * Loop through length of segment after first fragment,
355 	 * make new header and copy data of each part and link onto chain.
356 	 */
357 	m0 = m;
358 	mhlen = sizeof (struct ip);
359 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
360 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
361 		if (m == 0) {
362 			error = ENOBUFS;
363 			ipstat.ips_odropped++;
364 			goto sendorfree;
365 		}
366 		*mnext = m;
367 		mnext = &m->m_nextpkt;
368 		m->m_data += max_linkhdr;
369 		mhip = mtod(m, struct ip *);
370 		*mhip = *ip;
371 		if (hlen > sizeof (struct ip)) {
372 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
373 			mhip->ip_hl = mhlen >> 2;
374 		}
375 		m->m_len = mhlen;
376 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
377 		if (ip->ip_off & IP_MF)
378 			mhip->ip_off |= IP_MF;
379 		if (off + len >= (u_int16_t)ip->ip_len)
380 			len = (u_int16_t)ip->ip_len - off;
381 		else
382 			mhip->ip_off |= IP_MF;
383 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
384 		m->m_next = m_copy(m0, off, len);
385 		if (m->m_next == 0) {
386 			error = ENOBUFS;	/* ??? */
387 			ipstat.ips_odropped++;
388 			goto sendorfree;
389 		}
390 		m->m_pkthdr.len = mhlen + len;
391 		m->m_pkthdr.rcvif = (struct ifnet *)0;
392 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
393 		mhip->ip_sum = 0;
394 		mhip->ip_sum = in_cksum(m, mhlen);
395 		ipstat.ips_ofragments++;
396 	}
397 	/*
398 	 * Update first fragment by trimming what's been copied out
399 	 * and updating header, then send each fragment (in order).
400 	 */
401 	m = m0;
402 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
403 	m->m_pkthdr.len = hlen + firstlen;
404 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
405 	ip->ip_off = htons((u_int16_t)(ip->ip_off | IP_MF));
406 	ip->ip_sum = 0;
407 	ip->ip_sum = in_cksum(m, hlen);
408 sendorfree:
409 	for (m = m0; m; m = m0) {
410 		m0 = m->m_nextpkt;
411 		m->m_nextpkt = 0;
412 		if (error == 0)
413 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
414 			    ro->ro_rt);
415 		else
416 			m_freem(m);
417 	}
418 
419 	if (error == 0)
420 		ipstat.ips_fragmented++;
421     }
422 done:
423 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
424 		RTFREE(ro->ro_rt);
425 		ro->ro_rt = 0;
426 	}
427 	return (error);
428 bad:
429 	m_freem(m);
430 	goto done;
431 }
432 
433 /*
434  * Insert IP options into preformed packet.
435  * Adjust IP destination as required for IP source routing,
436  * as indicated by a non-zero in_addr at the start of the options.
437  */
438 static struct mbuf *
439 ip_insertoptions(m, opt, phlen)
440 	register struct mbuf *m;
441 	struct mbuf *opt;
442 	int *phlen;
443 {
444 	register struct ipoption *p = mtod(opt, struct ipoption *);
445 	struct mbuf *n;
446 	register struct ip *ip = mtod(m, struct ip *);
447 	unsigned optlen;
448 
449 	optlen = opt->m_len - sizeof(p->ipopt_dst);
450 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
451 		return (m);		/* XXX should fail */
452 	if (!in_nullhost(p->ipopt_dst))
453 		ip->ip_dst = p->ipopt_dst;
454 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
455 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
456 		if (n == 0)
457 			return (m);
458 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
459 		m->m_len -= sizeof(struct ip);
460 		m->m_data += sizeof(struct ip);
461 		n->m_next = m;
462 		m = n;
463 		m->m_len = optlen + sizeof(struct ip);
464 		m->m_data += max_linkhdr;
465 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
466 	} else {
467 		m->m_data -= optlen;
468 		m->m_len += optlen;
469 		m->m_pkthdr.len += optlen;
470 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
471 	}
472 	ip = mtod(m, struct ip *);
473 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
474 	*phlen = sizeof(struct ip) + optlen;
475 	ip->ip_len += optlen;
476 	return (m);
477 }
478 
479 /*
480  * Copy options from ip to jp,
481  * omitting those not copied during fragmentation.
482  */
483 int
484 ip_optcopy(ip, jp)
485 	struct ip *ip, *jp;
486 {
487 	register u_char *cp, *dp;
488 	int opt, optlen, cnt;
489 
490 	cp = (u_char *)(ip + 1);
491 	dp = (u_char *)(jp + 1);
492 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
493 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
494 		opt = cp[0];
495 		if (opt == IPOPT_EOL)
496 			break;
497 		if (opt == IPOPT_NOP) {
498 			/* Preserve for IP mcast tunnel's LSRR alignment. */
499 			*dp++ = IPOPT_NOP;
500 			optlen = 1;
501 			continue;
502 		} else
503 			optlen = cp[IPOPT_OLEN];
504 		/* bogus lengths should have been caught by ip_dooptions */
505 		if (optlen > cnt)
506 			optlen = cnt;
507 		if (IPOPT_COPIED(opt)) {
508 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
509 			dp += optlen;
510 		}
511 	}
512 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
513 		*dp++ = IPOPT_EOL;
514 	return (optlen);
515 }
516 
517 /*
518  * IP socket option processing.
519  */
520 int
521 ip_ctloutput(op, so, level, optname, mp)
522 	int op;
523 	struct socket *so;
524 	int level, optname;
525 	struct mbuf **mp;
526 {
527 	register struct inpcb *inp = sotoinpcb(so);
528 	register struct mbuf *m = *mp;
529 	register int optval = 0;
530 	int error = 0;
531 
532 	if (level != IPPROTO_IP) {
533 		error = EINVAL;
534 		if (op == PRCO_SETOPT && *mp)
535 			(void) m_free(*mp);
536 	} else switch (op) {
537 
538 	case PRCO_SETOPT:
539 		switch (optname) {
540 		case IP_OPTIONS:
541 #ifdef notyet
542 		case IP_RETOPTS:
543 			return (ip_pcbopts(optname, &inp->inp_options, m));
544 #else
545 			return (ip_pcbopts(&inp->inp_options, m));
546 #endif
547 
548 		case IP_TOS:
549 		case IP_TTL:
550 		case IP_RECVOPTS:
551 		case IP_RECVRETOPTS:
552 		case IP_RECVDSTADDR:
553 		case IP_RECVIF:
554 			if (m == NULL || m->m_len != sizeof(int))
555 				error = EINVAL;
556 			else {
557 				optval = *mtod(m, int *);
558 				switch (optname) {
559 
560 				case IP_TOS:
561 					inp->inp_ip.ip_tos = optval;
562 					break;
563 
564 				case IP_TTL:
565 					inp->inp_ip.ip_ttl = optval;
566 					break;
567 #define	OPTSET(bit) \
568 	if (optval) \
569 		inp->inp_flags |= bit; \
570 	else \
571 		inp->inp_flags &= ~bit;
572 
573 				case IP_RECVOPTS:
574 					OPTSET(INP_RECVOPTS);
575 					break;
576 
577 				case IP_RECVRETOPTS:
578 					OPTSET(INP_RECVRETOPTS);
579 					break;
580 
581 				case IP_RECVDSTADDR:
582 					OPTSET(INP_RECVDSTADDR);
583 					break;
584 
585 				case IP_RECVIF:
586 					OPTSET(INP_RECVIF);
587 					break;
588 				}
589 			}
590 			break;
591 #undef OPTSET
592 
593 		case IP_MULTICAST_IF:
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 		case IP_PORTRANGE:
602 			if (m == 0 || m->m_len != sizeof(int))
603 				error = EINVAL;
604 			else {
605 				optval = *mtod(m, int *);
606 
607 				switch (optval) {
608 
609 				case IP_PORTRANGE_DEFAULT:
610 				case IP_PORTRANGE_HIGH:
611 					inp->inp_flags &= ~(INP_LOWPORT);
612 					break;
613 
614 				case IP_PORTRANGE_LOW:
615 					inp->inp_flags |= INP_LOWPORT;
616 					break;
617 
618 				default:
619 					error = EINVAL;
620 					break;
621 				}
622 			}
623 			break;
624 
625 		default:
626 			error = ENOPROTOOPT;
627 			break;
628 		}
629 		if (m)
630 			(void)m_free(m);
631 		break;
632 
633 	case PRCO_GETOPT:
634 		switch (optname) {
635 		case IP_OPTIONS:
636 		case IP_RETOPTS:
637 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
638 			if (inp->inp_options) {
639 				m->m_len = inp->inp_options->m_len;
640 				bcopy(mtod(inp->inp_options, caddr_t),
641 				    mtod(m, caddr_t), (unsigned)m->m_len);
642 			} else
643 				m->m_len = 0;
644 			break;
645 
646 		case IP_TOS:
647 		case IP_TTL:
648 		case IP_RECVOPTS:
649 		case IP_RECVRETOPTS:
650 		case IP_RECVDSTADDR:
651 		case IP_RECVIF:
652 		case IP_ERRORMTU:
653 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
654 			m->m_len = sizeof(int);
655 			switch (optname) {
656 
657 			case IP_TOS:
658 				optval = inp->inp_ip.ip_tos;
659 				break;
660 
661 			case IP_TTL:
662 				optval = inp->inp_ip.ip_ttl;
663 				break;
664 
665 			case IP_ERRORMTU:
666 				optval = inp->inp_errormtu;
667 				break;
668 
669 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
670 
671 			case IP_RECVOPTS:
672 				optval = OPTBIT(INP_RECVOPTS);
673 				break;
674 
675 			case IP_RECVRETOPTS:
676 				optval = OPTBIT(INP_RECVRETOPTS);
677 				break;
678 
679 			case IP_RECVDSTADDR:
680 				optval = OPTBIT(INP_RECVDSTADDR);
681 				break;
682 
683 			case IP_RECVIF:
684 				optval = OPTBIT(INP_RECVIF);
685 				break;
686 			}
687 			*mtod(m, int *) = optval;
688 			break;
689 
690 		case IP_MULTICAST_IF:
691 		case IP_MULTICAST_TTL:
692 		case IP_MULTICAST_LOOP:
693 		case IP_ADD_MEMBERSHIP:
694 		case IP_DROP_MEMBERSHIP:
695 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
696 			break;
697 
698 		case IP_PORTRANGE:
699 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
700 			m->m_len = sizeof(int);
701 
702 			if (inp->inp_flags & INP_LOWPORT)
703 				optval = IP_PORTRANGE_LOW;
704 			else
705 				optval = IP_PORTRANGE_DEFAULT;
706 
707 			*mtod(m, int *) = optval;
708 			break;
709 
710 		default:
711 			error = ENOPROTOOPT;
712 			break;
713 		}
714 		break;
715 	}
716 	return (error);
717 }
718 
719 /*
720  * Set up IP options in pcb for insertion in output packets.
721  * Store in mbuf with pointer in pcbopt, adding pseudo-option
722  * with destination address if source routed.
723  */
724 int
725 #ifdef notyet
726 ip_pcbopts(optname, pcbopt, m)
727 	int optname;
728 #else
729 ip_pcbopts(pcbopt, m)
730 #endif
731 	struct mbuf **pcbopt;
732 	register struct mbuf *m;
733 {
734 	register cnt, optlen;
735 	register u_char *cp;
736 	u_char opt;
737 
738 	/* turn off any old options */
739 	if (*pcbopt)
740 		(void)m_free(*pcbopt);
741 	*pcbopt = 0;
742 	if (m == (struct mbuf *)0 || m->m_len == 0) {
743 		/*
744 		 * Only turning off any previous options.
745 		 */
746 		if (m)
747 			(void)m_free(m);
748 		return (0);
749 	}
750 
751 #ifndef	vax
752 	if (m->m_len % sizeof(int32_t))
753 		goto bad;
754 #endif
755 	/*
756 	 * IP first-hop destination address will be stored before
757 	 * actual options; move other options back
758 	 * and clear it when none present.
759 	 */
760 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
761 		goto bad;
762 	cnt = m->m_len;
763 	m->m_len += sizeof(struct in_addr);
764 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
765 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
766 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
767 
768 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
769 		opt = cp[IPOPT_OPTVAL];
770 		if (opt == IPOPT_EOL)
771 			break;
772 		if (opt == IPOPT_NOP)
773 			optlen = 1;
774 		else {
775 			optlen = cp[IPOPT_OLEN];
776 			if (optlen <= IPOPT_OLEN || optlen > cnt)
777 				goto bad;
778 		}
779 		switch (opt) {
780 
781 		default:
782 			break;
783 
784 		case IPOPT_LSRR:
785 		case IPOPT_SSRR:
786 			/*
787 			 * user process specifies route as:
788 			 *	->A->B->C->D
789 			 * D must be our final destination (but we can't
790 			 * check that since we may not have connected yet).
791 			 * A is first hop destination, which doesn't appear in
792 			 * actual IP option, but is stored before the options.
793 			 */
794 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
795 				goto bad;
796 			m->m_len -= sizeof(struct in_addr);
797 			cnt -= sizeof(struct in_addr);
798 			optlen -= sizeof(struct in_addr);
799 			cp[IPOPT_OLEN] = optlen;
800 			/*
801 			 * Move first hop before start of options.
802 			 */
803 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
804 			    sizeof(struct in_addr));
805 			/*
806 			 * Then copy rest of options back
807 			 * to close up the deleted entry.
808 			 */
809 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
810 			    sizeof(struct in_addr)),
811 			    (caddr_t)&cp[IPOPT_OFFSET+1],
812 			    (unsigned)cnt + sizeof(struct in_addr));
813 			break;
814 		}
815 	}
816 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
817 		goto bad;
818 	*pcbopt = m;
819 	return (0);
820 
821 bad:
822 	(void)m_free(m);
823 	return (EINVAL);
824 }
825 
826 /*
827  * Set the IP multicast options in response to user setsockopt().
828  */
829 int
830 ip_setmoptions(optname, imop, m)
831 	int optname;
832 	struct ip_moptions **imop;
833 	struct mbuf *m;
834 {
835 	register int error = 0;
836 	u_char loop;
837 	register int i;
838 	struct in_addr addr;
839 	register struct ip_mreq *mreq;
840 	register struct ifnet *ifp;
841 	register struct ip_moptions *imo = *imop;
842 	struct route ro;
843 	register struct sockaddr_in *dst;
844 
845 	if (imo == NULL) {
846 		/*
847 		 * No multicast option buffer attached to the pcb;
848 		 * allocate one and initialize to default values.
849 		 */
850 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
851 		    M_WAITOK);
852 
853 		if (imo == NULL)
854 			return (ENOBUFS);
855 		*imop = imo;
856 		imo->imo_multicast_ifp = NULL;
857 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
858 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
859 		imo->imo_num_memberships = 0;
860 	}
861 
862 	switch (optname) {
863 
864 	case IP_MULTICAST_IF:
865 		/*
866 		 * Select the interface for outgoing multicast packets.
867 		 */
868 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
869 			error = EINVAL;
870 			break;
871 		}
872 		addr = *(mtod(m, struct in_addr *));
873 		/*
874 		 * INADDR_ANY is used to remove a previous selection.
875 		 * When no interface is selected, a default one is
876 		 * chosen every time a multicast packet is sent.
877 		 */
878 		if (in_nullhost(addr)) {
879 			imo->imo_multicast_ifp = NULL;
880 			break;
881 		}
882 		/*
883 		 * The selected interface is identified by its local
884 		 * IP address.  Find the interface and confirm that
885 		 * it supports multicasting.
886 		 */
887 		INADDR_TO_IFP(addr, ifp);
888 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
889 			error = EADDRNOTAVAIL;
890 			break;
891 		}
892 		imo->imo_multicast_ifp = ifp;
893 		break;
894 
895 	case IP_MULTICAST_TTL:
896 		/*
897 		 * Set the IP time-to-live for outgoing multicast packets.
898 		 */
899 		if (m == NULL || m->m_len != 1) {
900 			error = EINVAL;
901 			break;
902 		}
903 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
904 		break;
905 
906 	case IP_MULTICAST_LOOP:
907 		/*
908 		 * Set the loopback flag for outgoing multicast packets.
909 		 * Must be zero or one.
910 		 */
911 		if (m == NULL || m->m_len != 1 ||
912 		   (loop = *(mtod(m, u_char *))) > 1) {
913 			error = EINVAL;
914 			break;
915 		}
916 		imo->imo_multicast_loop = loop;
917 		break;
918 
919 	case IP_ADD_MEMBERSHIP:
920 		/*
921 		 * Add a multicast group membership.
922 		 * Group must be a valid IP multicast address.
923 		 */
924 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
925 			error = EINVAL;
926 			break;
927 		}
928 		mreq = mtod(m, struct ip_mreq *);
929 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
930 			error = EINVAL;
931 			break;
932 		}
933 		/*
934 		 * If no interface address was provided, use the interface of
935 		 * the route to the given multicast address.
936 		 */
937 		if (in_nullhost(mreq->imr_interface)) {
938 			ro.ro_rt = NULL;
939 			dst = satosin(&ro.ro_dst);
940 			dst->sin_len = sizeof(*dst);
941 			dst->sin_family = AF_INET;
942 			dst->sin_addr = mreq->imr_multiaddr;
943 			rtalloc(&ro);
944 			if (ro.ro_rt == NULL) {
945 				error = EADDRNOTAVAIL;
946 				break;
947 			}
948 			ifp = ro.ro_rt->rt_ifp;
949 			rtfree(ro.ro_rt);
950 		} else {
951 			INADDR_TO_IFP(mreq->imr_interface, ifp);
952 		}
953 		/*
954 		 * See if we found an interface, and confirm that it
955 		 * supports multicast.
956 		 */
957 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
958 			error = EADDRNOTAVAIL;
959 			break;
960 		}
961 		/*
962 		 * See if the membership already exists or if all the
963 		 * membership slots are full.
964 		 */
965 		for (i = 0; i < imo->imo_num_memberships; ++i) {
966 			if (imo->imo_membership[i]->inm_ifp == ifp &&
967 			    in_hosteq(imo->imo_membership[i]->inm_addr,
968 				      mreq->imr_multiaddr))
969 				break;
970 		}
971 		if (i < imo->imo_num_memberships) {
972 			error = EADDRINUSE;
973 			break;
974 		}
975 		if (i == IP_MAX_MEMBERSHIPS) {
976 			error = ETOOMANYREFS;
977 			break;
978 		}
979 		/*
980 		 * Everything looks good; add a new record to the multicast
981 		 * address list for the given interface.
982 		 */
983 		if ((imo->imo_membership[i] =
984 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
985 			error = ENOBUFS;
986 			break;
987 		}
988 		++imo->imo_num_memberships;
989 		break;
990 
991 	case IP_DROP_MEMBERSHIP:
992 		/*
993 		 * Drop a multicast group membership.
994 		 * Group must be a valid IP multicast address.
995 		 */
996 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
997 			error = EINVAL;
998 			break;
999 		}
1000 		mreq = mtod(m, struct ip_mreq *);
1001 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1002 			error = EINVAL;
1003 			break;
1004 		}
1005 		/*
1006 		 * If an interface address was specified, get a pointer
1007 		 * to its ifnet structure.
1008 		 */
1009 		if (in_nullhost(mreq->imr_interface))
1010 			ifp = NULL;
1011 		else {
1012 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1013 			if (ifp == NULL) {
1014 				error = EADDRNOTAVAIL;
1015 				break;
1016 			}
1017 		}
1018 		/*
1019 		 * Find the membership in the membership array.
1020 		 */
1021 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1022 			if ((ifp == NULL ||
1023 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1024 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1025 				       mreq->imr_multiaddr))
1026 				break;
1027 		}
1028 		if (i == imo->imo_num_memberships) {
1029 			error = EADDRNOTAVAIL;
1030 			break;
1031 		}
1032 		/*
1033 		 * Give up the multicast address record to which the
1034 		 * membership points.
1035 		 */
1036 		in_delmulti(imo->imo_membership[i]);
1037 		/*
1038 		 * Remove the gap in the membership array.
1039 		 */
1040 		for (++i; i < imo->imo_num_memberships; ++i)
1041 			imo->imo_membership[i-1] = imo->imo_membership[i];
1042 		--imo->imo_num_memberships;
1043 		break;
1044 
1045 	default:
1046 		error = EOPNOTSUPP;
1047 		break;
1048 	}
1049 
1050 	/*
1051 	 * If all options have default values, no need to keep the mbuf.
1052 	 */
1053 	if (imo->imo_multicast_ifp == NULL &&
1054 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1055 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1056 	    imo->imo_num_memberships == 0) {
1057 		free(*imop, M_IPMOPTS);
1058 		*imop = NULL;
1059 	}
1060 
1061 	return (error);
1062 }
1063 
1064 /*
1065  * Return the IP multicast options in response to user getsockopt().
1066  */
1067 int
1068 ip_getmoptions(optname, imo, mp)
1069 	int optname;
1070 	register struct ip_moptions *imo;
1071 	register struct mbuf **mp;
1072 {
1073 	u_char *ttl;
1074 	u_char *loop;
1075 	struct in_addr *addr;
1076 	struct in_ifaddr *ia;
1077 
1078 	*mp = m_get(M_WAIT, MT_SOOPTS);
1079 
1080 	switch (optname) {
1081 
1082 	case IP_MULTICAST_IF:
1083 		addr = mtod(*mp, struct in_addr *);
1084 		(*mp)->m_len = sizeof(struct in_addr);
1085 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1086 			*addr = zeroin_addr;
1087 		else {
1088 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1089 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1090 		}
1091 		return (0);
1092 
1093 	case IP_MULTICAST_TTL:
1094 		ttl = mtod(*mp, u_char *);
1095 		(*mp)->m_len = 1;
1096 		*ttl = imo ? imo->imo_multicast_ttl
1097 			   : IP_DEFAULT_MULTICAST_TTL;
1098 		return (0);
1099 
1100 	case IP_MULTICAST_LOOP:
1101 		loop = mtod(*mp, u_char *);
1102 		(*mp)->m_len = 1;
1103 		*loop = imo ? imo->imo_multicast_loop
1104 			    : IP_DEFAULT_MULTICAST_LOOP;
1105 		return (0);
1106 
1107 	default:
1108 		return (EOPNOTSUPP);
1109 	}
1110 }
1111 
1112 /*
1113  * Discard the IP multicast options.
1114  */
1115 void
1116 ip_freemoptions(imo)
1117 	register struct ip_moptions *imo;
1118 {
1119 	register int i;
1120 
1121 	if (imo != NULL) {
1122 		for (i = 0; i < imo->imo_num_memberships; ++i)
1123 			in_delmulti(imo->imo_membership[i]);
1124 		free(imo, M_IPMOPTS);
1125 	}
1126 }
1127 
1128 /*
1129  * Routine called from ip_output() to loop back a copy of an IP multicast
1130  * packet to the input queue of a specified interface.  Note that this
1131  * calls the output routine of the loopback "driver", but with an interface
1132  * pointer that might NOT be &loif -- easier than replicating that code here.
1133  */
1134 static void
1135 ip_mloopback(ifp, m, dst)
1136 	struct ifnet *ifp;
1137 	register struct mbuf *m;
1138 	register struct sockaddr_in *dst;
1139 {
1140 	register struct ip *ip;
1141 	struct mbuf *copym;
1142 
1143 	copym = m_copy(m, 0, M_COPYALL);
1144 	if (copym != NULL) {
1145 		/*
1146 		 * We don't bother to fragment if the IP length is greater
1147 		 * than the interface's MTU.  Can this possibly matter?
1148 		 */
1149 		ip = mtod(copym, struct ip *);
1150 		ip->ip_len = htons((u_int16_t)ip->ip_len);
1151 		ip->ip_off = htons((u_int16_t)ip->ip_off);
1152 		ip->ip_sum = 0;
1153 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1154 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1155 	}
1156 }
1157