xref: /netbsd-src/sys/netinet/ip_output.c (revision 220b5c059a84c51ea44107ea8951a57ffaecdc8c)
1 /*	$NetBSD: ip_output.c,v 1.90 2001/11/21 06:28:08 itojun Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * 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. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1990, 1993
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.90 2001/11/21 06:28:08 itojun Exp $");
106 
107 #include "opt_pfil_hooks.h"
108 #include "opt_ipsec.h"
109 #include "opt_mrouting.h"
110 
111 #include <sys/param.h>
112 #include <sys/malloc.h>
113 #include <sys/mbuf.h>
114 #include <sys/errno.h>
115 #include <sys/protosw.h>
116 #include <sys/socket.h>
117 #include <sys/socketvar.h>
118 #include <sys/systm.h>
119 #include <sys/proc.h>
120 
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip_var.h>
131 
132 #ifdef MROUTING
133 #include <netinet/ip_mroute.h>
134 #endif
135 
136 #include <machine/stdarg.h>
137 
138 #ifdef IPSEC
139 #include <netinet6/ipsec.h>
140 #include <netkey/key.h>
141 #include <netkey/key_debug.h>
142 #endif /*IPSEC*/
143 
144 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
145 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
146 static void ip_mloopback
147 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
148 
149 #ifdef PFIL_HOOKS
150 extern struct pfil_head inet_pfil_hook;			/* XXX */
151 #endif
152 
153 /*
154  * IP output.  The packet in mbuf chain m contains a skeletal IP
155  * header (with len, off, ttl, proto, tos, src, dst).
156  * The mbuf chain containing the packet will be freed.
157  * The mbuf opt, if present, will not be freed.
158  */
159 int
160 #if __STDC__
161 ip_output(struct mbuf *m0, ...)
162 #else
163 ip_output(m0, va_alist)
164 	struct mbuf *m0;
165 	va_dcl
166 #endif
167 {
168 	struct ip *ip, *mhip;
169 	struct ifnet *ifp;
170 	struct mbuf *m = m0;
171 	int hlen = sizeof (struct ip);
172 	int len, off, error = 0;
173 	struct route iproute;
174 	struct sockaddr_in *dst;
175 	struct in_ifaddr *ia;
176 	struct mbuf *opt;
177 	struct route *ro;
178 	int flags, sw_csum;
179 	int *mtu_p;
180 	int mtu;
181 	struct ip_moptions *imo;
182 	va_list ap;
183 #ifdef IPSEC
184 	struct socket *so;
185 	struct secpolicy *sp = NULL;
186 #endif /*IPSEC*/
187 	u_int16_t ip_len;
188 
189 	va_start(ap, m0);
190 	opt = va_arg(ap, struct mbuf *);
191 	ro = va_arg(ap, struct route *);
192 	flags = va_arg(ap, int);
193 	imo = va_arg(ap, struct ip_moptions *);
194 	if (flags & IP_RETURNMTU)
195 		mtu_p = va_arg(ap, int *);
196 	else
197 		mtu_p = NULL;
198 	va_end(ap);
199 
200 #ifdef IPSEC
201 	so = ipsec_getsocket(m);
202 	(void)ipsec_setsocket(m, NULL);
203 #endif /*IPSEC*/
204 
205 #ifdef	DIAGNOSTIC
206 	if ((m->m_flags & M_PKTHDR) == 0)
207 		panic("ip_output no HDR");
208 #endif
209 	if (opt) {
210 		m = ip_insertoptions(m, opt, &len);
211 		hlen = len;
212 	}
213 	ip = mtod(m, struct ip *);
214 	/*
215 	 * Fill in IP header.
216 	 */
217 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
218 		ip->ip_v = IPVERSION;
219 		ip->ip_off = 0;
220 		ip->ip_id = htons(ip_id++);
221 		ip->ip_hl = hlen >> 2;
222 		ipstat.ips_localout++;
223 	} else {
224 		hlen = ip->ip_hl << 2;
225 	}
226 	/*
227 	 * Route packet.
228 	 */
229 	if (ro == 0) {
230 		ro = &iproute;
231 		bzero((caddr_t)ro, sizeof (*ro));
232 	}
233 	dst = satosin(&ro->ro_dst);
234 	/*
235 	 * If there is a cached route,
236 	 * check that it is to the same destination
237 	 * and is still up.  If not, free it and try again.
238 	 */
239 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
240 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
241 		RTFREE(ro->ro_rt);
242 		ro->ro_rt = (struct rtentry *)0;
243 	}
244 	if (ro->ro_rt == 0) {
245 		dst->sin_family = AF_INET;
246 		dst->sin_len = sizeof(*dst);
247 		dst->sin_addr = ip->ip_dst;
248 	}
249 	/*
250 	 * If routing to interface only,
251 	 * short circuit routing lookup.
252 	 */
253 	if (flags & IP_ROUTETOIF) {
254 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
255 			ipstat.ips_noroute++;
256 			error = ENETUNREACH;
257 			goto bad;
258 		}
259 		ifp = ia->ia_ifp;
260 		mtu = ifp->if_mtu;
261 		ip->ip_ttl = 1;
262 	} else {
263 		if (ro->ro_rt == 0)
264 			rtalloc(ro);
265 		if (ro->ro_rt == 0) {
266 			ipstat.ips_noroute++;
267 			error = EHOSTUNREACH;
268 			goto bad;
269 		}
270 		ia = ifatoia(ro->ro_rt->rt_ifa);
271 		ifp = ro->ro_rt->rt_ifp;
272 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
273 			mtu = ifp->if_mtu;
274 		ro->ro_rt->rt_use++;
275 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
276 			dst = satosin(ro->ro_rt->rt_gateway);
277 	}
278 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
279 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
280 		struct in_multi *inm;
281 
282 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
283 			M_BCAST : M_MCAST;
284 		/*
285 		 * IP destination address is multicast.  Make sure "dst"
286 		 * still points to the address in "ro".  (It may have been
287 		 * changed to point to a gateway address, above.)
288 		 */
289 		dst = satosin(&ro->ro_dst);
290 		/*
291 		 * See if the caller provided any multicast options
292 		 */
293 		if (imo != NULL) {
294 			ip->ip_ttl = imo->imo_multicast_ttl;
295 			if (imo->imo_multicast_ifp != NULL) {
296 				ifp = imo->imo_multicast_ifp;
297 				mtu = ifp->if_mtu;
298 			}
299 		} else
300 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
301 		/*
302 		 * Confirm that the outgoing interface supports multicast.
303 		 */
304 		if (((m->m_flags & M_MCAST) &&
305 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
306 		    ((m->m_flags & M_BCAST) &&
307 		     (ifp->if_flags & IFF_BROADCAST) == 0))  {
308 			ipstat.ips_noroute++;
309 			error = ENETUNREACH;
310 			goto bad;
311 		}
312 		/*
313 		 * If source address not specified yet, use an address
314 		 * of outgoing interface.
315 		 */
316 		if (in_nullhost(ip->ip_src)) {
317 			struct in_ifaddr *ia;
318 
319 			IFP_TO_IA(ifp, ia);
320 			ip->ip_src = ia->ia_addr.sin_addr;
321 		}
322 
323 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
324 		if (inm != NULL &&
325 		   (imo == NULL || imo->imo_multicast_loop)) {
326 			/*
327 			 * If we belong to the destination multicast group
328 			 * on the outgoing interface, and the caller did not
329 			 * forbid loopback, loop back a copy.
330 			 */
331 			ip_mloopback(ifp, m, dst);
332 		}
333 #ifdef MROUTING
334 		else {
335 			/*
336 			 * If we are acting as a multicast router, perform
337 			 * multicast forwarding as if the packet had just
338 			 * arrived on the interface to which we are about
339 			 * to send.  The multicast forwarding function
340 			 * recursively calls this function, using the
341 			 * IP_FORWARDING flag to prevent infinite recursion.
342 			 *
343 			 * Multicasts that are looped back by ip_mloopback(),
344 			 * above, will be forwarded by the ip_input() routine,
345 			 * if necessary.
346 			 */
347 			extern struct socket *ip_mrouter;
348 
349 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
350 				if (ip_mforward(m, ifp) != 0) {
351 					m_freem(m);
352 					goto done;
353 				}
354 			}
355 		}
356 #endif
357 		/*
358 		 * Multicasts with a time-to-live of zero may be looped-
359 		 * back, above, but must not be transmitted on a network.
360 		 * Also, multicasts addressed to the loopback interface
361 		 * are not sent -- the above call to ip_mloopback() will
362 		 * loop back a copy if this host actually belongs to the
363 		 * destination group on the loopback interface.
364 		 */
365 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
366 			m_freem(m);
367 			goto done;
368 		}
369 
370 		goto sendit;
371 	}
372 #ifndef notdef
373 	/*
374 	 * If source address not specified yet, use address
375 	 * of outgoing interface.
376 	 */
377 	if (in_nullhost(ip->ip_src))
378 		ip->ip_src = ia->ia_addr.sin_addr;
379 #endif
380 
381 	/*
382 	 * packets with Class-D address as source are not valid per
383 	 * RFC 1112
384 	 */
385 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
386 		ipstat.ips_odropped++;
387 		error = EADDRNOTAVAIL;
388 		goto bad;
389 	}
390 
391 	/*
392 	 * Look for broadcast address and
393 	 * and verify user is allowed to send
394 	 * such a packet.
395 	 */
396 	if (in_broadcast(dst->sin_addr, ifp)) {
397 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
398 			error = EADDRNOTAVAIL;
399 			goto bad;
400 		}
401 		if ((flags & IP_ALLOWBROADCAST) == 0) {
402 			error = EACCES;
403 			goto bad;
404 		}
405 		/* don't allow broadcast messages to be fragmented */
406 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
407 			error = EMSGSIZE;
408 			goto bad;
409 		}
410 		m->m_flags |= M_BCAST;
411 	} else
412 		m->m_flags &= ~M_BCAST;
413 
414 sendit:
415 	/*
416 	 * If we're doing Path MTU Discovery, we need to set DF unless
417 	 * the route's MTU is locked.
418 	 */
419 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
420 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
421 		ip->ip_off |= IP_DF;
422 
423 	/*
424 	 * Remember the current ip_len and ip_off, and swap them into
425 	 * network order.
426 	 */
427 	ip_len = ip->ip_len;
428 
429 	HTONS(ip->ip_len);
430 	HTONS(ip->ip_off);
431 
432 #ifdef IPSEC
433 	/* get SP for this packet */
434 	if (so == NULL)
435 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
436 	else
437 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
438 
439 	if (sp == NULL) {
440 		ipsecstat.out_inval++;
441 		goto bad;
442 	}
443 
444 	error = 0;
445 
446 	/* check policy */
447 	switch (sp->policy) {
448 	case IPSEC_POLICY_DISCARD:
449 		/*
450 		 * This packet is just discarded.
451 		 */
452 		ipsecstat.out_polvio++;
453 		goto bad;
454 
455 	case IPSEC_POLICY_BYPASS:
456 	case IPSEC_POLICY_NONE:
457 		/* no need to do IPsec. */
458 		goto skip_ipsec;
459 
460 	case IPSEC_POLICY_IPSEC:
461 		if (sp->req == NULL) {
462 			/* XXX should be panic ? */
463 			printf("ip_output: No IPsec request specified.\n");
464 			error = EINVAL;
465 			goto bad;
466 		}
467 		break;
468 
469 	case IPSEC_POLICY_ENTRUST:
470 	default:
471 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
472 	}
473 
474 	/*
475 	 * ipsec4_output() expects ip_len and ip_off in network
476 	 * order.  They have been set to network order above.
477 	 */
478 
479     {
480 	struct ipsec_output_state state;
481 	bzero(&state, sizeof(state));
482 	state.m = m;
483 	if (flags & IP_ROUTETOIF) {
484 		state.ro = &iproute;
485 		bzero(&iproute, sizeof(iproute));
486 	} else
487 		state.ro = ro;
488 	state.dst = (struct sockaddr *)dst;
489 
490 	/*
491 	 * We can't defer the checksum of payload data if
492 	 * we're about to encrypt/authenticate it.
493 	 *
494 	 * XXX When we support crypto offloading functions of
495 	 * XXX network interfaces, we need to reconsider this,
496 	 * XXX since it's likely that they'll support checksumming,
497 	 * XXX as well.
498 	 */
499 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
500 		in_delayed_cksum(m);
501 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
502 	}
503 
504 	error = ipsec4_output(&state, sp, flags);
505 
506 	m = state.m;
507 	if (flags & IP_ROUTETOIF) {
508 		/*
509 		 * if we have tunnel mode SA, we may need to ignore
510 		 * IP_ROUTETOIF.
511 		 */
512 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
513 			flags &= ~IP_ROUTETOIF;
514 			ro = state.ro;
515 		}
516 	} else
517 		ro = state.ro;
518 	dst = (struct sockaddr_in *)state.dst;
519 	if (error) {
520 		/* mbuf is already reclaimed in ipsec4_output. */
521 		m0 = NULL;
522 		switch (error) {
523 		case EHOSTUNREACH:
524 		case ENETUNREACH:
525 		case EMSGSIZE:
526 		case ENOBUFS:
527 		case ENOMEM:
528 			break;
529 		default:
530 			printf("ip4_output (ipsec): error code %d\n", error);
531 			/*fall through*/
532 		case ENOENT:
533 			/* don't show these error codes to the user */
534 			error = 0;
535 			break;
536 		}
537 		goto bad;
538 	}
539 
540 	/* be sure to update variables that are affected by ipsec4_output() */
541 	ip = mtod(m, struct ip *);
542 #ifdef _IP_VHL
543 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
544 #else
545 	hlen = ip->ip_hl << 2;
546 #endif
547 	ip_len = ntohs(ip->ip_len);
548 
549 	if (ro->ro_rt == NULL) {
550 		if ((flags & IP_ROUTETOIF) == 0) {
551 			printf("ip_output: "
552 				"can't update route after IPsec processing\n");
553 			error = EHOSTUNREACH;	/*XXX*/
554 			goto bad;
555 		}
556 	} else {
557 		/* nobody uses ia beyond here */
558 		if (state.encap)
559 			ifp = ro->ro_rt->rt_ifp;
560 	}
561     }
562 
563 skip_ipsec:
564 #endif /*IPSEC*/
565 
566 #ifdef PFIL_HOOKS
567 	/*
568 	 * Run through list of hooks for output packets.
569 	 */
570 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp,
571 				    PFIL_OUT)) != 0)
572 		goto done;
573 	if (m == NULL)
574 		goto done;
575 
576 	ip = mtod(m, struct ip *);
577 #endif /* PFIL_HOOKS */
578 
579 	/*
580 	 * If small enough for mtu of path, can just send directly.
581 	 */
582 	if (ip_len <= mtu) {
583 #if IFA_STATS
584 		/*
585 		 * search for the source address structure to
586 		 * maintain output statistics.
587 		 */
588 		INADDR_TO_IA(ip->ip_src, ia);
589 		if (ia)
590 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
591 #endif
592 		/*
593 		 * Always initialize the sum to 0!  Some HW assisted
594 		 * checksumming requires this.
595 		 */
596 		ip->ip_sum = 0;
597 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
598 
599 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
600 
601 		/*
602 		 * Perform any checksums that the hardware can't do
603 		 * for us.
604 		 *
605 		 * XXX Does any hardware require the {th,uh}_sum
606 		 * XXX fields to be 0?
607 		 */
608 		if (sw_csum & M_CSUM_IPv4)
609 			ip->ip_sum = in_cksum(m, hlen);
610 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
611 			in_delayed_cksum(m);
612 			sw_csum &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
613 		}
614 		m->m_pkthdr.csum_flags &= ifp->if_csum_flags_tx;
615 
616 #ifdef IPSEC
617 		/* clean ipsec history once it goes out of the node */
618 		ipsec_delaux(m);
619 #endif
620 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
621 		goto done;
622 	}
623 
624 	/*
625 	 * We can't use HW checksumming if we're about to
626 	 * to fragment the packet.
627 	 *
628 	 * XXX Some hardware can do this.
629 	 */
630 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
631 		in_delayed_cksum(m);
632 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
633 	}
634 
635 	/*
636 	 * Too large for interface; fragment if possible.
637 	 * Must be able to put at least 8 bytes per fragment.
638 	 *
639 	 * Note we swap ip_len and ip_off into host order to make
640 	 * the logic below a little simpler.
641 	 */
642 
643 	NTOHS(ip->ip_len);
644 	NTOHS(ip->ip_off);
645 
646 	if (ip->ip_off & IP_DF) {
647 		if (flags & IP_RETURNMTU)
648 			*mtu_p = mtu;
649 		error = EMSGSIZE;
650 		ipstat.ips_cantfrag++;
651 		goto bad;
652 	}
653 	len = (mtu - hlen) &~ 7;
654 	if (len < 8) {
655 		error = EMSGSIZE;
656 		goto bad;
657 	}
658 
659     {
660 	int mhlen, firstlen = len;
661 	struct mbuf **mnext = &m->m_nextpkt;
662 	int fragments = 0;
663 	int s;
664 
665 	/*
666 	 * Loop through length of segment after first fragment,
667 	 * make new header and copy data of each part and link onto chain.
668 	 */
669 	m0 = m;
670 	mhlen = sizeof (struct ip);
671 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
672 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
673 		if (m == 0) {
674 			error = ENOBUFS;
675 			ipstat.ips_odropped++;
676 			goto sendorfree;
677 		}
678 		*mnext = m;
679 		mnext = &m->m_nextpkt;
680 		m->m_data += max_linkhdr;
681 		mhip = mtod(m, struct ip *);
682 		*mhip = *ip;
683 		/* we must inherit MCAST and BCAST flags */
684 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
685 		if (hlen > sizeof (struct ip)) {
686 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
687 			mhip->ip_hl = mhlen >> 2;
688 		}
689 		m->m_len = mhlen;
690 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
691 		if (ip->ip_off & IP_MF)
692 			mhip->ip_off |= IP_MF;
693 		if (off + len >= (u_int16_t)ip->ip_len)
694 			len = (u_int16_t)ip->ip_len - off;
695 		else
696 			mhip->ip_off |= IP_MF;
697 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
698 		m->m_next = m_copy(m0, off, len);
699 		if (m->m_next == 0) {
700 			error = ENOBUFS;	/* ??? */
701 			ipstat.ips_odropped++;
702 			goto sendorfree;
703 		}
704 		m->m_pkthdr.len = mhlen + len;
705 		m->m_pkthdr.rcvif = (struct ifnet *)0;
706 		HTONS(mhip->ip_off);
707 		mhip->ip_sum = 0;
708 		mhip->ip_sum = in_cksum(m, mhlen);
709 		ipstat.ips_ofragments++;
710 		fragments++;
711 	}
712 	/*
713 	 * Update first fragment by trimming what's been copied out
714 	 * and updating header, then send each fragment (in order).
715 	 */
716 	m = m0;
717 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
718 	m->m_pkthdr.len = hlen + firstlen;
719 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
720 	ip->ip_off |= IP_MF;
721 	HTONS(ip->ip_off);
722 	ip->ip_sum = 0;
723 	ip->ip_sum = in_cksum(m, hlen);
724 sendorfree:
725 	/*
726 	 * If there is no room for all the fragments, don't queue
727 	 * any of them.
728 	 */
729 	s = splnet();
730 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
731 		error = ENOBUFS;
732 	splx(s);
733 	for (m = m0; m; m = m0) {
734 		m0 = m->m_nextpkt;
735 		m->m_nextpkt = 0;
736 		if (error == 0) {
737 #if IFA_STATS
738 			/*
739 			 * search for the source address structure to
740 			 * maintain output statistics.
741 			 */
742 			INADDR_TO_IA(ip->ip_src, ia);
743 			if (ia) {
744 				ia->ia_ifa.ifa_data.ifad_outbytes +=
745 					ntohs(ip->ip_len);
746 			}
747 #endif
748 #ifdef IPSEC
749 			/* clean ipsec history once it goes out of the node */
750 			ipsec_delaux(m);
751 #endif
752 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
753 			    ro->ro_rt);
754 		} else
755 			m_freem(m);
756 	}
757 
758 	if (error == 0)
759 		ipstat.ips_fragmented++;
760     }
761 done:
762 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
763 		RTFREE(ro->ro_rt);
764 		ro->ro_rt = 0;
765 	}
766 
767 #ifdef IPSEC
768 	if (sp != NULL) {
769 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
770 			printf("DP ip_output call free SP:%p\n", sp));
771 		key_freesp(sp);
772 	}
773 #endif /* IPSEC */
774 
775 	return (error);
776 bad:
777 	m_freem(m);
778 	goto done;
779 }
780 
781 /*
782  * Process a delayed payload checksum calculation.
783  */
784 void
785 in_delayed_cksum(struct mbuf *m)
786 {
787 	struct ip *ip;
788 	u_int16_t csum, offset;
789 
790 	ip = mtod(m, struct ip *);
791 	offset = ip->ip_hl << 2;
792 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
793 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
794 		csum = 0xffff;
795 
796 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
797 
798 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
799 		/* This happen when ip options were inserted
800 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
801 		    m->m_len, offset, ip->ip_p);
802 		 */
803 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
804 	} else
805 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
806 }
807 
808 /*
809  * Determine the maximum length of the options to be inserted;
810  * we would far rather allocate too much space rather than too little.
811  */
812 
813 u_int
814 ip_optlen(inp)
815 	struct inpcb *inp;
816 {
817 	struct mbuf *m = inp->inp_options;
818 
819 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
820 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
821 	else
822 		return 0;
823 }
824 
825 
826 /*
827  * Insert IP options into preformed packet.
828  * Adjust IP destination as required for IP source routing,
829  * as indicated by a non-zero in_addr at the start of the options.
830  */
831 static struct mbuf *
832 ip_insertoptions(m, opt, phlen)
833 	struct mbuf *m;
834 	struct mbuf *opt;
835 	int *phlen;
836 {
837 	struct ipoption *p = mtod(opt, struct ipoption *);
838 	struct mbuf *n;
839 	struct ip *ip = mtod(m, struct ip *);
840 	unsigned optlen;
841 
842 	optlen = opt->m_len - sizeof(p->ipopt_dst);
843 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
844 		return (m);		/* XXX should fail */
845 	if (!in_nullhost(p->ipopt_dst))
846 		ip->ip_dst = p->ipopt_dst;
847 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
848 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
849 		if (n == 0)
850 			return (m);
851 		M_COPY_PKTHDR(n, m);
852 		m->m_flags &= ~M_PKTHDR;
853 		m->m_len -= sizeof(struct ip);
854 		m->m_data += sizeof(struct ip);
855 		n->m_next = m;
856 		m = n;
857 		m->m_len = optlen + sizeof(struct ip);
858 		m->m_data += max_linkhdr;
859 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
860 	} else {
861 		m->m_data -= optlen;
862 		m->m_len += optlen;
863 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
864 	}
865 	m->m_pkthdr.len += optlen;
866 	ip = mtod(m, struct ip *);
867 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
868 	*phlen = sizeof(struct ip) + optlen;
869 	ip->ip_len += optlen;
870 	return (m);
871 }
872 
873 /*
874  * Copy options from ip to jp,
875  * omitting those not copied during fragmentation.
876  */
877 int
878 ip_optcopy(ip, jp)
879 	struct ip *ip, *jp;
880 {
881 	u_char *cp, *dp;
882 	int opt, optlen, cnt;
883 
884 	cp = (u_char *)(ip + 1);
885 	dp = (u_char *)(jp + 1);
886 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
887 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
888 		opt = cp[0];
889 		if (opt == IPOPT_EOL)
890 			break;
891 		if (opt == IPOPT_NOP) {
892 			/* Preserve for IP mcast tunnel's LSRR alignment. */
893 			*dp++ = IPOPT_NOP;
894 			optlen = 1;
895 			continue;
896 		}
897 #ifdef DIAGNOSTIC
898 		if (cnt < IPOPT_OLEN + sizeof(*cp))
899 			panic("malformed IPv4 option passed to ip_optcopy");
900 #endif
901 		optlen = cp[IPOPT_OLEN];
902 #ifdef DIAGNOSTIC
903 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
904 			panic("malformed IPv4 option passed to ip_optcopy");
905 #endif
906 		/* bogus lengths should have been caught by ip_dooptions */
907 		if (optlen > cnt)
908 			optlen = cnt;
909 		if (IPOPT_COPIED(opt)) {
910 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
911 			dp += optlen;
912 		}
913 	}
914 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
915 		*dp++ = IPOPT_EOL;
916 	return (optlen);
917 }
918 
919 /*
920  * IP socket option processing.
921  */
922 int
923 ip_ctloutput(op, so, level, optname, mp)
924 	int op;
925 	struct socket *so;
926 	int level, optname;
927 	struct mbuf **mp;
928 {
929 	struct inpcb *inp = sotoinpcb(so);
930 	struct mbuf *m = *mp;
931 	int optval = 0;
932 	int error = 0;
933 #ifdef IPSEC
934 #ifdef __NetBSD__
935 	struct proc *p = curproc;	/*XXX*/
936 #endif
937 #endif
938 
939 	if (level != IPPROTO_IP) {
940 		error = EINVAL;
941 		if (op == PRCO_SETOPT && *mp)
942 			(void) m_free(*mp);
943 	} else switch (op) {
944 
945 	case PRCO_SETOPT:
946 		switch (optname) {
947 		case IP_OPTIONS:
948 #ifdef notyet
949 		case IP_RETOPTS:
950 			return (ip_pcbopts(optname, &inp->inp_options, m));
951 #else
952 			return (ip_pcbopts(&inp->inp_options, m));
953 #endif
954 
955 		case IP_TOS:
956 		case IP_TTL:
957 		case IP_RECVOPTS:
958 		case IP_RECVRETOPTS:
959 		case IP_RECVDSTADDR:
960 		case IP_RECVIF:
961 			if (m == NULL || m->m_len != sizeof(int))
962 				error = EINVAL;
963 			else {
964 				optval = *mtod(m, int *);
965 				switch (optname) {
966 
967 				case IP_TOS:
968 					inp->inp_ip.ip_tos = optval;
969 					break;
970 
971 				case IP_TTL:
972 					inp->inp_ip.ip_ttl = optval;
973 					break;
974 #define	OPTSET(bit) \
975 	if (optval) \
976 		inp->inp_flags |= bit; \
977 	else \
978 		inp->inp_flags &= ~bit;
979 
980 				case IP_RECVOPTS:
981 					OPTSET(INP_RECVOPTS);
982 					break;
983 
984 				case IP_RECVRETOPTS:
985 					OPTSET(INP_RECVRETOPTS);
986 					break;
987 
988 				case IP_RECVDSTADDR:
989 					OPTSET(INP_RECVDSTADDR);
990 					break;
991 
992 				case IP_RECVIF:
993 					OPTSET(INP_RECVIF);
994 					break;
995 				}
996 			}
997 			break;
998 #undef OPTSET
999 
1000 		case IP_MULTICAST_IF:
1001 		case IP_MULTICAST_TTL:
1002 		case IP_MULTICAST_LOOP:
1003 		case IP_ADD_MEMBERSHIP:
1004 		case IP_DROP_MEMBERSHIP:
1005 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1006 			break;
1007 
1008 		case IP_PORTRANGE:
1009 			if (m == 0 || m->m_len != sizeof(int))
1010 				error = EINVAL;
1011 			else {
1012 				optval = *mtod(m, int *);
1013 
1014 				switch (optval) {
1015 
1016 				case IP_PORTRANGE_DEFAULT:
1017 				case IP_PORTRANGE_HIGH:
1018 					inp->inp_flags &= ~(INP_LOWPORT);
1019 					break;
1020 
1021 				case IP_PORTRANGE_LOW:
1022 					inp->inp_flags |= INP_LOWPORT;
1023 					break;
1024 
1025 				default:
1026 					error = EINVAL;
1027 					break;
1028 				}
1029 			}
1030 			break;
1031 
1032 #ifdef IPSEC
1033 		case IP_IPSEC_POLICY:
1034 		{
1035 			caddr_t req = NULL;
1036 			size_t len = 0;
1037 			int priv = 0;
1038 
1039 #ifdef __NetBSD__
1040 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1041 				priv = 0;
1042 			else
1043 				priv = 1;
1044 #else
1045 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1046 #endif
1047 			if (m) {
1048 				req = mtod(m, caddr_t);
1049 				len = m->m_len;
1050 			}
1051 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1052 			break;
1053 		    }
1054 #endif /*IPSEC*/
1055 
1056 		default:
1057 			error = ENOPROTOOPT;
1058 			break;
1059 		}
1060 		if (m)
1061 			(void)m_free(m);
1062 		break;
1063 
1064 	case PRCO_GETOPT:
1065 		switch (optname) {
1066 		case IP_OPTIONS:
1067 		case IP_RETOPTS:
1068 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1069 			if (inp->inp_options) {
1070 				m->m_len = inp->inp_options->m_len;
1071 				bcopy(mtod(inp->inp_options, caddr_t),
1072 				    mtod(m, caddr_t), (unsigned)m->m_len);
1073 			} else
1074 				m->m_len = 0;
1075 			break;
1076 
1077 		case IP_TOS:
1078 		case IP_TTL:
1079 		case IP_RECVOPTS:
1080 		case IP_RECVRETOPTS:
1081 		case IP_RECVDSTADDR:
1082 		case IP_RECVIF:
1083 		case IP_ERRORMTU:
1084 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1085 			m->m_len = sizeof(int);
1086 			switch (optname) {
1087 
1088 			case IP_TOS:
1089 				optval = inp->inp_ip.ip_tos;
1090 				break;
1091 
1092 			case IP_TTL:
1093 				optval = inp->inp_ip.ip_ttl;
1094 				break;
1095 
1096 			case IP_ERRORMTU:
1097 				optval = inp->inp_errormtu;
1098 				break;
1099 
1100 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1101 
1102 			case IP_RECVOPTS:
1103 				optval = OPTBIT(INP_RECVOPTS);
1104 				break;
1105 
1106 			case IP_RECVRETOPTS:
1107 				optval = OPTBIT(INP_RECVRETOPTS);
1108 				break;
1109 
1110 			case IP_RECVDSTADDR:
1111 				optval = OPTBIT(INP_RECVDSTADDR);
1112 				break;
1113 
1114 			case IP_RECVIF:
1115 				optval = OPTBIT(INP_RECVIF);
1116 				break;
1117 			}
1118 			*mtod(m, int *) = optval;
1119 			break;
1120 
1121 #ifdef IPSEC
1122 		case IP_IPSEC_POLICY:
1123 		{
1124 			caddr_t req = NULL;
1125 			size_t len = 0;
1126 
1127 			if (m) {
1128 				req = mtod(m, caddr_t);
1129 				len = m->m_len;
1130 			}
1131 			error = ipsec4_get_policy(inp, req, len, mp);
1132 			break;
1133 		}
1134 #endif /*IPSEC*/
1135 
1136 		case IP_MULTICAST_IF:
1137 		case IP_MULTICAST_TTL:
1138 		case IP_MULTICAST_LOOP:
1139 		case IP_ADD_MEMBERSHIP:
1140 		case IP_DROP_MEMBERSHIP:
1141 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1142 			break;
1143 
1144 		case IP_PORTRANGE:
1145 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1146 			m->m_len = sizeof(int);
1147 
1148 			if (inp->inp_flags & INP_LOWPORT)
1149 				optval = IP_PORTRANGE_LOW;
1150 			else
1151 				optval = IP_PORTRANGE_DEFAULT;
1152 
1153 			*mtod(m, int *) = optval;
1154 			break;
1155 
1156 		default:
1157 			error = ENOPROTOOPT;
1158 			break;
1159 		}
1160 		break;
1161 	}
1162 	return (error);
1163 }
1164 
1165 /*
1166  * Set up IP options in pcb for insertion in output packets.
1167  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1168  * with destination address if source routed.
1169  */
1170 int
1171 #ifdef notyet
1172 ip_pcbopts(optname, pcbopt, m)
1173 	int optname;
1174 #else
1175 ip_pcbopts(pcbopt, m)
1176 #endif
1177 	struct mbuf **pcbopt;
1178 	struct mbuf *m;
1179 {
1180 	int cnt, optlen;
1181 	u_char *cp;
1182 	u_char opt;
1183 
1184 	/* turn off any old options */
1185 	if (*pcbopt)
1186 		(void)m_free(*pcbopt);
1187 	*pcbopt = 0;
1188 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1189 		/*
1190 		 * Only turning off any previous options.
1191 		 */
1192 		if (m)
1193 			(void)m_free(m);
1194 		return (0);
1195 	}
1196 
1197 #ifndef	__vax__
1198 	if (m->m_len % sizeof(int32_t))
1199 		goto bad;
1200 #endif
1201 	/*
1202 	 * IP first-hop destination address will be stored before
1203 	 * actual options; move other options back
1204 	 * and clear it when none present.
1205 	 */
1206 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1207 		goto bad;
1208 	cnt = m->m_len;
1209 	m->m_len += sizeof(struct in_addr);
1210 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1211 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1212 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1213 
1214 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1215 		opt = cp[IPOPT_OPTVAL];
1216 		if (opt == IPOPT_EOL)
1217 			break;
1218 		if (opt == IPOPT_NOP)
1219 			optlen = 1;
1220 		else {
1221 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1222 				goto bad;
1223 			optlen = cp[IPOPT_OLEN];
1224 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1225 				goto bad;
1226 		}
1227 		switch (opt) {
1228 
1229 		default:
1230 			break;
1231 
1232 		case IPOPT_LSRR:
1233 		case IPOPT_SSRR:
1234 			/*
1235 			 * user process specifies route as:
1236 			 *	->A->B->C->D
1237 			 * D must be our final destination (but we can't
1238 			 * check that since we may not have connected yet).
1239 			 * A is first hop destination, which doesn't appear in
1240 			 * actual IP option, but is stored before the options.
1241 			 */
1242 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1243 				goto bad;
1244 			m->m_len -= sizeof(struct in_addr);
1245 			cnt -= sizeof(struct in_addr);
1246 			optlen -= sizeof(struct in_addr);
1247 			cp[IPOPT_OLEN] = optlen;
1248 			/*
1249 			 * Move first hop before start of options.
1250 			 */
1251 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1252 			    sizeof(struct in_addr));
1253 			/*
1254 			 * Then copy rest of options back
1255 			 * to close up the deleted entry.
1256 			 */
1257 			memmove(&cp[IPOPT_OFFSET+1],
1258                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1259 			    (unsigned)cnt + sizeof(struct in_addr));
1260 			break;
1261 		}
1262 	}
1263 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1264 		goto bad;
1265 	*pcbopt = m;
1266 	return (0);
1267 
1268 bad:
1269 	(void)m_free(m);
1270 	return (EINVAL);
1271 }
1272 
1273 /*
1274  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1275  */
1276 static struct ifnet *
1277 ip_multicast_if(a, ifindexp)
1278 	struct in_addr *a;
1279 	int *ifindexp;
1280 {
1281 	int ifindex;
1282 	struct ifnet *ifp;
1283 
1284 	if (ifindexp)
1285 		*ifindexp = 0;
1286 	if (ntohl(a->s_addr) >> 24 == 0) {
1287 		ifindex = ntohl(a->s_addr) & 0xffffff;
1288 		if (ifindex < 0 || if_index < ifindex)
1289 			return NULL;
1290 		ifp = ifindex2ifnet[ifindex];
1291 		if (ifindexp)
1292 			*ifindexp = ifindex;
1293 	} else {
1294 		INADDR_TO_IFP(*a, ifp);
1295 	}
1296 	return ifp;
1297 }
1298 
1299 /*
1300  * Set the IP multicast options in response to user setsockopt().
1301  */
1302 int
1303 ip_setmoptions(optname, imop, m)
1304 	int optname;
1305 	struct ip_moptions **imop;
1306 	struct mbuf *m;
1307 {
1308 	int error = 0;
1309 	u_char loop;
1310 	int i;
1311 	struct in_addr addr;
1312 	struct ip_mreq *mreq;
1313 	struct ifnet *ifp;
1314 	struct ip_moptions *imo = *imop;
1315 	struct route ro;
1316 	struct sockaddr_in *dst;
1317 	int ifindex;
1318 
1319 	if (imo == NULL) {
1320 		/*
1321 		 * No multicast option buffer attached to the pcb;
1322 		 * allocate one and initialize to default values.
1323 		 */
1324 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1325 		    M_WAITOK);
1326 
1327 		if (imo == NULL)
1328 			return (ENOBUFS);
1329 		*imop = imo;
1330 		imo->imo_multicast_ifp = NULL;
1331 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1332 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1333 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1334 		imo->imo_num_memberships = 0;
1335 	}
1336 
1337 	switch (optname) {
1338 
1339 	case IP_MULTICAST_IF:
1340 		/*
1341 		 * Select the interface for outgoing multicast packets.
1342 		 */
1343 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1344 			error = EINVAL;
1345 			break;
1346 		}
1347 		addr = *(mtod(m, struct in_addr *));
1348 		/*
1349 		 * INADDR_ANY is used to remove a previous selection.
1350 		 * When no interface is selected, a default one is
1351 		 * chosen every time a multicast packet is sent.
1352 		 */
1353 		if (in_nullhost(addr)) {
1354 			imo->imo_multicast_ifp = NULL;
1355 			break;
1356 		}
1357 		/*
1358 		 * The selected interface is identified by its local
1359 		 * IP address.  Find the interface and confirm that
1360 		 * it supports multicasting.
1361 		 */
1362 		ifp = ip_multicast_if(&addr, &ifindex);
1363 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1364 			error = EADDRNOTAVAIL;
1365 			break;
1366 		}
1367 		imo->imo_multicast_ifp = ifp;
1368 		if (ifindex)
1369 			imo->imo_multicast_addr = addr;
1370 		else
1371 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1372 		break;
1373 
1374 	case IP_MULTICAST_TTL:
1375 		/*
1376 		 * Set the IP time-to-live for outgoing multicast packets.
1377 		 */
1378 		if (m == NULL || m->m_len != 1) {
1379 			error = EINVAL;
1380 			break;
1381 		}
1382 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1383 		break;
1384 
1385 	case IP_MULTICAST_LOOP:
1386 		/*
1387 		 * Set the loopback flag for outgoing multicast packets.
1388 		 * Must be zero or one.
1389 		 */
1390 		if (m == NULL || m->m_len != 1 ||
1391 		   (loop = *(mtod(m, u_char *))) > 1) {
1392 			error = EINVAL;
1393 			break;
1394 		}
1395 		imo->imo_multicast_loop = loop;
1396 		break;
1397 
1398 	case IP_ADD_MEMBERSHIP:
1399 		/*
1400 		 * Add a multicast group membership.
1401 		 * Group must be a valid IP multicast address.
1402 		 */
1403 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1404 			error = EINVAL;
1405 			break;
1406 		}
1407 		mreq = mtod(m, struct ip_mreq *);
1408 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1409 			error = EINVAL;
1410 			break;
1411 		}
1412 		/*
1413 		 * If no interface address was provided, use the interface of
1414 		 * the route to the given multicast address.
1415 		 */
1416 		if (in_nullhost(mreq->imr_interface)) {
1417 			bzero((caddr_t)&ro, sizeof(ro));
1418 			ro.ro_rt = NULL;
1419 			dst = satosin(&ro.ro_dst);
1420 			dst->sin_len = sizeof(*dst);
1421 			dst->sin_family = AF_INET;
1422 			dst->sin_addr = mreq->imr_multiaddr;
1423 			rtalloc(&ro);
1424 			if (ro.ro_rt == NULL) {
1425 				error = EADDRNOTAVAIL;
1426 				break;
1427 			}
1428 			ifp = ro.ro_rt->rt_ifp;
1429 			rtfree(ro.ro_rt);
1430 		} else {
1431 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1432 		}
1433 		/*
1434 		 * See if we found an interface, and confirm that it
1435 		 * supports multicast.
1436 		 */
1437 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1438 			error = EADDRNOTAVAIL;
1439 			break;
1440 		}
1441 		/*
1442 		 * See if the membership already exists or if all the
1443 		 * membership slots are full.
1444 		 */
1445 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1446 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1447 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1448 				      mreq->imr_multiaddr))
1449 				break;
1450 		}
1451 		if (i < imo->imo_num_memberships) {
1452 			error = EADDRINUSE;
1453 			break;
1454 		}
1455 		if (i == IP_MAX_MEMBERSHIPS) {
1456 			error = ETOOMANYREFS;
1457 			break;
1458 		}
1459 		/*
1460 		 * Everything looks good; add a new record to the multicast
1461 		 * address list for the given interface.
1462 		 */
1463 		if ((imo->imo_membership[i] =
1464 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1465 			error = ENOBUFS;
1466 			break;
1467 		}
1468 		++imo->imo_num_memberships;
1469 		break;
1470 
1471 	case IP_DROP_MEMBERSHIP:
1472 		/*
1473 		 * Drop a multicast group membership.
1474 		 * Group must be a valid IP multicast address.
1475 		 */
1476 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1477 			error = EINVAL;
1478 			break;
1479 		}
1480 		mreq = mtod(m, struct ip_mreq *);
1481 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1482 			error = EINVAL;
1483 			break;
1484 		}
1485 		/*
1486 		 * If an interface address was specified, get a pointer
1487 		 * to its ifnet structure.
1488 		 */
1489 		if (in_nullhost(mreq->imr_interface))
1490 			ifp = NULL;
1491 		else {
1492 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1493 			if (ifp == NULL) {
1494 				error = EADDRNOTAVAIL;
1495 				break;
1496 			}
1497 		}
1498 		/*
1499 		 * Find the membership in the membership array.
1500 		 */
1501 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1502 			if ((ifp == NULL ||
1503 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1504 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1505 				       mreq->imr_multiaddr))
1506 				break;
1507 		}
1508 		if (i == imo->imo_num_memberships) {
1509 			error = EADDRNOTAVAIL;
1510 			break;
1511 		}
1512 		/*
1513 		 * Give up the multicast address record to which the
1514 		 * membership points.
1515 		 */
1516 		in_delmulti(imo->imo_membership[i]);
1517 		/*
1518 		 * Remove the gap in the membership array.
1519 		 */
1520 		for (++i; i < imo->imo_num_memberships; ++i)
1521 			imo->imo_membership[i-1] = imo->imo_membership[i];
1522 		--imo->imo_num_memberships;
1523 		break;
1524 
1525 	default:
1526 		error = EOPNOTSUPP;
1527 		break;
1528 	}
1529 
1530 	/*
1531 	 * If all options have default values, no need to keep the mbuf.
1532 	 */
1533 	if (imo->imo_multicast_ifp == NULL &&
1534 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1535 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1536 	    imo->imo_num_memberships == 0) {
1537 		free(*imop, M_IPMOPTS);
1538 		*imop = NULL;
1539 	}
1540 
1541 	return (error);
1542 }
1543 
1544 /*
1545  * Return the IP multicast options in response to user getsockopt().
1546  */
1547 int
1548 ip_getmoptions(optname, imo, mp)
1549 	int optname;
1550 	struct ip_moptions *imo;
1551 	struct mbuf **mp;
1552 {
1553 	u_char *ttl;
1554 	u_char *loop;
1555 	struct in_addr *addr;
1556 	struct in_ifaddr *ia;
1557 
1558 	*mp = m_get(M_WAIT, MT_SOOPTS);
1559 
1560 	switch (optname) {
1561 
1562 	case IP_MULTICAST_IF:
1563 		addr = mtod(*mp, struct in_addr *);
1564 		(*mp)->m_len = sizeof(struct in_addr);
1565 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1566 			*addr = zeroin_addr;
1567 		else if (imo->imo_multicast_addr.s_addr) {
1568 			/* return the value user has set */
1569 			*addr = imo->imo_multicast_addr;
1570 		} else {
1571 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1572 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1573 		}
1574 		return (0);
1575 
1576 	case IP_MULTICAST_TTL:
1577 		ttl = mtod(*mp, u_char *);
1578 		(*mp)->m_len = 1;
1579 		*ttl = imo ? imo->imo_multicast_ttl
1580 			   : IP_DEFAULT_MULTICAST_TTL;
1581 		return (0);
1582 
1583 	case IP_MULTICAST_LOOP:
1584 		loop = mtod(*mp, u_char *);
1585 		(*mp)->m_len = 1;
1586 		*loop = imo ? imo->imo_multicast_loop
1587 			    : IP_DEFAULT_MULTICAST_LOOP;
1588 		return (0);
1589 
1590 	default:
1591 		return (EOPNOTSUPP);
1592 	}
1593 }
1594 
1595 /*
1596  * Discard the IP multicast options.
1597  */
1598 void
1599 ip_freemoptions(imo)
1600 	struct ip_moptions *imo;
1601 {
1602 	int i;
1603 
1604 	if (imo != NULL) {
1605 		for (i = 0; i < imo->imo_num_memberships; ++i)
1606 			in_delmulti(imo->imo_membership[i]);
1607 		free(imo, M_IPMOPTS);
1608 	}
1609 }
1610 
1611 /*
1612  * Routine called from ip_output() to loop back a copy of an IP multicast
1613  * packet to the input queue of a specified interface.  Note that this
1614  * calls the output routine of the loopback "driver", but with an interface
1615  * pointer that might NOT be &loif -- easier than replicating that code here.
1616  */
1617 static void
1618 ip_mloopback(ifp, m, dst)
1619 	struct ifnet *ifp;
1620 	struct mbuf *m;
1621 	struct sockaddr_in *dst;
1622 {
1623 	struct ip *ip;
1624 	struct mbuf *copym;
1625 
1626 	copym = m_copy(m, 0, M_COPYALL);
1627 	if (copym != NULL
1628 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1629 		copym = m_pullup(copym, sizeof(struct ip));
1630 	if (copym != NULL) {
1631 		/*
1632 		 * We don't bother to fragment if the IP length is greater
1633 		 * than the interface's MTU.  Can this possibly matter?
1634 		 */
1635 		ip = mtod(copym, struct ip *);
1636 		HTONS(ip->ip_len);
1637 		HTONS(ip->ip_off);
1638 		ip->ip_sum = 0;
1639 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1640 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1641 	}
1642 }
1643