xref: /netbsd-src/sys/netinet/ip_output.c (revision 06be8101a16cc95f40783b3cb7afd12112103a9a)
1 /*	$NetBSD: ip_output.c,v 1.89 2001/11/13 00:32:39 lukem 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 1.89 2001/11/13 00:32:39 lukem 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 
541 	/* be sure to update variables that are affected by ipsec4_output() */
542 	ip = mtod(m, struct ip *);
543 #ifdef _IP_VHL
544 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
545 #else
546 	hlen = ip->ip_hl << 2;
547 #endif
548 	ip_len = ntohs(ip->ip_len);
549 
550 	if (ro->ro_rt == NULL) {
551 		if ((flags & IP_ROUTETOIF) == 0) {
552 			printf("ip_output: "
553 				"can't update route after IPsec processing\n");
554 			error = EHOSTUNREACH;	/*XXX*/
555 			goto bad;
556 		}
557 	} else {
558 		/* nobody uses ia beyond here */
559 		ifp = ro->ro_rt->rt_ifp;
560 	}
561 
562 skip_ipsec:
563 #endif /*IPSEC*/
564 
565 #ifdef PFIL_HOOKS
566 	/*
567 	 * Run through list of hooks for output packets.
568 	 */
569 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp,
570 				    PFIL_OUT)) != 0)
571 		goto done;
572 	if (m == NULL)
573 		goto done;
574 
575 	ip = mtod(m, struct ip *);
576 #endif /* PFIL_HOOKS */
577 
578 	/*
579 	 * If small enough for mtu of path, can just send directly.
580 	 */
581 	if (ip_len <= mtu) {
582 #if IFA_STATS
583 		/*
584 		 * search for the source address structure to
585 		 * maintain output statistics.
586 		 */
587 		INADDR_TO_IA(ip->ip_src, ia);
588 		if (ia)
589 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
590 #endif
591 		/*
592 		 * Always initialize the sum to 0!  Some HW assisted
593 		 * checksumming requires this.
594 		 */
595 		ip->ip_sum = 0;
596 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
597 
598 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
599 
600 		/*
601 		 * Perform any checksums that the hardware can't do
602 		 * for us.
603 		 *
604 		 * XXX Does any hardware require the {th,uh}_sum
605 		 * XXX fields to be 0?
606 		 */
607 		if (sw_csum & M_CSUM_IPv4)
608 			ip->ip_sum = in_cksum(m, hlen);
609 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
610 			in_delayed_cksum(m);
611 			sw_csum &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
612 		}
613 		m->m_pkthdr.csum_flags &= ifp->if_csum_flags_tx;
614 
615 #ifdef IPSEC
616 		/* clean ipsec history once it goes out of the node */
617 		ipsec_delaux(m);
618 #endif
619 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
620 		goto done;
621 	}
622 
623 	/*
624 	 * We can't use HW checksumming if we're about to
625 	 * to fragment the packet.
626 	 *
627 	 * XXX Some hardware can do this.
628 	 */
629 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
630 		in_delayed_cksum(m);
631 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
632 	}
633 
634 	/*
635 	 * Too large for interface; fragment if possible.
636 	 * Must be able to put at least 8 bytes per fragment.
637 	 *
638 	 * Note we swap ip_len and ip_off into host order to make
639 	 * the logic below a little simpler.
640 	 */
641 
642 	NTOHS(ip->ip_len);
643 	NTOHS(ip->ip_off);
644 
645 	if (ip->ip_off & IP_DF) {
646 		if (flags & IP_RETURNMTU)
647 			*mtu_p = mtu;
648 		error = EMSGSIZE;
649 		ipstat.ips_cantfrag++;
650 		goto bad;
651 	}
652 	len = (mtu - hlen) &~ 7;
653 	if (len < 8) {
654 		error = EMSGSIZE;
655 		goto bad;
656 	}
657 
658     {
659 	int mhlen, firstlen = len;
660 	struct mbuf **mnext = &m->m_nextpkt;
661 	int fragments = 0;
662 	int s;
663 
664 	/*
665 	 * Loop through length of segment after first fragment,
666 	 * make new header and copy data of each part and link onto chain.
667 	 */
668 	m0 = m;
669 	mhlen = sizeof (struct ip);
670 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
671 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
672 		if (m == 0) {
673 			error = ENOBUFS;
674 			ipstat.ips_odropped++;
675 			goto sendorfree;
676 		}
677 		*mnext = m;
678 		mnext = &m->m_nextpkt;
679 		m->m_data += max_linkhdr;
680 		mhip = mtod(m, struct ip *);
681 		*mhip = *ip;
682 		/* we must inherit MCAST and BCAST flags */
683 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
684 		if (hlen > sizeof (struct ip)) {
685 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
686 			mhip->ip_hl = mhlen >> 2;
687 		}
688 		m->m_len = mhlen;
689 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
690 		if (ip->ip_off & IP_MF)
691 			mhip->ip_off |= IP_MF;
692 		if (off + len >= (u_int16_t)ip->ip_len)
693 			len = (u_int16_t)ip->ip_len - off;
694 		else
695 			mhip->ip_off |= IP_MF;
696 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
697 		m->m_next = m_copy(m0, off, len);
698 		if (m->m_next == 0) {
699 			error = ENOBUFS;	/* ??? */
700 			ipstat.ips_odropped++;
701 			goto sendorfree;
702 		}
703 		m->m_pkthdr.len = mhlen + len;
704 		m->m_pkthdr.rcvif = (struct ifnet *)0;
705 		HTONS(mhip->ip_off);
706 		mhip->ip_sum = 0;
707 		mhip->ip_sum = in_cksum(m, mhlen);
708 		ipstat.ips_ofragments++;
709 		fragments++;
710 	}
711 	/*
712 	 * Update first fragment by trimming what's been copied out
713 	 * and updating header, then send each fragment (in order).
714 	 */
715 	m = m0;
716 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
717 	m->m_pkthdr.len = hlen + firstlen;
718 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
719 	ip->ip_off |= IP_MF;
720 	HTONS(ip->ip_off);
721 	ip->ip_sum = 0;
722 	ip->ip_sum = in_cksum(m, hlen);
723 sendorfree:
724 	/*
725 	 * If there is no room for all the fragments, don't queue
726 	 * any of them.
727 	 */
728 	s = splnet();
729 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
730 		error = ENOBUFS;
731 	splx(s);
732 	for (m = m0; m; m = m0) {
733 		m0 = m->m_nextpkt;
734 		m->m_nextpkt = 0;
735 		if (error == 0) {
736 #if IFA_STATS
737 			/*
738 			 * search for the source address structure to
739 			 * maintain output statistics.
740 			 */
741 			INADDR_TO_IA(ip->ip_src, ia);
742 			if (ia) {
743 				ia->ia_ifa.ifa_data.ifad_outbytes +=
744 					ntohs(ip->ip_len);
745 			}
746 #endif
747 #ifdef IPSEC
748 			/* clean ipsec history once it goes out of the node */
749 			ipsec_delaux(m);
750 #endif
751 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
752 			    ro->ro_rt);
753 		} else
754 			m_freem(m);
755 	}
756 
757 	if (error == 0)
758 		ipstat.ips_fragmented++;
759     }
760 done:
761 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
762 		RTFREE(ro->ro_rt);
763 		ro->ro_rt = 0;
764 	}
765 
766 #ifdef IPSEC
767 	if (sp != NULL) {
768 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
769 			printf("DP ip_output call free SP:%p\n", sp));
770 		key_freesp(sp);
771 	}
772 #endif /* IPSEC */
773 
774 	return (error);
775 bad:
776 	m_freem(m);
777 	goto done;
778 }
779 
780 /*
781  * Process a delayed payload checksum calculation.
782  */
783 void
784 in_delayed_cksum(struct mbuf *m)
785 {
786 	struct ip *ip;
787 	u_int16_t csum, offset;
788 
789 	ip = mtod(m, struct ip *);
790 	offset = ip->ip_hl << 2;
791 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
792 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
793 		csum = 0xffff;
794 
795 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
796 
797 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
798 		/* This happen when ip options were inserted
799 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
800 		    m->m_len, offset, ip->ip_p);
801 		 */
802 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
803 	} else
804 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
805 }
806 
807 /*
808  * Determine the maximum length of the options to be inserted;
809  * we would far rather allocate too much space rather than too little.
810  */
811 
812 u_int
813 ip_optlen(inp)
814 	struct inpcb *inp;
815 {
816 	struct mbuf *m = inp->inp_options;
817 
818 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
819 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
820 	else
821 		return 0;
822 }
823 
824 
825 /*
826  * Insert IP options into preformed packet.
827  * Adjust IP destination as required for IP source routing,
828  * as indicated by a non-zero in_addr at the start of the options.
829  */
830 static struct mbuf *
831 ip_insertoptions(m, opt, phlen)
832 	struct mbuf *m;
833 	struct mbuf *opt;
834 	int *phlen;
835 {
836 	struct ipoption *p = mtod(opt, struct ipoption *);
837 	struct mbuf *n;
838 	struct ip *ip = mtod(m, struct ip *);
839 	unsigned optlen;
840 
841 	optlen = opt->m_len - sizeof(p->ipopt_dst);
842 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
843 		return (m);		/* XXX should fail */
844 	if (!in_nullhost(p->ipopt_dst))
845 		ip->ip_dst = p->ipopt_dst;
846 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
847 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
848 		if (n == 0)
849 			return (m);
850 		M_COPY_PKTHDR(n, m);
851 		m->m_flags &= ~M_PKTHDR;
852 		m->m_len -= sizeof(struct ip);
853 		m->m_data += sizeof(struct ip);
854 		n->m_next = m;
855 		m = n;
856 		m->m_len = optlen + sizeof(struct ip);
857 		m->m_data += max_linkhdr;
858 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
859 	} else {
860 		m->m_data -= optlen;
861 		m->m_len += optlen;
862 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
863 	}
864 	m->m_pkthdr.len += optlen;
865 	ip = mtod(m, struct ip *);
866 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
867 	*phlen = sizeof(struct ip) + optlen;
868 	ip->ip_len += optlen;
869 	return (m);
870 }
871 
872 /*
873  * Copy options from ip to jp,
874  * omitting those not copied during fragmentation.
875  */
876 int
877 ip_optcopy(ip, jp)
878 	struct ip *ip, *jp;
879 {
880 	u_char *cp, *dp;
881 	int opt, optlen, cnt;
882 
883 	cp = (u_char *)(ip + 1);
884 	dp = (u_char *)(jp + 1);
885 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
886 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
887 		opt = cp[0];
888 		if (opt == IPOPT_EOL)
889 			break;
890 		if (opt == IPOPT_NOP) {
891 			/* Preserve for IP mcast tunnel's LSRR alignment. */
892 			*dp++ = IPOPT_NOP;
893 			optlen = 1;
894 			continue;
895 		}
896 #ifdef DIAGNOSTIC
897 		if (cnt < IPOPT_OLEN + sizeof(*cp))
898 			panic("malformed IPv4 option passed to ip_optcopy");
899 #endif
900 		optlen = cp[IPOPT_OLEN];
901 #ifdef DIAGNOSTIC
902 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
903 			panic("malformed IPv4 option passed to ip_optcopy");
904 #endif
905 		/* bogus lengths should have been caught by ip_dooptions */
906 		if (optlen > cnt)
907 			optlen = cnt;
908 		if (IPOPT_COPIED(opt)) {
909 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
910 			dp += optlen;
911 		}
912 	}
913 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
914 		*dp++ = IPOPT_EOL;
915 	return (optlen);
916 }
917 
918 /*
919  * IP socket option processing.
920  */
921 int
922 ip_ctloutput(op, so, level, optname, mp)
923 	int op;
924 	struct socket *so;
925 	int level, optname;
926 	struct mbuf **mp;
927 {
928 	struct inpcb *inp = sotoinpcb(so);
929 	struct mbuf *m = *mp;
930 	int optval = 0;
931 	int error = 0;
932 #ifdef IPSEC
933 #ifdef __NetBSD__
934 	struct proc *p = curproc;	/*XXX*/
935 #endif
936 #endif
937 
938 	if (level != IPPROTO_IP) {
939 		error = EINVAL;
940 		if (op == PRCO_SETOPT && *mp)
941 			(void) m_free(*mp);
942 	} else switch (op) {
943 
944 	case PRCO_SETOPT:
945 		switch (optname) {
946 		case IP_OPTIONS:
947 #ifdef notyet
948 		case IP_RETOPTS:
949 			return (ip_pcbopts(optname, &inp->inp_options, m));
950 #else
951 			return (ip_pcbopts(&inp->inp_options, m));
952 #endif
953 
954 		case IP_TOS:
955 		case IP_TTL:
956 		case IP_RECVOPTS:
957 		case IP_RECVRETOPTS:
958 		case IP_RECVDSTADDR:
959 		case IP_RECVIF:
960 			if (m == NULL || m->m_len != sizeof(int))
961 				error = EINVAL;
962 			else {
963 				optval = *mtod(m, int *);
964 				switch (optname) {
965 
966 				case IP_TOS:
967 					inp->inp_ip.ip_tos = optval;
968 					break;
969 
970 				case IP_TTL:
971 					inp->inp_ip.ip_ttl = optval;
972 					break;
973 #define	OPTSET(bit) \
974 	if (optval) \
975 		inp->inp_flags |= bit; \
976 	else \
977 		inp->inp_flags &= ~bit;
978 
979 				case IP_RECVOPTS:
980 					OPTSET(INP_RECVOPTS);
981 					break;
982 
983 				case IP_RECVRETOPTS:
984 					OPTSET(INP_RECVRETOPTS);
985 					break;
986 
987 				case IP_RECVDSTADDR:
988 					OPTSET(INP_RECVDSTADDR);
989 					break;
990 
991 				case IP_RECVIF:
992 					OPTSET(INP_RECVIF);
993 					break;
994 				}
995 			}
996 			break;
997 #undef OPTSET
998 
999 		case IP_MULTICAST_IF:
1000 		case IP_MULTICAST_TTL:
1001 		case IP_MULTICAST_LOOP:
1002 		case IP_ADD_MEMBERSHIP:
1003 		case IP_DROP_MEMBERSHIP:
1004 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1005 			break;
1006 
1007 		case IP_PORTRANGE:
1008 			if (m == 0 || m->m_len != sizeof(int))
1009 				error = EINVAL;
1010 			else {
1011 				optval = *mtod(m, int *);
1012 
1013 				switch (optval) {
1014 
1015 				case IP_PORTRANGE_DEFAULT:
1016 				case IP_PORTRANGE_HIGH:
1017 					inp->inp_flags &= ~(INP_LOWPORT);
1018 					break;
1019 
1020 				case IP_PORTRANGE_LOW:
1021 					inp->inp_flags |= INP_LOWPORT;
1022 					break;
1023 
1024 				default:
1025 					error = EINVAL;
1026 					break;
1027 				}
1028 			}
1029 			break;
1030 
1031 #ifdef IPSEC
1032 		case IP_IPSEC_POLICY:
1033 		{
1034 			caddr_t req = NULL;
1035 			size_t len = 0;
1036 			int priv = 0;
1037 
1038 #ifdef __NetBSD__
1039 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1040 				priv = 0;
1041 			else
1042 				priv = 1;
1043 #else
1044 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1045 #endif
1046 			if (m) {
1047 				req = mtod(m, caddr_t);
1048 				len = m->m_len;
1049 			}
1050 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1051 			break;
1052 		    }
1053 #endif /*IPSEC*/
1054 
1055 		default:
1056 			error = ENOPROTOOPT;
1057 			break;
1058 		}
1059 		if (m)
1060 			(void)m_free(m);
1061 		break;
1062 
1063 	case PRCO_GETOPT:
1064 		switch (optname) {
1065 		case IP_OPTIONS:
1066 		case IP_RETOPTS:
1067 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1068 			if (inp->inp_options) {
1069 				m->m_len = inp->inp_options->m_len;
1070 				bcopy(mtod(inp->inp_options, caddr_t),
1071 				    mtod(m, caddr_t), (unsigned)m->m_len);
1072 			} else
1073 				m->m_len = 0;
1074 			break;
1075 
1076 		case IP_TOS:
1077 		case IP_TTL:
1078 		case IP_RECVOPTS:
1079 		case IP_RECVRETOPTS:
1080 		case IP_RECVDSTADDR:
1081 		case IP_RECVIF:
1082 		case IP_ERRORMTU:
1083 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1084 			m->m_len = sizeof(int);
1085 			switch (optname) {
1086 
1087 			case IP_TOS:
1088 				optval = inp->inp_ip.ip_tos;
1089 				break;
1090 
1091 			case IP_TTL:
1092 				optval = inp->inp_ip.ip_ttl;
1093 				break;
1094 
1095 			case IP_ERRORMTU:
1096 				optval = inp->inp_errormtu;
1097 				break;
1098 
1099 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1100 
1101 			case IP_RECVOPTS:
1102 				optval = OPTBIT(INP_RECVOPTS);
1103 				break;
1104 
1105 			case IP_RECVRETOPTS:
1106 				optval = OPTBIT(INP_RECVRETOPTS);
1107 				break;
1108 
1109 			case IP_RECVDSTADDR:
1110 				optval = OPTBIT(INP_RECVDSTADDR);
1111 				break;
1112 
1113 			case IP_RECVIF:
1114 				optval = OPTBIT(INP_RECVIF);
1115 				break;
1116 			}
1117 			*mtod(m, int *) = optval;
1118 			break;
1119 
1120 #ifdef IPSEC
1121 		case IP_IPSEC_POLICY:
1122 		{
1123 			caddr_t req = NULL;
1124 			size_t len = 0;
1125 
1126 			if (m) {
1127 				req = mtod(m, caddr_t);
1128 				len = m->m_len;
1129 			}
1130 			error = ipsec4_get_policy(inp, req, len, mp);
1131 			break;
1132 		}
1133 #endif /*IPSEC*/
1134 
1135 		case IP_MULTICAST_IF:
1136 		case IP_MULTICAST_TTL:
1137 		case IP_MULTICAST_LOOP:
1138 		case IP_ADD_MEMBERSHIP:
1139 		case IP_DROP_MEMBERSHIP:
1140 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1141 			break;
1142 
1143 		case IP_PORTRANGE:
1144 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1145 			m->m_len = sizeof(int);
1146 
1147 			if (inp->inp_flags & INP_LOWPORT)
1148 				optval = IP_PORTRANGE_LOW;
1149 			else
1150 				optval = IP_PORTRANGE_DEFAULT;
1151 
1152 			*mtod(m, int *) = optval;
1153 			break;
1154 
1155 		default:
1156 			error = ENOPROTOOPT;
1157 			break;
1158 		}
1159 		break;
1160 	}
1161 	return (error);
1162 }
1163 
1164 /*
1165  * Set up IP options in pcb for insertion in output packets.
1166  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1167  * with destination address if source routed.
1168  */
1169 int
1170 #ifdef notyet
1171 ip_pcbopts(optname, pcbopt, m)
1172 	int optname;
1173 #else
1174 ip_pcbopts(pcbopt, m)
1175 #endif
1176 	struct mbuf **pcbopt;
1177 	struct mbuf *m;
1178 {
1179 	int cnt, optlen;
1180 	u_char *cp;
1181 	u_char opt;
1182 
1183 	/* turn off any old options */
1184 	if (*pcbopt)
1185 		(void)m_free(*pcbopt);
1186 	*pcbopt = 0;
1187 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1188 		/*
1189 		 * Only turning off any previous options.
1190 		 */
1191 		if (m)
1192 			(void)m_free(m);
1193 		return (0);
1194 	}
1195 
1196 #ifndef	__vax__
1197 	if (m->m_len % sizeof(int32_t))
1198 		goto bad;
1199 #endif
1200 	/*
1201 	 * IP first-hop destination address will be stored before
1202 	 * actual options; move other options back
1203 	 * and clear it when none present.
1204 	 */
1205 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1206 		goto bad;
1207 	cnt = m->m_len;
1208 	m->m_len += sizeof(struct in_addr);
1209 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1210 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1211 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1212 
1213 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1214 		opt = cp[IPOPT_OPTVAL];
1215 		if (opt == IPOPT_EOL)
1216 			break;
1217 		if (opt == IPOPT_NOP)
1218 			optlen = 1;
1219 		else {
1220 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1221 				goto bad;
1222 			optlen = cp[IPOPT_OLEN];
1223 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1224 				goto bad;
1225 		}
1226 		switch (opt) {
1227 
1228 		default:
1229 			break;
1230 
1231 		case IPOPT_LSRR:
1232 		case IPOPT_SSRR:
1233 			/*
1234 			 * user process specifies route as:
1235 			 *	->A->B->C->D
1236 			 * D must be our final destination (but we can't
1237 			 * check that since we may not have connected yet).
1238 			 * A is first hop destination, which doesn't appear in
1239 			 * actual IP option, but is stored before the options.
1240 			 */
1241 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1242 				goto bad;
1243 			m->m_len -= sizeof(struct in_addr);
1244 			cnt -= sizeof(struct in_addr);
1245 			optlen -= sizeof(struct in_addr);
1246 			cp[IPOPT_OLEN] = optlen;
1247 			/*
1248 			 * Move first hop before start of options.
1249 			 */
1250 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1251 			    sizeof(struct in_addr));
1252 			/*
1253 			 * Then copy rest of options back
1254 			 * to close up the deleted entry.
1255 			 */
1256 			memmove(&cp[IPOPT_OFFSET+1],
1257                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1258 			    (unsigned)cnt + sizeof(struct in_addr));
1259 			break;
1260 		}
1261 	}
1262 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1263 		goto bad;
1264 	*pcbopt = m;
1265 	return (0);
1266 
1267 bad:
1268 	(void)m_free(m);
1269 	return (EINVAL);
1270 }
1271 
1272 /*
1273  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1274  */
1275 static struct ifnet *
1276 ip_multicast_if(a, ifindexp)
1277 	struct in_addr *a;
1278 	int *ifindexp;
1279 {
1280 	int ifindex;
1281 	struct ifnet *ifp;
1282 
1283 	if (ifindexp)
1284 		*ifindexp = 0;
1285 	if (ntohl(a->s_addr) >> 24 == 0) {
1286 		ifindex = ntohl(a->s_addr) & 0xffffff;
1287 		if (ifindex < 0 || if_index < ifindex)
1288 			return NULL;
1289 		ifp = ifindex2ifnet[ifindex];
1290 		if (ifindexp)
1291 			*ifindexp = ifindex;
1292 	} else {
1293 		INADDR_TO_IFP(*a, ifp);
1294 	}
1295 	return ifp;
1296 }
1297 
1298 /*
1299  * Set the IP multicast options in response to user setsockopt().
1300  */
1301 int
1302 ip_setmoptions(optname, imop, m)
1303 	int optname;
1304 	struct ip_moptions **imop;
1305 	struct mbuf *m;
1306 {
1307 	int error = 0;
1308 	u_char loop;
1309 	int i;
1310 	struct in_addr addr;
1311 	struct ip_mreq *mreq;
1312 	struct ifnet *ifp;
1313 	struct ip_moptions *imo = *imop;
1314 	struct route ro;
1315 	struct sockaddr_in *dst;
1316 	int ifindex;
1317 
1318 	if (imo == NULL) {
1319 		/*
1320 		 * No multicast option buffer attached to the pcb;
1321 		 * allocate one and initialize to default values.
1322 		 */
1323 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1324 		    M_WAITOK);
1325 
1326 		if (imo == NULL)
1327 			return (ENOBUFS);
1328 		*imop = imo;
1329 		imo->imo_multicast_ifp = NULL;
1330 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1331 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1332 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1333 		imo->imo_num_memberships = 0;
1334 	}
1335 
1336 	switch (optname) {
1337 
1338 	case IP_MULTICAST_IF:
1339 		/*
1340 		 * Select the interface for outgoing multicast packets.
1341 		 */
1342 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1343 			error = EINVAL;
1344 			break;
1345 		}
1346 		addr = *(mtod(m, struct in_addr *));
1347 		/*
1348 		 * INADDR_ANY is used to remove a previous selection.
1349 		 * When no interface is selected, a default one is
1350 		 * chosen every time a multicast packet is sent.
1351 		 */
1352 		if (in_nullhost(addr)) {
1353 			imo->imo_multicast_ifp = NULL;
1354 			break;
1355 		}
1356 		/*
1357 		 * The selected interface is identified by its local
1358 		 * IP address.  Find the interface and confirm that
1359 		 * it supports multicasting.
1360 		 */
1361 		ifp = ip_multicast_if(&addr, &ifindex);
1362 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1363 			error = EADDRNOTAVAIL;
1364 			break;
1365 		}
1366 		imo->imo_multicast_ifp = ifp;
1367 		if (ifindex)
1368 			imo->imo_multicast_addr = addr;
1369 		else
1370 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1371 		break;
1372 
1373 	case IP_MULTICAST_TTL:
1374 		/*
1375 		 * Set the IP time-to-live for outgoing multicast packets.
1376 		 */
1377 		if (m == NULL || m->m_len != 1) {
1378 			error = EINVAL;
1379 			break;
1380 		}
1381 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1382 		break;
1383 
1384 	case IP_MULTICAST_LOOP:
1385 		/*
1386 		 * Set the loopback flag for outgoing multicast packets.
1387 		 * Must be zero or one.
1388 		 */
1389 		if (m == NULL || m->m_len != 1 ||
1390 		   (loop = *(mtod(m, u_char *))) > 1) {
1391 			error = EINVAL;
1392 			break;
1393 		}
1394 		imo->imo_multicast_loop = loop;
1395 		break;
1396 
1397 	case IP_ADD_MEMBERSHIP:
1398 		/*
1399 		 * Add a multicast group membership.
1400 		 * Group must be a valid IP multicast address.
1401 		 */
1402 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1403 			error = EINVAL;
1404 			break;
1405 		}
1406 		mreq = mtod(m, struct ip_mreq *);
1407 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1408 			error = EINVAL;
1409 			break;
1410 		}
1411 		/*
1412 		 * If no interface address was provided, use the interface of
1413 		 * the route to the given multicast address.
1414 		 */
1415 		if (in_nullhost(mreq->imr_interface)) {
1416 			bzero((caddr_t)&ro, sizeof(ro));
1417 			ro.ro_rt = NULL;
1418 			dst = satosin(&ro.ro_dst);
1419 			dst->sin_len = sizeof(*dst);
1420 			dst->sin_family = AF_INET;
1421 			dst->sin_addr = mreq->imr_multiaddr;
1422 			rtalloc(&ro);
1423 			if (ro.ro_rt == NULL) {
1424 				error = EADDRNOTAVAIL;
1425 				break;
1426 			}
1427 			ifp = ro.ro_rt->rt_ifp;
1428 			rtfree(ro.ro_rt);
1429 		} else {
1430 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1431 		}
1432 		/*
1433 		 * See if we found an interface, and confirm that it
1434 		 * supports multicast.
1435 		 */
1436 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1437 			error = EADDRNOTAVAIL;
1438 			break;
1439 		}
1440 		/*
1441 		 * See if the membership already exists or if all the
1442 		 * membership slots are full.
1443 		 */
1444 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1445 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1446 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1447 				      mreq->imr_multiaddr))
1448 				break;
1449 		}
1450 		if (i < imo->imo_num_memberships) {
1451 			error = EADDRINUSE;
1452 			break;
1453 		}
1454 		if (i == IP_MAX_MEMBERSHIPS) {
1455 			error = ETOOMANYREFS;
1456 			break;
1457 		}
1458 		/*
1459 		 * Everything looks good; add a new record to the multicast
1460 		 * address list for the given interface.
1461 		 */
1462 		if ((imo->imo_membership[i] =
1463 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1464 			error = ENOBUFS;
1465 			break;
1466 		}
1467 		++imo->imo_num_memberships;
1468 		break;
1469 
1470 	case IP_DROP_MEMBERSHIP:
1471 		/*
1472 		 * Drop a multicast group membership.
1473 		 * Group must be a valid IP multicast address.
1474 		 */
1475 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1476 			error = EINVAL;
1477 			break;
1478 		}
1479 		mreq = mtod(m, struct ip_mreq *);
1480 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1481 			error = EINVAL;
1482 			break;
1483 		}
1484 		/*
1485 		 * If an interface address was specified, get a pointer
1486 		 * to its ifnet structure.
1487 		 */
1488 		if (in_nullhost(mreq->imr_interface))
1489 			ifp = NULL;
1490 		else {
1491 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1492 			if (ifp == NULL) {
1493 				error = EADDRNOTAVAIL;
1494 				break;
1495 			}
1496 		}
1497 		/*
1498 		 * Find the membership in the membership array.
1499 		 */
1500 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1501 			if ((ifp == NULL ||
1502 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1503 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1504 				       mreq->imr_multiaddr))
1505 				break;
1506 		}
1507 		if (i == imo->imo_num_memberships) {
1508 			error = EADDRNOTAVAIL;
1509 			break;
1510 		}
1511 		/*
1512 		 * Give up the multicast address record to which the
1513 		 * membership points.
1514 		 */
1515 		in_delmulti(imo->imo_membership[i]);
1516 		/*
1517 		 * Remove the gap in the membership array.
1518 		 */
1519 		for (++i; i < imo->imo_num_memberships; ++i)
1520 			imo->imo_membership[i-1] = imo->imo_membership[i];
1521 		--imo->imo_num_memberships;
1522 		break;
1523 
1524 	default:
1525 		error = EOPNOTSUPP;
1526 		break;
1527 	}
1528 
1529 	/*
1530 	 * If all options have default values, no need to keep the mbuf.
1531 	 */
1532 	if (imo->imo_multicast_ifp == NULL &&
1533 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1534 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1535 	    imo->imo_num_memberships == 0) {
1536 		free(*imop, M_IPMOPTS);
1537 		*imop = NULL;
1538 	}
1539 
1540 	return (error);
1541 }
1542 
1543 /*
1544  * Return the IP multicast options in response to user getsockopt().
1545  */
1546 int
1547 ip_getmoptions(optname, imo, mp)
1548 	int optname;
1549 	struct ip_moptions *imo;
1550 	struct mbuf **mp;
1551 {
1552 	u_char *ttl;
1553 	u_char *loop;
1554 	struct in_addr *addr;
1555 	struct in_ifaddr *ia;
1556 
1557 	*mp = m_get(M_WAIT, MT_SOOPTS);
1558 
1559 	switch (optname) {
1560 
1561 	case IP_MULTICAST_IF:
1562 		addr = mtod(*mp, struct in_addr *);
1563 		(*mp)->m_len = sizeof(struct in_addr);
1564 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1565 			*addr = zeroin_addr;
1566 		else if (imo->imo_multicast_addr.s_addr) {
1567 			/* return the value user has set */
1568 			*addr = imo->imo_multicast_addr;
1569 		} else {
1570 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1571 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1572 		}
1573 		return (0);
1574 
1575 	case IP_MULTICAST_TTL:
1576 		ttl = mtod(*mp, u_char *);
1577 		(*mp)->m_len = 1;
1578 		*ttl = imo ? imo->imo_multicast_ttl
1579 			   : IP_DEFAULT_MULTICAST_TTL;
1580 		return (0);
1581 
1582 	case IP_MULTICAST_LOOP:
1583 		loop = mtod(*mp, u_char *);
1584 		(*mp)->m_len = 1;
1585 		*loop = imo ? imo->imo_multicast_loop
1586 			    : IP_DEFAULT_MULTICAST_LOOP;
1587 		return (0);
1588 
1589 	default:
1590 		return (EOPNOTSUPP);
1591 	}
1592 }
1593 
1594 /*
1595  * Discard the IP multicast options.
1596  */
1597 void
1598 ip_freemoptions(imo)
1599 	struct ip_moptions *imo;
1600 {
1601 	int i;
1602 
1603 	if (imo != NULL) {
1604 		for (i = 0; i < imo->imo_num_memberships; ++i)
1605 			in_delmulti(imo->imo_membership[i]);
1606 		free(imo, M_IPMOPTS);
1607 	}
1608 }
1609 
1610 /*
1611  * Routine called from ip_output() to loop back a copy of an IP multicast
1612  * packet to the input queue of a specified interface.  Note that this
1613  * calls the output routine of the loopback "driver", but with an interface
1614  * pointer that might NOT be &loif -- easier than replicating that code here.
1615  */
1616 static void
1617 ip_mloopback(ifp, m, dst)
1618 	struct ifnet *ifp;
1619 	struct mbuf *m;
1620 	struct sockaddr_in *dst;
1621 {
1622 	struct ip *ip;
1623 	struct mbuf *copym;
1624 
1625 	copym = m_copy(m, 0, M_COPYALL);
1626 	if (copym != NULL
1627 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1628 		copym = m_pullup(copym, sizeof(struct ip));
1629 	if (copym != NULL) {
1630 		/*
1631 		 * We don't bother to fragment if the IP length is greater
1632 		 * than the interface's MTU.  Can this possibly matter?
1633 		 */
1634 		ip = mtod(copym, struct ip *);
1635 		HTONS(ip->ip_len);
1636 		HTONS(ip->ip_off);
1637 		ip->ip_sum = 0;
1638 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1639 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1640 	}
1641 }
1642