xref: /netbsd-src/sys/netinet/ip_output.c (revision 27578b9aac214cc7796ead81dcc5427e79d5f2a0)
1 /*	$NetBSD: ip_output.c,v 1.88 2001/09/17 17:27:00 thorpej 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 "opt_pfil_hooks.h"
105 #include "opt_ipsec.h"
106 #include "opt_mrouting.h"
107 
108 #include <sys/param.h>
109 #include <sys/malloc.h>
110 #include <sys/mbuf.h>
111 #include <sys/errno.h>
112 #include <sys/protosw.h>
113 #include <sys/socket.h>
114 #include <sys/socketvar.h>
115 #include <sys/systm.h>
116 #include <sys/proc.h>
117 
118 #include <net/if.h>
119 #include <net/route.h>
120 #include <net/pfil.h>
121 
122 #include <netinet/in.h>
123 #include <netinet/in_systm.h>
124 #include <netinet/ip.h>
125 #include <netinet/in_pcb.h>
126 #include <netinet/in_var.h>
127 #include <netinet/ip_var.h>
128 
129 #ifdef MROUTING
130 #include <netinet/ip_mroute.h>
131 #endif
132 
133 #include <machine/stdarg.h>
134 
135 #ifdef IPSEC
136 #include <netinet6/ipsec.h>
137 #include <netkey/key.h>
138 #include <netkey/key_debug.h>
139 #endif /*IPSEC*/
140 
141 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
142 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
143 static void ip_mloopback
144 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
145 
146 #ifdef PFIL_HOOKS
147 extern struct pfil_head inet_pfil_hook;			/* XXX */
148 #endif
149 
150 /*
151  * IP output.  The packet in mbuf chain m contains a skeletal IP
152  * header (with len, off, ttl, proto, tos, src, dst).
153  * The mbuf chain containing the packet will be freed.
154  * The mbuf opt, if present, will not be freed.
155  */
156 int
157 #if __STDC__
158 ip_output(struct mbuf *m0, ...)
159 #else
160 ip_output(m0, va_alist)
161 	struct mbuf *m0;
162 	va_dcl
163 #endif
164 {
165 	struct ip *ip, *mhip;
166 	struct ifnet *ifp;
167 	struct mbuf *m = m0;
168 	int hlen = sizeof (struct ip);
169 	int len, off, error = 0;
170 	struct route iproute;
171 	struct sockaddr_in *dst;
172 	struct in_ifaddr *ia;
173 	struct mbuf *opt;
174 	struct route *ro;
175 	int flags, sw_csum;
176 	int *mtu_p;
177 	int mtu;
178 	struct ip_moptions *imo;
179 	va_list ap;
180 #ifdef IPSEC
181 	struct socket *so;
182 	struct secpolicy *sp = NULL;
183 #endif /*IPSEC*/
184 	u_int16_t ip_len;
185 
186 	va_start(ap, m0);
187 	opt = va_arg(ap, struct mbuf *);
188 	ro = va_arg(ap, struct route *);
189 	flags = va_arg(ap, int);
190 	imo = va_arg(ap, struct ip_moptions *);
191 	if (flags & IP_RETURNMTU)
192 		mtu_p = va_arg(ap, int *);
193 	else
194 		mtu_p = NULL;
195 	va_end(ap);
196 
197 #ifdef IPSEC
198 	so = ipsec_getsocket(m);
199 	(void)ipsec_setsocket(m, NULL);
200 #endif /*IPSEC*/
201 
202 #ifdef	DIAGNOSTIC
203 	if ((m->m_flags & M_PKTHDR) == 0)
204 		panic("ip_output no HDR");
205 #endif
206 	if (opt) {
207 		m = ip_insertoptions(m, opt, &len);
208 		hlen = len;
209 	}
210 	ip = mtod(m, struct ip *);
211 	/*
212 	 * Fill in IP header.
213 	 */
214 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
215 		ip->ip_v = IPVERSION;
216 		ip->ip_off = 0;
217 		ip->ip_id = htons(ip_id++);
218 		ip->ip_hl = hlen >> 2;
219 		ipstat.ips_localout++;
220 	} else {
221 		hlen = ip->ip_hl << 2;
222 	}
223 	/*
224 	 * Route packet.
225 	 */
226 	if (ro == 0) {
227 		ro = &iproute;
228 		bzero((caddr_t)ro, sizeof (*ro));
229 	}
230 	dst = satosin(&ro->ro_dst);
231 	/*
232 	 * If there is a cached route,
233 	 * check that it is to the same destination
234 	 * and is still up.  If not, free it and try again.
235 	 */
236 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
237 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
238 		RTFREE(ro->ro_rt);
239 		ro->ro_rt = (struct rtentry *)0;
240 	}
241 	if (ro->ro_rt == 0) {
242 		dst->sin_family = AF_INET;
243 		dst->sin_len = sizeof(*dst);
244 		dst->sin_addr = ip->ip_dst;
245 	}
246 	/*
247 	 * If routing to interface only,
248 	 * short circuit routing lookup.
249 	 */
250 	if (flags & IP_ROUTETOIF) {
251 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
252 			ipstat.ips_noroute++;
253 			error = ENETUNREACH;
254 			goto bad;
255 		}
256 		ifp = ia->ia_ifp;
257 		mtu = ifp->if_mtu;
258 		ip->ip_ttl = 1;
259 	} else {
260 		if (ro->ro_rt == 0)
261 			rtalloc(ro);
262 		if (ro->ro_rt == 0) {
263 			ipstat.ips_noroute++;
264 			error = EHOSTUNREACH;
265 			goto bad;
266 		}
267 		ia = ifatoia(ro->ro_rt->rt_ifa);
268 		ifp = ro->ro_rt->rt_ifp;
269 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
270 			mtu = ifp->if_mtu;
271 		ro->ro_rt->rt_use++;
272 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
273 			dst = satosin(ro->ro_rt->rt_gateway);
274 	}
275 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
276 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
277 		struct in_multi *inm;
278 
279 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
280 			M_BCAST : M_MCAST;
281 		/*
282 		 * IP destination address is multicast.  Make sure "dst"
283 		 * still points to the address in "ro".  (It may have been
284 		 * changed to point to a gateway address, above.)
285 		 */
286 		dst = satosin(&ro->ro_dst);
287 		/*
288 		 * See if the caller provided any multicast options
289 		 */
290 		if (imo != NULL) {
291 			ip->ip_ttl = imo->imo_multicast_ttl;
292 			if (imo->imo_multicast_ifp != NULL) {
293 				ifp = imo->imo_multicast_ifp;
294 				mtu = ifp->if_mtu;
295 			}
296 		} else
297 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
298 		/*
299 		 * Confirm that the outgoing interface supports multicast.
300 		 */
301 		if (((m->m_flags & M_MCAST) &&
302 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
303 		    ((m->m_flags & M_BCAST) &&
304 		     (ifp->if_flags & IFF_BROADCAST) == 0))  {
305 			ipstat.ips_noroute++;
306 			error = ENETUNREACH;
307 			goto bad;
308 		}
309 		/*
310 		 * If source address not specified yet, use an address
311 		 * of outgoing interface.
312 		 */
313 		if (in_nullhost(ip->ip_src)) {
314 			struct in_ifaddr *ia;
315 
316 			IFP_TO_IA(ifp, ia);
317 			ip->ip_src = ia->ia_addr.sin_addr;
318 		}
319 
320 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
321 		if (inm != NULL &&
322 		   (imo == NULL || imo->imo_multicast_loop)) {
323 			/*
324 			 * If we belong to the destination multicast group
325 			 * on the outgoing interface, and the caller did not
326 			 * forbid loopback, loop back a copy.
327 			 */
328 			ip_mloopback(ifp, m, dst);
329 		}
330 #ifdef MROUTING
331 		else {
332 			/*
333 			 * If we are acting as a multicast router, perform
334 			 * multicast forwarding as if the packet had just
335 			 * arrived on the interface to which we are about
336 			 * to send.  The multicast forwarding function
337 			 * recursively calls this function, using the
338 			 * IP_FORWARDING flag to prevent infinite recursion.
339 			 *
340 			 * Multicasts that are looped back by ip_mloopback(),
341 			 * above, will be forwarded by the ip_input() routine,
342 			 * if necessary.
343 			 */
344 			extern struct socket *ip_mrouter;
345 
346 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
347 				if (ip_mforward(m, ifp) != 0) {
348 					m_freem(m);
349 					goto done;
350 				}
351 			}
352 		}
353 #endif
354 		/*
355 		 * Multicasts with a time-to-live of zero may be looped-
356 		 * back, above, but must not be transmitted on a network.
357 		 * Also, multicasts addressed to the loopback interface
358 		 * are not sent -- the above call to ip_mloopback() will
359 		 * loop back a copy if this host actually belongs to the
360 		 * destination group on the loopback interface.
361 		 */
362 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
363 			m_freem(m);
364 			goto done;
365 		}
366 
367 		goto sendit;
368 	}
369 #ifndef notdef
370 	/*
371 	 * If source address not specified yet, use address
372 	 * of outgoing interface.
373 	 */
374 	if (in_nullhost(ip->ip_src))
375 		ip->ip_src = ia->ia_addr.sin_addr;
376 #endif
377 
378 	/*
379 	 * packets with Class-D address as source are not valid per
380 	 * RFC 1112
381 	 */
382 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
383 		ipstat.ips_odropped++;
384 		error = EADDRNOTAVAIL;
385 		goto bad;
386 	}
387 
388 	/*
389 	 * Look for broadcast address and
390 	 * and verify user is allowed to send
391 	 * such a packet.
392 	 */
393 	if (in_broadcast(dst->sin_addr, ifp)) {
394 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
395 			error = EADDRNOTAVAIL;
396 			goto bad;
397 		}
398 		if ((flags & IP_ALLOWBROADCAST) == 0) {
399 			error = EACCES;
400 			goto bad;
401 		}
402 		/* don't allow broadcast messages to be fragmented */
403 		if ((u_int16_t)ip->ip_len > ifp->if_mtu) {
404 			error = EMSGSIZE;
405 			goto bad;
406 		}
407 		m->m_flags |= M_BCAST;
408 	} else
409 		m->m_flags &= ~M_BCAST;
410 
411 sendit:
412 	/*
413 	 * If we're doing Path MTU Discovery, we need to set DF unless
414 	 * the route's MTU is locked.
415 	 */
416 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
417 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
418 		ip->ip_off |= IP_DF;
419 
420 	/*
421 	 * Remember the current ip_len and ip_off, and swap them into
422 	 * network order.
423 	 */
424 	ip_len = ip->ip_len;
425 
426 	HTONS(ip->ip_len);
427 	HTONS(ip->ip_off);
428 
429 #ifdef IPSEC
430 	/* get SP for this packet */
431 	if (so == NULL)
432 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
433 	else
434 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
435 
436 	if (sp == NULL) {
437 		ipsecstat.out_inval++;
438 		goto bad;
439 	}
440 
441 	error = 0;
442 
443 	/* check policy */
444 	switch (sp->policy) {
445 	case IPSEC_POLICY_DISCARD:
446 		/*
447 		 * This packet is just discarded.
448 		 */
449 		ipsecstat.out_polvio++;
450 		goto bad;
451 
452 	case IPSEC_POLICY_BYPASS:
453 	case IPSEC_POLICY_NONE:
454 		/* no need to do IPsec. */
455 		goto skip_ipsec;
456 
457 	case IPSEC_POLICY_IPSEC:
458 		if (sp->req == NULL) {
459 			/* XXX should be panic ? */
460 			printf("ip_output: No IPsec request specified.\n");
461 			error = EINVAL;
462 			goto bad;
463 		}
464 		break;
465 
466 	case IPSEC_POLICY_ENTRUST:
467 	default:
468 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
469 	}
470 
471 	/*
472 	 * ipsec4_output() expects ip_len and ip_off in network
473 	 * order.  They have been set to network order above.
474 	 */
475 
476     {
477 	struct ipsec_output_state state;
478 	bzero(&state, sizeof(state));
479 	state.m = m;
480 	if (flags & IP_ROUTETOIF) {
481 		state.ro = &iproute;
482 		bzero(&iproute, sizeof(iproute));
483 	} else
484 		state.ro = ro;
485 	state.dst = (struct sockaddr *)dst;
486 
487 	/*
488 	 * We can't defer the checksum of payload data if
489 	 * we're about to encrypt/authenticate it.
490 	 *
491 	 * XXX When we support crypto offloading functions of
492 	 * XXX network interfaces, we need to reconsider this,
493 	 * XXX since it's likely that they'll support checksumming,
494 	 * XXX as well.
495 	 */
496 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
497 		in_delayed_cksum(m);
498 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
499 	}
500 
501 	error = ipsec4_output(&state, sp, flags);
502 
503 	m = state.m;
504 	if (flags & IP_ROUTETOIF) {
505 		/*
506 		 * if we have tunnel mode SA, we may need to ignore
507 		 * IP_ROUTETOIF.
508 		 */
509 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
510 			flags &= ~IP_ROUTETOIF;
511 			ro = state.ro;
512 		}
513 	} else
514 		ro = state.ro;
515 	dst = (struct sockaddr_in *)state.dst;
516 	if (error) {
517 		/* mbuf is already reclaimed in ipsec4_output. */
518 		m0 = NULL;
519 		switch (error) {
520 		case EHOSTUNREACH:
521 		case ENETUNREACH:
522 		case EMSGSIZE:
523 		case ENOBUFS:
524 		case ENOMEM:
525 			break;
526 		default:
527 			printf("ip4_output (ipsec): error code %d\n", error);
528 			/*fall through*/
529 		case ENOENT:
530 			/* don't show these error codes to the user */
531 			error = 0;
532 			break;
533 		}
534 		goto bad;
535 	}
536     }
537 
538 	/* be sure to update variables that are affected by ipsec4_output() */
539 	ip = mtod(m, struct ip *);
540 #ifdef _IP_VHL
541 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
542 #else
543 	hlen = ip->ip_hl << 2;
544 #endif
545 	ip_len = ntohs(ip->ip_len);
546 
547 	if (ro->ro_rt == NULL) {
548 		if ((flags & IP_ROUTETOIF) == 0) {
549 			printf("ip_output: "
550 				"can't update route after IPsec processing\n");
551 			error = EHOSTUNREACH;	/*XXX*/
552 			goto bad;
553 		}
554 	} else {
555 		/* nobody uses ia beyond here */
556 		ifp = ro->ro_rt->rt_ifp;
557 	}
558 
559 skip_ipsec:
560 #endif /*IPSEC*/
561 
562 #ifdef PFIL_HOOKS
563 	/*
564 	 * Run through list of hooks for output packets.
565 	 */
566 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp,
567 				    PFIL_OUT)) != 0)
568 		goto done;
569 	if (m == NULL)
570 		goto done;
571 
572 	ip = mtod(m, struct ip *);
573 #endif /* PFIL_HOOKS */
574 
575 	/*
576 	 * If small enough for mtu of path, can just send directly.
577 	 */
578 	if (ip_len <= mtu) {
579 #if IFA_STATS
580 		/*
581 		 * search for the source address structure to
582 		 * maintain output statistics.
583 		 */
584 		INADDR_TO_IA(ip->ip_src, ia);
585 		if (ia)
586 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
587 #endif
588 		/*
589 		 * Always initialize the sum to 0!  Some HW assisted
590 		 * checksumming requires this.
591 		 */
592 		ip->ip_sum = 0;
593 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
594 
595 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
596 
597 		/*
598 		 * Perform any checksums that the hardware can't do
599 		 * for us.
600 		 *
601 		 * XXX Does any hardware require the {th,uh}_sum
602 		 * XXX fields to be 0?
603 		 */
604 		if (sw_csum & M_CSUM_IPv4)
605 			ip->ip_sum = in_cksum(m, hlen);
606 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
607 			in_delayed_cksum(m);
608 			sw_csum &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
609 		}
610 		m->m_pkthdr.csum_flags &= ifp->if_csum_flags_tx;
611 
612 #ifdef IPSEC
613 		/* clean ipsec history once it goes out of the node */
614 		ipsec_delaux(m);
615 #endif
616 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
617 		goto done;
618 	}
619 
620 	/*
621 	 * We can't use HW checksumming if we're about to
622 	 * to fragment the packet.
623 	 *
624 	 * XXX Some hardware can do this.
625 	 */
626 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
627 		in_delayed_cksum(m);
628 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
629 	}
630 
631 	/*
632 	 * Too large for interface; fragment if possible.
633 	 * Must be able to put at least 8 bytes per fragment.
634 	 *
635 	 * Note we swap ip_len and ip_off into host order to make
636 	 * the logic below a little simpler.
637 	 */
638 
639 	NTOHS(ip->ip_len);
640 	NTOHS(ip->ip_off);
641 
642 	if (ip->ip_off & IP_DF) {
643 		if (flags & IP_RETURNMTU)
644 			*mtu_p = mtu;
645 		error = EMSGSIZE;
646 		ipstat.ips_cantfrag++;
647 		goto bad;
648 	}
649 	len = (mtu - hlen) &~ 7;
650 	if (len < 8) {
651 		error = EMSGSIZE;
652 		goto bad;
653 	}
654 
655     {
656 	int mhlen, firstlen = len;
657 	struct mbuf **mnext = &m->m_nextpkt;
658 	int fragments = 0;
659 	int s;
660 
661 	/*
662 	 * Loop through length of segment after first fragment,
663 	 * make new header and copy data of each part and link onto chain.
664 	 */
665 	m0 = m;
666 	mhlen = sizeof (struct ip);
667 	for (off = hlen + len; off < (u_int16_t)ip->ip_len; off += len) {
668 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
669 		if (m == 0) {
670 			error = ENOBUFS;
671 			ipstat.ips_odropped++;
672 			goto sendorfree;
673 		}
674 		*mnext = m;
675 		mnext = &m->m_nextpkt;
676 		m->m_data += max_linkhdr;
677 		mhip = mtod(m, struct ip *);
678 		*mhip = *ip;
679 		/* we must inherit MCAST and BCAST flags */
680 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
681 		if (hlen > sizeof (struct ip)) {
682 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
683 			mhip->ip_hl = mhlen >> 2;
684 		}
685 		m->m_len = mhlen;
686 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
687 		if (ip->ip_off & IP_MF)
688 			mhip->ip_off |= IP_MF;
689 		if (off + len >= (u_int16_t)ip->ip_len)
690 			len = (u_int16_t)ip->ip_len - off;
691 		else
692 			mhip->ip_off |= IP_MF;
693 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
694 		m->m_next = m_copy(m0, off, len);
695 		if (m->m_next == 0) {
696 			error = ENOBUFS;	/* ??? */
697 			ipstat.ips_odropped++;
698 			goto sendorfree;
699 		}
700 		m->m_pkthdr.len = mhlen + len;
701 		m->m_pkthdr.rcvif = (struct ifnet *)0;
702 		HTONS(mhip->ip_off);
703 		mhip->ip_sum = 0;
704 		mhip->ip_sum = in_cksum(m, mhlen);
705 		ipstat.ips_ofragments++;
706 		fragments++;
707 	}
708 	/*
709 	 * Update first fragment by trimming what's been copied out
710 	 * and updating header, then send each fragment (in order).
711 	 */
712 	m = m0;
713 	m_adj(m, hlen + firstlen - (u_int16_t)ip->ip_len);
714 	m->m_pkthdr.len = hlen + firstlen;
715 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
716 	ip->ip_off |= IP_MF;
717 	HTONS(ip->ip_off);
718 	ip->ip_sum = 0;
719 	ip->ip_sum = in_cksum(m, hlen);
720 sendorfree:
721 	/*
722 	 * If there is no room for all the fragments, don't queue
723 	 * any of them.
724 	 */
725 	s = splnet();
726 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
727 		error = ENOBUFS;
728 	splx(s);
729 	for (m = m0; m; m = m0) {
730 		m0 = m->m_nextpkt;
731 		m->m_nextpkt = 0;
732 		if (error == 0) {
733 #if IFA_STATS
734 			/*
735 			 * search for the source address structure to
736 			 * maintain output statistics.
737 			 */
738 			INADDR_TO_IA(ip->ip_src, ia);
739 			if (ia) {
740 				ia->ia_ifa.ifa_data.ifad_outbytes +=
741 					ntohs(ip->ip_len);
742 			}
743 #endif
744 #ifdef IPSEC
745 			/* clean ipsec history once it goes out of the node */
746 			ipsec_delaux(m);
747 #endif
748 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
749 			    ro->ro_rt);
750 		} else
751 			m_freem(m);
752 	}
753 
754 	if (error == 0)
755 		ipstat.ips_fragmented++;
756     }
757 done:
758 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
759 		RTFREE(ro->ro_rt);
760 		ro->ro_rt = 0;
761 	}
762 
763 #ifdef IPSEC
764 	if (sp != NULL) {
765 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
766 			printf("DP ip_output call free SP:%p\n", sp));
767 		key_freesp(sp);
768 	}
769 #endif /* IPSEC */
770 
771 	return (error);
772 bad:
773 	m_freem(m);
774 	goto done;
775 }
776 
777 /*
778  * Process a delayed payload checksum calculation.
779  */
780 void
781 in_delayed_cksum(struct mbuf *m)
782 {
783 	struct ip *ip;
784 	u_int16_t csum, offset;
785 
786 	ip = mtod(m, struct ip *);
787 	offset = ip->ip_hl << 2;
788 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
789 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
790 		csum = 0xffff;
791 
792 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
793 
794 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
795 		/* This happen when ip options were inserted
796 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
797 		    m->m_len, offset, ip->ip_p);
798 		 */
799 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
800 	} else
801 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
802 }
803 
804 /*
805  * Determine the maximum length of the options to be inserted;
806  * we would far rather allocate too much space rather than too little.
807  */
808 
809 u_int
810 ip_optlen(inp)
811 	struct inpcb *inp;
812 {
813 	struct mbuf *m = inp->inp_options;
814 
815 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
816 		return(m->m_len - offsetof(struct ipoption, ipopt_dst));
817 	else
818 		return 0;
819 }
820 
821 
822 /*
823  * Insert IP options into preformed packet.
824  * Adjust IP destination as required for IP source routing,
825  * as indicated by a non-zero in_addr at the start of the options.
826  */
827 static struct mbuf *
828 ip_insertoptions(m, opt, phlen)
829 	struct mbuf *m;
830 	struct mbuf *opt;
831 	int *phlen;
832 {
833 	struct ipoption *p = mtod(opt, struct ipoption *);
834 	struct mbuf *n;
835 	struct ip *ip = mtod(m, struct ip *);
836 	unsigned optlen;
837 
838 	optlen = opt->m_len - sizeof(p->ipopt_dst);
839 	if (optlen + (u_int16_t)ip->ip_len > IP_MAXPACKET)
840 		return (m);		/* XXX should fail */
841 	if (!in_nullhost(p->ipopt_dst))
842 		ip->ip_dst = p->ipopt_dst;
843 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
844 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
845 		if (n == 0)
846 			return (m);
847 		M_COPY_PKTHDR(n, m);
848 		m->m_flags &= ~M_PKTHDR;
849 		m->m_len -= sizeof(struct ip);
850 		m->m_data += sizeof(struct ip);
851 		n->m_next = m;
852 		m = n;
853 		m->m_len = optlen + sizeof(struct ip);
854 		m->m_data += max_linkhdr;
855 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
856 	} else {
857 		m->m_data -= optlen;
858 		m->m_len += optlen;
859 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
860 	}
861 	m->m_pkthdr.len += optlen;
862 	ip = mtod(m, struct ip *);
863 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
864 	*phlen = sizeof(struct ip) + optlen;
865 	ip->ip_len += optlen;
866 	return (m);
867 }
868 
869 /*
870  * Copy options from ip to jp,
871  * omitting those not copied during fragmentation.
872  */
873 int
874 ip_optcopy(ip, jp)
875 	struct ip *ip, *jp;
876 {
877 	u_char *cp, *dp;
878 	int opt, optlen, cnt;
879 
880 	cp = (u_char *)(ip + 1);
881 	dp = (u_char *)(jp + 1);
882 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
883 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
884 		opt = cp[0];
885 		if (opt == IPOPT_EOL)
886 			break;
887 		if (opt == IPOPT_NOP) {
888 			/* Preserve for IP mcast tunnel's LSRR alignment. */
889 			*dp++ = IPOPT_NOP;
890 			optlen = 1;
891 			continue;
892 		}
893 #ifdef DIAGNOSTIC
894 		if (cnt < IPOPT_OLEN + sizeof(*cp))
895 			panic("malformed IPv4 option passed to ip_optcopy");
896 #endif
897 		optlen = cp[IPOPT_OLEN];
898 #ifdef DIAGNOSTIC
899 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
900 			panic("malformed IPv4 option passed to ip_optcopy");
901 #endif
902 		/* bogus lengths should have been caught by ip_dooptions */
903 		if (optlen > cnt)
904 			optlen = cnt;
905 		if (IPOPT_COPIED(opt)) {
906 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
907 			dp += optlen;
908 		}
909 	}
910 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
911 		*dp++ = IPOPT_EOL;
912 	return (optlen);
913 }
914 
915 /*
916  * IP socket option processing.
917  */
918 int
919 ip_ctloutput(op, so, level, optname, mp)
920 	int op;
921 	struct socket *so;
922 	int level, optname;
923 	struct mbuf **mp;
924 {
925 	struct inpcb *inp = sotoinpcb(so);
926 	struct mbuf *m = *mp;
927 	int optval = 0;
928 	int error = 0;
929 #ifdef IPSEC
930 #ifdef __NetBSD__
931 	struct proc *p = curproc;	/*XXX*/
932 #endif
933 #endif
934 
935 	if (level != IPPROTO_IP) {
936 		error = EINVAL;
937 		if (op == PRCO_SETOPT && *mp)
938 			(void) m_free(*mp);
939 	} else switch (op) {
940 
941 	case PRCO_SETOPT:
942 		switch (optname) {
943 		case IP_OPTIONS:
944 #ifdef notyet
945 		case IP_RETOPTS:
946 			return (ip_pcbopts(optname, &inp->inp_options, m));
947 #else
948 			return (ip_pcbopts(&inp->inp_options, m));
949 #endif
950 
951 		case IP_TOS:
952 		case IP_TTL:
953 		case IP_RECVOPTS:
954 		case IP_RECVRETOPTS:
955 		case IP_RECVDSTADDR:
956 		case IP_RECVIF:
957 			if (m == NULL || m->m_len != sizeof(int))
958 				error = EINVAL;
959 			else {
960 				optval = *mtod(m, int *);
961 				switch (optname) {
962 
963 				case IP_TOS:
964 					inp->inp_ip.ip_tos = optval;
965 					break;
966 
967 				case IP_TTL:
968 					inp->inp_ip.ip_ttl = optval;
969 					break;
970 #define	OPTSET(bit) \
971 	if (optval) \
972 		inp->inp_flags |= bit; \
973 	else \
974 		inp->inp_flags &= ~bit;
975 
976 				case IP_RECVOPTS:
977 					OPTSET(INP_RECVOPTS);
978 					break;
979 
980 				case IP_RECVRETOPTS:
981 					OPTSET(INP_RECVRETOPTS);
982 					break;
983 
984 				case IP_RECVDSTADDR:
985 					OPTSET(INP_RECVDSTADDR);
986 					break;
987 
988 				case IP_RECVIF:
989 					OPTSET(INP_RECVIF);
990 					break;
991 				}
992 			}
993 			break;
994 #undef OPTSET
995 
996 		case IP_MULTICAST_IF:
997 		case IP_MULTICAST_TTL:
998 		case IP_MULTICAST_LOOP:
999 		case IP_ADD_MEMBERSHIP:
1000 		case IP_DROP_MEMBERSHIP:
1001 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1002 			break;
1003 
1004 		case IP_PORTRANGE:
1005 			if (m == 0 || m->m_len != sizeof(int))
1006 				error = EINVAL;
1007 			else {
1008 				optval = *mtod(m, int *);
1009 
1010 				switch (optval) {
1011 
1012 				case IP_PORTRANGE_DEFAULT:
1013 				case IP_PORTRANGE_HIGH:
1014 					inp->inp_flags &= ~(INP_LOWPORT);
1015 					break;
1016 
1017 				case IP_PORTRANGE_LOW:
1018 					inp->inp_flags |= INP_LOWPORT;
1019 					break;
1020 
1021 				default:
1022 					error = EINVAL;
1023 					break;
1024 				}
1025 			}
1026 			break;
1027 
1028 #ifdef IPSEC
1029 		case IP_IPSEC_POLICY:
1030 		{
1031 			caddr_t req = NULL;
1032 			size_t len = 0;
1033 			int priv = 0;
1034 
1035 #ifdef __NetBSD__
1036 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1037 				priv = 0;
1038 			else
1039 				priv = 1;
1040 #else
1041 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1042 #endif
1043 			if (m) {
1044 				req = mtod(m, caddr_t);
1045 				len = m->m_len;
1046 			}
1047 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1048 			break;
1049 		    }
1050 #endif /*IPSEC*/
1051 
1052 		default:
1053 			error = ENOPROTOOPT;
1054 			break;
1055 		}
1056 		if (m)
1057 			(void)m_free(m);
1058 		break;
1059 
1060 	case PRCO_GETOPT:
1061 		switch (optname) {
1062 		case IP_OPTIONS:
1063 		case IP_RETOPTS:
1064 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1065 			if (inp->inp_options) {
1066 				m->m_len = inp->inp_options->m_len;
1067 				bcopy(mtod(inp->inp_options, caddr_t),
1068 				    mtod(m, caddr_t), (unsigned)m->m_len);
1069 			} else
1070 				m->m_len = 0;
1071 			break;
1072 
1073 		case IP_TOS:
1074 		case IP_TTL:
1075 		case IP_RECVOPTS:
1076 		case IP_RECVRETOPTS:
1077 		case IP_RECVDSTADDR:
1078 		case IP_RECVIF:
1079 		case IP_ERRORMTU:
1080 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1081 			m->m_len = sizeof(int);
1082 			switch (optname) {
1083 
1084 			case IP_TOS:
1085 				optval = inp->inp_ip.ip_tos;
1086 				break;
1087 
1088 			case IP_TTL:
1089 				optval = inp->inp_ip.ip_ttl;
1090 				break;
1091 
1092 			case IP_ERRORMTU:
1093 				optval = inp->inp_errormtu;
1094 				break;
1095 
1096 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1097 
1098 			case IP_RECVOPTS:
1099 				optval = OPTBIT(INP_RECVOPTS);
1100 				break;
1101 
1102 			case IP_RECVRETOPTS:
1103 				optval = OPTBIT(INP_RECVRETOPTS);
1104 				break;
1105 
1106 			case IP_RECVDSTADDR:
1107 				optval = OPTBIT(INP_RECVDSTADDR);
1108 				break;
1109 
1110 			case IP_RECVIF:
1111 				optval = OPTBIT(INP_RECVIF);
1112 				break;
1113 			}
1114 			*mtod(m, int *) = optval;
1115 			break;
1116 
1117 #ifdef IPSEC
1118 		case IP_IPSEC_POLICY:
1119 		{
1120 			caddr_t req = NULL;
1121 			size_t len = 0;
1122 
1123 			if (m) {
1124 				req = mtod(m, caddr_t);
1125 				len = m->m_len;
1126 			}
1127 			error = ipsec4_get_policy(inp, req, len, mp);
1128 			break;
1129 		}
1130 #endif /*IPSEC*/
1131 
1132 		case IP_MULTICAST_IF:
1133 		case IP_MULTICAST_TTL:
1134 		case IP_MULTICAST_LOOP:
1135 		case IP_ADD_MEMBERSHIP:
1136 		case IP_DROP_MEMBERSHIP:
1137 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1138 			break;
1139 
1140 		case IP_PORTRANGE:
1141 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1142 			m->m_len = sizeof(int);
1143 
1144 			if (inp->inp_flags & INP_LOWPORT)
1145 				optval = IP_PORTRANGE_LOW;
1146 			else
1147 				optval = IP_PORTRANGE_DEFAULT;
1148 
1149 			*mtod(m, int *) = optval;
1150 			break;
1151 
1152 		default:
1153 			error = ENOPROTOOPT;
1154 			break;
1155 		}
1156 		break;
1157 	}
1158 	return (error);
1159 }
1160 
1161 /*
1162  * Set up IP options in pcb for insertion in output packets.
1163  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1164  * with destination address if source routed.
1165  */
1166 int
1167 #ifdef notyet
1168 ip_pcbopts(optname, pcbopt, m)
1169 	int optname;
1170 #else
1171 ip_pcbopts(pcbopt, m)
1172 #endif
1173 	struct mbuf **pcbopt;
1174 	struct mbuf *m;
1175 {
1176 	int cnt, optlen;
1177 	u_char *cp;
1178 	u_char opt;
1179 
1180 	/* turn off any old options */
1181 	if (*pcbopt)
1182 		(void)m_free(*pcbopt);
1183 	*pcbopt = 0;
1184 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1185 		/*
1186 		 * Only turning off any previous options.
1187 		 */
1188 		if (m)
1189 			(void)m_free(m);
1190 		return (0);
1191 	}
1192 
1193 #ifndef	__vax__
1194 	if (m->m_len % sizeof(int32_t))
1195 		goto bad;
1196 #endif
1197 	/*
1198 	 * IP first-hop destination address will be stored before
1199 	 * actual options; move other options back
1200 	 * and clear it when none present.
1201 	 */
1202 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1203 		goto bad;
1204 	cnt = m->m_len;
1205 	m->m_len += sizeof(struct in_addr);
1206 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1207 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1208 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1209 
1210 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1211 		opt = cp[IPOPT_OPTVAL];
1212 		if (opt == IPOPT_EOL)
1213 			break;
1214 		if (opt == IPOPT_NOP)
1215 			optlen = 1;
1216 		else {
1217 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1218 				goto bad;
1219 			optlen = cp[IPOPT_OLEN];
1220 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1221 				goto bad;
1222 		}
1223 		switch (opt) {
1224 
1225 		default:
1226 			break;
1227 
1228 		case IPOPT_LSRR:
1229 		case IPOPT_SSRR:
1230 			/*
1231 			 * user process specifies route as:
1232 			 *	->A->B->C->D
1233 			 * D must be our final destination (but we can't
1234 			 * check that since we may not have connected yet).
1235 			 * A is first hop destination, which doesn't appear in
1236 			 * actual IP option, but is stored before the options.
1237 			 */
1238 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1239 				goto bad;
1240 			m->m_len -= sizeof(struct in_addr);
1241 			cnt -= sizeof(struct in_addr);
1242 			optlen -= sizeof(struct in_addr);
1243 			cp[IPOPT_OLEN] = optlen;
1244 			/*
1245 			 * Move first hop before start of options.
1246 			 */
1247 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1248 			    sizeof(struct in_addr));
1249 			/*
1250 			 * Then copy rest of options back
1251 			 * to close up the deleted entry.
1252 			 */
1253 			memmove(&cp[IPOPT_OFFSET+1],
1254                             (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1255 			    (unsigned)cnt + sizeof(struct in_addr));
1256 			break;
1257 		}
1258 	}
1259 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1260 		goto bad;
1261 	*pcbopt = m;
1262 	return (0);
1263 
1264 bad:
1265 	(void)m_free(m);
1266 	return (EINVAL);
1267 }
1268 
1269 /*
1270  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1271  */
1272 static struct ifnet *
1273 ip_multicast_if(a, ifindexp)
1274 	struct in_addr *a;
1275 	int *ifindexp;
1276 {
1277 	int ifindex;
1278 	struct ifnet *ifp;
1279 
1280 	if (ifindexp)
1281 		*ifindexp = 0;
1282 	if (ntohl(a->s_addr) >> 24 == 0) {
1283 		ifindex = ntohl(a->s_addr) & 0xffffff;
1284 		if (ifindex < 0 || if_index < ifindex)
1285 			return NULL;
1286 		ifp = ifindex2ifnet[ifindex];
1287 		if (ifindexp)
1288 			*ifindexp = ifindex;
1289 	} else {
1290 		INADDR_TO_IFP(*a, ifp);
1291 	}
1292 	return ifp;
1293 }
1294 
1295 /*
1296  * Set the IP multicast options in response to user setsockopt().
1297  */
1298 int
1299 ip_setmoptions(optname, imop, m)
1300 	int optname;
1301 	struct ip_moptions **imop;
1302 	struct mbuf *m;
1303 {
1304 	int error = 0;
1305 	u_char loop;
1306 	int i;
1307 	struct in_addr addr;
1308 	struct ip_mreq *mreq;
1309 	struct ifnet *ifp;
1310 	struct ip_moptions *imo = *imop;
1311 	struct route ro;
1312 	struct sockaddr_in *dst;
1313 	int ifindex;
1314 
1315 	if (imo == NULL) {
1316 		/*
1317 		 * No multicast option buffer attached to the pcb;
1318 		 * allocate one and initialize to default values.
1319 		 */
1320 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1321 		    M_WAITOK);
1322 
1323 		if (imo == NULL)
1324 			return (ENOBUFS);
1325 		*imop = imo;
1326 		imo->imo_multicast_ifp = NULL;
1327 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1328 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1329 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1330 		imo->imo_num_memberships = 0;
1331 	}
1332 
1333 	switch (optname) {
1334 
1335 	case IP_MULTICAST_IF:
1336 		/*
1337 		 * Select the interface for outgoing multicast packets.
1338 		 */
1339 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1340 			error = EINVAL;
1341 			break;
1342 		}
1343 		addr = *(mtod(m, struct in_addr *));
1344 		/*
1345 		 * INADDR_ANY is used to remove a previous selection.
1346 		 * When no interface is selected, a default one is
1347 		 * chosen every time a multicast packet is sent.
1348 		 */
1349 		if (in_nullhost(addr)) {
1350 			imo->imo_multicast_ifp = NULL;
1351 			break;
1352 		}
1353 		/*
1354 		 * The selected interface is identified by its local
1355 		 * IP address.  Find the interface and confirm that
1356 		 * it supports multicasting.
1357 		 */
1358 		ifp = ip_multicast_if(&addr, &ifindex);
1359 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1360 			error = EADDRNOTAVAIL;
1361 			break;
1362 		}
1363 		imo->imo_multicast_ifp = ifp;
1364 		if (ifindex)
1365 			imo->imo_multicast_addr = addr;
1366 		else
1367 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1368 		break;
1369 
1370 	case IP_MULTICAST_TTL:
1371 		/*
1372 		 * Set the IP time-to-live for outgoing multicast packets.
1373 		 */
1374 		if (m == NULL || m->m_len != 1) {
1375 			error = EINVAL;
1376 			break;
1377 		}
1378 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1379 		break;
1380 
1381 	case IP_MULTICAST_LOOP:
1382 		/*
1383 		 * Set the loopback flag for outgoing multicast packets.
1384 		 * Must be zero or one.
1385 		 */
1386 		if (m == NULL || m->m_len != 1 ||
1387 		   (loop = *(mtod(m, u_char *))) > 1) {
1388 			error = EINVAL;
1389 			break;
1390 		}
1391 		imo->imo_multicast_loop = loop;
1392 		break;
1393 
1394 	case IP_ADD_MEMBERSHIP:
1395 		/*
1396 		 * Add a multicast group membership.
1397 		 * Group must be a valid IP multicast address.
1398 		 */
1399 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1400 			error = EINVAL;
1401 			break;
1402 		}
1403 		mreq = mtod(m, struct ip_mreq *);
1404 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1405 			error = EINVAL;
1406 			break;
1407 		}
1408 		/*
1409 		 * If no interface address was provided, use the interface of
1410 		 * the route to the given multicast address.
1411 		 */
1412 		if (in_nullhost(mreq->imr_interface)) {
1413 			bzero((caddr_t)&ro, sizeof(ro));
1414 			ro.ro_rt = NULL;
1415 			dst = satosin(&ro.ro_dst);
1416 			dst->sin_len = sizeof(*dst);
1417 			dst->sin_family = AF_INET;
1418 			dst->sin_addr = mreq->imr_multiaddr;
1419 			rtalloc(&ro);
1420 			if (ro.ro_rt == NULL) {
1421 				error = EADDRNOTAVAIL;
1422 				break;
1423 			}
1424 			ifp = ro.ro_rt->rt_ifp;
1425 			rtfree(ro.ro_rt);
1426 		} else {
1427 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1428 		}
1429 		/*
1430 		 * See if we found an interface, and confirm that it
1431 		 * supports multicast.
1432 		 */
1433 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1434 			error = EADDRNOTAVAIL;
1435 			break;
1436 		}
1437 		/*
1438 		 * See if the membership already exists or if all the
1439 		 * membership slots are full.
1440 		 */
1441 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1442 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1443 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1444 				      mreq->imr_multiaddr))
1445 				break;
1446 		}
1447 		if (i < imo->imo_num_memberships) {
1448 			error = EADDRINUSE;
1449 			break;
1450 		}
1451 		if (i == IP_MAX_MEMBERSHIPS) {
1452 			error = ETOOMANYREFS;
1453 			break;
1454 		}
1455 		/*
1456 		 * Everything looks good; add a new record to the multicast
1457 		 * address list for the given interface.
1458 		 */
1459 		if ((imo->imo_membership[i] =
1460 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1461 			error = ENOBUFS;
1462 			break;
1463 		}
1464 		++imo->imo_num_memberships;
1465 		break;
1466 
1467 	case IP_DROP_MEMBERSHIP:
1468 		/*
1469 		 * Drop a multicast group membership.
1470 		 * Group must be a valid IP multicast address.
1471 		 */
1472 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1473 			error = EINVAL;
1474 			break;
1475 		}
1476 		mreq = mtod(m, struct ip_mreq *);
1477 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1478 			error = EINVAL;
1479 			break;
1480 		}
1481 		/*
1482 		 * If an interface address was specified, get a pointer
1483 		 * to its ifnet structure.
1484 		 */
1485 		if (in_nullhost(mreq->imr_interface))
1486 			ifp = NULL;
1487 		else {
1488 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1489 			if (ifp == NULL) {
1490 				error = EADDRNOTAVAIL;
1491 				break;
1492 			}
1493 		}
1494 		/*
1495 		 * Find the membership in the membership array.
1496 		 */
1497 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1498 			if ((ifp == NULL ||
1499 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1500 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1501 				       mreq->imr_multiaddr))
1502 				break;
1503 		}
1504 		if (i == imo->imo_num_memberships) {
1505 			error = EADDRNOTAVAIL;
1506 			break;
1507 		}
1508 		/*
1509 		 * Give up the multicast address record to which the
1510 		 * membership points.
1511 		 */
1512 		in_delmulti(imo->imo_membership[i]);
1513 		/*
1514 		 * Remove the gap in the membership array.
1515 		 */
1516 		for (++i; i < imo->imo_num_memberships; ++i)
1517 			imo->imo_membership[i-1] = imo->imo_membership[i];
1518 		--imo->imo_num_memberships;
1519 		break;
1520 
1521 	default:
1522 		error = EOPNOTSUPP;
1523 		break;
1524 	}
1525 
1526 	/*
1527 	 * If all options have default values, no need to keep the mbuf.
1528 	 */
1529 	if (imo->imo_multicast_ifp == NULL &&
1530 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1531 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1532 	    imo->imo_num_memberships == 0) {
1533 		free(*imop, M_IPMOPTS);
1534 		*imop = NULL;
1535 	}
1536 
1537 	return (error);
1538 }
1539 
1540 /*
1541  * Return the IP multicast options in response to user getsockopt().
1542  */
1543 int
1544 ip_getmoptions(optname, imo, mp)
1545 	int optname;
1546 	struct ip_moptions *imo;
1547 	struct mbuf **mp;
1548 {
1549 	u_char *ttl;
1550 	u_char *loop;
1551 	struct in_addr *addr;
1552 	struct in_ifaddr *ia;
1553 
1554 	*mp = m_get(M_WAIT, MT_SOOPTS);
1555 
1556 	switch (optname) {
1557 
1558 	case IP_MULTICAST_IF:
1559 		addr = mtod(*mp, struct in_addr *);
1560 		(*mp)->m_len = sizeof(struct in_addr);
1561 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1562 			*addr = zeroin_addr;
1563 		else if (imo->imo_multicast_addr.s_addr) {
1564 			/* return the value user has set */
1565 			*addr = imo->imo_multicast_addr;
1566 		} else {
1567 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1568 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1569 		}
1570 		return (0);
1571 
1572 	case IP_MULTICAST_TTL:
1573 		ttl = mtod(*mp, u_char *);
1574 		(*mp)->m_len = 1;
1575 		*ttl = imo ? imo->imo_multicast_ttl
1576 			   : IP_DEFAULT_MULTICAST_TTL;
1577 		return (0);
1578 
1579 	case IP_MULTICAST_LOOP:
1580 		loop = mtod(*mp, u_char *);
1581 		(*mp)->m_len = 1;
1582 		*loop = imo ? imo->imo_multicast_loop
1583 			    : IP_DEFAULT_MULTICAST_LOOP;
1584 		return (0);
1585 
1586 	default:
1587 		return (EOPNOTSUPP);
1588 	}
1589 }
1590 
1591 /*
1592  * Discard the IP multicast options.
1593  */
1594 void
1595 ip_freemoptions(imo)
1596 	struct ip_moptions *imo;
1597 {
1598 	int i;
1599 
1600 	if (imo != NULL) {
1601 		for (i = 0; i < imo->imo_num_memberships; ++i)
1602 			in_delmulti(imo->imo_membership[i]);
1603 		free(imo, M_IPMOPTS);
1604 	}
1605 }
1606 
1607 /*
1608  * Routine called from ip_output() to loop back a copy of an IP multicast
1609  * packet to the input queue of a specified interface.  Note that this
1610  * calls the output routine of the loopback "driver", but with an interface
1611  * pointer that might NOT be &loif -- easier than replicating that code here.
1612  */
1613 static void
1614 ip_mloopback(ifp, m, dst)
1615 	struct ifnet *ifp;
1616 	struct mbuf *m;
1617 	struct sockaddr_in *dst;
1618 {
1619 	struct ip *ip;
1620 	struct mbuf *copym;
1621 
1622 	copym = m_copy(m, 0, M_COPYALL);
1623 	if (copym != NULL
1624 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1625 		copym = m_pullup(copym, sizeof(struct ip));
1626 	if (copym != NULL) {
1627 		/*
1628 		 * We don't bother to fragment if the IP length is greater
1629 		 * than the interface's MTU.  Can this possibly matter?
1630 		 */
1631 		ip = mtod(copym, struct ip *);
1632 		HTONS(ip->ip_len);
1633 		HTONS(ip->ip_off);
1634 		ip->ip_sum = 0;
1635 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1636 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1637 	}
1638 }
1639