xref: /netbsd-src/sys/netinet/ip_output.c (revision 001c68bd94f75ce9270b69227c4199fbf34ee396)
1 /*	$NetBSD: ip_output.c,v 1.107 2003/06/30 02:08:28 itojun Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1990, 1993
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.107 2003/06/30 02:08:28 itojun Exp $");
106 
107 #include "opt_pfil_hooks.h"
108 #include "opt_ipsec.h"
109 #include "opt_mrouting.h"
110 
111 #include <sys/param.h>
112 #include <sys/malloc.h>
113 #include <sys/mbuf.h>
114 #include <sys/errno.h>
115 #include <sys/protosw.h>
116 #include <sys/socket.h>
117 #include <sys/socketvar.h>
118 #include <sys/systm.h>
119 #include <sys/proc.h>
120 
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip_var.h>
131 
132 #ifdef MROUTING
133 #include <netinet/ip_mroute.h>
134 #endif
135 
136 #include <machine/stdarg.h>
137 
138 #ifdef IPSEC
139 #include <netinet6/ipsec.h>
140 #include <netkey/key.h>
141 #include <netkey/key_debug.h>
142 #endif /*IPSEC*/
143 
144 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
145 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
146 static void ip_mloopback
147 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *));
148 
149 #ifdef PFIL_HOOKS
150 extern struct pfil_head inet_pfil_hook;			/* XXX */
151 #endif
152 
153 /*
154  * IP output.  The packet in mbuf chain m contains a skeletal IP
155  * header (with len, off, ttl, proto, tos, src, dst).
156  * The mbuf chain containing the packet will be freed.
157  * The mbuf opt, if present, will not be freed.
158  */
159 int
160 #if __STDC__
161 ip_output(struct mbuf *m0, ...)
162 #else
163 ip_output(m0, va_alist)
164 	struct mbuf *m0;
165 	va_dcl
166 #endif
167 {
168 	struct ip *ip, *mhip;
169 	struct ifnet *ifp;
170 	struct mbuf *m = m0;
171 	int hlen = sizeof (struct ip);
172 	int len, off, error = 0;
173 	struct route iproute;
174 	struct sockaddr_in *dst;
175 	struct in_ifaddr *ia;
176 	struct mbuf *opt;
177 	struct route *ro;
178 	int flags, sw_csum;
179 	int *mtu_p;
180 	u_long 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 	len = 0;
190 	va_start(ap, m0);
191 	opt = va_arg(ap, struct mbuf *);
192 	ro = va_arg(ap, struct route *);
193 	flags = va_arg(ap, int);
194 	imo = va_arg(ap, struct ip_moptions *);
195 	if (flags & IP_RETURNMTU)
196 		mtu_p = va_arg(ap, int *);
197 	else
198 		mtu_p = NULL;
199 	va_end(ap);
200 
201 	MCLAIM(m, &ip_tx_mowner);
202 #ifdef IPSEC
203 	so = ipsec_getsocket(m);
204 	(void)ipsec_setsocket(m, NULL);
205 #endif /*IPSEC*/
206 
207 #ifdef	DIAGNOSTIC
208 	if ((m->m_flags & M_PKTHDR) == 0)
209 		panic("ip_output no HDR");
210 #endif
211 	if (opt) {
212 		m = ip_insertoptions(m, opt, &len);
213 		if (len >= sizeof(struct ip))
214 			hlen = len;
215 	}
216 	ip = mtod(m, struct ip *);
217 	/*
218 	 * Fill in IP header.
219 	 */
220 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
221 		ip->ip_v = IPVERSION;
222 		ip->ip_off = htons(0);
223 		ip->ip_id = htons(ip_id++);
224 		ip->ip_hl = hlen >> 2;
225 		ipstat.ips_localout++;
226 	} else {
227 		hlen = ip->ip_hl << 2;
228 	}
229 	/*
230 	 * Route packet.
231 	 */
232 	if (ro == 0) {
233 		ro = &iproute;
234 		bzero((caddr_t)ro, sizeof (*ro));
235 	}
236 	dst = satosin(&ro->ro_dst);
237 	/*
238 	 * If there is a cached route,
239 	 * check that it is to the same destination
240 	 * and is still up.  If not, free it and try again.
241 	 * The address family should also be checked in case of sharing the
242 	 * cache with IPv6.
243 	 */
244 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
245 	    dst->sin_family != AF_INET ||
246 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
247 		RTFREE(ro->ro_rt);
248 		ro->ro_rt = (struct rtentry *)0;
249 	}
250 	if (ro->ro_rt == 0) {
251 		bzero(dst, sizeof(*dst));
252 		dst->sin_family = AF_INET;
253 		dst->sin_len = sizeof(*dst);
254 		dst->sin_addr = ip->ip_dst;
255 	}
256 	/*
257 	 * If routing to interface only,
258 	 * short circuit routing lookup.
259 	 */
260 	if (flags & IP_ROUTETOIF) {
261 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
262 			ipstat.ips_noroute++;
263 			error = ENETUNREACH;
264 			goto bad;
265 		}
266 		ifp = ia->ia_ifp;
267 		mtu = ifp->if_mtu;
268 		ip->ip_ttl = 1;
269 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
270 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
271 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
272 		ifp = imo->imo_multicast_ifp;
273 		mtu = ifp->if_mtu;
274 		IFP_TO_IA(ifp, ia);
275 	} else {
276 		if (ro->ro_rt == 0)
277 			rtalloc(ro);
278 		if (ro->ro_rt == 0) {
279 			ipstat.ips_noroute++;
280 			error = EHOSTUNREACH;
281 			goto bad;
282 		}
283 		ia = ifatoia(ro->ro_rt->rt_ifa);
284 		ifp = ro->ro_rt->rt_ifp;
285 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
286 			mtu = ifp->if_mtu;
287 		ro->ro_rt->rt_use++;
288 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
289 			dst = satosin(ro->ro_rt->rt_gateway);
290 	}
291 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
292 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
293 		struct in_multi *inm;
294 
295 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
296 			M_BCAST : M_MCAST;
297 		/*
298 		 * IP destination address is multicast.  Make sure "dst"
299 		 * still points to the address in "ro".  (It may have been
300 		 * changed to point to a gateway address, above.)
301 		 */
302 		dst = satosin(&ro->ro_dst);
303 		/*
304 		 * See if the caller provided any multicast options
305 		 */
306 		if (imo != NULL)
307 			ip->ip_ttl = imo->imo_multicast_ttl;
308 		else
309 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
310 
311 		/*
312 		 * if we don't know the outgoing ifp yet, we can't generate
313 		 * output
314 		 */
315 		if (!ifp) {
316 			ipstat.ips_noroute++;
317 			error = ENETUNREACH;
318 			goto bad;
319 		}
320 
321 		/*
322 		 * If the packet is multicast or broadcast, confirm that
323 		 * the outgoing interface can transmit it.
324 		 */
325 		if (((m->m_flags & M_MCAST) &&
326 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
327 		    ((m->m_flags & M_BCAST) &&
328 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
329 			ipstat.ips_noroute++;
330 			error = ENETUNREACH;
331 			goto bad;
332 		}
333 		/*
334 		 * If source address not specified yet, use an address
335 		 * of outgoing interface.
336 		 */
337 		if (in_nullhost(ip->ip_src)) {
338 			struct in_ifaddr *ia;
339 
340 			IFP_TO_IA(ifp, ia);
341 			if (!ia) {
342 				error = EADDRNOTAVAIL;
343 				goto bad;
344 			}
345 			ip->ip_src = ia->ia_addr.sin_addr;
346 		}
347 
348 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
349 		if (inm != NULL &&
350 		   (imo == NULL || imo->imo_multicast_loop)) {
351 			/*
352 			 * If we belong to the destination multicast group
353 			 * on the outgoing interface, and the caller did not
354 			 * forbid loopback, loop back a copy.
355 			 */
356 			ip_mloopback(ifp, m, dst);
357 		}
358 #ifdef MROUTING
359 		else {
360 			/*
361 			 * If we are acting as a multicast router, perform
362 			 * multicast forwarding as if the packet had just
363 			 * arrived on the interface to which we are about
364 			 * to send.  The multicast forwarding function
365 			 * recursively calls this function, using the
366 			 * IP_FORWARDING flag to prevent infinite recursion.
367 			 *
368 			 * Multicasts that are looped back by ip_mloopback(),
369 			 * above, will be forwarded by the ip_input() routine,
370 			 * if necessary.
371 			 */
372 			extern struct socket *ip_mrouter;
373 
374 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
375 				if (ip_mforward(m, ifp) != 0) {
376 					m_freem(m);
377 					goto done;
378 				}
379 			}
380 		}
381 #endif
382 		/*
383 		 * Multicasts with a time-to-live of zero may be looped-
384 		 * back, above, but must not be transmitted on a network.
385 		 * Also, multicasts addressed to the loopback interface
386 		 * are not sent -- the above call to ip_mloopback() will
387 		 * loop back a copy if this host actually belongs to the
388 		 * destination group on the loopback interface.
389 		 */
390 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
391 			m_freem(m);
392 			goto done;
393 		}
394 
395 		goto sendit;
396 	}
397 #ifndef notdef
398 	/*
399 	 * If source address not specified yet, use address
400 	 * of outgoing interface.
401 	 */
402 	if (in_nullhost(ip->ip_src))
403 		ip->ip_src = ia->ia_addr.sin_addr;
404 #endif
405 
406 	/*
407 	 * packets with Class-D address as source are not valid per
408 	 * RFC 1112
409 	 */
410 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
411 		ipstat.ips_odropped++;
412 		error = EADDRNOTAVAIL;
413 		goto bad;
414 	}
415 
416 	/*
417 	 * Look for broadcast address and
418 	 * and verify user is allowed to send
419 	 * such a packet.
420 	 */
421 	if (in_broadcast(dst->sin_addr, ifp)) {
422 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
423 			error = EADDRNOTAVAIL;
424 			goto bad;
425 		}
426 		if ((flags & IP_ALLOWBROADCAST) == 0) {
427 			error = EACCES;
428 			goto bad;
429 		}
430 		/* don't allow broadcast messages to be fragmented */
431 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
432 			error = EMSGSIZE;
433 			goto bad;
434 		}
435 		m->m_flags |= M_BCAST;
436 	} else
437 		m->m_flags &= ~M_BCAST;
438 
439 sendit:
440 	/*
441 	 * If we're doing Path MTU Discovery, we need to set DF unless
442 	 * the route's MTU is locked.
443 	 */
444 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
445 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
446 		ip->ip_off |= htons(IP_DF);
447 
448 	/* Remember the current ip_len */
449 	ip_len = ntohs(ip->ip_len);
450 
451 #ifdef IPSEC
452 	/* get SP for this packet */
453 	if (so == NULL)
454 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
455 	else
456 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
457 
458 	if (sp == NULL) {
459 		ipsecstat.out_inval++;
460 		goto bad;
461 	}
462 
463 	error = 0;
464 
465 	/* check policy */
466 	switch (sp->policy) {
467 	case IPSEC_POLICY_DISCARD:
468 		/*
469 		 * This packet is just discarded.
470 		 */
471 		ipsecstat.out_polvio++;
472 		goto bad;
473 
474 	case IPSEC_POLICY_BYPASS:
475 	case IPSEC_POLICY_NONE:
476 		/* no need to do IPsec. */
477 		goto skip_ipsec;
478 
479 	case IPSEC_POLICY_IPSEC:
480 		if (sp->req == NULL) {
481 			/* XXX should be panic ? */
482 			printf("ip_output: No IPsec request specified.\n");
483 			error = EINVAL;
484 			goto bad;
485 		}
486 		break;
487 
488 	case IPSEC_POLICY_ENTRUST:
489 	default:
490 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
491 	}
492 
493 	/*
494 	 * ipsec4_output() expects ip_len and ip_off in network
495 	 * order.  They have been set to network order above.
496 	 */
497 
498     {
499 	struct ipsec_output_state state;
500 	bzero(&state, sizeof(state));
501 	state.m = m;
502 	if (flags & IP_ROUTETOIF) {
503 		state.ro = &iproute;
504 		bzero(&iproute, sizeof(iproute));
505 	} else
506 		state.ro = ro;
507 	state.dst = (struct sockaddr *)dst;
508 
509 	/*
510 	 * We can't defer the checksum of payload data if
511 	 * we're about to encrypt/authenticate it.
512 	 *
513 	 * XXX When we support crypto offloading functions of
514 	 * XXX network interfaces, we need to reconsider this,
515 	 * XXX since it's likely that they'll support checksumming,
516 	 * XXX as well.
517 	 */
518 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
519 		in_delayed_cksum(m);
520 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
521 	}
522 
523 	error = ipsec4_output(&state, sp, flags);
524 
525 	m = state.m;
526 	if (flags & IP_ROUTETOIF) {
527 		/*
528 		 * if we have tunnel mode SA, we may need to ignore
529 		 * IP_ROUTETOIF.
530 		 */
531 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
532 			flags &= ~IP_ROUTETOIF;
533 			ro = state.ro;
534 		}
535 	} else
536 		ro = state.ro;
537 	dst = (struct sockaddr_in *)state.dst;
538 	if (error) {
539 		/* mbuf is already reclaimed in ipsec4_output. */
540 		m0 = NULL;
541 		switch (error) {
542 		case EHOSTUNREACH:
543 		case ENETUNREACH:
544 		case EMSGSIZE:
545 		case ENOBUFS:
546 		case ENOMEM:
547 			break;
548 		default:
549 			printf("ip4_output (ipsec): error code %d\n", error);
550 			/*fall through*/
551 		case ENOENT:
552 			/* don't show these error codes to the user */
553 			error = 0;
554 			break;
555 		}
556 		goto bad;
557 	}
558 
559 	/* be sure to update variables that are affected by ipsec4_output() */
560 	ip = mtod(m, struct ip *);
561 	hlen = ip->ip_hl << 2;
562 	ip_len = ntohs(ip->ip_len);
563 
564 	if (ro->ro_rt == NULL) {
565 		if ((flags & IP_ROUTETOIF) == 0) {
566 			printf("ip_output: "
567 				"can't update route after IPsec processing\n");
568 			error = EHOSTUNREACH;	/*XXX*/
569 			goto bad;
570 		}
571 	} else {
572 		/* nobody uses ia beyond here */
573 		if (state.encap)
574 			ifp = ro->ro_rt->rt_ifp;
575 	}
576     }
577 
578 skip_ipsec:
579 #endif /*IPSEC*/
580 
581 #ifdef PFIL_HOOKS
582 	/*
583 	 * Run through list of hooks for output packets.
584 	 */
585 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
586 		goto done;
587 	if (m == NULL)
588 		goto done;
589 
590 	ip = mtod(m, struct ip *);
591 	hlen = ip->ip_hl << 2;
592 #endif /* PFIL_HOOKS */
593 
594 	m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
595 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
596 	/*
597 	 * If small enough for mtu of path, can just send directly.
598 	 */
599 	if (ip_len <= mtu) {
600 #if IFA_STATS
601 		/*
602 		 * search for the source address structure to
603 		 * maintain output statistics.
604 		 */
605 		INADDR_TO_IA(ip->ip_src, ia);
606 		if (ia)
607 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
608 #endif
609 		/*
610 		 * Always initialize the sum to 0!  Some HW assisted
611 		 * checksumming requires this.
612 		 */
613 		ip->ip_sum = 0;
614 
615 		/*
616 		 * Perform any checksums that the hardware can't do
617 		 * for us.
618 		 *
619 		 * XXX Does any hardware require the {th,uh}_sum
620 		 * XXX fields to be 0?
621 		 */
622 		if (sw_csum & M_CSUM_IPv4) {
623 			ip->ip_sum = in_cksum(m, hlen);
624 			m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
625 		}
626 		if (sw_csum & (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 #ifdef IPSEC
632 		/* clean ipsec history once it goes out of the node */
633 		ipsec_delaux(m);
634 #endif
635 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
636 		goto done;
637 	}
638 
639 	/*
640 	 * We can't use HW checksumming if we're about to
641 	 * to fragment the packet.
642 	 *
643 	 * XXX Some hardware can do this.
644 	 */
645 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
646 		in_delayed_cksum(m);
647 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
648 	}
649 
650 	/*
651 	 * Too large for interface; fragment if possible.
652 	 * Must be able to put at least 8 bytes per fragment.
653 	 */
654 	if (ntohs(ip->ip_off) & IP_DF) {
655 		if (flags & IP_RETURNMTU)
656 			*mtu_p = mtu;
657 		error = EMSGSIZE;
658 		ipstat.ips_cantfrag++;
659 		goto bad;
660 	}
661 	len = (mtu - hlen) &~ 7;
662 	if (len < 8) {
663 		error = EMSGSIZE;
664 		goto bad;
665 	}
666 
667     {
668 	int mhlen, firstlen = len;
669 	struct mbuf **mnext = &m->m_nextpkt;
670 	int fragments = 0;
671 	int s;
672 
673 	/*
674 	 * Loop through length of segment after first fragment,
675 	 * make new header and copy data of each part and link onto chain.
676 	 */
677 	m0 = m;
678 	mhlen = sizeof (struct ip);
679 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
680 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
681 		if (m == 0) {
682 			error = ENOBUFS;
683 			ipstat.ips_odropped++;
684 			goto sendorfree;
685 		}
686 		MCLAIM(m, m0->m_owner);
687 		*mnext = m;
688 		mnext = &m->m_nextpkt;
689 		m->m_data += max_linkhdr;
690 		mhip = mtod(m, struct ip *);
691 		*mhip = *ip;
692 		/* we must inherit MCAST and BCAST flags */
693 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
694 		if (hlen > sizeof (struct ip)) {
695 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
696 			mhip->ip_hl = mhlen >> 2;
697 		}
698 		m->m_len = mhlen;
699 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
700 		if (ip->ip_off & IP_MF)
701 			mhip->ip_off |= IP_MF;
702 		if (off + len >= ntohs(ip->ip_len))
703 			len = ntohs(ip->ip_len) - off;
704 		else
705 			mhip->ip_off |= IP_MF;
706 		HTONS(mhip->ip_off);
707 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
708 		m->m_next = m_copy(m0, off, len);
709 		if (m->m_next == 0) {
710 			error = ENOBUFS;	/* ??? */
711 			ipstat.ips_odropped++;
712 			goto sendorfree;
713 		}
714 		m->m_pkthdr.len = mhlen + len;
715 		m->m_pkthdr.rcvif = (struct ifnet *)0;
716 		mhip->ip_sum = 0;
717 		if (sw_csum & M_CSUM_IPv4) {
718 			mhip->ip_sum = in_cksum(m, mhlen);
719 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
720 		} else {
721 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
722 		}
723 		ipstat.ips_ofragments++;
724 		fragments++;
725 	}
726 	/*
727 	 * Update first fragment by trimming what's been copied out
728 	 * and updating header, then send each fragment (in order).
729 	 */
730 	m = m0;
731 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
732 	m->m_pkthdr.len = hlen + firstlen;
733 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
734 	ip->ip_off |= htons(IP_MF);
735 	ip->ip_sum = 0;
736 	if (sw_csum & M_CSUM_IPv4) {
737 		ip->ip_sum = in_cksum(m, hlen);
738 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
739 	} else {
740 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
741 	}
742 sendorfree:
743 	/*
744 	 * If there is no room for all the fragments, don't queue
745 	 * any of them.
746 	 */
747 	s = splnet();
748 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
749 		error = ENOBUFS;
750 	splx(s);
751 	for (m = m0; m; m = m0) {
752 		m0 = m->m_nextpkt;
753 		m->m_nextpkt = 0;
754 		if (error == 0) {
755 #if IFA_STATS
756 			/*
757 			 * search for the source address structure to
758 			 * maintain output statistics.
759 			 */
760 			INADDR_TO_IA(ip->ip_src, ia);
761 			if (ia) {
762 				ia->ia_ifa.ifa_data.ifad_outbytes +=
763 				    ntohs(ip->ip_len);
764 			}
765 #endif
766 #ifdef IPSEC
767 			/* clean ipsec history once it goes out of the node */
768 			ipsec_delaux(m);
769 #endif
770 			KASSERT((m->m_pkthdr.csum_flags &
771 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
772 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
773 			    ro->ro_rt);
774 		} else
775 			m_freem(m);
776 	}
777 
778 	if (error == 0)
779 		ipstat.ips_fragmented++;
780     }
781 done:
782 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
783 		RTFREE(ro->ro_rt);
784 		ro->ro_rt = 0;
785 	}
786 
787 #ifdef IPSEC
788 	if (sp != NULL) {
789 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
790 			printf("DP ip_output call free SP:%p\n", sp));
791 		key_freesp(sp);
792 	}
793 #endif /* IPSEC */
794 
795 	return (error);
796 bad:
797 	m_freem(m);
798 	goto done;
799 }
800 
801 /*
802  * Process a delayed payload checksum calculation.
803  */
804 void
805 in_delayed_cksum(struct mbuf *m)
806 {
807 	struct ip *ip;
808 	u_int16_t csum, offset;
809 
810 	ip = mtod(m, struct ip *);
811 	offset = ip->ip_hl << 2;
812 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
813 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
814 		csum = 0xffff;
815 
816 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
817 
818 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
819 		/* This happen when ip options were inserted
820 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
821 		    m->m_len, offset, ip->ip_p);
822 		 */
823 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
824 	} else
825 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
826 }
827 
828 /*
829  * Determine the maximum length of the options to be inserted;
830  * we would far rather allocate too much space rather than too little.
831  */
832 
833 u_int
834 ip_optlen(inp)
835 	struct inpcb *inp;
836 {
837 	struct mbuf *m = inp->inp_options;
838 
839 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
840 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
841 	else
842 		return 0;
843 }
844 
845 
846 /*
847  * Insert IP options into preformed packet.
848  * Adjust IP destination as required for IP source routing,
849  * as indicated by a non-zero in_addr at the start of the options.
850  */
851 static struct mbuf *
852 ip_insertoptions(m, opt, phlen)
853 	struct mbuf *m;
854 	struct mbuf *opt;
855 	int *phlen;
856 {
857 	struct ipoption *p = mtod(opt, struct ipoption *);
858 	struct mbuf *n;
859 	struct ip *ip = mtod(m, struct ip *);
860 	unsigned optlen;
861 
862 	optlen = opt->m_len - sizeof(p->ipopt_dst);
863 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
864 		return (m);		/* XXX should fail */
865 	if (!in_nullhost(p->ipopt_dst))
866 		ip->ip_dst = p->ipopt_dst;
867 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
868 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
869 		if (n == 0)
870 			return (m);
871 		MCLAIM(n, m->m_owner);
872 		M_COPY_PKTHDR(n, m);
873 		m->m_flags &= ~M_PKTHDR;
874 		m->m_len -= sizeof(struct ip);
875 		m->m_data += sizeof(struct ip);
876 		n->m_next = m;
877 		m = n;
878 		m->m_len = optlen + sizeof(struct ip);
879 		m->m_data += max_linkhdr;
880 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
881 	} else {
882 		m->m_data -= optlen;
883 		m->m_len += optlen;
884 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
885 	}
886 	m->m_pkthdr.len += optlen;
887 	ip = mtod(m, struct ip *);
888 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
889 	*phlen = sizeof(struct ip) + optlen;
890 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
891 	return (m);
892 }
893 
894 /*
895  * Copy options from ip to jp,
896  * omitting those not copied during fragmentation.
897  */
898 int
899 ip_optcopy(ip, jp)
900 	struct ip *ip, *jp;
901 {
902 	u_char *cp, *dp;
903 	int opt, optlen, cnt;
904 
905 	cp = (u_char *)(ip + 1);
906 	dp = (u_char *)(jp + 1);
907 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
908 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
909 		opt = cp[0];
910 		if (opt == IPOPT_EOL)
911 			break;
912 		if (opt == IPOPT_NOP) {
913 			/* Preserve for IP mcast tunnel's LSRR alignment. */
914 			*dp++ = IPOPT_NOP;
915 			optlen = 1;
916 			continue;
917 		}
918 #ifdef DIAGNOSTIC
919 		if (cnt < IPOPT_OLEN + sizeof(*cp))
920 			panic("malformed IPv4 option passed to ip_optcopy");
921 #endif
922 		optlen = cp[IPOPT_OLEN];
923 #ifdef DIAGNOSTIC
924 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
925 			panic("malformed IPv4 option passed to ip_optcopy");
926 #endif
927 		/* bogus lengths should have been caught by ip_dooptions */
928 		if (optlen > cnt)
929 			optlen = cnt;
930 		if (IPOPT_COPIED(opt)) {
931 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
932 			dp += optlen;
933 		}
934 	}
935 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
936 		*dp++ = IPOPT_EOL;
937 	return (optlen);
938 }
939 
940 /*
941  * IP socket option processing.
942  */
943 int
944 ip_ctloutput(op, so, level, optname, mp)
945 	int op;
946 	struct socket *so;
947 	int level, optname;
948 	struct mbuf **mp;
949 {
950 	struct inpcb *inp = sotoinpcb(so);
951 	struct mbuf *m = *mp;
952 	int optval = 0;
953 	int error = 0;
954 #ifdef IPSEC
955 #ifdef __NetBSD__
956 	struct proc *p = curproc;	/*XXX*/
957 #endif
958 #endif
959 
960 	if (level != IPPROTO_IP) {
961 		error = EINVAL;
962 		if (op == PRCO_SETOPT && *mp)
963 			(void) m_free(*mp);
964 	} else switch (op) {
965 
966 	case PRCO_SETOPT:
967 		switch (optname) {
968 		case IP_OPTIONS:
969 #ifdef notyet
970 		case IP_RETOPTS:
971 			return (ip_pcbopts(optname, &inp->inp_options, m));
972 #else
973 			return (ip_pcbopts(&inp->inp_options, m));
974 #endif
975 
976 		case IP_TOS:
977 		case IP_TTL:
978 		case IP_RECVOPTS:
979 		case IP_RECVRETOPTS:
980 		case IP_RECVDSTADDR:
981 		case IP_RECVIF:
982 			if (m == NULL || m->m_len != sizeof(int))
983 				error = EINVAL;
984 			else {
985 				optval = *mtod(m, int *);
986 				switch (optname) {
987 
988 				case IP_TOS:
989 					inp->inp_ip.ip_tos = optval;
990 					break;
991 
992 				case IP_TTL:
993 					inp->inp_ip.ip_ttl = optval;
994 					break;
995 #define	OPTSET(bit) \
996 	if (optval) \
997 		inp->inp_flags |= bit; \
998 	else \
999 		inp->inp_flags &= ~bit;
1000 
1001 				case IP_RECVOPTS:
1002 					OPTSET(INP_RECVOPTS);
1003 					break;
1004 
1005 				case IP_RECVRETOPTS:
1006 					OPTSET(INP_RECVRETOPTS);
1007 					break;
1008 
1009 				case IP_RECVDSTADDR:
1010 					OPTSET(INP_RECVDSTADDR);
1011 					break;
1012 
1013 				case IP_RECVIF:
1014 					OPTSET(INP_RECVIF);
1015 					break;
1016 				}
1017 			}
1018 			break;
1019 #undef OPTSET
1020 
1021 		case IP_MULTICAST_IF:
1022 		case IP_MULTICAST_TTL:
1023 		case IP_MULTICAST_LOOP:
1024 		case IP_ADD_MEMBERSHIP:
1025 		case IP_DROP_MEMBERSHIP:
1026 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1027 			break;
1028 
1029 		case IP_PORTRANGE:
1030 			if (m == 0 || m->m_len != sizeof(int))
1031 				error = EINVAL;
1032 			else {
1033 				optval = *mtod(m, int *);
1034 
1035 				switch (optval) {
1036 
1037 				case IP_PORTRANGE_DEFAULT:
1038 				case IP_PORTRANGE_HIGH:
1039 					inp->inp_flags &= ~(INP_LOWPORT);
1040 					break;
1041 
1042 				case IP_PORTRANGE_LOW:
1043 					inp->inp_flags |= INP_LOWPORT;
1044 					break;
1045 
1046 				default:
1047 					error = EINVAL;
1048 					break;
1049 				}
1050 			}
1051 			break;
1052 
1053 #ifdef IPSEC
1054 		case IP_IPSEC_POLICY:
1055 		{
1056 			caddr_t req = NULL;
1057 			size_t len = 0;
1058 			int priv = 0;
1059 
1060 #ifdef __NetBSD__
1061 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1062 				priv = 0;
1063 			else
1064 				priv = 1;
1065 #else
1066 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1067 #endif
1068 			if (m) {
1069 				req = mtod(m, caddr_t);
1070 				len = m->m_len;
1071 			}
1072 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1073 			break;
1074 		    }
1075 #endif /*IPSEC*/
1076 
1077 		default:
1078 			error = ENOPROTOOPT;
1079 			break;
1080 		}
1081 		if (m)
1082 			(void)m_free(m);
1083 		break;
1084 
1085 	case PRCO_GETOPT:
1086 		switch (optname) {
1087 		case IP_OPTIONS:
1088 		case IP_RETOPTS:
1089 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1090 			MCLAIM(m, so->so_mowner);
1091 			if (inp->inp_options) {
1092 				m->m_len = inp->inp_options->m_len;
1093 				bcopy(mtod(inp->inp_options, caddr_t),
1094 				    mtod(m, caddr_t), (unsigned)m->m_len);
1095 			} else
1096 				m->m_len = 0;
1097 			break;
1098 
1099 		case IP_TOS:
1100 		case IP_TTL:
1101 		case IP_RECVOPTS:
1102 		case IP_RECVRETOPTS:
1103 		case IP_RECVDSTADDR:
1104 		case IP_RECVIF:
1105 		case IP_ERRORMTU:
1106 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1107 			MCLAIM(m, so->so_mowner);
1108 			m->m_len = sizeof(int);
1109 			switch (optname) {
1110 
1111 			case IP_TOS:
1112 				optval = inp->inp_ip.ip_tos;
1113 				break;
1114 
1115 			case IP_TTL:
1116 				optval = inp->inp_ip.ip_ttl;
1117 				break;
1118 
1119 			case IP_ERRORMTU:
1120 				optval = inp->inp_errormtu;
1121 				break;
1122 
1123 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1124 
1125 			case IP_RECVOPTS:
1126 				optval = OPTBIT(INP_RECVOPTS);
1127 				break;
1128 
1129 			case IP_RECVRETOPTS:
1130 				optval = OPTBIT(INP_RECVRETOPTS);
1131 				break;
1132 
1133 			case IP_RECVDSTADDR:
1134 				optval = OPTBIT(INP_RECVDSTADDR);
1135 				break;
1136 
1137 			case IP_RECVIF:
1138 				optval = OPTBIT(INP_RECVIF);
1139 				break;
1140 			}
1141 			*mtod(m, int *) = optval;
1142 			break;
1143 
1144 #ifdef IPSEC
1145 		case IP_IPSEC_POLICY:
1146 		{
1147 			caddr_t req = NULL;
1148 			size_t len = 0;
1149 
1150 			if (m) {
1151 				req = mtod(m, caddr_t);
1152 				len = m->m_len;
1153 			}
1154 			error = ipsec4_get_policy(inp, req, len, mp);
1155 			break;
1156 		}
1157 #endif /*IPSEC*/
1158 
1159 		case IP_MULTICAST_IF:
1160 		case IP_MULTICAST_TTL:
1161 		case IP_MULTICAST_LOOP:
1162 		case IP_ADD_MEMBERSHIP:
1163 		case IP_DROP_MEMBERSHIP:
1164 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1165 			if (*mp)
1166 				MCLAIM(*mp, so->so_mowner);
1167 			break;
1168 
1169 		case IP_PORTRANGE:
1170 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1171 			MCLAIM(m, so->so_mowner);
1172 			m->m_len = sizeof(int);
1173 
1174 			if (inp->inp_flags & INP_LOWPORT)
1175 				optval = IP_PORTRANGE_LOW;
1176 			else
1177 				optval = IP_PORTRANGE_DEFAULT;
1178 
1179 			*mtod(m, int *) = optval;
1180 			break;
1181 
1182 		default:
1183 			error = ENOPROTOOPT;
1184 			break;
1185 		}
1186 		break;
1187 	}
1188 	return (error);
1189 }
1190 
1191 /*
1192  * Set up IP options in pcb for insertion in output packets.
1193  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1194  * with destination address if source routed.
1195  */
1196 int
1197 #ifdef notyet
1198 ip_pcbopts(optname, pcbopt, m)
1199 	int optname;
1200 #else
1201 ip_pcbopts(pcbopt, m)
1202 #endif
1203 	struct mbuf **pcbopt;
1204 	struct mbuf *m;
1205 {
1206 	int cnt, optlen;
1207 	u_char *cp;
1208 	u_char opt;
1209 
1210 	/* turn off any old options */
1211 	if (*pcbopt)
1212 		(void)m_free(*pcbopt);
1213 	*pcbopt = 0;
1214 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1215 		/*
1216 		 * Only turning off any previous options.
1217 		 */
1218 		if (m)
1219 			(void)m_free(m);
1220 		return (0);
1221 	}
1222 
1223 #ifndef	__vax__
1224 	if (m->m_len % sizeof(int32_t))
1225 		goto bad;
1226 #endif
1227 	/*
1228 	 * IP first-hop destination address will be stored before
1229 	 * actual options; move other options back
1230 	 * and clear it when none present.
1231 	 */
1232 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1233 		goto bad;
1234 	cnt = m->m_len;
1235 	m->m_len += sizeof(struct in_addr);
1236 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1237 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1238 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1239 
1240 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1241 		opt = cp[IPOPT_OPTVAL];
1242 		if (opt == IPOPT_EOL)
1243 			break;
1244 		if (opt == IPOPT_NOP)
1245 			optlen = 1;
1246 		else {
1247 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1248 				goto bad;
1249 			optlen = cp[IPOPT_OLEN];
1250 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1251 				goto bad;
1252 		}
1253 		switch (opt) {
1254 
1255 		default:
1256 			break;
1257 
1258 		case IPOPT_LSRR:
1259 		case IPOPT_SSRR:
1260 			/*
1261 			 * user process specifies route as:
1262 			 *	->A->B->C->D
1263 			 * D must be our final destination (but we can't
1264 			 * check that since we may not have connected yet).
1265 			 * A is first hop destination, which doesn't appear in
1266 			 * actual IP option, but is stored before the options.
1267 			 */
1268 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1269 				goto bad;
1270 			m->m_len -= sizeof(struct in_addr);
1271 			cnt -= sizeof(struct in_addr);
1272 			optlen -= sizeof(struct in_addr);
1273 			cp[IPOPT_OLEN] = optlen;
1274 			/*
1275 			 * Move first hop before start of options.
1276 			 */
1277 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1278 			    sizeof(struct in_addr));
1279 			/*
1280 			 * Then copy rest of options back
1281 			 * to close up the deleted entry.
1282 			 */
1283 			memmove(&cp[IPOPT_OFFSET+1],
1284 			    (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1285 			    (unsigned)cnt + sizeof(struct in_addr));
1286 			break;
1287 		}
1288 	}
1289 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1290 		goto bad;
1291 	*pcbopt = m;
1292 	return (0);
1293 
1294 bad:
1295 	(void)m_free(m);
1296 	return (EINVAL);
1297 }
1298 
1299 /*
1300  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1301  */
1302 static struct ifnet *
1303 ip_multicast_if(a, ifindexp)
1304 	struct in_addr *a;
1305 	int *ifindexp;
1306 {
1307 	int ifindex;
1308 	struct ifnet *ifp;
1309 
1310 	if (ifindexp)
1311 		*ifindexp = 0;
1312 	if (ntohl(a->s_addr) >> 24 == 0) {
1313 		ifindex = ntohl(a->s_addr) & 0xffffff;
1314 		if (ifindex < 0 || if_index < ifindex)
1315 			return NULL;
1316 		ifp = ifindex2ifnet[ifindex];
1317 		if (ifindexp)
1318 			*ifindexp = ifindex;
1319 	} else {
1320 		INADDR_TO_IFP(*a, ifp);
1321 	}
1322 	return ifp;
1323 }
1324 
1325 /*
1326  * Set the IP multicast options in response to user setsockopt().
1327  */
1328 int
1329 ip_setmoptions(optname, imop, m)
1330 	int optname;
1331 	struct ip_moptions **imop;
1332 	struct mbuf *m;
1333 {
1334 	int error = 0;
1335 	u_char loop;
1336 	int i;
1337 	struct in_addr addr;
1338 	struct ip_mreq *mreq;
1339 	struct ifnet *ifp;
1340 	struct ip_moptions *imo = *imop;
1341 	struct route ro;
1342 	struct sockaddr_in *dst;
1343 	int ifindex;
1344 
1345 	if (imo == NULL) {
1346 		/*
1347 		 * No multicast option buffer attached to the pcb;
1348 		 * allocate one and initialize to default values.
1349 		 */
1350 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1351 		    M_WAITOK);
1352 
1353 		if (imo == NULL)
1354 			return (ENOBUFS);
1355 		*imop = imo;
1356 		imo->imo_multicast_ifp = NULL;
1357 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1358 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1359 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1360 		imo->imo_num_memberships = 0;
1361 	}
1362 
1363 	switch (optname) {
1364 
1365 	case IP_MULTICAST_IF:
1366 		/*
1367 		 * Select the interface for outgoing multicast packets.
1368 		 */
1369 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1370 			error = EINVAL;
1371 			break;
1372 		}
1373 		addr = *(mtod(m, struct in_addr *));
1374 		/*
1375 		 * INADDR_ANY is used to remove a previous selection.
1376 		 * When no interface is selected, a default one is
1377 		 * chosen every time a multicast packet is sent.
1378 		 */
1379 		if (in_nullhost(addr)) {
1380 			imo->imo_multicast_ifp = NULL;
1381 			break;
1382 		}
1383 		/*
1384 		 * The selected interface is identified by its local
1385 		 * IP address.  Find the interface and confirm that
1386 		 * it supports multicasting.
1387 		 */
1388 		ifp = ip_multicast_if(&addr, &ifindex);
1389 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1390 			error = EADDRNOTAVAIL;
1391 			break;
1392 		}
1393 		imo->imo_multicast_ifp = ifp;
1394 		if (ifindex)
1395 			imo->imo_multicast_addr = addr;
1396 		else
1397 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1398 		break;
1399 
1400 	case IP_MULTICAST_TTL:
1401 		/*
1402 		 * Set the IP time-to-live for outgoing multicast packets.
1403 		 */
1404 		if (m == NULL || m->m_len != 1) {
1405 			error = EINVAL;
1406 			break;
1407 		}
1408 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1409 		break;
1410 
1411 	case IP_MULTICAST_LOOP:
1412 		/*
1413 		 * Set the loopback flag for outgoing multicast packets.
1414 		 * Must be zero or one.
1415 		 */
1416 		if (m == NULL || m->m_len != 1 ||
1417 		   (loop = *(mtod(m, u_char *))) > 1) {
1418 			error = EINVAL;
1419 			break;
1420 		}
1421 		imo->imo_multicast_loop = loop;
1422 		break;
1423 
1424 	case IP_ADD_MEMBERSHIP:
1425 		/*
1426 		 * Add a multicast group membership.
1427 		 * Group must be a valid IP multicast address.
1428 		 */
1429 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1430 			error = EINVAL;
1431 			break;
1432 		}
1433 		mreq = mtod(m, struct ip_mreq *);
1434 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1435 			error = EINVAL;
1436 			break;
1437 		}
1438 		/*
1439 		 * If no interface address was provided, use the interface of
1440 		 * the route to the given multicast address.
1441 		 */
1442 		if (in_nullhost(mreq->imr_interface)) {
1443 			bzero((caddr_t)&ro, sizeof(ro));
1444 			ro.ro_rt = NULL;
1445 			dst = satosin(&ro.ro_dst);
1446 			dst->sin_len = sizeof(*dst);
1447 			dst->sin_family = AF_INET;
1448 			dst->sin_addr = mreq->imr_multiaddr;
1449 			rtalloc(&ro);
1450 			if (ro.ro_rt == NULL) {
1451 				error = EADDRNOTAVAIL;
1452 				break;
1453 			}
1454 			ifp = ro.ro_rt->rt_ifp;
1455 			rtfree(ro.ro_rt);
1456 		} else {
1457 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1458 		}
1459 		/*
1460 		 * See if we found an interface, and confirm that it
1461 		 * supports multicast.
1462 		 */
1463 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1464 			error = EADDRNOTAVAIL;
1465 			break;
1466 		}
1467 		/*
1468 		 * See if the membership already exists or if all the
1469 		 * membership slots are full.
1470 		 */
1471 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1472 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1473 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1474 				      mreq->imr_multiaddr))
1475 				break;
1476 		}
1477 		if (i < imo->imo_num_memberships) {
1478 			error = EADDRINUSE;
1479 			break;
1480 		}
1481 		if (i == IP_MAX_MEMBERSHIPS) {
1482 			error = ETOOMANYREFS;
1483 			break;
1484 		}
1485 		/*
1486 		 * Everything looks good; add a new record to the multicast
1487 		 * address list for the given interface.
1488 		 */
1489 		if ((imo->imo_membership[i] =
1490 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1491 			error = ENOBUFS;
1492 			break;
1493 		}
1494 		++imo->imo_num_memberships;
1495 		break;
1496 
1497 	case IP_DROP_MEMBERSHIP:
1498 		/*
1499 		 * Drop a multicast group membership.
1500 		 * Group must be a valid IP multicast address.
1501 		 */
1502 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1503 			error = EINVAL;
1504 			break;
1505 		}
1506 		mreq = mtod(m, struct ip_mreq *);
1507 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1508 			error = EINVAL;
1509 			break;
1510 		}
1511 		/*
1512 		 * If an interface address was specified, get a pointer
1513 		 * to its ifnet structure.
1514 		 */
1515 		if (in_nullhost(mreq->imr_interface))
1516 			ifp = NULL;
1517 		else {
1518 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1519 			if (ifp == NULL) {
1520 				error = EADDRNOTAVAIL;
1521 				break;
1522 			}
1523 		}
1524 		/*
1525 		 * Find the membership in the membership array.
1526 		 */
1527 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1528 			if ((ifp == NULL ||
1529 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1530 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1531 				       mreq->imr_multiaddr))
1532 				break;
1533 		}
1534 		if (i == imo->imo_num_memberships) {
1535 			error = EADDRNOTAVAIL;
1536 			break;
1537 		}
1538 		/*
1539 		 * Give up the multicast address record to which the
1540 		 * membership points.
1541 		 */
1542 		in_delmulti(imo->imo_membership[i]);
1543 		/*
1544 		 * Remove the gap in the membership array.
1545 		 */
1546 		for (++i; i < imo->imo_num_memberships; ++i)
1547 			imo->imo_membership[i-1] = imo->imo_membership[i];
1548 		--imo->imo_num_memberships;
1549 		break;
1550 
1551 	default:
1552 		error = EOPNOTSUPP;
1553 		break;
1554 	}
1555 
1556 	/*
1557 	 * If all options have default values, no need to keep the mbuf.
1558 	 */
1559 	if (imo->imo_multicast_ifp == NULL &&
1560 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1561 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1562 	    imo->imo_num_memberships == 0) {
1563 		free(*imop, M_IPMOPTS);
1564 		*imop = NULL;
1565 	}
1566 
1567 	return (error);
1568 }
1569 
1570 /*
1571  * Return the IP multicast options in response to user getsockopt().
1572  */
1573 int
1574 ip_getmoptions(optname, imo, mp)
1575 	int optname;
1576 	struct ip_moptions *imo;
1577 	struct mbuf **mp;
1578 {
1579 	u_char *ttl;
1580 	u_char *loop;
1581 	struct in_addr *addr;
1582 	struct in_ifaddr *ia;
1583 
1584 	*mp = m_get(M_WAIT, MT_SOOPTS);
1585 
1586 	switch (optname) {
1587 
1588 	case IP_MULTICAST_IF:
1589 		addr = mtod(*mp, struct in_addr *);
1590 		(*mp)->m_len = sizeof(struct in_addr);
1591 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1592 			*addr = zeroin_addr;
1593 		else if (imo->imo_multicast_addr.s_addr) {
1594 			/* return the value user has set */
1595 			*addr = imo->imo_multicast_addr;
1596 		} else {
1597 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1598 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1599 		}
1600 		return (0);
1601 
1602 	case IP_MULTICAST_TTL:
1603 		ttl = mtod(*mp, u_char *);
1604 		(*mp)->m_len = 1;
1605 		*ttl = imo ? imo->imo_multicast_ttl
1606 			   : IP_DEFAULT_MULTICAST_TTL;
1607 		return (0);
1608 
1609 	case IP_MULTICAST_LOOP:
1610 		loop = mtod(*mp, u_char *);
1611 		(*mp)->m_len = 1;
1612 		*loop = imo ? imo->imo_multicast_loop
1613 			    : IP_DEFAULT_MULTICAST_LOOP;
1614 		return (0);
1615 
1616 	default:
1617 		return (EOPNOTSUPP);
1618 	}
1619 }
1620 
1621 /*
1622  * Discard the IP multicast options.
1623  */
1624 void
1625 ip_freemoptions(imo)
1626 	struct ip_moptions *imo;
1627 {
1628 	int i;
1629 
1630 	if (imo != NULL) {
1631 		for (i = 0; i < imo->imo_num_memberships; ++i)
1632 			in_delmulti(imo->imo_membership[i]);
1633 		free(imo, M_IPMOPTS);
1634 	}
1635 }
1636 
1637 /*
1638  * Routine called from ip_output() to loop back a copy of an IP multicast
1639  * packet to the input queue of a specified interface.  Note that this
1640  * calls the output routine of the loopback "driver", but with an interface
1641  * pointer that might NOT be &loif -- easier than replicating that code here.
1642  */
1643 static void
1644 ip_mloopback(ifp, m, dst)
1645 	struct ifnet *ifp;
1646 	struct mbuf *m;
1647 	struct sockaddr_in *dst;
1648 {
1649 	struct ip *ip;
1650 	struct mbuf *copym;
1651 
1652 	copym = m_copy(m, 0, M_COPYALL);
1653 	if (copym != NULL
1654 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1655 		copym = m_pullup(copym, sizeof(struct ip));
1656 	if (copym != NULL) {
1657 		/*
1658 		 * We don't bother to fragment if the IP length is greater
1659 		 * than the interface's MTU.  Can this possibly matter?
1660 		 */
1661 		ip = mtod(copym, struct ip *);
1662 
1663 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1664 			in_delayed_cksum(copym);
1665 			copym->m_pkthdr.csum_flags &=
1666 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1667 		}
1668 
1669 		ip->ip_sum = 0;
1670 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1671 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1672 	}
1673 }
1674