xref: /netbsd-src/sys/netinet/ip_output.c (revision aaf4ece63a859a04e37cf3a7229b5fab0157cc06)
1 /*	$NetBSD: ip_output.c,v 1.159 2005/12/11 12:24:57 christos 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. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.159 2005/12/11 12:24:57 christos Exp $");
102 
103 #include "opt_pfil_hooks.h"
104 #include "opt_inet.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 #ifdef FAST_IPSEC
116 #include <sys/domain.h>
117 #endif
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 #include <netinet/in_offload.h>
132 
133 #ifdef MROUTING
134 #include <netinet/ip_mroute.h>
135 #endif
136 
137 #include <machine/stdarg.h>
138 
139 #ifdef IPSEC
140 #include <netinet6/ipsec.h>
141 #include <netkey/key.h>
142 #include <netkey/key_debug.h>
143 #ifdef IPSEC_NAT_T
144 #include <netinet/udp.h>
145 #endif
146 #endif /*IPSEC*/
147 
148 #ifdef FAST_IPSEC
149 #include <netipsec/ipsec.h>
150 #include <netipsec/key.h>
151 #include <netipsec/xform.h>
152 #endif	/* FAST_IPSEC*/
153 
154 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
155 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
156 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
157 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int);
158 
159 #ifdef PFIL_HOOKS
160 extern struct pfil_head inet_pfil_hook;			/* XXX */
161 #endif
162 
163 int	ip_do_loopback_cksum = 0;
164 
165 #define	IN_NEED_CHECKSUM(ifp, csum_flags) \
166 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
167 	(((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \
168 	(((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \
169 	(((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum)))
170 
171 struct ip_tso_output_args {
172 	struct ifnet *ifp;
173 	struct sockaddr *sa;
174 	struct rtentry *rt;
175 };
176 
177 static int ip_tso_output_callback(void *, struct mbuf *);
178 static int ip_tso_output(struct ifnet *, struct mbuf *, struct sockaddr *,
179     struct rtentry *);
180 
181 static int
182 ip_tso_output_callback(void *vp, struct mbuf *m)
183 {
184 	struct ip_tso_output_args *args = vp;
185 	struct ifnet *ifp = args->ifp;
186 
187 	return (*ifp->if_output)(ifp, m, args->sa, args->rt);
188 }
189 
190 static int
191 ip_tso_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
192     struct rtentry *rt)
193 {
194 	struct ip_tso_output_args args;
195 
196 	args.ifp = ifp;
197 	args.sa = sa;
198 	args.rt = rt;
199 
200 	return tcp4_segment(m, ip_tso_output_callback, &args);
201 }
202 
203 /*
204  * IP output.  The packet in mbuf chain m contains a skeletal IP
205  * header (with len, off, ttl, proto, tos, src, dst).
206  * The mbuf chain containing the packet will be freed.
207  * The mbuf opt, if present, will not be freed.
208  */
209 int
210 ip_output(struct mbuf *m0, ...)
211 {
212 	struct ip *ip;
213 	struct ifnet *ifp;
214 	struct mbuf *m = m0;
215 	int hlen = sizeof (struct ip);
216 	int len, error = 0;
217 	struct route iproute;
218 	struct sockaddr_in *dst;
219 	struct in_ifaddr *ia;
220 	struct mbuf *opt;
221 	struct route *ro;
222 	int flags, sw_csum;
223 	int *mtu_p;
224 	u_long mtu;
225 	struct ip_moptions *imo;
226 	struct socket *so;
227 	va_list ap;
228 #ifdef IPSEC
229 	struct secpolicy *sp = NULL;
230 #ifdef IPSEC_NAT_T
231 	int natt_frag = 0;
232 #endif
233 #endif /*IPSEC*/
234 #ifdef FAST_IPSEC
235 	struct inpcb *inp;
236 	struct m_tag *mtag;
237 	struct secpolicy *sp = NULL;
238 	struct tdb_ident *tdbi;
239 	int s;
240 #endif
241 	u_int16_t ip_len;
242 
243 	len = 0;
244 	va_start(ap, m0);
245 	opt = va_arg(ap, struct mbuf *);
246 	ro = va_arg(ap, struct route *);
247 	flags = va_arg(ap, int);
248 	imo = va_arg(ap, struct ip_moptions *);
249 	so = va_arg(ap, struct socket *);
250 	if (flags & IP_RETURNMTU)
251 		mtu_p = va_arg(ap, int *);
252 	else
253 		mtu_p = NULL;
254 	va_end(ap);
255 
256 	MCLAIM(m, &ip_tx_mowner);
257 #ifdef FAST_IPSEC
258 	if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET)
259 		inp = (struct inpcb *)so->so_pcb;
260 	else
261 		inp = NULL;
262 #endif /* FAST_IPSEC */
263 
264 #ifdef	DIAGNOSTIC
265 	if ((m->m_flags & M_PKTHDR) == 0)
266 		panic("ip_output no HDR");
267 #endif
268 	if (opt) {
269 		m = ip_insertoptions(m, opt, &len);
270 		if (len >= sizeof(struct ip))
271 			hlen = len;
272 	}
273 	ip = mtod(m, struct ip *);
274 	/*
275 	 * Fill in IP header.
276 	 */
277 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
278 		ip->ip_v = IPVERSION;
279 		ip->ip_off = htons(0);
280 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
281 			ip->ip_id = ip_newid();
282 		} else {
283 
284 			/*
285 			 * TSO capable interfaces (typically?) increment
286 			 * ip_id for each segment.
287 			 * "allocate" enough ids here to increase the chance
288 			 * for them to be unique.
289 			 *
290 			 * note that the following calculation is not
291 			 * needed to be precise.  wasting some ip_id is fine.
292 			 */
293 
294 			unsigned int segsz = m->m_pkthdr.segsz;
295 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
296 			unsigned int num = howmany(datasz, segsz);
297 
298 			ip->ip_id = ip_newid_range(num);
299 		}
300 		ip->ip_hl = hlen >> 2;
301 		ipstat.ips_localout++;
302 	} else {
303 		hlen = ip->ip_hl << 2;
304 	}
305 	/*
306 	 * Route packet.
307 	 */
308 	if (ro == 0) {
309 		ro = &iproute;
310 		bzero((caddr_t)ro, sizeof (*ro));
311 	}
312 	dst = satosin(&ro->ro_dst);
313 	/*
314 	 * If there is a cached route,
315 	 * check that it is to the same destination
316 	 * and is still up.  If not, free it and try again.
317 	 * The address family should also be checked in case of sharing the
318 	 * cache with IPv6.
319 	 */
320 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
321 	    dst->sin_family != AF_INET ||
322 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
323 		RTFREE(ro->ro_rt);
324 		ro->ro_rt = (struct rtentry *)0;
325 	}
326 	if (ro->ro_rt == 0) {
327 		bzero(dst, sizeof(*dst));
328 		dst->sin_family = AF_INET;
329 		dst->sin_len = sizeof(*dst);
330 		dst->sin_addr = ip->ip_dst;
331 	}
332 	/*
333 	 * If routing to interface only,
334 	 * short circuit routing lookup.
335 	 */
336 	if (flags & IP_ROUTETOIF) {
337 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
338 			ipstat.ips_noroute++;
339 			error = ENETUNREACH;
340 			goto bad;
341 		}
342 		ifp = ia->ia_ifp;
343 		mtu = ifp->if_mtu;
344 		ip->ip_ttl = 1;
345 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
346 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
347 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
348 		ifp = imo->imo_multicast_ifp;
349 		mtu = ifp->if_mtu;
350 		IFP_TO_IA(ifp, ia);
351 	} else {
352 		if (ro->ro_rt == 0)
353 			rtalloc(ro);
354 		if (ro->ro_rt == 0) {
355 			ipstat.ips_noroute++;
356 			error = EHOSTUNREACH;
357 			goto bad;
358 		}
359 		ia = ifatoia(ro->ro_rt->rt_ifa);
360 		ifp = ro->ro_rt->rt_ifp;
361 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
362 			mtu = ifp->if_mtu;
363 		ro->ro_rt->rt_use++;
364 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
365 			dst = satosin(ro->ro_rt->rt_gateway);
366 	}
367 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
368 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
369 		struct in_multi *inm;
370 
371 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
372 			M_BCAST : M_MCAST;
373 		/*
374 		 * IP destination address is multicast.  Make sure "dst"
375 		 * still points to the address in "ro".  (It may have been
376 		 * changed to point to a gateway address, above.)
377 		 */
378 		dst = satosin(&ro->ro_dst);
379 		/*
380 		 * See if the caller provided any multicast options
381 		 */
382 		if (imo != NULL)
383 			ip->ip_ttl = imo->imo_multicast_ttl;
384 		else
385 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
386 
387 		/*
388 		 * if we don't know the outgoing ifp yet, we can't generate
389 		 * output
390 		 */
391 		if (!ifp) {
392 			ipstat.ips_noroute++;
393 			error = ENETUNREACH;
394 			goto bad;
395 		}
396 
397 		/*
398 		 * If the packet is multicast or broadcast, confirm that
399 		 * the outgoing interface can transmit it.
400 		 */
401 		if (((m->m_flags & M_MCAST) &&
402 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
403 		    ((m->m_flags & M_BCAST) &&
404 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
405 			ipstat.ips_noroute++;
406 			error = ENETUNREACH;
407 			goto bad;
408 		}
409 		/*
410 		 * If source address not specified yet, use an address
411 		 * of outgoing interface.
412 		 */
413 		if (in_nullhost(ip->ip_src)) {
414 			struct in_ifaddr *xia;
415 
416 			IFP_TO_IA(ifp, xia);
417 			if (!xia) {
418 				error = EADDRNOTAVAIL;
419 				goto bad;
420 			}
421 			ip->ip_src = xia->ia_addr.sin_addr;
422 		}
423 
424 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
425 		if (inm != NULL &&
426 		   (imo == NULL || imo->imo_multicast_loop)) {
427 			/*
428 			 * If we belong to the destination multicast group
429 			 * on the outgoing interface, and the caller did not
430 			 * forbid loopback, loop back a copy.
431 			 */
432 			ip_mloopback(ifp, m, dst);
433 		}
434 #ifdef MROUTING
435 		else {
436 			/*
437 			 * If we are acting as a multicast router, perform
438 			 * multicast forwarding as if the packet had just
439 			 * arrived on the interface to which we are about
440 			 * to send.  The multicast forwarding function
441 			 * recursively calls this function, using the
442 			 * IP_FORWARDING flag to prevent infinite recursion.
443 			 *
444 			 * Multicasts that are looped back by ip_mloopback(),
445 			 * above, will be forwarded by the ip_input() routine,
446 			 * if necessary.
447 			 */
448 			extern struct socket *ip_mrouter;
449 
450 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
451 				if (ip_mforward(m, ifp) != 0) {
452 					m_freem(m);
453 					goto done;
454 				}
455 			}
456 		}
457 #endif
458 		/*
459 		 * Multicasts with a time-to-live of zero may be looped-
460 		 * back, above, but must not be transmitted on a network.
461 		 * Also, multicasts addressed to the loopback interface
462 		 * are not sent -- the above call to ip_mloopback() will
463 		 * loop back a copy if this host actually belongs to the
464 		 * destination group on the loopback interface.
465 		 */
466 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
467 			m_freem(m);
468 			goto done;
469 		}
470 
471 		goto sendit;
472 	}
473 	/*
474 	 * If source address not specified yet, use address
475 	 * of outgoing interface.
476 	 */
477 	if (in_nullhost(ip->ip_src))
478 		ip->ip_src = ia->ia_addr.sin_addr;
479 
480 	/*
481 	 * packets with Class-D address as source are not valid per
482 	 * RFC 1112
483 	 */
484 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
485 		ipstat.ips_odropped++;
486 		error = EADDRNOTAVAIL;
487 		goto bad;
488 	}
489 
490 	/*
491 	 * Look for broadcast address and
492 	 * and verify user is allowed to send
493 	 * such a packet.
494 	 */
495 	if (in_broadcast(dst->sin_addr, ifp)) {
496 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
497 			error = EADDRNOTAVAIL;
498 			goto bad;
499 		}
500 		if ((flags & IP_ALLOWBROADCAST) == 0) {
501 			error = EACCES;
502 			goto bad;
503 		}
504 		/* don't allow broadcast messages to be fragmented */
505 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
506 			error = EMSGSIZE;
507 			goto bad;
508 		}
509 		m->m_flags |= M_BCAST;
510 	} else
511 		m->m_flags &= ~M_BCAST;
512 
513 sendit:
514 	/*
515 	 * If we're doing Path MTU Discovery, we need to set DF unless
516 	 * the route's MTU is locked.
517 	 */
518 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
519 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
520 		ip->ip_off |= htons(IP_DF);
521 
522 	/* Remember the current ip_len */
523 	ip_len = ntohs(ip->ip_len);
524 
525 #ifdef IPSEC
526 	/* get SP for this packet */
527 	if (so == NULL)
528 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
529 		    flags, &error);
530 	else {
531 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
532 					 IPSEC_DIR_OUTBOUND))
533 			goto skip_ipsec;
534 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
535 	}
536 
537 	if (sp == NULL) {
538 		ipsecstat.out_inval++;
539 		goto bad;
540 	}
541 
542 	error = 0;
543 
544 	/* check policy */
545 	switch (sp->policy) {
546 	case IPSEC_POLICY_DISCARD:
547 		/*
548 		 * This packet is just discarded.
549 		 */
550 		ipsecstat.out_polvio++;
551 		goto bad;
552 
553 	case IPSEC_POLICY_BYPASS:
554 	case IPSEC_POLICY_NONE:
555 		/* no need to do IPsec. */
556 		goto skip_ipsec;
557 
558 	case IPSEC_POLICY_IPSEC:
559 		if (sp->req == NULL) {
560 			/* XXX should be panic ? */
561 			printf("ip_output: No IPsec request specified.\n");
562 			error = EINVAL;
563 			goto bad;
564 		}
565 		break;
566 
567 	case IPSEC_POLICY_ENTRUST:
568 	default:
569 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
570 	}
571 
572 #ifdef IPSEC_NAT_T
573 	/*
574 	 * NAT-T ESP fragmentation: don't do IPSec processing now,
575 	 * we'll do it on each fragmented packet.
576 	 */
577 	if (sp->req->sav &&
578 	    ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
579 	     (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
580 		if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
581 			natt_frag = 1;
582 			mtu = sp->req->sav->esp_frag;
583 			goto skip_ipsec;
584 		}
585 	}
586 #endif /* IPSEC_NAT_T */
587 
588 	/*
589 	 * ipsec4_output() expects ip_len and ip_off in network
590 	 * order.  They have been set to network order above.
591 	 */
592 
593     {
594 	struct ipsec_output_state state;
595 	bzero(&state, sizeof(state));
596 	state.m = m;
597 	if (flags & IP_ROUTETOIF) {
598 		state.ro = &iproute;
599 		bzero(&iproute, sizeof(iproute));
600 	} else
601 		state.ro = ro;
602 	state.dst = (struct sockaddr *)dst;
603 
604 	/*
605 	 * We can't defer the checksum of payload data if
606 	 * we're about to encrypt/authenticate it.
607 	 *
608 	 * XXX When we support crypto offloading functions of
609 	 * XXX network interfaces, we need to reconsider this,
610 	 * XXX since it's likely that they'll support checksumming,
611 	 * XXX as well.
612 	 */
613 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
614 		in_delayed_cksum(m);
615 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
616 	}
617 
618 	error = ipsec4_output(&state, sp, flags);
619 
620 	m = state.m;
621 	if (flags & IP_ROUTETOIF) {
622 		/*
623 		 * if we have tunnel mode SA, we may need to ignore
624 		 * IP_ROUTETOIF.
625 		 */
626 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
627 			flags &= ~IP_ROUTETOIF;
628 			ro = state.ro;
629 		}
630 	} else
631 		ro = state.ro;
632 	dst = (struct sockaddr_in *)state.dst;
633 	if (error) {
634 		/* mbuf is already reclaimed in ipsec4_output. */
635 		m0 = NULL;
636 		switch (error) {
637 		case EHOSTUNREACH:
638 		case ENETUNREACH:
639 		case EMSGSIZE:
640 		case ENOBUFS:
641 		case ENOMEM:
642 			break;
643 		default:
644 			printf("ip4_output (ipsec): error code %d\n", error);
645 			/*fall through*/
646 		case ENOENT:
647 			/* don't show these error codes to the user */
648 			error = 0;
649 			break;
650 		}
651 		goto bad;
652 	}
653 
654 	/* be sure to update variables that are affected by ipsec4_output() */
655 	ip = mtod(m, struct ip *);
656 	hlen = ip->ip_hl << 2;
657 	ip_len = ntohs(ip->ip_len);
658 
659 	if (ro->ro_rt == NULL) {
660 		if ((flags & IP_ROUTETOIF) == 0) {
661 			printf("ip_output: "
662 				"can't update route after IPsec processing\n");
663 			error = EHOSTUNREACH;	/*XXX*/
664 			goto bad;
665 		}
666 	} else {
667 		/* nobody uses ia beyond here */
668 		if (state.encap) {
669 			ifp = ro->ro_rt->rt_ifp;
670 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
671 				mtu = ifp->if_mtu;
672 		}
673 	}
674     }
675 skip_ipsec:
676 #endif /*IPSEC*/
677 #ifdef FAST_IPSEC
678 	/*
679 	 * Check the security policy (SP) for the packet and, if
680 	 * required, do IPsec-related processing.  There are two
681 	 * cases here; the first time a packet is sent through
682 	 * it will be untagged and handled by ipsec4_checkpolicy.
683 	 * If the packet is resubmitted to ip_output (e.g. after
684 	 * AH, ESP, etc. processing), there will be a tag to bypass
685 	 * the lookup and related policy checking.
686 	 */
687 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
688 	s = splsoftnet();
689 	if (mtag != NULL) {
690 		tdbi = (struct tdb_ident *)(mtag + 1);
691 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
692 		if (sp == NULL)
693 			error = -EINVAL;	/* force silent drop */
694 		m_tag_delete(m, mtag);
695 	} else {
696 		if (inp != NULL &&
697 		    IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND))
698 			goto spd_done;
699 		sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
700 					&error, inp);
701 	}
702 	/*
703 	 * There are four return cases:
704 	 *    sp != NULL	 	    apply IPsec policy
705 	 *    sp == NULL, error == 0	    no IPsec handling needed
706 	 *    sp == NULL, error == -EINVAL  discard packet w/o error
707 	 *    sp == NULL, error != 0	    discard packet, report error
708 	 */
709 	if (sp != NULL) {
710 		/* Loop detection, check if ipsec processing already done */
711 		IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
712 		for (mtag = m_tag_first(m); mtag != NULL;
713 		     mtag = m_tag_next(m, mtag)) {
714 #ifdef MTAG_ABI_COMPAT
715 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
716 				continue;
717 #endif
718 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
719 			    mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
720 				continue;
721 			/*
722 			 * Check if policy has an SA associated with it.
723 			 * This can happen when an SP has yet to acquire
724 			 * an SA; e.g. on first reference.  If it occurs,
725 			 * then we let ipsec4_process_packet do its thing.
726 			 */
727 			if (sp->req->sav == NULL)
728 				break;
729 			tdbi = (struct tdb_ident *)(mtag + 1);
730 			if (tdbi->spi == sp->req->sav->spi &&
731 			    tdbi->proto == sp->req->sav->sah->saidx.proto &&
732 			    bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
733 				 sizeof (union sockaddr_union)) == 0) {
734 				/*
735 				 * No IPsec processing is needed, free
736 				 * reference to SP.
737 				 *
738 				 * NB: null pointer to avoid free at
739 				 *     done: below.
740 				 */
741 				KEY_FREESP(&sp), sp = NULL;
742 				splx(s);
743 				goto spd_done;
744 			}
745 		}
746 
747 		/*
748 		 * Do delayed checksums now because we send before
749 		 * this is done in the normal processing path.
750 		 */
751 		if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
752 			in_delayed_cksum(m);
753 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
754 		}
755 
756 #ifdef __FreeBSD__
757 		ip->ip_len = htons(ip->ip_len);
758 		ip->ip_off = htons(ip->ip_off);
759 #endif
760 
761 		/* NB: callee frees mbuf */
762 		error = ipsec4_process_packet(m, sp->req, flags, 0);
763 		/*
764 		 * Preserve KAME behaviour: ENOENT can be returned
765 		 * when an SA acquire is in progress.  Don't propagate
766 		 * this to user-level; it confuses applications.
767 		 *
768 		 * XXX this will go away when the SADB is redone.
769 		 */
770 		if (error == ENOENT)
771 			error = 0;
772 		splx(s);
773 		goto done;
774 	} else {
775 		splx(s);
776 
777 		if (error != 0) {
778 			/*
779 			 * Hack: -EINVAL is used to signal that a packet
780 			 * should be silently discarded.  This is typically
781 			 * because we asked key management for an SA and
782 			 * it was delayed (e.g. kicked up to IKE).
783 			 */
784 			if (error == -EINVAL)
785 				error = 0;
786 			goto bad;
787 		} else {
788 			/* No IPsec processing for this packet. */
789 		}
790 #ifdef notyet
791 		/*
792 		 * If deferred crypto processing is needed, check that
793 		 * the interface supports it.
794 		 */
795 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
796 		if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
797 			/* notify IPsec to do its own crypto */
798 			ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
799 			error = EHOSTUNREACH;
800 			goto bad;
801 		}
802 #endif
803 	}
804 spd_done:
805 #endif /* FAST_IPSEC */
806 
807 #ifdef PFIL_HOOKS
808 	/*
809 	 * Run through list of hooks for output packets.
810 	 */
811 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
812 		goto done;
813 	if (m == NULL)
814 		goto done;
815 
816 	ip = mtod(m, struct ip *);
817 	hlen = ip->ip_hl << 2;
818 #endif /* PFIL_HOOKS */
819 
820 	m->m_pkthdr.csum_data |= hlen << 16;
821 
822 #if IFA_STATS
823 	/*
824 	 * search for the source address structure to
825 	 * maintain output statistics.
826 	 */
827 	INADDR_TO_IA(ip->ip_src, ia);
828 #endif
829 
830 	/* Maybe skip checksums on loopback interfaces. */
831 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
832 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
833 	}
834 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
835 	/*
836 	 * If small enough for mtu of path, or if using TCP segmentation
837 	 * offload, can just send directly.
838 	 */
839 	if (ip_len <= mtu ||
840 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
841 #if IFA_STATS
842 		if (ia)
843 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
844 #endif
845 		/*
846 		 * Always initialize the sum to 0!  Some HW assisted
847 		 * checksumming requires this.
848 		 */
849 		ip->ip_sum = 0;
850 
851 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
852 			/*
853 			 * Perform any checksums that the hardware can't do
854 			 * for us.
855 			 *
856 			 * XXX Does any hardware require the {th,uh}_sum
857 			 * XXX fields to be 0?
858 			 */
859 			if (sw_csum & M_CSUM_IPv4) {
860 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
861 				ip->ip_sum = in_cksum(m, hlen);
862 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
863 			}
864 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
865 				if (IN_NEED_CHECKSUM(ifp,
866 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
867 					in_delayed_cksum(m);
868 				}
869 				m->m_pkthdr.csum_flags &=
870 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
871 			}
872 		}
873 
874 #ifdef IPSEC
875 		/* clean ipsec history once it goes out of the node */
876 		ipsec_delaux(m);
877 #endif
878 
879 		if (__predict_true(
880 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
881 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
882 			error =
883 			    (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
884 		} else {
885 			error =
886 			    ip_tso_output(ifp, m, sintosa(dst), ro->ro_rt);
887 		}
888 		goto done;
889 	}
890 
891 	/*
892 	 * We can't use HW checksumming if we're about to
893 	 * to fragment the packet.
894 	 *
895 	 * XXX Some hardware can do this.
896 	 */
897 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
898 		if (IN_NEED_CHECKSUM(ifp,
899 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
900 			in_delayed_cksum(m);
901 		}
902 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
903 	}
904 
905 	/*
906 	 * Too large for interface; fragment if possible.
907 	 * Must be able to put at least 8 bytes per fragment.
908 	 */
909 	if (ntohs(ip->ip_off) & IP_DF) {
910 		if (flags & IP_RETURNMTU)
911 			*mtu_p = mtu;
912 		error = EMSGSIZE;
913 		ipstat.ips_cantfrag++;
914 		goto bad;
915 	}
916 
917 	error = ip_fragment(m, ifp, mtu);
918 	if (error) {
919 		m = NULL;
920 		goto bad;
921 	}
922 
923 	for (; m; m = m0) {
924 		m0 = m->m_nextpkt;
925 		m->m_nextpkt = 0;
926 		if (error == 0) {
927 #if IFA_STATS
928 			if (ia)
929 				ia->ia_ifa.ifa_data.ifad_outbytes +=
930 				    ntohs(ip->ip_len);
931 #endif
932 #ifdef IPSEC
933 			/* clean ipsec history once it goes out of the node */
934 			ipsec_delaux(m);
935 
936 #ifdef IPSEC_NAT_T
937 			/*
938 			 * If we get there, the packet has not been handeld by
939 			 * IPSec whereas it should have. Now that it has been
940 			 * fragmented, re-inject it in ip_output so that IPsec
941 			 * processing can occur.
942 			 */
943 			if (natt_frag) {
944 				error = ip_output(m, opt,
945 				    ro, flags, imo, so, mtu_p);
946 			} else
947 #endif /* IPSEC_NAT_T */
948 #endif /* IPSEC */
949 			{
950 				KASSERT((m->m_pkthdr.csum_flags &
951 				    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
952 				error = (*ifp->if_output)(ifp, m, sintosa(dst),
953 				    ro->ro_rt);
954 			}
955 		} else
956 			m_freem(m);
957 	}
958 
959 	if (error == 0)
960 		ipstat.ips_fragmented++;
961 done:
962 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
963 		RTFREE(ro->ro_rt);
964 		ro->ro_rt = 0;
965 	}
966 
967 #ifdef IPSEC
968 	if (sp != NULL) {
969 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
970 			printf("DP ip_output call free SP:%p\n", sp));
971 		key_freesp(sp);
972 	}
973 #endif /* IPSEC */
974 #ifdef FAST_IPSEC
975 	if (sp != NULL)
976 		KEY_FREESP(&sp);
977 #endif /* FAST_IPSEC */
978 
979 	return (error);
980 bad:
981 	m_freem(m);
982 	goto done;
983 }
984 
985 int
986 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
987 {
988 	struct ip *ip, *mhip;
989 	struct mbuf *m0;
990 	int len, hlen, off;
991 	int mhlen, firstlen;
992 	struct mbuf **mnext;
993 	int sw_csum = m->m_pkthdr.csum_flags;
994 	int fragments = 0;
995 	int s;
996 	int error = 0;
997 
998 	ip = mtod(m, struct ip *);
999 	hlen = ip->ip_hl << 2;
1000 	if (ifp != NULL)
1001 		sw_csum &= ~ifp->if_csum_flags_tx;
1002 
1003 	len = (mtu - hlen) &~ 7;
1004 	if (len < 8) {
1005 		m_freem(m);
1006 		return (EMSGSIZE);
1007 	}
1008 
1009 	firstlen = len;
1010 	mnext = &m->m_nextpkt;
1011 
1012 	/*
1013 	 * Loop through length of segment after first fragment,
1014 	 * make new header and copy data of each part and link onto chain.
1015 	 */
1016 	m0 = m;
1017 	mhlen = sizeof (struct ip);
1018 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
1019 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
1020 		if (m == 0) {
1021 			error = ENOBUFS;
1022 			ipstat.ips_odropped++;
1023 			goto sendorfree;
1024 		}
1025 		MCLAIM(m, m0->m_owner);
1026 		*mnext = m;
1027 		mnext = &m->m_nextpkt;
1028 		m->m_data += max_linkhdr;
1029 		mhip = mtod(m, struct ip *);
1030 		*mhip = *ip;
1031 		/* we must inherit MCAST and BCAST flags */
1032 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
1033 		if (hlen > sizeof (struct ip)) {
1034 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1035 			mhip->ip_hl = mhlen >> 2;
1036 		}
1037 		m->m_len = mhlen;
1038 		mhip->ip_off = ((off - hlen) >> 3) +
1039 		    (ntohs(ip->ip_off) & ~IP_MF);
1040 		if (ip->ip_off & htons(IP_MF))
1041 			mhip->ip_off |= IP_MF;
1042 		if (off + len >= ntohs(ip->ip_len))
1043 			len = ntohs(ip->ip_len) - off;
1044 		else
1045 			mhip->ip_off |= IP_MF;
1046 		HTONS(mhip->ip_off);
1047 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
1048 		m->m_next = m_copy(m0, off, len);
1049 		if (m->m_next == 0) {
1050 			error = ENOBUFS;	/* ??? */
1051 			ipstat.ips_odropped++;
1052 			goto sendorfree;
1053 		}
1054 		m->m_pkthdr.len = mhlen + len;
1055 		m->m_pkthdr.rcvif = (struct ifnet *)0;
1056 		mhip->ip_sum = 0;
1057 		if (sw_csum & M_CSUM_IPv4) {
1058 			mhip->ip_sum = in_cksum(m, mhlen);
1059 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
1060 		} else {
1061 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
1062 			m->m_pkthdr.csum_data |= mhlen << 16;
1063 		}
1064 		ipstat.ips_ofragments++;
1065 		fragments++;
1066 	}
1067 	/*
1068 	 * Update first fragment by trimming what's been copied out
1069 	 * and updating header, then send each fragment (in order).
1070 	 */
1071 	m = m0;
1072 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
1073 	m->m_pkthdr.len = hlen + firstlen;
1074 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
1075 	ip->ip_off |= htons(IP_MF);
1076 	ip->ip_sum = 0;
1077 	if (sw_csum & M_CSUM_IPv4) {
1078 		ip->ip_sum = in_cksum(m, hlen);
1079 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
1080 	} else {
1081 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
1082 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
1083 			sizeof(struct ip));
1084 	}
1085 sendorfree:
1086 	/*
1087 	 * If there is no room for all the fragments, don't queue
1088 	 * any of them.
1089 	 */
1090 	if (ifp != NULL) {
1091 		s = splnet();
1092 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
1093 		    error == 0) {
1094 			error = ENOBUFS;
1095 			ipstat.ips_odropped++;
1096 			IFQ_INC_DROPS(&ifp->if_snd);
1097 		}
1098 		splx(s);
1099 	}
1100 	if (error) {
1101 		for (m = m0; m; m = m0) {
1102 			m0 = m->m_nextpkt;
1103 			m->m_nextpkt = NULL;
1104 			m_freem(m);
1105 		}
1106 	}
1107 	return (error);
1108 }
1109 
1110 /*
1111  * Process a delayed payload checksum calculation.
1112  */
1113 void
1114 in_delayed_cksum(struct mbuf *m)
1115 {
1116 	struct ip *ip;
1117 	u_int16_t csum, offset;
1118 
1119 	ip = mtod(m, struct ip *);
1120 	offset = ip->ip_hl << 2;
1121 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
1122 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
1123 		csum = 0xffff;
1124 
1125 	offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
1126 
1127 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
1128 		/* This happen when ip options were inserted
1129 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
1130 		    m->m_len, offset, ip->ip_p);
1131 		 */
1132 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
1133 	} else
1134 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1135 }
1136 
1137 /*
1138  * Determine the maximum length of the options to be inserted;
1139  * we would far rather allocate too much space rather than too little.
1140  */
1141 
1142 u_int
1143 ip_optlen(struct inpcb *inp)
1144 {
1145 	struct mbuf *m = inp->inp_options;
1146 
1147 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
1148 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1149 	else
1150 		return 0;
1151 }
1152 
1153 
1154 /*
1155  * Insert IP options into preformed packet.
1156  * Adjust IP destination as required for IP source routing,
1157  * as indicated by a non-zero in_addr at the start of the options.
1158  */
1159 static struct mbuf *
1160 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1161 {
1162 	struct ipoption *p = mtod(opt, struct ipoption *);
1163 	struct mbuf *n;
1164 	struct ip *ip = mtod(m, struct ip *);
1165 	unsigned optlen;
1166 
1167 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1168 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1169 		return (m);		/* XXX should fail */
1170 	if (!in_nullhost(p->ipopt_dst))
1171 		ip->ip_dst = p->ipopt_dst;
1172 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1173 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1174 		if (n == 0)
1175 			return (m);
1176 		MCLAIM(n, m->m_owner);
1177 		M_MOVE_PKTHDR(n, m);
1178 		m->m_len -= sizeof(struct ip);
1179 		m->m_data += sizeof(struct ip);
1180 		n->m_next = m;
1181 		m = n;
1182 		m->m_len = optlen + sizeof(struct ip);
1183 		m->m_data += max_linkhdr;
1184 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1185 	} else {
1186 		m->m_data -= optlen;
1187 		m->m_len += optlen;
1188 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
1189 	}
1190 	m->m_pkthdr.len += optlen;
1191 	ip = mtod(m, struct ip *);
1192 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1193 	*phlen = sizeof(struct ip) + optlen;
1194 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1195 	return (m);
1196 }
1197 
1198 /*
1199  * Copy options from ip to jp,
1200  * omitting those not copied during fragmentation.
1201  */
1202 int
1203 ip_optcopy(struct ip *ip, struct ip *jp)
1204 {
1205 	u_char *cp, *dp;
1206 	int opt, optlen, cnt;
1207 
1208 	cp = (u_char *)(ip + 1);
1209 	dp = (u_char *)(jp + 1);
1210 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1211 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1212 		opt = cp[0];
1213 		if (opt == IPOPT_EOL)
1214 			break;
1215 		if (opt == IPOPT_NOP) {
1216 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1217 			*dp++ = IPOPT_NOP;
1218 			optlen = 1;
1219 			continue;
1220 		}
1221 #ifdef DIAGNOSTIC
1222 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1223 			panic("malformed IPv4 option passed to ip_optcopy");
1224 #endif
1225 		optlen = cp[IPOPT_OLEN];
1226 #ifdef DIAGNOSTIC
1227 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1228 			panic("malformed IPv4 option passed to ip_optcopy");
1229 #endif
1230 		/* bogus lengths should have been caught by ip_dooptions */
1231 		if (optlen > cnt)
1232 			optlen = cnt;
1233 		if (IPOPT_COPIED(opt)) {
1234 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1235 			dp += optlen;
1236 		}
1237 	}
1238 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1239 		*dp++ = IPOPT_EOL;
1240 	return (optlen);
1241 }
1242 
1243 /*
1244  * IP socket option processing.
1245  */
1246 int
1247 ip_ctloutput(int op, struct socket *so, int level, int optname,
1248     struct mbuf **mp)
1249 {
1250 	struct inpcb *inp = sotoinpcb(so);
1251 	struct mbuf *m = *mp;
1252 	int optval = 0;
1253 	int error = 0;
1254 #if defined(IPSEC) || defined(FAST_IPSEC)
1255 	struct proc *p = curproc;	/*XXX*/
1256 #endif
1257 
1258 	if (level != IPPROTO_IP) {
1259 		error = EINVAL;
1260 		if (op == PRCO_SETOPT && *mp)
1261 			(void) m_free(*mp);
1262 	} else switch (op) {
1263 
1264 	case PRCO_SETOPT:
1265 		switch (optname) {
1266 		case IP_OPTIONS:
1267 #ifdef notyet
1268 		case IP_RETOPTS:
1269 			return (ip_pcbopts(optname, &inp->inp_options, m));
1270 #else
1271 			return (ip_pcbopts(&inp->inp_options, m));
1272 #endif
1273 
1274 		case IP_TOS:
1275 		case IP_TTL:
1276 		case IP_RECVOPTS:
1277 		case IP_RECVRETOPTS:
1278 		case IP_RECVDSTADDR:
1279 		case IP_RECVIF:
1280 			if (m == NULL || m->m_len != sizeof(int))
1281 				error = EINVAL;
1282 			else {
1283 				optval = *mtod(m, int *);
1284 				switch (optname) {
1285 
1286 				case IP_TOS:
1287 					inp->inp_ip.ip_tos = optval;
1288 					break;
1289 
1290 				case IP_TTL:
1291 					inp->inp_ip.ip_ttl = optval;
1292 					break;
1293 #define	OPTSET(bit) \
1294 	if (optval) \
1295 		inp->inp_flags |= bit; \
1296 	else \
1297 		inp->inp_flags &= ~bit;
1298 
1299 				case IP_RECVOPTS:
1300 					OPTSET(INP_RECVOPTS);
1301 					break;
1302 
1303 				case IP_RECVRETOPTS:
1304 					OPTSET(INP_RECVRETOPTS);
1305 					break;
1306 
1307 				case IP_RECVDSTADDR:
1308 					OPTSET(INP_RECVDSTADDR);
1309 					break;
1310 
1311 				case IP_RECVIF:
1312 					OPTSET(INP_RECVIF);
1313 					break;
1314 				}
1315 			}
1316 			break;
1317 #undef OPTSET
1318 
1319 		case IP_MULTICAST_IF:
1320 		case IP_MULTICAST_TTL:
1321 		case IP_MULTICAST_LOOP:
1322 		case IP_ADD_MEMBERSHIP:
1323 		case IP_DROP_MEMBERSHIP:
1324 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1325 			break;
1326 
1327 		case IP_PORTRANGE:
1328 			if (m == 0 || m->m_len != sizeof(int))
1329 				error = EINVAL;
1330 			else {
1331 				optval = *mtod(m, int *);
1332 
1333 				switch (optval) {
1334 
1335 				case IP_PORTRANGE_DEFAULT:
1336 				case IP_PORTRANGE_HIGH:
1337 					inp->inp_flags &= ~(INP_LOWPORT);
1338 					break;
1339 
1340 				case IP_PORTRANGE_LOW:
1341 					inp->inp_flags |= INP_LOWPORT;
1342 					break;
1343 
1344 				default:
1345 					error = EINVAL;
1346 					break;
1347 				}
1348 			}
1349 			break;
1350 
1351 #if defined(IPSEC) || defined(FAST_IPSEC)
1352 		case IP_IPSEC_POLICY:
1353 		{
1354 			caddr_t req = NULL;
1355 			size_t len = 0;
1356 			int priv = 0;
1357 
1358 #ifdef __NetBSD__
1359 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1360 				priv = 0;
1361 			else
1362 				priv = 1;
1363 #else
1364 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1365 #endif
1366 			if (m) {
1367 				req = mtod(m, caddr_t);
1368 				len = m->m_len;
1369 			}
1370 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1371 			break;
1372 		    }
1373 #endif /*IPSEC*/
1374 
1375 		default:
1376 			error = ENOPROTOOPT;
1377 			break;
1378 		}
1379 		if (m)
1380 			(void)m_free(m);
1381 		break;
1382 
1383 	case PRCO_GETOPT:
1384 		switch (optname) {
1385 		case IP_OPTIONS:
1386 		case IP_RETOPTS:
1387 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1388 			MCLAIM(m, so->so_mowner);
1389 			if (inp->inp_options) {
1390 				m->m_len = inp->inp_options->m_len;
1391 				bcopy(mtod(inp->inp_options, caddr_t),
1392 				    mtod(m, caddr_t), (unsigned)m->m_len);
1393 			} else
1394 				m->m_len = 0;
1395 			break;
1396 
1397 		case IP_TOS:
1398 		case IP_TTL:
1399 		case IP_RECVOPTS:
1400 		case IP_RECVRETOPTS:
1401 		case IP_RECVDSTADDR:
1402 		case IP_RECVIF:
1403 		case IP_ERRORMTU:
1404 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1405 			MCLAIM(m, so->so_mowner);
1406 			m->m_len = sizeof(int);
1407 			switch (optname) {
1408 
1409 			case IP_TOS:
1410 				optval = inp->inp_ip.ip_tos;
1411 				break;
1412 
1413 			case IP_TTL:
1414 				optval = inp->inp_ip.ip_ttl;
1415 				break;
1416 
1417 			case IP_ERRORMTU:
1418 				optval = inp->inp_errormtu;
1419 				break;
1420 
1421 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1422 
1423 			case IP_RECVOPTS:
1424 				optval = OPTBIT(INP_RECVOPTS);
1425 				break;
1426 
1427 			case IP_RECVRETOPTS:
1428 				optval = OPTBIT(INP_RECVRETOPTS);
1429 				break;
1430 
1431 			case IP_RECVDSTADDR:
1432 				optval = OPTBIT(INP_RECVDSTADDR);
1433 				break;
1434 
1435 			case IP_RECVIF:
1436 				optval = OPTBIT(INP_RECVIF);
1437 				break;
1438 			}
1439 			*mtod(m, int *) = optval;
1440 			break;
1441 
1442 #if 0	/* defined(IPSEC) || defined(FAST_IPSEC) */
1443 		/* XXX: code broken */
1444 		case IP_IPSEC_POLICY:
1445 		{
1446 			caddr_t req = NULL;
1447 			size_t len = 0;
1448 
1449 			if (m) {
1450 				req = mtod(m, caddr_t);
1451 				len = m->m_len;
1452 			}
1453 			error = ipsec4_get_policy(inp, req, len, mp);
1454 			break;
1455 		}
1456 #endif /*IPSEC*/
1457 
1458 		case IP_MULTICAST_IF:
1459 		case IP_MULTICAST_TTL:
1460 		case IP_MULTICAST_LOOP:
1461 		case IP_ADD_MEMBERSHIP:
1462 		case IP_DROP_MEMBERSHIP:
1463 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1464 			if (*mp)
1465 				MCLAIM(*mp, so->so_mowner);
1466 			break;
1467 
1468 		case IP_PORTRANGE:
1469 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1470 			MCLAIM(m, so->so_mowner);
1471 			m->m_len = sizeof(int);
1472 
1473 			if (inp->inp_flags & INP_LOWPORT)
1474 				optval = IP_PORTRANGE_LOW;
1475 			else
1476 				optval = IP_PORTRANGE_DEFAULT;
1477 
1478 			*mtod(m, int *) = optval;
1479 			break;
1480 
1481 		default:
1482 			error = ENOPROTOOPT;
1483 			break;
1484 		}
1485 		break;
1486 	}
1487 	return (error);
1488 }
1489 
1490 /*
1491  * Set up IP options in pcb for insertion in output packets.
1492  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1493  * with destination address if source routed.
1494  */
1495 int
1496 #ifdef notyet
1497 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1498 #else
1499 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1500 #endif
1501 {
1502 	int cnt, optlen;
1503 	u_char *cp;
1504 	u_char opt;
1505 
1506 	/* turn off any old options */
1507 	if (*pcbopt)
1508 		(void)m_free(*pcbopt);
1509 	*pcbopt = 0;
1510 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1511 		/*
1512 		 * Only turning off any previous options.
1513 		 */
1514 		if (m)
1515 			(void)m_free(m);
1516 		return (0);
1517 	}
1518 
1519 #ifndef	__vax__
1520 	if (m->m_len % sizeof(int32_t))
1521 		goto bad;
1522 #endif
1523 	/*
1524 	 * IP first-hop destination address will be stored before
1525 	 * actual options; move other options back
1526 	 * and clear it when none present.
1527 	 */
1528 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1529 		goto bad;
1530 	cnt = m->m_len;
1531 	m->m_len += sizeof(struct in_addr);
1532 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1533 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1534 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1535 
1536 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1537 		opt = cp[IPOPT_OPTVAL];
1538 		if (opt == IPOPT_EOL)
1539 			break;
1540 		if (opt == IPOPT_NOP)
1541 			optlen = 1;
1542 		else {
1543 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1544 				goto bad;
1545 			optlen = cp[IPOPT_OLEN];
1546 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1547 				goto bad;
1548 		}
1549 		switch (opt) {
1550 
1551 		default:
1552 			break;
1553 
1554 		case IPOPT_LSRR:
1555 		case IPOPT_SSRR:
1556 			/*
1557 			 * user process specifies route as:
1558 			 *	->A->B->C->D
1559 			 * D must be our final destination (but we can't
1560 			 * check that since we may not have connected yet).
1561 			 * A is first hop destination, which doesn't appear in
1562 			 * actual IP option, but is stored before the options.
1563 			 */
1564 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1565 				goto bad;
1566 			m->m_len -= sizeof(struct in_addr);
1567 			cnt -= sizeof(struct in_addr);
1568 			optlen -= sizeof(struct in_addr);
1569 			cp[IPOPT_OLEN] = optlen;
1570 			/*
1571 			 * Move first hop before start of options.
1572 			 */
1573 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1574 			    sizeof(struct in_addr));
1575 			/*
1576 			 * Then copy rest of options back
1577 			 * to close up the deleted entry.
1578 			 */
1579 			(void)memmove(&cp[IPOPT_OFFSET+1],
1580 			    &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1581 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
1582 			break;
1583 		}
1584 	}
1585 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1586 		goto bad;
1587 	*pcbopt = m;
1588 	return (0);
1589 
1590 bad:
1591 	(void)m_free(m);
1592 	return (EINVAL);
1593 }
1594 
1595 /*
1596  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1597  */
1598 static struct ifnet *
1599 ip_multicast_if(struct in_addr *a, int *ifindexp)
1600 {
1601 	int ifindex;
1602 	struct ifnet *ifp = NULL;
1603 	struct in_ifaddr *ia;
1604 
1605 	if (ifindexp)
1606 		*ifindexp = 0;
1607 	if (ntohl(a->s_addr) >> 24 == 0) {
1608 		ifindex = ntohl(a->s_addr) & 0xffffff;
1609 		if (ifindex < 0 || if_indexlim <= ifindex)
1610 			return NULL;
1611 		ifp = ifindex2ifnet[ifindex];
1612 		if (!ifp)
1613 			return NULL;
1614 		if (ifindexp)
1615 			*ifindexp = ifindex;
1616 	} else {
1617 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1618 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1619 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1620 				ifp = ia->ia_ifp;
1621 				break;
1622 			}
1623 		}
1624 	}
1625 	return ifp;
1626 }
1627 
1628 static int
1629 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval)
1630 {
1631 	u_int tval;
1632 
1633 	if (m == NULL)
1634 		return EINVAL;
1635 
1636 	switch (m->m_len) {
1637 	case sizeof(u_char):
1638 		tval = *(mtod(m, u_char *));
1639 		break;
1640 	case sizeof(u_int):
1641 		tval = *(mtod(m, u_int *));
1642 		break;
1643 	default:
1644 		return EINVAL;
1645 	}
1646 
1647 	if (tval > maxval)
1648 		return EINVAL;
1649 
1650 	*val = tval;
1651 	return 0;
1652 }
1653 
1654 /*
1655  * Set the IP multicast options in response to user setsockopt().
1656  */
1657 int
1658 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m)
1659 {
1660 	int error = 0;
1661 	int i;
1662 	struct in_addr addr;
1663 	struct ip_mreq *mreq;
1664 	struct ifnet *ifp;
1665 	struct ip_moptions *imo = *imop;
1666 	struct route ro;
1667 	struct sockaddr_in *dst;
1668 	int ifindex;
1669 
1670 	if (imo == NULL) {
1671 		/*
1672 		 * No multicast option buffer attached to the pcb;
1673 		 * allocate one and initialize to default values.
1674 		 */
1675 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1676 		    M_WAITOK);
1677 
1678 		if (imo == NULL)
1679 			return (ENOBUFS);
1680 		*imop = imo;
1681 		imo->imo_multicast_ifp = NULL;
1682 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1683 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1684 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1685 		imo->imo_num_memberships = 0;
1686 	}
1687 
1688 	switch (optname) {
1689 
1690 	case IP_MULTICAST_IF:
1691 		/*
1692 		 * Select the interface for outgoing multicast packets.
1693 		 */
1694 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1695 			error = EINVAL;
1696 			break;
1697 		}
1698 		addr = *(mtod(m, struct in_addr *));
1699 		/*
1700 		 * INADDR_ANY is used to remove a previous selection.
1701 		 * When no interface is selected, a default one is
1702 		 * chosen every time a multicast packet is sent.
1703 		 */
1704 		if (in_nullhost(addr)) {
1705 			imo->imo_multicast_ifp = NULL;
1706 			break;
1707 		}
1708 		/*
1709 		 * The selected interface is identified by its local
1710 		 * IP address.  Find the interface and confirm that
1711 		 * it supports multicasting.
1712 		 */
1713 		ifp = ip_multicast_if(&addr, &ifindex);
1714 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1715 			error = EADDRNOTAVAIL;
1716 			break;
1717 		}
1718 		imo->imo_multicast_ifp = ifp;
1719 		if (ifindex)
1720 			imo->imo_multicast_addr = addr;
1721 		else
1722 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1723 		break;
1724 
1725 	case IP_MULTICAST_TTL:
1726 		/*
1727 		 * Set the IP time-to-live for outgoing multicast packets.
1728 		 */
1729 		error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL);
1730 		break;
1731 
1732 	case IP_MULTICAST_LOOP:
1733 		/*
1734 		 * Set the loopback flag for outgoing multicast packets.
1735 		 * Must be zero or one.
1736 		 */
1737 		error = ip_getoptval(m, &imo->imo_multicast_loop, 1);
1738 		break;
1739 
1740 	case IP_ADD_MEMBERSHIP:
1741 		/*
1742 		 * Add a multicast group membership.
1743 		 * Group must be a valid IP multicast address.
1744 		 */
1745 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1746 			error = EINVAL;
1747 			break;
1748 		}
1749 		mreq = mtod(m, struct ip_mreq *);
1750 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1751 			error = EINVAL;
1752 			break;
1753 		}
1754 		/*
1755 		 * If no interface address was provided, use the interface of
1756 		 * the route to the given multicast address.
1757 		 */
1758 		if (in_nullhost(mreq->imr_interface)) {
1759 			bzero((caddr_t)&ro, sizeof(ro));
1760 			ro.ro_rt = NULL;
1761 			dst = satosin(&ro.ro_dst);
1762 			dst->sin_len = sizeof(*dst);
1763 			dst->sin_family = AF_INET;
1764 			dst->sin_addr = mreq->imr_multiaddr;
1765 			rtalloc(&ro);
1766 			if (ro.ro_rt == NULL) {
1767 				error = EADDRNOTAVAIL;
1768 				break;
1769 			}
1770 			ifp = ro.ro_rt->rt_ifp;
1771 			rtfree(ro.ro_rt);
1772 		} else {
1773 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1774 		}
1775 		/*
1776 		 * See if we found an interface, and confirm that it
1777 		 * supports multicast.
1778 		 */
1779 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1780 			error = EADDRNOTAVAIL;
1781 			break;
1782 		}
1783 		/*
1784 		 * See if the membership already exists or if all the
1785 		 * membership slots are full.
1786 		 */
1787 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1788 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1789 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1790 				      mreq->imr_multiaddr))
1791 				break;
1792 		}
1793 		if (i < imo->imo_num_memberships) {
1794 			error = EADDRINUSE;
1795 			break;
1796 		}
1797 		if (i == IP_MAX_MEMBERSHIPS) {
1798 			error = ETOOMANYREFS;
1799 			break;
1800 		}
1801 		/*
1802 		 * Everything looks good; add a new record to the multicast
1803 		 * address list for the given interface.
1804 		 */
1805 		if ((imo->imo_membership[i] =
1806 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1807 			error = ENOBUFS;
1808 			break;
1809 		}
1810 		++imo->imo_num_memberships;
1811 		break;
1812 
1813 	case IP_DROP_MEMBERSHIP:
1814 		/*
1815 		 * Drop a multicast group membership.
1816 		 * Group must be a valid IP multicast address.
1817 		 */
1818 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1819 			error = EINVAL;
1820 			break;
1821 		}
1822 		mreq = mtod(m, struct ip_mreq *);
1823 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1824 			error = EINVAL;
1825 			break;
1826 		}
1827 		/*
1828 		 * If an interface address was specified, get a pointer
1829 		 * to its ifnet structure.
1830 		 */
1831 		if (in_nullhost(mreq->imr_interface))
1832 			ifp = NULL;
1833 		else {
1834 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1835 			if (ifp == NULL) {
1836 				error = EADDRNOTAVAIL;
1837 				break;
1838 			}
1839 		}
1840 		/*
1841 		 * Find the membership in the membership array.
1842 		 */
1843 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1844 			if ((ifp == NULL ||
1845 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1846 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1847 				       mreq->imr_multiaddr))
1848 				break;
1849 		}
1850 		if (i == imo->imo_num_memberships) {
1851 			error = EADDRNOTAVAIL;
1852 			break;
1853 		}
1854 		/*
1855 		 * Give up the multicast address record to which the
1856 		 * membership points.
1857 		 */
1858 		in_delmulti(imo->imo_membership[i]);
1859 		/*
1860 		 * Remove the gap in the membership array.
1861 		 */
1862 		for (++i; i < imo->imo_num_memberships; ++i)
1863 			imo->imo_membership[i-1] = imo->imo_membership[i];
1864 		--imo->imo_num_memberships;
1865 		break;
1866 
1867 	default:
1868 		error = EOPNOTSUPP;
1869 		break;
1870 	}
1871 
1872 	/*
1873 	 * If all options have default values, no need to keep the mbuf.
1874 	 */
1875 	if (imo->imo_multicast_ifp == NULL &&
1876 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1877 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1878 	    imo->imo_num_memberships == 0) {
1879 		free(*imop, M_IPMOPTS);
1880 		*imop = NULL;
1881 	}
1882 
1883 	return (error);
1884 }
1885 
1886 /*
1887  * Return the IP multicast options in response to user getsockopt().
1888  */
1889 int
1890 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
1891 {
1892 	u_char *ttl;
1893 	u_char *loop;
1894 	struct in_addr *addr;
1895 	struct in_ifaddr *ia;
1896 
1897 	*mp = m_get(M_WAIT, MT_SOOPTS);
1898 
1899 	switch (optname) {
1900 
1901 	case IP_MULTICAST_IF:
1902 		addr = mtod(*mp, struct in_addr *);
1903 		(*mp)->m_len = sizeof(struct in_addr);
1904 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1905 			*addr = zeroin_addr;
1906 		else if (imo->imo_multicast_addr.s_addr) {
1907 			/* return the value user has set */
1908 			*addr = imo->imo_multicast_addr;
1909 		} else {
1910 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1911 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1912 		}
1913 		return (0);
1914 
1915 	case IP_MULTICAST_TTL:
1916 		ttl = mtod(*mp, u_char *);
1917 		(*mp)->m_len = 1;
1918 		*ttl = imo ? imo->imo_multicast_ttl
1919 			   : IP_DEFAULT_MULTICAST_TTL;
1920 		return (0);
1921 
1922 	case IP_MULTICAST_LOOP:
1923 		loop = mtod(*mp, u_char *);
1924 		(*mp)->m_len = 1;
1925 		*loop = imo ? imo->imo_multicast_loop
1926 			    : IP_DEFAULT_MULTICAST_LOOP;
1927 		return (0);
1928 
1929 	default:
1930 		return (EOPNOTSUPP);
1931 	}
1932 }
1933 
1934 /*
1935  * Discard the IP multicast options.
1936  */
1937 void
1938 ip_freemoptions(struct ip_moptions *imo)
1939 {
1940 	int i;
1941 
1942 	if (imo != NULL) {
1943 		for (i = 0; i < imo->imo_num_memberships; ++i)
1944 			in_delmulti(imo->imo_membership[i]);
1945 		free(imo, M_IPMOPTS);
1946 	}
1947 }
1948 
1949 /*
1950  * Routine called from ip_output() to loop back a copy of an IP multicast
1951  * packet to the input queue of a specified interface.  Note that this
1952  * calls the output routine of the loopback "driver", but with an interface
1953  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
1954  */
1955 static void
1956 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1957 {
1958 	struct ip *ip;
1959 	struct mbuf *copym;
1960 
1961 	copym = m_copy(m, 0, M_COPYALL);
1962 	if (copym != NULL
1963 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1964 		copym = m_pullup(copym, sizeof(struct ip));
1965 	if (copym != NULL) {
1966 		/*
1967 		 * We don't bother to fragment if the IP length is greater
1968 		 * than the interface's MTU.  Can this possibly matter?
1969 		 */
1970 		ip = mtod(copym, struct ip *);
1971 
1972 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1973 			in_delayed_cksum(copym);
1974 			copym->m_pkthdr.csum_flags &=
1975 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1976 		}
1977 
1978 		ip->ip_sum = 0;
1979 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1980 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1981 	}
1982 }
1983