xref: /netbsd-src/sys/netinet/ip_output.c (revision fad4c9f71477ae11cea2ee75ec82151ac770a534)
1 /*	$NetBSD: ip_output.c,v 1.162 2006/05/15 00:05:17 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 1.162 2006/05/15 00:05:17 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 #include <sys/kauth.h>
116 #ifdef FAST_IPSEC
117 #include <sys/domain.h>
118 #endif
119 #include <sys/systm.h>
120 #include <sys/proc.h>
121 
122 #include <net/if.h>
123 #include <net/route.h>
124 #include <net/pfil.h>
125 
126 #include <netinet/in.h>
127 #include <netinet/in_systm.h>
128 #include <netinet/ip.h>
129 #include <netinet/in_pcb.h>
130 #include <netinet/in_var.h>
131 #include <netinet/ip_var.h>
132 #include <netinet/in_offload.h>
133 
134 #ifdef MROUTING
135 #include <netinet/ip_mroute.h>
136 #endif
137 
138 #include <machine/stdarg.h>
139 
140 #ifdef IPSEC
141 #include <netinet6/ipsec.h>
142 #include <netkey/key.h>
143 #include <netkey/key_debug.h>
144 #endif /*IPSEC*/
145 
146 #ifdef FAST_IPSEC
147 #include <netipsec/ipsec.h>
148 #include <netipsec/key.h>
149 #include <netipsec/xform.h>
150 #endif	/* FAST_IPSEC*/
151 
152 #ifdef IPSEC_NAT_T
153 #include <netinet/udp.h>
154 #endif
155 
156 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
157 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
158 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
159 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int);
160 
161 #ifdef PFIL_HOOKS
162 extern struct pfil_head inet_pfil_hook;			/* XXX */
163 #endif
164 
165 int	ip_do_loopback_cksum = 0;
166 
167 #define	IN_NEED_CHECKSUM(ifp, csum_flags) \
168 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
169 	(((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \
170 	(((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \
171 	(((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum)))
172 
173 struct ip_tso_output_args {
174 	struct ifnet *ifp;
175 	struct sockaddr *sa;
176 	struct rtentry *rt;
177 };
178 
179 static int ip_tso_output_callback(void *, struct mbuf *);
180 static int ip_tso_output(struct ifnet *, struct mbuf *, struct sockaddr *,
181     struct rtentry *);
182 
183 static int
184 ip_tso_output_callback(void *vp, struct mbuf *m)
185 {
186 	struct ip_tso_output_args *args = vp;
187 	struct ifnet *ifp = args->ifp;
188 
189 	return (*ifp->if_output)(ifp, m, args->sa, args->rt);
190 }
191 
192 static int
193 ip_tso_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
194     struct rtentry *rt)
195 {
196 	struct ip_tso_output_args args;
197 
198 	args.ifp = ifp;
199 	args.sa = sa;
200 	args.rt = rt;
201 
202 	return tcp4_segment(m, ip_tso_output_callback, &args);
203 }
204 
205 /*
206  * IP output.  The packet in mbuf chain m contains a skeletal IP
207  * header (with len, off, ttl, proto, tos, src, dst).
208  * The mbuf chain containing the packet will be freed.
209  * The mbuf opt, if present, will not be freed.
210  */
211 int
212 ip_output(struct mbuf *m0, ...)
213 {
214 	struct ip *ip;
215 	struct ifnet *ifp;
216 	struct mbuf *m = m0;
217 	int hlen = sizeof (struct ip);
218 	int len, error = 0;
219 	struct route iproute;
220 	struct sockaddr_in *dst;
221 	struct in_ifaddr *ia;
222 	struct mbuf *opt;
223 	struct route *ro;
224 	int flags, sw_csum;
225 	int *mtu_p;
226 	u_long mtu;
227 	struct ip_moptions *imo;
228 	struct socket *so;
229 	va_list ap;
230 #ifdef IPSEC_NAT_T
231 	int natt_frag = 0;
232 #endif
233 #ifdef IPSEC
234 	struct secpolicy *sp = NULL;
235 #endif /*IPSEC*/
236 #ifdef FAST_IPSEC
237 	struct inpcb *inp;
238 	struct m_tag *mtag;
239 	struct secpolicy *sp = NULL;
240 	struct tdb_ident *tdbi;
241 	int s;
242 #endif
243 	u_int16_t ip_len;
244 
245 	len = 0;
246 	va_start(ap, m0);
247 	opt = va_arg(ap, struct mbuf *);
248 	ro = va_arg(ap, struct route *);
249 	flags = va_arg(ap, int);
250 	imo = va_arg(ap, struct ip_moptions *);
251 	so = va_arg(ap, struct socket *);
252 	if (flags & IP_RETURNMTU)
253 		mtu_p = va_arg(ap, int *);
254 	else
255 		mtu_p = NULL;
256 	va_end(ap);
257 
258 	MCLAIM(m, &ip_tx_mowner);
259 #ifdef FAST_IPSEC
260 	if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET)
261 		inp = (struct inpcb *)so->so_pcb;
262 	else
263 		inp = NULL;
264 #endif /* FAST_IPSEC */
265 
266 #ifdef	DIAGNOSTIC
267 	if ((m->m_flags & M_PKTHDR) == 0)
268 		panic("ip_output no HDR");
269 #endif
270 	if (opt) {
271 		m = ip_insertoptions(m, opt, &len);
272 		if (len >= sizeof(struct ip))
273 			hlen = len;
274 	}
275 	ip = mtod(m, struct ip *);
276 	/*
277 	 * Fill in IP header.
278 	 */
279 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
280 		ip->ip_v = IPVERSION;
281 		ip->ip_off = htons(0);
282 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
283 			ip->ip_id = ip_newid();
284 		} else {
285 
286 			/*
287 			 * TSO capable interfaces (typically?) increment
288 			 * ip_id for each segment.
289 			 * "allocate" enough ids here to increase the chance
290 			 * for them to be unique.
291 			 *
292 			 * note that the following calculation is not
293 			 * needed to be precise.  wasting some ip_id is fine.
294 			 */
295 
296 			unsigned int segsz = m->m_pkthdr.segsz;
297 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
298 			unsigned int num = howmany(datasz, segsz);
299 
300 			ip->ip_id = ip_newid_range(num);
301 		}
302 		ip->ip_hl = hlen >> 2;
303 		ipstat.ips_localout++;
304 	} else {
305 		hlen = ip->ip_hl << 2;
306 	}
307 	/*
308 	 * Route packet.
309 	 */
310 	if (ro == 0) {
311 		ro = &iproute;
312 		bzero((caddr_t)ro, sizeof (*ro));
313 	}
314 	dst = satosin(&ro->ro_dst);
315 	/*
316 	 * If there is a cached route,
317 	 * check that it is to the same destination
318 	 * and is still up.  If not, free it and try again.
319 	 * The address family should also be checked in case of sharing the
320 	 * cache with IPv6.
321 	 */
322 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
323 	    dst->sin_family != AF_INET ||
324 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
325 		RTFREE(ro->ro_rt);
326 		ro->ro_rt = (struct rtentry *)0;
327 	}
328 	if (ro->ro_rt == 0) {
329 		bzero(dst, sizeof(*dst));
330 		dst->sin_family = AF_INET;
331 		dst->sin_len = sizeof(*dst);
332 		dst->sin_addr = ip->ip_dst;
333 	}
334 	/*
335 	 * If routing to interface only,
336 	 * short circuit routing lookup.
337 	 */
338 	if (flags & IP_ROUTETOIF) {
339 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
340 			ipstat.ips_noroute++;
341 			error = ENETUNREACH;
342 			goto bad;
343 		}
344 		ifp = ia->ia_ifp;
345 		mtu = ifp->if_mtu;
346 		ip->ip_ttl = 1;
347 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
348 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
349 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
350 		ifp = imo->imo_multicast_ifp;
351 		mtu = ifp->if_mtu;
352 		IFP_TO_IA(ifp, ia);
353 	} else {
354 		if (ro->ro_rt == 0)
355 			rtalloc(ro);
356 		if (ro->ro_rt == 0) {
357 			ipstat.ips_noroute++;
358 			error = EHOSTUNREACH;
359 			goto bad;
360 		}
361 		ia = ifatoia(ro->ro_rt->rt_ifa);
362 		ifp = ro->ro_rt->rt_ifp;
363 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
364 			mtu = ifp->if_mtu;
365 		ro->ro_rt->rt_use++;
366 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
367 			dst = satosin(ro->ro_rt->rt_gateway);
368 	}
369 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
370 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
371 		struct in_multi *inm;
372 
373 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
374 			M_BCAST : M_MCAST;
375 		/*
376 		 * IP destination address is multicast.  Make sure "dst"
377 		 * still points to the address in "ro".  (It may have been
378 		 * changed to point to a gateway address, above.)
379 		 */
380 		dst = satosin(&ro->ro_dst);
381 		/*
382 		 * See if the caller provided any multicast options
383 		 */
384 		if (imo != NULL)
385 			ip->ip_ttl = imo->imo_multicast_ttl;
386 		else
387 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
388 
389 		/*
390 		 * if we don't know the outgoing ifp yet, we can't generate
391 		 * output
392 		 */
393 		if (!ifp) {
394 			ipstat.ips_noroute++;
395 			error = ENETUNREACH;
396 			goto bad;
397 		}
398 
399 		/*
400 		 * If the packet is multicast or broadcast, confirm that
401 		 * the outgoing interface can transmit it.
402 		 */
403 		if (((m->m_flags & M_MCAST) &&
404 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
405 		    ((m->m_flags & M_BCAST) &&
406 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
407 			ipstat.ips_noroute++;
408 			error = ENETUNREACH;
409 			goto bad;
410 		}
411 		/*
412 		 * If source address not specified yet, use an address
413 		 * of outgoing interface.
414 		 */
415 		if (in_nullhost(ip->ip_src)) {
416 			struct in_ifaddr *xia;
417 
418 			IFP_TO_IA(ifp, xia);
419 			if (!xia) {
420 				error = EADDRNOTAVAIL;
421 				goto bad;
422 			}
423 			ip->ip_src = xia->ia_addr.sin_addr;
424 		}
425 
426 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
427 		if (inm != NULL &&
428 		   (imo == NULL || imo->imo_multicast_loop)) {
429 			/*
430 			 * If we belong to the destination multicast group
431 			 * on the outgoing interface, and the caller did not
432 			 * forbid loopback, loop back a copy.
433 			 */
434 			ip_mloopback(ifp, m, dst);
435 		}
436 #ifdef MROUTING
437 		else {
438 			/*
439 			 * If we are acting as a multicast router, perform
440 			 * multicast forwarding as if the packet had just
441 			 * arrived on the interface to which we are about
442 			 * to send.  The multicast forwarding function
443 			 * recursively calls this function, using the
444 			 * IP_FORWARDING flag to prevent infinite recursion.
445 			 *
446 			 * Multicasts that are looped back by ip_mloopback(),
447 			 * above, will be forwarded by the ip_input() routine,
448 			 * if necessary.
449 			 */
450 			extern struct socket *ip_mrouter;
451 
452 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
453 				if (ip_mforward(m, ifp) != 0) {
454 					m_freem(m);
455 					goto done;
456 				}
457 			}
458 		}
459 #endif
460 		/*
461 		 * Multicasts with a time-to-live of zero may be looped-
462 		 * back, above, but must not be transmitted on a network.
463 		 * Also, multicasts addressed to the loopback interface
464 		 * are not sent -- the above call to ip_mloopback() will
465 		 * loop back a copy if this host actually belongs to the
466 		 * destination group on the loopback interface.
467 		 */
468 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
469 			m_freem(m);
470 			goto done;
471 		}
472 
473 		goto sendit;
474 	}
475 	/*
476 	 * If source address not specified yet, use address
477 	 * of outgoing interface.
478 	 */
479 	if (in_nullhost(ip->ip_src))
480 		ip->ip_src = ia->ia_addr.sin_addr;
481 
482 	/*
483 	 * packets with Class-D address as source are not valid per
484 	 * RFC 1112
485 	 */
486 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
487 		ipstat.ips_odropped++;
488 		error = EADDRNOTAVAIL;
489 		goto bad;
490 	}
491 
492 	/*
493 	 * Look for broadcast address and
494 	 * and verify user is allowed to send
495 	 * such a packet.
496 	 */
497 	if (in_broadcast(dst->sin_addr, ifp)) {
498 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
499 			error = EADDRNOTAVAIL;
500 			goto bad;
501 		}
502 		if ((flags & IP_ALLOWBROADCAST) == 0) {
503 			error = EACCES;
504 			goto bad;
505 		}
506 		/* don't allow broadcast messages to be fragmented */
507 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
508 			error = EMSGSIZE;
509 			goto bad;
510 		}
511 		m->m_flags |= M_BCAST;
512 	} else
513 		m->m_flags &= ~M_BCAST;
514 
515 sendit:
516 	/*
517 	 * If we're doing Path MTU Discovery, we need to set DF unless
518 	 * the route's MTU is locked.
519 	 */
520 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
521 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
522 		ip->ip_off |= htons(IP_DF);
523 
524 	/* Remember the current ip_len */
525 	ip_len = ntohs(ip->ip_len);
526 
527 #ifdef IPSEC
528 	/* get SP for this packet */
529 	if (so == NULL)
530 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
531 		    flags, &error);
532 	else {
533 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
534 					 IPSEC_DIR_OUTBOUND))
535 			goto skip_ipsec;
536 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
537 	}
538 
539 	if (sp == NULL) {
540 		ipsecstat.out_inval++;
541 		goto bad;
542 	}
543 
544 	error = 0;
545 
546 	/* check policy */
547 	switch (sp->policy) {
548 	case IPSEC_POLICY_DISCARD:
549 		/*
550 		 * This packet is just discarded.
551 		 */
552 		ipsecstat.out_polvio++;
553 		goto bad;
554 
555 	case IPSEC_POLICY_BYPASS:
556 	case IPSEC_POLICY_NONE:
557 		/* no need to do IPsec. */
558 		goto skip_ipsec;
559 
560 	case IPSEC_POLICY_IPSEC:
561 		if (sp->req == NULL) {
562 			/* XXX should be panic ? */
563 			printf("ip_output: No IPsec request specified.\n");
564 			error = EINVAL;
565 			goto bad;
566 		}
567 		break;
568 
569 	case IPSEC_POLICY_ENTRUST:
570 	default:
571 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
572 	}
573 
574 #ifdef IPSEC_NAT_T
575 	/*
576 	 * NAT-T ESP fragmentation: don't do IPSec processing now,
577 	 * we'll do it on each fragmented packet.
578 	 */
579 	if (sp->req->sav &&
580 	    ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
581 	     (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
582 		if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
583 			natt_frag = 1;
584 			mtu = sp->req->sav->esp_frag;
585 			goto skip_ipsec;
586 		}
587 	}
588 #endif /* IPSEC_NAT_T */
589 
590 	/*
591 	 * ipsec4_output() expects ip_len and ip_off in network
592 	 * order.  They have been set to network order above.
593 	 */
594 
595     {
596 	struct ipsec_output_state state;
597 	bzero(&state, sizeof(state));
598 	state.m = m;
599 	if (flags & IP_ROUTETOIF) {
600 		state.ro = &iproute;
601 		bzero(&iproute, sizeof(iproute));
602 	} else
603 		state.ro = ro;
604 	state.dst = (struct sockaddr *)dst;
605 
606 	/*
607 	 * We can't defer the checksum of payload data if
608 	 * we're about to encrypt/authenticate it.
609 	 *
610 	 * XXX When we support crypto offloading functions of
611 	 * XXX network interfaces, we need to reconsider this,
612 	 * XXX since it's likely that they'll support checksumming,
613 	 * XXX as well.
614 	 */
615 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
616 		in_delayed_cksum(m);
617 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
618 	}
619 
620 	error = ipsec4_output(&state, sp, flags);
621 
622 	m = state.m;
623 	if (flags & IP_ROUTETOIF) {
624 		/*
625 		 * if we have tunnel mode SA, we may need to ignore
626 		 * IP_ROUTETOIF.
627 		 */
628 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
629 			flags &= ~IP_ROUTETOIF;
630 			ro = state.ro;
631 		}
632 	} else
633 		ro = state.ro;
634 	dst = (struct sockaddr_in *)state.dst;
635 	if (error) {
636 		/* mbuf is already reclaimed in ipsec4_output. */
637 		m0 = NULL;
638 		switch (error) {
639 		case EHOSTUNREACH:
640 		case ENETUNREACH:
641 		case EMSGSIZE:
642 		case ENOBUFS:
643 		case ENOMEM:
644 			break;
645 		default:
646 			printf("ip4_output (ipsec): error code %d\n", error);
647 			/*fall through*/
648 		case ENOENT:
649 			/* don't show these error codes to the user */
650 			error = 0;
651 			break;
652 		}
653 		goto bad;
654 	}
655 
656 	/* be sure to update variables that are affected by ipsec4_output() */
657 	ip = mtod(m, struct ip *);
658 	hlen = ip->ip_hl << 2;
659 	ip_len = ntohs(ip->ip_len);
660 
661 	if (ro->ro_rt == NULL) {
662 		if ((flags & IP_ROUTETOIF) == 0) {
663 			printf("ip_output: "
664 				"can't update route after IPsec processing\n");
665 			error = EHOSTUNREACH;	/*XXX*/
666 			goto bad;
667 		}
668 	} else {
669 		/* nobody uses ia beyond here */
670 		if (state.encap) {
671 			ifp = ro->ro_rt->rt_ifp;
672 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
673 				mtu = ifp->if_mtu;
674 		}
675 	}
676     }
677 skip_ipsec:
678 #endif /*IPSEC*/
679 #ifdef FAST_IPSEC
680 	/*
681 	 * Check the security policy (SP) for the packet and, if
682 	 * required, do IPsec-related processing.  There are two
683 	 * cases here; the first time a packet is sent through
684 	 * it will be untagged and handled by ipsec4_checkpolicy.
685 	 * If the packet is resubmitted to ip_output (e.g. after
686 	 * AH, ESP, etc. processing), there will be a tag to bypass
687 	 * the lookup and related policy checking.
688 	 */
689 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
690 	s = splsoftnet();
691 	if (mtag != NULL) {
692 		tdbi = (struct tdb_ident *)(mtag + 1);
693 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
694 		if (sp == NULL)
695 			error = -EINVAL;	/* force silent drop */
696 		m_tag_delete(m, mtag);
697 	} else {
698 		if (inp != NULL &&
699 		    IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND))
700 			goto spd_done;
701 		sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
702 					&error, inp);
703 	}
704 	/*
705 	 * There are four return cases:
706 	 *    sp != NULL	 	    apply IPsec policy
707 	 *    sp == NULL, error == 0	    no IPsec handling needed
708 	 *    sp == NULL, error == -EINVAL  discard packet w/o error
709 	 *    sp == NULL, error != 0	    discard packet, report error
710 	 */
711 	if (sp != NULL) {
712 #ifdef IPSEC_NAT_T
713 		/*
714 		 * NAT-T ESP fragmentation: don't do IPSec processing now,
715 		 * we'll do it on each fragmented packet.
716 		 */
717 		if (sp->req->sav &&
718 		    ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
719 		     (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
720 			if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
721 				natt_frag = 1;
722 				mtu = sp->req->sav->esp_frag;
723 				goto spd_done;
724 			}
725 		}
726 #endif /* IPSEC_NAT_T */
727 		/* Loop detection, check if ipsec processing already done */
728 		IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
729 		for (mtag = m_tag_first(m); mtag != NULL;
730 		     mtag = m_tag_next(m, mtag)) {
731 #ifdef MTAG_ABI_COMPAT
732 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
733 				continue;
734 #endif
735 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
736 			    mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
737 				continue;
738 			/*
739 			 * Check if policy has an SA associated with it.
740 			 * This can happen when an SP has yet to acquire
741 			 * an SA; e.g. on first reference.  If it occurs,
742 			 * then we let ipsec4_process_packet do its thing.
743 			 */
744 			if (sp->req->sav == NULL)
745 				break;
746 			tdbi = (struct tdb_ident *)(mtag + 1);
747 			if (tdbi->spi == sp->req->sav->spi &&
748 			    tdbi->proto == sp->req->sav->sah->saidx.proto &&
749 			    bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
750 				 sizeof (union sockaddr_union)) == 0) {
751 				/*
752 				 * No IPsec processing is needed, free
753 				 * reference to SP.
754 				 *
755 				 * NB: null pointer to avoid free at
756 				 *     done: below.
757 				 */
758 				KEY_FREESP(&sp), sp = NULL;
759 				splx(s);
760 				goto spd_done;
761 			}
762 		}
763 
764 		/*
765 		 * Do delayed checksums now because we send before
766 		 * this is done in the normal processing path.
767 		 */
768 		if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
769 			in_delayed_cksum(m);
770 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
771 		}
772 
773 #ifdef __FreeBSD__
774 		ip->ip_len = htons(ip->ip_len);
775 		ip->ip_off = htons(ip->ip_off);
776 #endif
777 
778 		/* NB: callee frees mbuf */
779 		error = ipsec4_process_packet(m, sp->req, flags, 0);
780 		/*
781 		 * Preserve KAME behaviour: ENOENT can be returned
782 		 * when an SA acquire is in progress.  Don't propagate
783 		 * this to user-level; it confuses applications.
784 		 *
785 		 * XXX this will go away when the SADB is redone.
786 		 */
787 		if (error == ENOENT)
788 			error = 0;
789 		splx(s);
790 		goto done;
791 	} else {
792 		splx(s);
793 
794 		if (error != 0) {
795 			/*
796 			 * Hack: -EINVAL is used to signal that a packet
797 			 * should be silently discarded.  This is typically
798 			 * because we asked key management for an SA and
799 			 * it was delayed (e.g. kicked up to IKE).
800 			 */
801 			if (error == -EINVAL)
802 				error = 0;
803 			goto bad;
804 		} else {
805 			/* No IPsec processing for this packet. */
806 		}
807 #ifdef notyet
808 		/*
809 		 * If deferred crypto processing is needed, check that
810 		 * the interface supports it.
811 		 */
812 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
813 		if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
814 			/* notify IPsec to do its own crypto */
815 			ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
816 			error = EHOSTUNREACH;
817 			goto bad;
818 		}
819 #endif
820 	}
821 spd_done:
822 #endif /* FAST_IPSEC */
823 
824 #ifdef PFIL_HOOKS
825 	/*
826 	 * Run through list of hooks for output packets.
827 	 */
828 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
829 		goto done;
830 	if (m == NULL)
831 		goto done;
832 
833 	ip = mtod(m, struct ip *);
834 	hlen = ip->ip_hl << 2;
835 #endif /* PFIL_HOOKS */
836 
837 	m->m_pkthdr.csum_data |= hlen << 16;
838 
839 #if IFA_STATS
840 	/*
841 	 * search for the source address structure to
842 	 * maintain output statistics.
843 	 */
844 	INADDR_TO_IA(ip->ip_src, ia);
845 #endif
846 
847 	/* Maybe skip checksums on loopback interfaces. */
848 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
849 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
850 	}
851 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
852 	/*
853 	 * If small enough for mtu of path, or if using TCP segmentation
854 	 * offload, can just send directly.
855 	 */
856 	if (ip_len <= mtu ||
857 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
858 #if IFA_STATS
859 		if (ia)
860 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
861 #endif
862 		/*
863 		 * Always initialize the sum to 0!  Some HW assisted
864 		 * checksumming requires this.
865 		 */
866 		ip->ip_sum = 0;
867 
868 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
869 			/*
870 			 * Perform any checksums that the hardware can't do
871 			 * for us.
872 			 *
873 			 * XXX Does any hardware require the {th,uh}_sum
874 			 * XXX fields to be 0?
875 			 */
876 			if (sw_csum & M_CSUM_IPv4) {
877 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
878 				ip->ip_sum = in_cksum(m, hlen);
879 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
880 			}
881 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
882 				if (IN_NEED_CHECKSUM(ifp,
883 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
884 					in_delayed_cksum(m);
885 				}
886 				m->m_pkthdr.csum_flags &=
887 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
888 			}
889 		}
890 
891 #ifdef IPSEC
892 		/* clean ipsec history once it goes out of the node */
893 		ipsec_delaux(m);
894 #endif
895 
896 		if (__predict_true(
897 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
898 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
899 			error =
900 			    (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
901 		} else {
902 			error =
903 			    ip_tso_output(ifp, m, sintosa(dst), ro->ro_rt);
904 		}
905 		goto done;
906 	}
907 
908 	/*
909 	 * We can't use HW checksumming if we're about to
910 	 * to fragment the packet.
911 	 *
912 	 * XXX Some hardware can do this.
913 	 */
914 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
915 		if (IN_NEED_CHECKSUM(ifp,
916 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
917 			in_delayed_cksum(m);
918 		}
919 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
920 	}
921 
922 	/*
923 	 * Too large for interface; fragment if possible.
924 	 * Must be able to put at least 8 bytes per fragment.
925 	 */
926 	if (ntohs(ip->ip_off) & IP_DF) {
927 		if (flags & IP_RETURNMTU)
928 			*mtu_p = mtu;
929 		error = EMSGSIZE;
930 		ipstat.ips_cantfrag++;
931 		goto bad;
932 	}
933 
934 	error = ip_fragment(m, ifp, mtu);
935 	if (error) {
936 		m = NULL;
937 		goto bad;
938 	}
939 
940 	for (; m; m = m0) {
941 		m0 = m->m_nextpkt;
942 		m->m_nextpkt = 0;
943 		if (error == 0) {
944 #if IFA_STATS
945 			if (ia)
946 				ia->ia_ifa.ifa_data.ifad_outbytes +=
947 				    ntohs(ip->ip_len);
948 #endif
949 #ifdef IPSEC
950 			/* clean ipsec history once it goes out of the node */
951 			ipsec_delaux(m);
952 #endif /* IPSEC */
953 
954 #ifdef IPSEC_NAT_T
955 			/*
956 			 * If we get there, the packet has not been handeld by
957 			 * IPSec whereas it should have. Now that it has been
958 			 * fragmented, re-inject it in ip_output so that IPsec
959 			 * processing can occur.
960 			 */
961 			if (natt_frag) {
962 				error = ip_output(m, opt,
963 				    ro, flags, imo, so, mtu_p);
964 			} else
965 #endif /* IPSEC_NAT_T */
966 			{
967 				KASSERT((m->m_pkthdr.csum_flags &
968 				    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
969 				error = (*ifp->if_output)(ifp, m, sintosa(dst),
970 				    ro->ro_rt);
971 			}
972 		} else
973 			m_freem(m);
974 	}
975 
976 	if (error == 0)
977 		ipstat.ips_fragmented++;
978 done:
979 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
980 		RTFREE(ro->ro_rt);
981 		ro->ro_rt = 0;
982 	}
983 
984 #ifdef IPSEC
985 	if (sp != NULL) {
986 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
987 			printf("DP ip_output call free SP:%p\n", sp));
988 		key_freesp(sp);
989 	}
990 #endif /* IPSEC */
991 #ifdef FAST_IPSEC
992 	if (sp != NULL)
993 		KEY_FREESP(&sp);
994 #endif /* FAST_IPSEC */
995 
996 	return (error);
997 bad:
998 	m_freem(m);
999 	goto done;
1000 }
1001 
1002 int
1003 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
1004 {
1005 	struct ip *ip, *mhip;
1006 	struct mbuf *m0;
1007 	int len, hlen, off;
1008 	int mhlen, firstlen;
1009 	struct mbuf **mnext;
1010 	int sw_csum = m->m_pkthdr.csum_flags;
1011 	int fragments = 0;
1012 	int s;
1013 	int error = 0;
1014 
1015 	ip = mtod(m, struct ip *);
1016 	hlen = ip->ip_hl << 2;
1017 	if (ifp != NULL)
1018 		sw_csum &= ~ifp->if_csum_flags_tx;
1019 
1020 	len = (mtu - hlen) &~ 7;
1021 	if (len < 8) {
1022 		m_freem(m);
1023 		return (EMSGSIZE);
1024 	}
1025 
1026 	firstlen = len;
1027 	mnext = &m->m_nextpkt;
1028 
1029 	/*
1030 	 * Loop through length of segment after first fragment,
1031 	 * make new header and copy data of each part and link onto chain.
1032 	 */
1033 	m0 = m;
1034 	mhlen = sizeof (struct ip);
1035 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
1036 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
1037 		if (m == 0) {
1038 			error = ENOBUFS;
1039 			ipstat.ips_odropped++;
1040 			goto sendorfree;
1041 		}
1042 		MCLAIM(m, m0->m_owner);
1043 		*mnext = m;
1044 		mnext = &m->m_nextpkt;
1045 		m->m_data += max_linkhdr;
1046 		mhip = mtod(m, struct ip *);
1047 		*mhip = *ip;
1048 		/* we must inherit MCAST and BCAST flags */
1049 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
1050 		if (hlen > sizeof (struct ip)) {
1051 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1052 			mhip->ip_hl = mhlen >> 2;
1053 		}
1054 		m->m_len = mhlen;
1055 		mhip->ip_off = ((off - hlen) >> 3) +
1056 		    (ntohs(ip->ip_off) & ~IP_MF);
1057 		if (ip->ip_off & htons(IP_MF))
1058 			mhip->ip_off |= IP_MF;
1059 		if (off + len >= ntohs(ip->ip_len))
1060 			len = ntohs(ip->ip_len) - off;
1061 		else
1062 			mhip->ip_off |= IP_MF;
1063 		HTONS(mhip->ip_off);
1064 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
1065 		m->m_next = m_copy(m0, off, len);
1066 		if (m->m_next == 0) {
1067 			error = ENOBUFS;	/* ??? */
1068 			ipstat.ips_odropped++;
1069 			goto sendorfree;
1070 		}
1071 		m->m_pkthdr.len = mhlen + len;
1072 		m->m_pkthdr.rcvif = (struct ifnet *)0;
1073 		mhip->ip_sum = 0;
1074 		if (sw_csum & M_CSUM_IPv4) {
1075 			mhip->ip_sum = in_cksum(m, mhlen);
1076 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
1077 		} else {
1078 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
1079 			m->m_pkthdr.csum_data |= mhlen << 16;
1080 		}
1081 		ipstat.ips_ofragments++;
1082 		fragments++;
1083 	}
1084 	/*
1085 	 * Update first fragment by trimming what's been copied out
1086 	 * and updating header, then send each fragment (in order).
1087 	 */
1088 	m = m0;
1089 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
1090 	m->m_pkthdr.len = hlen + firstlen;
1091 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
1092 	ip->ip_off |= htons(IP_MF);
1093 	ip->ip_sum = 0;
1094 	if (sw_csum & M_CSUM_IPv4) {
1095 		ip->ip_sum = in_cksum(m, hlen);
1096 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
1097 	} else {
1098 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
1099 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
1100 			sizeof(struct ip));
1101 	}
1102 sendorfree:
1103 	/*
1104 	 * If there is no room for all the fragments, don't queue
1105 	 * any of them.
1106 	 */
1107 	if (ifp != NULL) {
1108 		s = splnet();
1109 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
1110 		    error == 0) {
1111 			error = ENOBUFS;
1112 			ipstat.ips_odropped++;
1113 			IFQ_INC_DROPS(&ifp->if_snd);
1114 		}
1115 		splx(s);
1116 	}
1117 	if (error) {
1118 		for (m = m0; m; m = m0) {
1119 			m0 = m->m_nextpkt;
1120 			m->m_nextpkt = NULL;
1121 			m_freem(m);
1122 		}
1123 	}
1124 	return (error);
1125 }
1126 
1127 /*
1128  * Process a delayed payload checksum calculation.
1129  */
1130 void
1131 in_delayed_cksum(struct mbuf *m)
1132 {
1133 	struct ip *ip;
1134 	u_int16_t csum, offset;
1135 
1136 	ip = mtod(m, struct ip *);
1137 	offset = ip->ip_hl << 2;
1138 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
1139 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
1140 		csum = 0xffff;
1141 
1142 	offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
1143 
1144 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
1145 		/* This happen when ip options were inserted
1146 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
1147 		    m->m_len, offset, ip->ip_p);
1148 		 */
1149 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
1150 	} else
1151 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1152 }
1153 
1154 /*
1155  * Determine the maximum length of the options to be inserted;
1156  * we would far rather allocate too much space rather than too little.
1157  */
1158 
1159 u_int
1160 ip_optlen(struct inpcb *inp)
1161 {
1162 	struct mbuf *m = inp->inp_options;
1163 
1164 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
1165 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1166 	else
1167 		return 0;
1168 }
1169 
1170 
1171 /*
1172  * Insert IP options into preformed packet.
1173  * Adjust IP destination as required for IP source routing,
1174  * as indicated by a non-zero in_addr at the start of the options.
1175  */
1176 static struct mbuf *
1177 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1178 {
1179 	struct ipoption *p = mtod(opt, struct ipoption *);
1180 	struct mbuf *n;
1181 	struct ip *ip = mtod(m, struct ip *);
1182 	unsigned optlen;
1183 
1184 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1185 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1186 		return (m);		/* XXX should fail */
1187 	if (!in_nullhost(p->ipopt_dst))
1188 		ip->ip_dst = p->ipopt_dst;
1189 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1190 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1191 		if (n == 0)
1192 			return (m);
1193 		MCLAIM(n, m->m_owner);
1194 		M_MOVE_PKTHDR(n, m);
1195 		m->m_len -= sizeof(struct ip);
1196 		m->m_data += sizeof(struct ip);
1197 		n->m_next = m;
1198 		m = n;
1199 		m->m_len = optlen + sizeof(struct ip);
1200 		m->m_data += max_linkhdr;
1201 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1202 	} else {
1203 		m->m_data -= optlen;
1204 		m->m_len += optlen;
1205 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
1206 	}
1207 	m->m_pkthdr.len += optlen;
1208 	ip = mtod(m, struct ip *);
1209 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1210 	*phlen = sizeof(struct ip) + optlen;
1211 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1212 	return (m);
1213 }
1214 
1215 /*
1216  * Copy options from ip to jp,
1217  * omitting those not copied during fragmentation.
1218  */
1219 int
1220 ip_optcopy(struct ip *ip, struct ip *jp)
1221 {
1222 	u_char *cp, *dp;
1223 	int opt, optlen, cnt;
1224 
1225 	cp = (u_char *)(ip + 1);
1226 	dp = (u_char *)(jp + 1);
1227 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1228 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1229 		opt = cp[0];
1230 		if (opt == IPOPT_EOL)
1231 			break;
1232 		if (opt == IPOPT_NOP) {
1233 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1234 			*dp++ = IPOPT_NOP;
1235 			optlen = 1;
1236 			continue;
1237 		}
1238 #ifdef DIAGNOSTIC
1239 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1240 			panic("malformed IPv4 option passed to ip_optcopy");
1241 #endif
1242 		optlen = cp[IPOPT_OLEN];
1243 #ifdef DIAGNOSTIC
1244 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1245 			panic("malformed IPv4 option passed to ip_optcopy");
1246 #endif
1247 		/* bogus lengths should have been caught by ip_dooptions */
1248 		if (optlen > cnt)
1249 			optlen = cnt;
1250 		if (IPOPT_COPIED(opt)) {
1251 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1252 			dp += optlen;
1253 		}
1254 	}
1255 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1256 		*dp++ = IPOPT_EOL;
1257 	return (optlen);
1258 }
1259 
1260 /*
1261  * IP socket option processing.
1262  */
1263 int
1264 ip_ctloutput(int op, struct socket *so, int level, int optname,
1265     struct mbuf **mp)
1266 {
1267 	struct inpcb *inp = sotoinpcb(so);
1268 	struct mbuf *m = *mp;
1269 	int optval = 0;
1270 	int error = 0;
1271 #if defined(IPSEC) || defined(FAST_IPSEC)
1272 	struct proc *p = curproc;	/*XXX*/
1273 #endif
1274 
1275 	if (level != IPPROTO_IP) {
1276 		error = EINVAL;
1277 		if (op == PRCO_SETOPT && *mp)
1278 			(void) m_free(*mp);
1279 	} else switch (op) {
1280 
1281 	case PRCO_SETOPT:
1282 		switch (optname) {
1283 		case IP_OPTIONS:
1284 #ifdef notyet
1285 		case IP_RETOPTS:
1286 			return (ip_pcbopts(optname, &inp->inp_options, m));
1287 #else
1288 			return (ip_pcbopts(&inp->inp_options, m));
1289 #endif
1290 
1291 		case IP_TOS:
1292 		case IP_TTL:
1293 		case IP_RECVOPTS:
1294 		case IP_RECVRETOPTS:
1295 		case IP_RECVDSTADDR:
1296 		case IP_RECVIF:
1297 			if (m == NULL || m->m_len != sizeof(int))
1298 				error = EINVAL;
1299 			else {
1300 				optval = *mtod(m, int *);
1301 				switch (optname) {
1302 
1303 				case IP_TOS:
1304 					inp->inp_ip.ip_tos = optval;
1305 					break;
1306 
1307 				case IP_TTL:
1308 					inp->inp_ip.ip_ttl = optval;
1309 					break;
1310 #define	OPTSET(bit) \
1311 	if (optval) \
1312 		inp->inp_flags |= bit; \
1313 	else \
1314 		inp->inp_flags &= ~bit;
1315 
1316 				case IP_RECVOPTS:
1317 					OPTSET(INP_RECVOPTS);
1318 					break;
1319 
1320 				case IP_RECVRETOPTS:
1321 					OPTSET(INP_RECVRETOPTS);
1322 					break;
1323 
1324 				case IP_RECVDSTADDR:
1325 					OPTSET(INP_RECVDSTADDR);
1326 					break;
1327 
1328 				case IP_RECVIF:
1329 					OPTSET(INP_RECVIF);
1330 					break;
1331 				}
1332 			}
1333 			break;
1334 #undef OPTSET
1335 
1336 		case IP_MULTICAST_IF:
1337 		case IP_MULTICAST_TTL:
1338 		case IP_MULTICAST_LOOP:
1339 		case IP_ADD_MEMBERSHIP:
1340 		case IP_DROP_MEMBERSHIP:
1341 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1342 			break;
1343 
1344 		case IP_PORTRANGE:
1345 			if (m == 0 || m->m_len != sizeof(int))
1346 				error = EINVAL;
1347 			else {
1348 				optval = *mtod(m, int *);
1349 
1350 				switch (optval) {
1351 
1352 				case IP_PORTRANGE_DEFAULT:
1353 				case IP_PORTRANGE_HIGH:
1354 					inp->inp_flags &= ~(INP_LOWPORT);
1355 					break;
1356 
1357 				case IP_PORTRANGE_LOW:
1358 					inp->inp_flags |= INP_LOWPORT;
1359 					break;
1360 
1361 				default:
1362 					error = EINVAL;
1363 					break;
1364 				}
1365 			}
1366 			break;
1367 
1368 #if defined(IPSEC) || defined(FAST_IPSEC)
1369 		case IP_IPSEC_POLICY:
1370 		{
1371 			caddr_t req = NULL;
1372 			size_t len = 0;
1373 			int priv = 0;
1374 
1375 #ifdef __NetBSD__
1376 			if (p == 0 || kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
1377 							&p->p_acflag))
1378 				priv = 0;
1379 			else
1380 				priv = 1;
1381 #else
1382 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1383 #endif
1384 			if (m) {
1385 				req = mtod(m, caddr_t);
1386 				len = m->m_len;
1387 			}
1388 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1389 			break;
1390 		    }
1391 #endif /*IPSEC*/
1392 
1393 		default:
1394 			error = ENOPROTOOPT;
1395 			break;
1396 		}
1397 		if (m)
1398 			(void)m_free(m);
1399 		break;
1400 
1401 	case PRCO_GETOPT:
1402 		switch (optname) {
1403 		case IP_OPTIONS:
1404 		case IP_RETOPTS:
1405 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1406 			MCLAIM(m, so->so_mowner);
1407 			if (inp->inp_options) {
1408 				m->m_len = inp->inp_options->m_len;
1409 				bcopy(mtod(inp->inp_options, caddr_t),
1410 				    mtod(m, caddr_t), (unsigned)m->m_len);
1411 			} else
1412 				m->m_len = 0;
1413 			break;
1414 
1415 		case IP_TOS:
1416 		case IP_TTL:
1417 		case IP_RECVOPTS:
1418 		case IP_RECVRETOPTS:
1419 		case IP_RECVDSTADDR:
1420 		case IP_RECVIF:
1421 		case IP_ERRORMTU:
1422 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1423 			MCLAIM(m, so->so_mowner);
1424 			m->m_len = sizeof(int);
1425 			switch (optname) {
1426 
1427 			case IP_TOS:
1428 				optval = inp->inp_ip.ip_tos;
1429 				break;
1430 
1431 			case IP_TTL:
1432 				optval = inp->inp_ip.ip_ttl;
1433 				break;
1434 
1435 			case IP_ERRORMTU:
1436 				optval = inp->inp_errormtu;
1437 				break;
1438 
1439 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1440 
1441 			case IP_RECVOPTS:
1442 				optval = OPTBIT(INP_RECVOPTS);
1443 				break;
1444 
1445 			case IP_RECVRETOPTS:
1446 				optval = OPTBIT(INP_RECVRETOPTS);
1447 				break;
1448 
1449 			case IP_RECVDSTADDR:
1450 				optval = OPTBIT(INP_RECVDSTADDR);
1451 				break;
1452 
1453 			case IP_RECVIF:
1454 				optval = OPTBIT(INP_RECVIF);
1455 				break;
1456 			}
1457 			*mtod(m, int *) = optval;
1458 			break;
1459 
1460 #if 0	/* defined(IPSEC) || defined(FAST_IPSEC) */
1461 		/* XXX: code broken */
1462 		case IP_IPSEC_POLICY:
1463 		{
1464 			caddr_t req = NULL;
1465 			size_t len = 0;
1466 
1467 			if (m) {
1468 				req = mtod(m, caddr_t);
1469 				len = m->m_len;
1470 			}
1471 			error = ipsec4_get_policy(inp, req, len, mp);
1472 			break;
1473 		}
1474 #endif /*IPSEC*/
1475 
1476 		case IP_MULTICAST_IF:
1477 		case IP_MULTICAST_TTL:
1478 		case IP_MULTICAST_LOOP:
1479 		case IP_ADD_MEMBERSHIP:
1480 		case IP_DROP_MEMBERSHIP:
1481 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1482 			if (*mp)
1483 				MCLAIM(*mp, so->so_mowner);
1484 			break;
1485 
1486 		case IP_PORTRANGE:
1487 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1488 			MCLAIM(m, so->so_mowner);
1489 			m->m_len = sizeof(int);
1490 
1491 			if (inp->inp_flags & INP_LOWPORT)
1492 				optval = IP_PORTRANGE_LOW;
1493 			else
1494 				optval = IP_PORTRANGE_DEFAULT;
1495 
1496 			*mtod(m, int *) = optval;
1497 			break;
1498 
1499 		default:
1500 			error = ENOPROTOOPT;
1501 			break;
1502 		}
1503 		break;
1504 	}
1505 	return (error);
1506 }
1507 
1508 /*
1509  * Set up IP options in pcb for insertion in output packets.
1510  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1511  * with destination address if source routed.
1512  */
1513 int
1514 #ifdef notyet
1515 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1516 #else
1517 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1518 #endif
1519 {
1520 	int cnt, optlen;
1521 	u_char *cp;
1522 	u_char opt;
1523 
1524 	/* turn off any old options */
1525 	if (*pcbopt)
1526 		(void)m_free(*pcbopt);
1527 	*pcbopt = 0;
1528 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1529 		/*
1530 		 * Only turning off any previous options.
1531 		 */
1532 		if (m)
1533 			(void)m_free(m);
1534 		return (0);
1535 	}
1536 
1537 #ifndef	__vax__
1538 	if (m->m_len % sizeof(int32_t))
1539 		goto bad;
1540 #endif
1541 	/*
1542 	 * IP first-hop destination address will be stored before
1543 	 * actual options; move other options back
1544 	 * and clear it when none present.
1545 	 */
1546 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1547 		goto bad;
1548 	cnt = m->m_len;
1549 	m->m_len += sizeof(struct in_addr);
1550 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1551 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1552 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1553 
1554 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1555 		opt = cp[IPOPT_OPTVAL];
1556 		if (opt == IPOPT_EOL)
1557 			break;
1558 		if (opt == IPOPT_NOP)
1559 			optlen = 1;
1560 		else {
1561 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1562 				goto bad;
1563 			optlen = cp[IPOPT_OLEN];
1564 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1565 				goto bad;
1566 		}
1567 		switch (opt) {
1568 
1569 		default:
1570 			break;
1571 
1572 		case IPOPT_LSRR:
1573 		case IPOPT_SSRR:
1574 			/*
1575 			 * user process specifies route as:
1576 			 *	->A->B->C->D
1577 			 * D must be our final destination (but we can't
1578 			 * check that since we may not have connected yet).
1579 			 * A is first hop destination, which doesn't appear in
1580 			 * actual IP option, but is stored before the options.
1581 			 */
1582 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1583 				goto bad;
1584 			m->m_len -= sizeof(struct in_addr);
1585 			cnt -= sizeof(struct in_addr);
1586 			optlen -= sizeof(struct in_addr);
1587 			cp[IPOPT_OLEN] = optlen;
1588 			/*
1589 			 * Move first hop before start of options.
1590 			 */
1591 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1592 			    sizeof(struct in_addr));
1593 			/*
1594 			 * Then copy rest of options back
1595 			 * to close up the deleted entry.
1596 			 */
1597 			(void)memmove(&cp[IPOPT_OFFSET+1],
1598 			    &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1599 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
1600 			break;
1601 		}
1602 	}
1603 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1604 		goto bad;
1605 	*pcbopt = m;
1606 	return (0);
1607 
1608 bad:
1609 	(void)m_free(m);
1610 	return (EINVAL);
1611 }
1612 
1613 /*
1614  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1615  */
1616 static struct ifnet *
1617 ip_multicast_if(struct in_addr *a, int *ifindexp)
1618 {
1619 	int ifindex;
1620 	struct ifnet *ifp = NULL;
1621 	struct in_ifaddr *ia;
1622 
1623 	if (ifindexp)
1624 		*ifindexp = 0;
1625 	if (ntohl(a->s_addr) >> 24 == 0) {
1626 		ifindex = ntohl(a->s_addr) & 0xffffff;
1627 		if (ifindex < 0 || if_indexlim <= ifindex)
1628 			return NULL;
1629 		ifp = ifindex2ifnet[ifindex];
1630 		if (!ifp)
1631 			return NULL;
1632 		if (ifindexp)
1633 			*ifindexp = ifindex;
1634 	} else {
1635 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1636 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1637 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1638 				ifp = ia->ia_ifp;
1639 				break;
1640 			}
1641 		}
1642 	}
1643 	return ifp;
1644 }
1645 
1646 static int
1647 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval)
1648 {
1649 	u_int tval;
1650 
1651 	if (m == NULL)
1652 		return EINVAL;
1653 
1654 	switch (m->m_len) {
1655 	case sizeof(u_char):
1656 		tval = *(mtod(m, u_char *));
1657 		break;
1658 	case sizeof(u_int):
1659 		tval = *(mtod(m, u_int *));
1660 		break;
1661 	default:
1662 		return EINVAL;
1663 	}
1664 
1665 	if (tval > maxval)
1666 		return EINVAL;
1667 
1668 	*val = tval;
1669 	return 0;
1670 }
1671 
1672 /*
1673  * Set the IP multicast options in response to user setsockopt().
1674  */
1675 int
1676 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m)
1677 {
1678 	int error = 0;
1679 	int i;
1680 	struct in_addr addr;
1681 	struct ip_mreq *mreq;
1682 	struct ifnet *ifp;
1683 	struct ip_moptions *imo = *imop;
1684 	struct route ro;
1685 	struct sockaddr_in *dst;
1686 	int ifindex;
1687 
1688 	if (imo == NULL) {
1689 		/*
1690 		 * No multicast option buffer attached to the pcb;
1691 		 * allocate one and initialize to default values.
1692 		 */
1693 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1694 		    M_WAITOK);
1695 
1696 		if (imo == NULL)
1697 			return (ENOBUFS);
1698 		*imop = imo;
1699 		imo->imo_multicast_ifp = NULL;
1700 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1701 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1702 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1703 		imo->imo_num_memberships = 0;
1704 	}
1705 
1706 	switch (optname) {
1707 
1708 	case IP_MULTICAST_IF:
1709 		/*
1710 		 * Select the interface for outgoing multicast packets.
1711 		 */
1712 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1713 			error = EINVAL;
1714 			break;
1715 		}
1716 		addr = *(mtod(m, struct in_addr *));
1717 		/*
1718 		 * INADDR_ANY is used to remove a previous selection.
1719 		 * When no interface is selected, a default one is
1720 		 * chosen every time a multicast packet is sent.
1721 		 */
1722 		if (in_nullhost(addr)) {
1723 			imo->imo_multicast_ifp = NULL;
1724 			break;
1725 		}
1726 		/*
1727 		 * The selected interface is identified by its local
1728 		 * IP address.  Find the interface and confirm that
1729 		 * it supports multicasting.
1730 		 */
1731 		ifp = ip_multicast_if(&addr, &ifindex);
1732 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1733 			error = EADDRNOTAVAIL;
1734 			break;
1735 		}
1736 		imo->imo_multicast_ifp = ifp;
1737 		if (ifindex)
1738 			imo->imo_multicast_addr = addr;
1739 		else
1740 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1741 		break;
1742 
1743 	case IP_MULTICAST_TTL:
1744 		/*
1745 		 * Set the IP time-to-live for outgoing multicast packets.
1746 		 */
1747 		error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL);
1748 		break;
1749 
1750 	case IP_MULTICAST_LOOP:
1751 		/*
1752 		 * Set the loopback flag for outgoing multicast packets.
1753 		 * Must be zero or one.
1754 		 */
1755 		error = ip_getoptval(m, &imo->imo_multicast_loop, 1);
1756 		break;
1757 
1758 	case IP_ADD_MEMBERSHIP:
1759 		/*
1760 		 * Add a multicast group membership.
1761 		 * Group must be a valid IP multicast address.
1762 		 */
1763 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1764 			error = EINVAL;
1765 			break;
1766 		}
1767 		mreq = mtod(m, struct ip_mreq *);
1768 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1769 			error = EINVAL;
1770 			break;
1771 		}
1772 		/*
1773 		 * If no interface address was provided, use the interface of
1774 		 * the route to the given multicast address.
1775 		 */
1776 		if (in_nullhost(mreq->imr_interface)) {
1777 			bzero((caddr_t)&ro, sizeof(ro));
1778 			ro.ro_rt = NULL;
1779 			dst = satosin(&ro.ro_dst);
1780 			dst->sin_len = sizeof(*dst);
1781 			dst->sin_family = AF_INET;
1782 			dst->sin_addr = mreq->imr_multiaddr;
1783 			rtalloc(&ro);
1784 			if (ro.ro_rt == NULL) {
1785 				error = EADDRNOTAVAIL;
1786 				break;
1787 			}
1788 			ifp = ro.ro_rt->rt_ifp;
1789 			rtfree(ro.ro_rt);
1790 		} else {
1791 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1792 		}
1793 		/*
1794 		 * See if we found an interface, and confirm that it
1795 		 * supports multicast.
1796 		 */
1797 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1798 			error = EADDRNOTAVAIL;
1799 			break;
1800 		}
1801 		/*
1802 		 * See if the membership already exists or if all the
1803 		 * membership slots are full.
1804 		 */
1805 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1806 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1807 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1808 				      mreq->imr_multiaddr))
1809 				break;
1810 		}
1811 		if (i < imo->imo_num_memberships) {
1812 			error = EADDRINUSE;
1813 			break;
1814 		}
1815 		if (i == IP_MAX_MEMBERSHIPS) {
1816 			error = ETOOMANYREFS;
1817 			break;
1818 		}
1819 		/*
1820 		 * Everything looks good; add a new record to the multicast
1821 		 * address list for the given interface.
1822 		 */
1823 		if ((imo->imo_membership[i] =
1824 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1825 			error = ENOBUFS;
1826 			break;
1827 		}
1828 		++imo->imo_num_memberships;
1829 		break;
1830 
1831 	case IP_DROP_MEMBERSHIP:
1832 		/*
1833 		 * Drop a multicast group membership.
1834 		 * Group must be a valid IP multicast address.
1835 		 */
1836 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1837 			error = EINVAL;
1838 			break;
1839 		}
1840 		mreq = mtod(m, struct ip_mreq *);
1841 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1842 			error = EINVAL;
1843 			break;
1844 		}
1845 		/*
1846 		 * If an interface address was specified, get a pointer
1847 		 * to its ifnet structure.
1848 		 */
1849 		if (in_nullhost(mreq->imr_interface))
1850 			ifp = NULL;
1851 		else {
1852 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1853 			if (ifp == NULL) {
1854 				error = EADDRNOTAVAIL;
1855 				break;
1856 			}
1857 		}
1858 		/*
1859 		 * Find the membership in the membership array.
1860 		 */
1861 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1862 			if ((ifp == NULL ||
1863 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1864 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1865 				       mreq->imr_multiaddr))
1866 				break;
1867 		}
1868 		if (i == imo->imo_num_memberships) {
1869 			error = EADDRNOTAVAIL;
1870 			break;
1871 		}
1872 		/*
1873 		 * Give up the multicast address record to which the
1874 		 * membership points.
1875 		 */
1876 		in_delmulti(imo->imo_membership[i]);
1877 		/*
1878 		 * Remove the gap in the membership array.
1879 		 */
1880 		for (++i; i < imo->imo_num_memberships; ++i)
1881 			imo->imo_membership[i-1] = imo->imo_membership[i];
1882 		--imo->imo_num_memberships;
1883 		break;
1884 
1885 	default:
1886 		error = EOPNOTSUPP;
1887 		break;
1888 	}
1889 
1890 	/*
1891 	 * If all options have default values, no need to keep the mbuf.
1892 	 */
1893 	if (imo->imo_multicast_ifp == NULL &&
1894 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1895 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1896 	    imo->imo_num_memberships == 0) {
1897 		free(*imop, M_IPMOPTS);
1898 		*imop = NULL;
1899 	}
1900 
1901 	return (error);
1902 }
1903 
1904 /*
1905  * Return the IP multicast options in response to user getsockopt().
1906  */
1907 int
1908 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
1909 {
1910 	u_char *ttl;
1911 	u_char *loop;
1912 	struct in_addr *addr;
1913 	struct in_ifaddr *ia;
1914 
1915 	*mp = m_get(M_WAIT, MT_SOOPTS);
1916 
1917 	switch (optname) {
1918 
1919 	case IP_MULTICAST_IF:
1920 		addr = mtod(*mp, struct in_addr *);
1921 		(*mp)->m_len = sizeof(struct in_addr);
1922 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1923 			*addr = zeroin_addr;
1924 		else if (imo->imo_multicast_addr.s_addr) {
1925 			/* return the value user has set */
1926 			*addr = imo->imo_multicast_addr;
1927 		} else {
1928 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1929 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1930 		}
1931 		return (0);
1932 
1933 	case IP_MULTICAST_TTL:
1934 		ttl = mtod(*mp, u_char *);
1935 		(*mp)->m_len = 1;
1936 		*ttl = imo ? imo->imo_multicast_ttl
1937 			   : IP_DEFAULT_MULTICAST_TTL;
1938 		return (0);
1939 
1940 	case IP_MULTICAST_LOOP:
1941 		loop = mtod(*mp, u_char *);
1942 		(*mp)->m_len = 1;
1943 		*loop = imo ? imo->imo_multicast_loop
1944 			    : IP_DEFAULT_MULTICAST_LOOP;
1945 		return (0);
1946 
1947 	default:
1948 		return (EOPNOTSUPP);
1949 	}
1950 }
1951 
1952 /*
1953  * Discard the IP multicast options.
1954  */
1955 void
1956 ip_freemoptions(struct ip_moptions *imo)
1957 {
1958 	int i;
1959 
1960 	if (imo != NULL) {
1961 		for (i = 0; i < imo->imo_num_memberships; ++i)
1962 			in_delmulti(imo->imo_membership[i]);
1963 		free(imo, M_IPMOPTS);
1964 	}
1965 }
1966 
1967 /*
1968  * Routine called from ip_output() to loop back a copy of an IP multicast
1969  * packet to the input queue of a specified interface.  Note that this
1970  * calls the output routine of the loopback "driver", but with an interface
1971  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
1972  */
1973 static void
1974 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1975 {
1976 	struct ip *ip;
1977 	struct mbuf *copym;
1978 
1979 	copym = m_copy(m, 0, M_COPYALL);
1980 	if (copym != NULL
1981 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1982 		copym = m_pullup(copym, sizeof(struct ip));
1983 	if (copym != NULL) {
1984 		/*
1985 		 * We don't bother to fragment if the IP length is greater
1986 		 * than the interface's MTU.  Can this possibly matter?
1987 		 */
1988 		ip = mtod(copym, struct ip *);
1989 
1990 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1991 			in_delayed_cksum(copym);
1992 			copym->m_pkthdr.csum_flags &=
1993 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1994 		}
1995 
1996 		ip->ip_sum = 0;
1997 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1998 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1999 	}
2000 }
2001