xref: /netbsd-src/sys/netinet/ip_output.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: ip_output.c,v 1.285 2017/11/17 07:37:12 ozaki-r 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  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1988, 1990, 1993
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.285 2017/11/17 07:37:12 ozaki-r Exp $");
95 
96 #ifdef _KERNEL_OPT
97 #include "opt_inet.h"
98 #include "opt_ipsec.h"
99 #include "opt_mrouting.h"
100 #include "opt_net_mpsafe.h"
101 #include "opt_mpls.h"
102 #endif
103 
104 #include "arp.h"
105 
106 #include <sys/param.h>
107 #include <sys/kmem.h>
108 #include <sys/mbuf.h>
109 #include <sys/socket.h>
110 #include <sys/socketvar.h>
111 #include <sys/kauth.h>
112 #include <sys/systm.h>
113 #include <sys/syslog.h>
114 
115 #include <net/if.h>
116 #include <net/if_types.h>
117 #include <net/route.h>
118 #include <net/pfil.h>
119 
120 #include <netinet/in.h>
121 #include <netinet/in_systm.h>
122 #include <netinet/ip.h>
123 #include <netinet/in_pcb.h>
124 #include <netinet/in_var.h>
125 #include <netinet/ip_var.h>
126 #include <netinet/ip_private.h>
127 #include <netinet/in_offload.h>
128 #include <netinet/portalgo.h>
129 #include <netinet/udp.h>
130 #include <netinet/udp_var.h>
131 
132 #ifdef INET6
133 #include <netinet6/ip6_var.h>
134 #endif
135 
136 #ifdef MROUTING
137 #include <netinet/ip_mroute.h>
138 #endif
139 
140 #ifdef IPSEC
141 #include <netipsec/ipsec.h>
142 #include <netipsec/key.h>
143 #endif
144 
145 #ifdef MPLS
146 #include <netmpls/mpls.h>
147 #include <netmpls/mpls_var.h>
148 #endif
149 
150 static int ip_pcbopts(struct inpcb *, const struct sockopt *);
151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
152 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
153 static void ip_mloopback(struct ifnet *, struct mbuf *,
154     const struct sockaddr_in *);
155 static int ip_ifaddrvalid(const struct in_ifaddr *);
156 
157 extern pfil_head_t *inet_pfil_hook;			/* XXX */
158 
159 int	ip_do_loopback_cksum = 0;
160 
161 static int
162 ip_mark_mpls(struct ifnet * const ifp, struct mbuf * const m,
163     const struct rtentry *rt)
164 {
165 	int error = 0;
166 #ifdef MPLS
167 	union mpls_shim msh;
168 
169 	if (rt == NULL || rt_gettag(rt) == NULL ||
170 	    rt_gettag(rt)->sa_family != AF_MPLS ||
171 	    (m->m_flags & (M_MCAST | M_BCAST)) != 0 ||
172 	    ifp->if_type != IFT_ETHER)
173 		return 0;
174 
175 	msh.s_addr = MPLS_GETSADDR(rt);
176 	if (msh.shim.label != MPLS_LABEL_IMPLNULL) {
177 		struct m_tag *mtag;
178 		/*
179 		 * XXX tentative solution to tell ether_output
180 		 * it's MPLS. Need some more efficient solution.
181 		 */
182 		mtag = m_tag_get(PACKET_TAG_MPLS,
183 		    sizeof(int) /* dummy */,
184 		    M_NOWAIT);
185 		if (mtag == NULL)
186 			return ENOMEM;
187 		m_tag_prepend(m, mtag);
188 	}
189 #endif
190 	return error;
191 }
192 
193 /*
194  * Send an IP packet to a host.
195  */
196 int
197 ip_if_output(struct ifnet * const ifp, struct mbuf * const m,
198     const struct sockaddr * const dst, const struct rtentry *rt)
199 {
200 	int error = 0;
201 
202 	if (rt != NULL) {
203 		error = rt_check_reject_route(rt, ifp);
204 		if (error != 0) {
205 			m_freem(m);
206 			return error;
207 		}
208 	}
209 
210 	error = ip_mark_mpls(ifp, m, rt);
211 	if (error != 0) {
212 		m_freem(m);
213 		return error;
214 	}
215 
216 	error = if_output_lock(ifp, ifp, m, dst, rt);
217 
218 	return error;
219 }
220 
221 /*
222  * IP output.  The packet in mbuf chain m contains a skeletal IP
223  * header (with len, off, ttl, proto, tos, src, dst).
224  * The mbuf chain containing the packet will be freed.
225  * The mbuf opt, if present, will not be freed.
226  */
227 int
228 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags,
229     struct ip_moptions *imo, struct inpcb *inp)
230 {
231 	struct rtentry *rt;
232 	struct ip *ip;
233 	struct ifnet *ifp, *mifp = NULL;
234 	struct mbuf *m = m0;
235 	int hlen = sizeof (struct ip);
236 	int len, error = 0;
237 	struct route iproute;
238 	const struct sockaddr_in *dst;
239 	struct in_ifaddr *ia = NULL;
240 	struct ifaddr *ifa;
241 	int isbroadcast;
242 	int sw_csum;
243 	u_long mtu;
244 	bool natt_frag = false;
245 	bool rtmtu_nolock;
246 	union {
247 		struct sockaddr		sa;
248 		struct sockaddr_in	sin;
249 	} udst, usrc;
250 	struct sockaddr *rdst = &udst.sa;	/* real IP destination, as
251 						 * opposed to the nexthop
252 						 */
253 	struct psref psref, psref_ia;
254 	int bound;
255 	bool bind_need_restore = false;
256 
257 	len = 0;
258 
259 	MCLAIM(m, &ip_tx_mowner);
260 
261 	KASSERT((m->m_flags & M_PKTHDR) != 0);
262 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0);
263 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) !=
264 	    (M_CSUM_TCPv4|M_CSUM_UDPv4));
265 
266 	if (opt) {
267 		m = ip_insertoptions(m, opt, &len);
268 		if (len >= sizeof(struct ip))
269 			hlen = len;
270 	}
271 	ip = mtod(m, struct ip *);
272 
273 	/*
274 	 * Fill in IP header.
275 	 */
276 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
277 		ip->ip_v = IPVERSION;
278 		ip->ip_off = htons(0);
279 		/* ip->ip_id filled in after we find out source ia */
280 		ip->ip_hl = hlen >> 2;
281 		IP_STATINC(IP_STAT_LOCALOUT);
282 	} else {
283 		hlen = ip->ip_hl << 2;
284 	}
285 
286 	/*
287 	 * Route packet.
288 	 */
289 	if (ro == NULL) {
290 		memset(&iproute, 0, sizeof(iproute));
291 		ro = &iproute;
292 	}
293 	sockaddr_in_init(&udst.sin, &ip->ip_dst, 0);
294 	dst = satocsin(rtcache_getdst(ro));
295 
296 	/*
297 	 * If there is a cached route, check that it is to the same
298 	 * destination and is still up.  If not, free it and try again.
299 	 * The address family should also be checked in case of sharing
300 	 * the cache with IPv6.
301 	 */
302 	if (dst && (dst->sin_family != AF_INET ||
303 	    !in_hosteq(dst->sin_addr, ip->ip_dst)))
304 		rtcache_free(ro);
305 
306 	if ((rt = rtcache_validate(ro)) == NULL &&
307 	    (rt = rtcache_update(ro, 1)) == NULL) {
308 		dst = &udst.sin;
309 		error = rtcache_setdst(ro, &udst.sa);
310 		if (error != 0)
311 			goto bad;
312 	}
313 
314 	bound = curlwp_bind();
315 	bind_need_restore = true;
316 	/*
317 	 * If routing to interface only, short circuit routing lookup.
318 	 */
319 	if (flags & IP_ROUTETOIF) {
320 		ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia);
321 		if (ifa == NULL) {
322 			IP_STATINC(IP_STAT_NOROUTE);
323 			error = ENETUNREACH;
324 			goto bad;
325 		}
326 		/* ia is already referenced by psref_ia */
327 		ia = ifatoia(ifa);
328 
329 		ifp = ia->ia_ifp;
330 		mtu = ifp->if_mtu;
331 		ip->ip_ttl = 1;
332 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
333 	} else if (((IN_MULTICAST(ip->ip_dst.s_addr) ||
334 	    ip->ip_dst.s_addr == INADDR_BROADCAST) ||
335 	    (flags & IP_ROUTETOIFINDEX)) &&
336 	    imo != NULL && imo->imo_multicast_if_index != 0) {
337 		ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref);
338 		if (ifp == NULL) {
339 			IP_STATINC(IP_STAT_NOROUTE);
340 			error = ENETUNREACH;
341 			goto bad;
342 		}
343 		mtu = ifp->if_mtu;
344 		ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
345 		if (ia == NULL) {
346 			error = EADDRNOTAVAIL;
347 			goto bad;
348 		}
349 		if (IN_MULTICAST(ip->ip_dst.s_addr) ||
350 		    ip->ip_dst.s_addr == INADDR_BROADCAST) {
351 			isbroadcast = 0;
352 		} else {
353 			/* IP_ROUTETOIFINDEX */
354 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
355 			if ((isbroadcast == 0) && ((ifp->if_flags &
356 			    (IFF_LOOPBACK | IFF_POINTOPOINT)) == 0) &&
357 			    (in_direct(dst->sin_addr, ifp) == 0)) {
358 				/* gateway address required */
359 				if (rt == NULL)
360 					rt = rtcache_init(ro);
361 				if (rt == NULL || rt->rt_ifp != ifp) {
362 					IP_STATINC(IP_STAT_NOROUTE);
363 					error = EHOSTUNREACH;
364 					goto bad;
365 				}
366 				rt->rt_use++;
367 				if (rt->rt_flags & RTF_GATEWAY)
368 					dst = satosin(rt->rt_gateway);
369 				if (rt->rt_flags & RTF_HOST)
370 					isbroadcast =
371 					    rt->rt_flags & RTF_BROADCAST;
372 			}
373 		}
374 	} else {
375 		if (rt == NULL)
376 			rt = rtcache_init(ro);
377 		if (rt == NULL) {
378 			IP_STATINC(IP_STAT_NOROUTE);
379 			error = EHOSTUNREACH;
380 			goto bad;
381 		}
382 		if (ifa_is_destroying(rt->rt_ifa)) {
383 			rtcache_unref(rt, ro);
384 			rt = NULL;
385 			IP_STATINC(IP_STAT_NOROUTE);
386 			error = EHOSTUNREACH;
387 			goto bad;
388 		}
389 		ifa_acquire(rt->rt_ifa, &psref_ia);
390 		ia = ifatoia(rt->rt_ifa);
391 		ifp = rt->rt_ifp;
392 		if ((mtu = rt->rt_rmx.rmx_mtu) == 0)
393 			mtu = ifp->if_mtu;
394 		rt->rt_use++;
395 		if (rt->rt_flags & RTF_GATEWAY)
396 			dst = satosin(rt->rt_gateway);
397 		if (rt->rt_flags & RTF_HOST)
398 			isbroadcast = rt->rt_flags & RTF_BROADCAST;
399 		else
400 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
401 	}
402 	rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0;
403 
404 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
405 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
406 		bool inmgroup;
407 
408 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
409 		    M_BCAST : M_MCAST;
410 		/*
411 		 * See if the caller provided any multicast options
412 		 */
413 		if (imo != NULL)
414 			ip->ip_ttl = imo->imo_multicast_ttl;
415 		else
416 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
417 
418 		/*
419 		 * if we don't know the outgoing ifp yet, we can't generate
420 		 * output
421 		 */
422 		if (!ifp) {
423 			IP_STATINC(IP_STAT_NOROUTE);
424 			error = ENETUNREACH;
425 			goto bad;
426 		}
427 
428 		/*
429 		 * If the packet is multicast or broadcast, confirm that
430 		 * the outgoing interface can transmit it.
431 		 */
432 		if (((m->m_flags & M_MCAST) &&
433 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
434 		    ((m->m_flags & M_BCAST) &&
435 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
436 			IP_STATINC(IP_STAT_NOROUTE);
437 			error = ENETUNREACH;
438 			goto bad;
439 		}
440 		/*
441 		 * If source address not specified yet, use an address
442 		 * of outgoing interface.
443 		 */
444 		if (in_nullhost(ip->ip_src)) {
445 			struct in_ifaddr *xia;
446 			struct ifaddr *xifa;
447 			struct psref _psref;
448 
449 			xia = in_get_ia_from_ifp_psref(ifp, &_psref);
450 			if (!xia) {
451 				error = EADDRNOTAVAIL;
452 				goto bad;
453 			}
454 			xifa = &xia->ia_ifa;
455 			if (xifa->ifa_getifa != NULL) {
456 				ia4_release(xia, &_psref);
457 				/* FIXME ifa_getifa is NOMPSAFE */
458 				xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
459 				if (xia == NULL) {
460 					error = EADDRNOTAVAIL;
461 					goto bad;
462 				}
463 				ia4_acquire(xia, &_psref);
464 			}
465 			ip->ip_src = xia->ia_addr.sin_addr;
466 			ia4_release(xia, &_psref);
467 		}
468 
469 		inmgroup = in_multi_group(ip->ip_dst, ifp, flags);
470 		if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) {
471 			/*
472 			 * If we belong to the destination multicast group
473 			 * on the outgoing interface, and the caller did not
474 			 * forbid loopback, loop back a copy.
475 			 */
476 			ip_mloopback(ifp, m, &udst.sin);
477 		}
478 #ifdef MROUTING
479 		else {
480 			/*
481 			 * If we are acting as a multicast router, perform
482 			 * multicast forwarding as if the packet had just
483 			 * arrived on the interface to which we are about
484 			 * to send.  The multicast forwarding function
485 			 * recursively calls this function, using the
486 			 * IP_FORWARDING flag to prevent infinite recursion.
487 			 *
488 			 * Multicasts that are looped back by ip_mloopback(),
489 			 * above, will be forwarded by the ip_input() routine,
490 			 * if necessary.
491 			 */
492 			extern struct socket *ip_mrouter;
493 
494 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
495 				if (ip_mforward(m, ifp) != 0) {
496 					m_freem(m);
497 					goto done;
498 				}
499 			}
500 		}
501 #endif
502 		/*
503 		 * Multicasts with a time-to-live of zero may be looped-
504 		 * back, above, but must not be transmitted on a network.
505 		 * Also, multicasts addressed to the loopback interface
506 		 * are not sent -- the above call to ip_mloopback() will
507 		 * loop back a copy if this host actually belongs to the
508 		 * destination group on the loopback interface.
509 		 */
510 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
511 			m_freem(m);
512 			goto done;
513 		}
514 		goto sendit;
515 	}
516 
517 	/*
518 	 * If source address not specified yet, use address
519 	 * of outgoing interface.
520 	 */
521 	if (in_nullhost(ip->ip_src)) {
522 		struct ifaddr *xifa;
523 
524 		xifa = &ia->ia_ifa;
525 		if (xifa->ifa_getifa != NULL) {
526 			ia4_release(ia, &psref_ia);
527 			/* FIXME ifa_getifa is NOMPSAFE */
528 			ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
529 			if (ia == NULL) {
530 				error = EADDRNOTAVAIL;
531 				goto bad;
532 			}
533 			ia4_acquire(ia, &psref_ia);
534 		}
535 		ip->ip_src = ia->ia_addr.sin_addr;
536 	}
537 
538 	/*
539 	 * packets with Class-D address as source are not valid per
540 	 * RFC 1112
541 	 */
542 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
543 		IP_STATINC(IP_STAT_ODROPPED);
544 		error = EADDRNOTAVAIL;
545 		goto bad;
546 	}
547 
548 	/*
549 	 * Look for broadcast address and and verify user is allowed to
550 	 * send such a packet.
551 	 */
552 	if (isbroadcast) {
553 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
554 			error = EADDRNOTAVAIL;
555 			goto bad;
556 		}
557 		if ((flags & IP_ALLOWBROADCAST) == 0) {
558 			error = EACCES;
559 			goto bad;
560 		}
561 		/* don't allow broadcast messages to be fragmented */
562 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
563 			error = EMSGSIZE;
564 			goto bad;
565 		}
566 		m->m_flags |= M_BCAST;
567 	} else
568 		m->m_flags &= ~M_BCAST;
569 
570 sendit:
571 	if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) {
572 		if (m->m_pkthdr.len < IP_MINFRAGSIZE) {
573 			ip->ip_id = 0;
574 		} else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
575 			ip->ip_id = ip_newid(ia);
576 		} else {
577 
578 			/*
579 			 * TSO capable interfaces (typically?) increment
580 			 * ip_id for each segment.
581 			 * "allocate" enough ids here to increase the chance
582 			 * for them to be unique.
583 			 *
584 			 * note that the following calculation is not
585 			 * needed to be precise.  wasting some ip_id is fine.
586 			 */
587 
588 			unsigned int segsz = m->m_pkthdr.segsz;
589 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
590 			unsigned int num = howmany(datasz, segsz);
591 
592 			ip->ip_id = ip_newid_range(ia, num);
593 		}
594 	}
595 	if (ia != NULL) {
596 		ia4_release(ia, &psref_ia);
597 		ia = NULL;
598 	}
599 
600 	/*
601 	 * If we're doing Path MTU Discovery, we need to set DF unless
602 	 * the route's MTU is locked.
603 	 */
604 	if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) {
605 		ip->ip_off |= htons(IP_DF);
606 	}
607 
608 #ifdef IPSEC
609 	if (ipsec_used) {
610 		bool ipsec_done = false;
611 
612 		/* Perform IPsec processing, if any. */
613 		error = ipsec4_output(m, inp, flags, &mtu, &natt_frag,
614 		    &ipsec_done);
615 		if (error || ipsec_done)
616 			goto done;
617 	}
618 #endif
619 
620 	/*
621 	 * Run through list of hooks for output packets.
622 	 */
623 	error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT);
624 	if (error)
625 		goto done;
626 	if (m == NULL)
627 		goto done;
628 
629 	ip = mtod(m, struct ip *);
630 	hlen = ip->ip_hl << 2;
631 
632 	m->m_pkthdr.csum_data |= hlen << 16;
633 
634 	/*
635 	 * search for the source address structure to
636 	 * maintain output statistics, and verify address
637 	 * validity
638 	 */
639 	KASSERT(ia == NULL);
640 	sockaddr_in_init(&usrc.sin, &ip->ip_src, 0);
641 	ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia);
642 	if (ifa != NULL)
643 		ia = ifatoia(ifa);
644 
645 	/*
646 	 * Ensure we only send from a valid address.
647 	 * A NULL address is valid because the packet could be
648 	 * generated from a packet filter.
649 	 */
650 	if (ia != NULL && (flags & IP_FORWARDING) == 0 &&
651 	    (error = ip_ifaddrvalid(ia)) != 0)
652 	{
653 		ARPLOG(LOG_ERR,
654 		    "refusing to send from invalid address %s (pid %d)\n",
655 		    ARPLOGADDR(&ip->ip_src), curproc->p_pid);
656 		IP_STATINC(IP_STAT_ODROPPED);
657 		if (error == 1)
658 			/*
659 			 * Address exists, but is tentative or detached.
660 			 * We can't send from it because it's invalid,
661 			 * so we drop the packet.
662 			 */
663 			error = 0;
664 		else
665 			error = EADDRNOTAVAIL;
666 		goto bad;
667 	}
668 
669 	/* Maybe skip checksums on loopback interfaces. */
670 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
671 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
672 	}
673 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
674 	/*
675 	 * If small enough for mtu of path, or if using TCP segmentation
676 	 * offload, can just send directly.
677 	 */
678 	if (ntohs(ip->ip_len) <= mtu ||
679 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
680 		const struct sockaddr *sa;
681 
682 #if IFA_STATS
683 		if (ia)
684 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
685 #endif
686 		/*
687 		 * Always initialize the sum to 0!  Some HW assisted
688 		 * checksumming requires this.
689 		 */
690 		ip->ip_sum = 0;
691 
692 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
693 			/*
694 			 * Perform any checksums that the hardware can't do
695 			 * for us.
696 			 *
697 			 * XXX Does any hardware require the {th,uh}_sum
698 			 * XXX fields to be 0?
699 			 */
700 			if (sw_csum & M_CSUM_IPv4) {
701 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
702 				ip->ip_sum = in_cksum(m, hlen);
703 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
704 			}
705 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
706 				if (IN_NEED_CHECKSUM(ifp,
707 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
708 					in_delayed_cksum(m);
709 				}
710 				m->m_pkthdr.csum_flags &=
711 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
712 			}
713 		}
714 
715 		sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst);
716 		if (__predict_true(
717 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
718 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
719 			error = ip_if_output(ifp, m, sa, rt);
720 		} else {
721 			error = ip_tso_output(ifp, m, sa, rt);
722 		}
723 		goto done;
724 	}
725 
726 	/*
727 	 * We can't use HW checksumming if we're about to
728 	 * fragment the packet.
729 	 *
730 	 * XXX Some hardware can do this.
731 	 */
732 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
733 		if (IN_NEED_CHECKSUM(ifp,
734 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
735 			in_delayed_cksum(m);
736 		}
737 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
738 	}
739 
740 	/*
741 	 * Too large for interface; fragment if possible.
742 	 * Must be able to put at least 8 bytes per fragment.
743 	 */
744 	if (ntohs(ip->ip_off) & IP_DF) {
745 		if (flags & IP_RETURNMTU) {
746 			KASSERT(inp != NULL);
747 			inp->inp_errormtu = mtu;
748 		}
749 		error = EMSGSIZE;
750 		IP_STATINC(IP_STAT_CANTFRAG);
751 		goto bad;
752 	}
753 
754 	error = ip_fragment(m, ifp, mtu);
755 	if (error) {
756 		m = NULL;
757 		goto bad;
758 	}
759 
760 	for (; m; m = m0) {
761 		m0 = m->m_nextpkt;
762 		m->m_nextpkt = 0;
763 		if (error) {
764 			m_freem(m);
765 			continue;
766 		}
767 #if IFA_STATS
768 		if (ia)
769 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
770 #endif
771 		/*
772 		 * If we get there, the packet has not been handled by
773 		 * IPsec whereas it should have. Now that it has been
774 		 * fragmented, re-inject it in ip_output so that IPsec
775 		 * processing can occur.
776 		 */
777 		if (natt_frag) {
778 			error = ip_output(m, opt, ro,
779 			    flags | IP_RAWOUTPUT | IP_NOIPNEWID,
780 			    imo, inp);
781 		} else {
782 			KASSERT((m->m_pkthdr.csum_flags &
783 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
784 			error = ip_if_output(ifp, m,
785 			    (m->m_flags & M_MCAST) ?
786 			    sintocsa(rdst) : sintocsa(dst), rt);
787 		}
788 	}
789 	if (error == 0) {
790 		IP_STATINC(IP_STAT_FRAGMENTED);
791 	}
792 done:
793 	ia4_release(ia, &psref_ia);
794 	rtcache_unref(rt, ro);
795 	if (ro == &iproute) {
796 		rtcache_free(&iproute);
797 	}
798 	if (mifp != NULL) {
799 		if_put(mifp, &psref);
800 	}
801 	if (bind_need_restore)
802 		curlwp_bindx(bound);
803 	return error;
804 bad:
805 	m_freem(m);
806 	goto done;
807 }
808 
809 int
810 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
811 {
812 	struct ip *ip, *mhip;
813 	struct mbuf *m0;
814 	int len, hlen, off;
815 	int mhlen, firstlen;
816 	struct mbuf **mnext;
817 	int sw_csum = m->m_pkthdr.csum_flags;
818 	int fragments = 0;
819 	int error = 0;
820 
821 	ip = mtod(m, struct ip *);
822 	hlen = ip->ip_hl << 2;
823 	if (ifp != NULL)
824 		sw_csum &= ~ifp->if_csum_flags_tx;
825 
826 	len = (mtu - hlen) &~ 7;
827 	if (len < 8) {
828 		m_freem(m);
829 		return (EMSGSIZE);
830 	}
831 
832 	firstlen = len;
833 	mnext = &m->m_nextpkt;
834 
835 	/*
836 	 * Loop through length of segment after first fragment,
837 	 * make new header and copy data of each part and link onto chain.
838 	 */
839 	m0 = m;
840 	mhlen = sizeof (struct ip);
841 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
842 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
843 		if (m == 0) {
844 			error = ENOBUFS;
845 			IP_STATINC(IP_STAT_ODROPPED);
846 			goto sendorfree;
847 		}
848 		MCLAIM(m, m0->m_owner);
849 		*mnext = m;
850 		mnext = &m->m_nextpkt;
851 		m->m_data += max_linkhdr;
852 		mhip = mtod(m, struct ip *);
853 		*mhip = *ip;
854 		/* we must inherit MCAST and BCAST flags */
855 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
856 		if (hlen > sizeof (struct ip)) {
857 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
858 			mhip->ip_hl = mhlen >> 2;
859 		}
860 		m->m_len = mhlen;
861 		mhip->ip_off = ((off - hlen) >> 3) +
862 		    (ntohs(ip->ip_off) & ~IP_MF);
863 		if (ip->ip_off & htons(IP_MF))
864 			mhip->ip_off |= IP_MF;
865 		if (off + len >= ntohs(ip->ip_len))
866 			len = ntohs(ip->ip_len) - off;
867 		else
868 			mhip->ip_off |= IP_MF;
869 		HTONS(mhip->ip_off);
870 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
871 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
872 		if (m->m_next == 0) {
873 			error = ENOBUFS;	/* ??? */
874 			IP_STATINC(IP_STAT_ODROPPED);
875 			goto sendorfree;
876 		}
877 		m->m_pkthdr.len = mhlen + len;
878 		m_reset_rcvif(m);
879 		mhip->ip_sum = 0;
880 		KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
881 		if (sw_csum & M_CSUM_IPv4) {
882 			mhip->ip_sum = in_cksum(m, mhlen);
883 		} else {
884 			/*
885 			 * checksum is hw-offloaded or not necessary.
886 			 */
887 			m->m_pkthdr.csum_flags |=
888 			    m0->m_pkthdr.csum_flags & M_CSUM_IPv4;
889 			m->m_pkthdr.csum_data |= mhlen << 16;
890 			KASSERT(!(ifp != NULL &&
891 			    IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
892 			    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
893 		}
894 		IP_STATINC(IP_STAT_OFRAGMENTS);
895 		fragments++;
896 	}
897 	/*
898 	 * Update first fragment by trimming what's been copied out
899 	 * and updating header, then send each fragment (in order).
900 	 */
901 	m = m0;
902 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
903 	m->m_pkthdr.len = hlen + firstlen;
904 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
905 	ip->ip_off |= htons(IP_MF);
906 	ip->ip_sum = 0;
907 	if (sw_csum & M_CSUM_IPv4) {
908 		ip->ip_sum = in_cksum(m, hlen);
909 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
910 	} else {
911 		/*
912 		 * checksum is hw-offloaded or not necessary.
913 		 */
914 		KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
915 		    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
916 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
917 		    sizeof(struct ip));
918 	}
919 sendorfree:
920 	/*
921 	 * If there is no room for all the fragments, don't queue
922 	 * any of them.
923 	 */
924 	if (ifp != NULL) {
925 		IFQ_LOCK(&ifp->if_snd);
926 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
927 		    error == 0) {
928 			error = ENOBUFS;
929 			IP_STATINC(IP_STAT_ODROPPED);
930 			IFQ_INC_DROPS(&ifp->if_snd);
931 		}
932 		IFQ_UNLOCK(&ifp->if_snd);
933 	}
934 	if (error) {
935 		for (m = m0; m; m = m0) {
936 			m0 = m->m_nextpkt;
937 			m->m_nextpkt = NULL;
938 			m_freem(m);
939 		}
940 	}
941 	return (error);
942 }
943 
944 /*
945  * Process a delayed payload checksum calculation.
946  */
947 void
948 in_delayed_cksum(struct mbuf *m)
949 {
950 	struct ip *ip;
951 	u_int16_t csum, offset;
952 
953 	ip = mtod(m, struct ip *);
954 	offset = ip->ip_hl << 2;
955 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
956 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
957 		csum = 0xffff;
958 
959 	offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
960 
961 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
962 		/* This happen when ip options were inserted
963 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
964 		    m->m_len, offset, ip->ip_p);
965 		 */
966 		m_copyback(m, offset, sizeof(csum), (void *) &csum);
967 	} else
968 		*(u_int16_t *)(mtod(m, char *) + offset) = csum;
969 }
970 
971 /*
972  * Determine the maximum length of the options to be inserted;
973  * we would far rather allocate too much space rather than too little.
974  */
975 
976 u_int
977 ip_optlen(struct inpcb *inp)
978 {
979 	struct mbuf *m = inp->inp_options;
980 
981 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) {
982 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
983 	}
984 	return 0;
985 }
986 
987 /*
988  * Insert IP options into preformed packet.
989  * Adjust IP destination as required for IP source routing,
990  * as indicated by a non-zero in_addr at the start of the options.
991  */
992 static struct mbuf *
993 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
994 {
995 	struct ipoption *p = mtod(opt, struct ipoption *);
996 	struct mbuf *n;
997 	struct ip *ip = mtod(m, struct ip *);
998 	unsigned optlen;
999 
1000 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1001 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1002 		return (m);		/* XXX should fail */
1003 	if (!in_nullhost(p->ipopt_dst))
1004 		ip->ip_dst = p->ipopt_dst;
1005 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1006 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1007 		if (n == 0)
1008 			return (m);
1009 		MCLAIM(n, m->m_owner);
1010 		M_MOVE_PKTHDR(n, m);
1011 		m->m_len -= sizeof(struct ip);
1012 		m->m_data += sizeof(struct ip);
1013 		n->m_next = m;
1014 		m = n;
1015 		m->m_len = optlen + sizeof(struct ip);
1016 		m->m_data += max_linkhdr;
1017 		bcopy((void *)ip, mtod(m, void *), sizeof(struct ip));
1018 	} else {
1019 		m->m_data -= optlen;
1020 		m->m_len += optlen;
1021 		memmove(mtod(m, void *), ip, sizeof(struct ip));
1022 	}
1023 	m->m_pkthdr.len += optlen;
1024 	ip = mtod(m, struct ip *);
1025 	bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen);
1026 	*phlen = sizeof(struct ip) + optlen;
1027 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1028 	return (m);
1029 }
1030 
1031 /*
1032  * Copy options from ip to jp,
1033  * omitting those not copied during fragmentation.
1034  */
1035 int
1036 ip_optcopy(struct ip *ip, struct ip *jp)
1037 {
1038 	u_char *cp, *dp;
1039 	int opt, optlen, cnt;
1040 
1041 	cp = (u_char *)(ip + 1);
1042 	dp = (u_char *)(jp + 1);
1043 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1044 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1045 		opt = cp[0];
1046 		if (opt == IPOPT_EOL)
1047 			break;
1048 		if (opt == IPOPT_NOP) {
1049 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1050 			*dp++ = IPOPT_NOP;
1051 			optlen = 1;
1052 			continue;
1053 		}
1054 
1055 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp));
1056 		optlen = cp[IPOPT_OLEN];
1057 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt);
1058 
1059 		/* Invalid lengths should have been caught by ip_dooptions. */
1060 		if (optlen > cnt)
1061 			optlen = cnt;
1062 		if (IPOPT_COPIED(opt)) {
1063 			bcopy((void *)cp, (void *)dp, (unsigned)optlen);
1064 			dp += optlen;
1065 		}
1066 	}
1067 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1068 		*dp++ = IPOPT_EOL;
1069 	return (optlen);
1070 }
1071 
1072 /*
1073  * IP socket option processing.
1074  */
1075 int
1076 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1077 {
1078 	struct inpcb *inp = sotoinpcb(so);
1079 	struct ip *ip = &inp->inp_ip;
1080 	int inpflags = inp->inp_flags;
1081 	int optval = 0, error = 0;
1082 
1083 	KASSERT(solocked(so));
1084 
1085 	if (sopt->sopt_level != IPPROTO_IP) {
1086 		if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER)
1087 			return 0;
1088 		return ENOPROTOOPT;
1089 	}
1090 
1091 	switch (op) {
1092 	case PRCO_SETOPT:
1093 		switch (sopt->sopt_name) {
1094 		case IP_OPTIONS:
1095 #ifdef notyet
1096 		case IP_RETOPTS:
1097 #endif
1098 			error = ip_pcbopts(inp, sopt);
1099 			break;
1100 
1101 		case IP_TOS:
1102 		case IP_TTL:
1103 		case IP_MINTTL:
1104 		case IP_PKTINFO:
1105 		case IP_RECVOPTS:
1106 		case IP_RECVRETOPTS:
1107 		case IP_RECVDSTADDR:
1108 		case IP_RECVIF:
1109 		case IP_RECVPKTINFO:
1110 		case IP_RECVTTL:
1111 			error = sockopt_getint(sopt, &optval);
1112 			if (error)
1113 				break;
1114 
1115 			switch (sopt->sopt_name) {
1116 			case IP_TOS:
1117 				ip->ip_tos = optval;
1118 				break;
1119 
1120 			case IP_TTL:
1121 				ip->ip_ttl = optval;
1122 				break;
1123 
1124 			case IP_MINTTL:
1125 				if (optval > 0 && optval <= MAXTTL)
1126 					inp->inp_ip_minttl = optval;
1127 				else
1128 					error = EINVAL;
1129 				break;
1130 #define	OPTSET(bit) \
1131 	if (optval) \
1132 		inpflags |= bit; \
1133 	else \
1134 		inpflags &= ~bit;
1135 
1136 			case IP_PKTINFO:
1137 				OPTSET(INP_PKTINFO);
1138 				break;
1139 
1140 			case IP_RECVOPTS:
1141 				OPTSET(INP_RECVOPTS);
1142 				break;
1143 
1144 			case IP_RECVPKTINFO:
1145 				OPTSET(INP_RECVPKTINFO);
1146 				break;
1147 
1148 			case IP_RECVRETOPTS:
1149 				OPTSET(INP_RECVRETOPTS);
1150 				break;
1151 
1152 			case IP_RECVDSTADDR:
1153 				OPTSET(INP_RECVDSTADDR);
1154 				break;
1155 
1156 			case IP_RECVIF:
1157 				OPTSET(INP_RECVIF);
1158 				break;
1159 
1160 			case IP_RECVTTL:
1161 				OPTSET(INP_RECVTTL);
1162 				break;
1163 			}
1164 		break;
1165 #undef OPTSET
1166 
1167 		case IP_MULTICAST_IF:
1168 		case IP_MULTICAST_TTL:
1169 		case IP_MULTICAST_LOOP:
1170 		case IP_ADD_MEMBERSHIP:
1171 		case IP_DROP_MEMBERSHIP:
1172 			error = ip_setmoptions(&inp->inp_moptions, sopt);
1173 			break;
1174 
1175 		case IP_PORTRANGE:
1176 			error = sockopt_getint(sopt, &optval);
1177 			if (error)
1178 				break;
1179 
1180 			switch (optval) {
1181 			case IP_PORTRANGE_DEFAULT:
1182 			case IP_PORTRANGE_HIGH:
1183 				inpflags &= ~(INP_LOWPORT);
1184 				break;
1185 
1186 			case IP_PORTRANGE_LOW:
1187 				inpflags |= INP_LOWPORT;
1188 				break;
1189 
1190 			default:
1191 				error = EINVAL;
1192 				break;
1193 			}
1194 			break;
1195 
1196 		case IP_PORTALGO:
1197 			error = sockopt_getint(sopt, &optval);
1198 			if (error)
1199 				break;
1200 
1201 			error = portalgo_algo_index_select(
1202 			    (struct inpcb_hdr *)inp, optval);
1203 			break;
1204 
1205 #if defined(IPSEC)
1206 		case IP_IPSEC_POLICY:
1207 			if (ipsec_enabled) {
1208 				error = ipsec4_set_policy(inp, sopt->sopt_name,
1209 				    sopt->sopt_data, sopt->sopt_size,
1210 				    curlwp->l_cred);
1211 				break;
1212 			}
1213 			/*FALLTHROUGH*/
1214 #endif /* IPSEC */
1215 
1216 		default:
1217 			error = ENOPROTOOPT;
1218 			break;
1219 		}
1220 		break;
1221 
1222 	case PRCO_GETOPT:
1223 		switch (sopt->sopt_name) {
1224 		case IP_OPTIONS:
1225 		case IP_RETOPTS: {
1226 			struct mbuf *mopts = inp->inp_options;
1227 
1228 			if (mopts) {
1229 				struct mbuf *m;
1230 
1231 				m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT);
1232 				if (m == NULL) {
1233 					error = ENOBUFS;
1234 					break;
1235 				}
1236 				error = sockopt_setmbuf(sopt, m);
1237 			}
1238 			break;
1239 		}
1240 		case IP_PKTINFO:
1241 		case IP_TOS:
1242 		case IP_TTL:
1243 		case IP_MINTTL:
1244 		case IP_RECVOPTS:
1245 		case IP_RECVRETOPTS:
1246 		case IP_RECVDSTADDR:
1247 		case IP_RECVIF:
1248 		case IP_RECVPKTINFO:
1249 		case IP_RECVTTL:
1250 		case IP_ERRORMTU:
1251 			switch (sopt->sopt_name) {
1252 			case IP_TOS:
1253 				optval = ip->ip_tos;
1254 				break;
1255 
1256 			case IP_TTL:
1257 				optval = ip->ip_ttl;
1258 				break;
1259 
1260 			case IP_MINTTL:
1261 				optval = inp->inp_ip_minttl;
1262 				break;
1263 
1264 			case IP_ERRORMTU:
1265 				optval = inp->inp_errormtu;
1266 				break;
1267 
1268 #define	OPTBIT(bit)	(inpflags & bit ? 1 : 0)
1269 
1270 			case IP_PKTINFO:
1271 				optval = OPTBIT(INP_PKTINFO);
1272 				break;
1273 
1274 			case IP_RECVOPTS:
1275 				optval = OPTBIT(INP_RECVOPTS);
1276 				break;
1277 
1278 			case IP_RECVPKTINFO:
1279 				optval = OPTBIT(INP_RECVPKTINFO);
1280 				break;
1281 
1282 			case IP_RECVRETOPTS:
1283 				optval = OPTBIT(INP_RECVRETOPTS);
1284 				break;
1285 
1286 			case IP_RECVDSTADDR:
1287 				optval = OPTBIT(INP_RECVDSTADDR);
1288 				break;
1289 
1290 			case IP_RECVIF:
1291 				optval = OPTBIT(INP_RECVIF);
1292 				break;
1293 
1294 			case IP_RECVTTL:
1295 				optval = OPTBIT(INP_RECVTTL);
1296 				break;
1297 			}
1298 			error = sockopt_setint(sopt, optval);
1299 			break;
1300 
1301 #if 0	/* defined(IPSEC) */
1302 		case IP_IPSEC_POLICY:
1303 		{
1304 			struct mbuf *m = NULL;
1305 
1306 			/* XXX this will return EINVAL as sopt is empty */
1307 			error = ipsec4_get_policy(inp, sopt->sopt_data,
1308 			    sopt->sopt_size, &m);
1309 			if (error == 0)
1310 				error = sockopt_setmbuf(sopt, m);
1311 			break;
1312 		}
1313 #endif /*IPSEC*/
1314 
1315 		case IP_MULTICAST_IF:
1316 		case IP_MULTICAST_TTL:
1317 		case IP_MULTICAST_LOOP:
1318 		case IP_ADD_MEMBERSHIP:
1319 		case IP_DROP_MEMBERSHIP:
1320 			error = ip_getmoptions(inp->inp_moptions, sopt);
1321 			break;
1322 
1323 		case IP_PORTRANGE:
1324 			if (inpflags & INP_LOWPORT)
1325 				optval = IP_PORTRANGE_LOW;
1326 			else
1327 				optval = IP_PORTRANGE_DEFAULT;
1328 			error = sockopt_setint(sopt, optval);
1329 			break;
1330 
1331 		case IP_PORTALGO:
1332 			optval = inp->inp_portalgo;
1333 			error = sockopt_setint(sopt, optval);
1334 			break;
1335 
1336 		default:
1337 			error = ENOPROTOOPT;
1338 			break;
1339 		}
1340 		break;
1341 	}
1342 
1343 	if (!error) {
1344 		inp->inp_flags = inpflags;
1345 	}
1346 	return error;
1347 }
1348 
1349 static int
1350 ip_pktinfo_prepare(const struct in_pktinfo *pktinfo, struct ip_pktopts *pktopts,
1351     int *flags, kauth_cred_t cred)
1352 {
1353 	struct ip_moptions *imo;
1354 	int error = 0;
1355 	bool addrset = false;
1356 
1357 	if (!in_nullhost(pktinfo->ipi_addr)) {
1358 		pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr;
1359 		/* EADDRNOTAVAIL? */
1360 		error = in_pcbbindableaddr(&pktopts->ippo_laddr, cred);
1361 		if (error != 0)
1362 			return error;
1363 		addrset = true;
1364 	}
1365 
1366 	if (pktinfo->ipi_ifindex != 0) {
1367 		if (!addrset) {
1368 			struct ifnet *ifp;
1369 			struct in_ifaddr *ia;
1370 			int s;
1371 
1372 			/* pick up primary address */
1373 			s = pserialize_read_enter();
1374 			ifp = if_byindex(pktinfo->ipi_ifindex);
1375 			if (ifp == NULL) {
1376 				pserialize_read_exit(s);
1377 				return EADDRNOTAVAIL;
1378 			}
1379 			ia = in_get_ia_from_ifp(ifp);
1380 			if (ia == NULL) {
1381 				pserialize_read_exit(s);
1382 				return EADDRNOTAVAIL;
1383 			}
1384 			pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr;
1385 			pserialize_read_exit(s);
1386 		}
1387 
1388 		/*
1389 		 * If specified ipi_ifindex,
1390 		 * use copied or locally initialized ip_moptions.
1391 		 * Original ip_moptions must not be modified.
1392 		 */
1393 		imo = &pktopts->ippo_imobuf;	/* local buf in pktopts */
1394 		if (pktopts->ippo_imo != NULL) {
1395 			memcpy(imo, pktopts->ippo_imo, sizeof(*imo));
1396 		} else {
1397 			memset(imo, 0, sizeof(*imo));
1398 			imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1399 			imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1400 		}
1401 		imo->imo_multicast_if_index = pktinfo->ipi_ifindex;
1402 		pktopts->ippo_imo = imo;
1403 		*flags |= IP_ROUTETOIFINDEX;
1404 	}
1405 	return error;
1406 }
1407 
1408 /*
1409  * Set up IP outgoing packet options. Even if control is NULL,
1410  * pktopts->ippo_laddr and pktopts->ippo_imo are set and used.
1411  */
1412 int
1413 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags,
1414     struct inpcb *inp, kauth_cred_t cred, int uproto)
1415 {
1416 	struct inpcb *xinp;
1417 	struct cmsghdr *cm;
1418 	struct in_pktinfo *pktinfo;
1419 	int error;
1420 
1421 	pktopts->ippo_imo = inp->inp_moptions;
1422 	sockaddr_in_init(&pktopts->ippo_laddr, &inp->inp_laddr, 0);
1423 
1424 	if (control == NULL)
1425 		return 0;
1426 
1427 	/*
1428 	 * XXX: Currently, we assume all the optional information is
1429 	 * stored in a single mbuf.
1430 	 */
1431 	if (control->m_next)
1432 		return EINVAL;
1433 
1434 	for (; control->m_len > 0;
1435 	    control->m_data += CMSG_ALIGN(cm->cmsg_len),
1436 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1437 		cm = mtod(control, struct cmsghdr *);
1438 		if ((control->m_len < sizeof(*cm)) ||
1439 		    (cm->cmsg_len == 0) ||
1440 		    (cm->cmsg_len > control->m_len)) {
1441 			return EINVAL;
1442 		}
1443 		if (cm->cmsg_level != IPPROTO_IP)
1444 			continue;
1445 
1446 		switch (cm->cmsg_type) {
1447 		case IP_PKTINFO:
1448 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
1449 				return EINVAL;
1450 
1451 			pktinfo = (struct in_pktinfo *)CMSG_DATA(cm);
1452 			error = ip_pktinfo_prepare(pktinfo, pktopts, flags,
1453 			    cred);
1454 			if (error != 0)
1455 				return error;
1456 
1457 			if ((uproto == IPPROTO_UDP) &&
1458 			    !in_nullhost(pktopts->ippo_laddr.sin_addr)) {
1459 				/* Checking laddr:port already in use? */
1460 				xinp = in_pcblookup_bind(&udbtable,
1461 				    pktopts->ippo_laddr.sin_addr,
1462 				    inp->inp_lport);
1463 				if ((xinp != NULL) && (xinp != inp))
1464 					return EADDRINUSE;
1465 			}
1466 			break;
1467 		default:
1468 			return ENOPROTOOPT;
1469 		}
1470 	}
1471 	return 0;
1472 }
1473 
1474 /*
1475  * Set up IP options in pcb for insertion in output packets.
1476  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1477  * with destination address if source routed.
1478  */
1479 static int
1480 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt)
1481 {
1482 	struct mbuf *m;
1483 	const u_char *cp;
1484 	u_char *dp;
1485 	int cnt;
1486 
1487 	KASSERT(inp_locked(inp));
1488 
1489 	/* Turn off any old options. */
1490 	if (inp->inp_options) {
1491 		m_free(inp->inp_options);
1492 	}
1493 	inp->inp_options = NULL;
1494 	if ((cnt = sopt->sopt_size) == 0) {
1495 		/* Only turning off any previous options. */
1496 		return 0;
1497 	}
1498 	cp = sopt->sopt_data;
1499 
1500 #ifndef	__vax__
1501 	if (cnt % sizeof(int32_t))
1502 		return (EINVAL);
1503 #endif
1504 
1505 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1506 	if (m == NULL)
1507 		return (ENOBUFS);
1508 
1509 	dp = mtod(m, u_char *);
1510 	memset(dp, 0, sizeof(struct in_addr));
1511 	dp += sizeof(struct in_addr);
1512 	m->m_len = sizeof(struct in_addr);
1513 
1514 	/*
1515 	 * IP option list according to RFC791. Each option is of the form
1516 	 *
1517 	 *	[optval] [olen] [(olen - 2) data bytes]
1518 	 *
1519 	 * We validate the list and copy options to an mbuf for prepending
1520 	 * to data packets. The IP first-hop destination address will be
1521 	 * stored before actual options and is zero if unset.
1522 	 */
1523 	while (cnt > 0) {
1524 		uint8_t optval, olen, offset;
1525 
1526 		optval = cp[IPOPT_OPTVAL];
1527 
1528 		if (optval == IPOPT_EOL || optval == IPOPT_NOP) {
1529 			olen = 1;
1530 		} else {
1531 			if (cnt < IPOPT_OLEN + 1)
1532 				goto bad;
1533 
1534 			olen = cp[IPOPT_OLEN];
1535 			if (olen < IPOPT_OLEN + 1 || olen > cnt)
1536 				goto bad;
1537 		}
1538 
1539 		if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) {
1540 			/*
1541 			 * user process specifies route as:
1542 			 *	->A->B->C->D
1543 			 * D must be our final destination (but we can't
1544 			 * check that since we may not have connected yet).
1545 			 * A is first hop destination, which doesn't appear in
1546 			 * actual IP option, but is stored before the options.
1547 			 */
1548 			if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr))
1549 				goto bad;
1550 
1551 			offset = cp[IPOPT_OFFSET];
1552 			memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1,
1553 			    sizeof(struct in_addr));
1554 
1555 			cp += sizeof(struct in_addr);
1556 			cnt -= sizeof(struct in_addr);
1557 			olen -= sizeof(struct in_addr);
1558 
1559 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1560 				goto bad;
1561 
1562 			memcpy(dp, cp, olen);
1563 			dp[IPOPT_OPTVAL] = optval;
1564 			dp[IPOPT_OLEN] = olen;
1565 			dp[IPOPT_OFFSET] = offset;
1566 			break;
1567 		} else {
1568 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1569 				goto bad;
1570 
1571 			memcpy(dp, cp, olen);
1572 			break;
1573 		}
1574 
1575 		dp += olen;
1576 		m->m_len += olen;
1577 
1578 		if (optval == IPOPT_EOL)
1579 			break;
1580 
1581 		cp += olen;
1582 		cnt -= olen;
1583 	}
1584 
1585 	inp->inp_options = m;
1586 	return 0;
1587 bad:
1588 	(void)m_free(m);
1589 	return EINVAL;
1590 }
1591 
1592 /*
1593  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1594  * Must be called in a pserialize critical section.
1595  */
1596 static struct ifnet *
1597 ip_multicast_if(struct in_addr *a, int *ifindexp)
1598 {
1599 	int ifindex;
1600 	struct ifnet *ifp = NULL;
1601 	struct in_ifaddr *ia;
1602 
1603 	if (ifindexp)
1604 		*ifindexp = 0;
1605 	if (ntohl(a->s_addr) >> 24 == 0) {
1606 		ifindex = ntohl(a->s_addr) & 0xffffff;
1607 		ifp = if_byindex(ifindex);
1608 		if (!ifp)
1609 			return NULL;
1610 		if (ifindexp)
1611 			*ifindexp = ifindex;
1612 	} else {
1613 		IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) {
1614 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1615 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1616 				ifp = ia->ia_ifp;
1617 				if (if_is_deactivated(ifp))
1618 					ifp = NULL;
1619 				break;
1620 			}
1621 		}
1622 	}
1623 	return ifp;
1624 }
1625 
1626 static int
1627 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval)
1628 {
1629 	u_int tval;
1630 	u_char cval;
1631 	int error;
1632 
1633 	if (sopt == NULL)
1634 		return EINVAL;
1635 
1636 	switch (sopt->sopt_size) {
1637 	case sizeof(u_char):
1638 		error = sockopt_get(sopt, &cval, sizeof(u_char));
1639 		tval = cval;
1640 		break;
1641 
1642 	case sizeof(u_int):
1643 		error = sockopt_get(sopt, &tval, sizeof(u_int));
1644 		break;
1645 
1646 	default:
1647 		error = EINVAL;
1648 	}
1649 
1650 	if (error)
1651 		return error;
1652 
1653 	if (tval > maxval)
1654 		return EINVAL;
1655 
1656 	*val = tval;
1657 	return 0;
1658 }
1659 
1660 static int
1661 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
1662     struct psref *psref, struct in_addr *ia, bool add)
1663 {
1664 	int error;
1665 	struct ip_mreq mreq;
1666 
1667 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
1668 	if (error)
1669 		return error;
1670 
1671 	if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr))
1672 		return EINVAL;
1673 
1674 	memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia));
1675 
1676 	if (in_nullhost(mreq.imr_interface)) {
1677 		union {
1678 			struct sockaddr		dst;
1679 			struct sockaddr_in	dst4;
1680 		} u;
1681 		struct route ro;
1682 
1683 		if (!add) {
1684 			*ifp = NULL;
1685 			return 0;
1686 		}
1687 		/*
1688 		 * If no interface address was provided, use the interface of
1689 		 * the route to the given multicast address.
1690 		 */
1691 		struct rtentry *rt;
1692 		memset(&ro, 0, sizeof(ro));
1693 
1694 		sockaddr_in_init(&u.dst4, ia, 0);
1695 		error = rtcache_setdst(&ro, &u.dst);
1696 		if (error != 0)
1697 			return error;
1698 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
1699 		if (*ifp != NULL) {
1700 			if (if_is_deactivated(*ifp))
1701 				*ifp = NULL;
1702 			else
1703 				if_acquire(*ifp, psref);
1704 		}
1705 		rtcache_unref(rt, &ro);
1706 		rtcache_free(&ro);
1707 	} else {
1708 		int s = pserialize_read_enter();
1709 		*ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1710 		if (!add && *ifp == NULL) {
1711 			pserialize_read_exit(s);
1712 			return EADDRNOTAVAIL;
1713 		}
1714 		if (*ifp != NULL) {
1715 			if (if_is_deactivated(*ifp))
1716 				*ifp = NULL;
1717 			else
1718 				if_acquire(*ifp, psref);
1719 		}
1720 		pserialize_read_exit(s);
1721 	}
1722 	return 0;
1723 }
1724 
1725 /*
1726  * Add a multicast group membership.
1727  * Group must be a valid IP multicast address.
1728  */
1729 static int
1730 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1731 {
1732 	struct ifnet *ifp = NULL;	// XXX: gcc [ppc]
1733 	struct in_addr ia;
1734 	int i, error, bound;
1735 	struct psref psref;
1736 
1737 	/* imo is protected by solock or referenced only by the caller */
1738 
1739 	bound = curlwp_bind();
1740 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1741 		error = ip_get_membership(sopt, &ifp, &psref, &ia, true);
1742 	else
1743 #ifdef INET6
1744 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1745 #else
1746 		error = EINVAL;
1747 		goto out;
1748 #endif
1749 
1750 	if (error)
1751 		goto out;
1752 
1753 	/*
1754 	 * See if we found an interface, and confirm that it
1755 	 * supports multicast.
1756 	 */
1757 	if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1758 		error = EADDRNOTAVAIL;
1759 		goto out;
1760 	}
1761 
1762 	/*
1763 	 * See if the membership already exists or if all the
1764 	 * membership slots are full.
1765 	 */
1766 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1767 		if (imo->imo_membership[i]->inm_ifp == ifp &&
1768 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1769 			break;
1770 	}
1771 	if (i < imo->imo_num_memberships) {
1772 		error = EADDRINUSE;
1773 		goto out;
1774 	}
1775 
1776 	if (i == IP_MAX_MEMBERSHIPS) {
1777 		error = ETOOMANYREFS;
1778 		goto out;
1779 	}
1780 
1781 	/*
1782 	 * Everything looks good; add a new record to the multicast
1783 	 * address list for the given interface.
1784 	 */
1785 	if ((imo->imo_membership[i] = in_addmulti(&ia, ifp)) == NULL) {
1786 		error = ENOBUFS;
1787 		goto out;
1788 	}
1789 
1790 	++imo->imo_num_memberships;
1791 	error = 0;
1792 out:
1793 	if_put(ifp, &psref);
1794 	curlwp_bindx(bound);
1795 	return error;
1796 }
1797 
1798 /*
1799  * Drop a multicast group membership.
1800  * Group must be a valid IP multicast address.
1801  */
1802 static int
1803 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1804 {
1805 	struct in_addr ia = { .s_addr = 0 };	// XXX: gcc [ppc]
1806 	struct ifnet *ifp = NULL;		// XXX: gcc [ppc]
1807 	int i, error, bound;
1808 	struct psref psref;
1809 
1810 	/* imo is protected by solock or referenced only by the caller */
1811 
1812 	bound = curlwp_bind();
1813 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1814 		error = ip_get_membership(sopt, &ifp, &psref, &ia, false);
1815 	else
1816 #ifdef INET6
1817 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1818 #else
1819 		error = EINVAL;
1820 		goto out;
1821 #endif
1822 
1823 	if (error)
1824 		goto out;
1825 
1826 	/*
1827 	 * Find the membership in the membership array.
1828 	 */
1829 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1830 		if ((ifp == NULL ||
1831 		     imo->imo_membership[i]->inm_ifp == ifp) &&
1832 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1833 			break;
1834 	}
1835 	if (i == imo->imo_num_memberships) {
1836 		error = EADDRNOTAVAIL;
1837 		goto out;
1838 	}
1839 
1840 	/*
1841 	 * Give up the multicast address record to which the
1842 	 * membership points.
1843 	 */
1844 	in_delmulti(imo->imo_membership[i]);
1845 
1846 	/*
1847 	 * Remove the gap in the membership array.
1848 	 */
1849 	for (++i; i < imo->imo_num_memberships; ++i)
1850 		imo->imo_membership[i-1] = imo->imo_membership[i];
1851 	--imo->imo_num_memberships;
1852 	error = 0;
1853 out:
1854 	if_put(ifp, &psref);
1855 	curlwp_bindx(bound);
1856 	return error;
1857 }
1858 
1859 /*
1860  * Set the IP multicast options in response to user setsockopt().
1861  */
1862 int
1863 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt)
1864 {
1865 	struct ip_moptions *imo = *pimo;
1866 	struct in_addr addr;
1867 	struct ifnet *ifp;
1868 	int ifindex, error = 0;
1869 
1870 	/* The passed imo isn't NULL, it should be protected by solock */
1871 
1872 	if (!imo) {
1873 		/*
1874 		 * No multicast option buffer attached to the pcb;
1875 		 * allocate one and initialize to default values.
1876 		 */
1877 		imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP);
1878 		if (imo == NULL)
1879 			return ENOBUFS;
1880 
1881 		imo->imo_multicast_if_index = 0;
1882 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1883 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1884 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1885 		imo->imo_num_memberships = 0;
1886 		*pimo = imo;
1887 	}
1888 
1889 	switch (sopt->sopt_name) {
1890 	case IP_MULTICAST_IF: {
1891 		int s;
1892 		/*
1893 		 * Select the interface for outgoing multicast packets.
1894 		 */
1895 		error = sockopt_get(sopt, &addr, sizeof(addr));
1896 		if (error)
1897 			break;
1898 
1899 		/*
1900 		 * INADDR_ANY is used to remove a previous selection.
1901 		 * When no interface is selected, a default one is
1902 		 * chosen every time a multicast packet is sent.
1903 		 */
1904 		if (in_nullhost(addr)) {
1905 			imo->imo_multicast_if_index = 0;
1906 			break;
1907 		}
1908 		/*
1909 		 * The selected interface is identified by its local
1910 		 * IP address.  Find the interface and confirm that
1911 		 * it supports multicasting.
1912 		 */
1913 		s = pserialize_read_enter();
1914 		ifp = ip_multicast_if(&addr, &ifindex);
1915 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1916 			pserialize_read_exit(s);
1917 			error = EADDRNOTAVAIL;
1918 			break;
1919 		}
1920 		imo->imo_multicast_if_index = ifp->if_index;
1921 		pserialize_read_exit(s);
1922 		if (ifindex)
1923 			imo->imo_multicast_addr = addr;
1924 		else
1925 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1926 		break;
1927 	    }
1928 
1929 	case IP_MULTICAST_TTL:
1930 		/*
1931 		 * Set the IP time-to-live for outgoing multicast packets.
1932 		 */
1933 		error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL);
1934 		break;
1935 
1936 	case IP_MULTICAST_LOOP:
1937 		/*
1938 		 * Set the loopback flag for outgoing multicast packets.
1939 		 * Must be zero or one.
1940 		 */
1941 		error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1);
1942 		break;
1943 
1944 	case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */
1945 		error = ip_add_membership(imo, sopt);
1946 		break;
1947 
1948 	case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */
1949 		error = ip_drop_membership(imo, sopt);
1950 		break;
1951 
1952 	default:
1953 		error = EOPNOTSUPP;
1954 		break;
1955 	}
1956 
1957 	/*
1958 	 * If all options have default values, no need to keep the mbuf.
1959 	 */
1960 	if (imo->imo_multicast_if_index == 0 &&
1961 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1962 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1963 	    imo->imo_num_memberships == 0) {
1964 		kmem_intr_free(imo, sizeof(*imo));
1965 		*pimo = NULL;
1966 	}
1967 
1968 	return error;
1969 }
1970 
1971 /*
1972  * Return the IP multicast options in response to user getsockopt().
1973  */
1974 int
1975 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt)
1976 {
1977 	struct in_addr addr;
1978 	uint8_t optval;
1979 	int error = 0;
1980 
1981 	/* imo is protected by solock or refereced only by the caller */
1982 
1983 	switch (sopt->sopt_name) {
1984 	case IP_MULTICAST_IF:
1985 		if (imo == NULL || imo->imo_multicast_if_index == 0)
1986 			addr = zeroin_addr;
1987 		else if (imo->imo_multicast_addr.s_addr) {
1988 			/* return the value user has set */
1989 			addr = imo->imo_multicast_addr;
1990 		} else {
1991 			struct ifnet *ifp;
1992 			struct in_ifaddr *ia = NULL;
1993 			int s = pserialize_read_enter();
1994 
1995 			ifp = if_byindex(imo->imo_multicast_if_index);
1996 			if (ifp != NULL) {
1997 				ia = in_get_ia_from_ifp(ifp);
1998 			}
1999 			addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
2000 			pserialize_read_exit(s);
2001 		}
2002 		error = sockopt_set(sopt, &addr, sizeof(addr));
2003 		break;
2004 
2005 	case IP_MULTICAST_TTL:
2006 		optval = imo ? imo->imo_multicast_ttl
2007 		    : IP_DEFAULT_MULTICAST_TTL;
2008 
2009 		error = sockopt_set(sopt, &optval, sizeof(optval));
2010 		break;
2011 
2012 	case IP_MULTICAST_LOOP:
2013 		optval = imo ? imo->imo_multicast_loop
2014 		    : IP_DEFAULT_MULTICAST_LOOP;
2015 
2016 		error = sockopt_set(sopt, &optval, sizeof(optval));
2017 		break;
2018 
2019 	default:
2020 		error = EOPNOTSUPP;
2021 	}
2022 
2023 	return error;
2024 }
2025 
2026 /*
2027  * Discard the IP multicast options.
2028  */
2029 void
2030 ip_freemoptions(struct ip_moptions *imo)
2031 {
2032 	int i;
2033 
2034 	/* The owner of imo (inp) should be protected by solock */
2035 
2036 	if (imo != NULL) {
2037 		for (i = 0; i < imo->imo_num_memberships; ++i)
2038 			in_delmulti(imo->imo_membership[i]);
2039 		kmem_intr_free(imo, sizeof(*imo));
2040 	}
2041 }
2042 
2043 /*
2044  * Routine called from ip_output() to loop back a copy of an IP multicast
2045  * packet to the input queue of a specified interface.  Note that this
2046  * calls the output routine of the loopback "driver", but with an interface
2047  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2048  */
2049 static void
2050 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst)
2051 {
2052 	struct ip *ip;
2053 	struct mbuf *copym;
2054 
2055 	copym = m_copypacket(m, M_DONTWAIT);
2056 	if (copym != NULL &&
2057 	    (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
2058 		copym = m_pullup(copym, sizeof(struct ip));
2059 	if (copym == NULL)
2060 		return;
2061 	/*
2062 	 * We don't bother to fragment if the IP length is greater
2063 	 * than the interface's MTU.  Can this possibly matter?
2064 	 */
2065 	ip = mtod(copym, struct ip *);
2066 
2067 	if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
2068 		in_delayed_cksum(copym);
2069 		copym->m_pkthdr.csum_flags &=
2070 		    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
2071 	}
2072 
2073 	ip->ip_sum = 0;
2074 	ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2075 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
2076 	(void)looutput(ifp, copym, sintocsa(dst), NULL);
2077 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
2078 }
2079 
2080 /*
2081  * Ensure sending address is valid.
2082  * Returns 0 on success, -1 if an error should be sent back or 1
2083  * if the packet could be dropped without error (protocol dependent).
2084  */
2085 static int
2086 ip_ifaddrvalid(const struct in_ifaddr *ia)
2087 {
2088 
2089 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
2090 		return 0;
2091 
2092 	if (ia->ia4_flags & IN_IFF_DUPLICATED)
2093 		return -1;
2094 	else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED))
2095 		return 1;
2096 
2097 	return 0;
2098 }
2099