xref: /netbsd-src/sys/netinet/ip_output.c (revision 7330f729ccf0bd976a06f95fad452fe774fc7fd1)
1 /*	$NetBSD: ip_output.c,v 1.314 2019/06/05 01:31:04 knakahara 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 1.314 2019/06/05 01:31:04 knakahara 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 len, hlen, error = 0;
236 	struct route iproute;
237 	const struct sockaddr_in *dst;
238 	struct in_ifaddr *ia = NULL;
239 	struct ifaddr *ifa;
240 	int isbroadcast;
241 	int sw_csum;
242 	u_long mtu;
243 	bool natt_frag = false;
244 	bool rtmtu_nolock;
245 	union {
246 		struct sockaddr		sa;
247 		struct sockaddr_in	sin;
248 	} udst, usrc;
249 	struct sockaddr *rdst = &udst.sa;	/* real IP destination, as
250 						 * opposed to the nexthop
251 						 */
252 	struct psref psref, psref_ia;
253 	int bound;
254 	bool bind_need_restore = false;
255 
256 	len = 0;
257 
258 	MCLAIM(m, &ip_tx_mowner);
259 
260 	KASSERT((m->m_flags & M_PKTHDR) != 0);
261 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0);
262 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) !=
263 	    (M_CSUM_TCPv4|M_CSUM_UDPv4));
264 	KASSERT(m->m_len >= sizeof(struct ip));
265 
266 	hlen = sizeof(struct ip);
267 	if (opt) {
268 		m = ip_insertoptions(m, opt, &len);
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 	/* XXX must be before rtcache operations */
307 	bound = curlwp_bind();
308 	bind_need_restore = true;
309 
310 	if ((rt = rtcache_validate(ro)) == NULL &&
311 	    (rt = rtcache_update(ro, 1)) == NULL) {
312 		dst = &udst.sin;
313 		error = rtcache_setdst(ro, &udst.sa);
314 		if (error != 0)
315 			goto bad;
316 	}
317 
318 	/*
319 	 * If routing to interface only, short circuit routing lookup.
320 	 */
321 	if (flags & IP_ROUTETOIF) {
322 		ifa = ifa_ifwithladdr_psref(sintocsa(dst), &psref_ia);
323 		if (ifa == NULL) {
324 			IP_STATINC(IP_STAT_NOROUTE);
325 			error = ENETUNREACH;
326 			goto bad;
327 		}
328 		/* ia is already referenced by psref_ia */
329 		ia = ifatoia(ifa);
330 
331 		ifp = ia->ia_ifp;
332 		mtu = ifp->if_mtu;
333 		ip->ip_ttl = 1;
334 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
335 	} else if (((IN_MULTICAST(ip->ip_dst.s_addr) ||
336 	    ip->ip_dst.s_addr == INADDR_BROADCAST) ||
337 	    (flags & IP_ROUTETOIFINDEX)) &&
338 	    imo != NULL && imo->imo_multicast_if_index != 0) {
339 		ifp = mifp = if_get_byindex(imo->imo_multicast_if_index, &psref);
340 		if (ifp == NULL) {
341 			IP_STATINC(IP_STAT_NOROUTE);
342 			error = ENETUNREACH;
343 			goto bad;
344 		}
345 		mtu = ifp->if_mtu;
346 		ia = in_get_ia_from_ifp_psref(ifp, &psref_ia);
347 		if (ia == NULL) {
348 			error = EADDRNOTAVAIL;
349 			goto bad;
350 		}
351 		if (IN_MULTICAST(ip->ip_dst.s_addr) ||
352 		    ip->ip_dst.s_addr == INADDR_BROADCAST) {
353 			isbroadcast = 0;
354 		} else {
355 			/* IP_ROUTETOIFINDEX */
356 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
357 			if ((isbroadcast == 0) && ((ifp->if_flags &
358 			    (IFF_LOOPBACK | IFF_POINTOPOINT)) == 0) &&
359 			    (in_direct(dst->sin_addr, ifp) == 0)) {
360 				/* gateway address required */
361 				if (rt == NULL)
362 					rt = rtcache_init(ro);
363 				if (rt == NULL || rt->rt_ifp != ifp) {
364 					IP_STATINC(IP_STAT_NOROUTE);
365 					error = EHOSTUNREACH;
366 					goto bad;
367 				}
368 				rt->rt_use++;
369 				if (rt->rt_flags & RTF_GATEWAY)
370 					dst = satosin(rt->rt_gateway);
371 				if (rt->rt_flags & RTF_HOST)
372 					isbroadcast =
373 					    rt->rt_flags & RTF_BROADCAST;
374 			}
375 		}
376 	} else {
377 		if (rt == NULL)
378 			rt = rtcache_init(ro);
379 		if (rt == NULL) {
380 			IP_STATINC(IP_STAT_NOROUTE);
381 			error = EHOSTUNREACH;
382 			goto bad;
383 		}
384 		if (ifa_is_destroying(rt->rt_ifa)) {
385 			rtcache_unref(rt, ro);
386 			rt = NULL;
387 			IP_STATINC(IP_STAT_NOROUTE);
388 			error = EHOSTUNREACH;
389 			goto bad;
390 		}
391 		ifa_acquire(rt->rt_ifa, &psref_ia);
392 		ia = ifatoia(rt->rt_ifa);
393 		ifp = rt->rt_ifp;
394 		if ((mtu = rt->rt_rmx.rmx_mtu) == 0)
395 			mtu = ifp->if_mtu;
396 		rt->rt_use++;
397 		if (rt->rt_flags & RTF_GATEWAY)
398 			dst = satosin(rt->rt_gateway);
399 		if (rt->rt_flags & RTF_HOST)
400 			isbroadcast = rt->rt_flags & RTF_BROADCAST;
401 		else
402 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
403 	}
404 	rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0;
405 
406 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
407 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
408 		bool inmgroup;
409 
410 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
411 		    M_BCAST : M_MCAST;
412 		/*
413 		 * See if the caller provided any multicast options
414 		 */
415 		if (imo != NULL)
416 			ip->ip_ttl = imo->imo_multicast_ttl;
417 		else
418 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
419 
420 		/*
421 		 * if we don't know the outgoing ifp yet, we can't generate
422 		 * output
423 		 */
424 		if (!ifp) {
425 			IP_STATINC(IP_STAT_NOROUTE);
426 			error = ENETUNREACH;
427 			goto bad;
428 		}
429 
430 		/*
431 		 * If the packet is multicast or broadcast, confirm that
432 		 * the outgoing interface can transmit it.
433 		 */
434 		if (((m->m_flags & M_MCAST) &&
435 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
436 		    ((m->m_flags & M_BCAST) &&
437 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
438 			IP_STATINC(IP_STAT_NOROUTE);
439 			error = ENETUNREACH;
440 			goto bad;
441 		}
442 		/*
443 		 * If source address not specified yet, use an address
444 		 * of outgoing interface.
445 		 */
446 		if (in_nullhost(ip->ip_src)) {
447 			struct in_ifaddr *xia;
448 			struct ifaddr *xifa;
449 			struct psref _psref;
450 
451 			xia = in_get_ia_from_ifp_psref(ifp, &_psref);
452 			if (!xia) {
453 				error = EADDRNOTAVAIL;
454 				goto bad;
455 			}
456 			xifa = &xia->ia_ifa;
457 			if (xifa->ifa_getifa != NULL) {
458 				ia4_release(xia, &_psref);
459 				/* FIXME ifa_getifa is NOMPSAFE */
460 				xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
461 				if (xia == NULL) {
462 					error = EADDRNOTAVAIL;
463 					goto bad;
464 				}
465 				ia4_acquire(xia, &_psref);
466 			}
467 			ip->ip_src = xia->ia_addr.sin_addr;
468 			ia4_release(xia, &_psref);
469 		}
470 
471 		inmgroup = in_multi_group(ip->ip_dst, ifp, flags);
472 		if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) {
473 			/*
474 			 * If we belong to the destination multicast group
475 			 * on the outgoing interface, and the caller did not
476 			 * forbid loopback, loop back a copy.
477 			 */
478 			ip_mloopback(ifp, m, &udst.sin);
479 		}
480 #ifdef MROUTING
481 		else {
482 			/*
483 			 * If we are acting as a multicast router, perform
484 			 * multicast forwarding as if the packet had just
485 			 * arrived on the interface to which we are about
486 			 * to send.  The multicast forwarding function
487 			 * recursively calls this function, using the
488 			 * IP_FORWARDING flag to prevent infinite recursion.
489 			 *
490 			 * Multicasts that are looped back by ip_mloopback(),
491 			 * above, will be forwarded by the ip_input() routine,
492 			 * if necessary.
493 			 */
494 			extern struct socket *ip_mrouter;
495 
496 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
497 				if (ip_mforward(m, ifp) != 0) {
498 					m_freem(m);
499 					goto done;
500 				}
501 			}
502 		}
503 #endif
504 		/*
505 		 * Multicasts with a time-to-live of zero may be looped-
506 		 * back, above, but must not be transmitted on a network.
507 		 * Also, multicasts addressed to the loopback interface
508 		 * are not sent -- the above call to ip_mloopback() will
509 		 * loop back a copy if this host actually belongs to the
510 		 * destination group on the loopback interface.
511 		 */
512 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
513 			m_freem(m);
514 			goto done;
515 		}
516 		goto sendit;
517 	}
518 
519 	/*
520 	 * If source address not specified yet, use address
521 	 * of outgoing interface.
522 	 */
523 	if (in_nullhost(ip->ip_src)) {
524 		struct ifaddr *xifa;
525 
526 		xifa = &ia->ia_ifa;
527 		if (xifa->ifa_getifa != NULL) {
528 			ia4_release(ia, &psref_ia);
529 			/* FIXME ifa_getifa is NOMPSAFE */
530 			ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
531 			if (ia == NULL) {
532 				error = EADDRNOTAVAIL;
533 				goto bad;
534 			}
535 			ia4_acquire(ia, &psref_ia);
536 		}
537 		ip->ip_src = ia->ia_addr.sin_addr;
538 	}
539 
540 	/*
541 	 * Packets with Class-D address as source are not valid per
542 	 * RFC1112.
543 	 */
544 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
545 		IP_STATINC(IP_STAT_ODROPPED);
546 		error = EADDRNOTAVAIL;
547 		goto bad;
548 	}
549 
550 	/*
551 	 * Look for broadcast address and verify user is allowed to
552 	 * send such a packet.
553 	 */
554 	if (isbroadcast) {
555 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
556 			error = EADDRNOTAVAIL;
557 			goto bad;
558 		}
559 		if ((flags & IP_ALLOWBROADCAST) == 0) {
560 			error = EACCES;
561 			goto bad;
562 		}
563 		/* don't allow broadcast messages to be fragmented */
564 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
565 			error = EMSGSIZE;
566 			goto bad;
567 		}
568 		m->m_flags |= M_BCAST;
569 	} else
570 		m->m_flags &= ~M_BCAST;
571 
572 sendit:
573 	if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) {
574 		if (m->m_pkthdr.len < IP_MINFRAGSIZE) {
575 			ip->ip_id = 0;
576 		} else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
577 			ip->ip_id = ip_newid(ia);
578 		} else {
579 			/*
580 			 * TSO capable interfaces (typically?) increment
581 			 * ip_id for each segment.
582 			 * "allocate" enough ids here to increase the chance
583 			 * for them to be unique.
584 			 *
585 			 * note that the following calculation is not
586 			 * needed to be precise.  wasting some ip_id is fine.
587 			 */
588 
589 			unsigned int segsz = m->m_pkthdr.segsz;
590 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
591 			unsigned int num = howmany(datasz, segsz);
592 
593 			ip->ip_id = ip_newid_range(ia, num);
594 		}
595 	}
596 	if (ia != NULL) {
597 		ia4_release(ia, &psref_ia);
598 		ia = NULL;
599 	}
600 
601 	/*
602 	 * If we're doing Path MTU Discovery, we need to set DF unless
603 	 * the route's MTU is locked.
604 	 */
605 	if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) {
606 		ip->ip_off |= htons(IP_DF);
607 	}
608 
609 #ifdef IPSEC
610 	if (ipsec_used) {
611 		bool ipsec_done = false;
612 
613 		/* Perform IPsec processing, if any. */
614 		error = ipsec4_output(m, inp, flags, &mtu, &natt_frag,
615 		    &ipsec_done);
616 		if (error || ipsec_done)
617 			goto done;
618 	}
619 
620 	if (!ipsec_used || !natt_frag)
621 #endif
622 	{
623 		/*
624 		 * Run through list of hooks for output packets.
625 		 */
626 		error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT);
627 		if (error || m == NULL) {
628 			IP_STATINC(IP_STAT_PFILDROP_OUT);
629 			goto done;
630 		}
631 	}
632 
633 	ip = mtod(m, struct ip *);
634 	hlen = ip->ip_hl << 2;
635 
636 	m->m_pkthdr.csum_data |= hlen << 16;
637 
638 	/*
639 	 * search for the source address structure to
640 	 * maintain output statistics, and verify address
641 	 * validity
642 	 */
643 	KASSERT(ia == NULL);
644 	sockaddr_in_init(&usrc.sin, &ip->ip_src, 0);
645 	ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia);
646 	if (ifa != NULL)
647 		ia = ifatoia(ifa);
648 
649 	/*
650 	 * Ensure we only send from a valid address.
651 	 * A NULL address is valid because the packet could be
652 	 * generated from a packet filter.
653 	 */
654 	if (ia != NULL && (flags & IP_FORWARDING) == 0 &&
655 	    (error = ip_ifaddrvalid(ia)) != 0)
656 	{
657 		ARPLOG(LOG_ERR,
658 		    "refusing to send from invalid address %s (pid %d)\n",
659 		    ARPLOGADDR(&ip->ip_src), curproc->p_pid);
660 		IP_STATINC(IP_STAT_ODROPPED);
661 		if (error == 1)
662 			/*
663 			 * Address exists, but is tentative or detached.
664 			 * We can't send from it because it's invalid,
665 			 * so we drop the packet.
666 			 */
667 			error = 0;
668 		else
669 			error = EADDRNOTAVAIL;
670 		goto bad;
671 	}
672 
673 	/* Maybe skip checksums on loopback interfaces. */
674 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
675 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
676 	}
677 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
678 
679 	/*
680 	 * If small enough for mtu of path, or if using TCP segmentation
681 	 * offload, can just send directly.
682 	 */
683 	if (ntohs(ip->ip_len) <= mtu ||
684 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
685 		const struct sockaddr *sa;
686 
687 #if IFA_STATS
688 		if (ia)
689 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
690 #endif
691 		/*
692 		 * Always initialize the sum to 0!  Some HW assisted
693 		 * checksumming requires this.
694 		 */
695 		ip->ip_sum = 0;
696 
697 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
698 			/*
699 			 * Perform any checksums that the hardware can't do
700 			 * for us.
701 			 *
702 			 * XXX Does any hardware require the {th,uh}_sum
703 			 * XXX fields to be 0?
704 			 */
705 			if (sw_csum & M_CSUM_IPv4) {
706 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
707 				ip->ip_sum = in_cksum(m, hlen);
708 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
709 			}
710 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
711 				if (IN_NEED_CHECKSUM(ifp,
712 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
713 					in_undefer_cksum_tcpudp(m);
714 				}
715 				m->m_pkthdr.csum_flags &=
716 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
717 			}
718 		}
719 
720 		sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst);
721 		if (__predict_false(sw_csum & M_CSUM_TSOv4)) {
722 			/*
723 			 * TSO4 is required by a packet, but disabled for
724 			 * the interface.
725 			 */
726 			error = ip_tso_output(ifp, m, sa, rt);
727 		} else
728 			error = ip_if_output(ifp, m, sa, rt);
729 		goto done;
730 	}
731 
732 	/*
733 	 * We can't use HW checksumming if we're about to fragment the packet.
734 	 *
735 	 * XXX Some hardware can do this.
736 	 */
737 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
738 		if (IN_NEED_CHECKSUM(ifp,
739 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
740 			in_undefer_cksum_tcpudp(m);
741 		}
742 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
743 	}
744 
745 	/*
746 	 * Too large for interface; fragment if possible.
747 	 * Must be able to put at least 8 bytes per fragment.
748 	 */
749 	if (ntohs(ip->ip_off) & IP_DF) {
750 		if (flags & IP_RETURNMTU) {
751 			KASSERT(inp != NULL);
752 			inp->inp_errormtu = mtu;
753 		}
754 		error = EMSGSIZE;
755 		IP_STATINC(IP_STAT_CANTFRAG);
756 		goto bad;
757 	}
758 
759 	error = ip_fragment(m, ifp, mtu);
760 	if (error) {
761 		m = NULL;
762 		goto bad;
763 	}
764 
765 	for (; m; m = m0) {
766 		m0 = m->m_nextpkt;
767 		m->m_nextpkt = NULL;
768 		if (error) {
769 			m_freem(m);
770 			continue;
771 		}
772 #if IFA_STATS
773 		if (ia)
774 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
775 #endif
776 		/*
777 		 * If we get there, the packet has not been handled by
778 		 * IPsec whereas it should have. Now that it has been
779 		 * fragmented, re-inject it in ip_output so that IPsec
780 		 * processing can occur.
781 		 */
782 		if (natt_frag) {
783 			error = ip_output(m, opt, NULL,
784 			    flags | IP_RAWOUTPUT | IP_NOIPNEWID,
785 			    imo, inp);
786 		} else {
787 			KASSERT((m->m_pkthdr.csum_flags &
788 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
789 			error = ip_if_output(ifp, m,
790 			    (m->m_flags & M_MCAST) ?
791 			    sintocsa(rdst) : sintocsa(dst), rt);
792 		}
793 	}
794 	if (error == 0) {
795 		IP_STATINC(IP_STAT_FRAGMENTED);
796 	}
797 
798 done:
799 	ia4_release(ia, &psref_ia);
800 	rtcache_unref(rt, ro);
801 	if (ro == &iproute) {
802 		rtcache_free(&iproute);
803 	}
804 	if (mifp != NULL) {
805 		if_put(mifp, &psref);
806 	}
807 	if (bind_need_restore)
808 		curlwp_bindx(bound);
809 	return error;
810 
811 bad:
812 	m_freem(m);
813 	goto done;
814 }
815 
816 int
817 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
818 {
819 	struct ip *ip, *mhip;
820 	struct mbuf *m0;
821 	int len, hlen, off;
822 	int mhlen, firstlen;
823 	struct mbuf **mnext;
824 	int sw_csum = m->m_pkthdr.csum_flags;
825 	int fragments = 0;
826 	int error = 0;
827 	int ipoff, ipflg;
828 
829 	ip = mtod(m, struct ip *);
830 	hlen = ip->ip_hl << 2;
831 
832 	/* Preserve the offset and flags. */
833 	ipoff = ntohs(ip->ip_off) & IP_OFFMASK;
834 	ipflg = ntohs(ip->ip_off) & (IP_RF|IP_DF|IP_MF);
835 
836 	if (ifp != NULL)
837 		sw_csum &= ~ifp->if_csum_flags_tx;
838 
839 	len = (mtu - hlen) &~ 7;
840 	if (len < 8) {
841 		m_freem(m);
842 		return EMSGSIZE;
843 	}
844 
845 	firstlen = len;
846 	mnext = &m->m_nextpkt;
847 
848 	/*
849 	 * Loop through length of segment after first fragment,
850 	 * make new header and copy data of each part and link onto chain.
851 	 */
852 	m0 = m;
853 	mhlen = sizeof(struct ip);
854 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
855 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
856 		if (m == NULL) {
857 			error = ENOBUFS;
858 			IP_STATINC(IP_STAT_ODROPPED);
859 			goto sendorfree;
860 		}
861 		MCLAIM(m, m0->m_owner);
862 
863 		*mnext = m;
864 		mnext = &m->m_nextpkt;
865 
866 		m->m_data += max_linkhdr;
867 		mhip = mtod(m, struct ip *);
868 		*mhip = *ip;
869 
870 		/* we must inherit the flags */
871 		m->m_flags |= m0->m_flags & M_COPYFLAGS;
872 
873 		if (hlen > sizeof(struct ip)) {
874 			mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
875 			mhip->ip_hl = mhlen >> 2;
876 		}
877 		m->m_len = mhlen;
878 
879 		mhip->ip_off = ((off - hlen) >> 3) + ipoff;
880 		mhip->ip_off |= ipflg;
881 		if (off + len >= ntohs(ip->ip_len))
882 			len = ntohs(ip->ip_len) - off;
883 		else
884 			mhip->ip_off |= IP_MF;
885 		HTONS(mhip->ip_off);
886 
887 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
888 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
889 		if (m->m_next == NULL) {
890 			error = ENOBUFS;
891 			IP_STATINC(IP_STAT_ODROPPED);
892 			goto sendorfree;
893 		}
894 
895 		m->m_pkthdr.len = mhlen + len;
896 		m_reset_rcvif(m);
897 
898 		mhip->ip_sum = 0;
899 		KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
900 		if (sw_csum & M_CSUM_IPv4) {
901 			mhip->ip_sum = in_cksum(m, mhlen);
902 		} else {
903 			/*
904 			 * checksum is hw-offloaded or not necessary.
905 			 */
906 			m->m_pkthdr.csum_flags |=
907 			    m0->m_pkthdr.csum_flags & M_CSUM_IPv4;
908 			m->m_pkthdr.csum_data |= mhlen << 16;
909 			KASSERT(!(ifp != NULL &&
910 			    IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
911 			    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
912 		}
913 		IP_STATINC(IP_STAT_OFRAGMENTS);
914 		fragments++;
915 	}
916 
917 	/*
918 	 * Update first fragment by trimming what's been copied out
919 	 * and updating header, then send each fragment (in order).
920 	 */
921 	m = m0;
922 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
923 	m->m_pkthdr.len = hlen + firstlen;
924 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
925 	ip->ip_off |= htons(IP_MF);
926 	ip->ip_sum = 0;
927 	if (sw_csum & M_CSUM_IPv4) {
928 		ip->ip_sum = in_cksum(m, hlen);
929 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
930 	} else {
931 		/*
932 		 * checksum is hw-offloaded or not necessary.
933 		 */
934 		KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
935 		    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
936 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
937 		    sizeof(struct ip));
938 	}
939 
940 sendorfree:
941 	/*
942 	 * If there is no room for all the fragments, don't queue
943 	 * any of them.
944 	 */
945 	if (ifp != NULL) {
946 		IFQ_LOCK(&ifp->if_snd);
947 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
948 		    error == 0) {
949 			error = ENOBUFS;
950 			IP_STATINC(IP_STAT_ODROPPED);
951 			IFQ_INC_DROPS(&ifp->if_snd);
952 		}
953 		IFQ_UNLOCK(&ifp->if_snd);
954 	}
955 	if (error) {
956 		for (m = m0; m; m = m0) {
957 			m0 = m->m_nextpkt;
958 			m->m_nextpkt = NULL;
959 			m_freem(m);
960 		}
961 	}
962 
963 	return error;
964 }
965 
966 /*
967  * Determine the maximum length of the options to be inserted;
968  * we would far rather allocate too much space rather than too little.
969  */
970 u_int
971 ip_optlen(struct inpcb *inp)
972 {
973 	struct mbuf *m = inp->inp_options;
974 
975 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) {
976 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
977 	}
978 	return 0;
979 }
980 
981 /*
982  * Insert IP options into preformed packet.
983  * Adjust IP destination as required for IP source routing,
984  * as indicated by a non-zero in_addr at the start of the options.
985  */
986 static struct mbuf *
987 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
988 {
989 	struct ipoption *p = mtod(opt, struct ipoption *);
990 	struct mbuf *n;
991 	struct ip *ip = mtod(m, struct ip *);
992 	unsigned optlen;
993 
994 	optlen = opt->m_len - sizeof(p->ipopt_dst);
995 	KASSERT(optlen % 4 == 0);
996 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
997 		return m;		/* XXX should fail */
998 	if (!in_nullhost(p->ipopt_dst))
999 		ip->ip_dst = p->ipopt_dst;
1000 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1001 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1002 		if (n == NULL)
1003 			return m;
1004 		MCLAIM(n, m->m_owner);
1005 		m_move_pkthdr(n, m);
1006 		m->m_len -= sizeof(struct ip);
1007 		m->m_data += sizeof(struct ip);
1008 		n->m_next = m;
1009 		n->m_len = optlen + sizeof(struct ip);
1010 		n->m_data += max_linkhdr;
1011 		memcpy(mtod(n, void *), ip, sizeof(struct ip));
1012 		m = n;
1013 	} else {
1014 		m->m_data -= optlen;
1015 		m->m_len += optlen;
1016 		memmove(mtod(m, void *), ip, sizeof(struct ip));
1017 	}
1018 	m->m_pkthdr.len += optlen;
1019 	ip = mtod(m, struct ip *);
1020 	memcpy(ip + 1, p->ipopt_list, optlen);
1021 	*phlen = sizeof(struct ip) + optlen;
1022 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1023 	return m;
1024 }
1025 
1026 /*
1027  * Copy options from ipsrc to ipdst, omitting those not copied during
1028  * fragmentation.
1029  */
1030 int
1031 ip_optcopy(struct ip *ipsrc, struct ip *ipdst)
1032 {
1033 	u_char *cp, *dp;
1034 	int opt, optlen, cnt;
1035 
1036 	cp = (u_char *)(ipsrc + 1);
1037 	dp = (u_char *)(ipdst + 1);
1038 	cnt = (ipsrc->ip_hl << 2) - sizeof(struct ip);
1039 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1040 		opt = cp[0];
1041 		if (opt == IPOPT_EOL)
1042 			break;
1043 		if (opt == IPOPT_NOP) {
1044 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1045 			*dp++ = IPOPT_NOP;
1046 			optlen = 1;
1047 			continue;
1048 		}
1049 
1050 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp));
1051 		optlen = cp[IPOPT_OLEN];
1052 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt);
1053 
1054 		/* Invalid lengths should have been caught by ip_dooptions. */
1055 		if (optlen > cnt)
1056 			optlen = cnt;
1057 		if (IPOPT_COPIED(opt)) {
1058 			bcopy((void *)cp, (void *)dp, (unsigned)optlen);
1059 			dp += optlen;
1060 		}
1061 	}
1062 
1063 	for (optlen = dp - (u_char *)(ipdst+1); optlen & 0x3; optlen++) {
1064 		*dp++ = IPOPT_EOL;
1065 	}
1066 
1067 	return optlen;
1068 }
1069 
1070 /*
1071  * IP socket option processing.
1072  */
1073 int
1074 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1075 {
1076 	struct inpcb *inp = sotoinpcb(so);
1077 	struct ip *ip = &inp->inp_ip;
1078 	int inpflags = inp->inp_flags;
1079 	int optval = 0, error = 0;
1080 	struct in_pktinfo pktinfo;
1081 
1082 	KASSERT(solocked(so));
1083 
1084 	if (sopt->sopt_level != IPPROTO_IP) {
1085 		if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER)
1086 			return 0;
1087 		return ENOPROTOOPT;
1088 	}
1089 
1090 	switch (op) {
1091 	case PRCO_SETOPT:
1092 		switch (sopt->sopt_name) {
1093 		case IP_OPTIONS:
1094 #ifdef notyet
1095 		case IP_RETOPTS:
1096 #endif
1097 			error = ip_pcbopts(inp, sopt);
1098 			break;
1099 
1100 		case IP_TOS:
1101 		case IP_TTL:
1102 		case IP_MINTTL:
1103 		case IP_RECVOPTS:
1104 		case IP_RECVRETOPTS:
1105 		case IP_RECVDSTADDR:
1106 		case IP_RECVIF:
1107 		case IP_RECVPKTINFO:
1108 		case IP_RECVTTL:
1109 			error = sockopt_getint(sopt, &optval);
1110 			if (error)
1111 				break;
1112 
1113 			switch (sopt->sopt_name) {
1114 			case IP_TOS:
1115 				ip->ip_tos = optval;
1116 				break;
1117 
1118 			case IP_TTL:
1119 				ip->ip_ttl = optval;
1120 				break;
1121 
1122 			case IP_MINTTL:
1123 				if (optval > 0 && optval <= MAXTTL)
1124 					inp->inp_ip_minttl = optval;
1125 				else
1126 					error = EINVAL;
1127 				break;
1128 #define	OPTSET(bit) \
1129 	if (optval) \
1130 		inpflags |= bit; \
1131 	else \
1132 		inpflags &= ~bit;
1133 
1134 			case IP_RECVOPTS:
1135 				OPTSET(INP_RECVOPTS);
1136 				break;
1137 
1138 			case IP_RECVPKTINFO:
1139 				OPTSET(INP_RECVPKTINFO);
1140 				break;
1141 
1142 			case IP_RECVRETOPTS:
1143 				OPTSET(INP_RECVRETOPTS);
1144 				break;
1145 
1146 			case IP_RECVDSTADDR:
1147 				OPTSET(INP_RECVDSTADDR);
1148 				break;
1149 
1150 			case IP_RECVIF:
1151 				OPTSET(INP_RECVIF);
1152 				break;
1153 
1154 			case IP_RECVTTL:
1155 				OPTSET(INP_RECVTTL);
1156 				break;
1157 			}
1158 			break;
1159 		case IP_PKTINFO:
1160 			error = sockopt_getint(sopt, &optval);
1161 			if (!error) {
1162 				/* Linux compatibility */
1163 				OPTSET(INP_RECVPKTINFO);
1164 				break;
1165 			}
1166 			error = sockopt_get(sopt, &pktinfo, sizeof(pktinfo));
1167 			if (error)
1168 				break;
1169 
1170 			if (pktinfo.ipi_ifindex == 0) {
1171 				inp->inp_prefsrcip = pktinfo.ipi_addr;
1172 				break;
1173 			}
1174 
1175 			/* Solaris compatibility */
1176 			struct ifnet *ifp;
1177 			struct in_ifaddr *ia;
1178 			int s;
1179 
1180 			/* pick up primary address */
1181 			s = pserialize_read_enter();
1182 			ifp = if_byindex(pktinfo.ipi_ifindex);
1183 			if (ifp == NULL) {
1184 				pserialize_read_exit(s);
1185 				error = EADDRNOTAVAIL;
1186 				break;
1187 			}
1188 			ia = in_get_ia_from_ifp(ifp);
1189 			if (ia == NULL) {
1190 				pserialize_read_exit(s);
1191 				error = EADDRNOTAVAIL;
1192 				break;
1193 			}
1194 			inp->inp_prefsrcip = IA_SIN(ia)->sin_addr;
1195 			pserialize_read_exit(s);
1196 			break;
1197 		break;
1198 #undef OPTSET
1199 
1200 		case IP_MULTICAST_IF:
1201 		case IP_MULTICAST_TTL:
1202 		case IP_MULTICAST_LOOP:
1203 		case IP_ADD_MEMBERSHIP:
1204 		case IP_DROP_MEMBERSHIP:
1205 			error = ip_setmoptions(&inp->inp_moptions, sopt);
1206 			break;
1207 
1208 		case IP_PORTRANGE:
1209 			error = sockopt_getint(sopt, &optval);
1210 			if (error)
1211 				break;
1212 
1213 			switch (optval) {
1214 			case IP_PORTRANGE_DEFAULT:
1215 			case IP_PORTRANGE_HIGH:
1216 				inpflags &= ~(INP_LOWPORT);
1217 				break;
1218 
1219 			case IP_PORTRANGE_LOW:
1220 				inpflags |= INP_LOWPORT;
1221 				break;
1222 
1223 			default:
1224 				error = EINVAL;
1225 				break;
1226 			}
1227 			break;
1228 
1229 		case IP_PORTALGO:
1230 			error = sockopt_getint(sopt, &optval);
1231 			if (error)
1232 				break;
1233 
1234 			error = portalgo_algo_index_select(
1235 			    (struct inpcb_hdr *)inp, optval);
1236 			break;
1237 
1238 #if defined(IPSEC)
1239 		case IP_IPSEC_POLICY:
1240 			if (ipsec_enabled) {
1241 				error = ipsec_set_policy(inp,
1242 				    sopt->sopt_data, sopt->sopt_size,
1243 				    curlwp->l_cred);
1244 			} else
1245 				error = ENOPROTOOPT;
1246 			break;
1247 #endif /* IPSEC */
1248 
1249 		default:
1250 			error = ENOPROTOOPT;
1251 			break;
1252 		}
1253 		break;
1254 
1255 	case PRCO_GETOPT:
1256 		switch (sopt->sopt_name) {
1257 		case IP_OPTIONS:
1258 		case IP_RETOPTS: {
1259 			struct mbuf *mopts = inp->inp_options;
1260 
1261 			if (mopts) {
1262 				struct mbuf *m;
1263 
1264 				m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT);
1265 				if (m == NULL) {
1266 					error = ENOBUFS;
1267 					break;
1268 				}
1269 				error = sockopt_setmbuf(sopt, m);
1270 			}
1271 			break;
1272 		}
1273 		case IP_TOS:
1274 		case IP_TTL:
1275 		case IP_MINTTL:
1276 		case IP_RECVOPTS:
1277 		case IP_RECVRETOPTS:
1278 		case IP_RECVDSTADDR:
1279 		case IP_RECVIF:
1280 		case IP_RECVPKTINFO:
1281 		case IP_RECVTTL:
1282 		case IP_ERRORMTU:
1283 			switch (sopt->sopt_name) {
1284 			case IP_TOS:
1285 				optval = ip->ip_tos;
1286 				break;
1287 
1288 			case IP_TTL:
1289 				optval = ip->ip_ttl;
1290 				break;
1291 
1292 			case IP_MINTTL:
1293 				optval = inp->inp_ip_minttl;
1294 				break;
1295 
1296 			case IP_ERRORMTU:
1297 				optval = inp->inp_errormtu;
1298 				break;
1299 
1300 #define	OPTBIT(bit)	(inpflags & bit ? 1 : 0)
1301 
1302 			case IP_RECVOPTS:
1303 				optval = OPTBIT(INP_RECVOPTS);
1304 				break;
1305 
1306 			case IP_RECVPKTINFO:
1307 				optval = OPTBIT(INP_RECVPKTINFO);
1308 				break;
1309 
1310 			case IP_RECVRETOPTS:
1311 				optval = OPTBIT(INP_RECVRETOPTS);
1312 				break;
1313 
1314 			case IP_RECVDSTADDR:
1315 				optval = OPTBIT(INP_RECVDSTADDR);
1316 				break;
1317 
1318 			case IP_RECVIF:
1319 				optval = OPTBIT(INP_RECVIF);
1320 				break;
1321 
1322 			case IP_RECVTTL:
1323 				optval = OPTBIT(INP_RECVTTL);
1324 				break;
1325 			}
1326 			error = sockopt_setint(sopt, optval);
1327 			break;
1328 
1329 		case IP_PKTINFO:
1330 			switch (sopt->sopt_size) {
1331 			case sizeof(int):
1332 				/* Linux compatibility */
1333 				optval = OPTBIT(INP_RECVPKTINFO);
1334 				error = sockopt_setint(sopt, optval);
1335 				break;
1336 			case sizeof(struct in_pktinfo):
1337 				/* Solaris compatibility */
1338 				pktinfo.ipi_ifindex = 0;
1339 				pktinfo.ipi_addr = inp->inp_prefsrcip;
1340 				error = sockopt_set(sopt, &pktinfo,
1341 				    sizeof(pktinfo));
1342 				break;
1343 			default:
1344 				/*
1345 				 * While size is stuck at 0, and, later, if
1346 				 * the caller doesn't use an exactly sized
1347 				 * recipient for the data, default to Linux
1348 				 * compatibility
1349 				 */
1350 				optval = OPTBIT(INP_RECVPKTINFO);
1351 				error = sockopt_setint(sopt, optval);
1352 				break;
1353 			}
1354 			break;
1355 
1356 #if 0	/* defined(IPSEC) */
1357 		case IP_IPSEC_POLICY:
1358 		{
1359 			struct mbuf *m = NULL;
1360 
1361 			/* XXX this will return EINVAL as sopt is empty */
1362 			error = ipsec_get_policy(inp, sopt->sopt_data,
1363 			    sopt->sopt_size, &m);
1364 			if (error == 0)
1365 				error = sockopt_setmbuf(sopt, m);
1366 			break;
1367 		}
1368 #endif /*IPSEC*/
1369 
1370 		case IP_MULTICAST_IF:
1371 		case IP_MULTICAST_TTL:
1372 		case IP_MULTICAST_LOOP:
1373 		case IP_ADD_MEMBERSHIP:
1374 		case IP_DROP_MEMBERSHIP:
1375 			error = ip_getmoptions(inp->inp_moptions, sopt);
1376 			break;
1377 
1378 		case IP_PORTRANGE:
1379 			if (inpflags & INP_LOWPORT)
1380 				optval = IP_PORTRANGE_LOW;
1381 			else
1382 				optval = IP_PORTRANGE_DEFAULT;
1383 			error = sockopt_setint(sopt, optval);
1384 			break;
1385 
1386 		case IP_PORTALGO:
1387 			optval = inp->inp_portalgo;
1388 			error = sockopt_setint(sopt, optval);
1389 			break;
1390 
1391 		default:
1392 			error = ENOPROTOOPT;
1393 			break;
1394 		}
1395 		break;
1396 	}
1397 
1398 	if (!error) {
1399 		inp->inp_flags = inpflags;
1400 	}
1401 	return error;
1402 }
1403 
1404 static int
1405 ip_pktinfo_prepare(const struct in_pktinfo *pktinfo, struct ip_pktopts *pktopts,
1406     int *flags, kauth_cred_t cred)
1407 {
1408 	struct ip_moptions *imo;
1409 	int error = 0;
1410 	bool addrset = false;
1411 
1412 	if (!in_nullhost(pktinfo->ipi_addr)) {
1413 		pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr;
1414 		/* EADDRNOTAVAIL? */
1415 		error = in_pcbbindableaddr(&pktopts->ippo_laddr, cred);
1416 		if (error != 0)
1417 			return error;
1418 		addrset = true;
1419 	}
1420 
1421 	if (pktinfo->ipi_ifindex != 0) {
1422 		if (!addrset) {
1423 			struct ifnet *ifp;
1424 			struct in_ifaddr *ia;
1425 			int s;
1426 
1427 			/* pick up primary address */
1428 			s = pserialize_read_enter();
1429 			ifp = if_byindex(pktinfo->ipi_ifindex);
1430 			if (ifp == NULL) {
1431 				pserialize_read_exit(s);
1432 				return EADDRNOTAVAIL;
1433 			}
1434 			ia = in_get_ia_from_ifp(ifp);
1435 			if (ia == NULL) {
1436 				pserialize_read_exit(s);
1437 				return EADDRNOTAVAIL;
1438 			}
1439 			pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr;
1440 			pserialize_read_exit(s);
1441 		}
1442 
1443 		/*
1444 		 * If specified ipi_ifindex,
1445 		 * use copied or locally initialized ip_moptions.
1446 		 * Original ip_moptions must not be modified.
1447 		 */
1448 		imo = &pktopts->ippo_imobuf;	/* local buf in pktopts */
1449 		if (pktopts->ippo_imo != NULL) {
1450 			memcpy(imo, pktopts->ippo_imo, sizeof(*imo));
1451 		} else {
1452 			memset(imo, 0, sizeof(*imo));
1453 			imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1454 			imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1455 		}
1456 		imo->imo_multicast_if_index = pktinfo->ipi_ifindex;
1457 		pktopts->ippo_imo = imo;
1458 		*flags |= IP_ROUTETOIFINDEX;
1459 	}
1460 	return error;
1461 }
1462 
1463 /*
1464  * Set up IP outgoing packet options. Even if control is NULL,
1465  * pktopts->ippo_laddr and pktopts->ippo_imo are set and used.
1466  */
1467 int
1468 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags,
1469     struct inpcb *inp, kauth_cred_t cred)
1470 {
1471 	struct cmsghdr *cm;
1472 	struct in_pktinfo pktinfo;
1473 	int error;
1474 
1475 	pktopts->ippo_imo = inp->inp_moptions;
1476 
1477 	struct in_addr *ia = in_nullhost(inp->inp_prefsrcip) ? &inp->inp_laddr :
1478 	    &inp->inp_prefsrcip;
1479 	sockaddr_in_init(&pktopts->ippo_laddr, ia, 0);
1480 
1481 	if (control == NULL)
1482 		return 0;
1483 
1484 	/*
1485 	 * XXX: Currently, we assume all the optional information is
1486 	 * stored in a single mbuf.
1487 	 */
1488 	if (control->m_next)
1489 		return EINVAL;
1490 
1491 	for (; control->m_len > 0;
1492 	    control->m_data += CMSG_ALIGN(cm->cmsg_len),
1493 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1494 		cm = mtod(control, struct cmsghdr *);
1495 		if ((control->m_len < sizeof(*cm)) ||
1496 		    (cm->cmsg_len == 0) ||
1497 		    (cm->cmsg_len > control->m_len)) {
1498 			return EINVAL;
1499 		}
1500 		if (cm->cmsg_level != IPPROTO_IP)
1501 			continue;
1502 
1503 		switch (cm->cmsg_type) {
1504 		case IP_PKTINFO:
1505 			if (cm->cmsg_len != CMSG_LEN(sizeof(pktinfo)))
1506 				return EINVAL;
1507 			memcpy(&pktinfo, CMSG_DATA(cm), sizeof(pktinfo));
1508 			error = ip_pktinfo_prepare(&pktinfo, pktopts, flags,
1509 			    cred);
1510 			if (error)
1511 				return error;
1512 			break;
1513 		case IP_SENDSRCADDR: /* FreeBSD compatibility */
1514 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr)))
1515 				return EINVAL;
1516 			pktinfo.ipi_ifindex = 0;
1517 			pktinfo.ipi_addr =
1518 			    ((struct in_pktinfo *)CMSG_DATA(cm))->ipi_addr;
1519 			error = ip_pktinfo_prepare(&pktinfo, pktopts, flags,
1520 			    cred);
1521 			if (error)
1522 				return error;
1523 			break;
1524 		default:
1525 			return ENOPROTOOPT;
1526 		}
1527 	}
1528 	return 0;
1529 }
1530 
1531 /*
1532  * Set up IP options in pcb for insertion in output packets.
1533  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1534  * with destination address if source routed.
1535  */
1536 static int
1537 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt)
1538 {
1539 	struct mbuf *m;
1540 	const u_char *cp;
1541 	u_char *dp;
1542 	int cnt;
1543 
1544 	KASSERT(inp_locked(inp));
1545 
1546 	/* Turn off any old options. */
1547 	if (inp->inp_options) {
1548 		m_free(inp->inp_options);
1549 	}
1550 	inp->inp_options = NULL;
1551 	if ((cnt = sopt->sopt_size) == 0) {
1552 		/* Only turning off any previous options. */
1553 		return 0;
1554 	}
1555 	cp = sopt->sopt_data;
1556 
1557 	if (cnt % 4) {
1558 		/* Must be 4-byte aligned, because there's no padding. */
1559 		return EINVAL;
1560 	}
1561 
1562 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1563 	if (m == NULL)
1564 		return ENOBUFS;
1565 
1566 	dp = mtod(m, u_char *);
1567 	memset(dp, 0, sizeof(struct in_addr));
1568 	dp += sizeof(struct in_addr);
1569 	m->m_len = sizeof(struct in_addr);
1570 
1571 	/*
1572 	 * IP option list according to RFC791. Each option is of the form
1573 	 *
1574 	 *	[optval] [olen] [(olen - 2) data bytes]
1575 	 *
1576 	 * We validate the list and copy options to an mbuf for prepending
1577 	 * to data packets. The IP first-hop destination address will be
1578 	 * stored before actual options and is zero if unset.
1579 	 */
1580 	while (cnt > 0) {
1581 		uint8_t optval, olen, offset;
1582 
1583 		optval = cp[IPOPT_OPTVAL];
1584 
1585 		if (optval == IPOPT_EOL || optval == IPOPT_NOP) {
1586 			olen = 1;
1587 		} else {
1588 			if (cnt < IPOPT_OLEN + 1)
1589 				goto bad;
1590 
1591 			olen = cp[IPOPT_OLEN];
1592 			if (olen < IPOPT_OLEN + 1 || olen > cnt)
1593 				goto bad;
1594 		}
1595 
1596 		if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) {
1597 			/*
1598 			 * user process specifies route as:
1599 			 *	->A->B->C->D
1600 			 * D must be our final destination (but we can't
1601 			 * check that since we may not have connected yet).
1602 			 * A is first hop destination, which doesn't appear in
1603 			 * actual IP option, but is stored before the options.
1604 			 */
1605 			if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr))
1606 				goto bad;
1607 
1608 			offset = cp[IPOPT_OFFSET];
1609 			memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1,
1610 			    sizeof(struct in_addr));
1611 
1612 			cp += sizeof(struct in_addr);
1613 			cnt -= sizeof(struct in_addr);
1614 			olen -= sizeof(struct in_addr);
1615 
1616 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1617 				goto bad;
1618 
1619 			memcpy(dp, cp, olen);
1620 			dp[IPOPT_OPTVAL] = optval;
1621 			dp[IPOPT_OLEN] = olen;
1622 			dp[IPOPT_OFFSET] = offset;
1623 			break;
1624 		} else {
1625 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1626 				goto bad;
1627 
1628 			memcpy(dp, cp, olen);
1629 			break;
1630 		}
1631 
1632 		dp += olen;
1633 		m->m_len += olen;
1634 
1635 		if (optval == IPOPT_EOL)
1636 			break;
1637 
1638 		cp += olen;
1639 		cnt -= olen;
1640 	}
1641 
1642 	inp->inp_options = m;
1643 	return 0;
1644 
1645 bad:
1646 	(void)m_free(m);
1647 	return EINVAL;
1648 }
1649 
1650 /*
1651  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1652  * Must be called in a pserialize critical section.
1653  */
1654 static struct ifnet *
1655 ip_multicast_if(struct in_addr *a, int *ifindexp)
1656 {
1657 	int ifindex;
1658 	struct ifnet *ifp = NULL;
1659 	struct in_ifaddr *ia;
1660 
1661 	if (ifindexp)
1662 		*ifindexp = 0;
1663 	if (ntohl(a->s_addr) >> 24 == 0) {
1664 		ifindex = ntohl(a->s_addr) & 0xffffff;
1665 		ifp = if_byindex(ifindex);
1666 		if (!ifp)
1667 			return NULL;
1668 		if (ifindexp)
1669 			*ifindexp = ifindex;
1670 	} else {
1671 		IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) {
1672 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1673 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1674 				ifp = ia->ia_ifp;
1675 				if (if_is_deactivated(ifp))
1676 					ifp = NULL;
1677 				break;
1678 			}
1679 		}
1680 	}
1681 	return ifp;
1682 }
1683 
1684 static int
1685 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval)
1686 {
1687 	u_int tval;
1688 	u_char cval;
1689 	int error;
1690 
1691 	if (sopt == NULL)
1692 		return EINVAL;
1693 
1694 	switch (sopt->sopt_size) {
1695 	case sizeof(u_char):
1696 		error = sockopt_get(sopt, &cval, sizeof(u_char));
1697 		tval = cval;
1698 		break;
1699 
1700 	case sizeof(u_int):
1701 		error = sockopt_get(sopt, &tval, sizeof(u_int));
1702 		break;
1703 
1704 	default:
1705 		error = EINVAL;
1706 	}
1707 
1708 	if (error)
1709 		return error;
1710 
1711 	if (tval > maxval)
1712 		return EINVAL;
1713 
1714 	*val = tval;
1715 	return 0;
1716 }
1717 
1718 static int
1719 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
1720     struct psref *psref, struct in_addr *ia, bool add)
1721 {
1722 	int error;
1723 	struct ip_mreq mreq;
1724 
1725 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
1726 	if (error)
1727 		return error;
1728 
1729 	if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr))
1730 		return EINVAL;
1731 
1732 	memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia));
1733 
1734 	if (in_nullhost(mreq.imr_interface)) {
1735 		union {
1736 			struct sockaddr		dst;
1737 			struct sockaddr_in	dst4;
1738 		} u;
1739 		struct route ro;
1740 
1741 		if (!add) {
1742 			*ifp = NULL;
1743 			return 0;
1744 		}
1745 		/*
1746 		 * If no interface address was provided, use the interface of
1747 		 * the route to the given multicast address.
1748 		 */
1749 		struct rtentry *rt;
1750 		memset(&ro, 0, sizeof(ro));
1751 
1752 		sockaddr_in_init(&u.dst4, ia, 0);
1753 		error = rtcache_setdst(&ro, &u.dst);
1754 		if (error != 0)
1755 			return error;
1756 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
1757 		if (*ifp != NULL) {
1758 			if (if_is_deactivated(*ifp))
1759 				*ifp = NULL;
1760 			else
1761 				if_acquire(*ifp, psref);
1762 		}
1763 		rtcache_unref(rt, &ro);
1764 		rtcache_free(&ro);
1765 	} else {
1766 		int s = pserialize_read_enter();
1767 		*ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1768 		if (!add && *ifp == NULL) {
1769 			pserialize_read_exit(s);
1770 			return EADDRNOTAVAIL;
1771 		}
1772 		if (*ifp != NULL) {
1773 			if (if_is_deactivated(*ifp))
1774 				*ifp = NULL;
1775 			else
1776 				if_acquire(*ifp, psref);
1777 		}
1778 		pserialize_read_exit(s);
1779 	}
1780 	return 0;
1781 }
1782 
1783 /*
1784  * Add a multicast group membership.
1785  * Group must be a valid IP multicast address.
1786  */
1787 static int
1788 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1789 {
1790 	struct ifnet *ifp = NULL;	// XXX: gcc [ppc]
1791 	struct in_addr ia;
1792 	int i, error, bound;
1793 	struct psref psref;
1794 
1795 	/* imo is protected by solock or referenced only by the caller */
1796 
1797 	bound = curlwp_bind();
1798 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1799 		error = ip_get_membership(sopt, &ifp, &psref, &ia, true);
1800 	else {
1801 #ifdef INET6
1802 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1803 #else
1804 		error = EINVAL;
1805 #endif
1806 	}
1807 
1808 	if (error)
1809 		goto out;
1810 
1811 	/*
1812 	 * See if we found an interface, and confirm that it
1813 	 * supports multicast.
1814 	 */
1815 	if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1816 		error = EADDRNOTAVAIL;
1817 		goto out;
1818 	}
1819 
1820 	/*
1821 	 * See if the membership already exists or if all the
1822 	 * membership slots are full.
1823 	 */
1824 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1825 		if (imo->imo_membership[i]->inm_ifp == ifp &&
1826 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1827 			break;
1828 	}
1829 	if (i < imo->imo_num_memberships) {
1830 		error = EADDRINUSE;
1831 		goto out;
1832 	}
1833 
1834 	if (i == IP_MAX_MEMBERSHIPS) {
1835 		error = ETOOMANYREFS;
1836 		goto out;
1837 	}
1838 
1839 	/*
1840 	 * Everything looks good; add a new record to the multicast
1841 	 * address list for the given interface.
1842 	 */
1843 	imo->imo_membership[i] = in_addmulti(&ia, ifp);
1844 	if (imo->imo_membership[i] == NULL) {
1845 		error = ENOBUFS;
1846 		goto out;
1847 	}
1848 
1849 	++imo->imo_num_memberships;
1850 	error = 0;
1851 out:
1852 	if_put(ifp, &psref);
1853 	curlwp_bindx(bound);
1854 	return error;
1855 }
1856 
1857 /*
1858  * Drop a multicast group membership.
1859  * Group must be a valid IP multicast address.
1860  */
1861 static int
1862 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1863 {
1864 	struct in_addr ia = { .s_addr = 0 };	// XXX: gcc [ppc]
1865 	struct ifnet *ifp = NULL;		// XXX: gcc [ppc]
1866 	int i, error, bound;
1867 	struct psref psref;
1868 
1869 	/* imo is protected by solock or referenced only by the caller */
1870 
1871 	bound = curlwp_bind();
1872 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1873 		error = ip_get_membership(sopt, &ifp, &psref, &ia, false);
1874 	else {
1875 #ifdef INET6
1876 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1877 #else
1878 		error = EINVAL;
1879 #endif
1880 	}
1881 
1882 	if (error)
1883 		goto out;
1884 
1885 	/*
1886 	 * Find the membership in the membership array.
1887 	 */
1888 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1889 		if ((ifp == NULL ||
1890 		     imo->imo_membership[i]->inm_ifp == ifp) &&
1891 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1892 			break;
1893 	}
1894 	if (i == imo->imo_num_memberships) {
1895 		error = EADDRNOTAVAIL;
1896 		goto out;
1897 	}
1898 
1899 	/*
1900 	 * Give up the multicast address record to which the
1901 	 * membership points.
1902 	 */
1903 	in_delmulti(imo->imo_membership[i]);
1904 
1905 	/*
1906 	 * Remove the gap in the membership array.
1907 	 */
1908 	for (++i; i < imo->imo_num_memberships; ++i)
1909 		imo->imo_membership[i-1] = imo->imo_membership[i];
1910 	--imo->imo_num_memberships;
1911 	error = 0;
1912 out:
1913 	if_put(ifp, &psref);
1914 	curlwp_bindx(bound);
1915 	return error;
1916 }
1917 
1918 /*
1919  * Set the IP multicast options in response to user setsockopt().
1920  */
1921 int
1922 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt)
1923 {
1924 	struct ip_moptions *imo = *pimo;
1925 	struct in_addr addr;
1926 	struct ifnet *ifp;
1927 	int ifindex, error = 0;
1928 
1929 	/* The passed imo isn't NULL, it should be protected by solock */
1930 
1931 	if (!imo) {
1932 		/*
1933 		 * No multicast option buffer attached to the pcb;
1934 		 * allocate one and initialize to default values.
1935 		 */
1936 		imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP);
1937 		if (imo == NULL)
1938 			return ENOBUFS;
1939 
1940 		imo->imo_multicast_if_index = 0;
1941 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1942 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1943 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1944 		imo->imo_num_memberships = 0;
1945 		*pimo = imo;
1946 	}
1947 
1948 	switch (sopt->sopt_name) {
1949 	case IP_MULTICAST_IF: {
1950 		int s;
1951 		/*
1952 		 * Select the interface for outgoing multicast packets.
1953 		 */
1954 		error = sockopt_get(sopt, &addr, sizeof(addr));
1955 		if (error)
1956 			break;
1957 
1958 		/*
1959 		 * INADDR_ANY is used to remove a previous selection.
1960 		 * When no interface is selected, a default one is
1961 		 * chosen every time a multicast packet is sent.
1962 		 */
1963 		if (in_nullhost(addr)) {
1964 			imo->imo_multicast_if_index = 0;
1965 			break;
1966 		}
1967 		/*
1968 		 * The selected interface is identified by its local
1969 		 * IP address.  Find the interface and confirm that
1970 		 * it supports multicasting.
1971 		 */
1972 		s = pserialize_read_enter();
1973 		ifp = ip_multicast_if(&addr, &ifindex);
1974 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1975 			pserialize_read_exit(s);
1976 			error = EADDRNOTAVAIL;
1977 			break;
1978 		}
1979 		imo->imo_multicast_if_index = ifp->if_index;
1980 		pserialize_read_exit(s);
1981 		if (ifindex)
1982 			imo->imo_multicast_addr = addr;
1983 		else
1984 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1985 		break;
1986 	    }
1987 
1988 	case IP_MULTICAST_TTL:
1989 		/*
1990 		 * Set the IP time-to-live for outgoing multicast packets.
1991 		 */
1992 		error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL);
1993 		break;
1994 
1995 	case IP_MULTICAST_LOOP:
1996 		/*
1997 		 * Set the loopback flag for outgoing multicast packets.
1998 		 * Must be zero or one.
1999 		 */
2000 		error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1);
2001 		break;
2002 
2003 	case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */
2004 		error = ip_add_membership(imo, sopt);
2005 		break;
2006 
2007 	case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */
2008 		error = ip_drop_membership(imo, sopt);
2009 		break;
2010 
2011 	default:
2012 		error = EOPNOTSUPP;
2013 		break;
2014 	}
2015 
2016 	/*
2017 	 * If all options have default values, no need to keep the mbuf.
2018 	 */
2019 	if (imo->imo_multicast_if_index == 0 &&
2020 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2021 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2022 	    imo->imo_num_memberships == 0) {
2023 		kmem_intr_free(imo, sizeof(*imo));
2024 		*pimo = NULL;
2025 	}
2026 
2027 	return error;
2028 }
2029 
2030 /*
2031  * Return the IP multicast options in response to user getsockopt().
2032  */
2033 int
2034 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt)
2035 {
2036 	struct in_addr addr;
2037 	uint8_t optval;
2038 	int error = 0;
2039 
2040 	/* imo is protected by solock or refereced only by the caller */
2041 
2042 	switch (sopt->sopt_name) {
2043 	case IP_MULTICAST_IF:
2044 		if (imo == NULL || imo->imo_multicast_if_index == 0)
2045 			addr = zeroin_addr;
2046 		else if (imo->imo_multicast_addr.s_addr) {
2047 			/* return the value user has set */
2048 			addr = imo->imo_multicast_addr;
2049 		} else {
2050 			struct ifnet *ifp;
2051 			struct in_ifaddr *ia = NULL;
2052 			int s = pserialize_read_enter();
2053 
2054 			ifp = if_byindex(imo->imo_multicast_if_index);
2055 			if (ifp != NULL) {
2056 				ia = in_get_ia_from_ifp(ifp);
2057 			}
2058 			addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
2059 			pserialize_read_exit(s);
2060 		}
2061 		error = sockopt_set(sopt, &addr, sizeof(addr));
2062 		break;
2063 
2064 	case IP_MULTICAST_TTL:
2065 		optval = imo ? imo->imo_multicast_ttl
2066 		    : IP_DEFAULT_MULTICAST_TTL;
2067 
2068 		error = sockopt_set(sopt, &optval, sizeof(optval));
2069 		break;
2070 
2071 	case IP_MULTICAST_LOOP:
2072 		optval = imo ? imo->imo_multicast_loop
2073 		    : IP_DEFAULT_MULTICAST_LOOP;
2074 
2075 		error = sockopt_set(sopt, &optval, sizeof(optval));
2076 		break;
2077 
2078 	default:
2079 		error = EOPNOTSUPP;
2080 	}
2081 
2082 	return error;
2083 }
2084 
2085 /*
2086  * Discard the IP multicast options.
2087  */
2088 void
2089 ip_freemoptions(struct ip_moptions *imo)
2090 {
2091 	int i;
2092 
2093 	/* The owner of imo (inp) should be protected by solock */
2094 
2095 	if (imo != NULL) {
2096 		for (i = 0; i < imo->imo_num_memberships; ++i) {
2097 			struct in_multi *inm = imo->imo_membership[i];
2098 			in_delmulti(inm);
2099 			/* ifp should not leave thanks to solock */
2100 		}
2101 
2102 		kmem_intr_free(imo, sizeof(*imo));
2103 	}
2104 }
2105 
2106 /*
2107  * Routine called from ip_output() to loop back a copy of an IP multicast
2108  * packet to the input queue of a specified interface.  Note that this
2109  * calls the output routine of the loopback "driver", but with an interface
2110  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2111  */
2112 static void
2113 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst)
2114 {
2115 	struct ip *ip;
2116 	struct mbuf *copym;
2117 
2118 	copym = m_copypacket(m, M_DONTWAIT);
2119 	if (copym != NULL &&
2120 	    (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
2121 		copym = m_pullup(copym, sizeof(struct ip));
2122 	if (copym == NULL)
2123 		return;
2124 	/*
2125 	 * We don't bother to fragment if the IP length is greater
2126 	 * than the interface's MTU.  Can this possibly matter?
2127 	 */
2128 	ip = mtod(copym, struct ip *);
2129 
2130 	if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
2131 		in_undefer_cksum_tcpudp(copym);
2132 		copym->m_pkthdr.csum_flags &=
2133 		    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
2134 	}
2135 
2136 	ip->ip_sum = 0;
2137 	ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2138 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
2139 	(void)looutput(ifp, copym, sintocsa(dst), NULL);
2140 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
2141 }
2142 
2143 /*
2144  * Ensure sending address is valid.
2145  * Returns 0 on success, -1 if an error should be sent back or 1
2146  * if the packet could be dropped without error (protocol dependent).
2147  */
2148 static int
2149 ip_ifaddrvalid(const struct in_ifaddr *ia)
2150 {
2151 
2152 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
2153 		return 0;
2154 
2155 	if (ia->ia4_flags & IN_IFF_DUPLICATED)
2156 		return -1;
2157 	else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED))
2158 		return 1;
2159 
2160 	return 0;
2161 }
2162