xref: /netbsd-src/sys/netinet/ip_output.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: ip_output.c,v 1.307 2018/07/11 05:25:45 maxv 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 1.307 2018/07/11 05:25:45 maxv 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 #endif
620 
621 	/*
622 	 * Run through list of hooks for output packets.
623 	 */
624 	error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT);
625 	if (error)
626 		goto done;
627 	if (m == NULL)
628 		goto done;
629 
630 	ip = mtod(m, struct ip *);
631 	hlen = ip->ip_hl << 2;
632 
633 	m->m_pkthdr.csum_data |= hlen << 16;
634 
635 	/*
636 	 * search for the source address structure to
637 	 * maintain output statistics, and verify address
638 	 * validity
639 	 */
640 	KASSERT(ia == NULL);
641 	sockaddr_in_init(&usrc.sin, &ip->ip_src, 0);
642 	ifa = ifaof_ifpforaddr_psref(&usrc.sa, ifp, &psref_ia);
643 	if (ifa != NULL)
644 		ia = ifatoia(ifa);
645 
646 	/*
647 	 * Ensure we only send from a valid address.
648 	 * A NULL address is valid because the packet could be
649 	 * generated from a packet filter.
650 	 */
651 	if (ia != NULL && (flags & IP_FORWARDING) == 0 &&
652 	    (error = ip_ifaddrvalid(ia)) != 0)
653 	{
654 		ARPLOG(LOG_ERR,
655 		    "refusing to send from invalid address %s (pid %d)\n",
656 		    ARPLOGADDR(&ip->ip_src), curproc->p_pid);
657 		IP_STATINC(IP_STAT_ODROPPED);
658 		if (error == 1)
659 			/*
660 			 * Address exists, but is tentative or detached.
661 			 * We can't send from it because it's invalid,
662 			 * so we drop the packet.
663 			 */
664 			error = 0;
665 		else
666 			error = EADDRNOTAVAIL;
667 		goto bad;
668 	}
669 
670 	/* Maybe skip checksums on loopback interfaces. */
671 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
672 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
673 	}
674 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
675 
676 	/*
677 	 * If small enough for mtu of path, or if using TCP segmentation
678 	 * offload, can just send directly.
679 	 */
680 	if (ntohs(ip->ip_len) <= mtu ||
681 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
682 		const struct sockaddr *sa;
683 
684 #if IFA_STATS
685 		if (ia)
686 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
687 #endif
688 		/*
689 		 * Always initialize the sum to 0!  Some HW assisted
690 		 * checksumming requires this.
691 		 */
692 		ip->ip_sum = 0;
693 
694 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
695 			/*
696 			 * Perform any checksums that the hardware can't do
697 			 * for us.
698 			 *
699 			 * XXX Does any hardware require the {th,uh}_sum
700 			 * XXX fields to be 0?
701 			 */
702 			if (sw_csum & M_CSUM_IPv4) {
703 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
704 				ip->ip_sum = in_cksum(m, hlen);
705 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
706 			}
707 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
708 				if (IN_NEED_CHECKSUM(ifp,
709 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
710 					in_undefer_cksum_tcpudp(m);
711 				}
712 				m->m_pkthdr.csum_flags &=
713 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
714 			}
715 		}
716 
717 		sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst);
718 		if (__predict_true(
719 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
720 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
721 			error = ip_if_output(ifp, m, sa, rt);
722 		} else {
723 			error = ip_tso_output(ifp, m, sa, rt);
724 		}
725 		goto done;
726 	}
727 
728 	/*
729 	 * We can't use HW checksumming if we're about to fragment the packet.
730 	 *
731 	 * XXX Some hardware can do this.
732 	 */
733 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
734 		if (IN_NEED_CHECKSUM(ifp,
735 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
736 			in_undefer_cksum_tcpudp(m);
737 		}
738 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
739 	}
740 
741 	/*
742 	 * Too large for interface; fragment if possible.
743 	 * Must be able to put at least 8 bytes per fragment.
744 	 */
745 	if (ntohs(ip->ip_off) & IP_DF) {
746 		if (flags & IP_RETURNMTU) {
747 			KASSERT(inp != NULL);
748 			inp->inp_errormtu = mtu;
749 		}
750 		error = EMSGSIZE;
751 		IP_STATINC(IP_STAT_CANTFRAG);
752 		goto bad;
753 	}
754 
755 	error = ip_fragment(m, ifp, mtu);
756 	if (error) {
757 		m = NULL;
758 		goto bad;
759 	}
760 
761 	for (; m; m = m0) {
762 		m0 = m->m_nextpkt;
763 		m->m_nextpkt = NULL;
764 		if (error) {
765 			m_freem(m);
766 			continue;
767 		}
768 #if IFA_STATS
769 		if (ia)
770 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
771 #endif
772 		/*
773 		 * If we get there, the packet has not been handled by
774 		 * IPsec whereas it should have. Now that it has been
775 		 * fragmented, re-inject it in ip_output so that IPsec
776 		 * processing can occur.
777 		 */
778 		if (natt_frag) {
779 			error = ip_output(m, opt, ro,
780 			    flags | IP_RAWOUTPUT | IP_NOIPNEWID,
781 			    imo, inp);
782 		} else {
783 			KASSERT((m->m_pkthdr.csum_flags &
784 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
785 			error = ip_if_output(ifp, m,
786 			    (m->m_flags & M_MCAST) ?
787 			    sintocsa(rdst) : sintocsa(dst), rt);
788 		}
789 	}
790 	if (error == 0) {
791 		IP_STATINC(IP_STAT_FRAGMENTED);
792 	}
793 
794 done:
795 	ia4_release(ia, &psref_ia);
796 	rtcache_unref(rt, ro);
797 	if (ro == &iproute) {
798 		rtcache_free(&iproute);
799 	}
800 	if (mifp != NULL) {
801 		if_put(mifp, &psref);
802 	}
803 	if (bind_need_restore)
804 		curlwp_bindx(bound);
805 	return error;
806 
807 bad:
808 	m_freem(m);
809 	goto done;
810 }
811 
812 int
813 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
814 {
815 	struct ip *ip, *mhip;
816 	struct mbuf *m0;
817 	int len, hlen, off;
818 	int mhlen, firstlen;
819 	struct mbuf **mnext;
820 	int sw_csum = m->m_pkthdr.csum_flags;
821 	int fragments = 0;
822 	int error = 0;
823 	int ipoff, ipflg;
824 
825 	ip = mtod(m, struct ip *);
826 	hlen = ip->ip_hl << 2;
827 
828 	/* Preserve the offset and flags. */
829 	ipoff = ntohs(ip->ip_off) & IP_OFFMASK;
830 	ipflg = ntohs(ip->ip_off) & (IP_RF|IP_DF|IP_MF);
831 
832 	if (ifp != NULL)
833 		sw_csum &= ~ifp->if_csum_flags_tx;
834 
835 	len = (mtu - hlen) &~ 7;
836 	if (len < 8) {
837 		m_freem(m);
838 		return EMSGSIZE;
839 	}
840 
841 	firstlen = len;
842 	mnext = &m->m_nextpkt;
843 
844 	/*
845 	 * Loop through length of segment after first fragment,
846 	 * make new header and copy data of each part and link onto chain.
847 	 */
848 	m0 = m;
849 	mhlen = sizeof(struct ip);
850 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
851 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
852 		if (m == NULL) {
853 			error = ENOBUFS;
854 			IP_STATINC(IP_STAT_ODROPPED);
855 			goto sendorfree;
856 		}
857 		MCLAIM(m, m0->m_owner);
858 
859 		*mnext = m;
860 		mnext = &m->m_nextpkt;
861 
862 		m->m_data += max_linkhdr;
863 		mhip = mtod(m, struct ip *);
864 		*mhip = *ip;
865 
866 		/* we must inherit the flags */
867 		m->m_flags |= m0->m_flags & M_COPYFLAGS;
868 
869 		if (hlen > sizeof(struct ip)) {
870 			mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
871 			mhip->ip_hl = mhlen >> 2;
872 		}
873 		m->m_len = mhlen;
874 
875 		mhip->ip_off = ((off - hlen) >> 3) + ipoff;
876 		mhip->ip_off |= ipflg;
877 		if (off + len >= ntohs(ip->ip_len))
878 			len = ntohs(ip->ip_len) - off;
879 		else
880 			mhip->ip_off |= IP_MF;
881 		HTONS(mhip->ip_off);
882 
883 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
884 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
885 		if (m->m_next == NULL) {
886 			error = ENOBUFS;
887 			IP_STATINC(IP_STAT_ODROPPED);
888 			goto sendorfree;
889 		}
890 
891 		m->m_pkthdr.len = mhlen + len;
892 		m_reset_rcvif(m);
893 
894 		mhip->ip_sum = 0;
895 		KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
896 		if (sw_csum & M_CSUM_IPv4) {
897 			mhip->ip_sum = in_cksum(m, mhlen);
898 		} else {
899 			/*
900 			 * checksum is hw-offloaded or not necessary.
901 			 */
902 			m->m_pkthdr.csum_flags |=
903 			    m0->m_pkthdr.csum_flags & M_CSUM_IPv4;
904 			m->m_pkthdr.csum_data |= mhlen << 16;
905 			KASSERT(!(ifp != NULL &&
906 			    IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
907 			    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
908 		}
909 		IP_STATINC(IP_STAT_OFRAGMENTS);
910 		fragments++;
911 	}
912 
913 	/*
914 	 * Update first fragment by trimming what's been copied out
915 	 * and updating header, then send each fragment (in order).
916 	 */
917 	m = m0;
918 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
919 	m->m_pkthdr.len = hlen + firstlen;
920 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
921 	ip->ip_off |= htons(IP_MF);
922 	ip->ip_sum = 0;
923 	if (sw_csum & M_CSUM_IPv4) {
924 		ip->ip_sum = in_cksum(m, hlen);
925 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
926 	} else {
927 		/*
928 		 * checksum is hw-offloaded or not necessary.
929 		 */
930 		KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
931 		    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
932 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
933 		    sizeof(struct ip));
934 	}
935 
936 sendorfree:
937 	/*
938 	 * If there is no room for all the fragments, don't queue
939 	 * any of them.
940 	 */
941 	if (ifp != NULL) {
942 		IFQ_LOCK(&ifp->if_snd);
943 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
944 		    error == 0) {
945 			error = ENOBUFS;
946 			IP_STATINC(IP_STAT_ODROPPED);
947 			IFQ_INC_DROPS(&ifp->if_snd);
948 		}
949 		IFQ_UNLOCK(&ifp->if_snd);
950 	}
951 	if (error) {
952 		for (m = m0; m; m = m0) {
953 			m0 = m->m_nextpkt;
954 			m->m_nextpkt = NULL;
955 			m_freem(m);
956 		}
957 	}
958 
959 	return error;
960 }
961 
962 /*
963  * Determine the maximum length of the options to be inserted;
964  * we would far rather allocate too much space rather than too little.
965  */
966 u_int
967 ip_optlen(struct inpcb *inp)
968 {
969 	struct mbuf *m = inp->inp_options;
970 
971 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) {
972 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
973 	}
974 	return 0;
975 }
976 
977 /*
978  * Insert IP options into preformed packet.
979  * Adjust IP destination as required for IP source routing,
980  * as indicated by a non-zero in_addr at the start of the options.
981  */
982 static struct mbuf *
983 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
984 {
985 	struct ipoption *p = mtod(opt, struct ipoption *);
986 	struct mbuf *n;
987 	struct ip *ip = mtod(m, struct ip *);
988 	unsigned optlen;
989 
990 	optlen = opt->m_len - sizeof(p->ipopt_dst);
991 	KASSERT(optlen % 4 == 0);
992 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
993 		return m;		/* XXX should fail */
994 	if (!in_nullhost(p->ipopt_dst))
995 		ip->ip_dst = p->ipopt_dst;
996 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
997 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
998 		if (n == NULL)
999 			return m;
1000 		MCLAIM(n, m->m_owner);
1001 		M_MOVE_PKTHDR(n, m);
1002 		m->m_len -= sizeof(struct ip);
1003 		m->m_data += sizeof(struct ip);
1004 		n->m_next = m;
1005 		n->m_len = optlen + sizeof(struct ip);
1006 		n->m_data += max_linkhdr;
1007 		memcpy(mtod(n, void *), ip, sizeof(struct ip));
1008 		m = n;
1009 	} else {
1010 		m->m_data -= optlen;
1011 		m->m_len += optlen;
1012 		memmove(mtod(m, void *), ip, sizeof(struct ip));
1013 	}
1014 	m->m_pkthdr.len += optlen;
1015 	ip = mtod(m, struct ip *);
1016 	memcpy(ip + 1, p->ipopt_list, optlen);
1017 	*phlen = sizeof(struct ip) + optlen;
1018 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1019 	return m;
1020 }
1021 
1022 /*
1023  * Copy options from ipsrc to ipdst, omitting those not copied during
1024  * fragmentation.
1025  */
1026 int
1027 ip_optcopy(struct ip *ipsrc, struct ip *ipdst)
1028 {
1029 	u_char *cp, *dp;
1030 	int opt, optlen, cnt;
1031 
1032 	cp = (u_char *)(ipsrc + 1);
1033 	dp = (u_char *)(ipdst + 1);
1034 	cnt = (ipsrc->ip_hl << 2) - sizeof(struct ip);
1035 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1036 		opt = cp[0];
1037 		if (opt == IPOPT_EOL)
1038 			break;
1039 		if (opt == IPOPT_NOP) {
1040 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1041 			*dp++ = IPOPT_NOP;
1042 			optlen = 1;
1043 			continue;
1044 		}
1045 
1046 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp));
1047 		optlen = cp[IPOPT_OLEN];
1048 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt);
1049 
1050 		/* Invalid lengths should have been caught by ip_dooptions. */
1051 		if (optlen > cnt)
1052 			optlen = cnt;
1053 		if (IPOPT_COPIED(opt)) {
1054 			bcopy((void *)cp, (void *)dp, (unsigned)optlen);
1055 			dp += optlen;
1056 		}
1057 	}
1058 
1059 	for (optlen = dp - (u_char *)(ipdst+1); optlen & 0x3; optlen++) {
1060 		*dp++ = IPOPT_EOL;
1061 	}
1062 
1063 	return optlen;
1064 }
1065 
1066 /*
1067  * IP socket option processing.
1068  */
1069 int
1070 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1071 {
1072 	struct inpcb *inp = sotoinpcb(so);
1073 	struct ip *ip = &inp->inp_ip;
1074 	int inpflags = inp->inp_flags;
1075 	int optval = 0, error = 0;
1076 	struct in_pktinfo pktinfo;
1077 
1078 	KASSERT(solocked(so));
1079 
1080 	if (sopt->sopt_level != IPPROTO_IP) {
1081 		if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER)
1082 			return 0;
1083 		return ENOPROTOOPT;
1084 	}
1085 
1086 	switch (op) {
1087 	case PRCO_SETOPT:
1088 		switch (sopt->sopt_name) {
1089 		case IP_OPTIONS:
1090 #ifdef notyet
1091 		case IP_RETOPTS:
1092 #endif
1093 			error = ip_pcbopts(inp, sopt);
1094 			break;
1095 
1096 		case IP_TOS:
1097 		case IP_TTL:
1098 		case IP_MINTTL:
1099 		case IP_RECVOPTS:
1100 		case IP_RECVRETOPTS:
1101 		case IP_RECVDSTADDR:
1102 		case IP_RECVIF:
1103 		case IP_RECVPKTINFO:
1104 		case IP_RECVTTL:
1105 			error = sockopt_getint(sopt, &optval);
1106 			if (error)
1107 				break;
1108 
1109 			switch (sopt->sopt_name) {
1110 			case IP_TOS:
1111 				ip->ip_tos = optval;
1112 				break;
1113 
1114 			case IP_TTL:
1115 				ip->ip_ttl = optval;
1116 				break;
1117 
1118 			case IP_MINTTL:
1119 				if (optval > 0 && optval <= MAXTTL)
1120 					inp->inp_ip_minttl = optval;
1121 				else
1122 					error = EINVAL;
1123 				break;
1124 #define	OPTSET(bit) \
1125 	if (optval) \
1126 		inpflags |= bit; \
1127 	else \
1128 		inpflags &= ~bit;
1129 
1130 			case IP_RECVOPTS:
1131 				OPTSET(INP_RECVOPTS);
1132 				break;
1133 
1134 			case IP_RECVPKTINFO:
1135 				OPTSET(INP_RECVPKTINFO);
1136 				break;
1137 
1138 			case IP_RECVRETOPTS:
1139 				OPTSET(INP_RECVRETOPTS);
1140 				break;
1141 
1142 			case IP_RECVDSTADDR:
1143 				OPTSET(INP_RECVDSTADDR);
1144 				break;
1145 
1146 			case IP_RECVIF:
1147 				OPTSET(INP_RECVIF);
1148 				break;
1149 
1150 			case IP_RECVTTL:
1151 				OPTSET(INP_RECVTTL);
1152 				break;
1153 			}
1154 			break;
1155 		case IP_PKTINFO:
1156 			error = sockopt_getint(sopt, &optval);
1157 			if (!error) {
1158 				/* Linux compatibility */
1159 				OPTSET(INP_RECVPKTINFO);
1160 				break;
1161 			}
1162 			error = sockopt_get(sopt, &pktinfo, sizeof(pktinfo));
1163 			if (error)
1164 				break;
1165 
1166 			if (pktinfo.ipi_ifindex == 0) {
1167 				inp->inp_prefsrcip = pktinfo.ipi_addr;
1168 				break;
1169 			}
1170 
1171 			/* Solaris compatibility */
1172 			struct ifnet *ifp;
1173 			struct in_ifaddr *ia;
1174 			int s;
1175 
1176 			/* pick up primary address */
1177 			s = pserialize_read_enter();
1178 			ifp = if_byindex(pktinfo.ipi_ifindex);
1179 			if (ifp == NULL) {
1180 				pserialize_read_exit(s);
1181 				error = EADDRNOTAVAIL;
1182 				break;
1183 			}
1184 			ia = in_get_ia_from_ifp(ifp);
1185 			if (ia == NULL) {
1186 				pserialize_read_exit(s);
1187 				error = EADDRNOTAVAIL;
1188 				break;
1189 			}
1190 			inp->inp_prefsrcip = IA_SIN(ia)->sin_addr;
1191 			pserialize_read_exit(s);
1192 			break;
1193 		break;
1194 #undef OPTSET
1195 
1196 		case IP_MULTICAST_IF:
1197 		case IP_MULTICAST_TTL:
1198 		case IP_MULTICAST_LOOP:
1199 		case IP_ADD_MEMBERSHIP:
1200 		case IP_DROP_MEMBERSHIP:
1201 			error = ip_setmoptions(&inp->inp_moptions, sopt);
1202 			break;
1203 
1204 		case IP_PORTRANGE:
1205 			error = sockopt_getint(sopt, &optval);
1206 			if (error)
1207 				break;
1208 
1209 			switch (optval) {
1210 			case IP_PORTRANGE_DEFAULT:
1211 			case IP_PORTRANGE_HIGH:
1212 				inpflags &= ~(INP_LOWPORT);
1213 				break;
1214 
1215 			case IP_PORTRANGE_LOW:
1216 				inpflags |= INP_LOWPORT;
1217 				break;
1218 
1219 			default:
1220 				error = EINVAL;
1221 				break;
1222 			}
1223 			break;
1224 
1225 		case IP_PORTALGO:
1226 			error = sockopt_getint(sopt, &optval);
1227 			if (error)
1228 				break;
1229 
1230 			error = portalgo_algo_index_select(
1231 			    (struct inpcb_hdr *)inp, optval);
1232 			break;
1233 
1234 #if defined(IPSEC)
1235 		case IP_IPSEC_POLICY:
1236 			if (ipsec_enabled) {
1237 				error = ipsec_set_policy(inp,
1238 				    sopt->sopt_data, sopt->sopt_size,
1239 				    curlwp->l_cred);
1240 				break;
1241 			}
1242 			/*FALLTHROUGH*/
1243 #endif /* IPSEC */
1244 
1245 		default:
1246 			error = ENOPROTOOPT;
1247 			break;
1248 		}
1249 		break;
1250 
1251 	case PRCO_GETOPT:
1252 		switch (sopt->sopt_name) {
1253 		case IP_OPTIONS:
1254 		case IP_RETOPTS: {
1255 			struct mbuf *mopts = inp->inp_options;
1256 
1257 			if (mopts) {
1258 				struct mbuf *m;
1259 
1260 				m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT);
1261 				if (m == NULL) {
1262 					error = ENOBUFS;
1263 					break;
1264 				}
1265 				error = sockopt_setmbuf(sopt, m);
1266 			}
1267 			break;
1268 		}
1269 		case IP_TOS:
1270 		case IP_TTL:
1271 		case IP_MINTTL:
1272 		case IP_RECVOPTS:
1273 		case IP_RECVRETOPTS:
1274 		case IP_RECVDSTADDR:
1275 		case IP_RECVIF:
1276 		case IP_RECVPKTINFO:
1277 		case IP_RECVTTL:
1278 		case IP_ERRORMTU:
1279 			switch (sopt->sopt_name) {
1280 			case IP_TOS:
1281 				optval = ip->ip_tos;
1282 				break;
1283 
1284 			case IP_TTL:
1285 				optval = ip->ip_ttl;
1286 				break;
1287 
1288 			case IP_MINTTL:
1289 				optval = inp->inp_ip_minttl;
1290 				break;
1291 
1292 			case IP_ERRORMTU:
1293 				optval = inp->inp_errormtu;
1294 				break;
1295 
1296 #define	OPTBIT(bit)	(inpflags & bit ? 1 : 0)
1297 
1298 			case IP_RECVOPTS:
1299 				optval = OPTBIT(INP_RECVOPTS);
1300 				break;
1301 
1302 			case IP_RECVPKTINFO:
1303 				optval = OPTBIT(INP_RECVPKTINFO);
1304 				break;
1305 
1306 			case IP_RECVRETOPTS:
1307 				optval = OPTBIT(INP_RECVRETOPTS);
1308 				break;
1309 
1310 			case IP_RECVDSTADDR:
1311 				optval = OPTBIT(INP_RECVDSTADDR);
1312 				break;
1313 
1314 			case IP_RECVIF:
1315 				optval = OPTBIT(INP_RECVIF);
1316 				break;
1317 
1318 			case IP_RECVTTL:
1319 				optval = OPTBIT(INP_RECVTTL);
1320 				break;
1321 			}
1322 			error = sockopt_setint(sopt, optval);
1323 			break;
1324 
1325 		case IP_PKTINFO:
1326 			switch (sopt->sopt_size) {
1327 			case sizeof(int):
1328 				/* Linux compatibility */
1329 				optval = OPTBIT(INP_RECVPKTINFO);
1330 				error = sockopt_setint(sopt, optval);
1331 				break;
1332 			case sizeof(struct in_pktinfo):
1333 				/* Solaris compatibility */
1334 				pktinfo.ipi_ifindex = 0;
1335 				pktinfo.ipi_addr = inp->inp_prefsrcip;
1336 				error = sockopt_set(sopt, &pktinfo,
1337 				    sizeof(pktinfo));
1338 				break;
1339 			default:
1340 				/*
1341 				 * While size is stuck at 0, and, later, if
1342 				 * the caller doesn't use an exactly sized
1343 				 * recipient for the data, default to Linux
1344 				 * compatibility
1345 				 */
1346 				optval = OPTBIT(INP_RECVPKTINFO);
1347 				error = sockopt_setint(sopt, optval);
1348 				break;
1349 			}
1350 			break;
1351 
1352 #if 0	/* defined(IPSEC) */
1353 		case IP_IPSEC_POLICY:
1354 		{
1355 			struct mbuf *m = NULL;
1356 
1357 			/* XXX this will return EINVAL as sopt is empty */
1358 			error = ipsec_get_policy(inp, sopt->sopt_data,
1359 			    sopt->sopt_size, &m);
1360 			if (error == 0)
1361 				error = sockopt_setmbuf(sopt, m);
1362 			break;
1363 		}
1364 #endif /*IPSEC*/
1365 
1366 		case IP_MULTICAST_IF:
1367 		case IP_MULTICAST_TTL:
1368 		case IP_MULTICAST_LOOP:
1369 		case IP_ADD_MEMBERSHIP:
1370 		case IP_DROP_MEMBERSHIP:
1371 			error = ip_getmoptions(inp->inp_moptions, sopt);
1372 			break;
1373 
1374 		case IP_PORTRANGE:
1375 			if (inpflags & INP_LOWPORT)
1376 				optval = IP_PORTRANGE_LOW;
1377 			else
1378 				optval = IP_PORTRANGE_DEFAULT;
1379 			error = sockopt_setint(sopt, optval);
1380 			break;
1381 
1382 		case IP_PORTALGO:
1383 			optval = inp->inp_portalgo;
1384 			error = sockopt_setint(sopt, optval);
1385 			break;
1386 
1387 		default:
1388 			error = ENOPROTOOPT;
1389 			break;
1390 		}
1391 		break;
1392 	}
1393 
1394 	if (!error) {
1395 		inp->inp_flags = inpflags;
1396 	}
1397 	return error;
1398 }
1399 
1400 static int
1401 ip_pktinfo_prepare(const struct in_pktinfo *pktinfo, struct ip_pktopts *pktopts,
1402     int *flags, kauth_cred_t cred)
1403 {
1404 	struct ip_moptions *imo;
1405 	int error = 0;
1406 	bool addrset = false;
1407 
1408 	if (!in_nullhost(pktinfo->ipi_addr)) {
1409 		pktopts->ippo_laddr.sin_addr = pktinfo->ipi_addr;
1410 		/* EADDRNOTAVAIL? */
1411 		error = in_pcbbindableaddr(&pktopts->ippo_laddr, cred);
1412 		if (error != 0)
1413 			return error;
1414 		addrset = true;
1415 	}
1416 
1417 	if (pktinfo->ipi_ifindex != 0) {
1418 		if (!addrset) {
1419 			struct ifnet *ifp;
1420 			struct in_ifaddr *ia;
1421 			int s;
1422 
1423 			/* pick up primary address */
1424 			s = pserialize_read_enter();
1425 			ifp = if_byindex(pktinfo->ipi_ifindex);
1426 			if (ifp == NULL) {
1427 				pserialize_read_exit(s);
1428 				return EADDRNOTAVAIL;
1429 			}
1430 			ia = in_get_ia_from_ifp(ifp);
1431 			if (ia == NULL) {
1432 				pserialize_read_exit(s);
1433 				return EADDRNOTAVAIL;
1434 			}
1435 			pktopts->ippo_laddr.sin_addr = IA_SIN(ia)->sin_addr;
1436 			pserialize_read_exit(s);
1437 		}
1438 
1439 		/*
1440 		 * If specified ipi_ifindex,
1441 		 * use copied or locally initialized ip_moptions.
1442 		 * Original ip_moptions must not be modified.
1443 		 */
1444 		imo = &pktopts->ippo_imobuf;	/* local buf in pktopts */
1445 		if (pktopts->ippo_imo != NULL) {
1446 			memcpy(imo, pktopts->ippo_imo, sizeof(*imo));
1447 		} else {
1448 			memset(imo, 0, sizeof(*imo));
1449 			imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1450 			imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1451 		}
1452 		imo->imo_multicast_if_index = pktinfo->ipi_ifindex;
1453 		pktopts->ippo_imo = imo;
1454 		*flags |= IP_ROUTETOIFINDEX;
1455 	}
1456 	return error;
1457 }
1458 
1459 /*
1460  * Set up IP outgoing packet options. Even if control is NULL,
1461  * pktopts->ippo_laddr and pktopts->ippo_imo are set and used.
1462  */
1463 int
1464 ip_setpktopts(struct mbuf *control, struct ip_pktopts *pktopts, int *flags,
1465     struct inpcb *inp, kauth_cred_t cred)
1466 {
1467 	struct cmsghdr *cm;
1468 	struct in_pktinfo pktinfo;
1469 	int error;
1470 
1471 	pktopts->ippo_imo = inp->inp_moptions;
1472 
1473 	struct in_addr *ia = in_nullhost(inp->inp_prefsrcip) ? &inp->inp_laddr :
1474 	    &inp->inp_prefsrcip;
1475 	sockaddr_in_init(&pktopts->ippo_laddr, ia, 0);
1476 
1477 	if (control == NULL)
1478 		return 0;
1479 
1480 	/*
1481 	 * XXX: Currently, we assume all the optional information is
1482 	 * stored in a single mbuf.
1483 	 */
1484 	if (control->m_next)
1485 		return EINVAL;
1486 
1487 	for (; control->m_len > 0;
1488 	    control->m_data += CMSG_ALIGN(cm->cmsg_len),
1489 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1490 		cm = mtod(control, struct cmsghdr *);
1491 		if ((control->m_len < sizeof(*cm)) ||
1492 		    (cm->cmsg_len == 0) ||
1493 		    (cm->cmsg_len > control->m_len)) {
1494 			return EINVAL;
1495 		}
1496 		if (cm->cmsg_level != IPPROTO_IP)
1497 			continue;
1498 
1499 		switch (cm->cmsg_type) {
1500 		case IP_PKTINFO:
1501 			if (cm->cmsg_len != CMSG_LEN(sizeof(pktinfo)))
1502 				return EINVAL;
1503 			memcpy(&pktinfo, CMSG_DATA(cm), sizeof(pktinfo));
1504 			error = ip_pktinfo_prepare(&pktinfo, pktopts, flags,
1505 			    cred);
1506 			if (error)
1507 				return error;
1508 			break;
1509 		case IP_SENDSRCADDR: /* FreeBSD compatibility */
1510 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_addr)))
1511 				return EINVAL;
1512 			pktinfo.ipi_ifindex = 0;
1513 			pktinfo.ipi_addr =
1514 			    ((struct in_pktinfo *)CMSG_DATA(cm))->ipi_addr;
1515 			error = ip_pktinfo_prepare(&pktinfo, pktopts, flags,
1516 			    cred);
1517 			if (error)
1518 				return error;
1519 			break;
1520 		default:
1521 			return ENOPROTOOPT;
1522 		}
1523 	}
1524 	return 0;
1525 }
1526 
1527 /*
1528  * Set up IP options in pcb for insertion in output packets.
1529  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1530  * with destination address if source routed.
1531  */
1532 static int
1533 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt)
1534 {
1535 	struct mbuf *m;
1536 	const u_char *cp;
1537 	u_char *dp;
1538 	int cnt;
1539 
1540 	KASSERT(inp_locked(inp));
1541 
1542 	/* Turn off any old options. */
1543 	if (inp->inp_options) {
1544 		m_free(inp->inp_options);
1545 	}
1546 	inp->inp_options = NULL;
1547 	if ((cnt = sopt->sopt_size) == 0) {
1548 		/* Only turning off any previous options. */
1549 		return 0;
1550 	}
1551 	cp = sopt->sopt_data;
1552 
1553 	if (cnt % 4) {
1554 		/* Must be 4-byte aligned, because there's no padding. */
1555 		return EINVAL;
1556 	}
1557 
1558 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1559 	if (m == NULL)
1560 		return ENOBUFS;
1561 
1562 	dp = mtod(m, u_char *);
1563 	memset(dp, 0, sizeof(struct in_addr));
1564 	dp += sizeof(struct in_addr);
1565 	m->m_len = sizeof(struct in_addr);
1566 
1567 	/*
1568 	 * IP option list according to RFC791. Each option is of the form
1569 	 *
1570 	 *	[optval] [olen] [(olen - 2) data bytes]
1571 	 *
1572 	 * We validate the list and copy options to an mbuf for prepending
1573 	 * to data packets. The IP first-hop destination address will be
1574 	 * stored before actual options and is zero if unset.
1575 	 */
1576 	while (cnt > 0) {
1577 		uint8_t optval, olen, offset;
1578 
1579 		optval = cp[IPOPT_OPTVAL];
1580 
1581 		if (optval == IPOPT_EOL || optval == IPOPT_NOP) {
1582 			olen = 1;
1583 		} else {
1584 			if (cnt < IPOPT_OLEN + 1)
1585 				goto bad;
1586 
1587 			olen = cp[IPOPT_OLEN];
1588 			if (olen < IPOPT_OLEN + 1 || olen > cnt)
1589 				goto bad;
1590 		}
1591 
1592 		if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) {
1593 			/*
1594 			 * user process specifies route as:
1595 			 *	->A->B->C->D
1596 			 * D must be our final destination (but we can't
1597 			 * check that since we may not have connected yet).
1598 			 * A is first hop destination, which doesn't appear in
1599 			 * actual IP option, but is stored before the options.
1600 			 */
1601 			if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr))
1602 				goto bad;
1603 
1604 			offset = cp[IPOPT_OFFSET];
1605 			memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1,
1606 			    sizeof(struct in_addr));
1607 
1608 			cp += sizeof(struct in_addr);
1609 			cnt -= sizeof(struct in_addr);
1610 			olen -= sizeof(struct in_addr);
1611 
1612 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1613 				goto bad;
1614 
1615 			memcpy(dp, cp, olen);
1616 			dp[IPOPT_OPTVAL] = optval;
1617 			dp[IPOPT_OLEN] = olen;
1618 			dp[IPOPT_OFFSET] = offset;
1619 			break;
1620 		} else {
1621 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1622 				goto bad;
1623 
1624 			memcpy(dp, cp, olen);
1625 			break;
1626 		}
1627 
1628 		dp += olen;
1629 		m->m_len += olen;
1630 
1631 		if (optval == IPOPT_EOL)
1632 			break;
1633 
1634 		cp += olen;
1635 		cnt -= olen;
1636 	}
1637 
1638 	inp->inp_options = m;
1639 	return 0;
1640 
1641 bad:
1642 	(void)m_free(m);
1643 	return EINVAL;
1644 }
1645 
1646 /*
1647  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1648  * Must be called in a pserialize critical section.
1649  */
1650 static struct ifnet *
1651 ip_multicast_if(struct in_addr *a, int *ifindexp)
1652 {
1653 	int ifindex;
1654 	struct ifnet *ifp = NULL;
1655 	struct in_ifaddr *ia;
1656 
1657 	if (ifindexp)
1658 		*ifindexp = 0;
1659 	if (ntohl(a->s_addr) >> 24 == 0) {
1660 		ifindex = ntohl(a->s_addr) & 0xffffff;
1661 		ifp = if_byindex(ifindex);
1662 		if (!ifp)
1663 			return NULL;
1664 		if (ifindexp)
1665 			*ifindexp = ifindex;
1666 	} else {
1667 		IN_ADDRHASH_READER_FOREACH(ia, a->s_addr) {
1668 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1669 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1670 				ifp = ia->ia_ifp;
1671 				if (if_is_deactivated(ifp))
1672 					ifp = NULL;
1673 				break;
1674 			}
1675 		}
1676 	}
1677 	return ifp;
1678 }
1679 
1680 static int
1681 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval)
1682 {
1683 	u_int tval;
1684 	u_char cval;
1685 	int error;
1686 
1687 	if (sopt == NULL)
1688 		return EINVAL;
1689 
1690 	switch (sopt->sopt_size) {
1691 	case sizeof(u_char):
1692 		error = sockopt_get(sopt, &cval, sizeof(u_char));
1693 		tval = cval;
1694 		break;
1695 
1696 	case sizeof(u_int):
1697 		error = sockopt_get(sopt, &tval, sizeof(u_int));
1698 		break;
1699 
1700 	default:
1701 		error = EINVAL;
1702 	}
1703 
1704 	if (error)
1705 		return error;
1706 
1707 	if (tval > maxval)
1708 		return EINVAL;
1709 
1710 	*val = tval;
1711 	return 0;
1712 }
1713 
1714 static int
1715 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
1716     struct psref *psref, struct in_addr *ia, bool add)
1717 {
1718 	int error;
1719 	struct ip_mreq mreq;
1720 
1721 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
1722 	if (error)
1723 		return error;
1724 
1725 	if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr))
1726 		return EINVAL;
1727 
1728 	memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia));
1729 
1730 	if (in_nullhost(mreq.imr_interface)) {
1731 		union {
1732 			struct sockaddr		dst;
1733 			struct sockaddr_in	dst4;
1734 		} u;
1735 		struct route ro;
1736 
1737 		if (!add) {
1738 			*ifp = NULL;
1739 			return 0;
1740 		}
1741 		/*
1742 		 * If no interface address was provided, use the interface of
1743 		 * the route to the given multicast address.
1744 		 */
1745 		struct rtentry *rt;
1746 		memset(&ro, 0, sizeof(ro));
1747 
1748 		sockaddr_in_init(&u.dst4, ia, 0);
1749 		error = rtcache_setdst(&ro, &u.dst);
1750 		if (error != 0)
1751 			return error;
1752 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
1753 		if (*ifp != NULL) {
1754 			if (if_is_deactivated(*ifp))
1755 				*ifp = NULL;
1756 			else
1757 				if_acquire(*ifp, psref);
1758 		}
1759 		rtcache_unref(rt, &ro);
1760 		rtcache_free(&ro);
1761 	} else {
1762 		int s = pserialize_read_enter();
1763 		*ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1764 		if (!add && *ifp == NULL) {
1765 			pserialize_read_exit(s);
1766 			return EADDRNOTAVAIL;
1767 		}
1768 		if (*ifp != NULL) {
1769 			if (if_is_deactivated(*ifp))
1770 				*ifp = NULL;
1771 			else
1772 				if_acquire(*ifp, psref);
1773 		}
1774 		pserialize_read_exit(s);
1775 	}
1776 	return 0;
1777 }
1778 
1779 /*
1780  * Add a multicast group membership.
1781  * Group must be a valid IP multicast address.
1782  */
1783 static int
1784 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1785 {
1786 	struct ifnet *ifp = NULL;	// XXX: gcc [ppc]
1787 	struct in_addr ia;
1788 	int i, error, bound;
1789 	struct psref psref;
1790 
1791 	/* imo is protected by solock or referenced only by the caller */
1792 
1793 	bound = curlwp_bind();
1794 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1795 		error = ip_get_membership(sopt, &ifp, &psref, &ia, true);
1796 	else {
1797 #ifdef INET6
1798 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1799 #else
1800 		error = EINVAL;
1801 #endif
1802 	}
1803 
1804 	if (error)
1805 		goto out;
1806 
1807 	/*
1808 	 * See if we found an interface, and confirm that it
1809 	 * supports multicast.
1810 	 */
1811 	if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1812 		error = EADDRNOTAVAIL;
1813 		goto out;
1814 	}
1815 
1816 	/*
1817 	 * See if the membership already exists or if all the
1818 	 * membership slots are full.
1819 	 */
1820 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1821 		if (imo->imo_membership[i]->inm_ifp == ifp &&
1822 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1823 			break;
1824 	}
1825 	if (i < imo->imo_num_memberships) {
1826 		error = EADDRINUSE;
1827 		goto out;
1828 	}
1829 
1830 	if (i == IP_MAX_MEMBERSHIPS) {
1831 		error = ETOOMANYREFS;
1832 		goto out;
1833 	}
1834 
1835 	/*
1836 	 * Everything looks good; add a new record to the multicast
1837 	 * address list for the given interface.
1838 	 */
1839 	IFNET_LOCK(ifp);
1840 	imo->imo_membership[i] = in_addmulti(&ia, ifp);
1841 	IFNET_UNLOCK(ifp);
1842 	if (imo->imo_membership[i] == NULL) {
1843 		error = ENOBUFS;
1844 		goto out;
1845 	}
1846 
1847 	++imo->imo_num_memberships;
1848 	error = 0;
1849 out:
1850 	if_put(ifp, &psref);
1851 	curlwp_bindx(bound);
1852 	return error;
1853 }
1854 
1855 /*
1856  * Drop a multicast group membership.
1857  * Group must be a valid IP multicast address.
1858  */
1859 static int
1860 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1861 {
1862 	struct in_addr ia = { .s_addr = 0 };	// XXX: gcc [ppc]
1863 	struct ifnet *ifp = NULL;		// XXX: gcc [ppc]
1864 	int i, error, bound;
1865 	struct psref psref;
1866 
1867 	/* imo is protected by solock or referenced only by the caller */
1868 
1869 	bound = curlwp_bind();
1870 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1871 		error = ip_get_membership(sopt, &ifp, &psref, &ia, false);
1872 	else {
1873 #ifdef INET6
1874 		error = ip6_get_membership(sopt, &ifp, &psref, &ia, sizeof(ia));
1875 #else
1876 		error = EINVAL;
1877 #endif
1878 	}
1879 
1880 	if (error)
1881 		goto out;
1882 
1883 	/*
1884 	 * Find the membership in the membership array.
1885 	 */
1886 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1887 		if ((ifp == NULL ||
1888 		     imo->imo_membership[i]->inm_ifp == ifp) &&
1889 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1890 			break;
1891 	}
1892 	if (i == imo->imo_num_memberships) {
1893 		error = EADDRNOTAVAIL;
1894 		goto out;
1895 	}
1896 
1897 	/*
1898 	 * Give up the multicast address record to which the
1899 	 * membership points.
1900 	 */
1901 	struct ifnet *inm_ifp = imo->imo_membership[i]->inm_ifp;
1902 	IFNET_LOCK(inm_ifp);
1903 	in_delmulti(imo->imo_membership[i]);
1904 	IFNET_UNLOCK(inm_ifp);
1905 
1906 	/*
1907 	 * Remove the gap in the membership array.
1908 	 */
1909 	for (++i; i < imo->imo_num_memberships; ++i)
1910 		imo->imo_membership[i-1] = imo->imo_membership[i];
1911 	--imo->imo_num_memberships;
1912 	error = 0;
1913 out:
1914 	if_put(ifp, &psref);
1915 	curlwp_bindx(bound);
1916 	return error;
1917 }
1918 
1919 /*
1920  * Set the IP multicast options in response to user setsockopt().
1921  */
1922 int
1923 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt)
1924 {
1925 	struct ip_moptions *imo = *pimo;
1926 	struct in_addr addr;
1927 	struct ifnet *ifp;
1928 	int ifindex, error = 0;
1929 
1930 	/* The passed imo isn't NULL, it should be protected by solock */
1931 
1932 	if (!imo) {
1933 		/*
1934 		 * No multicast option buffer attached to the pcb;
1935 		 * allocate one and initialize to default values.
1936 		 */
1937 		imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP);
1938 		if (imo == NULL)
1939 			return ENOBUFS;
1940 
1941 		imo->imo_multicast_if_index = 0;
1942 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1943 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1944 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1945 		imo->imo_num_memberships = 0;
1946 		*pimo = imo;
1947 	}
1948 
1949 	switch (sopt->sopt_name) {
1950 	case IP_MULTICAST_IF: {
1951 		int s;
1952 		/*
1953 		 * Select the interface for outgoing multicast packets.
1954 		 */
1955 		error = sockopt_get(sopt, &addr, sizeof(addr));
1956 		if (error)
1957 			break;
1958 
1959 		/*
1960 		 * INADDR_ANY is used to remove a previous selection.
1961 		 * When no interface is selected, a default one is
1962 		 * chosen every time a multicast packet is sent.
1963 		 */
1964 		if (in_nullhost(addr)) {
1965 			imo->imo_multicast_if_index = 0;
1966 			break;
1967 		}
1968 		/*
1969 		 * The selected interface is identified by its local
1970 		 * IP address.  Find the interface and confirm that
1971 		 * it supports multicasting.
1972 		 */
1973 		s = pserialize_read_enter();
1974 		ifp = ip_multicast_if(&addr, &ifindex);
1975 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1976 			pserialize_read_exit(s);
1977 			error = EADDRNOTAVAIL;
1978 			break;
1979 		}
1980 		imo->imo_multicast_if_index = ifp->if_index;
1981 		pserialize_read_exit(s);
1982 		if (ifindex)
1983 			imo->imo_multicast_addr = addr;
1984 		else
1985 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1986 		break;
1987 	    }
1988 
1989 	case IP_MULTICAST_TTL:
1990 		/*
1991 		 * Set the IP time-to-live for outgoing multicast packets.
1992 		 */
1993 		error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL);
1994 		break;
1995 
1996 	case IP_MULTICAST_LOOP:
1997 		/*
1998 		 * Set the loopback flag for outgoing multicast packets.
1999 		 * Must be zero or one.
2000 		 */
2001 		error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1);
2002 		break;
2003 
2004 	case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */
2005 		error = ip_add_membership(imo, sopt);
2006 		break;
2007 
2008 	case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */
2009 		error = ip_drop_membership(imo, sopt);
2010 		break;
2011 
2012 	default:
2013 		error = EOPNOTSUPP;
2014 		break;
2015 	}
2016 
2017 	/*
2018 	 * If all options have default values, no need to keep the mbuf.
2019 	 */
2020 	if (imo->imo_multicast_if_index == 0 &&
2021 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2022 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2023 	    imo->imo_num_memberships == 0) {
2024 		kmem_intr_free(imo, sizeof(*imo));
2025 		*pimo = NULL;
2026 	}
2027 
2028 	return error;
2029 }
2030 
2031 /*
2032  * Return the IP multicast options in response to user getsockopt().
2033  */
2034 int
2035 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt)
2036 {
2037 	struct in_addr addr;
2038 	uint8_t optval;
2039 	int error = 0;
2040 
2041 	/* imo is protected by solock or refereced only by the caller */
2042 
2043 	switch (sopt->sopt_name) {
2044 	case IP_MULTICAST_IF:
2045 		if (imo == NULL || imo->imo_multicast_if_index == 0)
2046 			addr = zeroin_addr;
2047 		else if (imo->imo_multicast_addr.s_addr) {
2048 			/* return the value user has set */
2049 			addr = imo->imo_multicast_addr;
2050 		} else {
2051 			struct ifnet *ifp;
2052 			struct in_ifaddr *ia = NULL;
2053 			int s = pserialize_read_enter();
2054 
2055 			ifp = if_byindex(imo->imo_multicast_if_index);
2056 			if (ifp != NULL) {
2057 				ia = in_get_ia_from_ifp(ifp);
2058 			}
2059 			addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
2060 			pserialize_read_exit(s);
2061 		}
2062 		error = sockopt_set(sopt, &addr, sizeof(addr));
2063 		break;
2064 
2065 	case IP_MULTICAST_TTL:
2066 		optval = imo ? imo->imo_multicast_ttl
2067 		    : IP_DEFAULT_MULTICAST_TTL;
2068 
2069 		error = sockopt_set(sopt, &optval, sizeof(optval));
2070 		break;
2071 
2072 	case IP_MULTICAST_LOOP:
2073 		optval = imo ? imo->imo_multicast_loop
2074 		    : IP_DEFAULT_MULTICAST_LOOP;
2075 
2076 		error = sockopt_set(sopt, &optval, sizeof(optval));
2077 		break;
2078 
2079 	default:
2080 		error = EOPNOTSUPP;
2081 	}
2082 
2083 	return error;
2084 }
2085 
2086 /*
2087  * Discard the IP multicast options.
2088  */
2089 void
2090 ip_freemoptions(struct ip_moptions *imo)
2091 {
2092 	int i;
2093 
2094 	/* The owner of imo (inp) should be protected by solock */
2095 
2096 	if (imo != NULL) {
2097 		for (i = 0; i < imo->imo_num_memberships; ++i) {
2098 			struct in_multi *inm = imo->imo_membership[i];
2099 			struct ifnet *ifp = inm->inm_ifp;
2100 			IFNET_LOCK(ifp);
2101 			in_delmulti(inm);
2102 			/* ifp should not leave thanks to solock */
2103 			IFNET_UNLOCK(ifp);
2104 		}
2105 
2106 		kmem_intr_free(imo, sizeof(*imo));
2107 	}
2108 }
2109 
2110 /*
2111  * Routine called from ip_output() to loop back a copy of an IP multicast
2112  * packet to the input queue of a specified interface.  Note that this
2113  * calls the output routine of the loopback "driver", but with an interface
2114  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2115  */
2116 static void
2117 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst)
2118 {
2119 	struct ip *ip;
2120 	struct mbuf *copym;
2121 
2122 	copym = m_copypacket(m, M_DONTWAIT);
2123 	if (copym != NULL &&
2124 	    (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
2125 		copym = m_pullup(copym, sizeof(struct ip));
2126 	if (copym == NULL)
2127 		return;
2128 	/*
2129 	 * We don't bother to fragment if the IP length is greater
2130 	 * than the interface's MTU.  Can this possibly matter?
2131 	 */
2132 	ip = mtod(copym, struct ip *);
2133 
2134 	if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
2135 		in_undefer_cksum_tcpudp(copym);
2136 		copym->m_pkthdr.csum_flags &=
2137 		    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
2138 	}
2139 
2140 	ip->ip_sum = 0;
2141 	ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
2142 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
2143 	(void)looutput(ifp, copym, sintocsa(dst), NULL);
2144 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
2145 }
2146 
2147 /*
2148  * Ensure sending address is valid.
2149  * Returns 0 on success, -1 if an error should be sent back or 1
2150  * if the packet could be dropped without error (protocol dependent).
2151  */
2152 static int
2153 ip_ifaddrvalid(const struct in_ifaddr *ia)
2154 {
2155 
2156 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
2157 		return 0;
2158 
2159 	if (ia->ia4_flags & IN_IFF_DUPLICATED)
2160 		return -1;
2161 	else if (ia->ia4_flags & (IN_IFF_TENTATIVE | IN_IFF_DETACHED))
2162 		return 1;
2163 
2164 	return 0;
2165 }
2166