xref: /netbsd-src/sys/netinet/ip_output.c (revision 6cd39ddb8550f6fa1bff3fed32053d7f19fd0453)
1 /*	$NetBSD: ip_output.c,v 1.248 2016/01/20 22:12:22 riastradh 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 1.248 2016/01/20 22:12:22 riastradh 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 <sys/param.h>
105 #include <sys/kmem.h>
106 #include <sys/mbuf.h>
107 #include <sys/protosw.h>
108 #include <sys/socket.h>
109 #include <sys/socketvar.h>
110 #include <sys/kauth.h>
111 #ifdef IPSEC
112 #include <sys/domain.h>
113 #endif
114 #include <sys/systm.h>
115 
116 #include <net/if.h>
117 #include <net/if_types.h>
118 #include <net/route.h>
119 #include <net/pfil.h>
120 
121 #include <netinet/in.h>
122 #include <netinet/in_systm.h>
123 #include <netinet/ip.h>
124 #include <netinet/in_pcb.h>
125 #include <netinet/in_var.h>
126 #include <netinet/ip_var.h>
127 #include <netinet/ip_private.h>
128 #include <netinet/in_offload.h>
129 #include <netinet/portalgo.h>
130 #include <netinet/udp.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 
156 extern pfil_head_t *inet_pfil_hook;			/* XXX */
157 
158 int	ip_do_loopback_cksum = 0;
159 
160 static bool
161 ip_hresolv_needed(const struct ifnet * const ifp)
162 {
163 	switch (ifp->if_type) {
164 	case IFT_ARCNET:
165 	case IFT_ATM:
166 	case IFT_ECONET:
167 	case IFT_ETHER:
168 	case IFT_FDDI:
169 	case IFT_HIPPI:
170 	case IFT_IEEE1394:
171 	case IFT_ISO88025:
172 	case IFT_SLIP:
173 		return true;
174 	default:
175 		return false;
176 	}
177 }
178 
179 static int
180 klock_if_output(struct ifnet * const ifp, struct mbuf * const m,
181     const struct sockaddr * const dst, struct rtentry *rt)
182 {
183 	int error;
184 
185 #ifndef NET_MPSAFE
186 	KERNEL_LOCK(1, NULL);
187 #endif
188 
189 	error = (*ifp->if_output)(ifp, m, dst, rt);
190 
191 #ifndef NET_MPSAFE
192 	KERNEL_UNLOCK_ONE(NULL);
193 #endif
194 
195 	return error;
196 }
197 
198 /*
199  * Send an IP packet to a host.
200  *
201  * If necessary, resolve the arbitrary IP route, rt0, to an IP host route before
202  * calling ifp's output routine.
203  */
204 int
205 ip_hresolv_output(struct ifnet * const ifp0, struct mbuf * const m,
206     const struct sockaddr * const dst, struct rtentry *rt00)
207 {
208 	int error = 0;
209 	struct ifnet *ifp = ifp0;
210 	struct rtentry *rt, *rt0, *gwrt;
211 
212 #define RTFREE_IF_NEEDED(_rt) \
213 	if ((_rt) != NULL && (_rt) != rt00) \
214 		rtfree((_rt));
215 
216 	rt0 = rt00;
217 retry:
218 	if (!ip_hresolv_needed(ifp)) {
219 		rt = rt0;
220 		goto out;
221 	}
222 
223 	if (rt0 == NULL) {
224 		rt = NULL;
225 		goto out;
226 	}
227 
228 	rt = rt0;
229 
230 	/*
231 	 * The following block is highly questionable.  How did we get here
232 	 * with a !RTF_UP route?  Does rtalloc1() always return an RTF_UP
233 	 * route?
234 	 */
235 	if ((rt->rt_flags & RTF_UP) == 0) {
236 		rt = rtalloc1(dst, 1);
237 		if (rt == NULL) {
238 			error = EHOSTUNREACH;
239 			goto bad;
240 		}
241 		rt0 = rt;
242 		if (rt->rt_ifp != ifp) {
243 			ifp = rt->rt_ifp;
244 			goto retry;
245 		}
246 	}
247 
248 	if ((rt->rt_flags & RTF_GATEWAY) == 0)
249 		goto out;
250 
251 	gwrt = rt_get_gwroute(rt);
252 	RTFREE_IF_NEEDED(rt);
253 	rt = gwrt;
254 	if (rt == NULL || (rt->rt_flags & RTF_UP) == 0) {
255 		if (rt != NULL) {
256 			RTFREE_IF_NEEDED(rt);
257 			rt = rt0;
258 		}
259 		if (rt == NULL) {
260 			error = EHOSTUNREACH;
261 			goto bad;
262 		}
263 		gwrt = rtalloc1(rt->rt_gateway, 1);
264 		rt_set_gwroute(rt, gwrt);
265 		RTFREE_IF_NEEDED(rt);
266 		rt = gwrt;
267 		if (rt == NULL) {
268 			error = EHOSTUNREACH;
269 			goto bad;
270 		}
271 		/* the "G" test below also prevents rt == rt0 */
272 		if ((rt->rt_flags & RTF_GATEWAY) != 0 || rt->rt_ifp != ifp) {
273 			if (rt0->rt_gwroute != NULL)
274 				rtfree(rt0->rt_gwroute);
275 			rt0->rt_gwroute = NULL;
276 			error = EHOSTUNREACH;
277 			goto bad;
278 		}
279 	}
280 	if ((rt->rt_flags & RTF_REJECT) != 0) {
281 		if (rt->rt_rmx.rmx_expire == 0 ||
282 		    time_uptime < rt->rt_rmx.rmx_expire) {
283 			error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH;
284 			goto bad;
285 		}
286 	}
287 
288 out:
289 #ifdef MPLS
290 	if (rt0 != NULL && rt_gettag(rt0) != NULL &&
291 	    rt_gettag(rt0)->sa_family == AF_MPLS &&
292 	    (m->m_flags & (M_MCAST | M_BCAST)) == 0 &&
293 	    ifp->if_type == IFT_ETHER) {
294 		union mpls_shim msh;
295 		msh.s_addr = MPLS_GETSADDR(rt0);
296 		if (msh.shim.label != MPLS_LABEL_IMPLNULL) {
297 			struct m_tag *mtag;
298 			/*
299 			 * XXX tentative solution to tell ether_output
300 			 * it's MPLS. Need some more efficient solution.
301 			 */
302 			mtag = m_tag_get(PACKET_TAG_MPLS,
303 			    sizeof(int) /* dummy */,
304 			    M_NOWAIT);
305 			if (mtag == NULL) {
306 				error = ENOMEM;
307 				goto bad;
308 			}
309 			m_tag_prepend(m, mtag);
310 		}
311 	}
312 #endif
313 
314 	error = klock_if_output(ifp, m, dst, rt);
315 	goto exit;
316 
317 bad:
318 	if (m != NULL)
319 		m_freem(m);
320 exit:
321 	RTFREE_IF_NEEDED(rt);
322 
323 	return error;
324 
325 #undef RTFREE_IF_NEEDED
326 }
327 
328 /*
329  * IP output.  The packet in mbuf chain m contains a skeletal IP
330  * header (with len, off, ttl, proto, tos, src, dst).
331  * The mbuf chain containing the packet will be freed.
332  * The mbuf opt, if present, will not be freed.
333  */
334 int
335 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags,
336     struct ip_moptions *imo, struct socket *so)
337 {
338 	struct rtentry *rt;
339 	struct ip *ip;
340 	struct ifnet *ifp;
341 	struct mbuf *m = m0;
342 	int hlen = sizeof (struct ip);
343 	int len, error = 0;
344 	struct route iproute;
345 	const struct sockaddr_in *dst;
346 	struct in_ifaddr *ia;
347 	int isbroadcast;
348 	int sw_csum;
349 	u_long mtu;
350 #ifdef IPSEC
351 	struct secpolicy *sp = NULL;
352 #endif
353 	bool natt_frag = false;
354 	bool rtmtu_nolock;
355 	union {
356 		struct sockaddr		dst;
357 		struct sockaddr_in	dst4;
358 	} u;
359 	struct sockaddr *rdst = &u.dst;	/* real IP destination, as opposed
360 					 * to the nexthop
361 					 */
362 
363 	len = 0;
364 
365 	MCLAIM(m, &ip_tx_mowner);
366 
367 	KASSERT((m->m_flags & M_PKTHDR) != 0);
368 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) == 0);
369 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) !=
370 	    (M_CSUM_TCPv4|M_CSUM_UDPv4));
371 
372 	if (opt) {
373 		m = ip_insertoptions(m, opt, &len);
374 		if (len >= sizeof(struct ip))
375 			hlen = len;
376 	}
377 	ip = mtod(m, struct ip *);
378 
379 	/*
380 	 * Fill in IP header.
381 	 */
382 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
383 		ip->ip_v = IPVERSION;
384 		ip->ip_off = htons(0);
385 		/* ip->ip_id filled in after we find out source ia */
386 		ip->ip_hl = hlen >> 2;
387 		IP_STATINC(IP_STAT_LOCALOUT);
388 	} else {
389 		hlen = ip->ip_hl << 2;
390 	}
391 
392 	/*
393 	 * Route packet.
394 	 */
395 	if (ro == NULL) {
396 		memset(&iproute, 0, sizeof(iproute));
397 		ro = &iproute;
398 	}
399 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
400 	dst = satocsin(rtcache_getdst(ro));
401 
402 	/*
403 	 * If there is a cached route, check that it is to the same
404 	 * destination and is still up.  If not, free it and try again.
405 	 * The address family should also be checked in case of sharing
406 	 * the cache with IPv6.
407 	 */
408 	if (dst && (dst->sin_family != AF_INET ||
409 	    !in_hosteq(dst->sin_addr, ip->ip_dst)))
410 		rtcache_free(ro);
411 
412 	if ((rt = rtcache_validate(ro)) == NULL &&
413 	    (rt = rtcache_update(ro, 1)) == NULL) {
414 		dst = &u.dst4;
415 		error = rtcache_setdst(ro, &u.dst);
416 		if (error != 0)
417 			goto bad;
418 	}
419 
420 	/*
421 	 * If routing to interface only, short circuit routing lookup.
422 	 */
423 	if (flags & IP_ROUTETOIF) {
424 		if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) {
425 			IP_STATINC(IP_STAT_NOROUTE);
426 			error = ENETUNREACH;
427 			goto bad;
428 		}
429 		ifp = ia->ia_ifp;
430 		mtu = ifp->if_mtu;
431 		ip->ip_ttl = 1;
432 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
433 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
434 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
435 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
436 		ifp = imo->imo_multicast_ifp;
437 		mtu = ifp->if_mtu;
438 		IFP_TO_IA(ifp, ia);
439 		isbroadcast = 0;
440 	} else {
441 		if (rt == NULL)
442 			rt = rtcache_init(ro);
443 		if (rt == NULL) {
444 			IP_STATINC(IP_STAT_NOROUTE);
445 			error = EHOSTUNREACH;
446 			goto bad;
447 		}
448 		ia = ifatoia(rt->rt_ifa);
449 		ifp = rt->rt_ifp;
450 		if ((mtu = rt->rt_rmx.rmx_mtu) == 0)
451 			mtu = ifp->if_mtu;
452 		rt->rt_use++;
453 		if (rt->rt_flags & RTF_GATEWAY)
454 			dst = satosin(rt->rt_gateway);
455 		if (rt->rt_flags & RTF_HOST)
456 			isbroadcast = rt->rt_flags & RTF_BROADCAST;
457 		else
458 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
459 	}
460 	rtmtu_nolock = rt && (rt->rt_rmx.rmx_locks & RTV_MTU) == 0;
461 
462 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
463 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
464 		bool inmgroup;
465 
466 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
467 		    M_BCAST : M_MCAST;
468 		/*
469 		 * See if the caller provided any multicast options
470 		 */
471 		if (imo != NULL)
472 			ip->ip_ttl = imo->imo_multicast_ttl;
473 		else
474 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
475 
476 		/*
477 		 * if we don't know the outgoing ifp yet, we can't generate
478 		 * output
479 		 */
480 		if (!ifp) {
481 			IP_STATINC(IP_STAT_NOROUTE);
482 			error = ENETUNREACH;
483 			goto bad;
484 		}
485 
486 		/*
487 		 * If the packet is multicast or broadcast, confirm that
488 		 * the outgoing interface can transmit it.
489 		 */
490 		if (((m->m_flags & M_MCAST) &&
491 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
492 		    ((m->m_flags & M_BCAST) &&
493 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
494 			IP_STATINC(IP_STAT_NOROUTE);
495 			error = ENETUNREACH;
496 			goto bad;
497 		}
498 		/*
499 		 * If source address not specified yet, use an address
500 		 * of outgoing interface.
501 		 */
502 		if (in_nullhost(ip->ip_src)) {
503 			struct in_ifaddr *xia;
504 			struct ifaddr *xifa;
505 
506 			IFP_TO_IA(ifp, xia);
507 			if (!xia) {
508 				error = EADDRNOTAVAIL;
509 				goto bad;
510 			}
511 			xifa = &xia->ia_ifa;
512 			if (xifa->ifa_getifa != NULL) {
513 				xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
514 				if (xia == NULL) {
515 					error = EADDRNOTAVAIL;
516 					goto bad;
517 				}
518 			}
519 			ip->ip_src = xia->ia_addr.sin_addr;
520 		}
521 
522 		inmgroup = in_multi_group(ip->ip_dst, ifp, flags);
523 		if (inmgroup && (imo == NULL || imo->imo_multicast_loop)) {
524 			/*
525 			 * If we belong to the destination multicast group
526 			 * on the outgoing interface, and the caller did not
527 			 * forbid loopback, loop back a copy.
528 			 */
529 			ip_mloopback(ifp, m, &u.dst4);
530 		}
531 #ifdef MROUTING
532 		else {
533 			/*
534 			 * If we are acting as a multicast router, perform
535 			 * multicast forwarding as if the packet had just
536 			 * arrived on the interface to which we are about
537 			 * to send.  The multicast forwarding function
538 			 * recursively calls this function, using the
539 			 * IP_FORWARDING flag to prevent infinite recursion.
540 			 *
541 			 * Multicasts that are looped back by ip_mloopback(),
542 			 * above, will be forwarded by the ip_input() routine,
543 			 * if necessary.
544 			 */
545 			extern struct socket *ip_mrouter;
546 
547 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
548 				if (ip_mforward(m, ifp) != 0) {
549 					m_freem(m);
550 					goto done;
551 				}
552 			}
553 		}
554 #endif
555 		/*
556 		 * Multicasts with a time-to-live of zero may be looped-
557 		 * back, above, but must not be transmitted on a network.
558 		 * Also, multicasts addressed to the loopback interface
559 		 * are not sent -- the above call to ip_mloopback() will
560 		 * loop back a copy if this host actually belongs to the
561 		 * destination group on the loopback interface.
562 		 */
563 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
564 			m_freem(m);
565 			goto done;
566 		}
567 		goto sendit;
568 	}
569 
570 	/*
571 	 * If source address not specified yet, use address
572 	 * of outgoing interface.
573 	 */
574 	if (in_nullhost(ip->ip_src)) {
575 		struct ifaddr *xifa;
576 
577 		xifa = &ia->ia_ifa;
578 		if (xifa->ifa_getifa != NULL) {
579 			ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst));
580 			if (ia == NULL) {
581 				error = EADDRNOTAVAIL;
582 				goto bad;
583 			}
584 		}
585 		ip->ip_src = ia->ia_addr.sin_addr;
586 	}
587 
588 	/*
589 	 * packets with Class-D address as source are not valid per
590 	 * RFC 1112
591 	 */
592 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
593 		IP_STATINC(IP_STAT_ODROPPED);
594 		error = EADDRNOTAVAIL;
595 		goto bad;
596 	}
597 
598 	/*
599 	 * Look for broadcast address and and verify user is allowed to
600 	 * send such a packet.
601 	 */
602 	if (isbroadcast) {
603 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
604 			error = EADDRNOTAVAIL;
605 			goto bad;
606 		}
607 		if ((flags & IP_ALLOWBROADCAST) == 0) {
608 			error = EACCES;
609 			goto bad;
610 		}
611 		/* don't allow broadcast messages to be fragmented */
612 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
613 			error = EMSGSIZE;
614 			goto bad;
615 		}
616 		m->m_flags |= M_BCAST;
617 	} else
618 		m->m_flags &= ~M_BCAST;
619 
620 sendit:
621 	if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) {
622 		if (m->m_pkthdr.len < IP_MINFRAGSIZE) {
623 			ip->ip_id = 0;
624 		} else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
625 			ip->ip_id = ip_newid(ia);
626 		} else {
627 
628 			/*
629 			 * TSO capable interfaces (typically?) increment
630 			 * ip_id for each segment.
631 			 * "allocate" enough ids here to increase the chance
632 			 * for them to be unique.
633 			 *
634 			 * note that the following calculation is not
635 			 * needed to be precise.  wasting some ip_id is fine.
636 			 */
637 
638 			unsigned int segsz = m->m_pkthdr.segsz;
639 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
640 			unsigned int num = howmany(datasz, segsz);
641 
642 			ip->ip_id = ip_newid_range(ia, num);
643 		}
644 	}
645 
646 	/*
647 	 * If we're doing Path MTU Discovery, we need to set DF unless
648 	 * the route's MTU is locked.
649 	 */
650 	if ((flags & IP_MTUDISC) != 0 && rtmtu_nolock) {
651 		ip->ip_off |= htons(IP_DF);
652 	}
653 
654 #ifdef IPSEC
655 	if (ipsec_used) {
656 		bool ipsec_done = false;
657 
658 		/* Perform IPsec processing, if any. */
659 		error = ipsec4_output(m, so, flags, &sp, &mtu, &natt_frag,
660 		    &ipsec_done);
661 		if (error || ipsec_done)
662 			goto done;
663 	}
664 #endif
665 
666 	/*
667 	 * Run through list of hooks for output packets.
668 	 */
669 	error = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_OUT);
670 	if (error)
671 		goto done;
672 	if (m == NULL)
673 		goto done;
674 
675 	ip = mtod(m, struct ip *);
676 	hlen = ip->ip_hl << 2;
677 
678 	m->m_pkthdr.csum_data |= hlen << 16;
679 
680 #if IFA_STATS
681 	/*
682 	 * search for the source address structure to
683 	 * maintain output statistics.
684 	 */
685 	INADDR_TO_IA(ip->ip_src, ia);
686 #endif
687 
688 	/* Maybe skip checksums on loopback interfaces. */
689 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
690 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
691 	}
692 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
693 	/*
694 	 * If small enough for mtu of path, or if using TCP segmentation
695 	 * offload, can just send directly.
696 	 */
697 	if (ntohs(ip->ip_len) <= mtu ||
698 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
699 		const struct sockaddr *sa;
700 
701 #if IFA_STATS
702 		if (ia)
703 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
704 #endif
705 		/*
706 		 * Always initialize the sum to 0!  Some HW assisted
707 		 * checksumming requires this.
708 		 */
709 		ip->ip_sum = 0;
710 
711 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
712 			/*
713 			 * Perform any checksums that the hardware can't do
714 			 * for us.
715 			 *
716 			 * XXX Does any hardware require the {th,uh}_sum
717 			 * XXX fields to be 0?
718 			 */
719 			if (sw_csum & M_CSUM_IPv4) {
720 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
721 				ip->ip_sum = in_cksum(m, hlen);
722 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
723 			}
724 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
725 				if (IN_NEED_CHECKSUM(ifp,
726 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
727 					in_delayed_cksum(m);
728 				}
729 				m->m_pkthdr.csum_flags &=
730 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
731 			}
732 		}
733 
734 		sa = (m->m_flags & M_MCAST) ? sintocsa(rdst) : sintocsa(dst);
735 		if (__predict_true(
736 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
737 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
738 			error = ip_hresolv_output(ifp, m, sa, rt);
739 		} else {
740 			error = ip_tso_output(ifp, m, sa, rt);
741 		}
742 		goto done;
743 	}
744 
745 	/*
746 	 * We can't use HW checksumming if we're about to
747 	 * to fragment the packet.
748 	 *
749 	 * XXX Some hardware can do this.
750 	 */
751 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
752 		if (IN_NEED_CHECKSUM(ifp,
753 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
754 			in_delayed_cksum(m);
755 		}
756 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
757 	}
758 
759 	/*
760 	 * Too large for interface; fragment if possible.
761 	 * Must be able to put at least 8 bytes per fragment.
762 	 */
763 	if (ntohs(ip->ip_off) & IP_DF) {
764 		if (flags & IP_RETURNMTU) {
765 			struct inpcb *inp;
766 
767 			KASSERT(so && solocked(so));
768 			inp = sotoinpcb(so);
769 			inp->inp_errormtu = mtu;
770 		}
771 		error = EMSGSIZE;
772 		IP_STATINC(IP_STAT_CANTFRAG);
773 		goto bad;
774 	}
775 
776 	error = ip_fragment(m, ifp, mtu);
777 	if (error) {
778 		m = NULL;
779 		goto bad;
780 	}
781 
782 	for (; m; m = m0) {
783 		m0 = m->m_nextpkt;
784 		m->m_nextpkt = 0;
785 		if (error) {
786 			m_freem(m);
787 			continue;
788 		}
789 #if IFA_STATS
790 		if (ia)
791 			ia->ia_ifa.ifa_data.ifad_outbytes += ntohs(ip->ip_len);
792 #endif
793 		/*
794 		 * If we get there, the packet has not been handled by
795 		 * IPsec whereas it should have. Now that it has been
796 		 * fragmented, re-inject it in ip_output so that IPsec
797 		 * processing can occur.
798 		 */
799 		if (natt_frag) {
800 			error = ip_output(m, opt, ro,
801 			    flags | IP_RAWOUTPUT | IP_NOIPNEWID,
802 			    imo, so);
803 		} else {
804 			KASSERT((m->m_pkthdr.csum_flags &
805 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
806 			error = ip_hresolv_output(ifp, m,
807 			    (m->m_flags & M_MCAST) ?
808 			    sintocsa(rdst) : sintocsa(dst), rt);
809 		}
810 	}
811 	if (error == 0) {
812 		IP_STATINC(IP_STAT_FRAGMENTED);
813 	}
814 done:
815 	if (ro == &iproute) {
816 		rtcache_free(&iproute);
817 	}
818 #ifdef IPSEC
819 	if (sp) {
820 		KEY_FREESP(&sp);
821 	}
822 #endif
823 	return error;
824 bad:
825 	m_freem(m);
826 	goto done;
827 }
828 
829 int
830 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
831 {
832 	struct ip *ip, *mhip;
833 	struct mbuf *m0;
834 	int len, hlen, off;
835 	int mhlen, firstlen;
836 	struct mbuf **mnext;
837 	int sw_csum = m->m_pkthdr.csum_flags;
838 	int fragments = 0;
839 	int s;
840 	int error = 0;
841 
842 	ip = mtod(m, struct ip *);
843 	hlen = ip->ip_hl << 2;
844 	if (ifp != NULL)
845 		sw_csum &= ~ifp->if_csum_flags_tx;
846 
847 	len = (mtu - hlen) &~ 7;
848 	if (len < 8) {
849 		m_freem(m);
850 		return (EMSGSIZE);
851 	}
852 
853 	firstlen = len;
854 	mnext = &m->m_nextpkt;
855 
856 	/*
857 	 * Loop through length of segment after first fragment,
858 	 * make new header and copy data of each part and link onto chain.
859 	 */
860 	m0 = m;
861 	mhlen = sizeof (struct ip);
862 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
863 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
864 		if (m == 0) {
865 			error = ENOBUFS;
866 			IP_STATINC(IP_STAT_ODROPPED);
867 			goto sendorfree;
868 		}
869 		MCLAIM(m, m0->m_owner);
870 		*mnext = m;
871 		mnext = &m->m_nextpkt;
872 		m->m_data += max_linkhdr;
873 		mhip = mtod(m, struct ip *);
874 		*mhip = *ip;
875 		/* we must inherit MCAST and BCAST flags */
876 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
877 		if (hlen > sizeof (struct ip)) {
878 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
879 			mhip->ip_hl = mhlen >> 2;
880 		}
881 		m->m_len = mhlen;
882 		mhip->ip_off = ((off - hlen) >> 3) +
883 		    (ntohs(ip->ip_off) & ~IP_MF);
884 		if (ip->ip_off & htons(IP_MF))
885 			mhip->ip_off |= IP_MF;
886 		if (off + len >= ntohs(ip->ip_len))
887 			len = ntohs(ip->ip_len) - off;
888 		else
889 			mhip->ip_off |= IP_MF;
890 		HTONS(mhip->ip_off);
891 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
892 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
893 		if (m->m_next == 0) {
894 			error = ENOBUFS;	/* ??? */
895 			IP_STATINC(IP_STAT_ODROPPED);
896 			goto sendorfree;
897 		}
898 		m->m_pkthdr.len = mhlen + len;
899 		m->m_pkthdr.rcvif = NULL;
900 		mhip->ip_sum = 0;
901 		KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
902 		if (sw_csum & M_CSUM_IPv4) {
903 			mhip->ip_sum = in_cksum(m, mhlen);
904 		} else {
905 			/*
906 			 * checksum is hw-offloaded or not necessary.
907 			 */
908 			m->m_pkthdr.csum_flags |=
909 			    m0->m_pkthdr.csum_flags & M_CSUM_IPv4;
910 			m->m_pkthdr.csum_data |= mhlen << 16;
911 			KASSERT(!(ifp != NULL &&
912 			    IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
913 			    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
914 		}
915 		IP_STATINC(IP_STAT_OFRAGMENTS);
916 		fragments++;
917 	}
918 	/*
919 	 * Update first fragment by trimming what's been copied out
920 	 * and updating header, then send each fragment (in order).
921 	 */
922 	m = m0;
923 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
924 	m->m_pkthdr.len = hlen + firstlen;
925 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
926 	ip->ip_off |= htons(IP_MF);
927 	ip->ip_sum = 0;
928 	if (sw_csum & M_CSUM_IPv4) {
929 		ip->ip_sum = in_cksum(m, hlen);
930 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
931 	} else {
932 		/*
933 		 * checksum is hw-offloaded or not necessary.
934 		 */
935 		KASSERT(!(ifp != NULL && IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) ||
936 		    (m->m_pkthdr.csum_flags & M_CSUM_IPv4) != 0);
937 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
938 		    sizeof(struct ip));
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 		s = splnet();
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 		splx(s);
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 	return (error);
963 }
964 
965 /*
966  * Process a delayed payload checksum calculation.
967  */
968 void
969 in_delayed_cksum(struct mbuf *m)
970 {
971 	struct ip *ip;
972 	u_int16_t csum, offset;
973 
974 	ip = mtod(m, struct ip *);
975 	offset = ip->ip_hl << 2;
976 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
977 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
978 		csum = 0xffff;
979 
980 	offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
981 
982 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
983 		/* This happen when ip options were inserted
984 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
985 		    m->m_len, offset, ip->ip_p);
986 		 */
987 		m_copyback(m, offset, sizeof(csum), (void *) &csum);
988 	} else
989 		*(u_int16_t *)(mtod(m, char *) + offset) = csum;
990 }
991 
992 /*
993  * Determine the maximum length of the options to be inserted;
994  * we would far rather allocate too much space rather than too little.
995  */
996 
997 u_int
998 ip_optlen(struct inpcb *inp)
999 {
1000 	struct mbuf *m = inp->inp_options;
1001 
1002 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) {
1003 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1004 	}
1005 	return 0;
1006 }
1007 
1008 /*
1009  * Insert IP options into preformed packet.
1010  * Adjust IP destination as required for IP source routing,
1011  * as indicated by a non-zero in_addr at the start of the options.
1012  */
1013 static struct mbuf *
1014 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1015 {
1016 	struct ipoption *p = mtod(opt, struct ipoption *);
1017 	struct mbuf *n;
1018 	struct ip *ip = mtod(m, struct ip *);
1019 	unsigned optlen;
1020 
1021 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1022 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1023 		return (m);		/* XXX should fail */
1024 	if (!in_nullhost(p->ipopt_dst))
1025 		ip->ip_dst = p->ipopt_dst;
1026 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1027 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1028 		if (n == 0)
1029 			return (m);
1030 		MCLAIM(n, m->m_owner);
1031 		M_MOVE_PKTHDR(n, m);
1032 		m->m_len -= sizeof(struct ip);
1033 		m->m_data += sizeof(struct ip);
1034 		n->m_next = m;
1035 		m = n;
1036 		m->m_len = optlen + sizeof(struct ip);
1037 		m->m_data += max_linkhdr;
1038 		bcopy((void *)ip, mtod(m, void *), sizeof(struct ip));
1039 	} else {
1040 		m->m_data -= optlen;
1041 		m->m_len += optlen;
1042 		memmove(mtod(m, void *), ip, sizeof(struct ip));
1043 	}
1044 	m->m_pkthdr.len += optlen;
1045 	ip = mtod(m, struct ip *);
1046 	bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen);
1047 	*phlen = sizeof(struct ip) + optlen;
1048 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1049 	return (m);
1050 }
1051 
1052 /*
1053  * Copy options from ip to jp,
1054  * omitting those not copied during fragmentation.
1055  */
1056 int
1057 ip_optcopy(struct ip *ip, struct ip *jp)
1058 {
1059 	u_char *cp, *dp;
1060 	int opt, optlen, cnt;
1061 
1062 	cp = (u_char *)(ip + 1);
1063 	dp = (u_char *)(jp + 1);
1064 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1065 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1066 		opt = cp[0];
1067 		if (opt == IPOPT_EOL)
1068 			break;
1069 		if (opt == IPOPT_NOP) {
1070 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1071 			*dp++ = IPOPT_NOP;
1072 			optlen = 1;
1073 			continue;
1074 		}
1075 
1076 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp));
1077 		optlen = cp[IPOPT_OLEN];
1078 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen < cnt);
1079 
1080 		/* Invalid lengths should have been caught by ip_dooptions. */
1081 		if (optlen > cnt)
1082 			optlen = cnt;
1083 		if (IPOPT_COPIED(opt)) {
1084 			bcopy((void *)cp, (void *)dp, (unsigned)optlen);
1085 			dp += optlen;
1086 		}
1087 	}
1088 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1089 		*dp++ = IPOPT_EOL;
1090 	return (optlen);
1091 }
1092 
1093 /*
1094  * IP socket option processing.
1095  */
1096 int
1097 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1098 {
1099 	struct inpcb *inp = sotoinpcb(so);
1100 	struct ip *ip = &inp->inp_ip;
1101 	int inpflags = inp->inp_flags;
1102 	int optval = 0, error = 0;
1103 
1104 	if (sopt->sopt_level != IPPROTO_IP) {
1105 		if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER)
1106 			return 0;
1107 		return ENOPROTOOPT;
1108 	}
1109 
1110 	switch (op) {
1111 	case PRCO_SETOPT:
1112 		switch (sopt->sopt_name) {
1113 		case IP_OPTIONS:
1114 #ifdef notyet
1115 		case IP_RETOPTS:
1116 #endif
1117 			error = ip_pcbopts(inp, sopt);
1118 			break;
1119 
1120 		case IP_TOS:
1121 		case IP_TTL:
1122 		case IP_MINTTL:
1123 		case IP_PKTINFO:
1124 		case IP_RECVOPTS:
1125 		case IP_RECVRETOPTS:
1126 		case IP_RECVDSTADDR:
1127 		case IP_RECVIF:
1128 		case IP_RECVPKTINFO:
1129 		case IP_RECVTTL:
1130 			error = sockopt_getint(sopt, &optval);
1131 			if (error)
1132 				break;
1133 
1134 			switch (sopt->sopt_name) {
1135 			case IP_TOS:
1136 				ip->ip_tos = optval;
1137 				break;
1138 
1139 			case IP_TTL:
1140 				ip->ip_ttl = optval;
1141 				break;
1142 
1143 			case IP_MINTTL:
1144 				if (optval > 0 && optval <= MAXTTL)
1145 					inp->inp_ip_minttl = optval;
1146 				else
1147 					error = EINVAL;
1148 				break;
1149 #define	OPTSET(bit) \
1150 	if (optval) \
1151 		inpflags |= bit; \
1152 	else \
1153 		inpflags &= ~bit;
1154 
1155 			case IP_PKTINFO:
1156 				OPTSET(INP_PKTINFO);
1157 				break;
1158 
1159 			case IP_RECVOPTS:
1160 				OPTSET(INP_RECVOPTS);
1161 				break;
1162 
1163 			case IP_RECVPKTINFO:
1164 				OPTSET(INP_RECVPKTINFO);
1165 				break;
1166 
1167 			case IP_RECVRETOPTS:
1168 				OPTSET(INP_RECVRETOPTS);
1169 				break;
1170 
1171 			case IP_RECVDSTADDR:
1172 				OPTSET(INP_RECVDSTADDR);
1173 				break;
1174 
1175 			case IP_RECVIF:
1176 				OPTSET(INP_RECVIF);
1177 				break;
1178 
1179 			case IP_RECVTTL:
1180 				OPTSET(INP_RECVTTL);
1181 				break;
1182 			}
1183 		break;
1184 #undef OPTSET
1185 
1186 		case IP_MULTICAST_IF:
1187 		case IP_MULTICAST_TTL:
1188 		case IP_MULTICAST_LOOP:
1189 		case IP_ADD_MEMBERSHIP:
1190 		case IP_DROP_MEMBERSHIP:
1191 			error = ip_setmoptions(&inp->inp_moptions, sopt);
1192 			break;
1193 
1194 		case IP_PORTRANGE:
1195 			error = sockopt_getint(sopt, &optval);
1196 			if (error)
1197 				break;
1198 
1199 			switch (optval) {
1200 			case IP_PORTRANGE_DEFAULT:
1201 			case IP_PORTRANGE_HIGH:
1202 				inpflags &= ~(INP_LOWPORT);
1203 				break;
1204 
1205 			case IP_PORTRANGE_LOW:
1206 				inpflags |= INP_LOWPORT;
1207 				break;
1208 
1209 			default:
1210 				error = EINVAL;
1211 				break;
1212 			}
1213 			break;
1214 
1215 		case IP_PORTALGO:
1216 			error = sockopt_getint(sopt, &optval);
1217 			if (error)
1218 				break;
1219 
1220 			error = portalgo_algo_index_select(
1221 			    (struct inpcb_hdr *)inp, optval);
1222 			break;
1223 
1224 #if defined(IPSEC)
1225 		case IP_IPSEC_POLICY:
1226 			if (ipsec_enabled) {
1227 				error = ipsec4_set_policy(inp, sopt->sopt_name,
1228 				    sopt->sopt_data, sopt->sopt_size,
1229 				    curlwp->l_cred);
1230 				break;
1231 			}
1232 			/*FALLTHROUGH*/
1233 #endif /* IPSEC */
1234 
1235 		default:
1236 			error = ENOPROTOOPT;
1237 			break;
1238 		}
1239 		break;
1240 
1241 	case PRCO_GETOPT:
1242 		switch (sopt->sopt_name) {
1243 		case IP_OPTIONS:
1244 		case IP_RETOPTS: {
1245 			struct mbuf *mopts = inp->inp_options;
1246 
1247 			if (mopts) {
1248 				struct mbuf *m;
1249 
1250 				m = m_copym(mopts, 0, M_COPYALL, M_DONTWAIT);
1251 				if (m == NULL) {
1252 					error = ENOBUFS;
1253 					break;
1254 				}
1255 				error = sockopt_setmbuf(sopt, m);
1256 			}
1257 			break;
1258 		}
1259 		case IP_PKTINFO:
1260 		case IP_TOS:
1261 		case IP_TTL:
1262 		case IP_MINTTL:
1263 		case IP_RECVOPTS:
1264 		case IP_RECVRETOPTS:
1265 		case IP_RECVDSTADDR:
1266 		case IP_RECVIF:
1267 		case IP_RECVPKTINFO:
1268 		case IP_RECVTTL:
1269 		case IP_ERRORMTU:
1270 			switch (sopt->sopt_name) {
1271 			case IP_TOS:
1272 				optval = ip->ip_tos;
1273 				break;
1274 
1275 			case IP_TTL:
1276 				optval = ip->ip_ttl;
1277 				break;
1278 
1279 			case IP_MINTTL:
1280 				optval = inp->inp_ip_minttl;
1281 				break;
1282 
1283 			case IP_ERRORMTU:
1284 				optval = inp->inp_errormtu;
1285 				break;
1286 
1287 #define	OPTBIT(bit)	(inpflags & bit ? 1 : 0)
1288 
1289 			case IP_PKTINFO:
1290 				optval = OPTBIT(INP_PKTINFO);
1291 				break;
1292 
1293 			case IP_RECVOPTS:
1294 				optval = OPTBIT(INP_RECVOPTS);
1295 				break;
1296 
1297 			case IP_RECVPKTINFO:
1298 				optval = OPTBIT(INP_RECVPKTINFO);
1299 				break;
1300 
1301 			case IP_RECVRETOPTS:
1302 				optval = OPTBIT(INP_RECVRETOPTS);
1303 				break;
1304 
1305 			case IP_RECVDSTADDR:
1306 				optval = OPTBIT(INP_RECVDSTADDR);
1307 				break;
1308 
1309 			case IP_RECVIF:
1310 				optval = OPTBIT(INP_RECVIF);
1311 				break;
1312 
1313 			case IP_RECVTTL:
1314 				optval = OPTBIT(INP_RECVTTL);
1315 				break;
1316 			}
1317 			error = sockopt_setint(sopt, optval);
1318 			break;
1319 
1320 #if 0	/* defined(IPSEC) */
1321 		case IP_IPSEC_POLICY:
1322 		{
1323 			struct mbuf *m = NULL;
1324 
1325 			/* XXX this will return EINVAL as sopt is empty */
1326 			error = ipsec4_get_policy(inp, sopt->sopt_data,
1327 			    sopt->sopt_size, &m);
1328 			if (error == 0)
1329 				error = sockopt_setmbuf(sopt, m);
1330 			break;
1331 		}
1332 #endif /*IPSEC*/
1333 
1334 		case IP_MULTICAST_IF:
1335 		case IP_MULTICAST_TTL:
1336 		case IP_MULTICAST_LOOP:
1337 		case IP_ADD_MEMBERSHIP:
1338 		case IP_DROP_MEMBERSHIP:
1339 			error = ip_getmoptions(inp->inp_moptions, sopt);
1340 			break;
1341 
1342 		case IP_PORTRANGE:
1343 			if (inpflags & INP_LOWPORT)
1344 				optval = IP_PORTRANGE_LOW;
1345 			else
1346 				optval = IP_PORTRANGE_DEFAULT;
1347 			error = sockopt_setint(sopt, optval);
1348 			break;
1349 
1350 		case IP_PORTALGO:
1351 			optval = inp->inp_portalgo;
1352 			error = sockopt_setint(sopt, optval);
1353 			break;
1354 
1355 		default:
1356 			error = ENOPROTOOPT;
1357 			break;
1358 		}
1359 		break;
1360 	}
1361 
1362 	if (!error) {
1363 		inp->inp_flags = inpflags;
1364 	}
1365 	return error;
1366 }
1367 
1368 /*
1369  * Set up IP options in pcb for insertion in output packets.
1370  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1371  * with destination address if source routed.
1372  */
1373 static int
1374 ip_pcbopts(struct inpcb *inp, const struct sockopt *sopt)
1375 {
1376 	struct mbuf *m;
1377 	const u_char *cp;
1378 	u_char *dp;
1379 	int cnt;
1380 
1381 	/* Turn off any old options. */
1382 	if (inp->inp_options) {
1383 		m_free(inp->inp_options);
1384 	}
1385 	inp->inp_options = NULL;
1386 	if ((cnt = sopt->sopt_size) == 0) {
1387 		/* Only turning off any previous options. */
1388 		return 0;
1389 	}
1390 	cp = sopt->sopt_data;
1391 
1392 #ifndef	__vax__
1393 	if (cnt % sizeof(int32_t))
1394 		return (EINVAL);
1395 #endif
1396 
1397 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1398 	if (m == NULL)
1399 		return (ENOBUFS);
1400 
1401 	dp = mtod(m, u_char *);
1402 	memset(dp, 0, sizeof(struct in_addr));
1403 	dp += sizeof(struct in_addr);
1404 	m->m_len = sizeof(struct in_addr);
1405 
1406 	/*
1407 	 * IP option list according to RFC791. Each option is of the form
1408 	 *
1409 	 *	[optval] [olen] [(olen - 2) data bytes]
1410 	 *
1411 	 * We validate the list and copy options to an mbuf for prepending
1412 	 * to data packets. The IP first-hop destination address will be
1413 	 * stored before actual options and is zero if unset.
1414 	 */
1415 	while (cnt > 0) {
1416 		uint8_t optval, olen, offset;
1417 
1418 		optval = cp[IPOPT_OPTVAL];
1419 
1420 		if (optval == IPOPT_EOL || optval == IPOPT_NOP) {
1421 			olen = 1;
1422 		} else {
1423 			if (cnt < IPOPT_OLEN + 1)
1424 				goto bad;
1425 
1426 			olen = cp[IPOPT_OLEN];
1427 			if (olen < IPOPT_OLEN + 1 || olen > cnt)
1428 				goto bad;
1429 		}
1430 
1431 		if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) {
1432 			/*
1433 			 * user process specifies route as:
1434 			 *	->A->B->C->D
1435 			 * D must be our final destination (but we can't
1436 			 * check that since we may not have connected yet).
1437 			 * A is first hop destination, which doesn't appear in
1438 			 * actual IP option, but is stored before the options.
1439 			 */
1440 			if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr))
1441 				goto bad;
1442 
1443 			offset = cp[IPOPT_OFFSET];
1444 			memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1,
1445 			    sizeof(struct in_addr));
1446 
1447 			cp += sizeof(struct in_addr);
1448 			cnt -= sizeof(struct in_addr);
1449 			olen -= sizeof(struct in_addr);
1450 
1451 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1452 				goto bad;
1453 
1454 			memcpy(dp, cp, olen);
1455 			dp[IPOPT_OPTVAL] = optval;
1456 			dp[IPOPT_OLEN] = olen;
1457 			dp[IPOPT_OFFSET] = offset;
1458 			break;
1459 		} else {
1460 			if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr))
1461 				goto bad;
1462 
1463 			memcpy(dp, cp, olen);
1464 			break;
1465 		}
1466 
1467 		dp += olen;
1468 		m->m_len += olen;
1469 
1470 		if (optval == IPOPT_EOL)
1471 			break;
1472 
1473 		cp += olen;
1474 		cnt -= olen;
1475 	}
1476 
1477 	inp->inp_options = m;
1478 	return 0;
1479 bad:
1480 	(void)m_free(m);
1481 	return EINVAL;
1482 }
1483 
1484 /*
1485  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1486  */
1487 static struct ifnet *
1488 ip_multicast_if(struct in_addr *a, int *ifindexp)
1489 {
1490 	int ifindex;
1491 	struct ifnet *ifp = NULL;
1492 	struct in_ifaddr *ia;
1493 
1494 	if (ifindexp)
1495 		*ifindexp = 0;
1496 	if (ntohl(a->s_addr) >> 24 == 0) {
1497 		ifindex = ntohl(a->s_addr) & 0xffffff;
1498 		ifp = if_byindex(ifindex);
1499 		if (!ifp)
1500 			return NULL;
1501 		if (ifindexp)
1502 			*ifindexp = ifindex;
1503 	} else {
1504 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1505 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1506 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1507 				ifp = ia->ia_ifp;
1508 				break;
1509 			}
1510 		}
1511 	}
1512 	return ifp;
1513 }
1514 
1515 static int
1516 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval)
1517 {
1518 	u_int tval;
1519 	u_char cval;
1520 	int error;
1521 
1522 	if (sopt == NULL)
1523 		return EINVAL;
1524 
1525 	switch (sopt->sopt_size) {
1526 	case sizeof(u_char):
1527 		error = sockopt_get(sopt, &cval, sizeof(u_char));
1528 		tval = cval;
1529 		break;
1530 
1531 	case sizeof(u_int):
1532 		error = sockopt_get(sopt, &tval, sizeof(u_int));
1533 		break;
1534 
1535 	default:
1536 		error = EINVAL;
1537 	}
1538 
1539 	if (error)
1540 		return error;
1541 
1542 	if (tval > maxval)
1543 		return EINVAL;
1544 
1545 	*val = tval;
1546 	return 0;
1547 }
1548 
1549 static int
1550 ip_get_membership(const struct sockopt *sopt, struct ifnet **ifp,
1551     struct in_addr *ia, bool add)
1552 {
1553 	int error;
1554 	struct ip_mreq mreq;
1555 
1556 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
1557 	if (error)
1558 		return error;
1559 
1560 	if (!IN_MULTICAST(mreq.imr_multiaddr.s_addr))
1561 		return EINVAL;
1562 
1563 	memcpy(ia, &mreq.imr_multiaddr, sizeof(*ia));
1564 
1565 	if (in_nullhost(mreq.imr_interface)) {
1566 		union {
1567 			struct sockaddr		dst;
1568 			struct sockaddr_in	dst4;
1569 		} u;
1570 		struct route ro;
1571 
1572 		if (!add) {
1573 			*ifp = NULL;
1574 			return 0;
1575 		}
1576 		/*
1577 		 * If no interface address was provided, use the interface of
1578 		 * the route to the given multicast address.
1579 		 */
1580 		struct rtentry *rt;
1581 		memset(&ro, 0, sizeof(ro));
1582 
1583 		sockaddr_in_init(&u.dst4, ia, 0);
1584 		error = rtcache_setdst(&ro, &u.dst);
1585 		if (error != 0)
1586 			return error;
1587 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
1588 		rtcache_free(&ro);
1589 	} else {
1590 		*ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1591 		if (!add && *ifp == NULL)
1592 			return EADDRNOTAVAIL;
1593 	}
1594 	return 0;
1595 }
1596 
1597 /*
1598  * Add a multicast group membership.
1599  * Group must be a valid IP multicast address.
1600  */
1601 static int
1602 ip_add_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1603 {
1604 	struct ifnet *ifp;
1605 	struct in_addr ia;
1606 	int i, error;
1607 
1608 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1609 		error = ip_get_membership(sopt, &ifp, &ia, true);
1610 	else
1611 #ifdef INET6
1612 		error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia));
1613 #else
1614 		return EINVAL;
1615 #endif
1616 
1617 	if (error)
1618 		return error;
1619 
1620 	/*
1621 	 * See if we found an interface, and confirm that it
1622 	 * supports multicast.
1623 	 */
1624 	if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0)
1625 		return EADDRNOTAVAIL;
1626 
1627 	/*
1628 	 * See if the membership already exists or if all the
1629 	 * membership slots are full.
1630 	 */
1631 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1632 		if (imo->imo_membership[i]->inm_ifp == ifp &&
1633 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1634 			break;
1635 	}
1636 	if (i < imo->imo_num_memberships)
1637 		return EADDRINUSE;
1638 
1639 	if (i == IP_MAX_MEMBERSHIPS)
1640 		return ETOOMANYREFS;
1641 
1642 	/*
1643 	 * Everything looks good; add a new record to the multicast
1644 	 * address list for the given interface.
1645 	 */
1646 	if ((imo->imo_membership[i] = in_addmulti(&ia, ifp)) == NULL)
1647 		return ENOBUFS;
1648 
1649 	++imo->imo_num_memberships;
1650 	return 0;
1651 }
1652 
1653 /*
1654  * Drop a multicast group membership.
1655  * Group must be a valid IP multicast address.
1656  */
1657 static int
1658 ip_drop_membership(struct ip_moptions *imo, const struct sockopt *sopt)
1659 {
1660 	struct in_addr ia;
1661 	struct ifnet *ifp;
1662 	int i, error;
1663 
1664 	if (sopt->sopt_size == sizeof(struct ip_mreq))
1665 		error = ip_get_membership(sopt, &ifp, &ia, false);
1666 	else
1667 #ifdef INET6
1668 		error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia));
1669 #else
1670 		return EINVAL;
1671 #endif
1672 
1673 	if (error)
1674 		return error;
1675 
1676 	/*
1677 	 * Find the membership in the membership array.
1678 	 */
1679 	for (i = 0; i < imo->imo_num_memberships; ++i) {
1680 		if ((ifp == NULL ||
1681 		     imo->imo_membership[i]->inm_ifp == ifp) &&
1682 		    in_hosteq(imo->imo_membership[i]->inm_addr, ia))
1683 			break;
1684 	}
1685 	if (i == imo->imo_num_memberships)
1686 		return EADDRNOTAVAIL;
1687 
1688 	/*
1689 	 * Give up the multicast address record to which the
1690 	 * membership points.
1691 	 */
1692 	in_delmulti(imo->imo_membership[i]);
1693 
1694 	/*
1695 	 * Remove the gap in the membership array.
1696 	 */
1697 	for (++i; i < imo->imo_num_memberships; ++i)
1698 		imo->imo_membership[i-1] = imo->imo_membership[i];
1699 	--imo->imo_num_memberships;
1700 	return 0;
1701 }
1702 
1703 /*
1704  * Set the IP multicast options in response to user setsockopt().
1705  */
1706 int
1707 ip_setmoptions(struct ip_moptions **pimo, const struct sockopt *sopt)
1708 {
1709 	struct ip_moptions *imo = *pimo;
1710 	struct in_addr addr;
1711 	struct ifnet *ifp;
1712 	int ifindex, error = 0;
1713 
1714 	if (!imo) {
1715 		/*
1716 		 * No multicast option buffer attached to the pcb;
1717 		 * allocate one and initialize to default values.
1718 		 */
1719 		imo = kmem_intr_alloc(sizeof(*imo), KM_NOSLEEP);
1720 		if (imo == NULL)
1721 			return ENOBUFS;
1722 
1723 		imo->imo_multicast_ifp = NULL;
1724 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1725 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1726 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1727 		imo->imo_num_memberships = 0;
1728 		*pimo = imo;
1729 	}
1730 
1731 	switch (sopt->sopt_name) {
1732 	case IP_MULTICAST_IF:
1733 		/*
1734 		 * Select the interface for outgoing multicast packets.
1735 		 */
1736 		error = sockopt_get(sopt, &addr, sizeof(addr));
1737 		if (error)
1738 			break;
1739 
1740 		/*
1741 		 * INADDR_ANY is used to remove a previous selection.
1742 		 * When no interface is selected, a default one is
1743 		 * chosen every time a multicast packet is sent.
1744 		 */
1745 		if (in_nullhost(addr)) {
1746 			imo->imo_multicast_ifp = NULL;
1747 			break;
1748 		}
1749 		/*
1750 		 * The selected interface is identified by its local
1751 		 * IP address.  Find the interface and confirm that
1752 		 * it supports multicasting.
1753 		 */
1754 		ifp = ip_multicast_if(&addr, &ifindex);
1755 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1756 			error = EADDRNOTAVAIL;
1757 			break;
1758 		}
1759 		imo->imo_multicast_ifp = ifp;
1760 		if (ifindex)
1761 			imo->imo_multicast_addr = addr;
1762 		else
1763 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1764 		break;
1765 
1766 	case IP_MULTICAST_TTL:
1767 		/*
1768 		 * Set the IP time-to-live for outgoing multicast packets.
1769 		 */
1770 		error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL);
1771 		break;
1772 
1773 	case IP_MULTICAST_LOOP:
1774 		/*
1775 		 * Set the loopback flag for outgoing multicast packets.
1776 		 * Must be zero or one.
1777 		 */
1778 		error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1);
1779 		break;
1780 
1781 	case IP_ADD_MEMBERSHIP: /* IPV6_JOIN_GROUP */
1782 		error = ip_add_membership(imo, sopt);
1783 		break;
1784 
1785 	case IP_DROP_MEMBERSHIP: /* IPV6_LEAVE_GROUP */
1786 		error = ip_drop_membership(imo, sopt);
1787 		break;
1788 
1789 	default:
1790 		error = EOPNOTSUPP;
1791 		break;
1792 	}
1793 
1794 	/*
1795 	 * If all options have default values, no need to keep the mbuf.
1796 	 */
1797 	if (imo->imo_multicast_ifp == NULL &&
1798 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1799 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1800 	    imo->imo_num_memberships == 0) {
1801 		kmem_free(imo, sizeof(*imo));
1802 		*pimo = NULL;
1803 	}
1804 
1805 	return error;
1806 }
1807 
1808 /*
1809  * Return the IP multicast options in response to user getsockopt().
1810  */
1811 int
1812 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt)
1813 {
1814 	struct in_addr addr;
1815 	struct in_ifaddr *ia;
1816 	uint8_t optval;
1817 	int error = 0;
1818 
1819 	switch (sopt->sopt_name) {
1820 	case IP_MULTICAST_IF:
1821 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1822 			addr = zeroin_addr;
1823 		else if (imo->imo_multicast_addr.s_addr) {
1824 			/* return the value user has set */
1825 			addr = imo->imo_multicast_addr;
1826 		} else {
1827 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1828 			addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1829 		}
1830 		error = sockopt_set(sopt, &addr, sizeof(addr));
1831 		break;
1832 
1833 	case IP_MULTICAST_TTL:
1834 		optval = imo ? imo->imo_multicast_ttl
1835 		    : IP_DEFAULT_MULTICAST_TTL;
1836 
1837 		error = sockopt_set(sopt, &optval, sizeof(optval));
1838 		break;
1839 
1840 	case IP_MULTICAST_LOOP:
1841 		optval = imo ? imo->imo_multicast_loop
1842 		    : IP_DEFAULT_MULTICAST_LOOP;
1843 
1844 		error = sockopt_set(sopt, &optval, sizeof(optval));
1845 		break;
1846 
1847 	default:
1848 		error = EOPNOTSUPP;
1849 	}
1850 
1851 	return error;
1852 }
1853 
1854 /*
1855  * Discard the IP multicast options.
1856  */
1857 void
1858 ip_freemoptions(struct ip_moptions *imo)
1859 {
1860 	int i;
1861 
1862 	if (imo != NULL) {
1863 		for (i = 0; i < imo->imo_num_memberships; ++i)
1864 			in_delmulti(imo->imo_membership[i]);
1865 		kmem_free(imo, sizeof(*imo));
1866 	}
1867 }
1868 
1869 /*
1870  * Routine called from ip_output() to loop back a copy of an IP multicast
1871  * packet to the input queue of a specified interface.  Note that this
1872  * calls the output routine of the loopback "driver", but with an interface
1873  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
1874  */
1875 static void
1876 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst)
1877 {
1878 	struct ip *ip;
1879 	struct mbuf *copym;
1880 
1881 	copym = m_copypacket(m, M_DONTWAIT);
1882 	if (copym != NULL &&
1883 	    (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1884 		copym = m_pullup(copym, sizeof(struct ip));
1885 	if (copym == NULL)
1886 		return;
1887 	/*
1888 	 * We don't bother to fragment if the IP length is greater
1889 	 * than the interface's MTU.  Can this possibly matter?
1890 	 */
1891 	ip = mtod(copym, struct ip *);
1892 
1893 	if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1894 		in_delayed_cksum(copym);
1895 		copym->m_pkthdr.csum_flags &=
1896 		    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1897 	}
1898 
1899 	ip->ip_sum = 0;
1900 	ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1901 #ifndef NET_MPSAFE
1902 	KERNEL_LOCK(1, NULL);
1903 #endif
1904 	(void)looutput(ifp, copym, sintocsa(dst), NULL);
1905 #ifndef NET_MPSAFE
1906 	KERNEL_UNLOCK_ONE(NULL);
1907 #endif
1908 }
1909