xref: /netbsd-src/sys/netinet/ip_output.c (revision fd5cb0acea84d278e04e640d37ca2398f894991f)
1 /*	$NetBSD: ip_output.c,v 1.140 2005/02/03 23:13:20 perry Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1988, 1990, 1993
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.140 2005/02/03 23:13:20 perry Exp $");
102 
103 #include "opt_pfil_hooks.h"
104 #include "opt_inet.h"
105 #include "opt_ipsec.h"
106 #include "opt_mrouting.h"
107 
108 #include <sys/param.h>
109 #include <sys/malloc.h>
110 #include <sys/mbuf.h>
111 #include <sys/errno.h>
112 #include <sys/protosw.h>
113 #include <sys/socket.h>
114 #include <sys/socketvar.h>
115 #ifdef FAST_IPSEC
116 #include <sys/domain.h>
117 #endif
118 #include <sys/systm.h>
119 #include <sys/proc.h>
120 
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip_var.h>
131 
132 #ifdef MROUTING
133 #include <netinet/ip_mroute.h>
134 #endif
135 
136 #include <machine/stdarg.h>
137 
138 #ifdef IPSEC
139 #include <netinet6/ipsec.h>
140 #include <netkey/key.h>
141 #include <netkey/key_debug.h>
142 #endif /*IPSEC*/
143 
144 #ifdef FAST_IPSEC
145 #include <netipsec/ipsec.h>
146 #include <netipsec/key.h>
147 #include <netipsec/xform.h>
148 #endif	/* FAST_IPSEC*/
149 
150 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
151 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
152 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
153 
154 #ifdef PFIL_HOOKS
155 extern struct pfil_head inet_pfil_hook;			/* XXX */
156 #endif
157 
158 /*
159  * IP output.  The packet in mbuf chain m contains a skeletal IP
160  * header (with len, off, ttl, proto, tos, src, dst).
161  * The mbuf chain containing the packet will be freed.
162  * The mbuf opt, if present, will not be freed.
163  */
164 int
165 ip_output(struct mbuf *m0, ...)
166 {
167 	struct ip *ip;
168 	struct ifnet *ifp;
169 	struct mbuf *m = m0;
170 	int hlen = sizeof (struct ip);
171 	int len, error = 0;
172 	struct route iproute;
173 	struct sockaddr_in *dst;
174 	struct in_ifaddr *ia;
175 	struct mbuf *opt;
176 	struct route *ro;
177 	int flags, sw_csum;
178 	int *mtu_p;
179 	u_long mtu;
180 	struct ip_moptions *imo;
181 	struct socket *so;
182 	va_list ap;
183 #ifdef IPSEC
184 	struct secpolicy *sp = NULL;
185 #endif /*IPSEC*/
186 #ifdef FAST_IPSEC
187 	struct inpcb *inp;
188 	struct m_tag *mtag;
189 	struct secpolicy *sp = NULL;
190 	struct tdb_ident *tdbi;
191 	int s;
192 #endif
193 	u_int16_t ip_len;
194 
195 	len = 0;
196 	va_start(ap, m0);
197 	opt = va_arg(ap, struct mbuf *);
198 	ro = va_arg(ap, struct route *);
199 	flags = va_arg(ap, int);
200 	imo = va_arg(ap, struct ip_moptions *);
201 	so = va_arg(ap, struct socket *);
202 	if (flags & IP_RETURNMTU)
203 		mtu_p = va_arg(ap, int *);
204 	else
205 		mtu_p = NULL;
206 	va_end(ap);
207 
208 	MCLAIM(m, &ip_tx_mowner);
209 #ifdef FAST_IPSEC
210 	if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET)
211 		inp = (struct inpcb *)so->so_pcb;
212 	else
213 		inp = NULL;
214 #endif /* FAST_IPSEC */
215 
216 #ifdef	DIAGNOSTIC
217 	if ((m->m_flags & M_PKTHDR) == 0)
218 		panic("ip_output no HDR");
219 #endif
220 	if (opt) {
221 		m = ip_insertoptions(m, opt, &len);
222 		if (len >= sizeof(struct ip))
223 			hlen = len;
224 	}
225 	ip = mtod(m, struct ip *);
226 	/*
227 	 * Fill in IP header.
228 	 */
229 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
230 		ip->ip_v = IPVERSION;
231 		ip->ip_off = htons(0);
232 		ip->ip_id = ip_newid();
233 		ip->ip_hl = hlen >> 2;
234 		ipstat.ips_localout++;
235 	} else {
236 		hlen = ip->ip_hl << 2;
237 	}
238 	/*
239 	 * Route packet.
240 	 */
241 	if (ro == 0) {
242 		ro = &iproute;
243 		bzero((caddr_t)ro, sizeof (*ro));
244 	}
245 	dst = satosin(&ro->ro_dst);
246 	/*
247 	 * If there is a cached route,
248 	 * check that it is to the same destination
249 	 * and is still up.  If not, free it and try again.
250 	 * The address family should also be checked in case of sharing the
251 	 * cache with IPv6.
252 	 */
253 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
254 	    dst->sin_family != AF_INET ||
255 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
256 		RTFREE(ro->ro_rt);
257 		ro->ro_rt = (struct rtentry *)0;
258 	}
259 	if (ro->ro_rt == 0) {
260 		bzero(dst, sizeof(*dst));
261 		dst->sin_family = AF_INET;
262 		dst->sin_len = sizeof(*dst);
263 		dst->sin_addr = ip->ip_dst;
264 	}
265 	/*
266 	 * If routing to interface only,
267 	 * short circuit routing lookup.
268 	 */
269 	if (flags & IP_ROUTETOIF) {
270 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
271 			ipstat.ips_noroute++;
272 			error = ENETUNREACH;
273 			goto bad;
274 		}
275 		ifp = ia->ia_ifp;
276 		mtu = ifp->if_mtu;
277 		ip->ip_ttl = 1;
278 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
279 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
280 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
281 		ifp = imo->imo_multicast_ifp;
282 		mtu = ifp->if_mtu;
283 		IFP_TO_IA(ifp, ia);
284 	} else {
285 		if (ro->ro_rt == 0)
286 			rtalloc(ro);
287 		if (ro->ro_rt == 0) {
288 			ipstat.ips_noroute++;
289 			error = EHOSTUNREACH;
290 			goto bad;
291 		}
292 		ia = ifatoia(ro->ro_rt->rt_ifa);
293 		ifp = ro->ro_rt->rt_ifp;
294 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
295 			mtu = ifp->if_mtu;
296 		ro->ro_rt->rt_use++;
297 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
298 			dst = satosin(ro->ro_rt->rt_gateway);
299 	}
300 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
301 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
302 		struct in_multi *inm;
303 
304 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
305 			M_BCAST : M_MCAST;
306 		/*
307 		 * IP destination address is multicast.  Make sure "dst"
308 		 * still points to the address in "ro".  (It may have been
309 		 * changed to point to a gateway address, above.)
310 		 */
311 		dst = satosin(&ro->ro_dst);
312 		/*
313 		 * See if the caller provided any multicast options
314 		 */
315 		if (imo != NULL)
316 			ip->ip_ttl = imo->imo_multicast_ttl;
317 		else
318 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
319 
320 		/*
321 		 * if we don't know the outgoing ifp yet, we can't generate
322 		 * output
323 		 */
324 		if (!ifp) {
325 			ipstat.ips_noroute++;
326 			error = ENETUNREACH;
327 			goto bad;
328 		}
329 
330 		/*
331 		 * If the packet is multicast or broadcast, confirm that
332 		 * the outgoing interface can transmit it.
333 		 */
334 		if (((m->m_flags & M_MCAST) &&
335 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
336 		    ((m->m_flags & M_BCAST) &&
337 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
338 			ipstat.ips_noroute++;
339 			error = ENETUNREACH;
340 			goto bad;
341 		}
342 		/*
343 		 * If source address not specified yet, use an address
344 		 * of outgoing interface.
345 		 */
346 		if (in_nullhost(ip->ip_src)) {
347 			struct in_ifaddr *ia;
348 
349 			IFP_TO_IA(ifp, ia);
350 			if (!ia) {
351 				error = EADDRNOTAVAIL;
352 				goto bad;
353 			}
354 			ip->ip_src = ia->ia_addr.sin_addr;
355 		}
356 
357 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
358 		if (inm != NULL &&
359 		   (imo == NULL || imo->imo_multicast_loop)) {
360 			/*
361 			 * If we belong to the destination multicast group
362 			 * on the outgoing interface, and the caller did not
363 			 * forbid loopback, loop back a copy.
364 			 */
365 			ip_mloopback(ifp, m, dst);
366 		}
367 #ifdef MROUTING
368 		else {
369 			/*
370 			 * If we are acting as a multicast router, perform
371 			 * multicast forwarding as if the packet had just
372 			 * arrived on the interface to which we are about
373 			 * to send.  The multicast forwarding function
374 			 * recursively calls this function, using the
375 			 * IP_FORWARDING flag to prevent infinite recursion.
376 			 *
377 			 * Multicasts that are looped back by ip_mloopback(),
378 			 * above, will be forwarded by the ip_input() routine,
379 			 * if necessary.
380 			 */
381 			extern struct socket *ip_mrouter;
382 
383 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
384 				if (ip_mforward(m, ifp) != 0) {
385 					m_freem(m);
386 					goto done;
387 				}
388 			}
389 		}
390 #endif
391 		/*
392 		 * Multicasts with a time-to-live of zero may be looped-
393 		 * back, above, but must not be transmitted on a network.
394 		 * Also, multicasts addressed to the loopback interface
395 		 * are not sent -- the above call to ip_mloopback() will
396 		 * loop back a copy if this host actually belongs to the
397 		 * destination group on the loopback interface.
398 		 */
399 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
400 			m_freem(m);
401 			goto done;
402 		}
403 
404 		goto sendit;
405 	}
406 #ifndef notdef
407 	/*
408 	 * If source address not specified yet, use address
409 	 * of outgoing interface.
410 	 */
411 	if (in_nullhost(ip->ip_src))
412 		ip->ip_src = ia->ia_addr.sin_addr;
413 #endif
414 
415 	/*
416 	 * packets with Class-D address as source are not valid per
417 	 * RFC 1112
418 	 */
419 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
420 		ipstat.ips_odropped++;
421 		error = EADDRNOTAVAIL;
422 		goto bad;
423 	}
424 
425 	/*
426 	 * Look for broadcast address and
427 	 * and verify user is allowed to send
428 	 * such a packet.
429 	 */
430 	if (in_broadcast(dst->sin_addr, ifp)) {
431 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
432 			error = EADDRNOTAVAIL;
433 			goto bad;
434 		}
435 		if ((flags & IP_ALLOWBROADCAST) == 0) {
436 			error = EACCES;
437 			goto bad;
438 		}
439 		/* don't allow broadcast messages to be fragmented */
440 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
441 			error = EMSGSIZE;
442 			goto bad;
443 		}
444 		m->m_flags |= M_BCAST;
445 	} else
446 		m->m_flags &= ~M_BCAST;
447 
448 sendit:
449 	/*
450 	 * If we're doing Path MTU Discovery, we need to set DF unless
451 	 * the route's MTU is locked.
452 	 */
453 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
454 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
455 		ip->ip_off |= htons(IP_DF);
456 
457 	/* Remember the current ip_len */
458 	ip_len = ntohs(ip->ip_len);
459 
460 #ifdef IPSEC
461 	/* get SP for this packet */
462 	if (so == NULL)
463 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
464 		    flags, &error);
465 	else {
466 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
467 					 IPSEC_DIR_OUTBOUND))
468 			goto skip_ipsec;
469 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
470 	}
471 
472 	if (sp == NULL) {
473 		ipsecstat.out_inval++;
474 		goto bad;
475 	}
476 
477 	error = 0;
478 
479 	/* check policy */
480 	switch (sp->policy) {
481 	case IPSEC_POLICY_DISCARD:
482 		/*
483 		 * This packet is just discarded.
484 		 */
485 		ipsecstat.out_polvio++;
486 		goto bad;
487 
488 	case IPSEC_POLICY_BYPASS:
489 	case IPSEC_POLICY_NONE:
490 		/* no need to do IPsec. */
491 		goto skip_ipsec;
492 
493 	case IPSEC_POLICY_IPSEC:
494 		if (sp->req == NULL) {
495 			/* XXX should be panic ? */
496 			printf("ip_output: No IPsec request specified.\n");
497 			error = EINVAL;
498 			goto bad;
499 		}
500 		break;
501 
502 	case IPSEC_POLICY_ENTRUST:
503 	default:
504 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
505 	}
506 
507 	/*
508 	 * ipsec4_output() expects ip_len and ip_off in network
509 	 * order.  They have been set to network order above.
510 	 */
511 
512     {
513 	struct ipsec_output_state state;
514 	bzero(&state, sizeof(state));
515 	state.m = m;
516 	if (flags & IP_ROUTETOIF) {
517 		state.ro = &iproute;
518 		bzero(&iproute, sizeof(iproute));
519 	} else
520 		state.ro = ro;
521 	state.dst = (struct sockaddr *)dst;
522 
523 	/*
524 	 * We can't defer the checksum of payload data if
525 	 * we're about to encrypt/authenticate it.
526 	 *
527 	 * XXX When we support crypto offloading functions of
528 	 * XXX network interfaces, we need to reconsider this,
529 	 * XXX since it's likely that they'll support checksumming,
530 	 * XXX as well.
531 	 */
532 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
533 		in_delayed_cksum(m);
534 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
535 	}
536 
537 	error = ipsec4_output(&state, sp, flags);
538 
539 	m = state.m;
540 	if (flags & IP_ROUTETOIF) {
541 		/*
542 		 * if we have tunnel mode SA, we may need to ignore
543 		 * IP_ROUTETOIF.
544 		 */
545 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
546 			flags &= ~IP_ROUTETOIF;
547 			ro = state.ro;
548 		}
549 	} else
550 		ro = state.ro;
551 	dst = (struct sockaddr_in *)state.dst;
552 	if (error) {
553 		/* mbuf is already reclaimed in ipsec4_output. */
554 		m0 = NULL;
555 		switch (error) {
556 		case EHOSTUNREACH:
557 		case ENETUNREACH:
558 		case EMSGSIZE:
559 		case ENOBUFS:
560 		case ENOMEM:
561 			break;
562 		default:
563 			printf("ip4_output (ipsec): error code %d\n", error);
564 			/*fall through*/
565 		case ENOENT:
566 			/* don't show these error codes to the user */
567 			error = 0;
568 			break;
569 		}
570 		goto bad;
571 	}
572 
573 	/* be sure to update variables that are affected by ipsec4_output() */
574 	ip = mtod(m, struct ip *);
575 	hlen = ip->ip_hl << 2;
576 	ip_len = ntohs(ip->ip_len);
577 
578 	if (ro->ro_rt == NULL) {
579 		if ((flags & IP_ROUTETOIF) == 0) {
580 			printf("ip_output: "
581 				"can't update route after IPsec processing\n");
582 			error = EHOSTUNREACH;	/*XXX*/
583 			goto bad;
584 		}
585 	} else {
586 		/* nobody uses ia beyond here */
587 		if (state.encap) {
588 			ifp = ro->ro_rt->rt_ifp;
589 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
590 				mtu = ifp->if_mtu;
591 		}
592 	}
593     }
594 skip_ipsec:
595 #endif /*IPSEC*/
596 #ifdef FAST_IPSEC
597 	/*
598 	 * Check the security policy (SP) for the packet and, if
599 	 * required, do IPsec-related processing.  There are two
600 	 * cases here; the first time a packet is sent through
601 	 * it will be untagged and handled by ipsec4_checkpolicy.
602 	 * If the packet is resubmitted to ip_output (e.g. after
603 	 * AH, ESP, etc. processing), there will be a tag to bypass
604 	 * the lookup and related policy checking.
605 	 */
606 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
607 	s = splsoftnet();
608 	if (mtag != NULL) {
609 		tdbi = (struct tdb_ident *)(mtag + 1);
610 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
611 		if (sp == NULL)
612 			error = -EINVAL;	/* force silent drop */
613 		m_tag_delete(m, mtag);
614 	} else {
615 		if (inp != NULL &&
616 		    IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND))
617 			goto spd_done;
618 		sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
619 					&error, inp);
620 	}
621 	/*
622 	 * There are four return cases:
623 	 *    sp != NULL	 	    apply IPsec policy
624 	 *    sp == NULL, error == 0	    no IPsec handling needed
625 	 *    sp == NULL, error == -EINVAL  discard packet w/o error
626 	 *    sp == NULL, error != 0	    discard packet, report error
627 	 */
628 	if (sp != NULL) {
629 		/* Loop detection, check if ipsec processing already done */
630 		IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
631 		for (mtag = m_tag_first(m); mtag != NULL;
632 		     mtag = m_tag_next(m, mtag)) {
633 #ifdef MTAG_ABI_COMPAT
634 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
635 				continue;
636 #endif
637 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
638 			    mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
639 				continue;
640 			/*
641 			 * Check if policy has an SA associated with it.
642 			 * This can happen when an SP has yet to acquire
643 			 * an SA; e.g. on first reference.  If it occurs,
644 			 * then we let ipsec4_process_packet do its thing.
645 			 */
646 			if (sp->req->sav == NULL)
647 				break;
648 			tdbi = (struct tdb_ident *)(mtag + 1);
649 			if (tdbi->spi == sp->req->sav->spi &&
650 			    tdbi->proto == sp->req->sav->sah->saidx.proto &&
651 			    bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
652 				 sizeof (union sockaddr_union)) == 0) {
653 				/*
654 				 * No IPsec processing is needed, free
655 				 * reference to SP.
656 				 *
657 				 * NB: null pointer to avoid free at
658 				 *     done: below.
659 				 */
660 				KEY_FREESP(&sp), sp = NULL;
661 				splx(s);
662 				goto spd_done;
663 			}
664 		}
665 
666 		/*
667 		 * Do delayed checksums now because we send before
668 		 * this is done in the normal processing path.
669 		 */
670 		if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
671 			in_delayed_cksum(m);
672 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
673 		}
674 
675 #ifdef __FreeBSD__
676 		ip->ip_len = htons(ip->ip_len);
677 		ip->ip_off = htons(ip->ip_off);
678 #endif
679 
680 		/* NB: callee frees mbuf */
681 		error = ipsec4_process_packet(m, sp->req, flags, 0);
682 		/*
683 		 * Preserve KAME behaviour: ENOENT can be returned
684 		 * when an SA acquire is in progress.  Don't propagate
685 		 * this to user-level; it confuses applications.
686 		 *
687 		 * XXX this will go away when the SADB is redone.
688 		 */
689 		if (error == ENOENT)
690 			error = 0;
691 		splx(s);
692 		goto done;
693 	} else {
694 		splx(s);
695 
696 		if (error != 0) {
697 			/*
698 			 * Hack: -EINVAL is used to signal that a packet
699 			 * should be silently discarded.  This is typically
700 			 * because we asked key management for an SA and
701 			 * it was delayed (e.g. kicked up to IKE).
702 			 */
703 			if (error == -EINVAL)
704 				error = 0;
705 			goto bad;
706 		} else {
707 			/* No IPsec processing for this packet. */
708 		}
709 #ifdef notyet
710 		/*
711 		 * If deferred crypto processing is needed, check that
712 		 * the interface supports it.
713 		 */
714 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
715 		if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
716 			/* notify IPsec to do its own crypto */
717 			ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
718 			error = EHOSTUNREACH;
719 			goto bad;
720 		}
721 #endif
722 	}
723 spd_done:
724 #endif /* FAST_IPSEC */
725 
726 #ifdef PFIL_HOOKS
727 	/*
728 	 * Run through list of hooks for output packets.
729 	 */
730 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
731 		goto done;
732 	if (m == NULL)
733 		goto done;
734 
735 	ip = mtod(m, struct ip *);
736 	hlen = ip->ip_hl << 2;
737 #endif /* PFIL_HOOKS */
738 
739 #if IFA_STATS
740 	/*
741 	 * search for the source address structure to
742 	 * maintain output statistics.
743 	 */
744 	INADDR_TO_IA(ip->ip_src, ia);
745 #endif
746 
747 	/* Maybe skip checksums on loopback interfaces. */
748 	if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
749 			   ip_do_loopback_cksum))
750 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
751 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
752 	/*
753 	 * If small enough for mtu of path, can just send directly.
754 	 */
755 	if (ip_len <= mtu) {
756 #if IFA_STATS
757 		if (ia)
758 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
759 #endif
760 		/*
761 		 * Always initialize the sum to 0!  Some HW assisted
762 		 * checksumming requires this.
763 		 */
764 		ip->ip_sum = 0;
765 
766 		/*
767 		 * Perform any checksums that the hardware can't do
768 		 * for us.
769 		 *
770 		 * XXX Does any hardware require the {th,uh}_sum
771 		 * XXX fields to be 0?
772 		 */
773 		if (sw_csum & M_CSUM_IPv4) {
774 			ip->ip_sum = in_cksum(m, hlen);
775 			m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
776 		}
777 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
778 			in_delayed_cksum(m);
779 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
780 		}
781 
782 #ifdef IPSEC
783 		/* clean ipsec history once it goes out of the node */
784 		ipsec_delaux(m);
785 #endif
786 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
787 		goto done;
788 	}
789 
790 	/*
791 	 * We can't use HW checksumming if we're about to
792 	 * to fragment the packet.
793 	 *
794 	 * XXX Some hardware can do this.
795 	 */
796 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
797 		in_delayed_cksum(m);
798 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
799 	}
800 
801 	/*
802 	 * Too large for interface; fragment if possible.
803 	 * Must be able to put at least 8 bytes per fragment.
804 	 */
805 	if (ntohs(ip->ip_off) & IP_DF) {
806 		if (flags & IP_RETURNMTU)
807 			*mtu_p = mtu;
808 		error = EMSGSIZE;
809 		ipstat.ips_cantfrag++;
810 		goto bad;
811 	}
812 
813 	error = ip_fragment(m, ifp, mtu);
814 	if (error) {
815 		m = NULL;
816 		goto bad;
817 	}
818 
819 	for (; m; m = m0) {
820 		m0 = m->m_nextpkt;
821 		m->m_nextpkt = 0;
822 		if (error == 0) {
823 #if IFA_STATS
824 			if (ia)
825 				ia->ia_ifa.ifa_data.ifad_outbytes +=
826 				    ntohs(ip->ip_len);
827 #endif
828 #ifdef IPSEC
829 			/* clean ipsec history once it goes out of the node */
830 			ipsec_delaux(m);
831 #endif
832 			KASSERT((m->m_pkthdr.csum_flags &
833 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
834 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
835 			    ro->ro_rt);
836 		} else
837 			m_freem(m);
838 	}
839 
840 	if (error == 0)
841 		ipstat.ips_fragmented++;
842 done:
843 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
844 		RTFREE(ro->ro_rt);
845 		ro->ro_rt = 0;
846 	}
847 
848 #ifdef IPSEC
849 	if (sp != NULL) {
850 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
851 			printf("DP ip_output call free SP:%p\n", sp));
852 		key_freesp(sp);
853 	}
854 #endif /* IPSEC */
855 #ifdef FAST_IPSEC
856 	if (sp != NULL)
857 		KEY_FREESP(&sp);
858 #endif /* FAST_IPSEC */
859 
860 	return (error);
861 bad:
862 	m_freem(m);
863 	goto done;
864 }
865 
866 int
867 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
868 {
869 	struct ip *ip, *mhip;
870 	struct mbuf *m0;
871 	int len, hlen, off;
872 	int mhlen, firstlen;
873 	struct mbuf **mnext;
874 	int sw_csum = m->m_pkthdr.csum_flags;
875 	int fragments = 0;
876 	int s;
877 	int error = 0;
878 
879 	ip = mtod(m, struct ip *);
880 	hlen = ip->ip_hl << 2;
881 	if (ifp != NULL)
882 		sw_csum &= ~ifp->if_csum_flags_tx;
883 
884 	len = (mtu - hlen) &~ 7;
885 	if (len < 8) {
886 		m_freem(m);
887 		return (EMSGSIZE);
888 	}
889 
890 	firstlen = len;
891 	mnext = &m->m_nextpkt;
892 
893 	/*
894 	 * Loop through length of segment after first fragment,
895 	 * make new header and copy data of each part and link onto chain.
896 	 */
897 	m0 = m;
898 	mhlen = sizeof (struct ip);
899 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
900 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
901 		if (m == 0) {
902 			error = ENOBUFS;
903 			ipstat.ips_odropped++;
904 			goto sendorfree;
905 		}
906 		MCLAIM(m, m0->m_owner);
907 		*mnext = m;
908 		mnext = &m->m_nextpkt;
909 		m->m_data += max_linkhdr;
910 		mhip = mtod(m, struct ip *);
911 		*mhip = *ip;
912 		/* we must inherit MCAST and BCAST flags */
913 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
914 		if (hlen > sizeof (struct ip)) {
915 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
916 			mhip->ip_hl = mhlen >> 2;
917 		}
918 		m->m_len = mhlen;
919 		mhip->ip_off = ((off - hlen) >> 3) +
920 		    (ntohs(ip->ip_off) & ~IP_MF);
921 		if (ip->ip_off & htons(IP_MF))
922 			mhip->ip_off |= IP_MF;
923 		if (off + len >= ntohs(ip->ip_len))
924 			len = ntohs(ip->ip_len) - off;
925 		else
926 			mhip->ip_off |= IP_MF;
927 		HTONS(mhip->ip_off);
928 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
929 		m->m_next = m_copy(m0, off, len);
930 		if (m->m_next == 0) {
931 			error = ENOBUFS;	/* ??? */
932 			ipstat.ips_odropped++;
933 			goto sendorfree;
934 		}
935 		m->m_pkthdr.len = mhlen + len;
936 		m->m_pkthdr.rcvif = (struct ifnet *)0;
937 		mhip->ip_sum = 0;
938 		if (sw_csum & M_CSUM_IPv4) {
939 			mhip->ip_sum = in_cksum(m, mhlen);
940 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
941 		} else {
942 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
943 		}
944 		ipstat.ips_ofragments++;
945 		fragments++;
946 	}
947 	/*
948 	 * Update first fragment by trimming what's been copied out
949 	 * and updating header, then send each fragment (in order).
950 	 */
951 	m = m0;
952 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
953 	m->m_pkthdr.len = hlen + firstlen;
954 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
955 	ip->ip_off |= htons(IP_MF);
956 	ip->ip_sum = 0;
957 	if (sw_csum & M_CSUM_IPv4) {
958 		ip->ip_sum = in_cksum(m, hlen);
959 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
960 	} else {
961 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
962 	}
963 sendorfree:
964 	/*
965 	 * If there is no room for all the fragments, don't queue
966 	 * any of them.
967 	 */
968 	if (ifp != NULL) {
969 		s = splnet();
970 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
971 		    error == 0) {
972 			error = ENOBUFS;
973 			ipstat.ips_odropped++;
974 			IFQ_INC_DROPS(&ifp->if_snd);
975 		}
976 		splx(s);
977 	}
978 	if (error) {
979 		for (m = m0; m; m = m0) {
980 			m0 = m->m_nextpkt;
981 			m->m_nextpkt = NULL;
982 			m_freem(m);
983 		}
984 	}
985 	return (error);
986 }
987 
988 /*
989  * Process a delayed payload checksum calculation.
990  */
991 void
992 in_delayed_cksum(struct mbuf *m)
993 {
994 	struct ip *ip;
995 	u_int16_t csum, offset;
996 
997 	ip = mtod(m, struct ip *);
998 	offset = ip->ip_hl << 2;
999 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
1000 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
1001 		csum = 0xffff;
1002 
1003 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1004 
1005 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
1006 		/* This happen when ip options were inserted
1007 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
1008 		    m->m_len, offset, ip->ip_p);
1009 		 */
1010 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
1011 	} else
1012 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1013 }
1014 
1015 /*
1016  * Determine the maximum length of the options to be inserted;
1017  * we would far rather allocate too much space rather than too little.
1018  */
1019 
1020 u_int
1021 ip_optlen(struct inpcb *inp)
1022 {
1023 	struct mbuf *m = inp->inp_options;
1024 
1025 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
1026 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1027 	else
1028 		return 0;
1029 }
1030 
1031 
1032 /*
1033  * Insert IP options into preformed packet.
1034  * Adjust IP destination as required for IP source routing,
1035  * as indicated by a non-zero in_addr at the start of the options.
1036  */
1037 static struct mbuf *
1038 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1039 {
1040 	struct ipoption *p = mtod(opt, struct ipoption *);
1041 	struct mbuf *n;
1042 	struct ip *ip = mtod(m, struct ip *);
1043 	unsigned optlen;
1044 
1045 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1046 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1047 		return (m);		/* XXX should fail */
1048 	if (!in_nullhost(p->ipopt_dst))
1049 		ip->ip_dst = p->ipopt_dst;
1050 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1051 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1052 		if (n == 0)
1053 			return (m);
1054 		MCLAIM(n, m->m_owner);
1055 		M_COPY_PKTHDR(n, m);
1056 		m_tag_delete_chain(m, NULL);
1057 		m->m_flags &= ~M_PKTHDR;
1058 		m->m_len -= sizeof(struct ip);
1059 		m->m_data += sizeof(struct ip);
1060 		n->m_next = m;
1061 		m = n;
1062 		m->m_len = optlen + sizeof(struct ip);
1063 		m->m_data += max_linkhdr;
1064 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1065 	} else {
1066 		m->m_data -= optlen;
1067 		m->m_len += optlen;
1068 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
1069 	}
1070 	m->m_pkthdr.len += optlen;
1071 	ip = mtod(m, struct ip *);
1072 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1073 	*phlen = sizeof(struct ip) + optlen;
1074 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1075 	return (m);
1076 }
1077 
1078 /*
1079  * Copy options from ip to jp,
1080  * omitting those not copied during fragmentation.
1081  */
1082 int
1083 ip_optcopy(struct ip *ip, struct ip *jp)
1084 {
1085 	u_char *cp, *dp;
1086 	int opt, optlen, cnt;
1087 
1088 	cp = (u_char *)(ip + 1);
1089 	dp = (u_char *)(jp + 1);
1090 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1091 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1092 		opt = cp[0];
1093 		if (opt == IPOPT_EOL)
1094 			break;
1095 		if (opt == IPOPT_NOP) {
1096 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1097 			*dp++ = IPOPT_NOP;
1098 			optlen = 1;
1099 			continue;
1100 		}
1101 #ifdef DIAGNOSTIC
1102 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1103 			panic("malformed IPv4 option passed to ip_optcopy");
1104 #endif
1105 		optlen = cp[IPOPT_OLEN];
1106 #ifdef DIAGNOSTIC
1107 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1108 			panic("malformed IPv4 option passed to ip_optcopy");
1109 #endif
1110 		/* bogus lengths should have been caught by ip_dooptions */
1111 		if (optlen > cnt)
1112 			optlen = cnt;
1113 		if (IPOPT_COPIED(opt)) {
1114 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1115 			dp += optlen;
1116 		}
1117 	}
1118 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1119 		*dp++ = IPOPT_EOL;
1120 	return (optlen);
1121 }
1122 
1123 /*
1124  * IP socket option processing.
1125  */
1126 int
1127 ip_ctloutput(int op, struct socket *so, int level, int optname,
1128     struct mbuf **mp)
1129 {
1130 	struct inpcb *inp = sotoinpcb(so);
1131 	struct mbuf *m = *mp;
1132 	int optval = 0;
1133 	int error = 0;
1134 #if defined(IPSEC) || defined(FAST_IPSEC)
1135 	struct proc *p = curproc;	/*XXX*/
1136 #endif
1137 
1138 	if (level != IPPROTO_IP) {
1139 		error = EINVAL;
1140 		if (op == PRCO_SETOPT && *mp)
1141 			(void) m_free(*mp);
1142 	} else switch (op) {
1143 
1144 	case PRCO_SETOPT:
1145 		switch (optname) {
1146 		case IP_OPTIONS:
1147 #ifdef notyet
1148 		case IP_RETOPTS:
1149 			return (ip_pcbopts(optname, &inp->inp_options, m));
1150 #else
1151 			return (ip_pcbopts(&inp->inp_options, m));
1152 #endif
1153 
1154 		case IP_TOS:
1155 		case IP_TTL:
1156 		case IP_RECVOPTS:
1157 		case IP_RECVRETOPTS:
1158 		case IP_RECVDSTADDR:
1159 		case IP_RECVIF:
1160 			if (m == NULL || m->m_len != sizeof(int))
1161 				error = EINVAL;
1162 			else {
1163 				optval = *mtod(m, int *);
1164 				switch (optname) {
1165 
1166 				case IP_TOS:
1167 					inp->inp_ip.ip_tos = optval;
1168 					break;
1169 
1170 				case IP_TTL:
1171 					inp->inp_ip.ip_ttl = optval;
1172 					break;
1173 #define	OPTSET(bit) \
1174 	if (optval) \
1175 		inp->inp_flags |= bit; \
1176 	else \
1177 		inp->inp_flags &= ~bit;
1178 
1179 				case IP_RECVOPTS:
1180 					OPTSET(INP_RECVOPTS);
1181 					break;
1182 
1183 				case IP_RECVRETOPTS:
1184 					OPTSET(INP_RECVRETOPTS);
1185 					break;
1186 
1187 				case IP_RECVDSTADDR:
1188 					OPTSET(INP_RECVDSTADDR);
1189 					break;
1190 
1191 				case IP_RECVIF:
1192 					OPTSET(INP_RECVIF);
1193 					break;
1194 				}
1195 			}
1196 			break;
1197 #undef OPTSET
1198 
1199 		case IP_MULTICAST_IF:
1200 		case IP_MULTICAST_TTL:
1201 		case IP_MULTICAST_LOOP:
1202 		case IP_ADD_MEMBERSHIP:
1203 		case IP_DROP_MEMBERSHIP:
1204 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1205 			break;
1206 
1207 		case IP_PORTRANGE:
1208 			if (m == 0 || m->m_len != sizeof(int))
1209 				error = EINVAL;
1210 			else {
1211 				optval = *mtod(m, int *);
1212 
1213 				switch (optval) {
1214 
1215 				case IP_PORTRANGE_DEFAULT:
1216 				case IP_PORTRANGE_HIGH:
1217 					inp->inp_flags &= ~(INP_LOWPORT);
1218 					break;
1219 
1220 				case IP_PORTRANGE_LOW:
1221 					inp->inp_flags |= INP_LOWPORT;
1222 					break;
1223 
1224 				default:
1225 					error = EINVAL;
1226 					break;
1227 				}
1228 			}
1229 			break;
1230 
1231 #if defined(IPSEC) || defined(FAST_IPSEC)
1232 		case IP_IPSEC_POLICY:
1233 		{
1234 			caddr_t req = NULL;
1235 			size_t len = 0;
1236 			int priv = 0;
1237 
1238 #ifdef __NetBSD__
1239 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1240 				priv = 0;
1241 			else
1242 				priv = 1;
1243 #else
1244 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1245 #endif
1246 			if (m) {
1247 				req = mtod(m, caddr_t);
1248 				len = m->m_len;
1249 			}
1250 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1251 			break;
1252 		    }
1253 #endif /*IPSEC*/
1254 
1255 		default:
1256 			error = ENOPROTOOPT;
1257 			break;
1258 		}
1259 		if (m)
1260 			(void)m_free(m);
1261 		break;
1262 
1263 	case PRCO_GETOPT:
1264 		switch (optname) {
1265 		case IP_OPTIONS:
1266 		case IP_RETOPTS:
1267 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1268 			MCLAIM(m, so->so_mowner);
1269 			if (inp->inp_options) {
1270 				m->m_len = inp->inp_options->m_len;
1271 				bcopy(mtod(inp->inp_options, caddr_t),
1272 				    mtod(m, caddr_t), (unsigned)m->m_len);
1273 			} else
1274 				m->m_len = 0;
1275 			break;
1276 
1277 		case IP_TOS:
1278 		case IP_TTL:
1279 		case IP_RECVOPTS:
1280 		case IP_RECVRETOPTS:
1281 		case IP_RECVDSTADDR:
1282 		case IP_RECVIF:
1283 		case IP_ERRORMTU:
1284 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1285 			MCLAIM(m, so->so_mowner);
1286 			m->m_len = sizeof(int);
1287 			switch (optname) {
1288 
1289 			case IP_TOS:
1290 				optval = inp->inp_ip.ip_tos;
1291 				break;
1292 
1293 			case IP_TTL:
1294 				optval = inp->inp_ip.ip_ttl;
1295 				break;
1296 
1297 			case IP_ERRORMTU:
1298 				optval = inp->inp_errormtu;
1299 				break;
1300 
1301 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1302 
1303 			case IP_RECVOPTS:
1304 				optval = OPTBIT(INP_RECVOPTS);
1305 				break;
1306 
1307 			case IP_RECVRETOPTS:
1308 				optval = OPTBIT(INP_RECVRETOPTS);
1309 				break;
1310 
1311 			case IP_RECVDSTADDR:
1312 				optval = OPTBIT(INP_RECVDSTADDR);
1313 				break;
1314 
1315 			case IP_RECVIF:
1316 				optval = OPTBIT(INP_RECVIF);
1317 				break;
1318 			}
1319 			*mtod(m, int *) = optval;
1320 			break;
1321 
1322 #if 0	/* defined(IPSEC) || defined(FAST_IPSEC) */
1323 		/* XXX: code broken */
1324 		case IP_IPSEC_POLICY:
1325 		{
1326 			caddr_t req = NULL;
1327 			size_t len = 0;
1328 
1329 			if (m) {
1330 				req = mtod(m, caddr_t);
1331 				len = m->m_len;
1332 			}
1333 			error = ipsec4_get_policy(inp, req, len, mp);
1334 			break;
1335 		}
1336 #endif /*IPSEC*/
1337 
1338 		case IP_MULTICAST_IF:
1339 		case IP_MULTICAST_TTL:
1340 		case IP_MULTICAST_LOOP:
1341 		case IP_ADD_MEMBERSHIP:
1342 		case IP_DROP_MEMBERSHIP:
1343 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1344 			if (*mp)
1345 				MCLAIM(*mp, so->so_mowner);
1346 			break;
1347 
1348 		case IP_PORTRANGE:
1349 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1350 			MCLAIM(m, so->so_mowner);
1351 			m->m_len = sizeof(int);
1352 
1353 			if (inp->inp_flags & INP_LOWPORT)
1354 				optval = IP_PORTRANGE_LOW;
1355 			else
1356 				optval = IP_PORTRANGE_DEFAULT;
1357 
1358 			*mtod(m, int *) = optval;
1359 			break;
1360 
1361 		default:
1362 			error = ENOPROTOOPT;
1363 			break;
1364 		}
1365 		break;
1366 	}
1367 	return (error);
1368 }
1369 
1370 /*
1371  * Set up IP options in pcb for insertion in output packets.
1372  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1373  * with destination address if source routed.
1374  */
1375 int
1376 #ifdef notyet
1377 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1378 #else
1379 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1380 #endif
1381 {
1382 	int cnt, optlen;
1383 	u_char *cp;
1384 	u_char opt;
1385 
1386 	/* turn off any old options */
1387 	if (*pcbopt)
1388 		(void)m_free(*pcbopt);
1389 	*pcbopt = 0;
1390 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1391 		/*
1392 		 * Only turning off any previous options.
1393 		 */
1394 		if (m)
1395 			(void)m_free(m);
1396 		return (0);
1397 	}
1398 
1399 #ifndef	__vax__
1400 	if (m->m_len % sizeof(int32_t))
1401 		goto bad;
1402 #endif
1403 	/*
1404 	 * IP first-hop destination address will be stored before
1405 	 * actual options; move other options back
1406 	 * and clear it when none present.
1407 	 */
1408 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1409 		goto bad;
1410 	cnt = m->m_len;
1411 	m->m_len += sizeof(struct in_addr);
1412 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1413 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1414 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1415 
1416 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1417 		opt = cp[IPOPT_OPTVAL];
1418 		if (opt == IPOPT_EOL)
1419 			break;
1420 		if (opt == IPOPT_NOP)
1421 			optlen = 1;
1422 		else {
1423 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1424 				goto bad;
1425 			optlen = cp[IPOPT_OLEN];
1426 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1427 				goto bad;
1428 		}
1429 		switch (opt) {
1430 
1431 		default:
1432 			break;
1433 
1434 		case IPOPT_LSRR:
1435 		case IPOPT_SSRR:
1436 			/*
1437 			 * user process specifies route as:
1438 			 *	->A->B->C->D
1439 			 * D must be our final destination (but we can't
1440 			 * check that since we may not have connected yet).
1441 			 * A is first hop destination, which doesn't appear in
1442 			 * actual IP option, but is stored before the options.
1443 			 */
1444 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1445 				goto bad;
1446 			m->m_len -= sizeof(struct in_addr);
1447 			cnt -= sizeof(struct in_addr);
1448 			optlen -= sizeof(struct in_addr);
1449 			cp[IPOPT_OLEN] = optlen;
1450 			/*
1451 			 * Move first hop before start of options.
1452 			 */
1453 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1454 			    sizeof(struct in_addr));
1455 			/*
1456 			 * Then copy rest of options back
1457 			 * to close up the deleted entry.
1458 			 */
1459 			(void)memmove(&cp[IPOPT_OFFSET+1],
1460 			    &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1461 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
1462 			break;
1463 		}
1464 	}
1465 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1466 		goto bad;
1467 	*pcbopt = m;
1468 	return (0);
1469 
1470 bad:
1471 	(void)m_free(m);
1472 	return (EINVAL);
1473 }
1474 
1475 /*
1476  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1477  */
1478 static struct ifnet *
1479 ip_multicast_if(struct in_addr *a, int *ifindexp)
1480 {
1481 	int ifindex;
1482 	struct ifnet *ifp = NULL;
1483 	struct in_ifaddr *ia;
1484 
1485 	if (ifindexp)
1486 		*ifindexp = 0;
1487 	if (ntohl(a->s_addr) >> 24 == 0) {
1488 		ifindex = ntohl(a->s_addr) & 0xffffff;
1489 		if (ifindex < 0 || if_indexlim <= ifindex)
1490 			return NULL;
1491 		ifp = ifindex2ifnet[ifindex];
1492 		if (!ifp)
1493 			return NULL;
1494 		if (ifindexp)
1495 			*ifindexp = ifindex;
1496 	} else {
1497 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1498 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1499 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1500 				ifp = ia->ia_ifp;
1501 				break;
1502 			}
1503 		}
1504 	}
1505 	return ifp;
1506 }
1507 
1508 /*
1509  * Set the IP multicast options in response to user setsockopt().
1510  */
1511 int
1512 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m)
1513 {
1514 	int error = 0;
1515 	u_char loop;
1516 	int i;
1517 	struct in_addr addr;
1518 	struct ip_mreq *mreq;
1519 	struct ifnet *ifp;
1520 	struct ip_moptions *imo = *imop;
1521 	struct route ro;
1522 	struct sockaddr_in *dst;
1523 	int ifindex;
1524 
1525 	if (imo == NULL) {
1526 		/*
1527 		 * No multicast option buffer attached to the pcb;
1528 		 * allocate one and initialize to default values.
1529 		 */
1530 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1531 		    M_WAITOK);
1532 
1533 		if (imo == NULL)
1534 			return (ENOBUFS);
1535 		*imop = imo;
1536 		imo->imo_multicast_ifp = NULL;
1537 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1538 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1539 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1540 		imo->imo_num_memberships = 0;
1541 	}
1542 
1543 	switch (optname) {
1544 
1545 	case IP_MULTICAST_IF:
1546 		/*
1547 		 * Select the interface for outgoing multicast packets.
1548 		 */
1549 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1550 			error = EINVAL;
1551 			break;
1552 		}
1553 		addr = *(mtod(m, struct in_addr *));
1554 		/*
1555 		 * INADDR_ANY is used to remove a previous selection.
1556 		 * When no interface is selected, a default one is
1557 		 * chosen every time a multicast packet is sent.
1558 		 */
1559 		if (in_nullhost(addr)) {
1560 			imo->imo_multicast_ifp = NULL;
1561 			break;
1562 		}
1563 		/*
1564 		 * The selected interface is identified by its local
1565 		 * IP address.  Find the interface and confirm that
1566 		 * it supports multicasting.
1567 		 */
1568 		ifp = ip_multicast_if(&addr, &ifindex);
1569 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1570 			error = EADDRNOTAVAIL;
1571 			break;
1572 		}
1573 		imo->imo_multicast_ifp = ifp;
1574 		if (ifindex)
1575 			imo->imo_multicast_addr = addr;
1576 		else
1577 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1578 		break;
1579 
1580 	case IP_MULTICAST_TTL:
1581 		/*
1582 		 * Set the IP time-to-live for outgoing multicast packets.
1583 		 */
1584 		if (m == NULL || m->m_len != 1) {
1585 			error = EINVAL;
1586 			break;
1587 		}
1588 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1589 		break;
1590 
1591 	case IP_MULTICAST_LOOP:
1592 		/*
1593 		 * Set the loopback flag for outgoing multicast packets.
1594 		 * Must be zero or one.
1595 		 */
1596 		if (m == NULL || m->m_len != 1 ||
1597 		   (loop = *(mtod(m, u_char *))) > 1) {
1598 			error = EINVAL;
1599 			break;
1600 		}
1601 		imo->imo_multicast_loop = loop;
1602 		break;
1603 
1604 	case IP_ADD_MEMBERSHIP:
1605 		/*
1606 		 * Add a multicast group membership.
1607 		 * Group must be a valid IP multicast address.
1608 		 */
1609 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1610 			error = EINVAL;
1611 			break;
1612 		}
1613 		mreq = mtod(m, struct ip_mreq *);
1614 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1615 			error = EINVAL;
1616 			break;
1617 		}
1618 		/*
1619 		 * If no interface address was provided, use the interface of
1620 		 * the route to the given multicast address.
1621 		 */
1622 		if (in_nullhost(mreq->imr_interface)) {
1623 			bzero((caddr_t)&ro, sizeof(ro));
1624 			ro.ro_rt = NULL;
1625 			dst = satosin(&ro.ro_dst);
1626 			dst->sin_len = sizeof(*dst);
1627 			dst->sin_family = AF_INET;
1628 			dst->sin_addr = mreq->imr_multiaddr;
1629 			rtalloc(&ro);
1630 			if (ro.ro_rt == NULL) {
1631 				error = EADDRNOTAVAIL;
1632 				break;
1633 			}
1634 			ifp = ro.ro_rt->rt_ifp;
1635 			rtfree(ro.ro_rt);
1636 		} else {
1637 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1638 		}
1639 		/*
1640 		 * See if we found an interface, and confirm that it
1641 		 * supports multicast.
1642 		 */
1643 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1644 			error = EADDRNOTAVAIL;
1645 			break;
1646 		}
1647 		/*
1648 		 * See if the membership already exists or if all the
1649 		 * membership slots are full.
1650 		 */
1651 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1652 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1653 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1654 				      mreq->imr_multiaddr))
1655 				break;
1656 		}
1657 		if (i < imo->imo_num_memberships) {
1658 			error = EADDRINUSE;
1659 			break;
1660 		}
1661 		if (i == IP_MAX_MEMBERSHIPS) {
1662 			error = ETOOMANYREFS;
1663 			break;
1664 		}
1665 		/*
1666 		 * Everything looks good; add a new record to the multicast
1667 		 * address list for the given interface.
1668 		 */
1669 		if ((imo->imo_membership[i] =
1670 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1671 			error = ENOBUFS;
1672 			break;
1673 		}
1674 		++imo->imo_num_memberships;
1675 		break;
1676 
1677 	case IP_DROP_MEMBERSHIP:
1678 		/*
1679 		 * Drop a multicast group membership.
1680 		 * Group must be a valid IP multicast address.
1681 		 */
1682 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1683 			error = EINVAL;
1684 			break;
1685 		}
1686 		mreq = mtod(m, struct ip_mreq *);
1687 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1688 			error = EINVAL;
1689 			break;
1690 		}
1691 		/*
1692 		 * If an interface address was specified, get a pointer
1693 		 * to its ifnet structure.
1694 		 */
1695 		if (in_nullhost(mreq->imr_interface))
1696 			ifp = NULL;
1697 		else {
1698 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1699 			if (ifp == NULL) {
1700 				error = EADDRNOTAVAIL;
1701 				break;
1702 			}
1703 		}
1704 		/*
1705 		 * Find the membership in the membership array.
1706 		 */
1707 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1708 			if ((ifp == NULL ||
1709 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1710 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1711 				       mreq->imr_multiaddr))
1712 				break;
1713 		}
1714 		if (i == imo->imo_num_memberships) {
1715 			error = EADDRNOTAVAIL;
1716 			break;
1717 		}
1718 		/*
1719 		 * Give up the multicast address record to which the
1720 		 * membership points.
1721 		 */
1722 		in_delmulti(imo->imo_membership[i]);
1723 		/*
1724 		 * Remove the gap in the membership array.
1725 		 */
1726 		for (++i; i < imo->imo_num_memberships; ++i)
1727 			imo->imo_membership[i-1] = imo->imo_membership[i];
1728 		--imo->imo_num_memberships;
1729 		break;
1730 
1731 	default:
1732 		error = EOPNOTSUPP;
1733 		break;
1734 	}
1735 
1736 	/*
1737 	 * If all options have default values, no need to keep the mbuf.
1738 	 */
1739 	if (imo->imo_multicast_ifp == NULL &&
1740 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1741 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1742 	    imo->imo_num_memberships == 0) {
1743 		free(*imop, M_IPMOPTS);
1744 		*imop = NULL;
1745 	}
1746 
1747 	return (error);
1748 }
1749 
1750 /*
1751  * Return the IP multicast options in response to user getsockopt().
1752  */
1753 int
1754 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
1755 {
1756 	u_char *ttl;
1757 	u_char *loop;
1758 	struct in_addr *addr;
1759 	struct in_ifaddr *ia;
1760 
1761 	*mp = m_get(M_WAIT, MT_SOOPTS);
1762 
1763 	switch (optname) {
1764 
1765 	case IP_MULTICAST_IF:
1766 		addr = mtod(*mp, struct in_addr *);
1767 		(*mp)->m_len = sizeof(struct in_addr);
1768 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1769 			*addr = zeroin_addr;
1770 		else if (imo->imo_multicast_addr.s_addr) {
1771 			/* return the value user has set */
1772 			*addr = imo->imo_multicast_addr;
1773 		} else {
1774 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1775 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1776 		}
1777 		return (0);
1778 
1779 	case IP_MULTICAST_TTL:
1780 		ttl = mtod(*mp, u_char *);
1781 		(*mp)->m_len = 1;
1782 		*ttl = imo ? imo->imo_multicast_ttl
1783 			   : IP_DEFAULT_MULTICAST_TTL;
1784 		return (0);
1785 
1786 	case IP_MULTICAST_LOOP:
1787 		loop = mtod(*mp, u_char *);
1788 		(*mp)->m_len = 1;
1789 		*loop = imo ? imo->imo_multicast_loop
1790 			    : IP_DEFAULT_MULTICAST_LOOP;
1791 		return (0);
1792 
1793 	default:
1794 		return (EOPNOTSUPP);
1795 	}
1796 }
1797 
1798 /*
1799  * Discard the IP multicast options.
1800  */
1801 void
1802 ip_freemoptions(struct ip_moptions *imo)
1803 {
1804 	int i;
1805 
1806 	if (imo != NULL) {
1807 		for (i = 0; i < imo->imo_num_memberships; ++i)
1808 			in_delmulti(imo->imo_membership[i]);
1809 		free(imo, M_IPMOPTS);
1810 	}
1811 }
1812 
1813 /*
1814  * Routine called from ip_output() to loop back a copy of an IP multicast
1815  * packet to the input queue of a specified interface.  Note that this
1816  * calls the output routine of the loopback "driver", but with an interface
1817  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
1818  */
1819 static void
1820 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1821 {
1822 	struct ip *ip;
1823 	struct mbuf *copym;
1824 
1825 	copym = m_copy(m, 0, M_COPYALL);
1826 	if (copym != NULL
1827 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1828 		copym = m_pullup(copym, sizeof(struct ip));
1829 	if (copym != NULL) {
1830 		/*
1831 		 * We don't bother to fragment if the IP length is greater
1832 		 * than the interface's MTU.  Can this possibly matter?
1833 		 */
1834 		ip = mtod(copym, struct ip *);
1835 
1836 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1837 			in_delayed_cksum(copym);
1838 			copym->m_pkthdr.csum_flags &=
1839 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1840 		}
1841 
1842 		ip->ip_sum = 0;
1843 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1844 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1845 	}
1846 }
1847