xref: /netbsd-src/sys/netinet/ip_output.c (revision da5f4674a3fc214be3572d358b66af40ab9401e7)
1 /*	$NetBSD: ip_output.c,v 1.119 2003/08/27 02:09:59 itojun 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 1.119 2003/08/27 02:09:59 itojun Exp $");
102 
103 #include "opt_pfil_hooks.h"
104 #include "opt_ipsec.h"
105 #include "opt_mrouting.h"
106 
107 #include <sys/param.h>
108 #include <sys/malloc.h>
109 #include <sys/mbuf.h>
110 #include <sys/errno.h>
111 #include <sys/protosw.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #ifdef FAST_IPSEC
115 #include <sys/domain.h>
116 #endif
117 #include <sys/systm.h>
118 #include <sys/proc.h>
119 
120 #include <net/if.h>
121 #include <net/route.h>
122 #include <net/pfil.h>
123 
124 #include <netinet/in.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/in_pcb.h>
128 #include <netinet/in_var.h>
129 #include <netinet/ip_var.h>
130 
131 #ifdef MROUTING
132 #include <netinet/ip_mroute.h>
133 #endif
134 
135 #include <machine/stdarg.h>
136 
137 #ifdef IPSEC
138 #include <netinet6/ipsec.h>
139 #include <netkey/key.h>
140 #include <netkey/key_debug.h>
141 #endif /*IPSEC*/
142 
143 #ifdef FAST_IPSEC
144 #include <netipsec/ipsec.h>
145 #include <netipsec/key.h>
146 #include <netipsec/xform.h>
147 #endif	/* FAST_IPSEC*/
148 
149 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
150 static struct ifnet *ip_multicast_if __P((struct in_addr *, int *));
151 static void ip_mloopback
152 	__P((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 #if __STDC__
166 ip_output(struct mbuf *m0, ...)
167 #else
168 ip_output(m0, va_alist)
169 	struct mbuf *m0;
170 	va_dcl
171 #endif
172 {
173 	struct ip *ip;
174 	struct ifnet *ifp;
175 	struct mbuf *m = m0;
176 	int hlen = sizeof (struct ip);
177 	int len, error = 0;
178 	struct route iproute;
179 	struct sockaddr_in *dst;
180 	struct in_ifaddr *ia;
181 	struct mbuf *opt;
182 	struct route *ro;
183 	int flags, sw_csum;
184 	int *mtu_p;
185 	u_long mtu;
186 	struct ip_moptions *imo;
187 	struct socket *so;
188 	va_list ap;
189 #ifdef IPSEC
190 	struct secpolicy *sp = NULL;
191 #endif /*IPSEC*/
192 #ifdef FAST_IPSEC
193 	struct inpcb *inp;
194 	struct m_tag *mtag;
195 	struct secpolicy *sp = NULL;
196 	struct tdb_ident *tdbi;
197 	int s;
198 #endif
199 	u_int16_t ip_len;
200 
201 	len = 0;
202 	va_start(ap, m0);
203 	opt = va_arg(ap, struct mbuf *);
204 	ro = va_arg(ap, struct route *);
205 	flags = va_arg(ap, int);
206 	imo = va_arg(ap, struct ip_moptions *);
207 	so = va_arg(ap, struct socket *);
208 	if (flags & IP_RETURNMTU)
209 		mtu_p = va_arg(ap, int *);
210 	else
211 		mtu_p = NULL;
212 	va_end(ap);
213 
214 	MCLAIM(m, &ip_tx_mowner);
215 #ifdef FAST_IPSEC
216 	if (so->so_proto->pr_domain->dom_family == AF_INET)
217 		inp = (struct inpcb *)so->so_pcb;
218 	else
219 		inp = NULL;
220 #endif /*IPSEC*/
221 
222 #ifdef	DIAGNOSTIC
223 	if ((m->m_flags & M_PKTHDR) == 0)
224 		panic("ip_output no HDR");
225 #endif
226 	if (opt) {
227 		m = ip_insertoptions(m, opt, &len);
228 		if (len >= sizeof(struct ip))
229 			hlen = len;
230 	}
231 	ip = mtod(m, struct ip *);
232 	/*
233 	 * Fill in IP header.
234 	 */
235 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
236 		ip->ip_v = IPVERSION;
237 		ip->ip_off = htons(0);
238 		ip->ip_id = htons(ip_id++);
239 		ip->ip_hl = hlen >> 2;
240 		ipstat.ips_localout++;
241 	} else {
242 		hlen = ip->ip_hl << 2;
243 	}
244 	/*
245 	 * Route packet.
246 	 */
247 	if (ro == 0) {
248 		ro = &iproute;
249 		bzero((caddr_t)ro, sizeof (*ro));
250 	}
251 	dst = satosin(&ro->ro_dst);
252 	/*
253 	 * If there is a cached route,
254 	 * check that it is to the same destination
255 	 * and is still up.  If not, free it and try again.
256 	 * The address family should also be checked in case of sharing the
257 	 * cache with IPv6.
258 	 */
259 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
260 	    dst->sin_family != AF_INET ||
261 	    !in_hosteq(dst->sin_addr, ip->ip_dst))) {
262 		RTFREE(ro->ro_rt);
263 		ro->ro_rt = (struct rtentry *)0;
264 	}
265 	if (ro->ro_rt == 0) {
266 		bzero(dst, sizeof(*dst));
267 		dst->sin_family = AF_INET;
268 		dst->sin_len = sizeof(*dst);
269 		dst->sin_addr = ip->ip_dst;
270 	}
271 	/*
272 	 * If routing to interface only,
273 	 * short circuit routing lookup.
274 	 */
275 	if (flags & IP_ROUTETOIF) {
276 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
277 			ipstat.ips_noroute++;
278 			error = ENETUNREACH;
279 			goto bad;
280 		}
281 		ifp = ia->ia_ifp;
282 		mtu = ifp->if_mtu;
283 		ip->ip_ttl = 1;
284 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
285 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
286 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
287 		ifp = imo->imo_multicast_ifp;
288 		mtu = ifp->if_mtu;
289 		IFP_TO_IA(ifp, ia);
290 	} else {
291 		if (ro->ro_rt == 0)
292 			rtalloc(ro);
293 		if (ro->ro_rt == 0) {
294 			ipstat.ips_noroute++;
295 			error = EHOSTUNREACH;
296 			goto bad;
297 		}
298 		ia = ifatoia(ro->ro_rt->rt_ifa);
299 		ifp = ro->ro_rt->rt_ifp;
300 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
301 			mtu = ifp->if_mtu;
302 		ro->ro_rt->rt_use++;
303 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
304 			dst = satosin(ro->ro_rt->rt_gateway);
305 	}
306 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
307 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
308 		struct in_multi *inm;
309 
310 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
311 			M_BCAST : M_MCAST;
312 		/*
313 		 * IP destination address is multicast.  Make sure "dst"
314 		 * still points to the address in "ro".  (It may have been
315 		 * changed to point to a gateway address, above.)
316 		 */
317 		dst = satosin(&ro->ro_dst);
318 		/*
319 		 * See if the caller provided any multicast options
320 		 */
321 		if (imo != NULL)
322 			ip->ip_ttl = imo->imo_multicast_ttl;
323 		else
324 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
325 
326 		/*
327 		 * if we don't know the outgoing ifp yet, we can't generate
328 		 * output
329 		 */
330 		if (!ifp) {
331 			ipstat.ips_noroute++;
332 			error = ENETUNREACH;
333 			goto bad;
334 		}
335 
336 		/*
337 		 * If the packet is multicast or broadcast, confirm that
338 		 * the outgoing interface can transmit it.
339 		 */
340 		if (((m->m_flags & M_MCAST) &&
341 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
342 		    ((m->m_flags & M_BCAST) &&
343 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
344 			ipstat.ips_noroute++;
345 			error = ENETUNREACH;
346 			goto bad;
347 		}
348 		/*
349 		 * If source address not specified yet, use an address
350 		 * of outgoing interface.
351 		 */
352 		if (in_nullhost(ip->ip_src)) {
353 			struct in_ifaddr *ia;
354 
355 			IFP_TO_IA(ifp, ia);
356 			if (!ia) {
357 				error = EADDRNOTAVAIL;
358 				goto bad;
359 			}
360 			ip->ip_src = ia->ia_addr.sin_addr;
361 		}
362 
363 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
364 		if (inm != NULL &&
365 		   (imo == NULL || imo->imo_multicast_loop)) {
366 			/*
367 			 * If we belong to the destination multicast group
368 			 * on the outgoing interface, and the caller did not
369 			 * forbid loopback, loop back a copy.
370 			 */
371 			ip_mloopback(ifp, m, dst);
372 		}
373 #ifdef MROUTING
374 		else {
375 			/*
376 			 * If we are acting as a multicast router, perform
377 			 * multicast forwarding as if the packet had just
378 			 * arrived on the interface to which we are about
379 			 * to send.  The multicast forwarding function
380 			 * recursively calls this function, using the
381 			 * IP_FORWARDING flag to prevent infinite recursion.
382 			 *
383 			 * Multicasts that are looped back by ip_mloopback(),
384 			 * above, will be forwarded by the ip_input() routine,
385 			 * if necessary.
386 			 */
387 			extern struct socket *ip_mrouter;
388 
389 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
390 				if (ip_mforward(m, ifp) != 0) {
391 					m_freem(m);
392 					goto done;
393 				}
394 			}
395 		}
396 #endif
397 		/*
398 		 * Multicasts with a time-to-live of zero may be looped-
399 		 * back, above, but must not be transmitted on a network.
400 		 * Also, multicasts addressed to the loopback interface
401 		 * are not sent -- the above call to ip_mloopback() will
402 		 * loop back a copy if this host actually belongs to the
403 		 * destination group on the loopback interface.
404 		 */
405 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
406 			m_freem(m);
407 			goto done;
408 		}
409 
410 		goto sendit;
411 	}
412 #ifndef notdef
413 	/*
414 	 * If source address not specified yet, use address
415 	 * of outgoing interface.
416 	 */
417 	if (in_nullhost(ip->ip_src))
418 		ip->ip_src = ia->ia_addr.sin_addr;
419 #endif
420 
421 	/*
422 	 * packets with Class-D address as source are not valid per
423 	 * RFC 1112
424 	 */
425 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
426 		ipstat.ips_odropped++;
427 		error = EADDRNOTAVAIL;
428 		goto bad;
429 	}
430 
431 	/*
432 	 * Look for broadcast address and
433 	 * and verify user is allowed to send
434 	 * such a packet.
435 	 */
436 	if (in_broadcast(dst->sin_addr, ifp)) {
437 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
438 			error = EADDRNOTAVAIL;
439 			goto bad;
440 		}
441 		if ((flags & IP_ALLOWBROADCAST) == 0) {
442 			error = EACCES;
443 			goto bad;
444 		}
445 		/* don't allow broadcast messages to be fragmented */
446 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
447 			error = EMSGSIZE;
448 			goto bad;
449 		}
450 		m->m_flags |= M_BCAST;
451 	} else
452 		m->m_flags &= ~M_BCAST;
453 
454 sendit:
455 	/*
456 	 * If we're doing Path MTU Discovery, we need to set DF unless
457 	 * the route's MTU is locked.
458 	 */
459 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
460 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
461 		ip->ip_off |= htons(IP_DF);
462 
463 	/* Remember the current ip_len */
464 	ip_len = ntohs(ip->ip_len);
465 
466 #ifdef IPSEC
467 	/* get SP for this packet */
468 	if (so == NULL)
469 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
470 		    flags, &error);
471 	else
472 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
473 
474 	if (sp == NULL) {
475 		ipsecstat.out_inval++;
476 		goto bad;
477 	}
478 
479 	error = 0;
480 
481 	/* check policy */
482 	switch (sp->policy) {
483 	case IPSEC_POLICY_DISCARD:
484 		/*
485 		 * This packet is just discarded.
486 		 */
487 		ipsecstat.out_polvio++;
488 		goto bad;
489 
490 	case IPSEC_POLICY_BYPASS:
491 	case IPSEC_POLICY_NONE:
492 		/* no need to do IPsec. */
493 		goto skip_ipsec;
494 
495 	case IPSEC_POLICY_IPSEC:
496 		if (sp->req == NULL) {
497 			/* XXX should be panic ? */
498 			printf("ip_output: No IPsec request specified.\n");
499 			error = EINVAL;
500 			goto bad;
501 		}
502 		break;
503 
504 	case IPSEC_POLICY_ENTRUST:
505 	default:
506 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
507 	}
508 
509 	/*
510 	 * ipsec4_output() expects ip_len and ip_off in network
511 	 * order.  They have been set to network order above.
512 	 */
513 
514     {
515 	struct ipsec_output_state state;
516 	bzero(&state, sizeof(state));
517 	state.m = m;
518 	if (flags & IP_ROUTETOIF) {
519 		state.ro = &iproute;
520 		bzero(&iproute, sizeof(iproute));
521 	} else
522 		state.ro = ro;
523 	state.dst = (struct sockaddr *)dst;
524 
525 	/*
526 	 * We can't defer the checksum of payload data if
527 	 * we're about to encrypt/authenticate it.
528 	 *
529 	 * XXX When we support crypto offloading functions of
530 	 * XXX network interfaces, we need to reconsider this,
531 	 * XXX since it's likely that they'll support checksumming,
532 	 * XXX as well.
533 	 */
534 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
535 		in_delayed_cksum(m);
536 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
537 	}
538 
539 	error = ipsec4_output(&state, sp, flags);
540 
541 	m = state.m;
542 	if (flags & IP_ROUTETOIF) {
543 		/*
544 		 * if we have tunnel mode SA, we may need to ignore
545 		 * IP_ROUTETOIF.
546 		 */
547 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
548 			flags &= ~IP_ROUTETOIF;
549 			ro = state.ro;
550 		}
551 	} else
552 		ro = state.ro;
553 	dst = (struct sockaddr_in *)state.dst;
554 	if (error) {
555 		/* mbuf is already reclaimed in ipsec4_output. */
556 		m0 = NULL;
557 		switch (error) {
558 		case EHOSTUNREACH:
559 		case ENETUNREACH:
560 		case EMSGSIZE:
561 		case ENOBUFS:
562 		case ENOMEM:
563 			break;
564 		default:
565 			printf("ip4_output (ipsec): error code %d\n", error);
566 			/*fall through*/
567 		case ENOENT:
568 			/* don't show these error codes to the user */
569 			error = 0;
570 			break;
571 		}
572 		goto bad;
573 	}
574 
575 	/* be sure to update variables that are affected by ipsec4_output() */
576 	ip = mtod(m, struct ip *);
577 	hlen = ip->ip_hl << 2;
578 	ip_len = ntohs(ip->ip_len);
579 
580 	if (ro->ro_rt == NULL) {
581 		if ((flags & IP_ROUTETOIF) == 0) {
582 			printf("ip_output: "
583 				"can't update route after IPsec processing\n");
584 			error = EHOSTUNREACH;	/*XXX*/
585 			goto bad;
586 		}
587 	} else {
588 		/* nobody uses ia beyond here */
589 		if (state.encap)
590 			ifp = ro->ro_rt->rt_ifp;
591 	}
592     }
593 skip_ipsec:
594 #endif /*IPSEC*/
595 #ifdef FAST_IPSEC
596 	/*
597 	 * Check the security policy (SP) for the packet and, if
598 	 * required, do IPsec-related processing.  There are two
599 	 * cases here; the first time a packet is sent through
600 	 * it will be untagged and handled by ipsec4_checkpolicy.
601 	 * If the packet is resubmitted to ip_output (e.g. after
602 	 * AH, ESP, etc. processing), there will be a tag to bypass
603 	 * the lookup and related policy checking.
604 	 */
605 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
606 	s = splsoftnet();
607 	if (mtag != NULL) {
608 		tdbi = (struct tdb_ident *)(mtag + 1);
609 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
610 		if (sp == NULL)
611 			error = -EINVAL;	/* force silent drop */
612 		m_tag_delete(m, mtag);
613 	} else {
614 		sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
615 					&error, inp);
616 	}
617 	/*
618 	 * There are four return cases:
619 	 *    sp != NULL	 	    apply IPsec policy
620 	 *    sp == NULL, error == 0	    no IPsec handling needed
621 	 *    sp == NULL, error == -EINVAL  discard packet w/o error
622 	 *    sp == NULL, error != 0	    discard packet, report error
623 	 */
624 	if (sp != NULL) {
625 		/* Loop detection, check if ipsec processing already done */
626 		IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
627 		for (mtag = m_tag_first(m); mtag != NULL;
628 		     mtag = m_tag_next(m, mtag)) {
629 #ifdef MTAG_ABI_COMPAT
630 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
631 				continue;
632 #endif
633 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
634 			    mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
635 				continue;
636 			/*
637 			 * Check if policy has an SA associated with it.
638 			 * This can happen when an SP has yet to acquire
639 			 * an SA; e.g. on first reference.  If it occurs,
640 			 * then we let ipsec4_process_packet do its thing.
641 			 */
642 			if (sp->req->sav == NULL)
643 				break;
644 			tdbi = (struct tdb_ident *)(mtag + 1);
645 			if (tdbi->spi == sp->req->sav->spi &&
646 			    tdbi->proto == sp->req->sav->sah->saidx.proto &&
647 			    bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
648 				 sizeof (union sockaddr_union)) == 0) {
649 				/*
650 				 * No IPsec processing is needed, free
651 				 * reference to SP.
652 				 *
653 				 * NB: null pointer to avoid free at
654 				 *     done: below.
655 				 */
656 				KEY_FREESP(&sp), sp = NULL;
657 				splx(s);
658 				goto spd_done;
659 			}
660 		}
661 
662 		/*
663 		 * Do delayed checksums now because we send before
664 		 * this is done in the normal processing path.
665 		 */
666 		if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
667 			in_delayed_cksum(m);
668 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
669 		}
670 
671 #ifdef __FreeBSD__
672 		ip->ip_len = htons(ip->ip_len);
673 		ip->ip_off = htons(ip->ip_off);
674 #endif
675 
676 		/* NB: callee frees mbuf */
677 		error = ipsec4_process_packet(m, sp->req, flags, 0);
678 		/*
679 		 * Preserve KAME behaviour: ENOENT can be returned
680 		 * when an SA acquire is in progress.  Don't propagate
681 		 * this to user-level; it confuses applications.
682 		 *
683 		 * XXX this will go away when the SADB is redone.
684 		 */
685 		if (error == ENOENT)
686 			error = 0;
687 		splx(s);
688 		goto done;
689 	} else {
690 		splx(s);
691 
692 		if (error != 0) {
693 			/*
694 			 * Hack: -EINVAL is used to signal that a packet
695 			 * should be silently discarded.  This is typically
696 			 * because we asked key management for an SA and
697 			 * it was delayed (e.g. kicked up to IKE).
698 			 */
699 			if (error == -EINVAL)
700 				error = 0;
701 			goto bad;
702 		} else {
703 			/* No IPsec processing for this packet. */
704 		}
705 #ifdef notyet
706 		/*
707 		 * If deferred crypto processing is needed, check that
708 		 * the interface supports it.
709 		 */
710 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
711 		if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
712 			/* notify IPsec to do its own crypto */
713 			ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
714 			error = EHOSTUNREACH;
715 			goto bad;
716 		}
717 #endif
718 	}
719 spd_done:
720 #endif /* FAST_IPSEC */
721 
722 #ifdef PFIL_HOOKS
723 	/*
724 	 * Run through list of hooks for output packets.
725 	 */
726 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
727 		goto done;
728 	if (m == NULL)
729 		goto done;
730 
731 	ip = mtod(m, struct ip *);
732 	hlen = ip->ip_hl << 2;
733 #endif /* PFIL_HOOKS */
734 
735 	m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
736 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
737 	/*
738 	 * If small enough for mtu of path, can just send directly.
739 	 */
740 	if (ip_len <= mtu) {
741 #if IFA_STATS
742 		/*
743 		 * search for the source address structure to
744 		 * maintain output statistics.
745 		 */
746 		INADDR_TO_IA(ip->ip_src, ia);
747 		if (ia)
748 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
749 #endif
750 		/*
751 		 * Always initialize the sum to 0!  Some HW assisted
752 		 * checksumming requires this.
753 		 */
754 		ip->ip_sum = 0;
755 
756 		/*
757 		 * Perform any checksums that the hardware can't do
758 		 * for us.
759 		 *
760 		 * XXX Does any hardware require the {th,uh}_sum
761 		 * XXX fields to be 0?
762 		 */
763 		if (sw_csum & M_CSUM_IPv4) {
764 			ip->ip_sum = in_cksum(m, hlen);
765 			m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
766 		}
767 		if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
768 			in_delayed_cksum(m);
769 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
770 		}
771 
772 #ifdef IPSEC
773 		/* clean ipsec history once it goes out of the node */
774 		ipsec_delaux(m);
775 #endif
776 		error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
777 		goto done;
778 	}
779 
780 	/*
781 	 * We can't use HW checksumming if we're about to
782 	 * to fragment the packet.
783 	 *
784 	 * XXX Some hardware can do this.
785 	 */
786 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
787 		in_delayed_cksum(m);
788 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
789 	}
790 
791 	/*
792 	 * Too large for interface; fragment if possible.
793 	 * Must be able to put at least 8 bytes per fragment.
794 	 */
795 	if (ntohs(ip->ip_off) & IP_DF) {
796 		if (flags & IP_RETURNMTU)
797 			*mtu_p = mtu;
798 		error = EMSGSIZE;
799 		ipstat.ips_cantfrag++;
800 		goto bad;
801 	}
802 
803 	error = ip_fragment(m, ifp, mtu);
804 	if (error == EMSGSIZE)
805 		goto bad;
806 
807 	for (; m; m = m0) {
808 		m0 = m->m_nextpkt;
809 		m->m_nextpkt = 0;
810 		if (error == 0) {
811 #if IFA_STATS
812 			/*
813 			 * search for the source address structure to
814 			 * maintain output statistics.
815 			 */
816 			INADDR_TO_IA(ip->ip_src, ia);
817 			if (ia) {
818 				ia->ia_ifa.ifa_data.ifad_outbytes +=
819 				    ntohs(ip->ip_len);
820 			}
821 #endif
822 #ifdef IPSEC
823 			/* clean ipsec history once it goes out of the node */
824 			ipsec_delaux(m);
825 #endif
826 			KASSERT((m->m_pkthdr.csum_flags &
827 			    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
828 			error = (*ifp->if_output)(ifp, m, sintosa(dst),
829 			    ro->ro_rt);
830 		} else
831 			m_freem(m);
832 	}
833 
834 	if (error == 0)
835 		ipstat.ips_fragmented++;
836 done:
837 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) {
838 		RTFREE(ro->ro_rt);
839 		ro->ro_rt = 0;
840 	}
841 
842 #ifdef IPSEC
843 	if (sp != NULL) {
844 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
845 			printf("DP ip_output call free SP:%p\n", sp));
846 		key_freesp(sp);
847 	}
848 #endif /* IPSEC */
849 #ifdef FAST_IPSEC
850 	if (sp != NULL)
851 		KEY_FREESP(&sp);
852 #endif /* FAST_IPSEC */
853 
854 	return (error);
855 bad:
856 	m_freem(m);
857 	goto done;
858 }
859 
860 int
861 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
862 {
863 	struct ip *ip, *mhip;
864 	struct mbuf *m0;
865 	int len, hlen, off;
866 	int mhlen, firstlen;
867 	struct mbuf **mnext;
868 	int sw_csum;
869 	int fragments = 0;
870 	int s;
871 	int error = 0;
872 
873 	ip = mtod(m, struct ip *);
874 	hlen = ip->ip_hl << 2;
875 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
876 
877 	len = (mtu - hlen) &~ 7;
878 	if (len < 8)
879 		return (EMSGSIZE);
880 
881 	firstlen = len;
882 	mnext = &m->m_nextpkt;
883 
884 	/*
885 	 * Loop through length of segment after first fragment,
886 	 * make new header and copy data of each part and link onto chain.
887 	 */
888 	m0 = m;
889 	mhlen = sizeof (struct ip);
890 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
891 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
892 		if (m == 0) {
893 			error = ENOBUFS;
894 			ipstat.ips_odropped++;
895 			goto sendorfree;
896 		}
897 		MCLAIM(m, m0->m_owner);
898 		*mnext = m;
899 		mnext = &m->m_nextpkt;
900 		m->m_data += max_linkhdr;
901 		mhip = mtod(m, struct ip *);
902 		*mhip = *ip;
903 		/* we must inherit MCAST and BCAST flags */
904 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
905 		if (hlen > sizeof (struct ip)) {
906 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
907 			mhip->ip_hl = mhlen >> 2;
908 		}
909 		m->m_len = mhlen;
910 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
911 		if (ip->ip_off & IP_MF)
912 			mhip->ip_off |= IP_MF;
913 		if (off + len >= ntohs(ip->ip_len))
914 			len = ntohs(ip->ip_len) - off;
915 		else
916 			mhip->ip_off |= IP_MF;
917 		HTONS(mhip->ip_off);
918 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
919 		m->m_next = m_copy(m0, off, len);
920 		if (m->m_next == 0) {
921 			error = ENOBUFS;	/* ??? */
922 			ipstat.ips_odropped++;
923 			goto sendorfree;
924 		}
925 		m->m_pkthdr.len = mhlen + len;
926 		m->m_pkthdr.rcvif = (struct ifnet *)0;
927 		mhip->ip_sum = 0;
928 		if (sw_csum & M_CSUM_IPv4) {
929 			mhip->ip_sum = in_cksum(m, mhlen);
930 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
931 		} else {
932 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
933 		}
934 		ipstat.ips_ofragments++;
935 		fragments++;
936 	}
937 	/*
938 	 * Update first fragment by trimming what's been copied out
939 	 * and updating header, then send each fragment (in order).
940 	 */
941 	m = m0;
942 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
943 	m->m_pkthdr.len = hlen + firstlen;
944 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
945 	ip->ip_off |= htons(IP_MF);
946 	ip->ip_sum = 0;
947 	if (sw_csum & M_CSUM_IPv4) {
948 		ip->ip_sum = in_cksum(m, hlen);
949 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
950 	} else {
951 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
952 	}
953 sendorfree:
954 	/*
955 	 * If there is no room for all the fragments, don't queue
956 	 * any of them.
957 	 */
958 	s = splnet();
959 	if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments)
960 		error = ENOBUFS;
961 	splx(s);
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->m_pkthdr.csum_data;	/* checksum offset */
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), (caddr_t) &csum);
988 	} else
989 		*(u_int16_t *)(mtod(m, caddr_t) + 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(inp)
999 	struct inpcb *inp;
1000 {
1001 	struct mbuf *m = inp->inp_options;
1002 
1003 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
1004 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1005 	else
1006 		return 0;
1007 }
1008 
1009 
1010 /*
1011  * Insert IP options into preformed packet.
1012  * Adjust IP destination as required for IP source routing,
1013  * as indicated by a non-zero in_addr at the start of the options.
1014  */
1015 static struct mbuf *
1016 ip_insertoptions(m, opt, phlen)
1017 	struct mbuf *m;
1018 	struct mbuf *opt;
1019 	int *phlen;
1020 {
1021 	struct ipoption *p = mtod(opt, struct ipoption *);
1022 	struct mbuf *n;
1023 	struct ip *ip = mtod(m, struct ip *);
1024 	unsigned optlen;
1025 
1026 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1027 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1028 		return (m);		/* XXX should fail */
1029 	if (!in_nullhost(p->ipopt_dst))
1030 		ip->ip_dst = p->ipopt_dst;
1031 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1032 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1033 		if (n == 0)
1034 			return (m);
1035 		MCLAIM(n, m->m_owner);
1036 		M_COPY_PKTHDR(n, m);
1037 		m->m_flags &= ~M_PKTHDR;
1038 		m->m_len -= sizeof(struct ip);
1039 		m->m_data += sizeof(struct ip);
1040 		n->m_next = m;
1041 		m = n;
1042 		m->m_len = optlen + sizeof(struct ip);
1043 		m->m_data += max_linkhdr;
1044 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1045 	} else {
1046 		m->m_data -= optlen;
1047 		m->m_len += optlen;
1048 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
1049 	}
1050 	m->m_pkthdr.len += optlen;
1051 	ip = mtod(m, struct ip *);
1052 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1053 	*phlen = sizeof(struct ip) + optlen;
1054 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1055 	return (m);
1056 }
1057 
1058 /*
1059  * Copy options from ip to jp,
1060  * omitting those not copied during fragmentation.
1061  */
1062 int
1063 ip_optcopy(ip, jp)
1064 	struct ip *ip, *jp;
1065 {
1066 	u_char *cp, *dp;
1067 	int opt, optlen, cnt;
1068 
1069 	cp = (u_char *)(ip + 1);
1070 	dp = (u_char *)(jp + 1);
1071 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1072 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1073 		opt = cp[0];
1074 		if (opt == IPOPT_EOL)
1075 			break;
1076 		if (opt == IPOPT_NOP) {
1077 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1078 			*dp++ = IPOPT_NOP;
1079 			optlen = 1;
1080 			continue;
1081 		}
1082 #ifdef DIAGNOSTIC
1083 		if (cnt < IPOPT_OLEN + sizeof(*cp))
1084 			panic("malformed IPv4 option passed to ip_optcopy");
1085 #endif
1086 		optlen = cp[IPOPT_OLEN];
1087 #ifdef DIAGNOSTIC
1088 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1089 			panic("malformed IPv4 option passed to ip_optcopy");
1090 #endif
1091 		/* bogus lengths should have been caught by ip_dooptions */
1092 		if (optlen > cnt)
1093 			optlen = cnt;
1094 		if (IPOPT_COPIED(opt)) {
1095 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1096 			dp += optlen;
1097 		}
1098 	}
1099 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1100 		*dp++ = IPOPT_EOL;
1101 	return (optlen);
1102 }
1103 
1104 /*
1105  * IP socket option processing.
1106  */
1107 int
1108 ip_ctloutput(op, so, level, optname, mp)
1109 	int op;
1110 	struct socket *so;
1111 	int level, optname;
1112 	struct mbuf **mp;
1113 {
1114 	struct inpcb *inp = sotoinpcb(so);
1115 	struct mbuf *m = *mp;
1116 	int optval = 0;
1117 	int error = 0;
1118 #if defined(IPSEC) || defined(FAST_IPSEC)
1119 	struct proc *p = curproc;	/*XXX*/
1120 #endif
1121 
1122 	if (level != IPPROTO_IP) {
1123 		error = EINVAL;
1124 		if (op == PRCO_SETOPT && *mp)
1125 			(void) m_free(*mp);
1126 	} else switch (op) {
1127 
1128 	case PRCO_SETOPT:
1129 		switch (optname) {
1130 		case IP_OPTIONS:
1131 #ifdef notyet
1132 		case IP_RETOPTS:
1133 			return (ip_pcbopts(optname, &inp->inp_options, m));
1134 #else
1135 			return (ip_pcbopts(&inp->inp_options, m));
1136 #endif
1137 
1138 		case IP_TOS:
1139 		case IP_TTL:
1140 		case IP_RECVOPTS:
1141 		case IP_RECVRETOPTS:
1142 		case IP_RECVDSTADDR:
1143 		case IP_RECVIF:
1144 			if (m == NULL || m->m_len != sizeof(int))
1145 				error = EINVAL;
1146 			else {
1147 				optval = *mtod(m, int *);
1148 				switch (optname) {
1149 
1150 				case IP_TOS:
1151 					inp->inp_ip.ip_tos = optval;
1152 					break;
1153 
1154 				case IP_TTL:
1155 					inp->inp_ip.ip_ttl = optval;
1156 					break;
1157 #define	OPTSET(bit) \
1158 	if (optval) \
1159 		inp->inp_flags |= bit; \
1160 	else \
1161 		inp->inp_flags &= ~bit;
1162 
1163 				case IP_RECVOPTS:
1164 					OPTSET(INP_RECVOPTS);
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 			}
1180 			break;
1181 #undef OPTSET
1182 
1183 		case IP_MULTICAST_IF:
1184 		case IP_MULTICAST_TTL:
1185 		case IP_MULTICAST_LOOP:
1186 		case IP_ADD_MEMBERSHIP:
1187 		case IP_DROP_MEMBERSHIP:
1188 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
1189 			break;
1190 
1191 		case IP_PORTRANGE:
1192 			if (m == 0 || m->m_len != sizeof(int))
1193 				error = EINVAL;
1194 			else {
1195 				optval = *mtod(m, int *);
1196 
1197 				switch (optval) {
1198 
1199 				case IP_PORTRANGE_DEFAULT:
1200 				case IP_PORTRANGE_HIGH:
1201 					inp->inp_flags &= ~(INP_LOWPORT);
1202 					break;
1203 
1204 				case IP_PORTRANGE_LOW:
1205 					inp->inp_flags |= INP_LOWPORT;
1206 					break;
1207 
1208 				default:
1209 					error = EINVAL;
1210 					break;
1211 				}
1212 			}
1213 			break;
1214 
1215 #if defined(IPSEC) || defined(FAST_IPSEC)
1216 		case IP_IPSEC_POLICY:
1217 		{
1218 			caddr_t req = NULL;
1219 			size_t len = 0;
1220 			int priv = 0;
1221 
1222 #ifdef __NetBSD__
1223 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1224 				priv = 0;
1225 			else
1226 				priv = 1;
1227 #else
1228 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
1229 #endif
1230 			if (m) {
1231 				req = mtod(m, caddr_t);
1232 				len = m->m_len;
1233 			}
1234 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1235 			break;
1236 		    }
1237 #endif /*IPSEC*/
1238 
1239 		default:
1240 			error = ENOPROTOOPT;
1241 			break;
1242 		}
1243 		if (m)
1244 			(void)m_free(m);
1245 		break;
1246 
1247 	case PRCO_GETOPT:
1248 		switch (optname) {
1249 		case IP_OPTIONS:
1250 		case IP_RETOPTS:
1251 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1252 			MCLAIM(m, so->so_mowner);
1253 			if (inp->inp_options) {
1254 				m->m_len = inp->inp_options->m_len;
1255 				bcopy(mtod(inp->inp_options, caddr_t),
1256 				    mtod(m, caddr_t), (unsigned)m->m_len);
1257 			} else
1258 				m->m_len = 0;
1259 			break;
1260 
1261 		case IP_TOS:
1262 		case IP_TTL:
1263 		case IP_RECVOPTS:
1264 		case IP_RECVRETOPTS:
1265 		case IP_RECVDSTADDR:
1266 		case IP_RECVIF:
1267 		case IP_ERRORMTU:
1268 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1269 			MCLAIM(m, so->so_mowner);
1270 			m->m_len = sizeof(int);
1271 			switch (optname) {
1272 
1273 			case IP_TOS:
1274 				optval = inp->inp_ip.ip_tos;
1275 				break;
1276 
1277 			case IP_TTL:
1278 				optval = inp->inp_ip.ip_ttl;
1279 				break;
1280 
1281 			case IP_ERRORMTU:
1282 				optval = inp->inp_errormtu;
1283 				break;
1284 
1285 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1286 
1287 			case IP_RECVOPTS:
1288 				optval = OPTBIT(INP_RECVOPTS);
1289 				break;
1290 
1291 			case IP_RECVRETOPTS:
1292 				optval = OPTBIT(INP_RECVRETOPTS);
1293 				break;
1294 
1295 			case IP_RECVDSTADDR:
1296 				optval = OPTBIT(INP_RECVDSTADDR);
1297 				break;
1298 
1299 			case IP_RECVIF:
1300 				optval = OPTBIT(INP_RECVIF);
1301 				break;
1302 			}
1303 			*mtod(m, int *) = optval;
1304 			break;
1305 
1306 #if defined(IPSEC) || defined(FAST_IPSEC)
1307 		case IP_IPSEC_POLICY:
1308 		{
1309 			caddr_t req = NULL;
1310 			size_t len = 0;
1311 
1312 			if (m) {
1313 				req = mtod(m, caddr_t);
1314 				len = m->m_len;
1315 			}
1316 			error = ipsec4_get_policy(inp, req, len, mp);
1317 			break;
1318 		}
1319 #endif /*IPSEC*/
1320 
1321 		case IP_MULTICAST_IF:
1322 		case IP_MULTICAST_TTL:
1323 		case IP_MULTICAST_LOOP:
1324 		case IP_ADD_MEMBERSHIP:
1325 		case IP_DROP_MEMBERSHIP:
1326 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1327 			if (*mp)
1328 				MCLAIM(*mp, so->so_mowner);
1329 			break;
1330 
1331 		case IP_PORTRANGE:
1332 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1333 			MCLAIM(m, so->so_mowner);
1334 			m->m_len = sizeof(int);
1335 
1336 			if (inp->inp_flags & INP_LOWPORT)
1337 				optval = IP_PORTRANGE_LOW;
1338 			else
1339 				optval = IP_PORTRANGE_DEFAULT;
1340 
1341 			*mtod(m, int *) = optval;
1342 			break;
1343 
1344 		default:
1345 			error = ENOPROTOOPT;
1346 			break;
1347 		}
1348 		break;
1349 	}
1350 	return (error);
1351 }
1352 
1353 /*
1354  * Set up IP options in pcb for insertion in output packets.
1355  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1356  * with destination address if source routed.
1357  */
1358 int
1359 #ifdef notyet
1360 ip_pcbopts(optname, pcbopt, m)
1361 	int optname;
1362 #else
1363 ip_pcbopts(pcbopt, m)
1364 #endif
1365 	struct mbuf **pcbopt;
1366 	struct mbuf *m;
1367 {
1368 	int cnt, optlen;
1369 	u_char *cp;
1370 	u_char opt;
1371 
1372 	/* turn off any old options */
1373 	if (*pcbopt)
1374 		(void)m_free(*pcbopt);
1375 	*pcbopt = 0;
1376 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1377 		/*
1378 		 * Only turning off any previous options.
1379 		 */
1380 		if (m)
1381 			(void)m_free(m);
1382 		return (0);
1383 	}
1384 
1385 #ifndef	__vax__
1386 	if (m->m_len % sizeof(int32_t))
1387 		goto bad;
1388 #endif
1389 	/*
1390 	 * IP first-hop destination address will be stored before
1391 	 * actual options; move other options back
1392 	 * and clear it when none present.
1393 	 */
1394 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1395 		goto bad;
1396 	cnt = m->m_len;
1397 	m->m_len += sizeof(struct in_addr);
1398 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1399 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1400 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1401 
1402 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1403 		opt = cp[IPOPT_OPTVAL];
1404 		if (opt == IPOPT_EOL)
1405 			break;
1406 		if (opt == IPOPT_NOP)
1407 			optlen = 1;
1408 		else {
1409 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1410 				goto bad;
1411 			optlen = cp[IPOPT_OLEN];
1412 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1413 				goto bad;
1414 		}
1415 		switch (opt) {
1416 
1417 		default:
1418 			break;
1419 
1420 		case IPOPT_LSRR:
1421 		case IPOPT_SSRR:
1422 			/*
1423 			 * user process specifies route as:
1424 			 *	->A->B->C->D
1425 			 * D must be our final destination (but we can't
1426 			 * check that since we may not have connected yet).
1427 			 * A is first hop destination, which doesn't appear in
1428 			 * actual IP option, but is stored before the options.
1429 			 */
1430 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1431 				goto bad;
1432 			m->m_len -= sizeof(struct in_addr);
1433 			cnt -= sizeof(struct in_addr);
1434 			optlen -= sizeof(struct in_addr);
1435 			cp[IPOPT_OLEN] = optlen;
1436 			/*
1437 			 * Move first hop before start of options.
1438 			 */
1439 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1440 			    sizeof(struct in_addr));
1441 			/*
1442 			 * Then copy rest of options back
1443 			 * to close up the deleted entry.
1444 			 */
1445 			memmove(&cp[IPOPT_OFFSET+1],
1446 			    (caddr_t)(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1447 			    (unsigned)cnt + sizeof(struct in_addr));
1448 			break;
1449 		}
1450 	}
1451 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1452 		goto bad;
1453 	*pcbopt = m;
1454 	return (0);
1455 
1456 bad:
1457 	(void)m_free(m);
1458 	return (EINVAL);
1459 }
1460 
1461 /*
1462  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1463  */
1464 static struct ifnet *
1465 ip_multicast_if(a, ifindexp)
1466 	struct in_addr *a;
1467 	int *ifindexp;
1468 {
1469 	int ifindex;
1470 	struct ifnet *ifp = NULL;
1471 	struct in_ifaddr *ia;
1472 
1473 	if (ifindexp)
1474 		*ifindexp = 0;
1475 	if (ntohl(a->s_addr) >> 24 == 0) {
1476 		ifindex = ntohl(a->s_addr) & 0xffffff;
1477 		if (ifindex < 0 || if_index < ifindex)
1478 			return NULL;
1479 		ifp = ifindex2ifnet[ifindex];
1480 		if (ifindexp)
1481 			*ifindexp = ifindex;
1482 	} else {
1483 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1484 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1485 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1486 				ifp = ia->ia_ifp;
1487 				break;
1488 			}
1489 		}
1490 	}
1491 	return ifp;
1492 }
1493 
1494 /*
1495  * Set the IP multicast options in response to user setsockopt().
1496  */
1497 int
1498 ip_setmoptions(optname, imop, m)
1499 	int optname;
1500 	struct ip_moptions **imop;
1501 	struct mbuf *m;
1502 {
1503 	int error = 0;
1504 	u_char loop;
1505 	int i;
1506 	struct in_addr addr;
1507 	struct ip_mreq *mreq;
1508 	struct ifnet *ifp;
1509 	struct ip_moptions *imo = *imop;
1510 	struct route ro;
1511 	struct sockaddr_in *dst;
1512 	int ifindex;
1513 
1514 	if (imo == NULL) {
1515 		/*
1516 		 * No multicast option buffer attached to the pcb;
1517 		 * allocate one and initialize to default values.
1518 		 */
1519 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1520 		    M_WAITOK);
1521 
1522 		if (imo == NULL)
1523 			return (ENOBUFS);
1524 		*imop = imo;
1525 		imo->imo_multicast_ifp = NULL;
1526 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1527 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1528 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1529 		imo->imo_num_memberships = 0;
1530 	}
1531 
1532 	switch (optname) {
1533 
1534 	case IP_MULTICAST_IF:
1535 		/*
1536 		 * Select the interface for outgoing multicast packets.
1537 		 */
1538 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1539 			error = EINVAL;
1540 			break;
1541 		}
1542 		addr = *(mtod(m, struct in_addr *));
1543 		/*
1544 		 * INADDR_ANY is used to remove a previous selection.
1545 		 * When no interface is selected, a default one is
1546 		 * chosen every time a multicast packet is sent.
1547 		 */
1548 		if (in_nullhost(addr)) {
1549 			imo->imo_multicast_ifp = NULL;
1550 			break;
1551 		}
1552 		/*
1553 		 * The selected interface is identified by its local
1554 		 * IP address.  Find the interface and confirm that
1555 		 * it supports multicasting.
1556 		 */
1557 		ifp = ip_multicast_if(&addr, &ifindex);
1558 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1559 			error = EADDRNOTAVAIL;
1560 			break;
1561 		}
1562 		imo->imo_multicast_ifp = ifp;
1563 		if (ifindex)
1564 			imo->imo_multicast_addr = addr;
1565 		else
1566 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1567 		break;
1568 
1569 	case IP_MULTICAST_TTL:
1570 		/*
1571 		 * Set the IP time-to-live for outgoing multicast packets.
1572 		 */
1573 		if (m == NULL || m->m_len != 1) {
1574 			error = EINVAL;
1575 			break;
1576 		}
1577 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1578 		break;
1579 
1580 	case IP_MULTICAST_LOOP:
1581 		/*
1582 		 * Set the loopback flag for outgoing multicast packets.
1583 		 * Must be zero or one.
1584 		 */
1585 		if (m == NULL || m->m_len != 1 ||
1586 		   (loop = *(mtod(m, u_char *))) > 1) {
1587 			error = EINVAL;
1588 			break;
1589 		}
1590 		imo->imo_multicast_loop = loop;
1591 		break;
1592 
1593 	case IP_ADD_MEMBERSHIP:
1594 		/*
1595 		 * Add a multicast group membership.
1596 		 * Group must be a valid IP multicast address.
1597 		 */
1598 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1599 			error = EINVAL;
1600 			break;
1601 		}
1602 		mreq = mtod(m, struct ip_mreq *);
1603 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1604 			error = EINVAL;
1605 			break;
1606 		}
1607 		/*
1608 		 * If no interface address was provided, use the interface of
1609 		 * the route to the given multicast address.
1610 		 */
1611 		if (in_nullhost(mreq->imr_interface)) {
1612 			bzero((caddr_t)&ro, sizeof(ro));
1613 			ro.ro_rt = NULL;
1614 			dst = satosin(&ro.ro_dst);
1615 			dst->sin_len = sizeof(*dst);
1616 			dst->sin_family = AF_INET;
1617 			dst->sin_addr = mreq->imr_multiaddr;
1618 			rtalloc(&ro);
1619 			if (ro.ro_rt == NULL) {
1620 				error = EADDRNOTAVAIL;
1621 				break;
1622 			}
1623 			ifp = ro.ro_rt->rt_ifp;
1624 			rtfree(ro.ro_rt);
1625 		} else {
1626 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1627 		}
1628 		/*
1629 		 * See if we found an interface, and confirm that it
1630 		 * supports multicast.
1631 		 */
1632 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1633 			error = EADDRNOTAVAIL;
1634 			break;
1635 		}
1636 		/*
1637 		 * See if the membership already exists or if all the
1638 		 * membership slots are full.
1639 		 */
1640 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1641 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1642 			    in_hosteq(imo->imo_membership[i]->inm_addr,
1643 				      mreq->imr_multiaddr))
1644 				break;
1645 		}
1646 		if (i < imo->imo_num_memberships) {
1647 			error = EADDRINUSE;
1648 			break;
1649 		}
1650 		if (i == IP_MAX_MEMBERSHIPS) {
1651 			error = ETOOMANYREFS;
1652 			break;
1653 		}
1654 		/*
1655 		 * Everything looks good; add a new record to the multicast
1656 		 * address list for the given interface.
1657 		 */
1658 		if ((imo->imo_membership[i] =
1659 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1660 			error = ENOBUFS;
1661 			break;
1662 		}
1663 		++imo->imo_num_memberships;
1664 		break;
1665 
1666 	case IP_DROP_MEMBERSHIP:
1667 		/*
1668 		 * Drop a multicast group membership.
1669 		 * Group must be a valid IP multicast address.
1670 		 */
1671 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1672 			error = EINVAL;
1673 			break;
1674 		}
1675 		mreq = mtod(m, struct ip_mreq *);
1676 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1677 			error = EINVAL;
1678 			break;
1679 		}
1680 		/*
1681 		 * If an interface address was specified, get a pointer
1682 		 * to its ifnet structure.
1683 		 */
1684 		if (in_nullhost(mreq->imr_interface))
1685 			ifp = NULL;
1686 		else {
1687 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1688 			if (ifp == NULL) {
1689 				error = EADDRNOTAVAIL;
1690 				break;
1691 			}
1692 		}
1693 		/*
1694 		 * Find the membership in the membership array.
1695 		 */
1696 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1697 			if ((ifp == NULL ||
1698 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1699 			     in_hosteq(imo->imo_membership[i]->inm_addr,
1700 				       mreq->imr_multiaddr))
1701 				break;
1702 		}
1703 		if (i == imo->imo_num_memberships) {
1704 			error = EADDRNOTAVAIL;
1705 			break;
1706 		}
1707 		/*
1708 		 * Give up the multicast address record to which the
1709 		 * membership points.
1710 		 */
1711 		in_delmulti(imo->imo_membership[i]);
1712 		/*
1713 		 * Remove the gap in the membership array.
1714 		 */
1715 		for (++i; i < imo->imo_num_memberships; ++i)
1716 			imo->imo_membership[i-1] = imo->imo_membership[i];
1717 		--imo->imo_num_memberships;
1718 		break;
1719 
1720 	default:
1721 		error = EOPNOTSUPP;
1722 		break;
1723 	}
1724 
1725 	/*
1726 	 * If all options have default values, no need to keep the mbuf.
1727 	 */
1728 	if (imo->imo_multicast_ifp == NULL &&
1729 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1730 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1731 	    imo->imo_num_memberships == 0) {
1732 		free(*imop, M_IPMOPTS);
1733 		*imop = NULL;
1734 	}
1735 
1736 	return (error);
1737 }
1738 
1739 /*
1740  * Return the IP multicast options in response to user getsockopt().
1741  */
1742 int
1743 ip_getmoptions(optname, imo, mp)
1744 	int optname;
1745 	struct ip_moptions *imo;
1746 	struct mbuf **mp;
1747 {
1748 	u_char *ttl;
1749 	u_char *loop;
1750 	struct in_addr *addr;
1751 	struct in_ifaddr *ia;
1752 
1753 	*mp = m_get(M_WAIT, MT_SOOPTS);
1754 
1755 	switch (optname) {
1756 
1757 	case IP_MULTICAST_IF:
1758 		addr = mtod(*mp, struct in_addr *);
1759 		(*mp)->m_len = sizeof(struct in_addr);
1760 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1761 			*addr = zeroin_addr;
1762 		else if (imo->imo_multicast_addr.s_addr) {
1763 			/* return the value user has set */
1764 			*addr = imo->imo_multicast_addr;
1765 		} else {
1766 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1767 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1768 		}
1769 		return (0);
1770 
1771 	case IP_MULTICAST_TTL:
1772 		ttl = mtod(*mp, u_char *);
1773 		(*mp)->m_len = 1;
1774 		*ttl = imo ? imo->imo_multicast_ttl
1775 			   : IP_DEFAULT_MULTICAST_TTL;
1776 		return (0);
1777 
1778 	case IP_MULTICAST_LOOP:
1779 		loop = mtod(*mp, u_char *);
1780 		(*mp)->m_len = 1;
1781 		*loop = imo ? imo->imo_multicast_loop
1782 			    : IP_DEFAULT_MULTICAST_LOOP;
1783 		return (0);
1784 
1785 	default:
1786 		return (EOPNOTSUPP);
1787 	}
1788 }
1789 
1790 /*
1791  * Discard the IP multicast options.
1792  */
1793 void
1794 ip_freemoptions(imo)
1795 	struct ip_moptions *imo;
1796 {
1797 	int i;
1798 
1799 	if (imo != NULL) {
1800 		for (i = 0; i < imo->imo_num_memberships; ++i)
1801 			in_delmulti(imo->imo_membership[i]);
1802 		free(imo, M_IPMOPTS);
1803 	}
1804 }
1805 
1806 /*
1807  * Routine called from ip_output() to loop back a copy of an IP multicast
1808  * packet to the input queue of a specified interface.  Note that this
1809  * calls the output routine of the loopback "driver", but with an interface
1810  * pointer that might NOT be &loif -- easier than replicating that code here.
1811  */
1812 static void
1813 ip_mloopback(ifp, m, dst)
1814 	struct ifnet *ifp;
1815 	struct mbuf *m;
1816 	struct sockaddr_in *dst;
1817 {
1818 	struct ip *ip;
1819 	struct mbuf *copym;
1820 
1821 	copym = m_copy(m, 0, M_COPYALL);
1822 	if (copym != NULL
1823 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1824 		copym = m_pullup(copym, sizeof(struct ip));
1825 	if (copym != NULL) {
1826 		/*
1827 		 * We don't bother to fragment if the IP length is greater
1828 		 * than the interface's MTU.  Can this possibly matter?
1829 		 */
1830 		ip = mtod(copym, struct ip *);
1831 
1832 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1833 			in_delayed_cksum(copym);
1834 			copym->m_pkthdr.csum_flags &=
1835 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1836 		}
1837 
1838 		ip->ip_sum = 0;
1839 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1840 		(void) looutput(ifp, copym, sintosa(dst), NULL);
1841 	}
1842 }
1843