xref: /freebsd/sys/netinet/ip_output.c (revision b52b9d56d4e96089873a75f9e29062eec19fabba)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
34  * $FreeBSD$
35  */
36 
37 #define _IP_VHL
38 
39 #include "opt_ipfw.h"
40 #include "opt_ipdn.h"
41 #include "opt_ipdivert.h"
42 #include "opt_ipfilter.h"
43 #include "opt_ipsec.h"
44 #include "opt_pfil_hooks.h"
45 #include "opt_random_ip_id.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 
56 #include <net/if.h>
57 #include <net/route.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/ip.h>
62 #include <netinet/in_pcb.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip_var.h>
65 
66 #include <machine/in_cksum.h>
67 
68 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
69 
70 #ifdef IPSEC
71 #include <netinet6/ipsec.h>
72 #include <netkey/key.h>
73 #ifdef IPSEC_DEBUG
74 #include <netkey/key_debug.h>
75 #else
76 #define	KEYDEBUG(lev,arg)
77 #endif
78 #endif /*IPSEC*/
79 
80 #include <netinet/ip_fw.h>
81 #include <netinet/ip_dummynet.h>
82 
83 #define print_ip(x, a, y)	 printf("%s %d.%d.%d.%d%s",\
84 				x, (ntohl(a.s_addr)>>24)&0xFF,\
85 				  (ntohl(a.s_addr)>>16)&0xFF,\
86 				  (ntohl(a.s_addr)>>8)&0xFF,\
87 				  (ntohl(a.s_addr))&0xFF, y);
88 
89 u_short ip_id;
90 
91 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
92 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
93 static void	ip_mloopback
94 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
95 static int	ip_getmoptions
96 	(struct sockopt *, struct ip_moptions *);
97 static int	ip_pcbopts(int, struct mbuf **, struct mbuf *);
98 static int	ip_setmoptions
99 	(struct sockopt *, struct ip_moptions **);
100 
101 int	ip_optcopy(struct ip *, struct ip *);
102 
103 
104 extern	struct protosw inetsw[];
105 
106 /*
107  * IP output.  The packet in mbuf chain m contains a skeletal IP
108  * header (with len, off, ttl, proto, tos, src, dst).
109  * The mbuf chain containing the packet will be freed.
110  * The mbuf opt, if present, will not be freed.
111  */
112 int
113 ip_output(m0, opt, ro, flags, imo)
114 	struct mbuf *m0;
115 	struct mbuf *opt;
116 	struct route *ro;
117 	int flags;
118 	struct ip_moptions *imo;
119 {
120 	struct ip *ip, *mhip;
121 	struct ifnet *ifp = NULL;	/* keep compiler happy */
122 	struct mbuf *m;
123 	int hlen = sizeof (struct ip);
124 	int len, off, error = 0;
125 	struct sockaddr_in *dst = NULL;	/* keep compiler happy */
126 	struct in_ifaddr *ia = NULL;
127 	int isbroadcast, sw_csum;
128 	struct in_addr pkt_dst;
129 #ifdef IPSEC
130 	struct route iproute;
131 	struct socket *so = NULL;
132 	struct secpolicy *sp = NULL;
133 #endif
134 	struct ip_fw_args args;
135 	int src_was_INADDR_ANY = 0;	/* as the name says... */
136 #ifdef PFIL_HOOKS
137 	struct packet_filter_hook *pfh;
138 	struct mbuf *m1;
139 	int rv;
140 #endif /* PFIL_HOOKS */
141 
142 	args.eh = NULL;
143 	args.rule = NULL;
144 	args.next_hop = NULL;
145 	args.divert_rule = 0;			/* divert cookie */
146 
147 	/* Grab info from MT_TAG mbufs prepended to the chain. */
148 	for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
149 		switch(m0->m_tag_id) {
150 		default:
151 			printf("ip_output: unrecognised MT_TAG tag %d\n",
152 			    m0->m_tag_id);
153 			break;
154 
155 		case PACKET_TAG_DUMMYNET:
156 			/*
157 			 * the packet was already tagged, so part of the
158 			 * processing was already done, and we need to go down.
159 			 * Get parameters from the header.
160 			 */
161 			args.rule = ((struct dn_pkt *)m0)->rule;
162 			opt = NULL ;
163 			ro = & ( ((struct dn_pkt *)m0)->ro ) ;
164 			imo = NULL ;
165 			dst = ((struct dn_pkt *)m0)->dn_dst ;
166 			ifp = ((struct dn_pkt *)m0)->ifp ;
167 			flags = ((struct dn_pkt *)m0)->flags ;
168 			break;
169 
170 		case PACKET_TAG_DIVERT:
171 			args.divert_rule = (intptr_t)m0->m_data & 0xffff;
172 			break;
173 
174 		case PACKET_TAG_IPFORWARD:
175 			args.next_hop = (struct sockaddr_in *)m0->m_data;
176 			break;
177 		}
178 	}
179 	m = m0;
180 
181 	KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
182 
183 	KASSERT(ro != NULL, ("ip_output: no route, proto %d",
184 	    mtod(m, struct ip *)->ip_p));
185 
186 #ifdef IPSEC
187 	so = ipsec_getsocket(m);
188 	(void)ipsec_setsocket(m, NULL);
189 #endif
190 	if (args.rule != NULL) {	/* dummynet already saw us */
191 		ip = mtod(m, struct ip *);
192 		hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
193 		if (ro->ro_rt)
194 			ia = ifatoia(ro->ro_rt->rt_ifa);
195 		goto sendit;
196 	}
197 
198 	if (opt) {
199 		m = ip_insertoptions(m, opt, &len);
200 		hlen = len;
201 	}
202 	ip = mtod(m, struct ip *);
203 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
204 
205 	/*
206 	 * Fill in IP header.
207 	 */
208 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
209 		ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
210 		ip->ip_off &= IP_DF;
211 #ifdef RANDOM_IP_ID
212 		ip->ip_id = ip_randomid();
213 #else
214 		ip->ip_id = htons(ip_id++);
215 #endif
216 		ipstat.ips_localout++;
217 	} else {
218 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
219 	}
220 
221 	dst = (struct sockaddr_in *)&ro->ro_dst;
222 	/*
223 	 * If there is a cached route,
224 	 * check that it is to the same destination
225 	 * and is still up.  If not, free it and try again.
226 	 * The address family should also be checked in case of sharing the
227 	 * cache with IPv6.
228 	 */
229 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
230 			  dst->sin_family != AF_INET ||
231 			  dst->sin_addr.s_addr != pkt_dst.s_addr)) {
232 		RTFREE(ro->ro_rt);
233 		ro->ro_rt = (struct rtentry *)0;
234 	}
235 	if (ro->ro_rt == 0) {
236 		bzero(dst, sizeof(*dst));
237 		dst->sin_family = AF_INET;
238 		dst->sin_len = sizeof(*dst);
239 		dst->sin_addr = pkt_dst;
240 	}
241 	/*
242 	 * If routing to interface only,
243 	 * short circuit routing lookup.
244 	 */
245 	if (flags & IP_ROUTETOIF) {
246 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
247 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
248 			ipstat.ips_noroute++;
249 			error = ENETUNREACH;
250 			goto bad;
251 		}
252 		ifp = ia->ia_ifp;
253 		ip->ip_ttl = 1;
254 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
255 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
256 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
257 		/*
258 		 * Bypass the normal routing lookup for multicast
259 		 * packets if the interface is specified.
260 		 */
261 		ifp = imo->imo_multicast_ifp;
262 		IFP_TO_IA(ifp, ia);
263 		isbroadcast = 0;	/* fool gcc */
264 	} else {
265 		/*
266 		 * If this is the case, we probably don't want to allocate
267 		 * a protocol-cloned route since we didn't get one from the
268 		 * ULP.  This lets TCP do its thing, while not burdening
269 		 * forwarding or ICMP with the overhead of cloning a route.
270 		 * Of course, we still want to do any cloning requested by
271 		 * the link layer, as this is probably required in all cases
272 		 * for correct operation (as it is for ARP).
273 		 */
274 		if (ro->ro_rt == 0)
275 			rtalloc_ign(ro, RTF_PRCLONING);
276 		if (ro->ro_rt == 0) {
277 			ipstat.ips_noroute++;
278 			error = EHOSTUNREACH;
279 			goto bad;
280 		}
281 		ia = ifatoia(ro->ro_rt->rt_ifa);
282 		ifp = ro->ro_rt->rt_ifp;
283 		ro->ro_rt->rt_use++;
284 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
285 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
286 		if (ro->ro_rt->rt_flags & RTF_HOST)
287 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
288 		else
289 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
290 	}
291 	if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
292 		struct in_multi *inm;
293 
294 		m->m_flags |= M_MCAST;
295 		/*
296 		 * IP destination address is multicast.  Make sure "dst"
297 		 * still points to the address in "ro".  (It may have been
298 		 * changed to point to a gateway address, above.)
299 		 */
300 		dst = (struct sockaddr_in *)&ro->ro_dst;
301 		/*
302 		 * See if the caller provided any multicast options
303 		 */
304 		if (imo != NULL) {
305 			ip->ip_ttl = imo->imo_multicast_ttl;
306 			if (imo->imo_multicast_vif != -1)
307 				ip->ip_src.s_addr =
308 				    ip_mcast_src(imo->imo_multicast_vif);
309 		} else
310 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
311 		/*
312 		 * Confirm that the outgoing interface supports multicast.
313 		 */
314 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
315 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
316 				ipstat.ips_noroute++;
317 				error = ENETUNREACH;
318 				goto bad;
319 			}
320 		}
321 		/*
322 		 * If source address not specified yet, use address
323 		 * of outgoing interface.
324 		 */
325 		if (ip->ip_src.s_addr == INADDR_ANY) {
326 			/* Interface may have no addresses. */
327 			if (ia != NULL)
328 				ip->ip_src = IA_SIN(ia)->sin_addr;
329 		}
330 
331 		if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
332 			/*
333 			 * XXX
334 			 * delayed checksums are not currently
335 			 * compatible with IP multicast routing
336 			 */
337 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
338 				in_delayed_cksum(m);
339 				m->m_pkthdr.csum_flags &=
340 					~CSUM_DELAY_DATA;
341 			}
342 		}
343 		IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
344 		if (inm != NULL &&
345 		   (imo == NULL || imo->imo_multicast_loop)) {
346 			/*
347 			 * If we belong to the destination multicast group
348 			 * on the outgoing interface, and the caller did not
349 			 * forbid loopback, loop back a copy.
350 			 */
351 			ip_mloopback(ifp, m, dst, hlen);
352 		}
353 		else {
354 			/*
355 			 * If we are acting as a multicast router, perform
356 			 * multicast forwarding as if the packet had just
357 			 * arrived on the interface to which we are about
358 			 * to send.  The multicast forwarding function
359 			 * recursively calls this function, using the
360 			 * IP_FORWARDING flag to prevent infinite recursion.
361 			 *
362 			 * Multicasts that are looped back by ip_mloopback(),
363 			 * above, will be forwarded by the ip_input() routine,
364 			 * if necessary.
365 			 */
366 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
367 				/*
368 				 * Check if rsvp daemon is running. If not, don't
369 				 * set ip_moptions. This ensures that the packet
370 				 * is multicast and not just sent down one link
371 				 * as prescribed by rsvpd.
372 				 */
373 				if (!rsvp_on)
374 				  imo = NULL;
375 				if (ip_mforward(ip, ifp, m, imo) != 0) {
376 					m_freem(m);
377 					goto done;
378 				}
379 			}
380 		}
381 
382 		/*
383 		 * Multicasts with a time-to-live of zero may be looped-
384 		 * back, above, but must not be transmitted on a network.
385 		 * Also, multicasts addressed to the loopback interface
386 		 * are not sent -- the above call to ip_mloopback() will
387 		 * loop back a copy if this host actually belongs to the
388 		 * destination group on the loopback interface.
389 		 */
390 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
391 			m_freem(m);
392 			goto done;
393 		}
394 
395 		goto sendit;
396 	}
397 #ifndef notdef
398 	/*
399 	 * If the source address is not specified yet, use the address
400 	 * of the outoing interface. In case, keep note we did that, so
401 	 * if the the firewall changes the next-hop causing the output
402 	 * interface to change, we can fix that.
403 	 */
404 	if (ip->ip_src.s_addr == INADDR_ANY) {
405 		/* Interface may have no addresses. */
406 		if (ia != NULL) {
407 			ip->ip_src = IA_SIN(ia)->sin_addr;
408 			src_was_INADDR_ANY = 1;
409 		}
410 	}
411 #endif /* notdef */
412 	/*
413 	 * Verify that we have any chance at all of being able to queue
414 	 *      the packet or packet fragments
415 	 */
416 	if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
417 		ifp->if_snd.ifq_maxlen) {
418 			error = ENOBUFS;
419 			ipstat.ips_odropped++;
420 			goto bad;
421 	}
422 
423 	/*
424 	 * Look for broadcast address and
425 	 * verify user is allowed to send
426 	 * such a packet.
427 	 */
428 	if (isbroadcast) {
429 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
430 			error = EADDRNOTAVAIL;
431 			goto bad;
432 		}
433 		if ((flags & IP_ALLOWBROADCAST) == 0) {
434 			error = EACCES;
435 			goto bad;
436 		}
437 		/* don't allow broadcast messages to be fragmented */
438 		if ((u_short)ip->ip_len > ifp->if_mtu) {
439 			error = EMSGSIZE;
440 			goto bad;
441 		}
442 		m->m_flags |= M_BCAST;
443 	} else {
444 		m->m_flags &= ~M_BCAST;
445 	}
446 
447 sendit:
448 #ifdef IPSEC
449 	/* get SP for this packet */
450 	if (so == NULL)
451 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
452 	else
453 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
454 
455 	if (sp == NULL) {
456 		ipsecstat.out_inval++;
457 		goto bad;
458 	}
459 
460 	error = 0;
461 
462 	/* check policy */
463 	switch (sp->policy) {
464 	case IPSEC_POLICY_DISCARD:
465 		/*
466 		 * This packet is just discarded.
467 		 */
468 		ipsecstat.out_polvio++;
469 		goto bad;
470 
471 	case IPSEC_POLICY_BYPASS:
472 	case IPSEC_POLICY_NONE:
473 		/* no need to do IPsec. */
474 		goto skip_ipsec;
475 
476 	case IPSEC_POLICY_IPSEC:
477 		if (sp->req == NULL) {
478 			/* acquire a policy */
479 			error = key_spdacquire(sp);
480 			goto bad;
481 		}
482 		break;
483 
484 	case IPSEC_POLICY_ENTRUST:
485 	default:
486 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
487 	}
488     {
489 	struct ipsec_output_state state;
490 	bzero(&state, sizeof(state));
491 	state.m = m;
492 	if (flags & IP_ROUTETOIF) {
493 		state.ro = &iproute;
494 		bzero(&iproute, sizeof(iproute));
495 	} else
496 		state.ro = ro;
497 	state.dst = (struct sockaddr *)dst;
498 
499 	ip->ip_sum = 0;
500 
501 	/*
502 	 * XXX
503 	 * delayed checksums are not currently compatible with IPsec
504 	 */
505 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
506 		in_delayed_cksum(m);
507 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
508 	}
509 
510 	ip->ip_len = htons(ip->ip_len);
511 	ip->ip_off = htons(ip->ip_off);
512 
513 	error = ipsec4_output(&state, sp, flags);
514 
515 	m = state.m;
516 	if (flags & IP_ROUTETOIF) {
517 		/*
518 		 * if we have tunnel mode SA, we may need to ignore
519 		 * IP_ROUTETOIF.
520 		 */
521 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
522 			flags &= ~IP_ROUTETOIF;
523 			ro = state.ro;
524 		}
525 	} else
526 		ro = state.ro;
527 	dst = (struct sockaddr_in *)state.dst;
528 	if (error) {
529 		/* mbuf is already reclaimed in ipsec4_output. */
530 		m0 = NULL;
531 		switch (error) {
532 		case EHOSTUNREACH:
533 		case ENETUNREACH:
534 		case EMSGSIZE:
535 		case ENOBUFS:
536 		case ENOMEM:
537 			break;
538 		default:
539 			printf("ip4_output (ipsec): error code %d\n", error);
540 			/*fall through*/
541 		case ENOENT:
542 			/* don't show these error codes to the user */
543 			error = 0;
544 			break;
545 		}
546 		goto bad;
547 	}
548     }
549 
550 	/* be sure to update variables that are affected by ipsec4_output() */
551 	ip = mtod(m, struct ip *);
552 #ifdef _IP_VHL
553 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
554 #else
555 	hlen = ip->ip_hl << 2;
556 #endif
557 	if (ro->ro_rt == NULL) {
558 		if ((flags & IP_ROUTETOIF) == 0) {
559 			printf("ip_output: "
560 				"can't update route after IPsec processing\n");
561 			error = EHOSTUNREACH;	/*XXX*/
562 			goto bad;
563 		}
564 	} else {
565 		ia = ifatoia(ro->ro_rt->rt_ifa);
566 		ifp = ro->ro_rt->rt_ifp;
567 	}
568 
569 	/* make it flipped, again. */
570 	ip->ip_len = ntohs(ip->ip_len);
571 	ip->ip_off = ntohs(ip->ip_off);
572 skip_ipsec:
573 #endif /*IPSEC*/
574 
575 	/*
576 	 * IpHack's section.
577 	 * - Xlate: translate packet's addr/port (NAT).
578 	 * - Firewall: deny/allow/etc.
579 	 * - Wrap: fake packet's addr/port <unimpl.>
580 	 * - Encapsulate: put it in another IP and send out. <unimp.>
581 	 */
582 #ifdef PFIL_HOOKS
583 	/*
584 	 * Run through list of hooks for output packets.
585 	 */
586 	m1 = m;
587 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
588 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
589 		if (pfh->pfil_func) {
590 			rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
591 			if (rv) {
592 				error = EHOSTUNREACH;
593 				goto done;
594 			}
595 			m = m1;
596 			if (m == NULL)
597 				goto done;
598 			ip = mtod(m, struct ip *);
599 		}
600 #endif /* PFIL_HOOKS */
601 
602 	/*
603 	 * Check with the firewall...
604 	 * but not if we are already being fwd'd from a firewall.
605 	 */
606 	if (fw_enable && IPFW_LOADED && !args.next_hop) {
607 		struct sockaddr_in *old = dst;
608 
609 		args.m = m;
610 		args.next_hop = dst;
611 		args.oif = ifp;
612 		off = ip_fw_chk_ptr(&args);
613 		m = args.m;
614 		dst = args.next_hop;
615 
616                 /*
617 		 * On return we must do the following:
618 		 * m == NULL	-> drop the pkt (old interface, deprecated)
619 		 * (off & IP_FW_PORT_DENY_FLAG)	-> drop the pkt (new interface)
620 		 * 1<=off<= 0xffff		-> DIVERT
621 		 * (off & IP_FW_PORT_DYNT_FLAG)	-> send to a DUMMYNET pipe
622 		 * (off & IP_FW_PORT_TEE_FLAG)	-> TEE the packet
623 		 * dst != old			-> IPFIREWALL_FORWARD
624 		 * off==0, dst==old		-> accept
625 		 * If some of the above modules are not compiled in, then
626 		 * we should't have to check the corresponding condition
627 		 * (because the ipfw control socket should not accept
628 		 * unsupported rules), but better play safe and drop
629 		 * packets in case of doubt.
630 		 */
631 		if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
632 			if (m)
633 				m_freem(m);
634 			error = EACCES;
635 			goto done;
636 		}
637 		ip = mtod(m, struct ip *);
638 		if (off == 0 && dst == old)		/* common case */
639 			goto pass;
640                 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
641 			/*
642 			 * pass the pkt to dummynet. Need to include
643 			 * pipe number, m, ifp, ro, dst because these are
644 			 * not recomputed in the next pass.
645 			 * All other parameters have been already used and
646 			 * so they are not needed anymore.
647 			 * XXX note: if the ifp or ro entry are deleted
648 			 * while a pkt is in dummynet, we are in trouble!
649 			 */
650 			args.ro = ro;
651 			args.dst = dst;
652 			args.flags = flags;
653 
654 			error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
655 				&args);
656 			goto done;
657 		}
658 #ifdef IPDIVERT
659 		if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
660 			struct mbuf *clone = NULL;
661 
662 			/* Clone packet if we're doing a 'tee' */
663 			if ((off & IP_FW_PORT_TEE_FLAG) != 0)
664 				clone = m_dup(m, M_DONTWAIT);
665 
666 			/*
667 			 * XXX
668 			 * delayed checksums are not currently compatible
669 			 * with divert sockets.
670 			 */
671 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
672 				in_delayed_cksum(m);
673 				m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
674 			}
675 
676 			/* Restore packet header fields to original values */
677 			ip->ip_len = htons(ip->ip_len);
678 			ip->ip_off = htons(ip->ip_off);
679 
680 			/* Deliver packet to divert input routine */
681 			divert_packet(m, 0, off & 0xffff, args.divert_rule);
682 
683 			/* If 'tee', continue with original packet */
684 			if (clone != NULL) {
685 				m = clone;
686 				ip = mtod(m, struct ip *);
687 				goto pass;
688 			}
689 			goto done;
690 		}
691 #endif
692 
693 		/* IPFIREWALL_FORWARD */
694 		/*
695 		 * Check dst to make sure it is directly reachable on the
696 		 * interface we previously thought it was.
697 		 * If it isn't (which may be likely in some situations) we have
698 		 * to re-route it (ie, find a route for the next-hop and the
699 		 * associated interface) and set them here. This is nested
700 		 * forwarding which in most cases is undesirable, except where
701 		 * such control is nigh impossible. So we do it here.
702 		 * And I'm babbling.
703 		 */
704 		if (off == 0 && old != dst) { /* FORWARD, dst has changed */
705 #if 0
706 			/*
707 			 * XXX To improve readability, this block should be
708 			 * changed into a function call as below:
709 			 */
710 			error = ip_ipforward(&m, &dst, &ifp);
711 			if (error)
712 				goto bad;
713 			if (m == NULL) /* ip_input consumed the mbuf */
714 				goto done;
715 #else
716 			struct in_ifaddr *ia;
717 
718 			/*
719 			 * XXX sro_fwd below is static, and a pointer
720 			 * to it gets passed to routines downstream.
721 			 * This could have surprisingly bad results in
722 			 * practice, because its content is overwritten
723 			 * by subsequent packets.
724 			 */
725 			/* There must be a better way to do this next line... */
726 			static struct route sro_fwd;
727 			struct route *ro_fwd = &sro_fwd;
728 
729 #if 0
730 			print_ip("IPFIREWALL_FORWARD: New dst ip: ",
731 			    dst->sin_addr, "\n");
732 #endif
733 
734 			/*
735 			 * We need to figure out if we have been forwarded
736 			 * to a local socket. If so, then we should somehow
737 			 * "loop back" to ip_input, and get directed to the
738 			 * PCB as if we had received this packet. This is
739 			 * because it may be dificult to identify the packets
740 			 * you want to forward until they are being output
741 			 * and have selected an interface. (e.g. locally
742 			 * initiated packets) If we used the loopback inteface,
743 			 * we would not be able to control what happens
744 			 * as the packet runs through ip_input() as
745 			 * it is done through a ISR.
746 			 */
747 			LIST_FOREACH(ia,
748 			    INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
749 				/*
750 				 * If the addr to forward to is one
751 				 * of ours, we pretend to
752 				 * be the destination for this packet.
753 				 */
754 				if (IA_SIN(ia)->sin_addr.s_addr ==
755 						 dst->sin_addr.s_addr)
756 					break;
757 			}
758 			if (ia) {	/* tell ip_input "dont filter" */
759 				struct m_hdr tag;
760 
761 				tag.mh_type = MT_TAG;
762 				tag.mh_flags = PACKET_TAG_IPFORWARD;
763 				tag.mh_data = (caddr_t)args.next_hop;
764 				tag.mh_next = m;
765 
766 				if (m->m_pkthdr.rcvif == NULL)
767 					m->m_pkthdr.rcvif = ifunit("lo0");
768 				if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
769 					m->m_pkthdr.csum_flags |=
770 					    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
771 					m0->m_pkthdr.csum_data = 0xffff;
772 				}
773 				m->m_pkthdr.csum_flags |=
774 				    CSUM_IP_CHECKED | CSUM_IP_VALID;
775 				ip->ip_len = htons(ip->ip_len);
776 				ip->ip_off = htons(ip->ip_off);
777 				ip_input((struct mbuf *)&tag);
778 				goto done;
779 			}
780 			/* Some of the logic for this was
781 			 * nicked from above.
782 			 *
783 			 * This rewrites the cached route in a local PCB.
784 			 * Is this what we want to do?
785 			 */
786 			bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
787 
788 			ro_fwd->ro_rt = 0;
789 			rtalloc_ign(ro_fwd, RTF_PRCLONING);
790 
791 			if (ro_fwd->ro_rt == 0) {
792 				ipstat.ips_noroute++;
793 				error = EHOSTUNREACH;
794 				goto bad;
795 			}
796 
797 			ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
798 			ifp = ro_fwd->ro_rt->rt_ifp;
799 			ro_fwd->ro_rt->rt_use++;
800 			if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
801 				dst = (struct sockaddr_in *)
802 					ro_fwd->ro_rt->rt_gateway;
803 			if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
804 				isbroadcast =
805 				    (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
806 			else
807 				isbroadcast = in_broadcast(dst->sin_addr, ifp);
808 			if (ro->ro_rt)
809 				RTFREE(ro->ro_rt);
810 			ro->ro_rt = ro_fwd->ro_rt;
811 			dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
812 
813 #endif	/* ... block to be put into a function */
814 			/*
815 			 * If we added a default src ip earlier,
816 			 * which would have been gotten from the-then
817 			 * interface, do it again, from the new one.
818 			 */
819 			if (src_was_INADDR_ANY)
820 				ip->ip_src = IA_SIN(ia)->sin_addr;
821 			goto pass ;
822 		}
823 
824                 /*
825                  * if we get here, none of the above matches, and
826                  * we have to drop the pkt
827                  */
828 		m_freem(m);
829                 error = EACCES; /* not sure this is the right error msg */
830                 goto done;
831 	}
832 
833 pass:
834 	/* 127/8 must not appear on wire - RFC1122. */
835 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
836 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
837 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
838 			ipstat.ips_badaddr++;
839 			error = EADDRNOTAVAIL;
840 			goto bad;
841 		}
842 	}
843 
844 	m->m_pkthdr.csum_flags |= CSUM_IP;
845 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
846 	if (sw_csum & CSUM_DELAY_DATA) {
847 		in_delayed_cksum(m);
848 		sw_csum &= ~CSUM_DELAY_DATA;
849 	}
850 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
851 
852 	/*
853 	 * If small enough for interface, or the interface will take
854 	 * care of the fragmentation for us, can just send directly.
855 	 */
856 	if ((u_short)ip->ip_len <= ifp->if_mtu ||
857 	    ifp->if_hwassist & CSUM_FRAGMENT) {
858 		ip->ip_len = htons(ip->ip_len);
859 		ip->ip_off = htons(ip->ip_off);
860 		ip->ip_sum = 0;
861 		if (sw_csum & CSUM_DELAY_IP) {
862 			if (ip->ip_vhl == IP_VHL_BORING) {
863 				ip->ip_sum = in_cksum_hdr(ip);
864 			} else {
865 				ip->ip_sum = in_cksum(m, hlen);
866 			}
867 		}
868 
869 		/* Record statistics for this interface address. */
870 		if (!(flags & IP_FORWARDING) && ia) {
871 			ia->ia_ifa.if_opackets++;
872 			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
873 		}
874 
875 #ifdef IPSEC
876 		/* clean ipsec history once it goes out of the node */
877 		ipsec_delaux(m);
878 #endif
879 
880 		error = (*ifp->if_output)(ifp, m,
881 				(struct sockaddr *)dst, ro->ro_rt);
882 		goto done;
883 	}
884 	/*
885 	 * Too large for interface; fragment if possible.
886 	 * Must be able to put at least 8 bytes per fragment.
887 	 */
888 	if (ip->ip_off & IP_DF) {
889 		error = EMSGSIZE;
890 		/*
891 		 * This case can happen if the user changed the MTU
892 		 * of an interface after enabling IP on it.  Because
893 		 * most netifs don't keep track of routes pointing to
894 		 * them, there is no way for one to update all its
895 		 * routes when the MTU is changed.
896 		 */
897 		if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST))
898 		    && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU)
899 		    && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
900 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
901 		}
902 		ipstat.ips_cantfrag++;
903 		goto bad;
904 	}
905 	len = (ifp->if_mtu - hlen) &~ 7;
906 	if (len < 8) {
907 		error = EMSGSIZE;
908 		goto bad;
909 	}
910 
911 	/*
912 	 * if the interface will not calculate checksums on
913 	 * fragmented packets, then do it here.
914 	 */
915 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
916 	    (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
917 		in_delayed_cksum(m);
918 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
919 	}
920 
921 	if (len > PAGE_SIZE) {
922 		/*
923 		 * Fragement large datagrams such that each segment
924 		 * contains a multiple of PAGE_SIZE amount of data,
925 		 * plus headers. This enables a receiver to perform
926 		 * page-flipping zero-copy optimizations.
927 		 */
928 
929 		int newlen;
930 		struct mbuf *mtmp;
931 
932 		for (mtmp = m, off = 0;
933 		     mtmp && ((off + mtmp->m_len) <= ifp->if_mtu);
934 		     mtmp = mtmp->m_next) {
935 			off += mtmp->m_len;
936 		}
937 		/*
938 		 * firstlen (off - hlen) must be aligned on an
939 		 * 8-byte boundary
940 		 */
941 		if (off < hlen)
942 			goto smart_frag_failure;
943 		off = ((off - hlen) & ~7) + hlen;
944 		newlen = (~PAGE_MASK) & ifp->if_mtu;
945 		if ((newlen + sizeof (struct ip)) > ifp->if_mtu) {
946 			/* we failed, go back the default */
947 smart_frag_failure:
948 			newlen = len;
949 			off = hlen + len;
950 		}
951 
952 /*		printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n",
953 		len, hlen, sizeof (struct ip), newlen, off);*/
954 
955 		len = newlen;
956 
957 	} else {
958 		off = hlen + len;
959 	}
960 
961 
962 
963     {
964 	int mhlen, firstlen = off - hlen;
965 	struct mbuf **mnext = &m->m_nextpkt;
966 	int nfrags = 1;
967 
968 	/*
969 	 * Loop through length of segment after first fragment,
970 	 * make new header and copy data of each part and link onto chain.
971 	 */
972 	m0 = m;
973 	mhlen = sizeof (struct ip);
974 	for (; off < (u_short)ip->ip_len; off += len) {
975 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
976 		if (m == 0) {
977 			error = ENOBUFS;
978 			ipstat.ips_odropped++;
979 			goto sendorfree;
980 		}
981 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
982 		m->m_data += max_linkhdr;
983 		mhip = mtod(m, struct ip *);
984 		*mhip = *ip;
985 		if (hlen > sizeof (struct ip)) {
986 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
987 			mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
988 		}
989 		m->m_len = mhlen;
990 		mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
991 		if (off + len >= (u_short)ip->ip_len)
992 			len = (u_short)ip->ip_len - off;
993 		else
994 			mhip->ip_off |= IP_MF;
995 		mhip->ip_len = htons((u_short)(len + mhlen));
996 		m->m_next = m_copy(m0, off, len);
997 		if (m->m_next == 0) {
998 			(void) m_free(m);
999 			error = ENOBUFS;	/* ??? */
1000 			ipstat.ips_odropped++;
1001 			goto sendorfree;
1002 		}
1003 		m->m_pkthdr.len = mhlen + len;
1004 		m->m_pkthdr.rcvif = (struct ifnet *)0;
1005 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1006 		mhip->ip_off = htons(mhip->ip_off);
1007 		mhip->ip_sum = 0;
1008 		if (sw_csum & CSUM_DELAY_IP) {
1009 			if (mhip->ip_vhl == IP_VHL_BORING) {
1010 				mhip->ip_sum = in_cksum_hdr(mhip);
1011 			} else {
1012 				mhip->ip_sum = in_cksum(m, mhlen);
1013 			}
1014 		}
1015 		*mnext = m;
1016 		mnext = &m->m_nextpkt;
1017 		nfrags++;
1018 	}
1019 	ipstat.ips_ofragments += nfrags;
1020 
1021 	/* set first/last markers for fragment chain */
1022 	m->m_flags |= M_LASTFRAG;
1023 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1024 	m0->m_pkthdr.csum_data = nfrags;
1025 
1026 	/*
1027 	 * Update first fragment by trimming what's been copied out
1028 	 * and updating header, then send each fragment (in order).
1029 	 */
1030 	m = m0;
1031 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
1032 	m->m_pkthdr.len = hlen + firstlen;
1033 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
1034 	ip->ip_off |= IP_MF;
1035 	ip->ip_off = htons(ip->ip_off);
1036 	ip->ip_sum = 0;
1037 	if (sw_csum & CSUM_DELAY_IP) {
1038 		if (ip->ip_vhl == IP_VHL_BORING) {
1039 			ip->ip_sum = in_cksum_hdr(ip);
1040 		} else {
1041 			ip->ip_sum = in_cksum(m, hlen);
1042 		}
1043 	}
1044 sendorfree:
1045 	for (m = m0; m; m = m0) {
1046 		m0 = m->m_nextpkt;
1047 		m->m_nextpkt = 0;
1048 #ifdef IPSEC
1049 		/* clean ipsec history once it goes out of the node */
1050 		ipsec_delaux(m);
1051 #endif
1052 		if (error == 0) {
1053 			/* Record statistics for this interface address. */
1054 			if (ia != NULL) {
1055 				ia->ia_ifa.if_opackets++;
1056 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1057 			}
1058 
1059 			error = (*ifp->if_output)(ifp, m,
1060 			    (struct sockaddr *)dst, ro->ro_rt);
1061 		} else
1062 			m_freem(m);
1063 	}
1064 
1065 	if (error == 0)
1066 		ipstat.ips_fragmented++;
1067     }
1068 done:
1069 #ifdef IPSEC
1070 	if (ro == &iproute && ro->ro_rt) {
1071 		RTFREE(ro->ro_rt);
1072 		ro->ro_rt = NULL;
1073 	}
1074 	if (sp != NULL) {
1075 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1076 			printf("DP ip_output call free SP:%p\n", sp));
1077 		key_freesp(sp);
1078 	}
1079 #endif /* IPSEC */
1080 	return (error);
1081 bad:
1082 	m_freem(m);
1083 	goto done;
1084 }
1085 
1086 void
1087 in_delayed_cksum(struct mbuf *m)
1088 {
1089 	struct ip *ip;
1090 	u_short csum, offset;
1091 
1092 	ip = mtod(m, struct ip *);
1093 	offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1094 	csum = in_cksum_skip(m, ip->ip_len, offset);
1095 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1096 		csum = 0xffff;
1097 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1098 
1099 	if (offset + sizeof(u_short) > m->m_len) {
1100 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
1101 		    m->m_len, offset, ip->ip_p);
1102 		/*
1103 		 * XXX
1104 		 * this shouldn't happen, but if it does, the
1105 		 * correct behavior may be to insert the checksum
1106 		 * in the existing chain instead of rearranging it.
1107 		 */
1108 		m = m_pullup(m, offset + sizeof(u_short));
1109 	}
1110 	*(u_short *)(m->m_data + offset) = csum;
1111 }
1112 
1113 /*
1114  * Insert IP options into preformed packet.
1115  * Adjust IP destination as required for IP source routing,
1116  * as indicated by a non-zero in_addr at the start of the options.
1117  *
1118  * XXX This routine assumes that the packet has no options in place.
1119  */
1120 static struct mbuf *
1121 ip_insertoptions(m, opt, phlen)
1122 	register struct mbuf *m;
1123 	struct mbuf *opt;
1124 	int *phlen;
1125 {
1126 	register struct ipoption *p = mtod(opt, struct ipoption *);
1127 	struct mbuf *n;
1128 	register struct ip *ip = mtod(m, struct ip *);
1129 	unsigned optlen;
1130 
1131 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1132 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
1133 		return (m);		/* XXX should fail */
1134 	if (p->ipopt_dst.s_addr)
1135 		ip->ip_dst = p->ipopt_dst;
1136 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1137 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1138 		if (n == 0)
1139 			return (m);
1140 		n->m_pkthdr.rcvif = (struct ifnet *)0;
1141 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1142 		m->m_len -= sizeof(struct ip);
1143 		m->m_data += sizeof(struct ip);
1144 		n->m_next = m;
1145 		m = n;
1146 		m->m_len = optlen + sizeof(struct ip);
1147 		m->m_data += max_linkhdr;
1148 		(void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1149 	} else {
1150 		m->m_data -= optlen;
1151 		m->m_len += optlen;
1152 		m->m_pkthdr.len += optlen;
1153 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1154 	}
1155 	ip = mtod(m, struct ip *);
1156 	bcopy(p->ipopt_list, ip + 1, optlen);
1157 	*phlen = sizeof(struct ip) + optlen;
1158 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1159 	ip->ip_len += optlen;
1160 	return (m);
1161 }
1162 
1163 /*
1164  * Copy options from ip to jp,
1165  * omitting those not copied during fragmentation.
1166  */
1167 int
1168 ip_optcopy(ip, jp)
1169 	struct ip *ip, *jp;
1170 {
1171 	register u_char *cp, *dp;
1172 	int opt, optlen, cnt;
1173 
1174 	cp = (u_char *)(ip + 1);
1175 	dp = (u_char *)(jp + 1);
1176 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1177 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1178 		opt = cp[0];
1179 		if (opt == IPOPT_EOL)
1180 			break;
1181 		if (opt == IPOPT_NOP) {
1182 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1183 			*dp++ = IPOPT_NOP;
1184 			optlen = 1;
1185 			continue;
1186 		}
1187 
1188 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1189 		    ("ip_optcopy: malformed ipv4 option"));
1190 		optlen = cp[IPOPT_OLEN];
1191 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1192 		    ("ip_optcopy: malformed ipv4 option"));
1193 
1194 		/* bogus lengths should have been caught by ip_dooptions */
1195 		if (optlen > cnt)
1196 			optlen = cnt;
1197 		if (IPOPT_COPIED(opt)) {
1198 			bcopy(cp, dp, optlen);
1199 			dp += optlen;
1200 		}
1201 	}
1202 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1203 		*dp++ = IPOPT_EOL;
1204 	return (optlen);
1205 }
1206 
1207 /*
1208  * IP socket option processing.
1209  */
1210 int
1211 ip_ctloutput(so, sopt)
1212 	struct socket *so;
1213 	struct sockopt *sopt;
1214 {
1215 	struct	inpcb *inp = sotoinpcb(so);
1216 	int	error, optval;
1217 
1218 	error = optval = 0;
1219 	if (sopt->sopt_level != IPPROTO_IP) {
1220 		return (EINVAL);
1221 	}
1222 
1223 	switch (sopt->sopt_dir) {
1224 	case SOPT_SET:
1225 		switch (sopt->sopt_name) {
1226 		case IP_OPTIONS:
1227 #ifdef notyet
1228 		case IP_RETOPTS:
1229 #endif
1230 		{
1231 			struct mbuf *m;
1232 			if (sopt->sopt_valsize > MLEN) {
1233 				error = EMSGSIZE;
1234 				break;
1235 			}
1236 			MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1237 			if (m == 0) {
1238 				error = ENOBUFS;
1239 				break;
1240 			}
1241 			m->m_len = sopt->sopt_valsize;
1242 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1243 					    m->m_len);
1244 
1245 			return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1246 					   m));
1247 		}
1248 
1249 		case IP_TOS:
1250 		case IP_TTL:
1251 		case IP_RECVOPTS:
1252 		case IP_RECVRETOPTS:
1253 		case IP_RECVDSTADDR:
1254 		case IP_RECVIF:
1255 		case IP_FAITH:
1256 			error = sooptcopyin(sopt, &optval, sizeof optval,
1257 					    sizeof optval);
1258 			if (error)
1259 				break;
1260 
1261 			switch (sopt->sopt_name) {
1262 			case IP_TOS:
1263 				inp->inp_ip_tos = optval;
1264 				break;
1265 
1266 			case IP_TTL:
1267 				inp->inp_ip_ttl = optval;
1268 				break;
1269 #define	OPTSET(bit) \
1270 	if (optval) \
1271 		inp->inp_flags |= bit; \
1272 	else \
1273 		inp->inp_flags &= ~bit;
1274 
1275 			case IP_RECVOPTS:
1276 				OPTSET(INP_RECVOPTS);
1277 				break;
1278 
1279 			case IP_RECVRETOPTS:
1280 				OPTSET(INP_RECVRETOPTS);
1281 				break;
1282 
1283 			case IP_RECVDSTADDR:
1284 				OPTSET(INP_RECVDSTADDR);
1285 				break;
1286 
1287 			case IP_RECVIF:
1288 				OPTSET(INP_RECVIF);
1289 				break;
1290 
1291 			case IP_FAITH:
1292 				OPTSET(INP_FAITH);
1293 				break;
1294 			}
1295 			break;
1296 #undef OPTSET
1297 
1298 		case IP_MULTICAST_IF:
1299 		case IP_MULTICAST_VIF:
1300 		case IP_MULTICAST_TTL:
1301 		case IP_MULTICAST_LOOP:
1302 		case IP_ADD_MEMBERSHIP:
1303 		case IP_DROP_MEMBERSHIP:
1304 			error = ip_setmoptions(sopt, &inp->inp_moptions);
1305 			break;
1306 
1307 		case IP_PORTRANGE:
1308 			error = sooptcopyin(sopt, &optval, sizeof optval,
1309 					    sizeof optval);
1310 			if (error)
1311 				break;
1312 
1313 			switch (optval) {
1314 			case IP_PORTRANGE_DEFAULT:
1315 				inp->inp_flags &= ~(INP_LOWPORT);
1316 				inp->inp_flags &= ~(INP_HIGHPORT);
1317 				break;
1318 
1319 			case IP_PORTRANGE_HIGH:
1320 				inp->inp_flags &= ~(INP_LOWPORT);
1321 				inp->inp_flags |= INP_HIGHPORT;
1322 				break;
1323 
1324 			case IP_PORTRANGE_LOW:
1325 				inp->inp_flags &= ~(INP_HIGHPORT);
1326 				inp->inp_flags |= INP_LOWPORT;
1327 				break;
1328 
1329 			default:
1330 				error = EINVAL;
1331 				break;
1332 			}
1333 			break;
1334 
1335 #ifdef IPSEC
1336 		case IP_IPSEC_POLICY:
1337 		{
1338 			caddr_t req;
1339 			size_t len = 0;
1340 			int priv;
1341 			struct mbuf *m;
1342 			int optname;
1343 
1344 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1345 				break;
1346 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1347 				break;
1348 			priv = (sopt->sopt_td != NULL &&
1349 				suser(sopt->sopt_td) != 0) ? 0 : 1;
1350 			req = mtod(m, caddr_t);
1351 			len = m->m_len;
1352 			optname = sopt->sopt_name;
1353 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1354 			m_freem(m);
1355 			break;
1356 		}
1357 #endif /*IPSEC*/
1358 
1359 		default:
1360 			error = ENOPROTOOPT;
1361 			break;
1362 		}
1363 		break;
1364 
1365 	case SOPT_GET:
1366 		switch (sopt->sopt_name) {
1367 		case IP_OPTIONS:
1368 		case IP_RETOPTS:
1369 			if (inp->inp_options)
1370 				error = sooptcopyout(sopt,
1371 						     mtod(inp->inp_options,
1372 							  char *),
1373 						     inp->inp_options->m_len);
1374 			else
1375 				sopt->sopt_valsize = 0;
1376 			break;
1377 
1378 		case IP_TOS:
1379 		case IP_TTL:
1380 		case IP_RECVOPTS:
1381 		case IP_RECVRETOPTS:
1382 		case IP_RECVDSTADDR:
1383 		case IP_RECVIF:
1384 		case IP_PORTRANGE:
1385 		case IP_FAITH:
1386 			switch (sopt->sopt_name) {
1387 
1388 			case IP_TOS:
1389 				optval = inp->inp_ip_tos;
1390 				break;
1391 
1392 			case IP_TTL:
1393 				optval = inp->inp_ip_ttl;
1394 				break;
1395 
1396 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1397 
1398 			case IP_RECVOPTS:
1399 				optval = OPTBIT(INP_RECVOPTS);
1400 				break;
1401 
1402 			case IP_RECVRETOPTS:
1403 				optval = OPTBIT(INP_RECVRETOPTS);
1404 				break;
1405 
1406 			case IP_RECVDSTADDR:
1407 				optval = OPTBIT(INP_RECVDSTADDR);
1408 				break;
1409 
1410 			case IP_RECVIF:
1411 				optval = OPTBIT(INP_RECVIF);
1412 				break;
1413 
1414 			case IP_PORTRANGE:
1415 				if (inp->inp_flags & INP_HIGHPORT)
1416 					optval = IP_PORTRANGE_HIGH;
1417 				else if (inp->inp_flags & INP_LOWPORT)
1418 					optval = IP_PORTRANGE_LOW;
1419 				else
1420 					optval = 0;
1421 				break;
1422 
1423 			case IP_FAITH:
1424 				optval = OPTBIT(INP_FAITH);
1425 				break;
1426 			}
1427 			error = sooptcopyout(sopt, &optval, sizeof optval);
1428 			break;
1429 
1430 		case IP_MULTICAST_IF:
1431 		case IP_MULTICAST_VIF:
1432 		case IP_MULTICAST_TTL:
1433 		case IP_MULTICAST_LOOP:
1434 		case IP_ADD_MEMBERSHIP:
1435 		case IP_DROP_MEMBERSHIP:
1436 			error = ip_getmoptions(sopt, inp->inp_moptions);
1437 			break;
1438 
1439 #ifdef IPSEC
1440 		case IP_IPSEC_POLICY:
1441 		{
1442 			struct mbuf *m = NULL;
1443 			caddr_t req = NULL;
1444 			size_t len = 0;
1445 
1446 			if (m != 0) {
1447 				req = mtod(m, caddr_t);
1448 				len = m->m_len;
1449 			}
1450 			error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1451 			if (error == 0)
1452 				error = soopt_mcopyout(sopt, m); /* XXX */
1453 			if (error == 0)
1454 				m_freem(m);
1455 			break;
1456 		}
1457 #endif /*IPSEC*/
1458 
1459 		default:
1460 			error = ENOPROTOOPT;
1461 			break;
1462 		}
1463 		break;
1464 	}
1465 	return (error);
1466 }
1467 
1468 /*
1469  * Set up IP options in pcb for insertion in output packets.
1470  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1471  * with destination address if source routed.
1472  */
1473 static int
1474 ip_pcbopts(optname, pcbopt, m)
1475 	int optname;
1476 	struct mbuf **pcbopt;
1477 	register struct mbuf *m;
1478 {
1479 	register int cnt, optlen;
1480 	register u_char *cp;
1481 	u_char opt;
1482 
1483 	/* turn off any old options */
1484 	if (*pcbopt)
1485 		(void)m_free(*pcbopt);
1486 	*pcbopt = 0;
1487 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1488 		/*
1489 		 * Only turning off any previous options.
1490 		 */
1491 		if (m)
1492 			(void)m_free(m);
1493 		return (0);
1494 	}
1495 
1496 	if (m->m_len % sizeof(int32_t))
1497 		goto bad;
1498 	/*
1499 	 * IP first-hop destination address will be stored before
1500 	 * actual options; move other options back
1501 	 * and clear it when none present.
1502 	 */
1503 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1504 		goto bad;
1505 	cnt = m->m_len;
1506 	m->m_len += sizeof(struct in_addr);
1507 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1508 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1509 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1510 
1511 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1512 		opt = cp[IPOPT_OPTVAL];
1513 		if (opt == IPOPT_EOL)
1514 			break;
1515 		if (opt == IPOPT_NOP)
1516 			optlen = 1;
1517 		else {
1518 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1519 				goto bad;
1520 			optlen = cp[IPOPT_OLEN];
1521 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1522 				goto bad;
1523 		}
1524 		switch (opt) {
1525 
1526 		default:
1527 			break;
1528 
1529 		case IPOPT_LSRR:
1530 		case IPOPT_SSRR:
1531 			/*
1532 			 * user process specifies route as:
1533 			 *	->A->B->C->D
1534 			 * D must be our final destination (but we can't
1535 			 * check that since we may not have connected yet).
1536 			 * A is first hop destination, which doesn't appear in
1537 			 * actual IP option, but is stored before the options.
1538 			 */
1539 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1540 				goto bad;
1541 			m->m_len -= sizeof(struct in_addr);
1542 			cnt -= sizeof(struct in_addr);
1543 			optlen -= sizeof(struct in_addr);
1544 			cp[IPOPT_OLEN] = optlen;
1545 			/*
1546 			 * Move first hop before start of options.
1547 			 */
1548 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1549 			    sizeof(struct in_addr));
1550 			/*
1551 			 * Then copy rest of options back
1552 			 * to close up the deleted entry.
1553 			 */
1554 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1555 			    sizeof(struct in_addr)),
1556 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1557 			    (unsigned)cnt + sizeof(struct in_addr));
1558 			break;
1559 		}
1560 	}
1561 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1562 		goto bad;
1563 	*pcbopt = m;
1564 	return (0);
1565 
1566 bad:
1567 	(void)m_free(m);
1568 	return (EINVAL);
1569 }
1570 
1571 /*
1572  * XXX
1573  * The whole multicast option thing needs to be re-thought.
1574  * Several of these options are equally applicable to non-multicast
1575  * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1576  * standard option (IP_TTL).
1577  */
1578 
1579 /*
1580  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1581  */
1582 static struct ifnet *
1583 ip_multicast_if(a, ifindexp)
1584 	struct in_addr *a;
1585 	int *ifindexp;
1586 {
1587 	int ifindex;
1588 	struct ifnet *ifp;
1589 
1590 	if (ifindexp)
1591 		*ifindexp = 0;
1592 	if (ntohl(a->s_addr) >> 24 == 0) {
1593 		ifindex = ntohl(a->s_addr) & 0xffffff;
1594 		if (ifindex < 0 || if_index < ifindex)
1595 			return NULL;
1596 		ifp = ifnet_byindex(ifindex);
1597 		if (ifindexp)
1598 			*ifindexp = ifindex;
1599 	} else {
1600 		INADDR_TO_IFP(*a, ifp);
1601 	}
1602 	return ifp;
1603 }
1604 
1605 /*
1606  * Set the IP multicast options in response to user setsockopt().
1607  */
1608 static int
1609 ip_setmoptions(sopt, imop)
1610 	struct sockopt *sopt;
1611 	struct ip_moptions **imop;
1612 {
1613 	int error = 0;
1614 	int i;
1615 	struct in_addr addr;
1616 	struct ip_mreq mreq;
1617 	struct ifnet *ifp;
1618 	struct ip_moptions *imo = *imop;
1619 	struct route ro;
1620 	struct sockaddr_in *dst;
1621 	int ifindex;
1622 	int s;
1623 
1624 	if (imo == NULL) {
1625 		/*
1626 		 * No multicast option buffer attached to the pcb;
1627 		 * allocate one and initialize to default values.
1628 		 */
1629 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1630 		    M_WAITOK);
1631 
1632 		if (imo == NULL)
1633 			return (ENOBUFS);
1634 		*imop = imo;
1635 		imo->imo_multicast_ifp = NULL;
1636 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1637 		imo->imo_multicast_vif = -1;
1638 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1639 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1640 		imo->imo_num_memberships = 0;
1641 	}
1642 
1643 	switch (sopt->sopt_name) {
1644 	/* store an index number for the vif you wanna use in the send */
1645 	case IP_MULTICAST_VIF:
1646 		if (legal_vif_num == 0) {
1647 			error = EOPNOTSUPP;
1648 			break;
1649 		}
1650 		error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1651 		if (error)
1652 			break;
1653 		if (!legal_vif_num(i) && (i != -1)) {
1654 			error = EINVAL;
1655 			break;
1656 		}
1657 		imo->imo_multicast_vif = i;
1658 		break;
1659 
1660 	case IP_MULTICAST_IF:
1661 		/*
1662 		 * Select the interface for outgoing multicast packets.
1663 		 */
1664 		error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1665 		if (error)
1666 			break;
1667 		/*
1668 		 * INADDR_ANY is used to remove a previous selection.
1669 		 * When no interface is selected, a default one is
1670 		 * chosen every time a multicast packet is sent.
1671 		 */
1672 		if (addr.s_addr == INADDR_ANY) {
1673 			imo->imo_multicast_ifp = NULL;
1674 			break;
1675 		}
1676 		/*
1677 		 * The selected interface is identified by its local
1678 		 * IP address.  Find the interface and confirm that
1679 		 * it supports multicasting.
1680 		 */
1681 		s = splimp();
1682 		ifp = ip_multicast_if(&addr, &ifindex);
1683 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1684 			splx(s);
1685 			error = EADDRNOTAVAIL;
1686 			break;
1687 		}
1688 		imo->imo_multicast_ifp = ifp;
1689 		if (ifindex)
1690 			imo->imo_multicast_addr = addr;
1691 		else
1692 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1693 		splx(s);
1694 		break;
1695 
1696 	case IP_MULTICAST_TTL:
1697 		/*
1698 		 * Set the IP time-to-live for outgoing multicast packets.
1699 		 * The original multicast API required a char argument,
1700 		 * which is inconsistent with the rest of the socket API.
1701 		 * We allow either a char or an int.
1702 		 */
1703 		if (sopt->sopt_valsize == 1) {
1704 			u_char ttl;
1705 			error = sooptcopyin(sopt, &ttl, 1, 1);
1706 			if (error)
1707 				break;
1708 			imo->imo_multicast_ttl = ttl;
1709 		} else {
1710 			u_int ttl;
1711 			error = sooptcopyin(sopt, &ttl, sizeof ttl,
1712 					    sizeof ttl);
1713 			if (error)
1714 				break;
1715 			if (ttl > 255)
1716 				error = EINVAL;
1717 			else
1718 				imo->imo_multicast_ttl = ttl;
1719 		}
1720 		break;
1721 
1722 	case IP_MULTICAST_LOOP:
1723 		/*
1724 		 * Set the loopback flag for outgoing multicast packets.
1725 		 * Must be zero or one.  The original multicast API required a
1726 		 * char argument, which is inconsistent with the rest
1727 		 * of the socket API.  We allow either a char or an int.
1728 		 */
1729 		if (sopt->sopt_valsize == 1) {
1730 			u_char loop;
1731 			error = sooptcopyin(sopt, &loop, 1, 1);
1732 			if (error)
1733 				break;
1734 			imo->imo_multicast_loop = !!loop;
1735 		} else {
1736 			u_int loop;
1737 			error = sooptcopyin(sopt, &loop, sizeof loop,
1738 					    sizeof loop);
1739 			if (error)
1740 				break;
1741 			imo->imo_multicast_loop = !!loop;
1742 		}
1743 		break;
1744 
1745 	case IP_ADD_MEMBERSHIP:
1746 		/*
1747 		 * Add a multicast group membership.
1748 		 * Group must be a valid IP multicast address.
1749 		 */
1750 		error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1751 		if (error)
1752 			break;
1753 
1754 		if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1755 			error = EINVAL;
1756 			break;
1757 		}
1758 		s = splimp();
1759 		/*
1760 		 * If no interface address was provided, use the interface of
1761 		 * the route to the given multicast address.
1762 		 */
1763 		if (mreq.imr_interface.s_addr == INADDR_ANY) {
1764 			bzero((caddr_t)&ro, sizeof(ro));
1765 			dst = (struct sockaddr_in *)&ro.ro_dst;
1766 			dst->sin_len = sizeof(*dst);
1767 			dst->sin_family = AF_INET;
1768 			dst->sin_addr = mreq.imr_multiaddr;
1769 			rtalloc(&ro);
1770 			if (ro.ro_rt == NULL) {
1771 				error = EADDRNOTAVAIL;
1772 				splx(s);
1773 				break;
1774 			}
1775 			ifp = ro.ro_rt->rt_ifp;
1776 			rtfree(ro.ro_rt);
1777 		}
1778 		else {
1779 			ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1780 		}
1781 
1782 		/*
1783 		 * See if we found an interface, and confirm that it
1784 		 * supports multicast.
1785 		 */
1786 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1787 			error = EADDRNOTAVAIL;
1788 			splx(s);
1789 			break;
1790 		}
1791 		/*
1792 		 * See if the membership already exists or if all the
1793 		 * membership slots are full.
1794 		 */
1795 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1796 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1797 			    imo->imo_membership[i]->inm_addr.s_addr
1798 						== mreq.imr_multiaddr.s_addr)
1799 				break;
1800 		}
1801 		if (i < imo->imo_num_memberships) {
1802 			error = EADDRINUSE;
1803 			splx(s);
1804 			break;
1805 		}
1806 		if (i == IP_MAX_MEMBERSHIPS) {
1807 			error = ETOOMANYREFS;
1808 			splx(s);
1809 			break;
1810 		}
1811 		/*
1812 		 * Everything looks good; add a new record to the multicast
1813 		 * address list for the given interface.
1814 		 */
1815 		if ((imo->imo_membership[i] =
1816 		    in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1817 			error = ENOBUFS;
1818 			splx(s);
1819 			break;
1820 		}
1821 		++imo->imo_num_memberships;
1822 		splx(s);
1823 		break;
1824 
1825 	case IP_DROP_MEMBERSHIP:
1826 		/*
1827 		 * Drop a multicast group membership.
1828 		 * Group must be a valid IP multicast address.
1829 		 */
1830 		error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1831 		if (error)
1832 			break;
1833 
1834 		if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1835 			error = EINVAL;
1836 			break;
1837 		}
1838 
1839 		s = splimp();
1840 		/*
1841 		 * If an interface address was specified, get a pointer
1842 		 * to its ifnet structure.
1843 		 */
1844 		if (mreq.imr_interface.s_addr == INADDR_ANY)
1845 			ifp = NULL;
1846 		else {
1847 			ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1848 			if (ifp == NULL) {
1849 				error = EADDRNOTAVAIL;
1850 				splx(s);
1851 				break;
1852 			}
1853 		}
1854 		/*
1855 		 * Find the membership in the membership array.
1856 		 */
1857 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1858 			if ((ifp == NULL ||
1859 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1860 			     imo->imo_membership[i]->inm_addr.s_addr ==
1861 			     mreq.imr_multiaddr.s_addr)
1862 				break;
1863 		}
1864 		if (i == imo->imo_num_memberships) {
1865 			error = EADDRNOTAVAIL;
1866 			splx(s);
1867 			break;
1868 		}
1869 		/*
1870 		 * Give up the multicast address record to which the
1871 		 * membership points.
1872 		 */
1873 		in_delmulti(imo->imo_membership[i]);
1874 		/*
1875 		 * Remove the gap in the membership array.
1876 		 */
1877 		for (++i; i < imo->imo_num_memberships; ++i)
1878 			imo->imo_membership[i-1] = imo->imo_membership[i];
1879 		--imo->imo_num_memberships;
1880 		splx(s);
1881 		break;
1882 
1883 	default:
1884 		error = EOPNOTSUPP;
1885 		break;
1886 	}
1887 
1888 	/*
1889 	 * If all options have default values, no need to keep the mbuf.
1890 	 */
1891 	if (imo->imo_multicast_ifp == NULL &&
1892 	    imo->imo_multicast_vif == -1 &&
1893 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1894 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1895 	    imo->imo_num_memberships == 0) {
1896 		free(*imop, M_IPMOPTS);
1897 		*imop = NULL;
1898 	}
1899 
1900 	return (error);
1901 }
1902 
1903 /*
1904  * Return the IP multicast options in response to user getsockopt().
1905  */
1906 static int
1907 ip_getmoptions(sopt, imo)
1908 	struct sockopt *sopt;
1909 	register struct ip_moptions *imo;
1910 {
1911 	struct in_addr addr;
1912 	struct in_ifaddr *ia;
1913 	int error, optval;
1914 	u_char coptval;
1915 
1916 	error = 0;
1917 	switch (sopt->sopt_name) {
1918 	case IP_MULTICAST_VIF:
1919 		if (imo != NULL)
1920 			optval = imo->imo_multicast_vif;
1921 		else
1922 			optval = -1;
1923 		error = sooptcopyout(sopt, &optval, sizeof optval);
1924 		break;
1925 
1926 	case IP_MULTICAST_IF:
1927 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1928 			addr.s_addr = INADDR_ANY;
1929 		else if (imo->imo_multicast_addr.s_addr) {
1930 			/* return the value user has set */
1931 			addr = imo->imo_multicast_addr;
1932 		} else {
1933 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1934 			addr.s_addr = (ia == NULL) ? INADDR_ANY
1935 				: IA_SIN(ia)->sin_addr.s_addr;
1936 		}
1937 		error = sooptcopyout(sopt, &addr, sizeof addr);
1938 		break;
1939 
1940 	case IP_MULTICAST_TTL:
1941 		if (imo == 0)
1942 			optval = coptval = IP_DEFAULT_MULTICAST_TTL;
1943 		else
1944 			optval = coptval = imo->imo_multicast_ttl;
1945 		if (sopt->sopt_valsize == 1)
1946 			error = sooptcopyout(sopt, &coptval, 1);
1947 		else
1948 			error = sooptcopyout(sopt, &optval, sizeof optval);
1949 		break;
1950 
1951 	case IP_MULTICAST_LOOP:
1952 		if (imo == 0)
1953 			optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
1954 		else
1955 			optval = coptval = imo->imo_multicast_loop;
1956 		if (sopt->sopt_valsize == 1)
1957 			error = sooptcopyout(sopt, &coptval, 1);
1958 		else
1959 			error = sooptcopyout(sopt, &optval, sizeof optval);
1960 		break;
1961 
1962 	default:
1963 		error = ENOPROTOOPT;
1964 		break;
1965 	}
1966 	return (error);
1967 }
1968 
1969 /*
1970  * Discard the IP multicast options.
1971  */
1972 void
1973 ip_freemoptions(imo)
1974 	register struct ip_moptions *imo;
1975 {
1976 	register int i;
1977 
1978 	if (imo != NULL) {
1979 		for (i = 0; i < imo->imo_num_memberships; ++i)
1980 			in_delmulti(imo->imo_membership[i]);
1981 		free(imo, M_IPMOPTS);
1982 	}
1983 }
1984 
1985 /*
1986  * Routine called from ip_output() to loop back a copy of an IP multicast
1987  * packet to the input queue of a specified interface.  Note that this
1988  * calls the output routine of the loopback "driver", but with an interface
1989  * pointer that might NOT be a loopback interface -- evil, but easier than
1990  * replicating that code here.
1991  */
1992 static void
1993 ip_mloopback(ifp, m, dst, hlen)
1994 	struct ifnet *ifp;
1995 	register struct mbuf *m;
1996 	register struct sockaddr_in *dst;
1997 	int hlen;
1998 {
1999 	register struct ip *ip;
2000 	struct mbuf *copym;
2001 
2002 	copym = m_copy(m, 0, M_COPYALL);
2003 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2004 		copym = m_pullup(copym, hlen);
2005 	if (copym != NULL) {
2006 		/*
2007 		 * We don't bother to fragment if the IP length is greater
2008 		 * than the interface's MTU.  Can this possibly matter?
2009 		 */
2010 		ip = mtod(copym, struct ip *);
2011 		ip->ip_len = htons(ip->ip_len);
2012 		ip->ip_off = htons(ip->ip_off);
2013 		ip->ip_sum = 0;
2014 		if (ip->ip_vhl == IP_VHL_BORING) {
2015 			ip->ip_sum = in_cksum_hdr(ip);
2016 		} else {
2017 			ip->ip_sum = in_cksum(copym, hlen);
2018 		}
2019 		/*
2020 		 * NB:
2021 		 * It's not clear whether there are any lingering
2022 		 * reentrancy problems in other areas which might
2023 		 * be exposed by using ip_input directly (in
2024 		 * particular, everything which modifies the packet
2025 		 * in-place).  Yet another option is using the
2026 		 * protosw directly to deliver the looped back
2027 		 * packet.  For the moment, we'll err on the side
2028 		 * of safety by using if_simloop().
2029 		 */
2030 #if 1 /* XXX */
2031 		if (dst->sin_family != AF_INET) {
2032 			printf("ip_mloopback: bad address family %d\n",
2033 						dst->sin_family);
2034 			dst->sin_family = AF_INET;
2035 		}
2036 #endif
2037 
2038 #ifdef notdef
2039 		copym->m_pkthdr.rcvif = ifp;
2040 		ip_input(copym);
2041 #else
2042 		/* if the checksum hasn't been computed, mark it as valid */
2043 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2044 			copym->m_pkthdr.csum_flags |=
2045 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2046 			copym->m_pkthdr.csum_data = 0xffff;
2047 		}
2048 		if_simloop(ifp, copym, dst->sin_family, 0);
2049 #endif
2050 	}
2051 }
2052