xref: /freebsd/sys/netinet/ip_output.c (revision 11afcc8f9f96d657b8e6f7547c02c1957331fc96)
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  *	$Id: ip_output.c,v 1.79 1998/07/13 12:12:25 bde Exp $
35  */
36 
37 #define _IP_VHL
38 
39 #include "opt_ipfw.h"
40 #include "opt_ipdivert.h"
41 #include "opt_ipfilter.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 
51 #include <net/if.h>
52 #include <net/route.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/ip.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_var.h>
59 #include <netinet/ip_var.h>
60 
61 #ifdef vax
62 #include <machine/mtpr.h>
63 #endif
64 #include <machine/in_cksum.h>
65 
66 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
67 
68 #if !defined(COMPAT_IPFW) || COMPAT_IPFW == 1
69 #undef COMPAT_IPFW
70 #define COMPAT_IPFW 1
71 #else
72 #undef COMPAT_IPFW
73 #endif
74 
75 #ifdef IPFIREWALL_FORWARD_DEBUG
76 #define print_ip(a)	 printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\
77 				 		  (ntohl(a.s_addr)>>16)&0xFF,\
78 						  (ntohl(a.s_addr)>>8)&0xFF,\
79 						  (ntohl(a.s_addr))&0xFF);
80 #endif
81 
82 u_short ip_id;
83 
84 static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *));
85 static void	ip_mloopback
86 	__P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int));
87 static int	ip_getmoptions
88 	__P((int, struct ip_moptions *, struct mbuf **));
89 static int	ip_pcbopts __P((struct mbuf **, struct mbuf *));
90 static int	ip_setmoptions
91 	__P((int, struct ip_moptions **, struct mbuf *));
92 
93 #if defined(IPFILTER_LKM) || defined(IPFILTER)
94 int	ip_optcopy __P((struct ip *, struct ip *));
95 extern int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **));
96 #else
97 static int	ip_optcopy __P((struct ip *, struct ip *));
98 #endif
99 
100 
101 extern	struct protosw inetsw[];
102 
103 /*
104  * IP output.  The packet in mbuf chain m contains a skeletal IP
105  * header (with len, off, ttl, proto, tos, src, dst).
106  * The mbuf chain containing the packet will be freed.
107  * The mbuf opt, if present, will not be freed.
108  */
109 int
110 ip_output(m0, opt, ro, flags, imo)
111 	struct mbuf *m0;
112 	struct mbuf *opt;
113 	struct route *ro;
114 	int flags;
115 	struct ip_moptions *imo;
116 {
117 	struct ip *ip, *mhip;
118 	struct ifnet *ifp;
119 	struct mbuf *m = m0;
120 	int hlen = sizeof (struct ip);
121 	int len, off, error = 0;
122 	struct sockaddr_in *dst;
123 	struct in_ifaddr *ia;
124 	int isbroadcast;
125 #ifdef IPFIREWALL_FORWARD
126 	int fwd_rewrite_src = 0;
127 #endif
128 
129 #ifdef	DIAGNOSTIC
130 	if ((m->m_flags & M_PKTHDR) == 0)
131 		panic("ip_output no HDR");
132 	if (!ro)
133 		panic("ip_output no route, proto = %d",
134 		      mtod(m, struct ip *)->ip_p);
135 #endif
136 	if (opt) {
137 		m = ip_insertoptions(m, opt, &len);
138 		hlen = len;
139 	}
140 	ip = mtod(m, struct ip *);
141 	/*
142 	 * Fill in IP header.
143 	 */
144 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
145 		ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
146 		ip->ip_off &= IP_DF;
147 		ip->ip_id = htons(ip_id++);
148 		ipstat.ips_localout++;
149 	} else {
150 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
151 	}
152 
153 	dst = (struct sockaddr_in *)&ro->ro_dst;
154 	/*
155 	 * If there is a cached route,
156 	 * check that it is to the same destination
157 	 * and is still up.  If not, free it and try again.
158 	 */
159 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
160 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
161 		RTFREE(ro->ro_rt);
162 		ro->ro_rt = (struct rtentry *)0;
163 	}
164 	if (ro->ro_rt == 0) {
165 		dst->sin_family = AF_INET;
166 		dst->sin_len = sizeof(*dst);
167 		dst->sin_addr = ip->ip_dst;
168 	}
169 	/*
170 	 * If routing to interface only,
171 	 * short circuit routing lookup.
172 	 */
173 #define ifatoia(ifa)	((struct in_ifaddr *)(ifa))
174 #define sintosa(sin)	((struct sockaddr *)(sin))
175 	if (flags & IP_ROUTETOIF) {
176 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
177 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
178 			ipstat.ips_noroute++;
179 			error = ENETUNREACH;
180 			goto bad;
181 		}
182 		ifp = ia->ia_ifp;
183 		ip->ip_ttl = 1;
184 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
185 	} else {
186 		/*
187 		 * If this is the case, we probably don't want to allocate
188 		 * a protocol-cloned route since we didn't get one from the
189 		 * ULP.  This lets TCP do its thing, while not burdening
190 		 * forwarding or ICMP with the overhead of cloning a route.
191 		 * Of course, we still want to do any cloning requested by
192 		 * the link layer, as this is probably required in all cases
193 		 * for correct operation (as it is for ARP).
194 		 */
195 		if (ro->ro_rt == 0)
196 			rtalloc_ign(ro, RTF_PRCLONING);
197 		if (ro->ro_rt == 0) {
198 			ipstat.ips_noroute++;
199 			error = EHOSTUNREACH;
200 			goto bad;
201 		}
202 		ia = ifatoia(ro->ro_rt->rt_ifa);
203 		ifp = ro->ro_rt->rt_ifp;
204 		ro->ro_rt->rt_use++;
205 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
206 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
207 		if (ro->ro_rt->rt_flags & RTF_HOST)
208 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
209 		else
210 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
211 	}
212 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
213 		struct in_multi *inm;
214 
215 		m->m_flags |= M_MCAST;
216 		/*
217 		 * IP destination address is multicast.  Make sure "dst"
218 		 * still points to the address in "ro".  (It may have been
219 		 * changed to point to a gateway address, above.)
220 		 */
221 		dst = (struct sockaddr_in *)&ro->ro_dst;
222 		/*
223 		 * See if the caller provided any multicast options
224 		 */
225 		if (imo != NULL) {
226 			ip->ip_ttl = imo->imo_multicast_ttl;
227 			if (imo->imo_multicast_ifp != NULL)
228 				ifp = imo->imo_multicast_ifp;
229 			if (imo->imo_multicast_vif != -1)
230 				ip->ip_src.s_addr =
231 				    ip_mcast_src(imo->imo_multicast_vif);
232 		} else
233 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
234 		/*
235 		 * Confirm that the outgoing interface supports multicast.
236 		 */
237 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
238 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
239 				ipstat.ips_noroute++;
240 				error = ENETUNREACH;
241 				goto bad;
242 			}
243 		}
244 		/*
245 		 * If source address not specified yet, use address
246 		 * of outgoing interface.
247 		 */
248 		if (ip->ip_src.s_addr == INADDR_ANY) {
249 			register struct in_ifaddr *ia1;
250 
251 			for (ia1 = in_ifaddrhead.tqh_first; ia1;
252 			     ia1 = ia1->ia_link.tqe_next)
253 				if (ia1->ia_ifp == ifp) {
254 					ip->ip_src = IA_SIN(ia1)->sin_addr;
255 					break;
256 				}
257 		}
258 
259 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
260 		if (inm != NULL &&
261 		   (imo == NULL || imo->imo_multicast_loop)) {
262 			/*
263 			 * If we belong to the destination multicast group
264 			 * on the outgoing interface, and the caller did not
265 			 * forbid loopback, loop back a copy.
266 			 */
267 			ip_mloopback(ifp, m, dst, hlen);
268 		}
269 		else {
270 			/*
271 			 * If we are acting as a multicast router, perform
272 			 * multicast forwarding as if the packet had just
273 			 * arrived on the interface to which we are about
274 			 * to send.  The multicast forwarding function
275 			 * recursively calls this function, using the
276 			 * IP_FORWARDING flag to prevent infinite recursion.
277 			 *
278 			 * Multicasts that are looped back by ip_mloopback(),
279 			 * above, will be forwarded by the ip_input() routine,
280 			 * if necessary.
281 			 */
282 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
283 				/*
284 				 * Check if rsvp daemon is running. If not, don't
285 				 * set ip_moptions. This ensures that the packet
286 				 * is multicast and not just sent down one link
287 				 * as prescribed by rsvpd.
288 				 */
289 				if (!rsvp_on)
290 				  imo = NULL;
291 				if (ip_mforward(ip, ifp, m, imo) != 0) {
292 					m_freem(m);
293 					goto done;
294 				}
295 			}
296 		}
297 
298 		/*
299 		 * Multicasts with a time-to-live of zero may be looped-
300 		 * back, above, but must not be transmitted on a network.
301 		 * Also, multicasts addressed to the loopback interface
302 		 * are not sent -- the above call to ip_mloopback() will
303 		 * loop back a copy if this host actually belongs to the
304 		 * destination group on the loopback interface.
305 		 */
306 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
307 			m_freem(m);
308 			goto done;
309 		}
310 
311 		goto sendit;
312 	}
313 #ifndef notdef
314 	/*
315 	 * If source address not specified yet, use address
316 	 * of outgoing interface.
317 	 */
318 	if (ip->ip_src.s_addr == INADDR_ANY) {
319 		ip->ip_src = IA_SIN(ia)->sin_addr;
320 #ifdef IPFIREWALL_FORWARD
321 		/* Keep note that we did this - if the firewall changes
322 		 * the next-hop, our interface may change, changing the
323 		 * default source IP. It's a shame so much effort happens
324 		 * twice. Oh well.
325 		 */
326 		fwd_rewrite_src++;
327 #endif /* IPFIREWALL_FORWARD */
328 	}
329 #endif /* notdef */
330 	/*
331 	 * Verify that we have any chance at all of being able to queue
332 	 *      the packet or packet fragments
333 	 */
334 	if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
335 		ifp->if_snd.ifq_maxlen) {
336 			error = ENOBUFS;
337 			goto bad;
338 	}
339 
340 	/*
341 	 * Look for broadcast address and
342 	 * and verify user is allowed to send
343 	 * such a packet.
344 	 */
345 	if (isbroadcast) {
346 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
347 			error = EADDRNOTAVAIL;
348 			goto bad;
349 		}
350 		if ((flags & IP_ALLOWBROADCAST) == 0) {
351 			error = EACCES;
352 			goto bad;
353 		}
354 		/* don't allow broadcast messages to be fragmented */
355 		if ((u_short)ip->ip_len > ifp->if_mtu) {
356 			error = EMSGSIZE;
357 			goto bad;
358 		}
359 		m->m_flags |= M_BCAST;
360 	} else {
361 		m->m_flags &= ~M_BCAST;
362 	}
363 
364 sendit:
365 	/*
366 	 * IpHack's section.
367 	 * - Xlate: translate packet's addr/port (NAT).
368 	 * - Firewall: deny/allow/etc.
369 	 * - Wrap: fake packet's addr/port <unimpl.>
370 	 * - Encapsulate: put it in another IP and send out. <unimp.>
371 	 */
372 #if defined(IPFILTER) || defined(IPFILTER_LKM)
373 	if (fr_checkp) {
374 		struct  mbuf    *m1 = m;
375 
376 		if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1)
377 			goto done;
378 		ip = mtod(m = m1, struct ip *);
379 	}
380 #endif
381 
382 #ifdef COMPAT_IPFW
383         if (ip_nat_ptr && !(*ip_nat_ptr)(&ip, &m, ifp, IP_NAT_OUT)) {
384 		error = EACCES;
385 		goto done;
386 	}
387 
388 	/*
389 	 * Check with the firewall...
390 	 */
391 	if (ip_fw_chk_ptr) {
392 #ifdef IPFIREWALL_FORWARD
393 		struct sockaddr_in *old = dst;
394 #endif
395 #ifdef IPDIVERT
396 		ip_divert_port = (*ip_fw_chk_ptr)(&ip,
397 		    hlen, ifp, &ip_divert_cookie, &m, &dst);
398 		if (ip_divert_port) {		/* Divert packet */
399 			(*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, 0);
400 			goto done;
401 		}
402 #else	/* !IPDIVERT */
403 		u_int16_t 	dummy = 0;
404 		/* If ipfw says divert, we have to just drop packet */
405 		if ((*ip_fw_chk_ptr)(&ip, hlen, ifp, &dummy, &m, &dst)) {
406 			m_freem(m);
407 			goto done;
408 		}
409 #endif	/* !IPDIVERT */
410 		if (!m) {
411 			error = EACCES;
412 			goto done;
413 		}
414 #ifdef IPFIREWALL_FORWARD
415 		/* Here we check dst to make sure it's directly reachable on the
416 		 * interface we previously thought it was.
417 		 * If it isn't (which may be likely in some situations) we have
418 		 * to re-route it (ie, find a route for the next-hop and the
419 		 * associated interface) and set them here. This is nested
420 		 * forwarding which in most cases is undesirable, except where
421 		 * such control is nigh impossible. So we do it here.
422 		 * And I'm babbling.
423 		 */
424 		if (old != dst) {
425 			struct in_ifaddr *ia;
426 
427 			/* It's changed... */
428 			/* There must be a better way to do this next line... */
429 			static struct route sro_fwd, *ro_fwd = &sro_fwd;
430 #ifdef IPFIREWALL_FORWARD_DEBUG
431 			printf("IPFIREWALL_FORWARD: New dst ip: ");
432 			print_ip(dst->sin_addr);
433 			printf("\n");
434 #endif
435 			/*
436 			 * We need to figure out if we have been forwarded
437 			 * to a local socket. If so then we should somehow
438 			 * "loop back" to ip_input, and get directed to the
439 			 * PCB as if we had received this packet. This is
440 			 * because it may be dificult to identify the packets
441 			 * you want to forward until they are being output
442 			 * and have selected an interface. (e.g. locally
443 			 * initiated packets) If we used the loopback inteface,
444 			 * we would not be able to control what happens
445 			 * as the packet runs through ip_input() as
446 			 * it is done through a ISR.
447 			 */
448 			for (ia = TAILQ_FIRST(&in_ifaddrhead); ia;
449 					ia = TAILQ_NEXT(ia, ia_link)) {
450 				/*
451 				 * If the addr to forward to is one
452 				 * of ours, we pretend to
453 				 * be the destination for this packet.
454 				 */
455 				if (IA_SIN(ia)->sin_addr.s_addr ==
456 						 dst->sin_addr.s_addr)
457 					break;
458 			}
459 			if (ia) {
460 				/* tell ip_input "dont filter" */
461 				ip_fw_fwd_addr = dst;
462 				if (m->m_pkthdr.rcvif == NULL)
463 					m->m_pkthdr.rcvif = ifunit("lo0");
464 				ip->ip_len = htons((u_short)ip->ip_len);
465 				ip->ip_off = htons((u_short)ip->ip_off);
466 				ip->ip_sum = 0;
467 				if (ip->ip_vhl == IP_VHL_BORING) {
468 					ip->ip_sum = in_cksum_hdr(ip);
469 				} else {
470 					ip->ip_sum = in_cksum(m, hlen);
471 				}
472 				ip_input(m);
473 				goto done;
474 			}
475 			/* Some of the logic for this was
476 			 * nicked from above.
477 			 *
478 			 * This rewrites the cached route in a local PCB.
479 			 * Is this what we want to do?
480 			 */
481 			bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
482 
483 			ro_fwd->ro_rt = 0;
484 			rtalloc_ign(ro_fwd, RTF_PRCLONING);
485 
486 			if (ro_fwd->ro_rt == 0) {
487 				ipstat.ips_noroute++;
488 				error = EHOSTUNREACH;
489 				goto bad;
490 			}
491 
492 			ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
493 			ifp = ro_fwd->ro_rt->rt_ifp;
494 			ro_fwd->ro_rt->rt_use++;
495 			if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
496 				dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway;
497 			if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
498 				isbroadcast =
499 				    (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
500 			else
501 				isbroadcast = in_broadcast(dst->sin_addr, ifp);
502 			RTFREE(ro->ro_rt);
503 			ro->ro_rt = ro_fwd->ro_rt;
504 			dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
505 
506 			/*
507 			 * If we added a default src ip earlier,
508 			 * which would have been gotten from the-then
509 			 * interface, do it again, from the new one.
510 			 */
511 			if (fwd_rewrite_src)
512 				ip->ip_src = IA_SIN(ia)->sin_addr;
513 		}
514 #endif /* IPFIREWALL_FORWARD */
515 	}
516 #endif /* COMPAT_IPFW */
517 
518 
519 	/*
520 	 * If small enough for interface, can just send directly.
521 	 */
522 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
523 		ip->ip_len = htons((u_short)ip->ip_len);
524 		ip->ip_off = htons((u_short)ip->ip_off);
525 		ip->ip_sum = 0;
526 		if (ip->ip_vhl == IP_VHL_BORING) {
527 			ip->ip_sum = in_cksum_hdr(ip);
528 		} else {
529 			ip->ip_sum = in_cksum(m, hlen);
530 		}
531 		error = (*ifp->if_output)(ifp, m,
532 				(struct sockaddr *)dst, ro->ro_rt);
533 		goto done;
534 	}
535 	/*
536 	 * Too large for interface; fragment if possible.
537 	 * Must be able to put at least 8 bytes per fragment.
538 	 */
539 	if (ip->ip_off & IP_DF) {
540 		error = EMSGSIZE;
541 		/*
542 		 * This case can happen if the user changed the MTU
543 		 * of an interface after enabling IP on it.  Because
544 		 * most netifs don't keep track of routes pointing to
545 		 * them, there is no way for one to update all its
546 		 * routes when the MTU is changed.
547 		 */
548 		if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST))
549 		    && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU)
550 		    && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
551 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
552 		}
553 		ipstat.ips_cantfrag++;
554 		goto bad;
555 	}
556 	len = (ifp->if_mtu - hlen) &~ 7;
557 	if (len < 8) {
558 		error = EMSGSIZE;
559 		goto bad;
560 	}
561 
562     {
563 	int mhlen, firstlen = len;
564 	struct mbuf **mnext = &m->m_nextpkt;
565 
566 	/*
567 	 * Loop through length of segment after first fragment,
568 	 * make new header and copy data of each part and link onto chain.
569 	 */
570 	m0 = m;
571 	mhlen = sizeof (struct ip);
572 	for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
573 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
574 		if (m == 0) {
575 			error = ENOBUFS;
576 			ipstat.ips_odropped++;
577 			goto sendorfree;
578 		}
579 		m->m_data += max_linkhdr;
580 		mhip = mtod(m, struct ip *);
581 		*mhip = *ip;
582 		if (hlen > sizeof (struct ip)) {
583 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
584 			mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
585 		}
586 		m->m_len = mhlen;
587 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
588 		if (ip->ip_off & IP_MF)
589 			mhip->ip_off |= IP_MF;
590 		if (off + len >= (u_short)ip->ip_len)
591 			len = (u_short)ip->ip_len - off;
592 		else
593 			mhip->ip_off |= IP_MF;
594 		mhip->ip_len = htons((u_short)(len + mhlen));
595 		m->m_next = m_copy(m0, off, len);
596 		if (m->m_next == 0) {
597 			(void) m_free(m);
598 			error = ENOBUFS;	/* ??? */
599 			ipstat.ips_odropped++;
600 			goto sendorfree;
601 		}
602 		m->m_pkthdr.len = mhlen + len;
603 		m->m_pkthdr.rcvif = (struct ifnet *)0;
604 		mhip->ip_off = htons((u_short)mhip->ip_off);
605 		mhip->ip_sum = 0;
606 		if (mhip->ip_vhl == IP_VHL_BORING) {
607 			mhip->ip_sum = in_cksum_hdr(mhip);
608 		} else {
609 			mhip->ip_sum = in_cksum(m, mhlen);
610 		}
611 		*mnext = m;
612 		mnext = &m->m_nextpkt;
613 		ipstat.ips_ofragments++;
614 	}
615 	/*
616 	 * Update first fragment by trimming what's been copied out
617 	 * and updating header, then send each fragment (in order).
618 	 */
619 	m = m0;
620 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
621 	m->m_pkthdr.len = hlen + firstlen;
622 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
623 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
624 	ip->ip_sum = 0;
625 	if (ip->ip_vhl == IP_VHL_BORING) {
626 		ip->ip_sum = in_cksum_hdr(ip);
627 	} else {
628 		ip->ip_sum = in_cksum(m, hlen);
629 	}
630 sendorfree:
631 	for (m = m0; m; m = m0) {
632 		m0 = m->m_nextpkt;
633 		m->m_nextpkt = 0;
634 		if (error == 0)
635 			error = (*ifp->if_output)(ifp, m,
636 			    (struct sockaddr *)dst, ro->ro_rt);
637 		else
638 			m_freem(m);
639 	}
640 
641 	if (error == 0)
642 		ipstat.ips_fragmented++;
643     }
644 done:
645 	return (error);
646 bad:
647 	m_freem(m0);
648 	goto done;
649 }
650 
651 /*
652  * Insert IP options into preformed packet.
653  * Adjust IP destination as required for IP source routing,
654  * as indicated by a non-zero in_addr at the start of the options.
655  *
656  * XXX This routine assumes that the packet has no options in place.
657  */
658 static struct mbuf *
659 ip_insertoptions(m, opt, phlen)
660 	register struct mbuf *m;
661 	struct mbuf *opt;
662 	int *phlen;
663 {
664 	register struct ipoption *p = mtod(opt, struct ipoption *);
665 	struct mbuf *n;
666 	register struct ip *ip = mtod(m, struct ip *);
667 	unsigned optlen;
668 
669 	optlen = opt->m_len - sizeof(p->ipopt_dst);
670 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
671 		return (m);		/* XXX should fail */
672 	if (p->ipopt_dst.s_addr)
673 		ip->ip_dst = p->ipopt_dst;
674 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
675 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
676 		if (n == 0)
677 			return (m);
678 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
679 		m->m_len -= sizeof(struct ip);
680 		m->m_data += sizeof(struct ip);
681 		n->m_next = m;
682 		m = n;
683 		m->m_len = optlen + sizeof(struct ip);
684 		m->m_data += max_linkhdr;
685 		(void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
686 	} else {
687 		m->m_data -= optlen;
688 		m->m_len += optlen;
689 		m->m_pkthdr.len += optlen;
690 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
691 	}
692 	ip = mtod(m, struct ip *);
693 	bcopy(p->ipopt_list, ip + 1, optlen);
694 	*phlen = sizeof(struct ip) + optlen;
695 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
696 	ip->ip_len += optlen;
697 	return (m);
698 }
699 
700 /*
701  * Copy options from ip to jp,
702  * omitting those not copied during fragmentation.
703  */
704 #if !defined(IPFILTER) && !defined(IPFILTER_LKM)
705 static
706 #endif
707 int
708 ip_optcopy(ip, jp)
709 	struct ip *ip, *jp;
710 {
711 	register u_char *cp, *dp;
712 	int opt, optlen, cnt;
713 
714 	cp = (u_char *)(ip + 1);
715 	dp = (u_char *)(jp + 1);
716 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
717 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
718 		opt = cp[0];
719 		if (opt == IPOPT_EOL)
720 			break;
721 		if (opt == IPOPT_NOP) {
722 			/* Preserve for IP mcast tunnel's LSRR alignment. */
723 			*dp++ = IPOPT_NOP;
724 			optlen = 1;
725 			continue;
726 		} else
727 			optlen = cp[IPOPT_OLEN];
728 		/* bogus lengths should have been caught by ip_dooptions */
729 		if (optlen > cnt)
730 			optlen = cnt;
731 		if (IPOPT_COPIED(opt)) {
732 			bcopy(cp, dp, optlen);
733 			dp += optlen;
734 		}
735 	}
736 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
737 		*dp++ = IPOPT_EOL;
738 	return (optlen);
739 }
740 
741 /*
742  * IP socket option processing.
743  */
744 int
745 ip_ctloutput(op, so, level, optname, mp, p)
746 	int op;
747 	struct socket *so;
748 	int level, optname;
749 	struct mbuf **mp;
750 	struct proc *p;
751 {
752 	register struct inpcb *inp = sotoinpcb(so);
753 	register struct mbuf *m = *mp;
754 	register int optval = 0;
755 	int error = 0;
756 
757 	if (level != IPPROTO_IP) {
758 		error = EINVAL;
759 		if (op == PRCO_SETOPT && *mp)
760 			(void) m_free(*mp);
761 	} else switch (op) {
762 
763 	case PRCO_SETOPT:
764 		switch (optname) {
765 		case IP_OPTIONS:
766 #ifdef notyet
767 		case IP_RETOPTS:
768 			return (ip_pcbopts(optname, &inp->inp_options, m));
769 #else
770 			return (ip_pcbopts(&inp->inp_options, m));
771 #endif
772 
773 		case IP_TOS:
774 		case IP_TTL:
775 		case IP_RECVOPTS:
776 		case IP_RECVRETOPTS:
777 		case IP_RECVDSTADDR:
778 		case IP_RECVIF:
779 			if (m == 0 || m->m_len != sizeof(int))
780 				error = EINVAL;
781 			else {
782 				optval = *mtod(m, int *);
783 				switch (optname) {
784 
785 				case IP_TOS:
786 					inp->inp_ip_tos = optval;
787 					break;
788 
789 				case IP_TTL:
790 					inp->inp_ip_ttl = optval;
791 					break;
792 #define	OPTSET(bit) \
793 	if (optval) \
794 		inp->inp_flags |= bit; \
795 	else \
796 		inp->inp_flags &= ~bit;
797 
798 				case IP_RECVOPTS:
799 					OPTSET(INP_RECVOPTS);
800 					break;
801 
802 				case IP_RECVRETOPTS:
803 					OPTSET(INP_RECVRETOPTS);
804 					break;
805 
806 				case IP_RECVDSTADDR:
807 					OPTSET(INP_RECVDSTADDR);
808 					break;
809 
810 				case IP_RECVIF:
811 					OPTSET(INP_RECVIF);
812 					break;
813 				}
814 			}
815 			break;
816 #undef OPTSET
817 
818 		case IP_MULTICAST_IF:
819 		case IP_MULTICAST_VIF:
820 		case IP_MULTICAST_TTL:
821 		case IP_MULTICAST_LOOP:
822 		case IP_ADD_MEMBERSHIP:
823 		case IP_DROP_MEMBERSHIP:
824 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
825 			break;
826 
827 		case IP_PORTRANGE:
828 			if (m == 0 || m->m_len != sizeof(int))
829 				error = EINVAL;
830 			else {
831 				optval = *mtod(m, int *);
832 
833 				switch (optval) {
834 
835 				case IP_PORTRANGE_DEFAULT:
836 					inp->inp_flags &= ~(INP_LOWPORT);
837 					inp->inp_flags &= ~(INP_HIGHPORT);
838 					break;
839 
840 				case IP_PORTRANGE_HIGH:
841 					inp->inp_flags &= ~(INP_LOWPORT);
842 					inp->inp_flags |= INP_HIGHPORT;
843 					break;
844 
845 				case IP_PORTRANGE_LOW:
846 					inp->inp_flags &= ~(INP_HIGHPORT);
847 					inp->inp_flags |= INP_LOWPORT;
848 					break;
849 
850 				default:
851 					error = EINVAL;
852 					break;
853 				}
854 			}
855 			break;
856 
857 		default:
858 			error = ENOPROTOOPT;
859 			break;
860 		}
861 		if (m)
862 			(void)m_free(m);
863 		break;
864 
865 	case PRCO_GETOPT:
866 		switch (optname) {
867 		case IP_OPTIONS:
868 		case IP_RETOPTS:
869 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
870 			if (inp->inp_options) {
871 				m->m_len = inp->inp_options->m_len;
872 				bcopy(mtod(inp->inp_options, void *),
873 				    mtod(m, void *), m->m_len);
874 			} else
875 				m->m_len = 0;
876 			break;
877 
878 		case IP_TOS:
879 		case IP_TTL:
880 		case IP_RECVOPTS:
881 		case IP_RECVRETOPTS:
882 		case IP_RECVDSTADDR:
883 		case IP_RECVIF:
884 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
885 			m->m_len = sizeof(int);
886 			switch (optname) {
887 
888 			case IP_TOS:
889 				optval = inp->inp_ip_tos;
890 				break;
891 
892 			case IP_TTL:
893 				optval = inp->inp_ip_ttl;
894 				break;
895 
896 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
897 
898 			case IP_RECVOPTS:
899 				optval = OPTBIT(INP_RECVOPTS);
900 				break;
901 
902 			case IP_RECVRETOPTS:
903 				optval = OPTBIT(INP_RECVRETOPTS);
904 				break;
905 
906 			case IP_RECVDSTADDR:
907 				optval = OPTBIT(INP_RECVDSTADDR);
908 				break;
909 
910 			case IP_RECVIF:
911 				optval = OPTBIT(INP_RECVIF);
912 				break;
913 			}
914 			*mtod(m, int *) = optval;
915 			break;
916 
917 		case IP_MULTICAST_IF:
918 		case IP_MULTICAST_VIF:
919 		case IP_MULTICAST_TTL:
920 		case IP_MULTICAST_LOOP:
921 		case IP_ADD_MEMBERSHIP:
922 		case IP_DROP_MEMBERSHIP:
923 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
924 			break;
925 
926 		case IP_PORTRANGE:
927 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
928 			m->m_len = sizeof(int);
929 
930 			if (inp->inp_flags & INP_HIGHPORT)
931 				optval = IP_PORTRANGE_HIGH;
932 			else if (inp->inp_flags & INP_LOWPORT)
933 				optval = IP_PORTRANGE_LOW;
934 			else
935 				optval = 0;
936 
937 			*mtod(m, int *) = optval;
938 			break;
939 
940 		default:
941 			error = ENOPROTOOPT;
942 			break;
943 		}
944 		break;
945 	}
946 	return (error);
947 }
948 
949 /*
950  * Set up IP options in pcb for insertion in output packets.
951  * Store in mbuf with pointer in pcbopt, adding pseudo-option
952  * with destination address if source routed.
953  */
954 static int
955 #ifdef notyet
956 ip_pcbopts(optname, pcbopt, m)
957 	int optname;
958 #else
959 ip_pcbopts(pcbopt, m)
960 #endif
961 	struct mbuf **pcbopt;
962 	register struct mbuf *m;
963 {
964 	register int cnt, optlen;
965 	register u_char *cp;
966 	u_char opt;
967 
968 	/* turn off any old options */
969 	if (*pcbopt)
970 		(void)m_free(*pcbopt);
971 	*pcbopt = 0;
972 	if (m == (struct mbuf *)0 || m->m_len == 0) {
973 		/*
974 		 * Only turning off any previous options.
975 		 */
976 		if (m)
977 			(void)m_free(m);
978 		return (0);
979 	}
980 
981 #ifndef	vax
982 	if (m->m_len % sizeof(int32_t))
983 		goto bad;
984 #endif
985 	/*
986 	 * IP first-hop destination address will be stored before
987 	 * actual options; move other options back
988 	 * and clear it when none present.
989 	 */
990 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
991 		goto bad;
992 	cnt = m->m_len;
993 	m->m_len += sizeof(struct in_addr);
994 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
995 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
996 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
997 
998 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
999 		opt = cp[IPOPT_OPTVAL];
1000 		if (opt == IPOPT_EOL)
1001 			break;
1002 		if (opt == IPOPT_NOP)
1003 			optlen = 1;
1004 		else {
1005 			optlen = cp[IPOPT_OLEN];
1006 			if (optlen <= IPOPT_OLEN || optlen > cnt)
1007 				goto bad;
1008 		}
1009 		switch (opt) {
1010 
1011 		default:
1012 			break;
1013 
1014 		case IPOPT_LSRR:
1015 		case IPOPT_SSRR:
1016 			/*
1017 			 * user process specifies route as:
1018 			 *	->A->B->C->D
1019 			 * D must be our final destination (but we can't
1020 			 * check that since we may not have connected yet).
1021 			 * A is first hop destination, which doesn't appear in
1022 			 * actual IP option, but is stored before the options.
1023 			 */
1024 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1025 				goto bad;
1026 			m->m_len -= sizeof(struct in_addr);
1027 			cnt -= sizeof(struct in_addr);
1028 			optlen -= sizeof(struct in_addr);
1029 			cp[IPOPT_OLEN] = optlen;
1030 			/*
1031 			 * Move first hop before start of options.
1032 			 */
1033 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1034 			    sizeof(struct in_addr));
1035 			/*
1036 			 * Then copy rest of options back
1037 			 * to close up the deleted entry.
1038 			 */
1039 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1040 			    sizeof(struct in_addr)),
1041 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1042 			    (unsigned)cnt + sizeof(struct in_addr));
1043 			break;
1044 		}
1045 	}
1046 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1047 		goto bad;
1048 	*pcbopt = m;
1049 	return (0);
1050 
1051 bad:
1052 	(void)m_free(m);
1053 	return (EINVAL);
1054 }
1055 
1056 /*
1057  * Set the IP multicast options in response to user setsockopt().
1058  */
1059 static int
1060 ip_setmoptions(optname, imop, m)
1061 	int optname;
1062 	struct ip_moptions **imop;
1063 	struct mbuf *m;
1064 {
1065 	register int error = 0;
1066 	u_char loop;
1067 	register int i;
1068 	struct in_addr addr;
1069 	register struct ip_mreq *mreq;
1070 	register struct ifnet *ifp;
1071 	register struct ip_moptions *imo = *imop;
1072 	struct route ro;
1073 	register struct sockaddr_in *dst;
1074 	int s;
1075 
1076 	if (imo == NULL) {
1077 		/*
1078 		 * No multicast option buffer attached to the pcb;
1079 		 * allocate one and initialize to default values.
1080 		 */
1081 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1082 		    M_WAITOK);
1083 
1084 		if (imo == NULL)
1085 			return (ENOBUFS);
1086 		*imop = imo;
1087 		imo->imo_multicast_ifp = NULL;
1088 		imo->imo_multicast_vif = -1;
1089 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1090 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1091 		imo->imo_num_memberships = 0;
1092 	}
1093 
1094 	switch (optname) {
1095 	/* store an index number for the vif you wanna use in the send */
1096 	case IP_MULTICAST_VIF:
1097 		if (!legal_vif_num) {
1098 			error = EOPNOTSUPP;
1099 			break;
1100 		}
1101 		if (m == NULL || m->m_len != sizeof(int)) {
1102 			error = EINVAL;
1103 			break;
1104 		}
1105 		i = *(mtod(m, int *));
1106 		if (!legal_vif_num(i) && (i != -1)) {
1107 			error = EINVAL;
1108 			break;
1109 		}
1110 		imo->imo_multicast_vif = i;
1111 		break;
1112 
1113 	case IP_MULTICAST_IF:
1114 		/*
1115 		 * Select the interface for outgoing multicast packets.
1116 		 */
1117 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1118 			error = EINVAL;
1119 			break;
1120 		}
1121 		addr = *(mtod(m, struct in_addr *));
1122 		/*
1123 		 * INADDR_ANY is used to remove a previous selection.
1124 		 * When no interface is selected, a default one is
1125 		 * chosen every time a multicast packet is sent.
1126 		 */
1127 		if (addr.s_addr == INADDR_ANY) {
1128 			imo->imo_multicast_ifp = NULL;
1129 			break;
1130 		}
1131 		/*
1132 		 * The selected interface is identified by its local
1133 		 * IP address.  Find the interface and confirm that
1134 		 * it supports multicasting.
1135 		 */
1136 		s = splimp();
1137 		INADDR_TO_IFP(addr, ifp);
1138 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1139 			splx(s);
1140 			error = EADDRNOTAVAIL;
1141 			break;
1142 		}
1143 		imo->imo_multicast_ifp = ifp;
1144 		splx(s);
1145 		break;
1146 
1147 	case IP_MULTICAST_TTL:
1148 		/*
1149 		 * Set the IP time-to-live for outgoing multicast packets.
1150 		 */
1151 		if (m == NULL || m->m_len != 1) {
1152 			error = EINVAL;
1153 			break;
1154 		}
1155 		imo->imo_multicast_ttl = *(mtod(m, u_char *));
1156 		break;
1157 
1158 	case IP_MULTICAST_LOOP:
1159 		/*
1160 		 * Set the loopback flag for outgoing multicast packets.
1161 		 * Must be zero or one.
1162 		 */
1163 		if (m == NULL || m->m_len != 1 ||
1164 		   (loop = *(mtod(m, u_char *))) > 1) {
1165 			error = EINVAL;
1166 			break;
1167 		}
1168 		imo->imo_multicast_loop = loop;
1169 		break;
1170 
1171 	case IP_ADD_MEMBERSHIP:
1172 		/*
1173 		 * Add a multicast group membership.
1174 		 * Group must be a valid IP multicast address.
1175 		 */
1176 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1177 			error = EINVAL;
1178 			break;
1179 		}
1180 		mreq = mtod(m, struct ip_mreq *);
1181 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
1182 			error = EINVAL;
1183 			break;
1184 		}
1185 		s = splimp();
1186 		/*
1187 		 * If no interface address was provided, use the interface of
1188 		 * the route to the given multicast address.
1189 		 */
1190 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1191 			bzero((caddr_t)&ro, sizeof(ro));
1192 			dst = (struct sockaddr_in *)&ro.ro_dst;
1193 			dst->sin_len = sizeof(*dst);
1194 			dst->sin_family = AF_INET;
1195 			dst->sin_addr = mreq->imr_multiaddr;
1196 			rtalloc(&ro);
1197 			if (ro.ro_rt == NULL) {
1198 				error = EADDRNOTAVAIL;
1199 				splx(s);
1200 				break;
1201 			}
1202 			ifp = ro.ro_rt->rt_ifp;
1203 			rtfree(ro.ro_rt);
1204 		}
1205 		else {
1206 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1207 		}
1208 
1209 		/*
1210 		 * See if we found an interface, and confirm that it
1211 		 * supports multicast.
1212 		 */
1213 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1214 			error = EADDRNOTAVAIL;
1215 			splx(s);
1216 			break;
1217 		}
1218 		/*
1219 		 * See if the membership already exists or if all the
1220 		 * membership slots are full.
1221 		 */
1222 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1223 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1224 			    imo->imo_membership[i]->inm_addr.s_addr
1225 						== mreq->imr_multiaddr.s_addr)
1226 				break;
1227 		}
1228 		if (i < imo->imo_num_memberships) {
1229 			error = EADDRINUSE;
1230 			splx(s);
1231 			break;
1232 		}
1233 		if (i == IP_MAX_MEMBERSHIPS) {
1234 			error = ETOOMANYREFS;
1235 			splx(s);
1236 			break;
1237 		}
1238 		/*
1239 		 * Everything looks good; add a new record to the multicast
1240 		 * address list for the given interface.
1241 		 */
1242 		if ((imo->imo_membership[i] =
1243 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1244 			error = ENOBUFS;
1245 			splx(s);
1246 			break;
1247 		}
1248 		++imo->imo_num_memberships;
1249 		splx(s);
1250 		break;
1251 
1252 	case IP_DROP_MEMBERSHIP:
1253 		/*
1254 		 * Drop a multicast group membership.
1255 		 * Group must be a valid IP multicast address.
1256 		 */
1257 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1258 			error = EINVAL;
1259 			break;
1260 		}
1261 		mreq = mtod(m, struct ip_mreq *);
1262 		if (!IN_MULTICAST(ntohl(mreq->imr_multiaddr.s_addr))) {
1263 			error = EINVAL;
1264 			break;
1265 		}
1266 
1267 		s = splimp();
1268 		/*
1269 		 * If an interface address was specified, get a pointer
1270 		 * to its ifnet structure.
1271 		 */
1272 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1273 			ifp = NULL;
1274 		else {
1275 			INADDR_TO_IFP(mreq->imr_interface, ifp);
1276 			if (ifp == NULL) {
1277 				error = EADDRNOTAVAIL;
1278 				splx(s);
1279 				break;
1280 			}
1281 		}
1282 		/*
1283 		 * Find the membership in the membership array.
1284 		 */
1285 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1286 			if ((ifp == NULL ||
1287 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1288 			     imo->imo_membership[i]->inm_addr.s_addr ==
1289 			     mreq->imr_multiaddr.s_addr)
1290 				break;
1291 		}
1292 		if (i == imo->imo_num_memberships) {
1293 			error = EADDRNOTAVAIL;
1294 			splx(s);
1295 			break;
1296 		}
1297 		/*
1298 		 * Give up the multicast address record to which the
1299 		 * membership points.
1300 		 */
1301 		in_delmulti(imo->imo_membership[i]);
1302 		/*
1303 		 * Remove the gap in the membership array.
1304 		 */
1305 		for (++i; i < imo->imo_num_memberships; ++i)
1306 			imo->imo_membership[i-1] = imo->imo_membership[i];
1307 		--imo->imo_num_memberships;
1308 		splx(s);
1309 		break;
1310 
1311 	default:
1312 		error = EOPNOTSUPP;
1313 		break;
1314 	}
1315 
1316 	/*
1317 	 * If all options have default values, no need to keep the mbuf.
1318 	 */
1319 	if (imo->imo_multicast_ifp == NULL &&
1320 	    imo->imo_multicast_vif == -1 &&
1321 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1322 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1323 	    imo->imo_num_memberships == 0) {
1324 		free(*imop, M_IPMOPTS);
1325 		*imop = NULL;
1326 	}
1327 
1328 	return (error);
1329 }
1330 
1331 /*
1332  * Return the IP multicast options in response to user getsockopt().
1333  */
1334 static int
1335 ip_getmoptions(optname, imo, mp)
1336 	int optname;
1337 	register struct ip_moptions *imo;
1338 	register struct mbuf **mp;
1339 {
1340 	u_char *ttl;
1341 	u_char *loop;
1342 	struct in_addr *addr;
1343 	struct in_ifaddr *ia;
1344 
1345 	*mp = m_get(M_WAIT, MT_SOOPTS);
1346 
1347 	switch (optname) {
1348 
1349 	case IP_MULTICAST_VIF:
1350 		if (imo != NULL)
1351 			*(mtod(*mp, int *)) = imo->imo_multicast_vif;
1352 		else
1353 			*(mtod(*mp, int *)) = -1;
1354 		(*mp)->m_len = sizeof(int);
1355 		return(0);
1356 
1357 	case IP_MULTICAST_IF:
1358 		addr = mtod(*mp, struct in_addr *);
1359 		(*mp)->m_len = sizeof(struct in_addr);
1360 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1361 			addr->s_addr = INADDR_ANY;
1362 		else {
1363 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1364 			addr->s_addr = (ia == NULL) ? INADDR_ANY
1365 					: IA_SIN(ia)->sin_addr.s_addr;
1366 		}
1367 		return (0);
1368 
1369 	case IP_MULTICAST_TTL:
1370 		ttl = mtod(*mp, u_char *);
1371 		(*mp)->m_len = 1;
1372 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1373 				     : imo->imo_multicast_ttl;
1374 		return (0);
1375 
1376 	case IP_MULTICAST_LOOP:
1377 		loop = mtod(*mp, u_char *);
1378 		(*mp)->m_len = 1;
1379 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
1380 				      : imo->imo_multicast_loop;
1381 		return (0);
1382 
1383 	default:
1384 		return (EOPNOTSUPP);
1385 	}
1386 }
1387 
1388 /*
1389  * Discard the IP multicast options.
1390  */
1391 void
1392 ip_freemoptions(imo)
1393 	register struct ip_moptions *imo;
1394 {
1395 	register int i;
1396 
1397 	if (imo != NULL) {
1398 		for (i = 0; i < imo->imo_num_memberships; ++i)
1399 			in_delmulti(imo->imo_membership[i]);
1400 		free(imo, M_IPMOPTS);
1401 	}
1402 }
1403 
1404 /*
1405  * Routine called from ip_output() to loop back a copy of an IP multicast
1406  * packet to the input queue of a specified interface.  Note that this
1407  * calls the output routine of the loopback "driver", but with an interface
1408  * pointer that might NOT be a loopback interface -- evil, but easier than
1409  * replicating that code here.
1410  */
1411 static void
1412 ip_mloopback(ifp, m, dst, hlen)
1413 	struct ifnet *ifp;
1414 	register struct mbuf *m;
1415 	register struct sockaddr_in *dst;
1416 	int hlen;
1417 {
1418 	register struct ip *ip;
1419 	struct mbuf *copym;
1420 
1421 	copym = m_copy(m, 0, M_COPYALL);
1422 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1423 		copym = m_pullup(copym, hlen);
1424 	if (copym != NULL) {
1425 		/*
1426 		 * We don't bother to fragment if the IP length is greater
1427 		 * than the interface's MTU.  Can this possibly matter?
1428 		 */
1429 		ip = mtod(copym, struct ip *);
1430 		ip->ip_len = htons((u_short)ip->ip_len);
1431 		ip->ip_off = htons((u_short)ip->ip_off);
1432 		ip->ip_sum = 0;
1433 		if (ip->ip_vhl == IP_VHL_BORING) {
1434 			ip->ip_sum = in_cksum_hdr(ip);
1435 		} else {
1436 			ip->ip_sum = in_cksum(copym, hlen);
1437 		}
1438 		/*
1439 		 * NB:
1440 		 * It's not clear whether there are any lingering
1441 		 * reentrancy problems in other areas which might
1442 		 * be exposed by using ip_input directly (in
1443 		 * particular, everything which modifies the packet
1444 		 * in-place).  Yet another option is using the
1445 		 * protosw directly to deliver the looped back
1446 		 * packet.  For the moment, we'll err on the side
1447 		 * of safety by using if_simloop().
1448 		 */
1449 #if 1 /* XXX */
1450 		if (dst->sin_family != AF_INET) {
1451 			printf("ip_mloopback: bad address family %d\n",
1452 						dst->sin_family);
1453 			dst->sin_family = AF_INET;
1454 		}
1455 #endif
1456 
1457 #ifdef notdef
1458 		copym->m_pkthdr.rcvif = ifp;
1459 		ip_input(copym);
1460 #else
1461 		if_simloop(ifp, copym, (struct sockaddr *)dst, 0);
1462 #endif
1463 	}
1464 }
1465