xref: /freebsd/sys/netinet/ip_output.c (revision 224af215a6fe8d5e5e2c91cc97c48bdd67c991c7)
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_mac.h"
45 #include "opt_pfil_hooks.h"
46 #include "opt_random_ip_id.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/kernel.h>
51 #include <sys/mac.h>
52 #include <sys/malloc.h>
53 #include <sys/mbuf.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 
58 #include <net/if.h>
59 #include <net/route.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
67 
68 #include <machine/in_cksum.h>
69 
70 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
71 
72 #ifdef IPSEC
73 #include <netinet6/ipsec.h>
74 #include <netkey/key.h>
75 #ifdef IPSEC_DEBUG
76 #include <netkey/key_debug.h>
77 #else
78 #define	KEYDEBUG(lev,arg)
79 #endif
80 #endif /*IPSEC*/
81 
82 #include <netinet/ip_fw.h>
83 #include <netinet/ip_dummynet.h>
84 
85 #define print_ip(x, a, y)	 printf("%s %d.%d.%d.%d%s",\
86 				x, (ntohl(a.s_addr)>>24)&0xFF,\
87 				  (ntohl(a.s_addr)>>16)&0xFF,\
88 				  (ntohl(a.s_addr)>>8)&0xFF,\
89 				  (ntohl(a.s_addr))&0xFF, y);
90 
91 u_short ip_id;
92 
93 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
94 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
95 static void	ip_mloopback
96 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
97 static int	ip_getmoptions
98 	(struct sockopt *, struct ip_moptions *);
99 static int	ip_pcbopts(int, struct mbuf **, struct mbuf *);
100 static int	ip_setmoptions
101 	(struct sockopt *, struct ip_moptions **);
102 
103 int	ip_optcopy(struct ip *, struct ip *);
104 
105 
106 extern	struct protosw inetsw[];
107 
108 /*
109  * IP output.  The packet in mbuf chain m contains a skeletal IP
110  * header (with len, off, ttl, proto, tos, src, dst).
111  * The mbuf chain containing the packet will be freed.
112  * The mbuf opt, if present, will not be freed.
113  */
114 int
115 ip_output(m0, opt, ro, flags, imo)
116 	struct mbuf *m0;
117 	struct mbuf *opt;
118 	struct route *ro;
119 	int flags;
120 	struct ip_moptions *imo;
121 {
122 	struct ip *ip, *mhip;
123 	struct ifnet *ifp = NULL;	/* keep compiler happy */
124 	struct mbuf *m;
125 	int hlen = sizeof (struct ip);
126 	int len, off, error = 0;
127 	struct sockaddr_in *dst = NULL;	/* keep compiler happy */
128 	struct in_ifaddr *ia = NULL;
129 	int isbroadcast, sw_csum;
130 	struct in_addr pkt_dst;
131 #ifdef IPSEC
132 	struct route iproute;
133 	struct socket *so = NULL;
134 	struct secpolicy *sp = NULL;
135 #endif
136 	struct ip_fw_args args;
137 	int src_was_INADDR_ANY = 0;	/* as the name says... */
138 #ifdef PFIL_HOOKS
139 	struct packet_filter_hook *pfh;
140 	struct mbuf *m1;
141 	int rv;
142 #endif /* PFIL_HOOKS */
143 
144 	args.eh = NULL;
145 	args.rule = NULL;
146 	args.next_hop = NULL;
147 	args.divert_rule = 0;			/* divert cookie */
148 
149 	/* Grab info from MT_TAG mbufs prepended to the chain. */
150 	for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
151 		switch(m0->m_tag_id) {
152 		default:
153 			printf("ip_output: unrecognised MT_TAG tag %d\n",
154 			    m0->m_tag_id);
155 			break;
156 
157 		case PACKET_TAG_DUMMYNET:
158 			/*
159 			 * the packet was already tagged, so part of the
160 			 * processing was already done, and we need to go down.
161 			 * Get parameters from the header.
162 			 */
163 			args.rule = ((struct dn_pkt *)m0)->rule;
164 			opt = NULL ;
165 			ro = & ( ((struct dn_pkt *)m0)->ro ) ;
166 			imo = NULL ;
167 			dst = ((struct dn_pkt *)m0)->dn_dst ;
168 			ifp = ((struct dn_pkt *)m0)->ifp ;
169 			flags = ((struct dn_pkt *)m0)->flags ;
170 			break;
171 
172 		case PACKET_TAG_DIVERT:
173 			args.divert_rule = (intptr_t)m0->m_data & 0xffff;
174 			break;
175 
176 		case PACKET_TAG_IPFORWARD:
177 			args.next_hop = (struct sockaddr_in *)m0->m_data;
178 			break;
179 		}
180 	}
181 	m = m0;
182 
183 	KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
184 
185 	KASSERT(ro != NULL, ("ip_output: no route, proto %d",
186 	    mtod(m, struct ip *)->ip_p));
187 
188 #ifdef IPSEC
189 	so = ipsec_getsocket(m);
190 	(void)ipsec_setsocket(m, NULL);
191 #endif
192 	if (args.rule != NULL) {	/* dummynet already saw us */
193 		ip = mtod(m, struct ip *);
194 		hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
195 		if (ro->ro_rt)
196 			ia = ifatoia(ro->ro_rt->rt_ifa);
197 		goto sendit;
198 	}
199 
200 	if (opt) {
201 		len = 0;
202 		m = ip_insertoptions(m, opt, &len);
203 		if (len != 0)
204 			hlen = len;
205 	}
206 	ip = mtod(m, struct ip *);
207 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
208 
209 	/*
210 	 * Fill in IP header.
211 	 */
212 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
213 		ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
214 		ip->ip_off &= IP_DF;
215 #ifdef RANDOM_IP_ID
216 		ip->ip_id = ip_randomid();
217 #else
218 		ip->ip_id = htons(ip_id++);
219 #endif
220 		ipstat.ips_localout++;
221 	} else {
222 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
223 	}
224 
225 	dst = (struct sockaddr_in *)&ro->ro_dst;
226 	/*
227 	 * If there is a cached route,
228 	 * check that it is to the same destination
229 	 * and is still up.  If not, free it and try again.
230 	 * The address family should also be checked in case of sharing the
231 	 * cache with IPv6.
232 	 */
233 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
234 			  dst->sin_family != AF_INET ||
235 			  dst->sin_addr.s_addr != pkt_dst.s_addr)) {
236 		RTFREE(ro->ro_rt);
237 		ro->ro_rt = (struct rtentry *)0;
238 	}
239 	if (ro->ro_rt == 0) {
240 		bzero(dst, sizeof(*dst));
241 		dst->sin_family = AF_INET;
242 		dst->sin_len = sizeof(*dst);
243 		dst->sin_addr = pkt_dst;
244 	}
245 	/*
246 	 * If routing to interface only,
247 	 * short circuit routing lookup.
248 	 */
249 	if (flags & IP_ROUTETOIF) {
250 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
251 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
252 			ipstat.ips_noroute++;
253 			error = ENETUNREACH;
254 			goto bad;
255 		}
256 		ifp = ia->ia_ifp;
257 		ip->ip_ttl = 1;
258 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
259 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
260 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
261 		/*
262 		 * Bypass the normal routing lookup for multicast
263 		 * packets if the interface is specified.
264 		 */
265 		ifp = imo->imo_multicast_ifp;
266 		IFP_TO_IA(ifp, ia);
267 		isbroadcast = 0;	/* fool gcc */
268 	} else {
269 		/*
270 		 * If this is the case, we probably don't want to allocate
271 		 * a protocol-cloned route since we didn't get one from the
272 		 * ULP.  This lets TCP do its thing, while not burdening
273 		 * forwarding or ICMP with the overhead of cloning a route.
274 		 * Of course, we still want to do any cloning requested by
275 		 * the link layer, as this is probably required in all cases
276 		 * for correct operation (as it is for ARP).
277 		 */
278 		if (ro->ro_rt == 0)
279 			rtalloc_ign(ro, RTF_PRCLONING);
280 		if (ro->ro_rt == 0) {
281 			ipstat.ips_noroute++;
282 			error = EHOSTUNREACH;
283 			goto bad;
284 		}
285 		ia = ifatoia(ro->ro_rt->rt_ifa);
286 		ifp = ro->ro_rt->rt_ifp;
287 		ro->ro_rt->rt_use++;
288 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
289 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
290 		if (ro->ro_rt->rt_flags & RTF_HOST)
291 			isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
292 		else
293 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
294 	}
295 	if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
296 		struct in_multi *inm;
297 
298 		m->m_flags |= M_MCAST;
299 		/*
300 		 * IP destination address is multicast.  Make sure "dst"
301 		 * still points to the address in "ro".  (It may have been
302 		 * changed to point to a gateway address, above.)
303 		 */
304 		dst = (struct sockaddr_in *)&ro->ro_dst;
305 		/*
306 		 * See if the caller provided any multicast options
307 		 */
308 		if (imo != NULL) {
309 			ip->ip_ttl = imo->imo_multicast_ttl;
310 			if (imo->imo_multicast_vif != -1)
311 				ip->ip_src.s_addr =
312 				    ip_mcast_src(imo->imo_multicast_vif);
313 		} else
314 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
315 		/*
316 		 * Confirm that the outgoing interface supports multicast.
317 		 */
318 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
319 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
320 				ipstat.ips_noroute++;
321 				error = ENETUNREACH;
322 				goto bad;
323 			}
324 		}
325 		/*
326 		 * If source address not specified yet, use address
327 		 * of outgoing interface.
328 		 */
329 		if (ip->ip_src.s_addr == INADDR_ANY) {
330 			/* Interface may have no addresses. */
331 			if (ia != NULL)
332 				ip->ip_src = IA_SIN(ia)->sin_addr;
333 		}
334 
335 		if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
336 			/*
337 			 * XXX
338 			 * delayed checksums are not currently
339 			 * compatible with IP multicast routing
340 			 */
341 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
342 				in_delayed_cksum(m);
343 				m->m_pkthdr.csum_flags &=
344 					~CSUM_DELAY_DATA;
345 			}
346 		}
347 		IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
348 		if (inm != NULL &&
349 		   (imo == NULL || imo->imo_multicast_loop)) {
350 			/*
351 			 * If we belong to the destination multicast group
352 			 * on the outgoing interface, and the caller did not
353 			 * forbid loopback, loop back a copy.
354 			 */
355 			ip_mloopback(ifp, m, dst, hlen);
356 		}
357 		else {
358 			/*
359 			 * If we are acting as a multicast router, perform
360 			 * multicast forwarding as if the packet had just
361 			 * arrived on the interface to which we are about
362 			 * to send.  The multicast forwarding function
363 			 * recursively calls this function, using the
364 			 * IP_FORWARDING flag to prevent infinite recursion.
365 			 *
366 			 * Multicasts that are looped back by ip_mloopback(),
367 			 * above, will be forwarded by the ip_input() routine,
368 			 * if necessary.
369 			 */
370 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
371 				/*
372 				 * Check if rsvp daemon is running. If not, don't
373 				 * set ip_moptions. This ensures that the packet
374 				 * is multicast and not just sent down one link
375 				 * as prescribed by rsvpd.
376 				 */
377 				if (!rsvp_on)
378 				  imo = NULL;
379 				if (ip_mforward(ip, ifp, m, imo) != 0) {
380 					m_freem(m);
381 					goto done;
382 				}
383 			}
384 		}
385 
386 		/*
387 		 * Multicasts with a time-to-live of zero may be looped-
388 		 * back, above, but must not be transmitted on a network.
389 		 * Also, multicasts addressed to the loopback interface
390 		 * are not sent -- the above call to ip_mloopback() will
391 		 * loop back a copy if this host actually belongs to the
392 		 * destination group on the loopback interface.
393 		 */
394 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
395 			m_freem(m);
396 			goto done;
397 		}
398 
399 		goto sendit;
400 	}
401 #ifndef notdef
402 	/*
403 	 * If the source address is not specified yet, use the address
404 	 * of the outoing interface. In case, keep note we did that, so
405 	 * if the the firewall changes the next-hop causing the output
406 	 * interface to change, we can fix that.
407 	 */
408 	if (ip->ip_src.s_addr == INADDR_ANY) {
409 		/* Interface may have no addresses. */
410 		if (ia != NULL) {
411 			ip->ip_src = IA_SIN(ia)->sin_addr;
412 			src_was_INADDR_ANY = 1;
413 		}
414 	}
415 #endif /* notdef */
416 	/*
417 	 * Verify that we have any chance at all of being able to queue
418 	 *      the packet or packet fragments
419 	 */
420 	if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
421 		ifp->if_snd.ifq_maxlen) {
422 			error = ENOBUFS;
423 			ipstat.ips_odropped++;
424 			goto bad;
425 	}
426 
427 	/*
428 	 * Look for broadcast address and
429 	 * verify user is allowed to send
430 	 * such a packet.
431 	 */
432 	if (isbroadcast) {
433 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
434 			error = EADDRNOTAVAIL;
435 			goto bad;
436 		}
437 		if ((flags & IP_ALLOWBROADCAST) == 0) {
438 			error = EACCES;
439 			goto bad;
440 		}
441 		/* don't allow broadcast messages to be fragmented */
442 		if ((u_short)ip->ip_len > ifp->if_mtu) {
443 			error = EMSGSIZE;
444 			goto bad;
445 		}
446 		m->m_flags |= M_BCAST;
447 	} else {
448 		m->m_flags &= ~M_BCAST;
449 	}
450 
451 sendit:
452 #ifdef IPSEC
453 	/* get SP for this packet */
454 	if (so == NULL)
455 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
456 	else
457 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
458 
459 	if (sp == NULL) {
460 		ipsecstat.out_inval++;
461 		goto bad;
462 	}
463 
464 	error = 0;
465 
466 	/* check policy */
467 	switch (sp->policy) {
468 	case IPSEC_POLICY_DISCARD:
469 		/*
470 		 * This packet is just discarded.
471 		 */
472 		ipsecstat.out_polvio++;
473 		goto bad;
474 
475 	case IPSEC_POLICY_BYPASS:
476 	case IPSEC_POLICY_NONE:
477 		/* no need to do IPsec. */
478 		goto skip_ipsec;
479 
480 	case IPSEC_POLICY_IPSEC:
481 		if (sp->req == NULL) {
482 			/* acquire a policy */
483 			error = key_spdacquire(sp);
484 			goto bad;
485 		}
486 		break;
487 
488 	case IPSEC_POLICY_ENTRUST:
489 	default:
490 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
491 	}
492     {
493 	struct ipsec_output_state state;
494 	bzero(&state, sizeof(state));
495 	state.m = m;
496 	if (flags & IP_ROUTETOIF) {
497 		state.ro = &iproute;
498 		bzero(&iproute, sizeof(iproute));
499 	} else
500 		state.ro = ro;
501 	state.dst = (struct sockaddr *)dst;
502 
503 	ip->ip_sum = 0;
504 
505 	/*
506 	 * XXX
507 	 * delayed checksums are not currently compatible with IPsec
508 	 */
509 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
510 		in_delayed_cksum(m);
511 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
512 	}
513 
514 	ip->ip_len = htons(ip->ip_len);
515 	ip->ip_off = htons(ip->ip_off);
516 
517 	error = ipsec4_output(&state, sp, flags);
518 
519 	m = state.m;
520 	if (flags & IP_ROUTETOIF) {
521 		/*
522 		 * if we have tunnel mode SA, we may need to ignore
523 		 * IP_ROUTETOIF.
524 		 */
525 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
526 			flags &= ~IP_ROUTETOIF;
527 			ro = state.ro;
528 		}
529 	} else
530 		ro = state.ro;
531 	dst = (struct sockaddr_in *)state.dst;
532 	if (error) {
533 		/* mbuf is already reclaimed in ipsec4_output. */
534 		m0 = NULL;
535 		switch (error) {
536 		case EHOSTUNREACH:
537 		case ENETUNREACH:
538 		case EMSGSIZE:
539 		case ENOBUFS:
540 		case ENOMEM:
541 			break;
542 		default:
543 			printf("ip4_output (ipsec): error code %d\n", error);
544 			/*fall through*/
545 		case ENOENT:
546 			/* don't show these error codes to the user */
547 			error = 0;
548 			break;
549 		}
550 		goto bad;
551 	}
552     }
553 
554 	/* be sure to update variables that are affected by ipsec4_output() */
555 	ip = mtod(m, struct ip *);
556 #ifdef _IP_VHL
557 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
558 #else
559 	hlen = ip->ip_hl << 2;
560 #endif
561 	if (ro->ro_rt == NULL) {
562 		if ((flags & IP_ROUTETOIF) == 0) {
563 			printf("ip_output: "
564 				"can't update route after IPsec processing\n");
565 			error = EHOSTUNREACH;	/*XXX*/
566 			goto bad;
567 		}
568 	} else {
569 		ia = ifatoia(ro->ro_rt->rt_ifa);
570 		ifp = ro->ro_rt->rt_ifp;
571 	}
572 
573 	/* make it flipped, again. */
574 	ip->ip_len = ntohs(ip->ip_len);
575 	ip->ip_off = ntohs(ip->ip_off);
576 skip_ipsec:
577 #endif /*IPSEC*/
578 
579 	/*
580 	 * IpHack's section.
581 	 * - Xlate: translate packet's addr/port (NAT).
582 	 * - Firewall: deny/allow/etc.
583 	 * - Wrap: fake packet's addr/port <unimpl.>
584 	 * - Encapsulate: put it in another IP and send out. <unimp.>
585 	 */
586 #ifdef PFIL_HOOKS
587 	/*
588 	 * Run through list of hooks for output packets.
589 	 */
590 	m1 = m;
591 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
592 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
593 		if (pfh->pfil_func) {
594 			rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1);
595 			if (rv) {
596 				error = EHOSTUNREACH;
597 				goto done;
598 			}
599 			m = m1;
600 			if (m == NULL)
601 				goto done;
602 			ip = mtod(m, struct ip *);
603 		}
604 #endif /* PFIL_HOOKS */
605 
606 	/*
607 	 * Check with the firewall...
608 	 * but not if we are already being fwd'd from a firewall.
609 	 */
610 	if (fw_enable && IPFW_LOADED && !args.next_hop) {
611 		struct sockaddr_in *old = dst;
612 
613 		args.m = m;
614 		args.next_hop = dst;
615 		args.oif = ifp;
616 		off = ip_fw_chk_ptr(&args);
617 		m = args.m;
618 		dst = args.next_hop;
619 
620                 /*
621 		 * On return we must do the following:
622 		 * m == NULL	-> drop the pkt (old interface, deprecated)
623 		 * (off & IP_FW_PORT_DENY_FLAG)	-> drop the pkt (new interface)
624 		 * 1<=off<= 0xffff		-> DIVERT
625 		 * (off & IP_FW_PORT_DYNT_FLAG)	-> send to a DUMMYNET pipe
626 		 * (off & IP_FW_PORT_TEE_FLAG)	-> TEE the packet
627 		 * dst != old			-> IPFIREWALL_FORWARD
628 		 * off==0, dst==old		-> accept
629 		 * If some of the above modules are not compiled in, then
630 		 * we should't have to check the corresponding condition
631 		 * (because the ipfw control socket should not accept
632 		 * unsupported rules), but better play safe and drop
633 		 * packets in case of doubt.
634 		 */
635 		if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
636 			if (m)
637 				m_freem(m);
638 			error = EACCES;
639 			goto done;
640 		}
641 		ip = mtod(m, struct ip *);
642 		if (off == 0 && dst == old)		/* common case */
643 			goto pass;
644                 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
645 			/*
646 			 * pass the pkt to dummynet. Need to include
647 			 * pipe number, m, ifp, ro, dst because these are
648 			 * not recomputed in the next pass.
649 			 * All other parameters have been already used and
650 			 * so they are not needed anymore.
651 			 * XXX note: if the ifp or ro entry are deleted
652 			 * while a pkt is in dummynet, we are in trouble!
653 			 */
654 			args.ro = ro;
655 			args.dst = dst;
656 			args.flags = flags;
657 
658 			error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
659 				&args);
660 			goto done;
661 		}
662 #ifdef IPDIVERT
663 		if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
664 			struct mbuf *clone = NULL;
665 
666 			/* Clone packet if we're doing a 'tee' */
667 			if ((off & IP_FW_PORT_TEE_FLAG) != 0)
668 				clone = m_dup(m, M_DONTWAIT);
669 
670 			/*
671 			 * XXX
672 			 * delayed checksums are not currently compatible
673 			 * with divert sockets.
674 			 */
675 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
676 				in_delayed_cksum(m);
677 				m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
678 			}
679 
680 			/* Restore packet header fields to original values */
681 			ip->ip_len = htons(ip->ip_len);
682 			ip->ip_off = htons(ip->ip_off);
683 
684 			/* Deliver packet to divert input routine */
685 			divert_packet(m, 0, off & 0xffff, args.divert_rule);
686 
687 			/* If 'tee', continue with original packet */
688 			if (clone != NULL) {
689 				m = clone;
690 				ip = mtod(m, struct ip *);
691 				goto pass;
692 			}
693 			goto done;
694 		}
695 #endif
696 
697 		/* IPFIREWALL_FORWARD */
698 		/*
699 		 * Check dst to make sure it is directly reachable on the
700 		 * interface we previously thought it was.
701 		 * If it isn't (which may be likely in some situations) we have
702 		 * to re-route it (ie, find a route for the next-hop and the
703 		 * associated interface) and set them here. This is nested
704 		 * forwarding which in most cases is undesirable, except where
705 		 * such control is nigh impossible. So we do it here.
706 		 * And I'm babbling.
707 		 */
708 		if (off == 0 && old != dst) { /* FORWARD, dst has changed */
709 #if 0
710 			/*
711 			 * XXX To improve readability, this block should be
712 			 * changed into a function call as below:
713 			 */
714 			error = ip_ipforward(&m, &dst, &ifp);
715 			if (error)
716 				goto bad;
717 			if (m == NULL) /* ip_input consumed the mbuf */
718 				goto done;
719 #else
720 			struct in_ifaddr *ia;
721 
722 			/*
723 			 * XXX sro_fwd below is static, and a pointer
724 			 * to it gets passed to routines downstream.
725 			 * This could have surprisingly bad results in
726 			 * practice, because its content is overwritten
727 			 * by subsequent packets.
728 			 */
729 			/* There must be a better way to do this next line... */
730 			static struct route sro_fwd;
731 			struct route *ro_fwd = &sro_fwd;
732 
733 #if 0
734 			print_ip("IPFIREWALL_FORWARD: New dst ip: ",
735 			    dst->sin_addr, "\n");
736 #endif
737 
738 			/*
739 			 * We need to figure out if we have been forwarded
740 			 * to a local socket. If so, then we should somehow
741 			 * "loop back" to ip_input, and get directed to the
742 			 * PCB as if we had received this packet. This is
743 			 * because it may be dificult to identify the packets
744 			 * you want to forward until they are being output
745 			 * and have selected an interface. (e.g. locally
746 			 * initiated packets) If we used the loopback inteface,
747 			 * we would not be able to control what happens
748 			 * as the packet runs through ip_input() as
749 			 * it is done through a ISR.
750 			 */
751 			LIST_FOREACH(ia,
752 			    INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
753 				/*
754 				 * If the addr to forward to is one
755 				 * of ours, we pretend to
756 				 * be the destination for this packet.
757 				 */
758 				if (IA_SIN(ia)->sin_addr.s_addr ==
759 						 dst->sin_addr.s_addr)
760 					break;
761 			}
762 			if (ia) {	/* tell ip_input "dont filter" */
763 				struct m_hdr tag;
764 
765 				tag.mh_type = MT_TAG;
766 				tag.mh_flags = PACKET_TAG_IPFORWARD;
767 				tag.mh_data = (caddr_t)args.next_hop;
768 				tag.mh_next = m;
769 
770 				if (m->m_pkthdr.rcvif == NULL)
771 					m->m_pkthdr.rcvif = ifunit("lo0");
772 				if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
773 					m->m_pkthdr.csum_flags |=
774 					    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
775 					m0->m_pkthdr.csum_data = 0xffff;
776 				}
777 				m->m_pkthdr.csum_flags |=
778 				    CSUM_IP_CHECKED | CSUM_IP_VALID;
779 				ip->ip_len = htons(ip->ip_len);
780 				ip->ip_off = htons(ip->ip_off);
781 				ip_input((struct mbuf *)&tag);
782 				goto done;
783 			}
784 			/* Some of the logic for this was
785 			 * nicked from above.
786 			 *
787 			 * This rewrites the cached route in a local PCB.
788 			 * Is this what we want to do?
789 			 */
790 			bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
791 
792 			ro_fwd->ro_rt = 0;
793 			rtalloc_ign(ro_fwd, RTF_PRCLONING);
794 
795 			if (ro_fwd->ro_rt == 0) {
796 				ipstat.ips_noroute++;
797 				error = EHOSTUNREACH;
798 				goto bad;
799 			}
800 
801 			ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
802 			ifp = ro_fwd->ro_rt->rt_ifp;
803 			ro_fwd->ro_rt->rt_use++;
804 			if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
805 				dst = (struct sockaddr_in *)
806 					ro_fwd->ro_rt->rt_gateway;
807 			if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
808 				isbroadcast =
809 				    (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
810 			else
811 				isbroadcast = in_broadcast(dst->sin_addr, ifp);
812 			if (ro->ro_rt)
813 				RTFREE(ro->ro_rt);
814 			ro->ro_rt = ro_fwd->ro_rt;
815 			dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
816 
817 #endif	/* ... block to be put into a function */
818 			/*
819 			 * If we added a default src ip earlier,
820 			 * which would have been gotten from the-then
821 			 * interface, do it again, from the new one.
822 			 */
823 			if (src_was_INADDR_ANY)
824 				ip->ip_src = IA_SIN(ia)->sin_addr;
825 			goto pass ;
826 		}
827 
828                 /*
829                  * if we get here, none of the above matches, and
830                  * we have to drop the pkt
831                  */
832 		m_freem(m);
833                 error = EACCES; /* not sure this is the right error msg */
834                 goto done;
835 	}
836 
837 pass:
838 	/* 127/8 must not appear on wire - RFC1122. */
839 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
840 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
841 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
842 			ipstat.ips_badaddr++;
843 			error = EADDRNOTAVAIL;
844 			goto bad;
845 		}
846 	}
847 
848 	m->m_pkthdr.csum_flags |= CSUM_IP;
849 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
850 	if (sw_csum & CSUM_DELAY_DATA) {
851 		in_delayed_cksum(m);
852 		sw_csum &= ~CSUM_DELAY_DATA;
853 	}
854 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
855 
856 	/*
857 	 * If small enough for interface, or the interface will take
858 	 * care of the fragmentation for us, can just send directly.
859 	 */
860 	if ((u_short)ip->ip_len <= ifp->if_mtu ||
861 	    ifp->if_hwassist & CSUM_FRAGMENT) {
862 		ip->ip_len = htons(ip->ip_len);
863 		ip->ip_off = htons(ip->ip_off);
864 		ip->ip_sum = 0;
865 		if (sw_csum & CSUM_DELAY_IP) {
866 			if (ip->ip_vhl == IP_VHL_BORING) {
867 				ip->ip_sum = in_cksum_hdr(ip);
868 			} else {
869 				ip->ip_sum = in_cksum(m, hlen);
870 			}
871 		}
872 
873 		/* Record statistics for this interface address. */
874 		if (!(flags & IP_FORWARDING) && ia) {
875 			ia->ia_ifa.if_opackets++;
876 			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
877 		}
878 
879 #ifdef IPSEC
880 		/* clean ipsec history once it goes out of the node */
881 		ipsec_delaux(m);
882 #endif
883 
884 		error = (*ifp->if_output)(ifp, m,
885 				(struct sockaddr *)dst, ro->ro_rt);
886 		goto done;
887 	}
888 	/*
889 	 * Too large for interface; fragment if possible.
890 	 * Must be able to put at least 8 bytes per fragment.
891 	 */
892 	if (ip->ip_off & IP_DF) {
893 		error = EMSGSIZE;
894 		/*
895 		 * This case can happen if the user changed the MTU
896 		 * of an interface after enabling IP on it.  Because
897 		 * most netifs don't keep track of routes pointing to
898 		 * them, there is no way for one to update all its
899 		 * routes when the MTU is changed.
900 		 */
901 		if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST))
902 		    && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU)
903 		    && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
904 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
905 		}
906 		ipstat.ips_cantfrag++;
907 		goto bad;
908 	}
909 	len = (ifp->if_mtu - hlen) &~ 7;
910 	if (len < 8) {
911 		error = EMSGSIZE;
912 		goto bad;
913 	}
914 
915 	/*
916 	 * if the interface will not calculate checksums on
917 	 * fragmented packets, then do it here.
918 	 */
919 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
920 	    (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) {
921 		in_delayed_cksum(m);
922 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
923 	}
924 
925 	if (len > PAGE_SIZE) {
926 		/*
927 		 * Fragement large datagrams such that each segment
928 		 * contains a multiple of PAGE_SIZE amount of data,
929 		 * plus headers. This enables a receiver to perform
930 		 * page-flipping zero-copy optimizations.
931 		 */
932 
933 		int newlen;
934 		struct mbuf *mtmp;
935 
936 		for (mtmp = m, off = 0;
937 		     mtmp && ((off + mtmp->m_len) <= ifp->if_mtu);
938 		     mtmp = mtmp->m_next) {
939 			off += mtmp->m_len;
940 		}
941 		/*
942 		 * firstlen (off - hlen) must be aligned on an
943 		 * 8-byte boundary
944 		 */
945 		if (off < hlen)
946 			goto smart_frag_failure;
947 		off = ((off - hlen) & ~7) + hlen;
948 		newlen = (~PAGE_MASK) & ifp->if_mtu;
949 		if ((newlen + sizeof (struct ip)) > ifp->if_mtu) {
950 			/* we failed, go back the default */
951 smart_frag_failure:
952 			newlen = len;
953 			off = hlen + len;
954 		}
955 
956 /*		printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n",
957 		len, hlen, sizeof (struct ip), newlen, off);*/
958 
959 		len = newlen;
960 
961 	} else {
962 		off = hlen + len;
963 	}
964 
965 
966 
967     {
968 	int mhlen, firstlen = off - hlen;
969 	struct mbuf **mnext = &m->m_nextpkt;
970 	int nfrags = 1;
971 
972 	/*
973 	 * Loop through length of segment after first fragment,
974 	 * make new header and copy data of each part and link onto chain.
975 	 */
976 	m0 = m;
977 	mhlen = sizeof (struct ip);
978 	for (; off < (u_short)ip->ip_len; off += len) {
979 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
980 		if (m == 0) {
981 			error = ENOBUFS;
982 			ipstat.ips_odropped++;
983 			goto sendorfree;
984 		}
985 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
986 		m->m_data += max_linkhdr;
987 		mhip = mtod(m, struct ip *);
988 		*mhip = *ip;
989 		if (hlen > sizeof (struct ip)) {
990 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
991 			mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
992 		}
993 		m->m_len = mhlen;
994 		mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
995 		if (off + len >= (u_short)ip->ip_len)
996 			len = (u_short)ip->ip_len - off;
997 		else
998 			mhip->ip_off |= IP_MF;
999 		mhip->ip_len = htons((u_short)(len + mhlen));
1000 		m->m_next = m_copy(m0, off, len);
1001 		if (m->m_next == 0) {
1002 			(void) m_free(m);
1003 			error = ENOBUFS;	/* ??? */
1004 			ipstat.ips_odropped++;
1005 			goto sendorfree;
1006 		}
1007 		m->m_pkthdr.len = mhlen + len;
1008 		m->m_pkthdr.rcvif = (struct ifnet *)0;
1009 #ifdef MAC
1010 		mac_create_fragment(m0, m);
1011 #endif
1012 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1013 		mhip->ip_off = htons(mhip->ip_off);
1014 		mhip->ip_sum = 0;
1015 		if (sw_csum & CSUM_DELAY_IP) {
1016 			if (mhip->ip_vhl == IP_VHL_BORING) {
1017 				mhip->ip_sum = in_cksum_hdr(mhip);
1018 			} else {
1019 				mhip->ip_sum = in_cksum(m, mhlen);
1020 			}
1021 		}
1022 		*mnext = m;
1023 		mnext = &m->m_nextpkt;
1024 		nfrags++;
1025 	}
1026 	ipstat.ips_ofragments += nfrags;
1027 
1028 	/* set first/last markers for fragment chain */
1029 	m->m_flags |= M_LASTFRAG;
1030 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1031 	m0->m_pkthdr.csum_data = nfrags;
1032 
1033 	/*
1034 	 * Update first fragment by trimming what's been copied out
1035 	 * and updating header, then send each fragment (in order).
1036 	 */
1037 	m = m0;
1038 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
1039 	m->m_pkthdr.len = hlen + firstlen;
1040 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
1041 	ip->ip_off |= IP_MF;
1042 	ip->ip_off = htons(ip->ip_off);
1043 	ip->ip_sum = 0;
1044 	if (sw_csum & CSUM_DELAY_IP) {
1045 		if (ip->ip_vhl == IP_VHL_BORING) {
1046 			ip->ip_sum = in_cksum_hdr(ip);
1047 		} else {
1048 			ip->ip_sum = in_cksum(m, hlen);
1049 		}
1050 	}
1051 sendorfree:
1052 	for (m = m0; m; m = m0) {
1053 		m0 = m->m_nextpkt;
1054 		m->m_nextpkt = 0;
1055 #ifdef IPSEC
1056 		/* clean ipsec history once it goes out of the node */
1057 		ipsec_delaux(m);
1058 #endif
1059 		if (error == 0) {
1060 			/* Record statistics for this interface address. */
1061 			if (ia != NULL) {
1062 				ia->ia_ifa.if_opackets++;
1063 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1064 			}
1065 
1066 			error = (*ifp->if_output)(ifp, m,
1067 			    (struct sockaddr *)dst, ro->ro_rt);
1068 		} else
1069 			m_freem(m);
1070 	}
1071 
1072 	if (error == 0)
1073 		ipstat.ips_fragmented++;
1074     }
1075 done:
1076 #ifdef IPSEC
1077 	if (ro == &iproute && ro->ro_rt) {
1078 		RTFREE(ro->ro_rt);
1079 		ro->ro_rt = NULL;
1080 	}
1081 	if (sp != NULL) {
1082 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1083 			printf("DP ip_output call free SP:%p\n", sp));
1084 		key_freesp(sp);
1085 	}
1086 #endif /* IPSEC */
1087 	return (error);
1088 bad:
1089 	m_freem(m);
1090 	goto done;
1091 }
1092 
1093 void
1094 in_delayed_cksum(struct mbuf *m)
1095 {
1096 	struct ip *ip;
1097 	u_short csum, offset;
1098 
1099 	ip = mtod(m, struct ip *);
1100 	offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1101 	csum = in_cksum_skip(m, ip->ip_len, offset);
1102 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1103 		csum = 0xffff;
1104 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
1105 
1106 	if (offset + sizeof(u_short) > m->m_len) {
1107 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
1108 		    m->m_len, offset, ip->ip_p);
1109 		/*
1110 		 * XXX
1111 		 * this shouldn't happen, but if it does, the
1112 		 * correct behavior may be to insert the checksum
1113 		 * in the existing chain instead of rearranging it.
1114 		 */
1115 		m = m_pullup(m, offset + sizeof(u_short));
1116 	}
1117 	*(u_short *)(m->m_data + offset) = csum;
1118 }
1119 
1120 /*
1121  * Insert IP options into preformed packet.
1122  * Adjust IP destination as required for IP source routing,
1123  * as indicated by a non-zero in_addr at the start of the options.
1124  *
1125  * XXX This routine assumes that the packet has no options in place.
1126  */
1127 static struct mbuf *
1128 ip_insertoptions(m, opt, phlen)
1129 	register struct mbuf *m;
1130 	struct mbuf *opt;
1131 	int *phlen;
1132 {
1133 	register struct ipoption *p = mtod(opt, struct ipoption *);
1134 	struct mbuf *n;
1135 	register struct ip *ip = mtod(m, struct ip *);
1136 	unsigned optlen;
1137 
1138 	optlen = opt->m_len - sizeof(p->ipopt_dst);
1139 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1140 		*phlen = 0;
1141 		return (m);		/* XXX should fail */
1142 	}
1143 	if (p->ipopt_dst.s_addr)
1144 		ip->ip_dst = p->ipopt_dst;
1145 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1146 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
1147 		if (n == 0) {
1148 			*phlen = 0;
1149 			return (m);
1150 		}
1151 		n->m_pkthdr.rcvif = (struct ifnet *)0;
1152 #ifdef MAC
1153 		mac_create_mbuf_from_mbuf(m, n);
1154 #endif
1155 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1156 		m->m_len -= sizeof(struct ip);
1157 		m->m_data += sizeof(struct ip);
1158 		n->m_next = m;
1159 		m = n;
1160 		m->m_len = optlen + sizeof(struct ip);
1161 		m->m_data += max_linkhdr;
1162 		(void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1163 	} else {
1164 		m->m_data -= optlen;
1165 		m->m_len += optlen;
1166 		m->m_pkthdr.len += optlen;
1167 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1168 	}
1169 	ip = mtod(m, struct ip *);
1170 	bcopy(p->ipopt_list, ip + 1, optlen);
1171 	*phlen = sizeof(struct ip) + optlen;
1172 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1173 	ip->ip_len += optlen;
1174 	return (m);
1175 }
1176 
1177 /*
1178  * Copy options from ip to jp,
1179  * omitting those not copied during fragmentation.
1180  */
1181 int
1182 ip_optcopy(ip, jp)
1183 	struct ip *ip, *jp;
1184 {
1185 	register u_char *cp, *dp;
1186 	int opt, optlen, cnt;
1187 
1188 	cp = (u_char *)(ip + 1);
1189 	dp = (u_char *)(jp + 1);
1190 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1191 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1192 		opt = cp[0];
1193 		if (opt == IPOPT_EOL)
1194 			break;
1195 		if (opt == IPOPT_NOP) {
1196 			/* Preserve for IP mcast tunnel's LSRR alignment. */
1197 			*dp++ = IPOPT_NOP;
1198 			optlen = 1;
1199 			continue;
1200 		}
1201 
1202 		KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1203 		    ("ip_optcopy: malformed ipv4 option"));
1204 		optlen = cp[IPOPT_OLEN];
1205 		KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1206 		    ("ip_optcopy: malformed ipv4 option"));
1207 
1208 		/* bogus lengths should have been caught by ip_dooptions */
1209 		if (optlen > cnt)
1210 			optlen = cnt;
1211 		if (IPOPT_COPIED(opt)) {
1212 			bcopy(cp, dp, optlen);
1213 			dp += optlen;
1214 		}
1215 	}
1216 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1217 		*dp++ = IPOPT_EOL;
1218 	return (optlen);
1219 }
1220 
1221 /*
1222  * IP socket option processing.
1223  */
1224 int
1225 ip_ctloutput(so, sopt)
1226 	struct socket *so;
1227 	struct sockopt *sopt;
1228 {
1229 	struct	inpcb *inp = sotoinpcb(so);
1230 	int	error, optval;
1231 
1232 	error = optval = 0;
1233 	if (sopt->sopt_level != IPPROTO_IP) {
1234 		return (EINVAL);
1235 	}
1236 
1237 	switch (sopt->sopt_dir) {
1238 	case SOPT_SET:
1239 		switch (sopt->sopt_name) {
1240 		case IP_OPTIONS:
1241 #ifdef notyet
1242 		case IP_RETOPTS:
1243 #endif
1244 		{
1245 			struct mbuf *m;
1246 			if (sopt->sopt_valsize > MLEN) {
1247 				error = EMSGSIZE;
1248 				break;
1249 			}
1250 			MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1251 			if (m == 0) {
1252 				error = ENOBUFS;
1253 				break;
1254 			}
1255 			m->m_len = sopt->sopt_valsize;
1256 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1257 					    m->m_len);
1258 
1259 			return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1260 					   m));
1261 		}
1262 
1263 		case IP_TOS:
1264 		case IP_TTL:
1265 		case IP_RECVOPTS:
1266 		case IP_RECVRETOPTS:
1267 		case IP_RECVDSTADDR:
1268 		case IP_RECVIF:
1269 		case IP_FAITH:
1270 			error = sooptcopyin(sopt, &optval, sizeof optval,
1271 					    sizeof optval);
1272 			if (error)
1273 				break;
1274 
1275 			switch (sopt->sopt_name) {
1276 			case IP_TOS:
1277 				inp->inp_ip_tos = optval;
1278 				break;
1279 
1280 			case IP_TTL:
1281 				inp->inp_ip_ttl = optval;
1282 				break;
1283 #define	OPTSET(bit) \
1284 	if (optval) \
1285 		inp->inp_flags |= bit; \
1286 	else \
1287 		inp->inp_flags &= ~bit;
1288 
1289 			case IP_RECVOPTS:
1290 				OPTSET(INP_RECVOPTS);
1291 				break;
1292 
1293 			case IP_RECVRETOPTS:
1294 				OPTSET(INP_RECVRETOPTS);
1295 				break;
1296 
1297 			case IP_RECVDSTADDR:
1298 				OPTSET(INP_RECVDSTADDR);
1299 				break;
1300 
1301 			case IP_RECVIF:
1302 				OPTSET(INP_RECVIF);
1303 				break;
1304 
1305 			case IP_FAITH:
1306 				OPTSET(INP_FAITH);
1307 				break;
1308 			}
1309 			break;
1310 #undef OPTSET
1311 
1312 		case IP_MULTICAST_IF:
1313 		case IP_MULTICAST_VIF:
1314 		case IP_MULTICAST_TTL:
1315 		case IP_MULTICAST_LOOP:
1316 		case IP_ADD_MEMBERSHIP:
1317 		case IP_DROP_MEMBERSHIP:
1318 			error = ip_setmoptions(sopt, &inp->inp_moptions);
1319 			break;
1320 
1321 		case IP_PORTRANGE:
1322 			error = sooptcopyin(sopt, &optval, sizeof optval,
1323 					    sizeof optval);
1324 			if (error)
1325 				break;
1326 
1327 			switch (optval) {
1328 			case IP_PORTRANGE_DEFAULT:
1329 				inp->inp_flags &= ~(INP_LOWPORT);
1330 				inp->inp_flags &= ~(INP_HIGHPORT);
1331 				break;
1332 
1333 			case IP_PORTRANGE_HIGH:
1334 				inp->inp_flags &= ~(INP_LOWPORT);
1335 				inp->inp_flags |= INP_HIGHPORT;
1336 				break;
1337 
1338 			case IP_PORTRANGE_LOW:
1339 				inp->inp_flags &= ~(INP_HIGHPORT);
1340 				inp->inp_flags |= INP_LOWPORT;
1341 				break;
1342 
1343 			default:
1344 				error = EINVAL;
1345 				break;
1346 			}
1347 			break;
1348 
1349 #ifdef IPSEC
1350 		case IP_IPSEC_POLICY:
1351 		{
1352 			caddr_t req;
1353 			size_t len = 0;
1354 			int priv;
1355 			struct mbuf *m;
1356 			int optname;
1357 
1358 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1359 				break;
1360 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1361 				break;
1362 			priv = (sopt->sopt_td != NULL &&
1363 				suser(sopt->sopt_td) != 0) ? 0 : 1;
1364 			req = mtod(m, caddr_t);
1365 			len = m->m_len;
1366 			optname = sopt->sopt_name;
1367 			error = ipsec4_set_policy(inp, optname, req, len, priv);
1368 			m_freem(m);
1369 			break;
1370 		}
1371 #endif /*IPSEC*/
1372 
1373 		default:
1374 			error = ENOPROTOOPT;
1375 			break;
1376 		}
1377 		break;
1378 
1379 	case SOPT_GET:
1380 		switch (sopt->sopt_name) {
1381 		case IP_OPTIONS:
1382 		case IP_RETOPTS:
1383 			if (inp->inp_options)
1384 				error = sooptcopyout(sopt,
1385 						     mtod(inp->inp_options,
1386 							  char *),
1387 						     inp->inp_options->m_len);
1388 			else
1389 				sopt->sopt_valsize = 0;
1390 			break;
1391 
1392 		case IP_TOS:
1393 		case IP_TTL:
1394 		case IP_RECVOPTS:
1395 		case IP_RECVRETOPTS:
1396 		case IP_RECVDSTADDR:
1397 		case IP_RECVIF:
1398 		case IP_PORTRANGE:
1399 		case IP_FAITH:
1400 			switch (sopt->sopt_name) {
1401 
1402 			case IP_TOS:
1403 				optval = inp->inp_ip_tos;
1404 				break;
1405 
1406 			case IP_TTL:
1407 				optval = inp->inp_ip_ttl;
1408 				break;
1409 
1410 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1411 
1412 			case IP_RECVOPTS:
1413 				optval = OPTBIT(INP_RECVOPTS);
1414 				break;
1415 
1416 			case IP_RECVRETOPTS:
1417 				optval = OPTBIT(INP_RECVRETOPTS);
1418 				break;
1419 
1420 			case IP_RECVDSTADDR:
1421 				optval = OPTBIT(INP_RECVDSTADDR);
1422 				break;
1423 
1424 			case IP_RECVIF:
1425 				optval = OPTBIT(INP_RECVIF);
1426 				break;
1427 
1428 			case IP_PORTRANGE:
1429 				if (inp->inp_flags & INP_HIGHPORT)
1430 					optval = IP_PORTRANGE_HIGH;
1431 				else if (inp->inp_flags & INP_LOWPORT)
1432 					optval = IP_PORTRANGE_LOW;
1433 				else
1434 					optval = 0;
1435 				break;
1436 
1437 			case IP_FAITH:
1438 				optval = OPTBIT(INP_FAITH);
1439 				break;
1440 			}
1441 			error = sooptcopyout(sopt, &optval, sizeof optval);
1442 			break;
1443 
1444 		case IP_MULTICAST_IF:
1445 		case IP_MULTICAST_VIF:
1446 		case IP_MULTICAST_TTL:
1447 		case IP_MULTICAST_LOOP:
1448 		case IP_ADD_MEMBERSHIP:
1449 		case IP_DROP_MEMBERSHIP:
1450 			error = ip_getmoptions(sopt, inp->inp_moptions);
1451 			break;
1452 
1453 #ifdef IPSEC
1454 		case IP_IPSEC_POLICY:
1455 		{
1456 			struct mbuf *m = NULL;
1457 			caddr_t req = NULL;
1458 			size_t len = 0;
1459 
1460 			if (m != 0) {
1461 				req = mtod(m, caddr_t);
1462 				len = m->m_len;
1463 			}
1464 			error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1465 			if (error == 0)
1466 				error = soopt_mcopyout(sopt, m); /* XXX */
1467 			if (error == 0)
1468 				m_freem(m);
1469 			break;
1470 		}
1471 #endif /*IPSEC*/
1472 
1473 		default:
1474 			error = ENOPROTOOPT;
1475 			break;
1476 		}
1477 		break;
1478 	}
1479 	return (error);
1480 }
1481 
1482 /*
1483  * Set up IP options in pcb for insertion in output packets.
1484  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1485  * with destination address if source routed.
1486  */
1487 static int
1488 ip_pcbopts(optname, pcbopt, m)
1489 	int optname;
1490 	struct mbuf **pcbopt;
1491 	register struct mbuf *m;
1492 {
1493 	register int cnt, optlen;
1494 	register u_char *cp;
1495 	u_char opt;
1496 
1497 	/* turn off any old options */
1498 	if (*pcbopt)
1499 		(void)m_free(*pcbopt);
1500 	*pcbopt = 0;
1501 	if (m == (struct mbuf *)0 || m->m_len == 0) {
1502 		/*
1503 		 * Only turning off any previous options.
1504 		 */
1505 		if (m)
1506 			(void)m_free(m);
1507 		return (0);
1508 	}
1509 
1510 	if (m->m_len % sizeof(int32_t))
1511 		goto bad;
1512 	/*
1513 	 * IP first-hop destination address will be stored before
1514 	 * actual options; move other options back
1515 	 * and clear it when none present.
1516 	 */
1517 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1518 		goto bad;
1519 	cnt = m->m_len;
1520 	m->m_len += sizeof(struct in_addr);
1521 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1522 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1523 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1524 
1525 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1526 		opt = cp[IPOPT_OPTVAL];
1527 		if (opt == IPOPT_EOL)
1528 			break;
1529 		if (opt == IPOPT_NOP)
1530 			optlen = 1;
1531 		else {
1532 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1533 				goto bad;
1534 			optlen = cp[IPOPT_OLEN];
1535 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1536 				goto bad;
1537 		}
1538 		switch (opt) {
1539 
1540 		default:
1541 			break;
1542 
1543 		case IPOPT_LSRR:
1544 		case IPOPT_SSRR:
1545 			/*
1546 			 * user process specifies route as:
1547 			 *	->A->B->C->D
1548 			 * D must be our final destination (but we can't
1549 			 * check that since we may not have connected yet).
1550 			 * A is first hop destination, which doesn't appear in
1551 			 * actual IP option, but is stored before the options.
1552 			 */
1553 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1554 				goto bad;
1555 			m->m_len -= sizeof(struct in_addr);
1556 			cnt -= sizeof(struct in_addr);
1557 			optlen -= sizeof(struct in_addr);
1558 			cp[IPOPT_OLEN] = optlen;
1559 			/*
1560 			 * Move first hop before start of options.
1561 			 */
1562 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1563 			    sizeof(struct in_addr));
1564 			/*
1565 			 * Then copy rest of options back
1566 			 * to close up the deleted entry.
1567 			 */
1568 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1569 			    sizeof(struct in_addr)),
1570 			    (caddr_t)&cp[IPOPT_OFFSET+1],
1571 			    (unsigned)cnt + sizeof(struct in_addr));
1572 			break;
1573 		}
1574 	}
1575 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1576 		goto bad;
1577 	*pcbopt = m;
1578 	return (0);
1579 
1580 bad:
1581 	(void)m_free(m);
1582 	return (EINVAL);
1583 }
1584 
1585 /*
1586  * XXX
1587  * The whole multicast option thing needs to be re-thought.
1588  * Several of these options are equally applicable to non-multicast
1589  * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1590  * standard option (IP_TTL).
1591  */
1592 
1593 /*
1594  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1595  */
1596 static struct ifnet *
1597 ip_multicast_if(a, ifindexp)
1598 	struct in_addr *a;
1599 	int *ifindexp;
1600 {
1601 	int ifindex;
1602 	struct ifnet *ifp;
1603 
1604 	if (ifindexp)
1605 		*ifindexp = 0;
1606 	if (ntohl(a->s_addr) >> 24 == 0) {
1607 		ifindex = ntohl(a->s_addr) & 0xffffff;
1608 		if (ifindex < 0 || if_index < ifindex)
1609 			return NULL;
1610 		ifp = ifnet_byindex(ifindex);
1611 		if (ifindexp)
1612 			*ifindexp = ifindex;
1613 	} else {
1614 		INADDR_TO_IFP(*a, ifp);
1615 	}
1616 	return ifp;
1617 }
1618 
1619 /*
1620  * Set the IP multicast options in response to user setsockopt().
1621  */
1622 static int
1623 ip_setmoptions(sopt, imop)
1624 	struct sockopt *sopt;
1625 	struct ip_moptions **imop;
1626 {
1627 	int error = 0;
1628 	int i;
1629 	struct in_addr addr;
1630 	struct ip_mreq mreq;
1631 	struct ifnet *ifp;
1632 	struct ip_moptions *imo = *imop;
1633 	struct route ro;
1634 	struct sockaddr_in *dst;
1635 	int ifindex;
1636 	int s;
1637 
1638 	if (imo == NULL) {
1639 		/*
1640 		 * No multicast option buffer attached to the pcb;
1641 		 * allocate one and initialize to default values.
1642 		 */
1643 		imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1644 		    M_WAITOK);
1645 
1646 		if (imo == NULL)
1647 			return (ENOBUFS);
1648 		*imop = imo;
1649 		imo->imo_multicast_ifp = NULL;
1650 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
1651 		imo->imo_multicast_vif = -1;
1652 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1653 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1654 		imo->imo_num_memberships = 0;
1655 	}
1656 
1657 	switch (sopt->sopt_name) {
1658 	/* store an index number for the vif you wanna use in the send */
1659 	case IP_MULTICAST_VIF:
1660 		if (legal_vif_num == 0) {
1661 			error = EOPNOTSUPP;
1662 			break;
1663 		}
1664 		error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1665 		if (error)
1666 			break;
1667 		if (!legal_vif_num(i) && (i != -1)) {
1668 			error = EINVAL;
1669 			break;
1670 		}
1671 		imo->imo_multicast_vif = i;
1672 		break;
1673 
1674 	case IP_MULTICAST_IF:
1675 		/*
1676 		 * Select the interface for outgoing multicast packets.
1677 		 */
1678 		error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1679 		if (error)
1680 			break;
1681 		/*
1682 		 * INADDR_ANY is used to remove a previous selection.
1683 		 * When no interface is selected, a default one is
1684 		 * chosen every time a multicast packet is sent.
1685 		 */
1686 		if (addr.s_addr == INADDR_ANY) {
1687 			imo->imo_multicast_ifp = NULL;
1688 			break;
1689 		}
1690 		/*
1691 		 * The selected interface is identified by its local
1692 		 * IP address.  Find the interface and confirm that
1693 		 * it supports multicasting.
1694 		 */
1695 		s = splimp();
1696 		ifp = ip_multicast_if(&addr, &ifindex);
1697 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1698 			splx(s);
1699 			error = EADDRNOTAVAIL;
1700 			break;
1701 		}
1702 		imo->imo_multicast_ifp = ifp;
1703 		if (ifindex)
1704 			imo->imo_multicast_addr = addr;
1705 		else
1706 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
1707 		splx(s);
1708 		break;
1709 
1710 	case IP_MULTICAST_TTL:
1711 		/*
1712 		 * Set the IP time-to-live for outgoing multicast packets.
1713 		 * The original multicast API required a char argument,
1714 		 * which is inconsistent with the rest of the socket API.
1715 		 * We allow either a char or an int.
1716 		 */
1717 		if (sopt->sopt_valsize == 1) {
1718 			u_char ttl;
1719 			error = sooptcopyin(sopt, &ttl, 1, 1);
1720 			if (error)
1721 				break;
1722 			imo->imo_multicast_ttl = ttl;
1723 		} else {
1724 			u_int ttl;
1725 			error = sooptcopyin(sopt, &ttl, sizeof ttl,
1726 					    sizeof ttl);
1727 			if (error)
1728 				break;
1729 			if (ttl > 255)
1730 				error = EINVAL;
1731 			else
1732 				imo->imo_multicast_ttl = ttl;
1733 		}
1734 		break;
1735 
1736 	case IP_MULTICAST_LOOP:
1737 		/*
1738 		 * Set the loopback flag for outgoing multicast packets.
1739 		 * Must be zero or one.  The original multicast API required a
1740 		 * char argument, which is inconsistent with the rest
1741 		 * of the socket API.  We allow either a char or an int.
1742 		 */
1743 		if (sopt->sopt_valsize == 1) {
1744 			u_char loop;
1745 			error = sooptcopyin(sopt, &loop, 1, 1);
1746 			if (error)
1747 				break;
1748 			imo->imo_multicast_loop = !!loop;
1749 		} else {
1750 			u_int loop;
1751 			error = sooptcopyin(sopt, &loop, sizeof loop,
1752 					    sizeof loop);
1753 			if (error)
1754 				break;
1755 			imo->imo_multicast_loop = !!loop;
1756 		}
1757 		break;
1758 
1759 	case IP_ADD_MEMBERSHIP:
1760 		/*
1761 		 * Add a multicast group membership.
1762 		 * Group must be a valid IP multicast address.
1763 		 */
1764 		error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1765 		if (error)
1766 			break;
1767 
1768 		if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1769 			error = EINVAL;
1770 			break;
1771 		}
1772 		s = splimp();
1773 		/*
1774 		 * If no interface address was provided, use the interface of
1775 		 * the route to the given multicast address.
1776 		 */
1777 		if (mreq.imr_interface.s_addr == INADDR_ANY) {
1778 			bzero((caddr_t)&ro, sizeof(ro));
1779 			dst = (struct sockaddr_in *)&ro.ro_dst;
1780 			dst->sin_len = sizeof(*dst);
1781 			dst->sin_family = AF_INET;
1782 			dst->sin_addr = mreq.imr_multiaddr;
1783 			rtalloc(&ro);
1784 			if (ro.ro_rt == NULL) {
1785 				error = EADDRNOTAVAIL;
1786 				splx(s);
1787 				break;
1788 			}
1789 			ifp = ro.ro_rt->rt_ifp;
1790 			rtfree(ro.ro_rt);
1791 		}
1792 		else {
1793 			ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1794 		}
1795 
1796 		/*
1797 		 * See if we found an interface, and confirm that it
1798 		 * supports multicast.
1799 		 */
1800 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1801 			error = EADDRNOTAVAIL;
1802 			splx(s);
1803 			break;
1804 		}
1805 		/*
1806 		 * See if the membership already exists or if all the
1807 		 * membership slots are full.
1808 		 */
1809 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1810 			if (imo->imo_membership[i]->inm_ifp == ifp &&
1811 			    imo->imo_membership[i]->inm_addr.s_addr
1812 						== mreq.imr_multiaddr.s_addr)
1813 				break;
1814 		}
1815 		if (i < imo->imo_num_memberships) {
1816 			error = EADDRINUSE;
1817 			splx(s);
1818 			break;
1819 		}
1820 		if (i == IP_MAX_MEMBERSHIPS) {
1821 			error = ETOOMANYREFS;
1822 			splx(s);
1823 			break;
1824 		}
1825 		/*
1826 		 * Everything looks good; add a new record to the multicast
1827 		 * address list for the given interface.
1828 		 */
1829 		if ((imo->imo_membership[i] =
1830 		    in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1831 			error = ENOBUFS;
1832 			splx(s);
1833 			break;
1834 		}
1835 		++imo->imo_num_memberships;
1836 		splx(s);
1837 		break;
1838 
1839 	case IP_DROP_MEMBERSHIP:
1840 		/*
1841 		 * Drop a multicast group membership.
1842 		 * Group must be a valid IP multicast address.
1843 		 */
1844 		error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1845 		if (error)
1846 			break;
1847 
1848 		if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1849 			error = EINVAL;
1850 			break;
1851 		}
1852 
1853 		s = splimp();
1854 		/*
1855 		 * If an interface address was specified, get a pointer
1856 		 * to its ifnet structure.
1857 		 */
1858 		if (mreq.imr_interface.s_addr == INADDR_ANY)
1859 			ifp = NULL;
1860 		else {
1861 			ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1862 			if (ifp == NULL) {
1863 				error = EADDRNOTAVAIL;
1864 				splx(s);
1865 				break;
1866 			}
1867 		}
1868 		/*
1869 		 * Find the membership in the membership array.
1870 		 */
1871 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1872 			if ((ifp == NULL ||
1873 			     imo->imo_membership[i]->inm_ifp == ifp) &&
1874 			     imo->imo_membership[i]->inm_addr.s_addr ==
1875 			     mreq.imr_multiaddr.s_addr)
1876 				break;
1877 		}
1878 		if (i == imo->imo_num_memberships) {
1879 			error = EADDRNOTAVAIL;
1880 			splx(s);
1881 			break;
1882 		}
1883 		/*
1884 		 * Give up the multicast address record to which the
1885 		 * membership points.
1886 		 */
1887 		in_delmulti(imo->imo_membership[i]);
1888 		/*
1889 		 * Remove the gap in the membership array.
1890 		 */
1891 		for (++i; i < imo->imo_num_memberships; ++i)
1892 			imo->imo_membership[i-1] = imo->imo_membership[i];
1893 		--imo->imo_num_memberships;
1894 		splx(s);
1895 		break;
1896 
1897 	default:
1898 		error = EOPNOTSUPP;
1899 		break;
1900 	}
1901 
1902 	/*
1903 	 * If all options have default values, no need to keep the mbuf.
1904 	 */
1905 	if (imo->imo_multicast_ifp == NULL &&
1906 	    imo->imo_multicast_vif == -1 &&
1907 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1908 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1909 	    imo->imo_num_memberships == 0) {
1910 		free(*imop, M_IPMOPTS);
1911 		*imop = NULL;
1912 	}
1913 
1914 	return (error);
1915 }
1916 
1917 /*
1918  * Return the IP multicast options in response to user getsockopt().
1919  */
1920 static int
1921 ip_getmoptions(sopt, imo)
1922 	struct sockopt *sopt;
1923 	register struct ip_moptions *imo;
1924 {
1925 	struct in_addr addr;
1926 	struct in_ifaddr *ia;
1927 	int error, optval;
1928 	u_char coptval;
1929 
1930 	error = 0;
1931 	switch (sopt->sopt_name) {
1932 	case IP_MULTICAST_VIF:
1933 		if (imo != NULL)
1934 			optval = imo->imo_multicast_vif;
1935 		else
1936 			optval = -1;
1937 		error = sooptcopyout(sopt, &optval, sizeof optval);
1938 		break;
1939 
1940 	case IP_MULTICAST_IF:
1941 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
1942 			addr.s_addr = INADDR_ANY;
1943 		else if (imo->imo_multicast_addr.s_addr) {
1944 			/* return the value user has set */
1945 			addr = imo->imo_multicast_addr;
1946 		} else {
1947 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
1948 			addr.s_addr = (ia == NULL) ? INADDR_ANY
1949 				: IA_SIN(ia)->sin_addr.s_addr;
1950 		}
1951 		error = sooptcopyout(sopt, &addr, sizeof addr);
1952 		break;
1953 
1954 	case IP_MULTICAST_TTL:
1955 		if (imo == 0)
1956 			optval = coptval = IP_DEFAULT_MULTICAST_TTL;
1957 		else
1958 			optval = coptval = imo->imo_multicast_ttl;
1959 		if (sopt->sopt_valsize == 1)
1960 			error = sooptcopyout(sopt, &coptval, 1);
1961 		else
1962 			error = sooptcopyout(sopt, &optval, sizeof optval);
1963 		break;
1964 
1965 	case IP_MULTICAST_LOOP:
1966 		if (imo == 0)
1967 			optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
1968 		else
1969 			optval = coptval = imo->imo_multicast_loop;
1970 		if (sopt->sopt_valsize == 1)
1971 			error = sooptcopyout(sopt, &coptval, 1);
1972 		else
1973 			error = sooptcopyout(sopt, &optval, sizeof optval);
1974 		break;
1975 
1976 	default:
1977 		error = ENOPROTOOPT;
1978 		break;
1979 	}
1980 	return (error);
1981 }
1982 
1983 /*
1984  * Discard the IP multicast options.
1985  */
1986 void
1987 ip_freemoptions(imo)
1988 	register struct ip_moptions *imo;
1989 {
1990 	register int i;
1991 
1992 	if (imo != NULL) {
1993 		for (i = 0; i < imo->imo_num_memberships; ++i)
1994 			in_delmulti(imo->imo_membership[i]);
1995 		free(imo, M_IPMOPTS);
1996 	}
1997 }
1998 
1999 /*
2000  * Routine called from ip_output() to loop back a copy of an IP multicast
2001  * packet to the input queue of a specified interface.  Note that this
2002  * calls the output routine of the loopback "driver", but with an interface
2003  * pointer that might NOT be a loopback interface -- evil, but easier than
2004  * replicating that code here.
2005  */
2006 static void
2007 ip_mloopback(ifp, m, dst, hlen)
2008 	struct ifnet *ifp;
2009 	register struct mbuf *m;
2010 	register struct sockaddr_in *dst;
2011 	int hlen;
2012 {
2013 	register struct ip *ip;
2014 	struct mbuf *copym;
2015 
2016 	copym = m_copy(m, 0, M_COPYALL);
2017 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2018 		copym = m_pullup(copym, hlen);
2019 	if (copym != NULL) {
2020 		/*
2021 		 * We don't bother to fragment if the IP length is greater
2022 		 * than the interface's MTU.  Can this possibly matter?
2023 		 */
2024 		ip = mtod(copym, struct ip *);
2025 		ip->ip_len = htons(ip->ip_len);
2026 		ip->ip_off = htons(ip->ip_off);
2027 		ip->ip_sum = 0;
2028 		if (ip->ip_vhl == IP_VHL_BORING) {
2029 			ip->ip_sum = in_cksum_hdr(ip);
2030 		} else {
2031 			ip->ip_sum = in_cksum(copym, hlen);
2032 		}
2033 		/*
2034 		 * NB:
2035 		 * It's not clear whether there are any lingering
2036 		 * reentrancy problems in other areas which might
2037 		 * be exposed by using ip_input directly (in
2038 		 * particular, everything which modifies the packet
2039 		 * in-place).  Yet another option is using the
2040 		 * protosw directly to deliver the looped back
2041 		 * packet.  For the moment, we'll err on the side
2042 		 * of safety by using if_simloop().
2043 		 */
2044 #if 1 /* XXX */
2045 		if (dst->sin_family != AF_INET) {
2046 			printf("ip_mloopback: bad address family %d\n",
2047 						dst->sin_family);
2048 			dst->sin_family = AF_INET;
2049 		}
2050 #endif
2051 
2052 #ifdef notdef
2053 		copym->m_pkthdr.rcvif = ifp;
2054 		ip_input(copym);
2055 #else
2056 		/* if the checksum hasn't been computed, mark it as valid */
2057 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2058 			copym->m_pkthdr.csum_flags |=
2059 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2060 			copym->m_pkthdr.csum_data = 0xffff;
2061 		}
2062 		if_simloop(ifp, copym, dst->sin_family, 0);
2063 #endif
2064 	}
2065 }
2066