xref: /openbsd-src/sys/netinet/ip_output.c (revision 0b7734b3d77bb9b21afec6f4621cae6c805dbd45)
1 /*	$OpenBSD: ip_output.c,v 1.325 2016/07/01 18:28:58 jca Exp $	*/
2 /*	$NetBSD: ip_output.c,v 1.28 1996/02/13 23:43:07 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
33  */
34 
35 #include "pf.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45 
46 #include <net/if.h>
47 #include <net/if_var.h>
48 #include <net/if_enc.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/tcp.h>
58 #include <netinet/udp.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/udp_var.h>
62 
63 #if NPF > 0
64 #include <net/pfvar.h>
65 #endif
66 
67 #ifdef IPSEC
68 #ifdef ENCDEBUG
69 #define DPRINTF(x)    do { if (encdebug) printf x ; } while (0)
70 #else
71 #define DPRINTF(x)
72 #endif
73 #endif /* IPSEC */
74 
75 void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
76 static __inline u_int16_t __attribute__((__unused__))
77     in_cksum_phdr(u_int32_t, u_int32_t, u_int32_t);
78 void in_delayed_cksum(struct mbuf *);
79 
80 #ifdef IPSEC
81 struct tdb *
82 ip_output_ipsec_lookup(struct mbuf *m, int hlen, int *error, struct inpcb *inp,
83     int ipsecflowinfo);
84 int
85 ip_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct ifnet *ifp,
86     struct route *ro);
87 #endif /* IPSEC */
88 
89 /*
90  * IP output.  The packet in mbuf chain m contains a skeletal IP
91  * header (with len, off, ttl, proto, tos, src, dst).
92  * The mbuf chain containing the packet will be freed.
93  * The mbuf opt, if present, will not be freed.
94  */
95 int
96 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, int flags,
97     struct ip_moptions *imo, struct inpcb *inp, u_int32_t ipsecflowinfo)
98 {
99 	struct ip *ip;
100 	struct ifnet *ifp = NULL;
101 	struct mbuf *m = m0;
102 	int hlen = sizeof (struct ip);
103 	int len, error = 0;
104 	struct route iproute;
105 	struct sockaddr_in *dst;
106 	struct tdb *tdb = NULL;
107 	u_long mtu;
108 #if defined(MROUTING)
109 	int rv;
110 #endif
111 
112 #ifdef IPSEC
113 	if (inp && (inp->inp_flags & INP_IPV6) != 0)
114 		panic("ip_output: IPv6 pcb is passed");
115 #endif /* IPSEC */
116 
117 #ifdef	DIAGNOSTIC
118 	if ((m->m_flags & M_PKTHDR) == 0)
119 		panic("ip_output no HDR");
120 #endif
121 	if (opt) {
122 		m = ip_insertoptions(m, opt, &len);
123 		hlen = len;
124 	}
125 
126 	ip = mtod(m, struct ip *);
127 
128 	/*
129 	 * Fill in IP header.
130 	 */
131 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
132 		ip->ip_v = IPVERSION;
133 		ip->ip_off &= htons(IP_DF);
134 		ip->ip_id = htons(ip_randomid());
135 		ip->ip_hl = hlen >> 2;
136 		ipstat.ips_localout++;
137 	} else {
138 		hlen = ip->ip_hl << 2;
139 	}
140 
141 	/*
142 	 * We should not send traffic to 0/8 say both Stevens and RFCs
143 	 * 5735 section 3 and 1122 sections 3.2.1.3 and 3.3.6.
144 	 */
145 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == 0) {
146 		error = ENETUNREACH;
147 		goto bad;
148 	}
149 
150 #if NPF > 0
151 reroute:
152 #endif
153 
154 	/*
155 	 * Do a route lookup now in case we need the source address to
156 	 * do an SPD lookup in IPsec; for most packets, the source address
157 	 * is set at a higher level protocol. ICMPs and other packets
158 	 * though (e.g., traceroute) have a source address of zeroes.
159 	 */
160 	if (ro == NULL) {
161 		ro = &iproute;
162 		memset(ro, 0, sizeof(*ro));
163 	}
164 
165 	dst = satosin(&ro->ro_dst);
166 
167 	/*
168 	 * If there is a cached route, check that it is to the same
169 	 * destination and is still up.  If not, free it and try again.
170 	 */
171 	if (!rtisvalid(ro->ro_rt) ||
172 	    dst->sin_addr.s_addr != ip->ip_dst.s_addr ||
173 	    ro->ro_tableid != m->m_pkthdr.ph_rtableid) {
174 		rtfree(ro->ro_rt);
175 		ro->ro_rt = NULL;
176 	}
177 
178 	if (ro->ro_rt == NULL) {
179 		dst->sin_family = AF_INET;
180 		dst->sin_len = sizeof(*dst);
181 		dst->sin_addr = ip->ip_dst;
182 		ro->ro_tableid = m->m_pkthdr.ph_rtableid;
183 	}
184 
185 	if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
186 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
187 	    imo != NULL && (ifp = if_get(imo->imo_ifidx)) != NULL) {
188 
189 		mtu = ifp->if_mtu;
190 		if (ip->ip_src.s_addr == INADDR_ANY) {
191 			struct in_ifaddr *ia;
192 
193 			KERNEL_LOCK();
194 			IFP_TO_IA(ifp, ia);
195 			if (ia != NULL)
196 				ip->ip_src = ia->ia_addr.sin_addr;
197 			KERNEL_UNLOCK();
198 		}
199 	} else {
200 		struct in_ifaddr *ia;
201 
202 		if (ro->ro_rt == NULL)
203 			ro->ro_rt = rtalloc_mpath(&ro->ro_dst,
204 			    &ip->ip_src.s_addr, ro->ro_tableid);
205 
206 		if (ro->ro_rt == NULL) {
207 			ipstat.ips_noroute++;
208 			error = EHOSTUNREACH;
209 			goto bad;
210 		}
211 
212 		ia = ifatoia(ro->ro_rt->rt_ifa);
213 		if (ISSET(ro->ro_rt->rt_flags, RTF_LOCAL))
214 			ifp = if_get(lo0ifidx);
215 		else
216 			ifp = if_get(ro->ro_rt->rt_ifidx);
217 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
218 			mtu = ifp->if_mtu;
219 
220 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
221 			dst = satosin(ro->ro_rt->rt_gateway);
222 
223 		/* Set the source IP address */
224 		if (ip->ip_src.s_addr == INADDR_ANY && ia)
225 			ip->ip_src = ia->ia_addr.sin_addr;
226 	}
227 
228 #ifdef IPSEC
229 	if (ipsec_in_use || inp != NULL) {
230 		KERNEL_LOCK();
231 		/* Do we have any pending SAs to apply ? */
232 		tdb = ip_output_ipsec_lookup(m, hlen, &error, inp,
233 		    ipsecflowinfo);
234 		KERNEL_UNLOCK();
235 		if (error != 0) {
236 			/* Should silently drop packet */
237 			if (error == -EINVAL)
238 				error = 0;
239 			m_freem(m);
240 			goto done;
241 		}
242 		if (tdb != NULL) {
243 			/*
244 			 * If it needs TCP/UDP hardware-checksumming, do the
245 			 * computation now.
246 			 */
247 			in_proto_cksum_out(m, NULL);
248 
249 			/* If it's not a multicast packet, try to fast-path */
250 			if (!IN_MULTICAST(ip->ip_dst.s_addr)) {
251 				goto sendit;
252 			}
253 		}
254 	}
255 #endif /* IPSEC */
256 
257 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
258 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
259 
260 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
261 			M_BCAST : M_MCAST;
262 
263 		/*
264 		 * IP destination address is multicast.  Make sure "dst"
265 		 * still points to the address in "ro".  (It may have been
266 		 * changed to point to a gateway address, above.)
267 		 */
268 		dst = satosin(&ro->ro_dst);
269 
270 		/*
271 		 * See if the caller provided any multicast options
272 		 */
273 		if (imo != NULL)
274 			ip->ip_ttl = imo->imo_ttl;
275 		else
276 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
277 
278 		/*
279 		 * if we don't know the outgoing ifp yet, we can't generate
280 		 * output
281 		 */
282 		if (!ifp) {
283 			ipstat.ips_noroute++;
284 			error = EHOSTUNREACH;
285 			goto bad;
286 		}
287 
288 		/*
289 		 * Confirm that the outgoing interface supports multicast,
290 		 * but only if the packet actually is going out on that
291 		 * interface (i.e., no IPsec is applied).
292 		 */
293 		if ((((m->m_flags & M_MCAST) &&
294 		      (ifp->if_flags & IFF_MULTICAST) == 0) ||
295 		     ((m->m_flags & M_BCAST) &&
296 		      (ifp->if_flags & IFF_BROADCAST) == 0)) && (tdb == NULL)) {
297 			ipstat.ips_noroute++;
298 			error = ENETUNREACH;
299 			goto bad;
300 		}
301 
302 		/*
303 		 * If source address not specified yet, use address
304 		 * of outgoing interface.
305 		 */
306 		if (ip->ip_src.s_addr == INADDR_ANY) {
307 			struct in_ifaddr *ia;
308 
309 			KERNEL_LOCK();
310 			IFP_TO_IA(ifp, ia);
311 			if (ia != NULL)
312 				ip->ip_src = ia->ia_addr.sin_addr;
313 			KERNEL_UNLOCK();
314 		}
315 
316 		if ((imo == NULL || imo->imo_loop) &&
317 		    in_hasmulti(&ip->ip_dst, ifp)) {
318 			/*
319 			 * If we belong to the destination multicast group
320 			 * on the outgoing interface, and the caller did not
321 			 * forbid loopback, loop back a copy.
322 			 * Can't defer TCP/UDP checksumming, do the
323 			 * computation now.
324 			 */
325 			in_proto_cksum_out(m, NULL);
326 			ip_mloopback(ifp, m, dst);
327 		}
328 #ifdef MROUTING
329 		else {
330 			/*
331 			 * If we are acting as a multicast router, perform
332 			 * multicast forwarding as if the packet had just
333 			 * arrived on the interface to which we are about
334 			 * to send.  The multicast forwarding function
335 			 * recursively calls this function, using the
336 			 * IP_FORWARDING flag to prevent infinite recursion.
337 			 *
338 			 * Multicasts that are looped back by ip_mloopback(),
339 			 * above, will be forwarded by the ip_input() routine,
340 			 * if necessary.
341 			 */
342 			if (ipmforwarding && ip_mrouter &&
343 			    (flags & IP_FORWARDING) == 0) {
344 				KERNEL_LOCK();
345 				rv = ip_mforward(m, ifp);
346 				KERNEL_UNLOCK();
347 				if (rv != 0) {
348 					m_freem(m);
349 					goto done;
350 				}
351 			}
352 		}
353 #endif
354 		/*
355 		 * Multicasts with a time-to-live of zero may be looped-
356 		 * back, above, but must not be transmitted on a network.
357 		 * Also, multicasts addressed to the loopback interface
358 		 * are not sent -- the above call to ip_mloopback() will
359 		 * loop back a copy if this host actually belongs to the
360 		 * destination group on the loopback interface.
361 		 */
362 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
363 			m_freem(m);
364 			goto done;
365 		}
366 
367 		goto sendit;
368 	}
369 
370 	/*
371 	 * Look for broadcast address and verify user is allowed to send
372 	 * such a packet; if the packet is going in an IPsec tunnel, skip
373 	 * this check.
374 	 */
375 	if ((tdb == NULL) && ((dst->sin_addr.s_addr == INADDR_BROADCAST) ||
376 	    (ro && ro->ro_rt && ISSET(ro->ro_rt->rt_flags, RTF_BROADCAST)))) {
377 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
378 			error = EADDRNOTAVAIL;
379 			goto bad;
380 		}
381 		if ((flags & IP_ALLOWBROADCAST) == 0) {
382 			error = EACCES;
383 			goto bad;
384 		}
385 
386 		/* Don't allow broadcast messages to be fragmented */
387 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
388 			error = EMSGSIZE;
389 			goto bad;
390 		}
391 		m->m_flags |= M_BCAST;
392 	} else
393 		m->m_flags &= ~M_BCAST;
394 
395 sendit:
396 	/*
397 	 * If we're doing Path MTU discovery, we need to set DF unless
398 	 * the route's MTU is locked.
399 	 */
400 	if ((flags & IP_MTUDISC) && ro && ro->ro_rt &&
401 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
402 		ip->ip_off |= htons(IP_DF);
403 
404 #ifdef IPSEC
405 	/*
406 	 * Check if the packet needs encapsulation.
407 	 */
408 	if (tdb != NULL) {
409 		KERNEL_LOCK();
410 		/* Callee frees mbuf */
411 		error = ip_output_ipsec_send(tdb, m, ifp, ro);
412 		KERNEL_UNLOCK();
413 		goto done;
414 	}
415 #endif /* IPSEC */
416 
417 	/*
418 	 * Packet filter
419 	 */
420 #if NPF > 0
421 	if (pf_test(AF_INET, PF_OUT, ifp, &m) != PF_PASS) {
422 		error = EACCES;
423 		m_freem(m);
424 		goto done;
425 	}
426 	if (m == NULL)
427 		goto done;
428 	ip = mtod(m, struct ip *);
429 	hlen = ip->ip_hl << 2;
430 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
431 	    (PF_TAG_REROUTE | PF_TAG_GENERATED))
432 		/* already rerun the route lookup, go on */
433 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
434 	else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
435 		/* tag as generated to skip over pf_test on rerun */
436 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
437 		ro = NULL;
438 		if_put(ifp); /* drop reference since target changed */
439 		ifp = NULL;
440 		goto reroute;
441 	}
442 #endif
443 	in_proto_cksum_out(m, ifp);
444 
445 #ifdef IPSEC
446 	if (ipsec_in_use && (flags & IP_FORWARDING) && (ipforwarding == 2) &&
447 	    (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) == NULL)) {
448 		error = EHOSTUNREACH;
449 		m_freem(m);
450 		goto done;
451 	}
452 #endif
453 
454 	/*
455 	 * If small enough for interface, can just send directly.
456 	 */
457 	if (ntohs(ip->ip_len) <= mtu) {
458 		ip->ip_sum = 0;
459 		if ((ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
460 		    (ifp->if_bridgeport == NULL))
461 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
462 		else {
463 			ipstat.ips_outswcsum++;
464 			ip->ip_sum = in_cksum(m, hlen);
465 		}
466 
467 		error = ifp->if_output(ifp, m, sintosa(dst), ro->ro_rt);
468 		goto done;
469 	}
470 
471 	/*
472 	 * Too large for interface; fragment if possible.
473 	 * Must be able to put at least 8 bytes per fragment.
474 	 */
475 	if (ip->ip_off & htons(IP_DF)) {
476 #ifdef IPSEC
477 		if (ip_mtudisc)
478 			ipsec_adjust_mtu(m, ifp->if_mtu);
479 #endif
480 		error = EMSGSIZE;
481 		/*
482 		 * This case can happen if the user changed the MTU
483 		 * of an interface after enabling IP on it.  Because
484 		 * most netifs don't keep track of routes pointing to
485 		 * them, there is no way for one to update all its
486 		 * routes when the MTU is changed.
487 		 */
488 		if (rtisvalid(ro->ro_rt) &&
489 		    ISSET(ro->ro_rt->rt_flags, RTF_HOST) &&
490 		    !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
491 		    (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
492 			ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
493 		}
494 		ipstat.ips_cantfrag++;
495 		goto bad;
496 	}
497 
498 	error = ip_fragment(m, ifp, mtu);
499 	if (error) {
500 		m = m0 = NULL;
501 		goto bad;
502 	}
503 
504 	for (; m; m = m0) {
505 		m0 = m->m_nextpkt;
506 		m->m_nextpkt = 0;
507 		if (error == 0)
508 			error = ifp->if_output(ifp, m, sintosa(dst), ro->ro_rt);
509 		else
510 			m_freem(m);
511 	}
512 
513 	if (error == 0)
514 		ipstat.ips_fragmented++;
515 
516 done:
517 	if (ro == &iproute && ro->ro_rt)
518 		rtfree(ro->ro_rt);
519 	if_put(ifp);
520 	return (error);
521 bad:
522 	m_freem(m0);
523 	goto done;
524 }
525 
526 #ifdef IPSEC
527 struct tdb *
528 ip_output_ipsec_lookup(struct mbuf *m, int hlen, int *error, struct inpcb *inp,
529     int ipsecflowinfo)
530 {
531 	struct m_tag *mtag;
532 	struct tdb_ident *tdbi;
533 	struct tdb *tdb;
534 
535 	/* Do we have any pending SAs to apply ? */
536 	tdb = ipsp_spd_lookup(m, AF_INET, hlen, error, IPSP_DIRECTION_OUT,
537 	    NULL, inp, ipsecflowinfo);
538 	if (tdb == NULL)
539 		return NULL;
540 	/* Loop detection */
541 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
542 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
543 			continue;
544 		tdbi = (struct tdb_ident *)(mtag + 1);
545 		if (tdbi->spi == tdb->tdb_spi &&
546 		    tdbi->proto == tdb->tdb_sproto &&
547 		    tdbi->rdomain == tdb->tdb_rdomain &&
548 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
549 		    sizeof(union sockaddr_union))) {
550 			/* no IPsec needed */
551 			return NULL;
552 		}
553 	}
554 	return tdb;
555 }
556 
557 int
558 ip_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct ifnet *ifp,
559     struct route *ro)
560 {
561 #if NPF > 0
562 	struct ifnet *encif;
563 #endif
564 	struct ip *ip;
565 
566 #if NPF > 0
567 	/*
568 	 * Packet filter
569 	 */
570 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
571 	    pf_test(AF_INET, PF_OUT, encif, &m) != PF_PASS) {
572 		m_freem(m);
573 		return EACCES;
574 	}
575 	if (m == NULL)
576 		return 0;
577 	/*
578 	 * PF_TAG_REROUTE handling or not...
579 	 * Packet is entering IPsec so the routing is
580 	 * already overruled by the IPsec policy.
581 	 * Until now the change was not reconsidered.
582 	 * What's the behaviour?
583 	 */
584 	in_proto_cksum_out(m, encif);
585 #endif
586 
587 	/* Check if we are allowed to fragment */
588 	ip = mtod(m, struct ip *);
589 	if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
590 	    ntohs(ip->ip_len) > tdb->tdb_mtu &&
591 	    tdb->tdb_mtutimeout > time_second) {
592 		struct rtentry *rt = NULL;
593 		int rt_mtucloned = 0;
594 		int transportmode = 0;
595 
596 		transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
597 		    (tdb->tdb_dst.sin.sin_addr.s_addr == ip->ip_dst.s_addr);
598 
599 		/* Find a host route to store the mtu in */
600 		if (ro != NULL)
601 			rt = ro->ro_rt;
602 		/* but don't add a PMTU route for transport mode SAs */
603 		if (transportmode)
604 			rt = NULL;
605 		else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
606 			rt = icmp_mtudisc_clone(ip->ip_dst,
607 			    m->m_pkthdr.ph_rtableid);
608 			rt_mtucloned = 1;
609 		}
610 		DPRINTF(("%s: spi %08x mtu %d rt %p cloned %d\n", __func__,
611 		    ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned));
612 		if (rt != NULL) {
613 			rt->rt_rmx.rmx_mtu = tdb->tdb_mtu;
614 			if (ro && ro->ro_rt != NULL) {
615 				rtfree(ro->ro_rt);
616 				ro->ro_rt = rtalloc(&ro->ro_dst, RT_RESOLVE,
617 				    m->m_pkthdr.ph_rtableid);
618 			}
619 			if (rt_mtucloned)
620 				rtfree(rt);
621 		}
622 		ipsec_adjust_mtu(m, tdb->tdb_mtu);
623 		m_freem(m);
624 		return EMSGSIZE;
625 	}
626 
627 	/*
628 	 * Clear these -- they'll be set in the recursive invocation
629 	 * as needed.
630 	 */
631 	m->m_flags &= ~(M_MCAST | M_BCAST);
632 
633 	/* Callee frees mbuf */
634 	return ipsp_process_packet(m, tdb, AF_INET, 0);
635 }
636 #endif /* IPSEC */
637 
638 int
639 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
640 {
641 	struct ip *ip, *mhip;
642 	struct mbuf *m0;
643 	int len, hlen, off;
644 	int mhlen, firstlen;
645 	struct mbuf **mnext;
646 	int fragments = 0;
647 	int error = 0;
648 
649 	ip = mtod(m, struct ip *);
650 	hlen = ip->ip_hl << 2;
651 
652 	len = (mtu - hlen) &~ 7;
653 	if (len < 8) {
654 		m_freem(m);
655 		return (EMSGSIZE);
656 	}
657 
658 	/*
659 	 * If we are doing fragmentation, we can't defer TCP/UDP
660 	 * checksumming; compute the checksum and clear the flag.
661 	 */
662 	in_proto_cksum_out(m, NULL);
663 	firstlen = len;
664 	mnext = &m->m_nextpkt;
665 
666 	/*
667 	 * Loop through length of segment after first fragment,
668 	 * make new header and copy data of each part and link onto chain.
669 	 */
670 	m0 = m;
671 	mhlen = sizeof (struct ip);
672 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
673 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
674 		if (m == NULL) {
675 			ipstat.ips_odropped++;
676 			error = ENOBUFS;
677 			goto sendorfree;
678 		}
679 		*mnext = m;
680 		mnext = &m->m_nextpkt;
681 		m->m_data += max_linkhdr;
682 		mhip = mtod(m, struct ip *);
683 		*mhip = *ip;
684 		/* we must inherit MCAST/BCAST flags, routing table and prio */
685 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
686 		m->m_pkthdr.ph_rtableid = m0->m_pkthdr.ph_rtableid;
687 		m->m_pkthdr.pf.prio = m0->m_pkthdr.pf.prio;
688 		if (hlen > sizeof (struct ip)) {
689 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
690 			mhip->ip_hl = mhlen >> 2;
691 		}
692 		m->m_len = mhlen;
693 		mhip->ip_off = ((off - hlen) >> 3) +
694 		    (ntohs(ip->ip_off) & ~IP_MF);
695 		if (ip->ip_off & htons(IP_MF))
696 			mhip->ip_off |= IP_MF;
697 		if (off + len >= ntohs(ip->ip_len))
698 			len = ntohs(ip->ip_len) - off;
699 		else
700 			mhip->ip_off |= IP_MF;
701 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
702 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
703 		if (m->m_next == 0) {
704 			ipstat.ips_odropped++;
705 			error = ENOBUFS;
706 			goto sendorfree;
707 		}
708 		m->m_pkthdr.len = mhlen + len;
709 		m->m_pkthdr.ph_ifidx = 0;
710 		mhip->ip_off = htons((u_int16_t)mhip->ip_off);
711 		mhip->ip_sum = 0;
712 		if ((ifp != NULL) &&
713 		    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
714 		    (ifp->if_bridgeport == NULL))
715 			m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
716 		else {
717 			ipstat.ips_outswcsum++;
718 			mhip->ip_sum = in_cksum(m, mhlen);
719 		}
720 		ipstat.ips_ofragments++;
721 		fragments++;
722 	}
723 	/*
724 	 * Update first fragment by trimming what's been copied out
725 	 * and updating header, then send each fragment (in order).
726 	 */
727 	m = m0;
728 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
729 	m->m_pkthdr.len = hlen + firstlen;
730 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
731 	ip->ip_off |= htons(IP_MF);
732 	ip->ip_sum = 0;
733 	if ((ifp != NULL) &&
734 	    (ifp->if_capabilities & IFCAP_CSUM_IPv4) &&
735 	    (ifp->if_bridgeport == NULL))
736 		m->m_pkthdr.csum_flags |= M_IPV4_CSUM_OUT;
737 	else {
738 		ipstat.ips_outswcsum++;
739 		ip->ip_sum = in_cksum(m, hlen);
740 	}
741 sendorfree:
742 	if (error) {
743 		for (m = m0; m; m = m0) {
744 			m0 = m->m_nextpkt;
745 			m->m_nextpkt = NULL;
746 			m_freem(m);
747 		}
748 	}
749 
750 	return (error);
751 }
752 
753 /*
754  * Insert IP options into preformed packet.
755  * Adjust IP destination as required for IP source routing,
756  * as indicated by a non-zero in_addr at the start of the options.
757  */
758 struct mbuf *
759 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
760 {
761 	struct ipoption *p = mtod(opt, struct ipoption *);
762 	struct mbuf *n;
763 	struct ip *ip = mtod(m, struct ip *);
764 	unsigned int optlen;
765 
766 	optlen = opt->m_len - sizeof(p->ipopt_dst);
767 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
768 		return (m);		/* XXX should fail */
769 	if (p->ipopt_dst.s_addr)
770 		ip->ip_dst = p->ipopt_dst;
771 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
772 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
773 		if (n == NULL)
774 			return (m);
775 		M_MOVE_HDR(n, m);
776 		n->m_pkthdr.len += optlen;
777 		m->m_len -= sizeof(struct ip);
778 		m->m_data += sizeof(struct ip);
779 		n->m_next = m;
780 		m = n;
781 		m->m_len = optlen + sizeof(struct ip);
782 		m->m_data += max_linkhdr;
783 		memcpy(mtod(m, caddr_t), ip, sizeof(struct ip));
784 	} else {
785 		m->m_data -= optlen;
786 		m->m_len += optlen;
787 		m->m_pkthdr.len += optlen;
788 		memmove(mtod(m, caddr_t), (caddr_t)ip, sizeof(struct ip));
789 	}
790 	ip = mtod(m, struct ip *);
791 	memcpy(ip + 1, p->ipopt_list, optlen);
792 	*phlen = sizeof(struct ip) + optlen;
793 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
794 	return (m);
795 }
796 
797 /*
798  * Copy options from ip to jp,
799  * omitting those not copied during fragmentation.
800  */
801 int
802 ip_optcopy(struct ip *ip, struct ip *jp)
803 {
804 	u_char *cp, *dp;
805 	int opt, optlen, cnt;
806 
807 	cp = (u_char *)(ip + 1);
808 	dp = (u_char *)(jp + 1);
809 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
810 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
811 		opt = cp[0];
812 		if (opt == IPOPT_EOL)
813 			break;
814 		if (opt == IPOPT_NOP) {
815 			/* Preserve for IP mcast tunnel's LSRR alignment. */
816 			*dp++ = IPOPT_NOP;
817 			optlen = 1;
818 			continue;
819 		}
820 #ifdef DIAGNOSTIC
821 		if (cnt < IPOPT_OLEN + sizeof(*cp))
822 			panic("malformed IPv4 option passed to ip_optcopy");
823 #endif
824 		optlen = cp[IPOPT_OLEN];
825 #ifdef DIAGNOSTIC
826 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
827 			panic("malformed IPv4 option passed to ip_optcopy");
828 #endif
829 		/* bogus lengths should have been caught by ip_dooptions */
830 		if (optlen > cnt)
831 			optlen = cnt;
832 		if (IPOPT_COPIED(opt)) {
833 			memcpy(dp, cp, optlen);
834 			dp += optlen;
835 		}
836 	}
837 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
838 		*dp++ = IPOPT_EOL;
839 	return (optlen);
840 }
841 
842 /*
843  * IP socket option processing.
844  */
845 int
846 ip_ctloutput(int op, struct socket *so, int level, int optname,
847     struct mbuf **mp)
848 {
849 	struct inpcb *inp = sotoinpcb(so);
850 	struct mbuf *m = *mp;
851 	int optval = 0;
852 	struct proc *p = curproc; /* XXX */
853 	int error = 0;
854 	u_int rtid = 0;
855 
856 	if (level != IPPROTO_IP) {
857 		error = EINVAL;
858 		if (op == PRCO_SETOPT)
859 			(void) m_free(*mp);
860 	} else switch (op) {
861 	case PRCO_SETOPT:
862 		switch (optname) {
863 		case IP_OPTIONS:
864 			return (ip_pcbopts(&inp->inp_options, m));
865 
866 		case IP_TOS:
867 		case IP_TTL:
868 		case IP_MINTTL:
869 		case IP_RECVOPTS:
870 		case IP_RECVRETOPTS:
871 		case IP_RECVDSTADDR:
872 		case IP_RECVIF:
873 		case IP_RECVTTL:
874 		case IP_RECVDSTPORT:
875 		case IP_RECVRTABLE:
876 		case IP_IPSECFLOWINFO:
877 			if (m == NULL || m->m_len != sizeof(int))
878 				error = EINVAL;
879 			else {
880 				optval = *mtod(m, int *);
881 				switch (optname) {
882 
883 				case IP_TOS:
884 					inp->inp_ip.ip_tos = optval;
885 					break;
886 
887 				case IP_TTL:
888 					if (optval > 0 && optval <= MAXTTL)
889 						inp->inp_ip.ip_ttl = optval;
890 					else if (optval == -1)
891 						inp->inp_ip.ip_ttl = ip_defttl;
892 					else
893 						error = EINVAL;
894 					break;
895 
896 				case IP_MINTTL:
897 					if (optval >= 0 && optval <= MAXTTL)
898 						inp->inp_ip_minttl = optval;
899 					else
900 						error = EINVAL;
901 					break;
902 #define	OPTSET(bit) \
903 	if (optval) \
904 		inp->inp_flags |= bit; \
905 	else \
906 		inp->inp_flags &= ~bit;
907 
908 				case IP_RECVOPTS:
909 					OPTSET(INP_RECVOPTS);
910 					break;
911 
912 				case IP_RECVRETOPTS:
913 					OPTSET(INP_RECVRETOPTS);
914 					break;
915 
916 				case IP_RECVDSTADDR:
917 					OPTSET(INP_RECVDSTADDR);
918 					break;
919 				case IP_RECVIF:
920 					OPTSET(INP_RECVIF);
921 					break;
922 				case IP_RECVTTL:
923 					OPTSET(INP_RECVTTL);
924 					break;
925 				case IP_RECVDSTPORT:
926 					OPTSET(INP_RECVDSTPORT);
927 					break;
928 				case IP_RECVRTABLE:
929 					OPTSET(INP_RECVRTABLE);
930 					break;
931 				case IP_IPSECFLOWINFO:
932 					OPTSET(INP_IPSECFLOWINFO);
933 					break;
934 				}
935 			}
936 			break;
937 #undef OPTSET
938 
939 		case IP_MULTICAST_IF:
940 		case IP_MULTICAST_TTL:
941 		case IP_MULTICAST_LOOP:
942 		case IP_ADD_MEMBERSHIP:
943 		case IP_DROP_MEMBERSHIP:
944 			error = ip_setmoptions(optname, &inp->inp_moptions, m,
945 			    inp->inp_rtableid);
946 			break;
947 
948 		case IP_PORTRANGE:
949 			if (m == NULL || m->m_len != sizeof(int))
950 				error = EINVAL;
951 			else {
952 				optval = *mtod(m, int *);
953 
954 				switch (optval) {
955 
956 				case IP_PORTRANGE_DEFAULT:
957 					inp->inp_flags &= ~(INP_LOWPORT);
958 					inp->inp_flags &= ~(INP_HIGHPORT);
959 					break;
960 
961 				case IP_PORTRANGE_HIGH:
962 					inp->inp_flags &= ~(INP_LOWPORT);
963 					inp->inp_flags |= INP_HIGHPORT;
964 					break;
965 
966 				case IP_PORTRANGE_LOW:
967 					inp->inp_flags &= ~(INP_HIGHPORT);
968 					inp->inp_flags |= INP_LOWPORT;
969 					break;
970 
971 				default:
972 
973 					error = EINVAL;
974 					break;
975 				}
976 			}
977 			break;
978 		case IP_AUTH_LEVEL:
979 		case IP_ESP_TRANS_LEVEL:
980 		case IP_ESP_NETWORK_LEVEL:
981 		case IP_IPCOMP_LEVEL:
982 #ifndef IPSEC
983 			error = EOPNOTSUPP;
984 #else
985 			if (m == NULL || m->m_len != sizeof(int)) {
986 				error = EINVAL;
987 				break;
988 			}
989 			optval = *mtod(m, int *);
990 
991 			if (optval < IPSEC_LEVEL_BYPASS ||
992 			    optval > IPSEC_LEVEL_UNIQUE) {
993 				error = EINVAL;
994 				break;
995 			}
996 
997 			switch (optname) {
998 			case IP_AUTH_LEVEL:
999 				if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1000 				    suser(p, 0)) {
1001 					error = EACCES;
1002 					break;
1003 				}
1004 				inp->inp_seclevel[SL_AUTH] = optval;
1005 				break;
1006 
1007 			case IP_ESP_TRANS_LEVEL:
1008 				if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1009 				    suser(p, 0)) {
1010 					error = EACCES;
1011 					break;
1012 				}
1013 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1014 				break;
1015 
1016 			case IP_ESP_NETWORK_LEVEL:
1017 				if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1018 				    suser(p, 0)) {
1019 					error = EACCES;
1020 					break;
1021 				}
1022 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1023 				break;
1024 			case IP_IPCOMP_LEVEL:
1025 				if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1026 				    suser(p, 0)) {
1027 					error = EACCES;
1028 					break;
1029 				}
1030 				inp->inp_seclevel[SL_IPCOMP] = optval;
1031 				break;
1032 			}
1033 #endif
1034 			break;
1035 
1036 		case IP_IPSEC_LOCAL_ID:
1037 		case IP_IPSEC_REMOTE_ID:
1038 			error = EOPNOTSUPP;
1039 			break;
1040 		case SO_RTABLE:
1041 			if (m == NULL || m->m_len < sizeof(u_int)) {
1042 				error = EINVAL;
1043 				break;
1044 			}
1045 			rtid = *mtod(m, u_int *);
1046 			if (inp->inp_rtableid == rtid)
1047 				break;
1048 			/* needs privileges to switch when already set */
1049 			if (p->p_p->ps_rtableid != rtid &&
1050 			    p->p_p->ps_rtableid != 0 &&
1051 			    (error = suser(p, 0)) != 0)
1052 				break;
1053 			/* table must exist */
1054 			if (!rtable_exists(rtid)) {
1055 				error = EINVAL;
1056 				break;
1057 			}
1058 			if (inp->inp_lport) {
1059 				error = EBUSY;
1060 				break;
1061 			}
1062 			inp->inp_rtableid = rtid;
1063 			in_pcbrehash(inp);
1064 			break;
1065 		case IP_PIPEX:
1066 			if (m != NULL && m->m_len == sizeof(int))
1067 				inp->inp_pipex = *mtod(m, int *);
1068 			else
1069 				error = EINVAL;
1070 			break;
1071 
1072 		default:
1073 			error = ENOPROTOOPT;
1074 			break;
1075 		}
1076 		if (m)
1077 			(void)m_free(m);
1078 		break;
1079 
1080 	case PRCO_GETOPT:
1081 		switch (optname) {
1082 		case IP_OPTIONS:
1083 		case IP_RETOPTS:
1084 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1085 			if (inp->inp_options) {
1086 				m->m_len = inp->inp_options->m_len;
1087 				memcpy(mtod(m, caddr_t),
1088 				    mtod(inp->inp_options, caddr_t), m->m_len);
1089 			} else
1090 				m->m_len = 0;
1091 			break;
1092 
1093 		case IP_TOS:
1094 		case IP_TTL:
1095 		case IP_MINTTL:
1096 		case IP_RECVOPTS:
1097 		case IP_RECVRETOPTS:
1098 		case IP_RECVDSTADDR:
1099 		case IP_RECVIF:
1100 		case IP_RECVTTL:
1101 		case IP_RECVDSTPORT:
1102 		case IP_RECVRTABLE:
1103 		case IP_IPSECFLOWINFO:
1104 		case IP_IPDEFTTL:
1105 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1106 			m->m_len = sizeof(int);
1107 			switch (optname) {
1108 
1109 			case IP_TOS:
1110 				optval = inp->inp_ip.ip_tos;
1111 				break;
1112 
1113 			case IP_TTL:
1114 				optval = inp->inp_ip.ip_ttl;
1115 				break;
1116 
1117 			case IP_MINTTL:
1118 				optval = inp->inp_ip_minttl;
1119 				break;
1120 
1121 			case IP_IPDEFTTL:
1122 				optval = ip_defttl;
1123 				break;
1124 
1125 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1126 
1127 			case IP_RECVOPTS:
1128 				optval = OPTBIT(INP_RECVOPTS);
1129 				break;
1130 
1131 			case IP_RECVRETOPTS:
1132 				optval = OPTBIT(INP_RECVRETOPTS);
1133 				break;
1134 
1135 			case IP_RECVDSTADDR:
1136 				optval = OPTBIT(INP_RECVDSTADDR);
1137 				break;
1138 			case IP_RECVIF:
1139 				optval = OPTBIT(INP_RECVIF);
1140 				break;
1141 			case IP_RECVTTL:
1142 				optval = OPTBIT(INP_RECVTTL);
1143 				break;
1144 			case IP_RECVDSTPORT:
1145 				optval = OPTBIT(INP_RECVDSTPORT);
1146 				break;
1147 			case IP_RECVRTABLE:
1148 				optval = OPTBIT(INP_RECVRTABLE);
1149 				break;
1150 			case IP_IPSECFLOWINFO:
1151 				optval = OPTBIT(INP_IPSECFLOWINFO);
1152 				break;
1153 			}
1154 			*mtod(m, int *) = optval;
1155 			break;
1156 
1157 		case IP_MULTICAST_IF:
1158 		case IP_MULTICAST_TTL:
1159 		case IP_MULTICAST_LOOP:
1160 		case IP_ADD_MEMBERSHIP:
1161 		case IP_DROP_MEMBERSHIP:
1162 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
1163 			break;
1164 
1165 		case IP_PORTRANGE:
1166 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1167 			m->m_len = sizeof(int);
1168 
1169 			if (inp->inp_flags & INP_HIGHPORT)
1170 				optval = IP_PORTRANGE_HIGH;
1171 			else if (inp->inp_flags & INP_LOWPORT)
1172 				optval = IP_PORTRANGE_LOW;
1173 			else
1174 				optval = 0;
1175 
1176 			*mtod(m, int *) = optval;
1177 			break;
1178 
1179 		case IP_AUTH_LEVEL:
1180 		case IP_ESP_TRANS_LEVEL:
1181 		case IP_ESP_NETWORK_LEVEL:
1182 		case IP_IPCOMP_LEVEL:
1183 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1184 #ifndef IPSEC
1185 			m->m_len = sizeof(int);
1186 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1187 #else
1188 			m->m_len = sizeof(int);
1189 			switch (optname) {
1190 			case IP_AUTH_LEVEL:
1191 				optval = inp->inp_seclevel[SL_AUTH];
1192 				break;
1193 
1194 			case IP_ESP_TRANS_LEVEL:
1195 				optval = inp->inp_seclevel[SL_ESP_TRANS];
1196 				break;
1197 
1198 			case IP_ESP_NETWORK_LEVEL:
1199 				optval = inp->inp_seclevel[SL_ESP_NETWORK];
1200 				break;
1201 			case IP_IPCOMP_LEVEL:
1202 				optval = inp->inp_seclevel[SL_IPCOMP];
1203 				break;
1204 			}
1205 			*mtod(m, int *) = optval;
1206 #endif
1207 			break;
1208 		case IP_IPSEC_LOCAL_ID:
1209 		case IP_IPSEC_REMOTE_ID:
1210 			error = EOPNOTSUPP;
1211 			break;
1212 		case SO_RTABLE:
1213 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1214 			m->m_len = sizeof(u_int);
1215 			*mtod(m, u_int *) = inp->inp_rtableid;
1216 			break;
1217 		case IP_PIPEX:
1218 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1219 			m->m_len = sizeof(int);
1220 			*mtod(m, int *) = inp->inp_pipex;
1221 			break;
1222 		default:
1223 			error = ENOPROTOOPT;
1224 			break;
1225 		}
1226 		break;
1227 	}
1228 	return (error);
1229 }
1230 
1231 /*
1232  * Set up IP options in pcb for insertion in output packets.
1233  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1234  * with destination address if source routed.
1235  */
1236 int
1237 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1238 {
1239 	int cnt, optlen;
1240 	u_char *cp;
1241 	u_char opt;
1242 
1243 	/* turn off any old options */
1244 	if (*pcbopt)
1245 		(void)m_free(*pcbopt);
1246 	*pcbopt = 0;
1247 	if (m == NULL || m->m_len == 0) {
1248 		/*
1249 		 * Only turning off any previous options.
1250 		 */
1251 		if (m)
1252 			(void)m_free(m);
1253 		return (0);
1254 	}
1255 
1256 	if (m->m_len % sizeof(int32_t))
1257 		goto bad;
1258 
1259 	/*
1260 	 * IP first-hop destination address will be stored before
1261 	 * actual options; move other options back
1262 	 * and clear it when none present.
1263 	 */
1264 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1265 		goto bad;
1266 	cnt = m->m_len;
1267 	m->m_len += sizeof(struct in_addr);
1268 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
1269 	memmove((caddr_t)cp, mtod(m, caddr_t), (unsigned)cnt);
1270 	memset(mtod(m, caddr_t), 0, sizeof(struct in_addr));
1271 
1272 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1273 		opt = cp[IPOPT_OPTVAL];
1274 		if (opt == IPOPT_EOL)
1275 			break;
1276 		if (opt == IPOPT_NOP)
1277 			optlen = 1;
1278 		else {
1279 			if (cnt < IPOPT_OLEN + sizeof(*cp))
1280 				goto bad;
1281 			optlen = cp[IPOPT_OLEN];
1282 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
1283 				goto bad;
1284 		}
1285 		switch (opt) {
1286 
1287 		default:
1288 			break;
1289 
1290 		case IPOPT_LSRR:
1291 		case IPOPT_SSRR:
1292 			/*
1293 			 * user process specifies route as:
1294 			 *	->A->B->C->D
1295 			 * D must be our final destination (but we can't
1296 			 * check that since we may not have connected yet).
1297 			 * A is first hop destination, which doesn't appear in
1298 			 * actual IP option, but is stored before the options.
1299 			 */
1300 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1301 				goto bad;
1302 			m->m_len -= sizeof(struct in_addr);
1303 			cnt -= sizeof(struct in_addr);
1304 			optlen -= sizeof(struct in_addr);
1305 			cp[IPOPT_OLEN] = optlen;
1306 			/*
1307 			 * Move first hop before start of options.
1308 			 */
1309 			memcpy(mtod(m, caddr_t), &cp[IPOPT_OFFSET+1],
1310 			    sizeof(struct in_addr));
1311 			/*
1312 			 * Then copy rest of options back
1313 			 * to close up the deleted entry.
1314 			 */
1315 			memmove((caddr_t)&cp[IPOPT_OFFSET+1],
1316 			    (caddr_t)(&cp[IPOPT_OFFSET+1] +
1317 			    sizeof(struct in_addr)),
1318 			    (unsigned)cnt - (IPOPT_OFFSET+1));
1319 			break;
1320 		}
1321 	}
1322 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1323 		goto bad;
1324 	*pcbopt = m;
1325 	return (0);
1326 
1327 bad:
1328 	(void)m_free(m);
1329 	return (EINVAL);
1330 }
1331 
1332 /*
1333  * Set the IP multicast options in response to user setsockopt().
1334  */
1335 int
1336 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m,
1337     u_int rtableid)
1338 {
1339 	struct in_addr addr;
1340 	struct in_ifaddr *ia;
1341 	struct ip_mreq *mreq;
1342 	struct ifnet *ifp = NULL;
1343 	struct ip_moptions *imo = *imop;
1344 	struct in_multi **immp;
1345 	struct rtentry *rt;
1346 	struct sockaddr_in sin;
1347 	int i, error = 0;
1348 	u_char loop;
1349 
1350 	if (imo == NULL) {
1351 		/*
1352 		 * No multicast option buffer attached to the pcb;
1353 		 * allocate one and initialize to default values.
1354 		 */
1355 		imo = malloc(sizeof(*imo), M_IPMOPTS, M_WAITOK|M_ZERO);
1356 		immp = (struct in_multi **)malloc(
1357 		    (sizeof(*immp) * IP_MIN_MEMBERSHIPS), M_IPMOPTS,
1358 		    M_WAITOK|M_ZERO);
1359 		*imop = imo;
1360 		imo->imo_ifidx = 0;
1361 		imo->imo_ttl = IP_DEFAULT_MULTICAST_TTL;
1362 		imo->imo_loop = IP_DEFAULT_MULTICAST_LOOP;
1363 		imo->imo_num_memberships = 0;
1364 		imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1365 		imo->imo_membership = immp;
1366 	}
1367 
1368 	switch (optname) {
1369 
1370 	case IP_MULTICAST_IF:
1371 		/*
1372 		 * Select the interface for outgoing multicast packets.
1373 		 */
1374 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1375 			error = EINVAL;
1376 			break;
1377 		}
1378 		addr = *(mtod(m, struct in_addr *));
1379 		/*
1380 		 * INADDR_ANY is used to remove a previous selection.
1381 		 * When no interface is selected, a default one is
1382 		 * chosen every time a multicast packet is sent.
1383 		 */
1384 		if (addr.s_addr == INADDR_ANY) {
1385 			imo->imo_ifidx = 0;
1386 			break;
1387 		}
1388 		/*
1389 		 * The selected interface is identified by its local
1390 		 * IP address.  Find the interface and confirm that
1391 		 * it supports multicasting.
1392 		 */
1393 		memset(&sin, 0, sizeof(sin));
1394 		sin.sin_len = sizeof(sin);
1395 		sin.sin_family = AF_INET;
1396 		sin.sin_addr = addr;
1397 		ia = ifatoia(ifa_ifwithaddr(sintosa(&sin), rtableid));
1398 		if (ia == NULL ||
1399 		    (ia->ia_ifp->if_flags & IFF_MULTICAST) == 0) {
1400 			error = EADDRNOTAVAIL;
1401 			break;
1402 		}
1403 		imo->imo_ifidx = ia->ia_ifp->if_index;
1404 		break;
1405 
1406 	case IP_MULTICAST_TTL:
1407 		/*
1408 		 * Set the IP time-to-live for outgoing multicast packets.
1409 		 */
1410 		if (m == NULL || m->m_len != 1) {
1411 			error = EINVAL;
1412 			break;
1413 		}
1414 		imo->imo_ttl = *(mtod(m, u_char *));
1415 		break;
1416 
1417 	case IP_MULTICAST_LOOP:
1418 		/*
1419 		 * Set the loopback flag for outgoing multicast packets.
1420 		 * Must be zero or one.
1421 		 */
1422 		if (m == NULL || m->m_len != 1 ||
1423 		   (loop = *(mtod(m, u_char *))) > 1) {
1424 			error = EINVAL;
1425 			break;
1426 		}
1427 		imo->imo_loop = loop;
1428 		break;
1429 
1430 	case IP_ADD_MEMBERSHIP:
1431 		/*
1432 		 * Add a multicast group membership.
1433 		 * Group must be a valid IP multicast address.
1434 		 */
1435 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1436 			error = EINVAL;
1437 			break;
1438 		}
1439 		mreq = mtod(m, struct ip_mreq *);
1440 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1441 			error = EINVAL;
1442 			break;
1443 		}
1444 		/*
1445 		 * If no interface address was provided, use the interface of
1446 		 * the route to the given multicast address.
1447 		 */
1448 		if (mreq->imr_interface.s_addr == INADDR_ANY) {
1449 			memset(&sin, 0, sizeof(sin));
1450 			sin.sin_len = sizeof(sin);
1451 			sin.sin_family = AF_INET;
1452 			sin.sin_addr = mreq->imr_multiaddr;
1453 			rt = rtalloc(sintosa(&sin), RT_RESOLVE, rtableid);
1454 			if (!rtisvalid(rt)) {
1455 				rtfree(rt);
1456 				error = EADDRNOTAVAIL;
1457 				break;
1458 			}
1459 		} else {
1460 			memset(&sin, 0, sizeof(sin));
1461 			sin.sin_len = sizeof(sin);
1462 			sin.sin_family = AF_INET;
1463 			sin.sin_addr = mreq->imr_interface;
1464 			rt = rtalloc(sintosa(&sin), 0, rtableid);
1465 			if (!rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_LOCAL)) {
1466 				rtfree(rt);
1467 				error = EADDRNOTAVAIL;
1468 				break;
1469 			}
1470 		}
1471 		ifp = if_get(rt->rt_ifidx);
1472 		rtfree(rt);
1473 
1474 		/*
1475 		 * See if we found an interface, and confirm that it
1476 		 * supports multicast.
1477 		 */
1478 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1479 			error = EADDRNOTAVAIL;
1480 			if_put(ifp);
1481 			break;
1482 		}
1483 		/*
1484 		 * See if the membership already exists or if all the
1485 		 * membership slots are full.
1486 		 */
1487 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1488 			if (imo->imo_membership[i]->inm_ifidx
1489 						== ifp->if_index &&
1490 			    imo->imo_membership[i]->inm_addr.s_addr
1491 						== mreq->imr_multiaddr.s_addr)
1492 				break;
1493 		}
1494 		if (i < imo->imo_num_memberships) {
1495 			error = EADDRINUSE;
1496 			if_put(ifp);
1497 			break;
1498 		}
1499 		if (imo->imo_num_memberships == imo->imo_max_memberships) {
1500 			struct in_multi **nmships, **omships;
1501 			size_t newmax;
1502 			/*
1503 			 * Resize the vector to next power-of-two minus 1. If the
1504 			 * size would exceed the maximum then we know we've really
1505 			 * run out of entries. Otherwise, we reallocate the vector.
1506 			 */
1507 			nmships = NULL;
1508 			omships = imo->imo_membership;
1509 			newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1510 			if (newmax <= IP_MAX_MEMBERSHIPS) {
1511 				nmships = (struct in_multi **)malloc(
1512 				    sizeof(*nmships) * newmax, M_IPMOPTS,
1513 				    M_NOWAIT|M_ZERO);
1514 				if (nmships != NULL) {
1515 					memcpy(nmships, omships,
1516 					    sizeof(*omships) *
1517 					    imo->imo_max_memberships);
1518 					free(omships, M_IPMOPTS,
1519 					    sizeof(*omships) *
1520 					    imo->imo_max_memberships);
1521 					imo->imo_membership = nmships;
1522 					imo->imo_max_memberships = newmax;
1523 				}
1524 			}
1525 			if (nmships == NULL) {
1526 				error = ENOBUFS;
1527 				if_put(ifp);
1528 				break;
1529 			}
1530 		}
1531 		/*
1532 		 * Everything looks good; add a new record to the multicast
1533 		 * address list for the given interface.
1534 		 */
1535 		if ((imo->imo_membership[i] =
1536 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1537 			error = ENOBUFS;
1538 			if_put(ifp);
1539 			break;
1540 		}
1541 		++imo->imo_num_memberships;
1542 		if_put(ifp);
1543 		break;
1544 
1545 	case IP_DROP_MEMBERSHIP:
1546 		/*
1547 		 * Drop a multicast group membership.
1548 		 * Group must be a valid IP multicast address.
1549 		 */
1550 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1551 			error = EINVAL;
1552 			break;
1553 		}
1554 		mreq = mtod(m, struct ip_mreq *);
1555 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1556 			error = EINVAL;
1557 			break;
1558 		}
1559 		/*
1560 		 * If an interface address was specified, get a pointer
1561 		 * to its ifnet structure.
1562 		 */
1563 		if (mreq->imr_interface.s_addr == INADDR_ANY)
1564 			ifp = NULL;
1565 		else {
1566 			memset(&sin, 0, sizeof(sin));
1567 			sin.sin_len = sizeof(sin);
1568 			sin.sin_family = AF_INET;
1569 			sin.sin_addr = mreq->imr_interface;
1570 			ia = ifatoia(ifa_ifwithaddr(sintosa(&sin), rtableid));
1571 			if (ia == NULL) {
1572 				error = EADDRNOTAVAIL;
1573 				break;
1574 			}
1575 			ifp = ia->ia_ifp;
1576 		}
1577 		/*
1578 		 * Find the membership in the membership array.
1579 		 */
1580 		for (i = 0; i < imo->imo_num_memberships; ++i) {
1581 			if ((ifp == NULL ||
1582 			    imo->imo_membership[i]->inm_ifidx ==
1583 			        ifp->if_index) &&
1584 			     imo->imo_membership[i]->inm_addr.s_addr ==
1585 			     mreq->imr_multiaddr.s_addr)
1586 				break;
1587 		}
1588 		if (i == imo->imo_num_memberships) {
1589 			error = EADDRNOTAVAIL;
1590 			break;
1591 		}
1592 		/*
1593 		 * Give up the multicast address record to which the
1594 		 * membership points.
1595 		 */
1596 		in_delmulti(imo->imo_membership[i]);
1597 		/*
1598 		 * Remove the gap in the membership array.
1599 		 */
1600 		for (++i; i < imo->imo_num_memberships; ++i)
1601 			imo->imo_membership[i-1] = imo->imo_membership[i];
1602 		--imo->imo_num_memberships;
1603 		break;
1604 
1605 	default:
1606 		error = EOPNOTSUPP;
1607 		break;
1608 	}
1609 
1610 	/*
1611 	 * If all options have default values, no need to keep the data.
1612 	 */
1613 	if (imo->imo_ifidx == 0 &&
1614 	    imo->imo_ttl == IP_DEFAULT_MULTICAST_TTL &&
1615 	    imo->imo_loop == IP_DEFAULT_MULTICAST_LOOP &&
1616 	    imo->imo_num_memberships == 0) {
1617 		free(imo->imo_membership , M_IPMOPTS, 0);
1618 		free(*imop, M_IPMOPTS, sizeof(**imop));
1619 		*imop = NULL;
1620 	}
1621 
1622 	return (error);
1623 }
1624 
1625 /*
1626  * Return the IP multicast options in response to user getsockopt().
1627  */
1628 int
1629 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
1630 {
1631 	u_char *ttl;
1632 	u_char *loop;
1633 	struct in_addr *addr;
1634 	struct in_ifaddr *ia;
1635 	struct ifnet *ifp;
1636 
1637 	*mp = m_get(M_WAIT, MT_SOOPTS);
1638 
1639 	switch (optname) {
1640 
1641 	case IP_MULTICAST_IF:
1642 		addr = mtod(*mp, struct in_addr *);
1643 		(*mp)->m_len = sizeof(struct in_addr);
1644 		if (imo == NULL || (ifp = if_get(imo->imo_ifidx)) == NULL)
1645 			addr->s_addr = INADDR_ANY;
1646 		else {
1647 			IFP_TO_IA(ifp, ia);
1648 			if_put(ifp);
1649 			addr->s_addr = (ia == NULL) ? INADDR_ANY
1650 					: ia->ia_addr.sin_addr.s_addr;
1651 		}
1652 		return (0);
1653 
1654 	case IP_MULTICAST_TTL:
1655 		ttl = mtod(*mp, u_char *);
1656 		(*mp)->m_len = 1;
1657 		*ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1658 				     : imo->imo_ttl;
1659 		return (0);
1660 
1661 	case IP_MULTICAST_LOOP:
1662 		loop = mtod(*mp, u_char *);
1663 		(*mp)->m_len = 1;
1664 		*loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
1665 				      : imo->imo_loop;
1666 		return (0);
1667 
1668 	default:
1669 		return (EOPNOTSUPP);
1670 	}
1671 }
1672 
1673 /*
1674  * Discard the IP multicast options.
1675  */
1676 void
1677 ip_freemoptions(struct ip_moptions *imo)
1678 {
1679 	int i;
1680 
1681 	if (imo != NULL) {
1682 		for (i = 0; i < imo->imo_num_memberships; ++i)
1683 			in_delmulti(imo->imo_membership[i]);
1684 		free(imo->imo_membership, M_IPMOPTS, 0);
1685 		free(imo, M_IPMOPTS, sizeof(*imo));
1686 	}
1687 }
1688 
1689 /*
1690  * Routine called from ip_output() to loop back a copy of an IP multicast
1691  * packet to the input queue of a specified interface.
1692  */
1693 void
1694 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1695 {
1696 	struct ip *ip;
1697 	struct mbuf *copym;
1698 
1699 	copym = m_copym2(m, 0, M_COPYALL, M_DONTWAIT);
1700 	if (copym != NULL) {
1701 		/*
1702 		 * We don't bother to fragment if the IP length is greater
1703 		 * than the interface's MTU.  Can this possibly matter?
1704 		 */
1705 		ip = mtod(copym, struct ip *);
1706 		ip->ip_sum = 0;
1707 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1708 		if_input_local(ifp, copym, dst->sin_family);
1709 	}
1710 }
1711 
1712 /*
1713  *	Compute significant parts of the IPv4 checksum pseudo-header
1714  *	for use in a delayed TCP/UDP checksum calculation.
1715  */
1716 static __inline u_int16_t __attribute__((__unused__))
1717 in_cksum_phdr(u_int32_t src, u_int32_t dst, u_int32_t lenproto)
1718 {
1719 	u_int32_t sum;
1720 
1721 	sum = lenproto +
1722 	      (u_int16_t)(src >> 16) +
1723 	      (u_int16_t)(src /*& 0xffff*/) +
1724 	      (u_int16_t)(dst >> 16) +
1725 	      (u_int16_t)(dst /*& 0xffff*/);
1726 
1727 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
1728 
1729 	if (sum > 0xffff)
1730 		sum -= 0xffff;
1731 
1732 	return (sum);
1733 }
1734 
1735 /*
1736  * Process a delayed payload checksum calculation.
1737  */
1738 void
1739 in_delayed_cksum(struct mbuf *m)
1740 {
1741 	struct ip *ip;
1742 	u_int16_t csum, offset;
1743 
1744 	ip = mtod(m, struct ip *);
1745 	offset = ip->ip_hl << 2;
1746 	csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1747 	if (csum == 0 && ip->ip_p == IPPROTO_UDP)
1748 		csum = 0xffff;
1749 
1750 	switch (ip->ip_p) {
1751 	case IPPROTO_TCP:
1752 		offset += offsetof(struct tcphdr, th_sum);
1753 		break;
1754 
1755 	case IPPROTO_UDP:
1756 		offset += offsetof(struct udphdr, uh_sum);
1757 		break;
1758 
1759 	case IPPROTO_ICMP:
1760 		offset += offsetof(struct icmp, icmp_cksum);
1761 		break;
1762 
1763 	default:
1764 		return;
1765 	}
1766 
1767 	if ((offset + sizeof(u_int16_t)) > m->m_len)
1768 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1769 	else
1770 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1771 }
1772 
1773 void
1774 in_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
1775 {
1776 	struct ip *ip = mtod(m, struct ip *);
1777 
1778 	/* some hw and in_delayed_cksum need the pseudo header cksum */
1779 	if (m->m_pkthdr.csum_flags &
1780 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
1781 		u_int16_t csum = 0, offset;
1782 
1783 		offset = ip->ip_hl << 2;
1784 		if (m->m_pkthdr.csum_flags & (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT))
1785 			csum = in_cksum_phdr(ip->ip_src.s_addr,
1786 			    ip->ip_dst.s_addr, htonl(ntohs(ip->ip_len) -
1787 			    offset + ip->ip_p));
1788 		if (ip->ip_p == IPPROTO_TCP)
1789 			offset += offsetof(struct tcphdr, th_sum);
1790 		else if (ip->ip_p == IPPROTO_UDP)
1791 			offset += offsetof(struct udphdr, uh_sum);
1792 		else if (ip->ip_p == IPPROTO_ICMP)
1793 			offset += offsetof(struct icmp, icmp_cksum);
1794 		if ((offset + sizeof(u_int16_t)) > m->m_len)
1795 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1796 		else
1797 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1798 	}
1799 
1800 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
1801 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv4) ||
1802 		    ip->ip_hl != 5 || ifp->if_bridgeport != NULL) {
1803 			tcpstat.tcps_outswcsum++;
1804 			in_delayed_cksum(m);
1805 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
1806 		}
1807 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
1808 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv4) ||
1809 		    ip->ip_hl != 5 || ifp->if_bridgeport != NULL) {
1810 			udpstat.udps_outswcsum++;
1811 			in_delayed_cksum(m);
1812 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
1813 		}
1814 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
1815 		in_delayed_cksum(m);
1816 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
1817 	}
1818 }
1819