xref: /freebsd/sys/netinet/ip_output.c (revision 298cf604ccf133b101c6fad42d1a078a1fac58ca)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_ipfw.h"
36 #include "opt_ipsec.h"
37 #include "opt_route.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_sctp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/ucred.h>
54 
55 #include <net/if.h>
56 #include <net/if_llatbl.h>
57 #include <net/netisr.h>
58 #include <net/pfil.h>
59 #include <net/route.h>
60 #include <net/flowtable.h>
61 #ifdef RADIX_MPATH
62 #include <net/radix_mpath.h>
63 #endif
64 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_options.h>
73 #ifdef SCTP
74 #include <netinet/sctp.h>
75 #include <netinet/sctp_crc32.h>
76 #endif
77 
78 #ifdef IPSEC
79 #include <netinet/ip_ipsec.h>
80 #include <netipsec/ipsec.h>
81 #endif /* IPSEC*/
82 
83 #include <machine/in_cksum.h>
84 
85 #include <security/mac/mac_framework.h>
86 
87 VNET_DEFINE(u_short, ip_id);
88 
89 #ifdef MBUF_STRESS_TEST
90 static int mbuf_frag_size = 0;
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
92 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
93 #endif
94 
95 static void	ip_mloopback
96 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
97 
98 
99 extern int in_mcast_loop;
100 extern	struct protosw inetsw[];
101 
102 /*
103  * IP output.  The packet in mbuf chain m contains a skeletal IP
104  * header (with len, off, ttl, proto, tos, src, dst).
105  * The mbuf chain containing the packet will be freed.
106  * The mbuf opt, if present, will not be freed.
107  * If route ro is present and has ro_rt initialized, route lookup would be
108  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
109  * then result of route lookup is stored in ro->ro_rt.
110  *
111  * In the IP forwarding case, the packet will arrive with options already
112  * inserted, so must have a NULL opt pointer.
113  */
114 int
115 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
116     struct ip_moptions *imo, struct inpcb *inp)
117 {
118 	struct ip *ip;
119 	struct ifnet *ifp = NULL;	/* keep compiler happy */
120 	struct mbuf *m0;
121 	int hlen = sizeof (struct ip);
122 	int mtu;
123 	int n;	/* scratchpad */
124 	int error = 0;
125 	struct sockaddr_in *dst;
126 	struct in_ifaddr *ia;
127 	int isbroadcast;
128 	uint16_t ip_len, ip_off;
129 	struct route iproute;
130 	struct rtentry *rte;	/* cache for ro->ro_rt */
131 	struct in_addr odst;
132 	struct m_tag *fwd_tag = NULL;
133 #ifdef IPSEC
134 	int no_route_but_check_spd = 0;
135 #endif
136 	M_ASSERTPKTHDR(m);
137 
138 	if (inp != NULL) {
139 		INP_LOCK_ASSERT(inp);
140 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
141 		if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
142 			m->m_pkthdr.flowid = inp->inp_flowid;
143 			m->m_flags |= M_FLOWID;
144 		}
145 	}
146 
147 	if (ro == NULL) {
148 		ro = &iproute;
149 		bzero(ro, sizeof (*ro));
150 	}
151 
152 #ifdef FLOWTABLE
153 	if (ro->ro_rt == NULL) {
154 		struct flentry *fle;
155 
156 		/*
157 		 * The flow table returns route entries valid for up to 30
158 		 * seconds; we rely on the remainder of ip_output() taking no
159 		 * longer than that long for the stability of ro_rt. The
160 		 * flow ID assignment must have happened before this point.
161 		 */
162 		fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
163 		if (fle != NULL)
164 			flow_to_route(fle, ro);
165 	}
166 #endif
167 
168 	if (opt) {
169 		int len = 0;
170 		m = ip_insertoptions(m, opt, &len);
171 		if (len != 0)
172 			hlen = len; /* ip->ip_hl is updated above */
173 	}
174 	ip = mtod(m, struct ip *);
175 	ip_len = ntohs(ip->ip_len);
176 	ip_off = ntohs(ip->ip_off);
177 
178 	/*
179 	 * Fill in IP header.  If we are not allowing fragmentation,
180 	 * then the ip_id field is meaningless, but we don't set it
181 	 * to zero.  Doing so causes various problems when devices along
182 	 * the path (routers, load balancers, firewalls, etc.) illegally
183 	 * disable DF on our packet.  Note that a 16-bit counter
184 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
185 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
186 	 * for Counting NATted Hosts", Proc. IMW'02, available at
187 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
188 	 */
189 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
190 		ip->ip_v = IPVERSION;
191 		ip->ip_hl = hlen >> 2;
192 		ip->ip_id = ip_newid();
193 		IPSTAT_INC(ips_localout);
194 	} else {
195 		/* Header already set, fetch hlen from there */
196 		hlen = ip->ip_hl << 2;
197 	}
198 
199 	dst = (struct sockaddr_in *)&ro->ro_dst;
200 again:
201 	ia = NULL;
202 	/*
203 	 * If there is a cached route,
204 	 * check that it is to the same destination
205 	 * and is still up.  If not, free it and try again.
206 	 * The address family should also be checked in case of sharing the
207 	 * cache with IPv6.
208 	 */
209 	rte = ro->ro_rt;
210 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
211 		    rte->rt_ifp == NULL ||
212 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
213 			  dst->sin_family != AF_INET ||
214 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
215 		RO_RTFREE(ro);
216 		ro->ro_lle = NULL;
217 		rte = NULL;
218 	}
219 	if (rte == NULL && fwd_tag == NULL) {
220 		bzero(dst, sizeof(*dst));
221 		dst->sin_family = AF_INET;
222 		dst->sin_len = sizeof(*dst);
223 		dst->sin_addr = ip->ip_dst;
224 	}
225 	/*
226 	 * If routing to interface only, short circuit routing lookup.
227 	 * The use of an all-ones broadcast address implies this; an
228 	 * interface is specified by the broadcast address of an interface,
229 	 * or the destination address of a ptp interface.
230 	 */
231 	if (flags & IP_SENDONES) {
232 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
233 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
234 			IPSTAT_INC(ips_noroute);
235 			error = ENETUNREACH;
236 			goto bad;
237 		}
238 		ip->ip_dst.s_addr = INADDR_BROADCAST;
239 		dst->sin_addr = ip->ip_dst;
240 		ifp = ia->ia_ifp;
241 		ip->ip_ttl = 1;
242 		isbroadcast = 1;
243 	} else if (flags & IP_ROUTETOIF) {
244 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
245 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
246 			IPSTAT_INC(ips_noroute);
247 			error = ENETUNREACH;
248 			goto bad;
249 		}
250 		ifp = ia->ia_ifp;
251 		ip->ip_ttl = 1;
252 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
253 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
254 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
255 		/*
256 		 * Bypass the normal routing lookup for multicast
257 		 * packets if the interface is specified.
258 		 */
259 		ifp = imo->imo_multicast_ifp;
260 		IFP_TO_IA(ifp, ia);
261 		isbroadcast = 0;	/* fool gcc */
262 	} else {
263 		/*
264 		 * We want to do any cloning requested by the link layer,
265 		 * as this is probably required in all cases for correct
266 		 * operation (as it is for ARP).
267 		 */
268 		if (rte == NULL) {
269 #ifdef RADIX_MPATH
270 			rtalloc_mpath_fib(ro,
271 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
272 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
273 #else
274 			in_rtalloc_ign(ro, 0,
275 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
276 #endif
277 			rte = ro->ro_rt;
278 		}
279 		if (rte == NULL ||
280 		    rte->rt_ifp == NULL ||
281 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
282 #ifdef IPSEC
283 			/*
284 			 * There is no route for this packet, but it is
285 			 * possible that a matching SPD entry exists.
286 			 */
287 			no_route_but_check_spd = 1;
288 			mtu = 0; /* Silence GCC warning. */
289 			goto sendit;
290 #endif
291 			IPSTAT_INC(ips_noroute);
292 			error = EHOSTUNREACH;
293 			goto bad;
294 		}
295 		ia = ifatoia(rte->rt_ifa);
296 		ifa_ref(&ia->ia_ifa);
297 		ifp = rte->rt_ifp;
298 		rte->rt_rmx.rmx_pksent++;
299 		if (rte->rt_flags & RTF_GATEWAY)
300 			dst = (struct sockaddr_in *)rte->rt_gateway;
301 		if (rte->rt_flags & RTF_HOST)
302 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
303 		else
304 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
305 	}
306 	/*
307 	 * Calculate MTU.  If we have a route that is up, use that,
308 	 * otherwise use the interface's MTU.
309 	 */
310 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
311 		/*
312 		 * This case can happen if the user changed the MTU
313 		 * of an interface after enabling IP on it.  Because
314 		 * most netifs don't keep track of routes pointing to
315 		 * them, there is no way for one to update all its
316 		 * routes when the MTU is changed.
317 		 */
318 		if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
319 			rte->rt_rmx.rmx_mtu = ifp->if_mtu;
320 		mtu = rte->rt_rmx.rmx_mtu;
321 	} else {
322 		mtu = ifp->if_mtu;
323 	}
324 	/* Catch a possible divide by zero later. */
325 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
326 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
327 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
328 		m->m_flags |= M_MCAST;
329 		/*
330 		 * IP destination address is multicast.  Make sure "dst"
331 		 * still points to the address in "ro".  (It may have been
332 		 * changed to point to a gateway address, above.)
333 		 */
334 		dst = (struct sockaddr_in *)&ro->ro_dst;
335 		/*
336 		 * See if the caller provided any multicast options
337 		 */
338 		if (imo != NULL) {
339 			ip->ip_ttl = imo->imo_multicast_ttl;
340 			if (imo->imo_multicast_vif != -1)
341 				ip->ip_src.s_addr =
342 				    ip_mcast_src ?
343 				    ip_mcast_src(imo->imo_multicast_vif) :
344 				    INADDR_ANY;
345 		} else
346 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
347 		/*
348 		 * Confirm that the outgoing interface supports multicast.
349 		 */
350 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
351 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
352 				IPSTAT_INC(ips_noroute);
353 				error = ENETUNREACH;
354 				goto bad;
355 			}
356 		}
357 		/*
358 		 * If source address not specified yet, use address
359 		 * of outgoing interface.
360 		 */
361 		if (ip->ip_src.s_addr == INADDR_ANY) {
362 			/* Interface may have no addresses. */
363 			if (ia != NULL)
364 				ip->ip_src = IA_SIN(ia)->sin_addr;
365 		}
366 
367 		if ((imo == NULL && in_mcast_loop) ||
368 		    (imo && imo->imo_multicast_loop)) {
369 			/*
370 			 * Loop back multicast datagram if not expressly
371 			 * forbidden to do so, even if we are not a member
372 			 * of the group; ip_input() will filter it later,
373 			 * thus deferring a hash lookup and mutex acquisition
374 			 * at the expense of a cheap copy using m_copym().
375 			 */
376 			ip_mloopback(ifp, m, dst, hlen);
377 		} else {
378 			/*
379 			 * If we are acting as a multicast router, perform
380 			 * multicast forwarding as if the packet had just
381 			 * arrived on the interface to which we are about
382 			 * to send.  The multicast forwarding function
383 			 * recursively calls this function, using the
384 			 * IP_FORWARDING flag to prevent infinite recursion.
385 			 *
386 			 * Multicasts that are looped back by ip_mloopback(),
387 			 * above, will be forwarded by the ip_input() routine,
388 			 * if necessary.
389 			 */
390 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
391 				/*
392 				 * If rsvp daemon is not running, do not
393 				 * set ip_moptions. This ensures that the packet
394 				 * is multicast and not just sent down one link
395 				 * as prescribed by rsvpd.
396 				 */
397 				if (!V_rsvp_on)
398 					imo = NULL;
399 				if (ip_mforward &&
400 				    ip_mforward(ip, ifp, m, imo) != 0) {
401 					m_freem(m);
402 					goto done;
403 				}
404 			}
405 		}
406 
407 		/*
408 		 * Multicasts with a time-to-live of zero may be looped-
409 		 * back, above, but must not be transmitted on a network.
410 		 * Also, multicasts addressed to the loopback interface
411 		 * are not sent -- the above call to ip_mloopback() will
412 		 * loop back a copy. ip_input() will drop the copy if
413 		 * this host does not belong to the destination group on
414 		 * the loopback interface.
415 		 */
416 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
417 			m_freem(m);
418 			goto done;
419 		}
420 
421 		goto sendit;
422 	}
423 
424 	/*
425 	 * If the source address is not specified yet, use the address
426 	 * of the outoing interface.
427 	 */
428 	if (ip->ip_src.s_addr == INADDR_ANY) {
429 		/* Interface may have no addresses. */
430 		if (ia != NULL) {
431 			ip->ip_src = IA_SIN(ia)->sin_addr;
432 		}
433 	}
434 
435 	/*
436 	 * Verify that we have any chance at all of being able to queue the
437 	 * packet or packet fragments, unless ALTQ is enabled on the given
438 	 * interface in which case packetdrop should be done by queueing.
439 	 */
440 	n = ip_len / mtu + 1; /* how many fragments ? */
441 	if (
442 #ifdef ALTQ
443 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
444 #endif /* ALTQ */
445 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
446 		error = ENOBUFS;
447 		IPSTAT_INC(ips_odropped);
448 		ifp->if_snd.ifq_drops += n;
449 		goto bad;
450 	}
451 
452 	/*
453 	 * Look for broadcast address and
454 	 * verify user is allowed to send
455 	 * such a packet.
456 	 */
457 	if (isbroadcast) {
458 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
459 			error = EADDRNOTAVAIL;
460 			goto bad;
461 		}
462 		if ((flags & IP_ALLOWBROADCAST) == 0) {
463 			error = EACCES;
464 			goto bad;
465 		}
466 		/* don't allow broadcast messages to be fragmented */
467 		if (ip_len > mtu) {
468 			error = EMSGSIZE;
469 			goto bad;
470 		}
471 		m->m_flags |= M_BCAST;
472 	} else {
473 		m->m_flags &= ~M_BCAST;
474 	}
475 
476 sendit:
477 #ifdef IPSEC
478 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
479 	case 1:
480 		goto bad;
481 	case -1:
482 		goto done;
483 	case 0:
484 	default:
485 		break;	/* Continue with packet processing. */
486 	}
487 	/*
488 	 * Check if there was a route for this packet; return error if not.
489 	 */
490 	if (no_route_but_check_spd) {
491 		IPSTAT_INC(ips_noroute);
492 		error = EHOSTUNREACH;
493 		goto bad;
494 	}
495 	/* Update variables that are affected by ipsec4_output(). */
496 	ip = mtod(m, struct ip *);
497 	hlen = ip->ip_hl << 2;
498 #endif /* IPSEC */
499 
500 	/* Jump over all PFIL processing if hooks are not active. */
501 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
502 		goto passout;
503 
504 	/* Run through list of hooks for output packets. */
505 	odst.s_addr = ip->ip_dst.s_addr;
506 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
507 	if (error != 0 || m == NULL)
508 		goto done;
509 
510 	ip = mtod(m, struct ip *);
511 
512 	/* See if destination IP address was changed by packet filter. */
513 	if (odst.s_addr != ip->ip_dst.s_addr) {
514 		m->m_flags |= M_SKIP_FIREWALL;
515 		/* If destination is now ourself drop to ip_input(). */
516 		if (in_localip(ip->ip_dst)) {
517 			m->m_flags |= M_FASTFWD_OURS;
518 			if (m->m_pkthdr.rcvif == NULL)
519 				m->m_pkthdr.rcvif = V_loif;
520 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
521 				m->m_pkthdr.csum_flags |=
522 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
523 				m->m_pkthdr.csum_data = 0xffff;
524 			}
525 			m->m_pkthdr.csum_flags |=
526 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
527 #ifdef SCTP
528 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
529 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
530 #endif
531 			error = netisr_queue(NETISR_IP, m);
532 			goto done;
533 		} else {
534 			if (ia != NULL)
535 				ifa_free(&ia->ia_ifa);
536 			goto again;	/* Redo the routing table lookup. */
537 		}
538 	}
539 
540 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
541 	if (m->m_flags & M_FASTFWD_OURS) {
542 		if (m->m_pkthdr.rcvif == NULL)
543 			m->m_pkthdr.rcvif = V_loif;
544 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
545 			m->m_pkthdr.csum_flags |=
546 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
547 			m->m_pkthdr.csum_data = 0xffff;
548 		}
549 #ifdef SCTP
550 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
551 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
552 #endif
553 		m->m_pkthdr.csum_flags |=
554 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
555 
556 		error = netisr_queue(NETISR_IP, m);
557 		goto done;
558 	}
559 	/* Or forward to some other address? */
560 	if ((m->m_flags & M_IP_NEXTHOP) &&
561 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
562 		dst = (struct sockaddr_in *)&ro->ro_dst;
563 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
564 		m->m_flags |= M_SKIP_FIREWALL;
565 		m->m_flags &= ~M_IP_NEXTHOP;
566 		m_tag_delete(m, fwd_tag);
567 		if (ia != NULL)
568 			ifa_free(&ia->ia_ifa);
569 		goto again;
570 	}
571 
572 passout:
573 	/* 127/8 must not appear on wire - RFC1122. */
574 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
575 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
576 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
577 			IPSTAT_INC(ips_badaddr);
578 			error = EADDRNOTAVAIL;
579 			goto bad;
580 		}
581 	}
582 
583 	m->m_pkthdr.csum_flags |= CSUM_IP;
584 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
585 		in_delayed_cksum(m);
586 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
587 	}
588 #ifdef SCTP
589 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
590 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
591 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
592 	}
593 #endif
594 
595 	/*
596 	 * If small enough for interface, or the interface will take
597 	 * care of the fragmentation for us, we can just send directly.
598 	 */
599 	if (ip_len <= mtu ||
600 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
601 	    ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
602 		ip->ip_sum = 0;
603 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
604 			ip->ip_sum = in_cksum(m, hlen);
605 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
606 		}
607 
608 		/*
609 		 * Record statistics for this interface address.
610 		 * With CSUM_TSO the byte/packet count will be slightly
611 		 * incorrect because we count the IP+TCP headers only
612 		 * once instead of for every generated packet.
613 		 */
614 		if (!(flags & IP_FORWARDING) && ia) {
615 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
616 				ia->ia_ifa.if_opackets +=
617 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
618 			else
619 				ia->ia_ifa.if_opackets++;
620 			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
621 		}
622 #ifdef MBUF_STRESS_TEST
623 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
624 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
625 #endif
626 		/*
627 		 * Reset layer specific mbuf flags
628 		 * to avoid confusing lower layers.
629 		 */
630 		m->m_flags &= ~(M_PROTOFLAGS);
631 		error = (*ifp->if_output)(ifp, m,
632 		    		(struct sockaddr *)dst, ro);
633 		goto done;
634 	}
635 
636 	/* Balk when DF bit is set or the interface didn't support TSO. */
637 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
638 		error = EMSGSIZE;
639 		IPSTAT_INC(ips_cantfrag);
640 		goto bad;
641 	}
642 
643 	/*
644 	 * Too large for interface; fragment if possible. If successful,
645 	 * on return, m will point to a list of packets to be sent.
646 	 */
647 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
648 	if (error)
649 		goto bad;
650 	for (; m; m = m0) {
651 		m0 = m->m_nextpkt;
652 		m->m_nextpkt = 0;
653 		if (error == 0) {
654 			/* Record statistics for this interface address. */
655 			if (ia != NULL) {
656 				ia->ia_ifa.if_opackets++;
657 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
658 			}
659 			/*
660 			 * Reset layer specific mbuf flags
661 			 * to avoid confusing upper layers.
662 			 */
663 			m->m_flags &= ~(M_PROTOFLAGS);
664 
665 			error = (*ifp->if_output)(ifp, m,
666 			    (struct sockaddr *)dst, ro);
667 		} else
668 			m_freem(m);
669 	}
670 
671 	if (error == 0)
672 		IPSTAT_INC(ips_fragmented);
673 
674 done:
675 	if (ro == &iproute)
676 		RO_RTFREE(ro);
677 	if (ia != NULL)
678 		ifa_free(&ia->ia_ifa);
679 	return (error);
680 bad:
681 	m_freem(m);
682 	goto done;
683 }
684 
685 /*
686  * Create a chain of fragments which fit the given mtu. m_frag points to the
687  * mbuf to be fragmented; on return it points to the chain with the fragments.
688  * Return 0 if no error. If error, m_frag may contain a partially built
689  * chain of fragments that should be freed by the caller.
690  *
691  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
692  */
693 int
694 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
695     u_long if_hwassist_flags)
696 {
697 	int error = 0;
698 	int hlen = ip->ip_hl << 2;
699 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
700 	int off;
701 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
702 	int firstlen;
703 	struct mbuf **mnext;
704 	int nfrags;
705 	uint16_t ip_len, ip_off;
706 
707 	ip_len = ntohs(ip->ip_len);
708 	ip_off = ntohs(ip->ip_off);
709 
710 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
711 		IPSTAT_INC(ips_cantfrag);
712 		return EMSGSIZE;
713 	}
714 
715 	/*
716 	 * Must be able to put at least 8 bytes per fragment.
717 	 */
718 	if (len < 8)
719 		return EMSGSIZE;
720 
721 	/*
722 	 * If the interface will not calculate checksums on
723 	 * fragmented packets, then do it here.
724 	 */
725 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
726 		in_delayed_cksum(m0);
727 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
728 	}
729 #ifdef SCTP
730 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
731 		sctp_delayed_cksum(m0, hlen);
732 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
733 	}
734 #endif
735 	if (len > PAGE_SIZE) {
736 		/*
737 		 * Fragment large datagrams such that each segment
738 		 * contains a multiple of PAGE_SIZE amount of data,
739 		 * plus headers. This enables a receiver to perform
740 		 * page-flipping zero-copy optimizations.
741 		 *
742 		 * XXX When does this help given that sender and receiver
743 		 * could have different page sizes, and also mtu could
744 		 * be less than the receiver's page size ?
745 		 */
746 		int newlen;
747 		struct mbuf *m;
748 
749 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
750 			off += m->m_len;
751 
752 		/*
753 		 * firstlen (off - hlen) must be aligned on an
754 		 * 8-byte boundary
755 		 */
756 		if (off < hlen)
757 			goto smart_frag_failure;
758 		off = ((off - hlen) & ~7) + hlen;
759 		newlen = (~PAGE_MASK) & mtu;
760 		if ((newlen + sizeof (struct ip)) > mtu) {
761 			/* we failed, go back the default */
762 smart_frag_failure:
763 			newlen = len;
764 			off = hlen + len;
765 		}
766 		len = newlen;
767 
768 	} else {
769 		off = hlen + len;
770 	}
771 
772 	firstlen = off - hlen;
773 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
774 
775 	/*
776 	 * Loop through length of segment after first fragment,
777 	 * make new header and copy data of each part and link onto chain.
778 	 * Here, m0 is the original packet, m is the fragment being created.
779 	 * The fragments are linked off the m_nextpkt of the original
780 	 * packet, which after processing serves as the first fragment.
781 	 */
782 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
783 		struct ip *mhip;	/* ip header on the fragment */
784 		struct mbuf *m;
785 		int mhlen = sizeof (struct ip);
786 
787 		MGETHDR(m, M_NOWAIT, MT_DATA);
788 		if (m == NULL) {
789 			error = ENOBUFS;
790 			IPSTAT_INC(ips_odropped);
791 			goto done;
792 		}
793 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
794 		/*
795 		 * In the first mbuf, leave room for the link header, then
796 		 * copy the original IP header including options. The payload
797 		 * goes into an additional mbuf chain returned by m_copym().
798 		 */
799 		m->m_data += max_linkhdr;
800 		mhip = mtod(m, struct ip *);
801 		*mhip = *ip;
802 		if (hlen > sizeof (struct ip)) {
803 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
804 			mhip->ip_v = IPVERSION;
805 			mhip->ip_hl = mhlen >> 2;
806 		}
807 		m->m_len = mhlen;
808 		/* XXX do we need to add ip_off below ? */
809 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
810 		if (off + len >= ip_len) {	/* last fragment */
811 			len = ip_len - off;
812 			m->m_flags |= M_LASTFRAG;
813 		} else
814 			mhip->ip_off |= IP_MF;
815 		mhip->ip_len = htons((u_short)(len + mhlen));
816 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
817 		if (m->m_next == NULL) {	/* copy failed */
818 			m_free(m);
819 			error = ENOBUFS;	/* ??? */
820 			IPSTAT_INC(ips_odropped);
821 			goto done;
822 		}
823 		m->m_pkthdr.len = mhlen + len;
824 		m->m_pkthdr.rcvif = NULL;
825 #ifdef MAC
826 		mac_netinet_fragment(m0, m);
827 #endif
828 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
829 		mhip->ip_off = htons(mhip->ip_off);
830 		mhip->ip_sum = 0;
831 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
832 			mhip->ip_sum = in_cksum(m, mhlen);
833 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
834 		}
835 		*mnext = m;
836 		mnext = &m->m_nextpkt;
837 	}
838 	IPSTAT_ADD(ips_ofragments, nfrags);
839 
840 	/* set first marker for fragment chain */
841 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
842 	m0->m_pkthdr.csum_data = nfrags;
843 
844 	/*
845 	 * Update first fragment by trimming what's been copied out
846 	 * and updating header.
847 	 */
848 	m_adj(m0, hlen + firstlen - ip_len);
849 	m0->m_pkthdr.len = hlen + firstlen;
850 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
851 	ip->ip_off = htons(ip_off | IP_MF);
852 	ip->ip_sum = 0;
853 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
854 		ip->ip_sum = in_cksum(m0, hlen);
855 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
856 	}
857 
858 done:
859 	*m_frag = m0;
860 	return error;
861 }
862 
863 void
864 in_delayed_cksum(struct mbuf *m)
865 {
866 	struct ip *ip;
867 	uint16_t csum, offset, ip_len;
868 
869 	ip = mtod(m, struct ip *);
870 	offset = ip->ip_hl << 2 ;
871 	ip_len = ntohs(ip->ip_len);
872 	csum = in_cksum_skip(m, ip_len, offset);
873 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
874 		csum = 0xffff;
875 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
876 
877 	if (offset + sizeof(u_short) > m->m_len) {
878 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
879 		    m->m_len, offset, ip->ip_p);
880 		/*
881 		 * XXX
882 		 * this shouldn't happen, but if it does, the
883 		 * correct behavior may be to insert the checksum
884 		 * in the appropriate next mbuf in the chain.
885 		 */
886 		return;
887 	}
888 	*(u_short *)(m->m_data + offset) = csum;
889 }
890 
891 /*
892  * IP socket option processing.
893  */
894 int
895 ip_ctloutput(struct socket *so, struct sockopt *sopt)
896 {
897 	struct	inpcb *inp = sotoinpcb(so);
898 	int	error, optval;
899 
900 	error = optval = 0;
901 	if (sopt->sopt_level != IPPROTO_IP) {
902 		error = EINVAL;
903 
904 		if (sopt->sopt_level == SOL_SOCKET &&
905 		    sopt->sopt_dir == SOPT_SET) {
906 			switch (sopt->sopt_name) {
907 			case SO_REUSEADDR:
908 				INP_WLOCK(inp);
909 				if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
910 					if ((so->so_options &
911 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
912 						inp->inp_flags2 |= INP_REUSEPORT;
913 					else
914 						inp->inp_flags2 &= ~INP_REUSEPORT;
915 				}
916 				INP_WUNLOCK(inp);
917 				error = 0;
918 				break;
919 			case SO_REUSEPORT:
920 				INP_WLOCK(inp);
921 				if ((so->so_options & SO_REUSEPORT) != 0)
922 					inp->inp_flags2 |= INP_REUSEPORT;
923 				else
924 					inp->inp_flags2 &= ~INP_REUSEPORT;
925 				INP_WUNLOCK(inp);
926 				error = 0;
927 				break;
928 			case SO_SETFIB:
929 				INP_WLOCK(inp);
930 				inp->inp_inc.inc_fibnum = so->so_fibnum;
931 				INP_WUNLOCK(inp);
932 				error = 0;
933 				break;
934 			default:
935 				break;
936 			}
937 		}
938 		return (error);
939 	}
940 
941 	switch (sopt->sopt_dir) {
942 	case SOPT_SET:
943 		switch (sopt->sopt_name) {
944 		case IP_OPTIONS:
945 #ifdef notyet
946 		case IP_RETOPTS:
947 #endif
948 		{
949 			struct mbuf *m;
950 			if (sopt->sopt_valsize > MLEN) {
951 				error = EMSGSIZE;
952 				break;
953 			}
954 			MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
955 			if (m == NULL) {
956 				error = ENOBUFS;
957 				break;
958 			}
959 			m->m_len = sopt->sopt_valsize;
960 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
961 					    m->m_len);
962 			if (error) {
963 				m_free(m);
964 				break;
965 			}
966 			INP_WLOCK(inp);
967 			error = ip_pcbopts(inp, sopt->sopt_name, m);
968 			INP_WUNLOCK(inp);
969 			return (error);
970 		}
971 
972 		case IP_BINDANY:
973 			if (sopt->sopt_td != NULL) {
974 				error = priv_check(sopt->sopt_td,
975 				    PRIV_NETINET_BINDANY);
976 				if (error)
977 					break;
978 			}
979 			/* FALLTHROUGH */
980 		case IP_TOS:
981 		case IP_TTL:
982 		case IP_MINTTL:
983 		case IP_RECVOPTS:
984 		case IP_RECVRETOPTS:
985 		case IP_RECVDSTADDR:
986 		case IP_RECVTTL:
987 		case IP_RECVIF:
988 		case IP_FAITH:
989 		case IP_ONESBCAST:
990 		case IP_DONTFRAG:
991 		case IP_RECVTOS:
992 			error = sooptcopyin(sopt, &optval, sizeof optval,
993 					    sizeof optval);
994 			if (error)
995 				break;
996 
997 			switch (sopt->sopt_name) {
998 			case IP_TOS:
999 				inp->inp_ip_tos = optval;
1000 				break;
1001 
1002 			case IP_TTL:
1003 				inp->inp_ip_ttl = optval;
1004 				break;
1005 
1006 			case IP_MINTTL:
1007 				if (optval >= 0 && optval <= MAXTTL)
1008 					inp->inp_ip_minttl = optval;
1009 				else
1010 					error = EINVAL;
1011 				break;
1012 
1013 #define	OPTSET(bit) do {						\
1014 	INP_WLOCK(inp);							\
1015 	if (optval)							\
1016 		inp->inp_flags |= bit;					\
1017 	else								\
1018 		inp->inp_flags &= ~bit;					\
1019 	INP_WUNLOCK(inp);						\
1020 } while (0)
1021 
1022 			case IP_RECVOPTS:
1023 				OPTSET(INP_RECVOPTS);
1024 				break;
1025 
1026 			case IP_RECVRETOPTS:
1027 				OPTSET(INP_RECVRETOPTS);
1028 				break;
1029 
1030 			case IP_RECVDSTADDR:
1031 				OPTSET(INP_RECVDSTADDR);
1032 				break;
1033 
1034 			case IP_RECVTTL:
1035 				OPTSET(INP_RECVTTL);
1036 				break;
1037 
1038 			case IP_RECVIF:
1039 				OPTSET(INP_RECVIF);
1040 				break;
1041 
1042 			case IP_FAITH:
1043 				OPTSET(INP_FAITH);
1044 				break;
1045 
1046 			case IP_ONESBCAST:
1047 				OPTSET(INP_ONESBCAST);
1048 				break;
1049 			case IP_DONTFRAG:
1050 				OPTSET(INP_DONTFRAG);
1051 				break;
1052 			case IP_BINDANY:
1053 				OPTSET(INP_BINDANY);
1054 				break;
1055 			case IP_RECVTOS:
1056 				OPTSET(INP_RECVTOS);
1057 				break;
1058 			}
1059 			break;
1060 #undef OPTSET
1061 
1062 		/*
1063 		 * Multicast socket options are processed by the in_mcast
1064 		 * module.
1065 		 */
1066 		case IP_MULTICAST_IF:
1067 		case IP_MULTICAST_VIF:
1068 		case IP_MULTICAST_TTL:
1069 		case IP_MULTICAST_LOOP:
1070 		case IP_ADD_MEMBERSHIP:
1071 		case IP_DROP_MEMBERSHIP:
1072 		case IP_ADD_SOURCE_MEMBERSHIP:
1073 		case IP_DROP_SOURCE_MEMBERSHIP:
1074 		case IP_BLOCK_SOURCE:
1075 		case IP_UNBLOCK_SOURCE:
1076 		case IP_MSFILTER:
1077 		case MCAST_JOIN_GROUP:
1078 		case MCAST_LEAVE_GROUP:
1079 		case MCAST_JOIN_SOURCE_GROUP:
1080 		case MCAST_LEAVE_SOURCE_GROUP:
1081 		case MCAST_BLOCK_SOURCE:
1082 		case MCAST_UNBLOCK_SOURCE:
1083 			error = inp_setmoptions(inp, sopt);
1084 			break;
1085 
1086 		case IP_PORTRANGE:
1087 			error = sooptcopyin(sopt, &optval, sizeof optval,
1088 					    sizeof optval);
1089 			if (error)
1090 				break;
1091 
1092 			INP_WLOCK(inp);
1093 			switch (optval) {
1094 			case IP_PORTRANGE_DEFAULT:
1095 				inp->inp_flags &= ~(INP_LOWPORT);
1096 				inp->inp_flags &= ~(INP_HIGHPORT);
1097 				break;
1098 
1099 			case IP_PORTRANGE_HIGH:
1100 				inp->inp_flags &= ~(INP_LOWPORT);
1101 				inp->inp_flags |= INP_HIGHPORT;
1102 				break;
1103 
1104 			case IP_PORTRANGE_LOW:
1105 				inp->inp_flags &= ~(INP_HIGHPORT);
1106 				inp->inp_flags |= INP_LOWPORT;
1107 				break;
1108 
1109 			default:
1110 				error = EINVAL;
1111 				break;
1112 			}
1113 			INP_WUNLOCK(inp);
1114 			break;
1115 
1116 #ifdef IPSEC
1117 		case IP_IPSEC_POLICY:
1118 		{
1119 			caddr_t req;
1120 			struct mbuf *m;
1121 
1122 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1123 				break;
1124 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1125 				break;
1126 			req = mtod(m, caddr_t);
1127 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1128 			    m->m_len, (sopt->sopt_td != NULL) ?
1129 			    sopt->sopt_td->td_ucred : NULL);
1130 			m_freem(m);
1131 			break;
1132 		}
1133 #endif /* IPSEC */
1134 
1135 		default:
1136 			error = ENOPROTOOPT;
1137 			break;
1138 		}
1139 		break;
1140 
1141 	case SOPT_GET:
1142 		switch (sopt->sopt_name) {
1143 		case IP_OPTIONS:
1144 		case IP_RETOPTS:
1145 			if (inp->inp_options)
1146 				error = sooptcopyout(sopt,
1147 						     mtod(inp->inp_options,
1148 							  char *),
1149 						     inp->inp_options->m_len);
1150 			else
1151 				sopt->sopt_valsize = 0;
1152 			break;
1153 
1154 		case IP_TOS:
1155 		case IP_TTL:
1156 		case IP_MINTTL:
1157 		case IP_RECVOPTS:
1158 		case IP_RECVRETOPTS:
1159 		case IP_RECVDSTADDR:
1160 		case IP_RECVTTL:
1161 		case IP_RECVIF:
1162 		case IP_PORTRANGE:
1163 		case IP_FAITH:
1164 		case IP_ONESBCAST:
1165 		case IP_DONTFRAG:
1166 		case IP_BINDANY:
1167 		case IP_RECVTOS:
1168 			switch (sopt->sopt_name) {
1169 
1170 			case IP_TOS:
1171 				optval = inp->inp_ip_tos;
1172 				break;
1173 
1174 			case IP_TTL:
1175 				optval = inp->inp_ip_ttl;
1176 				break;
1177 
1178 			case IP_MINTTL:
1179 				optval = inp->inp_ip_minttl;
1180 				break;
1181 
1182 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1183 
1184 			case IP_RECVOPTS:
1185 				optval = OPTBIT(INP_RECVOPTS);
1186 				break;
1187 
1188 			case IP_RECVRETOPTS:
1189 				optval = OPTBIT(INP_RECVRETOPTS);
1190 				break;
1191 
1192 			case IP_RECVDSTADDR:
1193 				optval = OPTBIT(INP_RECVDSTADDR);
1194 				break;
1195 
1196 			case IP_RECVTTL:
1197 				optval = OPTBIT(INP_RECVTTL);
1198 				break;
1199 
1200 			case IP_RECVIF:
1201 				optval = OPTBIT(INP_RECVIF);
1202 				break;
1203 
1204 			case IP_PORTRANGE:
1205 				if (inp->inp_flags & INP_HIGHPORT)
1206 					optval = IP_PORTRANGE_HIGH;
1207 				else if (inp->inp_flags & INP_LOWPORT)
1208 					optval = IP_PORTRANGE_LOW;
1209 				else
1210 					optval = 0;
1211 				break;
1212 
1213 			case IP_FAITH:
1214 				optval = OPTBIT(INP_FAITH);
1215 				break;
1216 
1217 			case IP_ONESBCAST:
1218 				optval = OPTBIT(INP_ONESBCAST);
1219 				break;
1220 			case IP_DONTFRAG:
1221 				optval = OPTBIT(INP_DONTFRAG);
1222 				break;
1223 			case IP_BINDANY:
1224 				optval = OPTBIT(INP_BINDANY);
1225 				break;
1226 			case IP_RECVTOS:
1227 				optval = OPTBIT(INP_RECVTOS);
1228 				break;
1229 			}
1230 			error = sooptcopyout(sopt, &optval, sizeof optval);
1231 			break;
1232 
1233 		/*
1234 		 * Multicast socket options are processed by the in_mcast
1235 		 * module.
1236 		 */
1237 		case IP_MULTICAST_IF:
1238 		case IP_MULTICAST_VIF:
1239 		case IP_MULTICAST_TTL:
1240 		case IP_MULTICAST_LOOP:
1241 		case IP_MSFILTER:
1242 			error = inp_getmoptions(inp, sopt);
1243 			break;
1244 
1245 #ifdef IPSEC
1246 		case IP_IPSEC_POLICY:
1247 		{
1248 			struct mbuf *m = NULL;
1249 			caddr_t req = NULL;
1250 			size_t len = 0;
1251 
1252 			if (m != 0) {
1253 				req = mtod(m, caddr_t);
1254 				len = m->m_len;
1255 			}
1256 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1257 			if (error == 0)
1258 				error = soopt_mcopyout(sopt, m); /* XXX */
1259 			if (error == 0)
1260 				m_freem(m);
1261 			break;
1262 		}
1263 #endif /* IPSEC */
1264 
1265 		default:
1266 			error = ENOPROTOOPT;
1267 			break;
1268 		}
1269 		break;
1270 	}
1271 	return (error);
1272 }
1273 
1274 /*
1275  * Routine called from ip_output() to loop back a copy of an IP multicast
1276  * packet to the input queue of a specified interface.  Note that this
1277  * calls the output routine of the loopback "driver", but with an interface
1278  * pointer that might NOT be a loopback interface -- evil, but easier than
1279  * replicating that code here.
1280  */
1281 static void
1282 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1283     int hlen)
1284 {
1285 	register struct ip *ip;
1286 	struct mbuf *copym;
1287 
1288 	/*
1289 	 * Make a deep copy of the packet because we're going to
1290 	 * modify the pack in order to generate checksums.
1291 	 */
1292 	copym = m_dup(m, M_NOWAIT);
1293 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1294 		copym = m_pullup(copym, hlen);
1295 	if (copym != NULL) {
1296 		/* If needed, compute the checksum and mark it as valid. */
1297 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1298 			in_delayed_cksum(copym);
1299 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1300 			copym->m_pkthdr.csum_flags |=
1301 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1302 			copym->m_pkthdr.csum_data = 0xffff;
1303 		}
1304 		/*
1305 		 * We don't bother to fragment if the IP length is greater
1306 		 * than the interface's MTU.  Can this possibly matter?
1307 		 */
1308 		ip = mtod(copym, struct ip *);
1309 		ip->ip_sum = 0;
1310 		ip->ip_sum = in_cksum(copym, hlen);
1311 #if 1 /* XXX */
1312 		if (dst->sin_family != AF_INET) {
1313 			printf("ip_mloopback: bad address family %d\n",
1314 						dst->sin_family);
1315 			dst->sin_family = AF_INET;
1316 		}
1317 #endif
1318 		if_simloop(ifp, copym, dst->sin_family, 0);
1319 	}
1320 }
1321