xref: /freebsd/sys/netinet/ip_output.c (revision a10cee30c94cf5944826d2a495e9cdf339dfbcc8)
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  * ip_len and ip_off are in host format.
106  * The mbuf chain containing the packet will be freed.
107  * The mbuf opt, if present, will not be freed.
108  * If route ro is present and has ro_rt initialized, route lookup would be
109  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
110  * then result of route lookup is stored in ro->ro_rt.
111  *
112  * In the IP forwarding case, the packet will arrive with options already
113  * inserted, so must have a NULL opt pointer.
114  */
115 int
116 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
117     struct ip_moptions *imo, struct inpcb *inp)
118 {
119 	struct ip *ip;
120 	struct ifnet *ifp = NULL;	/* keep compiler happy */
121 	struct mbuf *m0;
122 	int hlen = sizeof (struct ip);
123 	int mtu;
124 	int n;	/* scratchpad */
125 	int error = 0;
126 	struct sockaddr_in *dst;
127 	struct in_ifaddr *ia;
128 	int isbroadcast;
129 	uint16_t ip_len, ip_off, sw_csum;
130 	struct route iproute;
131 	struct rtentry *rte;	/* cache for ro->ro_rt */
132 	struct in_addr odst;
133 #ifdef IPFIREWALL_FORWARD
134 	struct m_tag *fwd_tag = NULL;
135 #endif
136 #ifdef IPSEC
137 	int no_route_but_check_spd = 0;
138 #endif
139 	M_ASSERTPKTHDR(m);
140 
141 	if (inp != NULL) {
142 		INP_LOCK_ASSERT(inp);
143 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
144 		if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
145 			m->m_pkthdr.flowid = inp->inp_flowid;
146 			m->m_flags |= M_FLOWID;
147 		}
148 	}
149 
150 	if (ro == NULL) {
151 		ro = &iproute;
152 		bzero(ro, sizeof (*ro));
153 	}
154 
155 #ifdef FLOWTABLE
156 	if (ro->ro_rt == NULL) {
157 		struct flentry *fle;
158 
159 		/*
160 		 * The flow table returns route entries valid for up to 30
161 		 * seconds; we rely on the remainder of ip_output() taking no
162 		 * longer than that long for the stability of ro_rt. The
163 		 * flow ID assignment must have happened before this point.
164 		 */
165 		fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
166 		if (fle != NULL)
167 			flow_to_route(fle, ro);
168 	}
169 #endif
170 
171 	if (opt) {
172 		int len = 0;
173 		m = ip_insertoptions(m, opt, &len);
174 		if (len != 0)
175 			hlen = len; /* ip->ip_hl is updated above */
176 	}
177 	ip = mtod(m, struct ip *);
178 
179 	/*
180 	 * Fill in IP header.  If we are not allowing fragmentation,
181 	 * then the ip_id field is meaningless, but we don't set it
182 	 * to zero.  Doing so causes various problems when devices along
183 	 * the path (routers, load balancers, firewalls, etc.) illegally
184 	 * disable DF on our packet.  Note that a 16-bit counter
185 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
186 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
187 	 * for Counting NATted Hosts", Proc. IMW'02, available at
188 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
189 	 */
190 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
191 		ip->ip_v = IPVERSION;
192 		ip->ip_hl = hlen >> 2;
193 		ip->ip_id = ip_newid();
194 		IPSTAT_INC(ips_localout);
195 	} else {
196 		/* Header already set, fetch hlen from there */
197 		hlen = ip->ip_hl << 2;
198 	}
199 
200 	dst = (struct sockaddr_in *)&ro->ro_dst;
201 again:
202 	ia = NULL;
203 	/*
204 	 * If there is a cached route,
205 	 * check that it is to the same destination
206 	 * and is still up.  If not, free it and try again.
207 	 * The address family should also be checked in case of sharing the
208 	 * cache with IPv6.
209 	 */
210 	rte = ro->ro_rt;
211 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
212 		    rte->rt_ifp == NULL ||
213 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
214 			  dst->sin_family != AF_INET ||
215 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
216 		RO_RTFREE(ro);
217 		ro->ro_lle = NULL;
218 		rte = NULL;
219 	}
220 #ifdef IPFIREWALL_FORWARD
221 	if (rte == NULL && fwd_tag == NULL) {
222 #else
223 	if (rte == NULL) {
224 #endif
225 		bzero(dst, sizeof(*dst));
226 		dst->sin_family = AF_INET;
227 		dst->sin_len = sizeof(*dst);
228 		dst->sin_addr = ip->ip_dst;
229 	}
230 	/*
231 	 * If routing to interface only, short circuit routing lookup.
232 	 * The use of an all-ones broadcast address implies this; an
233 	 * interface is specified by the broadcast address of an interface,
234 	 * or the destination address of a ptp interface.
235 	 */
236 	if (flags & IP_SENDONES) {
237 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
238 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
239 			IPSTAT_INC(ips_noroute);
240 			error = ENETUNREACH;
241 			goto bad;
242 		}
243 		ip->ip_dst.s_addr = INADDR_BROADCAST;
244 		dst->sin_addr = ip->ip_dst;
245 		ifp = ia->ia_ifp;
246 		ip->ip_ttl = 1;
247 		isbroadcast = 1;
248 	} else if (flags & IP_ROUTETOIF) {
249 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
250 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
251 			IPSTAT_INC(ips_noroute);
252 			error = ENETUNREACH;
253 			goto bad;
254 		}
255 		ifp = ia->ia_ifp;
256 		ip->ip_ttl = 1;
257 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
258 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
259 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
260 		/*
261 		 * Bypass the normal routing lookup for multicast
262 		 * packets if the interface is specified.
263 		 */
264 		ifp = imo->imo_multicast_ifp;
265 		IFP_TO_IA(ifp, ia);
266 		isbroadcast = 0;	/* fool gcc */
267 	} else {
268 		/*
269 		 * We want to do any cloning requested by the link layer,
270 		 * as this is probably required in all cases for correct
271 		 * operation (as it is for ARP).
272 		 */
273 		if (rte == NULL) {
274 #ifdef RADIX_MPATH
275 			rtalloc_mpath_fib(ro,
276 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
277 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
278 #else
279 			in_rtalloc_ign(ro, 0,
280 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
281 #endif
282 			rte = ro->ro_rt;
283 		}
284 		if (rte == NULL ||
285 		    rte->rt_ifp == NULL ||
286 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
287 #ifdef IPSEC
288 			/*
289 			 * There is no route for this packet, but it is
290 			 * possible that a matching SPD entry exists.
291 			 */
292 			no_route_but_check_spd = 1;
293 			mtu = 0; /* Silence GCC warning. */
294 			goto sendit;
295 #endif
296 			IPSTAT_INC(ips_noroute);
297 			error = EHOSTUNREACH;
298 			goto bad;
299 		}
300 		ia = ifatoia(rte->rt_ifa);
301 		ifa_ref(&ia->ia_ifa);
302 		ifp = rte->rt_ifp;
303 		rte->rt_rmx.rmx_pksent++;
304 		if (rte->rt_flags & RTF_GATEWAY)
305 			dst = (struct sockaddr_in *)rte->rt_gateway;
306 		if (rte->rt_flags & RTF_HOST)
307 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
308 		else
309 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
310 	}
311 	/*
312 	 * Calculate MTU.  If we have a route that is up, use that,
313 	 * otherwise use the interface's MTU.
314 	 */
315 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
316 		/*
317 		 * This case can happen if the user changed the MTU
318 		 * of an interface after enabling IP on it.  Because
319 		 * most netifs don't keep track of routes pointing to
320 		 * them, there is no way for one to update all its
321 		 * routes when the MTU is changed.
322 		 */
323 		if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
324 			rte->rt_rmx.rmx_mtu = ifp->if_mtu;
325 		mtu = rte->rt_rmx.rmx_mtu;
326 	} else {
327 		mtu = ifp->if_mtu;
328 	}
329 	/* Catch a possible divide by zero later. */
330 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
331 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
332 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
333 		m->m_flags |= M_MCAST;
334 		/*
335 		 * IP destination address is multicast.  Make sure "dst"
336 		 * still points to the address in "ro".  (It may have been
337 		 * changed to point to a gateway address, above.)
338 		 */
339 		dst = (struct sockaddr_in *)&ro->ro_dst;
340 		/*
341 		 * See if the caller provided any multicast options
342 		 */
343 		if (imo != NULL) {
344 			ip->ip_ttl = imo->imo_multicast_ttl;
345 			if (imo->imo_multicast_vif != -1)
346 				ip->ip_src.s_addr =
347 				    ip_mcast_src ?
348 				    ip_mcast_src(imo->imo_multicast_vif) :
349 				    INADDR_ANY;
350 		} else
351 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
352 		/*
353 		 * Confirm that the outgoing interface supports multicast.
354 		 */
355 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
356 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
357 				IPSTAT_INC(ips_noroute);
358 				error = ENETUNREACH;
359 				goto bad;
360 			}
361 		}
362 		/*
363 		 * If source address not specified yet, use address
364 		 * of outgoing interface.
365 		 */
366 		if (ip->ip_src.s_addr == INADDR_ANY) {
367 			/* Interface may have no addresses. */
368 			if (ia != NULL)
369 				ip->ip_src = IA_SIN(ia)->sin_addr;
370 		}
371 
372 		if ((imo == NULL && in_mcast_loop) ||
373 		    (imo && imo->imo_multicast_loop)) {
374 			/*
375 			 * Loop back multicast datagram if not expressly
376 			 * forbidden to do so, even if we are not a member
377 			 * of the group; ip_input() will filter it later,
378 			 * thus deferring a hash lookup and mutex acquisition
379 			 * at the expense of a cheap copy using m_copym().
380 			 */
381 			ip_mloopback(ifp, m, dst, hlen);
382 		} else {
383 			/*
384 			 * If we are acting as a multicast router, perform
385 			 * multicast forwarding as if the packet had just
386 			 * arrived on the interface to which we are about
387 			 * to send.  The multicast forwarding function
388 			 * recursively calls this function, using the
389 			 * IP_FORWARDING flag to prevent infinite recursion.
390 			 *
391 			 * Multicasts that are looped back by ip_mloopback(),
392 			 * above, will be forwarded by the ip_input() routine,
393 			 * if necessary.
394 			 */
395 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
396 				/*
397 				 * If rsvp daemon is not running, do not
398 				 * set ip_moptions. This ensures that the packet
399 				 * is multicast and not just sent down one link
400 				 * as prescribed by rsvpd.
401 				 */
402 				if (!V_rsvp_on)
403 					imo = NULL;
404 				if (ip_mforward &&
405 				    ip_mforward(ip, ifp, m, imo) != 0) {
406 					m_freem(m);
407 					goto done;
408 				}
409 			}
410 		}
411 
412 		/*
413 		 * Multicasts with a time-to-live of zero may be looped-
414 		 * back, above, but must not be transmitted on a network.
415 		 * Also, multicasts addressed to the loopback interface
416 		 * are not sent -- the above call to ip_mloopback() will
417 		 * loop back a copy. ip_input() will drop the copy if
418 		 * this host does not belong to the destination group on
419 		 * the loopback interface.
420 		 */
421 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
422 			m_freem(m);
423 			goto done;
424 		}
425 
426 		goto sendit;
427 	}
428 
429 	/*
430 	 * If the source address is not specified yet, use the address
431 	 * of the outoing interface.
432 	 */
433 	if (ip->ip_src.s_addr == INADDR_ANY) {
434 		/* Interface may have no addresses. */
435 		if (ia != NULL) {
436 			ip->ip_src = IA_SIN(ia)->sin_addr;
437 		}
438 	}
439 
440 	/*
441 	 * Verify that we have any chance at all of being able to queue the
442 	 * packet or packet fragments, unless ALTQ is enabled on the given
443 	 * interface in which case packetdrop should be done by queueing.
444 	 */
445 	n = ip->ip_len / mtu + 1; /* how many fragments ? */
446 	if (
447 #ifdef ALTQ
448 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
449 #endif /* ALTQ */
450 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
451 		error = ENOBUFS;
452 		IPSTAT_INC(ips_odropped);
453 		ifp->if_snd.ifq_drops += n;
454 		goto bad;
455 	}
456 
457 	/*
458 	 * Look for broadcast address and
459 	 * verify user is allowed to send
460 	 * such a packet.
461 	 */
462 	if (isbroadcast) {
463 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
464 			error = EADDRNOTAVAIL;
465 			goto bad;
466 		}
467 		if ((flags & IP_ALLOWBROADCAST) == 0) {
468 			error = EACCES;
469 			goto bad;
470 		}
471 		/* don't allow broadcast messages to be fragmented */
472 		if (ip->ip_len > mtu) {
473 			error = EMSGSIZE;
474 			goto bad;
475 		}
476 		m->m_flags |= M_BCAST;
477 	} else {
478 		m->m_flags &= ~M_BCAST;
479 	}
480 
481 sendit:
482 #ifdef IPSEC
483 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
484 	case 1:
485 		goto bad;
486 	case -1:
487 		goto done;
488 	case 0:
489 	default:
490 		break;	/* Continue with packet processing. */
491 	}
492 	/*
493 	 * Check if there was a route for this packet; return error if not.
494 	 */
495 	if (no_route_but_check_spd) {
496 		IPSTAT_INC(ips_noroute);
497 		error = EHOSTUNREACH;
498 		goto bad;
499 	}
500 	/* Update variables that are affected by ipsec4_output(). */
501 	ip = mtod(m, struct ip *);
502 	hlen = ip->ip_hl << 2;
503 #endif /* IPSEC */
504 
505 	/*
506 	 * To network byte order. pfil(9) hooks and ip_fragment() expect this.
507 	 */
508 	ip->ip_len = htons(ip->ip_len);
509 	ip->ip_off = htons(ip->ip_off);
510 
511 	/* Jump over all PFIL processing if hooks are not active. */
512 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
513 		goto passout;
514 
515 	/* Run through list of hooks for output packets. */
516 	odst.s_addr = ip->ip_dst.s_addr;
517 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
518 	if (error != 0 || m == NULL)
519 		goto done;
520 
521 	ip = mtod(m, struct ip *);
522 
523 	/* See if destination IP address was changed by packet filter. */
524 	if (odst.s_addr != ip->ip_dst.s_addr) {
525 		m->m_flags |= M_SKIP_FIREWALL;
526 		/* If destination is now ourself drop to ip_input(). */
527 		if (in_localip(ip->ip_dst)) {
528 			m->m_flags |= M_FASTFWD_OURS;
529 			if (m->m_pkthdr.rcvif == NULL)
530 				m->m_pkthdr.rcvif = V_loif;
531 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
532 				m->m_pkthdr.csum_flags |=
533 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
534 				m->m_pkthdr.csum_data = 0xffff;
535 			}
536 			m->m_pkthdr.csum_flags |=
537 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
538 #ifdef SCTP
539 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
540 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
541 #endif
542 			error = netisr_queue(NETISR_IP, m);
543 			goto done;
544 		} else {
545 			if (ia != NULL)
546 				ifa_free(&ia->ia_ifa);
547 			ip->ip_len = ntohs(ip->ip_len);
548 			ip->ip_off = ntohs(ip->ip_off);
549 			goto again;	/* Redo the routing table lookup. */
550 		}
551 	}
552 
553 #ifdef IPFIREWALL_FORWARD
554 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
555 	if (m->m_flags & M_FASTFWD_OURS) {
556 		if (m->m_pkthdr.rcvif == NULL)
557 			m->m_pkthdr.rcvif = V_loif;
558 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
559 			m->m_pkthdr.csum_flags |=
560 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
561 			m->m_pkthdr.csum_data = 0xffff;
562 		}
563 #ifdef SCTP
564 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
565 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
566 #endif
567 		m->m_pkthdr.csum_flags |=
568 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
569 
570 		error = netisr_queue(NETISR_IP, m);
571 		goto done;
572 	}
573 	/* Or forward to some other address? */
574 	fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
575 	if (fwd_tag) {
576 		dst = (struct sockaddr_in *)&ro->ro_dst;
577 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
578 		m->m_flags |= M_SKIP_FIREWALL;
579 		m_tag_delete(m, fwd_tag);
580 		if (ia != NULL)
581 			ifa_free(&ia->ia_ifa);
582 		ip->ip_len = ntohs(ip->ip_len);
583 		ip->ip_off = ntohs(ip->ip_off);
584 		goto again;
585 	}
586 #endif /* IPFIREWALL_FORWARD */
587 
588 passout:
589 	ip_len = ntohs(ip->ip_len);
590 	ip_off = ntohs(ip->ip_off);
591 
592 	/* 127/8 must not appear on wire - RFC1122. */
593 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
594 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
595 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
596 			IPSTAT_INC(ips_badaddr);
597 			error = EADDRNOTAVAIL;
598 			goto bad;
599 		}
600 	}
601 
602 	m->m_pkthdr.csum_flags |= CSUM_IP;
603 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
604 	if (sw_csum & CSUM_DELAY_DATA) {
605 		in_delayed_cksum(m);
606 		sw_csum &= ~CSUM_DELAY_DATA;
607 	}
608 #ifdef SCTP
609 	if (sw_csum & CSUM_SCTP) {
610 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
611 		sw_csum &= ~CSUM_SCTP;
612 	}
613 #endif
614 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
615 
616 	/*
617 	 * If small enough for interface, or the interface will take
618 	 * care of the fragmentation for us, we can just send directly.
619 	 */
620 	if (ip_len <= mtu ||
621 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
622 	    ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
623 		ip->ip_sum = 0;
624 		if (sw_csum & CSUM_DELAY_IP)
625 			ip->ip_sum = in_cksum(m, hlen);
626 
627 		/*
628 		 * Record statistics for this interface address.
629 		 * With CSUM_TSO the byte/packet count will be slightly
630 		 * incorrect because we count the IP+TCP headers only
631 		 * once instead of for every generated packet.
632 		 */
633 		if (!(flags & IP_FORWARDING) && ia) {
634 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
635 				ia->ia_ifa.if_opackets +=
636 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
637 			else
638 				ia->ia_ifa.if_opackets++;
639 			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
640 		}
641 #ifdef MBUF_STRESS_TEST
642 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
643 			m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
644 #endif
645 		/*
646 		 * Reset layer specific mbuf flags
647 		 * to avoid confusing lower layers.
648 		 */
649 		m->m_flags &= ~(M_PROTOFLAGS);
650 		error = (*ifp->if_output)(ifp, m,
651 		    		(struct sockaddr *)dst, ro);
652 		goto done;
653 	}
654 
655 	/* Balk when DF bit is set or the interface didn't support TSO. */
656 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
657 		error = EMSGSIZE;
658 		IPSTAT_INC(ips_cantfrag);
659 		goto bad;
660 	}
661 
662 	/*
663 	 * Too large for interface; fragment if possible. If successful,
664 	 * on return, m will point to a list of packets to be sent.
665 	 */
666 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
667 	if (error)
668 		goto bad;
669 	for (; m; m = m0) {
670 		m0 = m->m_nextpkt;
671 		m->m_nextpkt = 0;
672 		if (error == 0) {
673 			/* Record statistics for this interface address. */
674 			if (ia != NULL) {
675 				ia->ia_ifa.if_opackets++;
676 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
677 			}
678 			/*
679 			 * Reset layer specific mbuf flags
680 			 * to avoid confusing upper layers.
681 			 */
682 			m->m_flags &= ~(M_PROTOFLAGS);
683 
684 			error = (*ifp->if_output)(ifp, m,
685 			    (struct sockaddr *)dst, ro);
686 		} else
687 			m_freem(m);
688 	}
689 
690 	if (error == 0)
691 		IPSTAT_INC(ips_fragmented);
692 
693 done:
694 	if (ro == &iproute)
695 		RO_RTFREE(ro);
696 	if (ia != NULL)
697 		ifa_free(&ia->ia_ifa);
698 	return (error);
699 bad:
700 	m_freem(m);
701 	goto done;
702 }
703 
704 /*
705  * Create a chain of fragments which fit the given mtu. m_frag points to the
706  * mbuf to be fragmented; on return it points to the chain with the fragments.
707  * Return 0 if no error. If error, m_frag may contain a partially built
708  * chain of fragments that should be freed by the caller.
709  *
710  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
711  * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
712  */
713 int
714 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
715     u_long if_hwassist_flags, int sw_csum)
716 {
717 	int error = 0;
718 	int hlen = ip->ip_hl << 2;
719 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
720 	int off;
721 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
722 	int firstlen;
723 	struct mbuf **mnext;
724 	int nfrags;
725 	uint16_t ip_len, ip_off;
726 
727 	ip_len = ntohs(ip->ip_len);
728 	ip_off = ntohs(ip->ip_off);
729 
730 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
731 		IPSTAT_INC(ips_cantfrag);
732 		return EMSGSIZE;
733 	}
734 
735 	/*
736 	 * Must be able to put at least 8 bytes per fragment.
737 	 */
738 	if (len < 8)
739 		return EMSGSIZE;
740 
741 	/*
742 	 * If the interface will not calculate checksums on
743 	 * fragmented packets, then do it here.
744 	 */
745 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
746 	    (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
747 		in_delayed_cksum(m0);
748 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
749 	}
750 #ifdef SCTP
751 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
752 	    (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
753 		sctp_delayed_cksum(m0, hlen);
754 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
755 	}
756 #endif
757 	if (len > PAGE_SIZE) {
758 		/*
759 		 * Fragment large datagrams such that each segment
760 		 * contains a multiple of PAGE_SIZE amount of data,
761 		 * plus headers. This enables a receiver to perform
762 		 * page-flipping zero-copy optimizations.
763 		 *
764 		 * XXX When does this help given that sender and receiver
765 		 * could have different page sizes, and also mtu could
766 		 * be less than the receiver's page size ?
767 		 */
768 		int newlen;
769 		struct mbuf *m;
770 
771 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
772 			off += m->m_len;
773 
774 		/*
775 		 * firstlen (off - hlen) must be aligned on an
776 		 * 8-byte boundary
777 		 */
778 		if (off < hlen)
779 			goto smart_frag_failure;
780 		off = ((off - hlen) & ~7) + hlen;
781 		newlen = (~PAGE_MASK) & mtu;
782 		if ((newlen + sizeof (struct ip)) > mtu) {
783 			/* we failed, go back the default */
784 smart_frag_failure:
785 			newlen = len;
786 			off = hlen + len;
787 		}
788 		len = newlen;
789 
790 	} else {
791 		off = hlen + len;
792 	}
793 
794 	firstlen = off - hlen;
795 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
796 
797 	/*
798 	 * Loop through length of segment after first fragment,
799 	 * make new header and copy data of each part and link onto chain.
800 	 * Here, m0 is the original packet, m is the fragment being created.
801 	 * The fragments are linked off the m_nextpkt of the original
802 	 * packet, which after processing serves as the first fragment.
803 	 */
804 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
805 		struct ip *mhip;	/* ip header on the fragment */
806 		struct mbuf *m;
807 		int mhlen = sizeof (struct ip);
808 
809 		MGETHDR(m, M_DONTWAIT, MT_DATA);
810 		if (m == NULL) {
811 			error = ENOBUFS;
812 			IPSTAT_INC(ips_odropped);
813 			goto done;
814 		}
815 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
816 		/*
817 		 * In the first mbuf, leave room for the link header, then
818 		 * copy the original IP header including options. The payload
819 		 * goes into an additional mbuf chain returned by m_copym().
820 		 */
821 		m->m_data += max_linkhdr;
822 		mhip = mtod(m, struct ip *);
823 		*mhip = *ip;
824 		if (hlen > sizeof (struct ip)) {
825 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
826 			mhip->ip_v = IPVERSION;
827 			mhip->ip_hl = mhlen >> 2;
828 		}
829 		m->m_len = mhlen;
830 		/* XXX do we need to add ip_off below ? */
831 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
832 		if (off + len >= ip_len) {	/* last fragment */
833 			len = ip_len - off;
834 			m->m_flags |= M_LASTFRAG;
835 		} else
836 			mhip->ip_off |= IP_MF;
837 		mhip->ip_len = htons((u_short)(len + mhlen));
838 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
839 		if (m->m_next == NULL) {	/* copy failed */
840 			m_free(m);
841 			error = ENOBUFS;	/* ??? */
842 			IPSTAT_INC(ips_odropped);
843 			goto done;
844 		}
845 		m->m_pkthdr.len = mhlen + len;
846 		m->m_pkthdr.rcvif = NULL;
847 #ifdef MAC
848 		mac_netinet_fragment(m0, m);
849 #endif
850 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
851 		mhip->ip_off = htons(mhip->ip_off);
852 		mhip->ip_sum = 0;
853 		if (sw_csum & CSUM_DELAY_IP)
854 			mhip->ip_sum = in_cksum(m, mhlen);
855 		*mnext = m;
856 		mnext = &m->m_nextpkt;
857 	}
858 	IPSTAT_ADD(ips_ofragments, nfrags);
859 
860 	/* set first marker for fragment chain */
861 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
862 	m0->m_pkthdr.csum_data = nfrags;
863 
864 	/*
865 	 * Update first fragment by trimming what's been copied out
866 	 * and updating header.
867 	 */
868 	m_adj(m0, hlen + firstlen - ip_len);
869 	m0->m_pkthdr.len = hlen + firstlen;
870 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
871 	ip->ip_off = htons(ip_off | IP_MF);
872 	ip->ip_sum = 0;
873 	if (sw_csum & CSUM_DELAY_IP)
874 		ip->ip_sum = in_cksum(m0, hlen);
875 
876 done:
877 	*m_frag = m0;
878 	return error;
879 }
880 
881 void
882 in_delayed_cksum(struct mbuf *m)
883 {
884 	struct ip *ip;
885 	uint16_t csum, offset, ip_len;
886 
887 	ip = mtod(m, struct ip *);
888 	offset = ip->ip_hl << 2 ;
889 	ip_len = ntohs(ip->ip_len);
890 	csum = in_cksum_skip(m, ip_len, offset);
891 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
892 		csum = 0xffff;
893 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
894 
895 	if (offset + sizeof(u_short) > m->m_len) {
896 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
897 		    m->m_len, offset, ip->ip_p);
898 		/*
899 		 * XXX
900 		 * this shouldn't happen, but if it does, the
901 		 * correct behavior may be to insert the checksum
902 		 * in the appropriate next mbuf in the chain.
903 		 */
904 		return;
905 	}
906 	*(u_short *)(m->m_data + offset) = csum;
907 }
908 
909 /*
910  * IP socket option processing.
911  */
912 int
913 ip_ctloutput(struct socket *so, struct sockopt *sopt)
914 {
915 	struct	inpcb *inp = sotoinpcb(so);
916 	int	error, optval;
917 
918 	error = optval = 0;
919 	if (sopt->sopt_level != IPPROTO_IP) {
920 		error = EINVAL;
921 
922 		if (sopt->sopt_level == SOL_SOCKET &&
923 		    sopt->sopt_dir == SOPT_SET) {
924 			switch (sopt->sopt_name) {
925 			case SO_REUSEADDR:
926 				INP_WLOCK(inp);
927 				if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
928 					if ((so->so_options &
929 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
930 						inp->inp_flags2 |= INP_REUSEPORT;
931 					else
932 						inp->inp_flags2 &= ~INP_REUSEPORT;
933 				}
934 				INP_WUNLOCK(inp);
935 				error = 0;
936 				break;
937 			case SO_REUSEPORT:
938 				INP_WLOCK(inp);
939 				if ((so->so_options & SO_REUSEPORT) != 0)
940 					inp->inp_flags2 |= INP_REUSEPORT;
941 				else
942 					inp->inp_flags2 &= ~INP_REUSEPORT;
943 				INP_WUNLOCK(inp);
944 				error = 0;
945 				break;
946 			case SO_SETFIB:
947 				INP_WLOCK(inp);
948 				inp->inp_inc.inc_fibnum = so->so_fibnum;
949 				INP_WUNLOCK(inp);
950 				error = 0;
951 				break;
952 			default:
953 				break;
954 			}
955 		}
956 		return (error);
957 	}
958 
959 	switch (sopt->sopt_dir) {
960 	case SOPT_SET:
961 		switch (sopt->sopt_name) {
962 		case IP_OPTIONS:
963 #ifdef notyet
964 		case IP_RETOPTS:
965 #endif
966 		{
967 			struct mbuf *m;
968 			if (sopt->sopt_valsize > MLEN) {
969 				error = EMSGSIZE;
970 				break;
971 			}
972 			MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
973 			if (m == NULL) {
974 				error = ENOBUFS;
975 				break;
976 			}
977 			m->m_len = sopt->sopt_valsize;
978 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
979 					    m->m_len);
980 			if (error) {
981 				m_free(m);
982 				break;
983 			}
984 			INP_WLOCK(inp);
985 			error = ip_pcbopts(inp, sopt->sopt_name, m);
986 			INP_WUNLOCK(inp);
987 			return (error);
988 		}
989 
990 		case IP_BINDANY:
991 			if (sopt->sopt_td != NULL) {
992 				error = priv_check(sopt->sopt_td,
993 				    PRIV_NETINET_BINDANY);
994 				if (error)
995 					break;
996 			}
997 			/* FALLTHROUGH */
998 		case IP_TOS:
999 		case IP_TTL:
1000 		case IP_MINTTL:
1001 		case IP_RECVOPTS:
1002 		case IP_RECVRETOPTS:
1003 		case IP_RECVDSTADDR:
1004 		case IP_RECVTTL:
1005 		case IP_RECVIF:
1006 		case IP_FAITH:
1007 		case IP_ONESBCAST:
1008 		case IP_DONTFRAG:
1009 		case IP_RECVTOS:
1010 			error = sooptcopyin(sopt, &optval, sizeof optval,
1011 					    sizeof optval);
1012 			if (error)
1013 				break;
1014 
1015 			switch (sopt->sopt_name) {
1016 			case IP_TOS:
1017 				inp->inp_ip_tos = optval;
1018 				break;
1019 
1020 			case IP_TTL:
1021 				inp->inp_ip_ttl = optval;
1022 				break;
1023 
1024 			case IP_MINTTL:
1025 				if (optval >= 0 && optval <= MAXTTL)
1026 					inp->inp_ip_minttl = optval;
1027 				else
1028 					error = EINVAL;
1029 				break;
1030 
1031 #define	OPTSET(bit) do {						\
1032 	INP_WLOCK(inp);							\
1033 	if (optval)							\
1034 		inp->inp_flags |= bit;					\
1035 	else								\
1036 		inp->inp_flags &= ~bit;					\
1037 	INP_WUNLOCK(inp);						\
1038 } while (0)
1039 
1040 			case IP_RECVOPTS:
1041 				OPTSET(INP_RECVOPTS);
1042 				break;
1043 
1044 			case IP_RECVRETOPTS:
1045 				OPTSET(INP_RECVRETOPTS);
1046 				break;
1047 
1048 			case IP_RECVDSTADDR:
1049 				OPTSET(INP_RECVDSTADDR);
1050 				break;
1051 
1052 			case IP_RECVTTL:
1053 				OPTSET(INP_RECVTTL);
1054 				break;
1055 
1056 			case IP_RECVIF:
1057 				OPTSET(INP_RECVIF);
1058 				break;
1059 
1060 			case IP_FAITH:
1061 				OPTSET(INP_FAITH);
1062 				break;
1063 
1064 			case IP_ONESBCAST:
1065 				OPTSET(INP_ONESBCAST);
1066 				break;
1067 			case IP_DONTFRAG:
1068 				OPTSET(INP_DONTFRAG);
1069 				break;
1070 			case IP_BINDANY:
1071 				OPTSET(INP_BINDANY);
1072 				break;
1073 			case IP_RECVTOS:
1074 				OPTSET(INP_RECVTOS);
1075 				break;
1076 			}
1077 			break;
1078 #undef OPTSET
1079 
1080 		/*
1081 		 * Multicast socket options are processed by the in_mcast
1082 		 * module.
1083 		 */
1084 		case IP_MULTICAST_IF:
1085 		case IP_MULTICAST_VIF:
1086 		case IP_MULTICAST_TTL:
1087 		case IP_MULTICAST_LOOP:
1088 		case IP_ADD_MEMBERSHIP:
1089 		case IP_DROP_MEMBERSHIP:
1090 		case IP_ADD_SOURCE_MEMBERSHIP:
1091 		case IP_DROP_SOURCE_MEMBERSHIP:
1092 		case IP_BLOCK_SOURCE:
1093 		case IP_UNBLOCK_SOURCE:
1094 		case IP_MSFILTER:
1095 		case MCAST_JOIN_GROUP:
1096 		case MCAST_LEAVE_GROUP:
1097 		case MCAST_JOIN_SOURCE_GROUP:
1098 		case MCAST_LEAVE_SOURCE_GROUP:
1099 		case MCAST_BLOCK_SOURCE:
1100 		case MCAST_UNBLOCK_SOURCE:
1101 			error = inp_setmoptions(inp, sopt);
1102 			break;
1103 
1104 		case IP_PORTRANGE:
1105 			error = sooptcopyin(sopt, &optval, sizeof optval,
1106 					    sizeof optval);
1107 			if (error)
1108 				break;
1109 
1110 			INP_WLOCK(inp);
1111 			switch (optval) {
1112 			case IP_PORTRANGE_DEFAULT:
1113 				inp->inp_flags &= ~(INP_LOWPORT);
1114 				inp->inp_flags &= ~(INP_HIGHPORT);
1115 				break;
1116 
1117 			case IP_PORTRANGE_HIGH:
1118 				inp->inp_flags &= ~(INP_LOWPORT);
1119 				inp->inp_flags |= INP_HIGHPORT;
1120 				break;
1121 
1122 			case IP_PORTRANGE_LOW:
1123 				inp->inp_flags &= ~(INP_HIGHPORT);
1124 				inp->inp_flags |= INP_LOWPORT;
1125 				break;
1126 
1127 			default:
1128 				error = EINVAL;
1129 				break;
1130 			}
1131 			INP_WUNLOCK(inp);
1132 			break;
1133 
1134 #ifdef IPSEC
1135 		case IP_IPSEC_POLICY:
1136 		{
1137 			caddr_t req;
1138 			struct mbuf *m;
1139 
1140 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1141 				break;
1142 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1143 				break;
1144 			req = mtod(m, caddr_t);
1145 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1146 			    m->m_len, (sopt->sopt_td != NULL) ?
1147 			    sopt->sopt_td->td_ucred : NULL);
1148 			m_freem(m);
1149 			break;
1150 		}
1151 #endif /* IPSEC */
1152 
1153 		default:
1154 			error = ENOPROTOOPT;
1155 			break;
1156 		}
1157 		break;
1158 
1159 	case SOPT_GET:
1160 		switch (sopt->sopt_name) {
1161 		case IP_OPTIONS:
1162 		case IP_RETOPTS:
1163 			if (inp->inp_options)
1164 				error = sooptcopyout(sopt,
1165 						     mtod(inp->inp_options,
1166 							  char *),
1167 						     inp->inp_options->m_len);
1168 			else
1169 				sopt->sopt_valsize = 0;
1170 			break;
1171 
1172 		case IP_TOS:
1173 		case IP_TTL:
1174 		case IP_MINTTL:
1175 		case IP_RECVOPTS:
1176 		case IP_RECVRETOPTS:
1177 		case IP_RECVDSTADDR:
1178 		case IP_RECVTTL:
1179 		case IP_RECVIF:
1180 		case IP_PORTRANGE:
1181 		case IP_FAITH:
1182 		case IP_ONESBCAST:
1183 		case IP_DONTFRAG:
1184 		case IP_BINDANY:
1185 		case IP_RECVTOS:
1186 			switch (sopt->sopt_name) {
1187 
1188 			case IP_TOS:
1189 				optval = inp->inp_ip_tos;
1190 				break;
1191 
1192 			case IP_TTL:
1193 				optval = inp->inp_ip_ttl;
1194 				break;
1195 
1196 			case IP_MINTTL:
1197 				optval = inp->inp_ip_minttl;
1198 				break;
1199 
1200 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1201 
1202 			case IP_RECVOPTS:
1203 				optval = OPTBIT(INP_RECVOPTS);
1204 				break;
1205 
1206 			case IP_RECVRETOPTS:
1207 				optval = OPTBIT(INP_RECVRETOPTS);
1208 				break;
1209 
1210 			case IP_RECVDSTADDR:
1211 				optval = OPTBIT(INP_RECVDSTADDR);
1212 				break;
1213 
1214 			case IP_RECVTTL:
1215 				optval = OPTBIT(INP_RECVTTL);
1216 				break;
1217 
1218 			case IP_RECVIF:
1219 				optval = OPTBIT(INP_RECVIF);
1220 				break;
1221 
1222 			case IP_PORTRANGE:
1223 				if (inp->inp_flags & INP_HIGHPORT)
1224 					optval = IP_PORTRANGE_HIGH;
1225 				else if (inp->inp_flags & INP_LOWPORT)
1226 					optval = IP_PORTRANGE_LOW;
1227 				else
1228 					optval = 0;
1229 				break;
1230 
1231 			case IP_FAITH:
1232 				optval = OPTBIT(INP_FAITH);
1233 				break;
1234 
1235 			case IP_ONESBCAST:
1236 				optval = OPTBIT(INP_ONESBCAST);
1237 				break;
1238 			case IP_DONTFRAG:
1239 				optval = OPTBIT(INP_DONTFRAG);
1240 				break;
1241 			case IP_BINDANY:
1242 				optval = OPTBIT(INP_BINDANY);
1243 				break;
1244 			case IP_RECVTOS:
1245 				optval = OPTBIT(INP_RECVTOS);
1246 				break;
1247 			}
1248 			error = sooptcopyout(sopt, &optval, sizeof optval);
1249 			break;
1250 
1251 		/*
1252 		 * Multicast socket options are processed by the in_mcast
1253 		 * module.
1254 		 */
1255 		case IP_MULTICAST_IF:
1256 		case IP_MULTICAST_VIF:
1257 		case IP_MULTICAST_TTL:
1258 		case IP_MULTICAST_LOOP:
1259 		case IP_MSFILTER:
1260 			error = inp_getmoptions(inp, sopt);
1261 			break;
1262 
1263 #ifdef IPSEC
1264 		case IP_IPSEC_POLICY:
1265 		{
1266 			struct mbuf *m = NULL;
1267 			caddr_t req = NULL;
1268 			size_t len = 0;
1269 
1270 			if (m != 0) {
1271 				req = mtod(m, caddr_t);
1272 				len = m->m_len;
1273 			}
1274 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1275 			if (error == 0)
1276 				error = soopt_mcopyout(sopt, m); /* XXX */
1277 			if (error == 0)
1278 				m_freem(m);
1279 			break;
1280 		}
1281 #endif /* IPSEC */
1282 
1283 		default:
1284 			error = ENOPROTOOPT;
1285 			break;
1286 		}
1287 		break;
1288 	}
1289 	return (error);
1290 }
1291 
1292 /*
1293  * Routine called from ip_output() to loop back a copy of an IP multicast
1294  * packet to the input queue of a specified interface.  Note that this
1295  * calls the output routine of the loopback "driver", but with an interface
1296  * pointer that might NOT be a loopback interface -- evil, but easier than
1297  * replicating that code here.
1298  *
1299  * IP header in host byte order.
1300  */
1301 static void
1302 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1303     int hlen)
1304 {
1305 	register struct ip *ip;
1306 	struct mbuf *copym;
1307 
1308 	/*
1309 	 * Make a deep copy of the packet because we're going to
1310 	 * modify the pack in order to generate checksums.
1311 	 */
1312 	copym = m_dup(m, M_DONTWAIT);
1313 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1314 		copym = m_pullup(copym, hlen);
1315 	if (copym != NULL) {
1316 		/* If needed, compute the checksum and mark it as valid. */
1317 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1318 			in_delayed_cksum(copym);
1319 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1320 			copym->m_pkthdr.csum_flags |=
1321 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1322 			copym->m_pkthdr.csum_data = 0xffff;
1323 		}
1324 		/*
1325 		 * We don't bother to fragment if the IP length is greater
1326 		 * than the interface's MTU.  Can this possibly matter?
1327 		 */
1328 		ip = mtod(copym, struct ip *);
1329 		ip->ip_len = htons(ip->ip_len);
1330 		ip->ip_off = htons(ip->ip_off);
1331 		ip->ip_sum = 0;
1332 		ip->ip_sum = in_cksum(copym, hlen);
1333 #if 1 /* XXX */
1334 		if (dst->sin_family != AF_INET) {
1335 			printf("ip_mloopback: bad address family %d\n",
1336 						dst->sin_family);
1337 			dst->sin_family = AF_INET;
1338 		}
1339 #endif
1340 		if_simloop(ifp, copym, dst->sin_family, 0);
1341 	}
1342 }
1343