xref: /freebsd/sys/netinet/ip_output.c (revision f5f47d5068fb97df18eb114a66ae8ef51a0b3c8c)
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, sw_csum;
129 	struct route iproute;
130 	struct rtentry *rte;	/* cache for ro->ro_rt */
131 	struct in_addr odst;
132 #ifdef IPFIREWALL_FORWARD
133 	struct m_tag *fwd_tag = NULL;
134 #endif
135 #ifdef IPSEC
136 	int no_route_but_check_spd = 0;
137 #endif
138 	M_ASSERTPKTHDR(m);
139 
140 	if (inp != NULL) {
141 		INP_LOCK_ASSERT(inp);
142 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
143 		if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
144 			m->m_pkthdr.flowid = inp->inp_flowid;
145 			m->m_flags |= M_FLOWID;
146 		}
147 	}
148 
149 	if (ro == NULL) {
150 		ro = &iproute;
151 		bzero(ro, sizeof (*ro));
152 	}
153 
154 #ifdef FLOWTABLE
155 	if (ro->ro_rt == NULL) {
156 		struct flentry *fle;
157 
158 		/*
159 		 * The flow table returns route entries valid for up to 30
160 		 * seconds; we rely on the remainder of ip_output() taking no
161 		 * longer than that long for the stability of ro_rt. The
162 		 * flow ID assignment must have happened before this point.
163 		 */
164 		fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET);
165 		if (fle != NULL)
166 			flow_to_route(fle, ro);
167 	}
168 #endif
169 
170 	if (opt) {
171 		int len = 0;
172 		m = ip_insertoptions(m, opt, &len);
173 		if (len != 0)
174 			hlen = len; /* ip->ip_hl is updated above */
175 	}
176 	ip = mtod(m, struct ip *);
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 #ifdef IPFIREWALL_FORWARD
220 	if (rte == NULL && fwd_tag == NULL) {
221 #else
222 	if (rte == NULL) {
223 #endif
224 		bzero(dst, sizeof(*dst));
225 		dst->sin_family = AF_INET;
226 		dst->sin_len = sizeof(*dst);
227 		dst->sin_addr = ip->ip_dst;
228 	}
229 	/*
230 	 * If routing to interface only, short circuit routing lookup.
231 	 * The use of an all-ones broadcast address implies this; an
232 	 * interface is specified by the broadcast address of an interface,
233 	 * or the destination address of a ptp interface.
234 	 */
235 	if (flags & IP_SENDONES) {
236 		if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL &&
237 		    (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) {
238 			IPSTAT_INC(ips_noroute);
239 			error = ENETUNREACH;
240 			goto bad;
241 		}
242 		ip->ip_dst.s_addr = INADDR_BROADCAST;
243 		dst->sin_addr = ip->ip_dst;
244 		ifp = ia->ia_ifp;
245 		ip->ip_ttl = 1;
246 		isbroadcast = 1;
247 	} else if (flags & IP_ROUTETOIF) {
248 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
249 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) {
250 			IPSTAT_INC(ips_noroute);
251 			error = ENETUNREACH;
252 			goto bad;
253 		}
254 		ifp = ia->ia_ifp;
255 		ip->ip_ttl = 1;
256 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
257 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
258 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
259 		/*
260 		 * Bypass the normal routing lookup for multicast
261 		 * packets if the interface is specified.
262 		 */
263 		ifp = imo->imo_multicast_ifp;
264 		IFP_TO_IA(ifp, ia);
265 		isbroadcast = 0;	/* fool gcc */
266 	} else {
267 		/*
268 		 * We want to do any cloning requested by the link layer,
269 		 * as this is probably required in all cases for correct
270 		 * operation (as it is for ARP).
271 		 */
272 		if (rte == NULL) {
273 #ifdef RADIX_MPATH
274 			rtalloc_mpath_fib(ro,
275 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
276 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
277 #else
278 			in_rtalloc_ign(ro, 0,
279 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
280 #endif
281 			rte = ro->ro_rt;
282 		}
283 		if (rte == NULL ||
284 		    rte->rt_ifp == NULL ||
285 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
286 #ifdef IPSEC
287 			/*
288 			 * There is no route for this packet, but it is
289 			 * possible that a matching SPD entry exists.
290 			 */
291 			no_route_but_check_spd = 1;
292 			mtu = 0; /* Silence GCC warning. */
293 			goto sendit;
294 #endif
295 			IPSTAT_INC(ips_noroute);
296 			error = EHOSTUNREACH;
297 			goto bad;
298 		}
299 		ia = ifatoia(rte->rt_ifa);
300 		ifa_ref(&ia->ia_ifa);
301 		ifp = rte->rt_ifp;
302 		rte->rt_rmx.rmx_pksent++;
303 		if (rte->rt_flags & RTF_GATEWAY)
304 			dst = (struct sockaddr_in *)rte->rt_gateway;
305 		if (rte->rt_flags & RTF_HOST)
306 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
307 		else
308 			isbroadcast = in_broadcast(dst->sin_addr, ifp);
309 	}
310 	/*
311 	 * Calculate MTU.  If we have a route that is up, use that,
312 	 * otherwise use the interface's MTU.
313 	 */
314 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
315 		/*
316 		 * This case can happen if the user changed the MTU
317 		 * of an interface after enabling IP on it.  Because
318 		 * most netifs don't keep track of routes pointing to
319 		 * them, there is no way for one to update all its
320 		 * routes when the MTU is changed.
321 		 */
322 		if (rte->rt_rmx.rmx_mtu > ifp->if_mtu)
323 			rte->rt_rmx.rmx_mtu = ifp->if_mtu;
324 		mtu = rte->rt_rmx.rmx_mtu;
325 	} else {
326 		mtu = ifp->if_mtu;
327 	}
328 	/* Catch a possible divide by zero later. */
329 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
330 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
331 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
332 		m->m_flags |= M_MCAST;
333 		/*
334 		 * IP destination address is multicast.  Make sure "dst"
335 		 * still points to the address in "ro".  (It may have been
336 		 * changed to point to a gateway address, above.)
337 		 */
338 		dst = (struct sockaddr_in *)&ro->ro_dst;
339 		/*
340 		 * See if the caller provided any multicast options
341 		 */
342 		if (imo != NULL) {
343 			ip->ip_ttl = imo->imo_multicast_ttl;
344 			if (imo->imo_multicast_vif != -1)
345 				ip->ip_src.s_addr =
346 				    ip_mcast_src ?
347 				    ip_mcast_src(imo->imo_multicast_vif) :
348 				    INADDR_ANY;
349 		} else
350 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
351 		/*
352 		 * Confirm that the outgoing interface supports multicast.
353 		 */
354 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
355 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
356 				IPSTAT_INC(ips_noroute);
357 				error = ENETUNREACH;
358 				goto bad;
359 			}
360 		}
361 		/*
362 		 * If source address not specified yet, use address
363 		 * of outgoing interface.
364 		 */
365 		if (ip->ip_src.s_addr == INADDR_ANY) {
366 			/* Interface may have no addresses. */
367 			if (ia != NULL)
368 				ip->ip_src = IA_SIN(ia)->sin_addr;
369 		}
370 
371 		if ((imo == NULL && in_mcast_loop) ||
372 		    (imo && imo->imo_multicast_loop)) {
373 			/*
374 			 * Loop back multicast datagram if not expressly
375 			 * forbidden to do so, even if we are not a member
376 			 * of the group; ip_input() will filter it later,
377 			 * thus deferring a hash lookup and mutex acquisition
378 			 * at the expense of a cheap copy using m_copym().
379 			 */
380 			ip_mloopback(ifp, m, dst, hlen);
381 		} else {
382 			/*
383 			 * If we are acting as a multicast router, perform
384 			 * multicast forwarding as if the packet had just
385 			 * arrived on the interface to which we are about
386 			 * to send.  The multicast forwarding function
387 			 * recursively calls this function, using the
388 			 * IP_FORWARDING flag to prevent infinite recursion.
389 			 *
390 			 * Multicasts that are looped back by ip_mloopback(),
391 			 * above, will be forwarded by the ip_input() routine,
392 			 * if necessary.
393 			 */
394 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
395 				/*
396 				 * If rsvp daemon is not running, do not
397 				 * set ip_moptions. This ensures that the packet
398 				 * is multicast and not just sent down one link
399 				 * as prescribed by rsvpd.
400 				 */
401 				if (!V_rsvp_on)
402 					imo = NULL;
403 				if (ip_mforward &&
404 				    ip_mforward(ip, ifp, m, imo) != 0) {
405 					m_freem(m);
406 					goto done;
407 				}
408 			}
409 		}
410 
411 		/*
412 		 * Multicasts with a time-to-live of zero may be looped-
413 		 * back, above, but must not be transmitted on a network.
414 		 * Also, multicasts addressed to the loopback interface
415 		 * are not sent -- the above call to ip_mloopback() will
416 		 * loop back a copy. ip_input() will drop the copy if
417 		 * this host does not belong to the destination group on
418 		 * the loopback interface.
419 		 */
420 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
421 			m_freem(m);
422 			goto done;
423 		}
424 
425 		goto sendit;
426 	}
427 
428 	/*
429 	 * If the source address is not specified yet, use the address
430 	 * of the outoing interface.
431 	 */
432 	if (ip->ip_src.s_addr == INADDR_ANY) {
433 		/* Interface may have no addresses. */
434 		if (ia != NULL) {
435 			ip->ip_src = IA_SIN(ia)->sin_addr;
436 		}
437 	}
438 
439 	/*
440 	 * Verify that we have any chance at all of being able to queue the
441 	 * packet or packet fragments, unless ALTQ is enabled on the given
442 	 * interface in which case packetdrop should be done by queueing.
443 	 */
444 	n = ip->ip_len / mtu + 1; /* how many fragments ? */
445 	if (
446 #ifdef ALTQ
447 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
448 #endif /* ALTQ */
449 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
450 		error = ENOBUFS;
451 		IPSTAT_INC(ips_odropped);
452 		ifp->if_snd.ifq_drops += n;
453 		goto bad;
454 	}
455 
456 	/*
457 	 * Look for broadcast address and
458 	 * verify user is allowed to send
459 	 * such a packet.
460 	 */
461 	if (isbroadcast) {
462 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
463 			error = EADDRNOTAVAIL;
464 			goto bad;
465 		}
466 		if ((flags & IP_ALLOWBROADCAST) == 0) {
467 			error = EACCES;
468 			goto bad;
469 		}
470 		/* don't allow broadcast messages to be fragmented */
471 		if (ip->ip_len > mtu) {
472 			error = EMSGSIZE;
473 			goto bad;
474 		}
475 		m->m_flags |= M_BCAST;
476 	} else {
477 		m->m_flags &= ~M_BCAST;
478 	}
479 
480 sendit:
481 #ifdef IPSEC
482 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
483 	case 1:
484 		goto bad;
485 	case -1:
486 		goto done;
487 	case 0:
488 	default:
489 		break;	/* Continue with packet processing. */
490 	}
491 	/*
492 	 * Check if there was a route for this packet; return error if not.
493 	 */
494 	if (no_route_but_check_spd) {
495 		IPSTAT_INC(ips_noroute);
496 		error = EHOSTUNREACH;
497 		goto bad;
498 	}
499 	/* Update variables that are affected by ipsec4_output(). */
500 	ip = mtod(m, struct ip *);
501 	hlen = ip->ip_hl << 2;
502 #endif /* IPSEC */
503 
504 	/* Jump over all PFIL processing if hooks are not active. */
505 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
506 		goto passout;
507 
508 	/* Run through list of hooks for output packets. */
509 	odst.s_addr = ip->ip_dst.s_addr;
510 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
511 	if (error != 0 || m == NULL)
512 		goto done;
513 
514 	ip = mtod(m, struct ip *);
515 
516 	/* See if destination IP address was changed by packet filter. */
517 	if (odst.s_addr != ip->ip_dst.s_addr) {
518 		m->m_flags |= M_SKIP_FIREWALL;
519 		/* If destination is now ourself drop to ip_input(). */
520 		if (in_localip(ip->ip_dst)) {
521 			m->m_flags |= M_FASTFWD_OURS;
522 			if (m->m_pkthdr.rcvif == NULL)
523 				m->m_pkthdr.rcvif = V_loif;
524 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
525 				m->m_pkthdr.csum_flags |=
526 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
527 				m->m_pkthdr.csum_data = 0xffff;
528 			}
529 			m->m_pkthdr.csum_flags |=
530 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
531 #ifdef SCTP
532 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
533 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
534 #endif
535 			error = netisr_queue(NETISR_IP, m);
536 			goto done;
537 		} else {
538 			if (ia != NULL)
539 				ifa_free(&ia->ia_ifa);
540 			goto again;	/* Redo the routing table lookup. */
541 		}
542 	}
543 
544 #ifdef IPFIREWALL_FORWARD
545 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
546 	if (m->m_flags & M_FASTFWD_OURS) {
547 		if (m->m_pkthdr.rcvif == NULL)
548 			m->m_pkthdr.rcvif = V_loif;
549 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
550 			m->m_pkthdr.csum_flags |=
551 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
552 			m->m_pkthdr.csum_data = 0xffff;
553 		}
554 #ifdef SCTP
555 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
556 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
557 #endif
558 		m->m_pkthdr.csum_flags |=
559 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
560 
561 		error = netisr_queue(NETISR_IP, m);
562 		goto done;
563 	}
564 	/* Or forward to some other address? */
565 	fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
566 	if (fwd_tag) {
567 		dst = (struct sockaddr_in *)&ro->ro_dst;
568 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
569 		m->m_flags |= M_SKIP_FIREWALL;
570 		m_tag_delete(m, fwd_tag);
571 		if (ia != NULL)
572 			ifa_free(&ia->ia_ifa);
573 		goto again;
574 	}
575 #endif /* IPFIREWALL_FORWARD */
576 
577 passout:
578 	/* 127/8 must not appear on wire - RFC1122. */
579 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
580 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
581 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
582 			IPSTAT_INC(ips_badaddr);
583 			error = EADDRNOTAVAIL;
584 			goto bad;
585 		}
586 	}
587 
588 	m->m_pkthdr.csum_flags |= CSUM_IP;
589 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
590 	if (sw_csum & CSUM_DELAY_DATA) {
591 		in_delayed_cksum(m);
592 		sw_csum &= ~CSUM_DELAY_DATA;
593 	}
594 #ifdef SCTP
595 	if (sw_csum & CSUM_SCTP) {
596 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
597 		sw_csum &= ~CSUM_SCTP;
598 	}
599 #endif
600 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
601 
602 	/*
603 	 * If small enough for interface, or the interface will take
604 	 * care of the fragmentation for us, we can just send directly.
605 	 */
606 	if (ip->ip_len <= mtu ||
607 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
608 	    ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
609 		ip->ip_len = htons(ip->ip_len);
610 		ip->ip_off = htons(ip->ip_off);
611 		ip->ip_sum = 0;
612 		if (sw_csum & CSUM_DELAY_IP)
613 			ip->ip_sum = in_cksum(m, hlen);
614 
615 		/*
616 		 * Record statistics for this interface address.
617 		 * With CSUM_TSO the byte/packet count will be slightly
618 		 * incorrect because we count the IP+TCP headers only
619 		 * once instead of for every generated packet.
620 		 */
621 		if (!(flags & IP_FORWARDING) && ia) {
622 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
623 				ia->ia_ifa.if_opackets +=
624 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz;
625 			else
626 				ia->ia_ifa.if_opackets++;
627 			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
628 		}
629 #ifdef MBUF_STRESS_TEST
630 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
631 			m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
632 #endif
633 		/*
634 		 * Reset layer specific mbuf flags
635 		 * to avoid confusing lower layers.
636 		 */
637 		m->m_flags &= ~(M_PROTOFLAGS);
638 		error = (*ifp->if_output)(ifp, m,
639 		    		(struct sockaddr *)dst, ro);
640 		goto done;
641 	}
642 
643 	/* Balk when DF bit is set or the interface didn't support TSO. */
644 	if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
645 		error = EMSGSIZE;
646 		IPSTAT_INC(ips_cantfrag);
647 		goto bad;
648 	}
649 
650 	/*
651 	 * Too large for interface; fragment if possible. If successful,
652 	 * on return, m will point to a list of packets to be sent.
653 	 */
654 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum);
655 	if (error)
656 		goto bad;
657 	for (; m; m = m0) {
658 		m0 = m->m_nextpkt;
659 		m->m_nextpkt = 0;
660 		if (error == 0) {
661 			/* Record statistics for this interface address. */
662 			if (ia != NULL) {
663 				ia->ia_ifa.if_opackets++;
664 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
665 			}
666 			/*
667 			 * Reset layer specific mbuf flags
668 			 * to avoid confusing upper layers.
669 			 */
670 			m->m_flags &= ~(M_PROTOFLAGS);
671 
672 			error = (*ifp->if_output)(ifp, m,
673 			    (struct sockaddr *)dst, ro);
674 		} else
675 			m_freem(m);
676 	}
677 
678 	if (error == 0)
679 		IPSTAT_INC(ips_fragmented);
680 
681 done:
682 	if (ro == &iproute)
683 		RO_RTFREE(ro);
684 	if (ia != NULL)
685 		ifa_free(&ia->ia_ifa);
686 	return (error);
687 bad:
688 	m_freem(m);
689 	goto done;
690 }
691 
692 /*
693  * Create a chain of fragments which fit the given mtu. m_frag points to the
694  * mbuf to be fragmented; on return it points to the chain with the fragments.
695  * Return 0 if no error. If error, m_frag may contain a partially built
696  * chain of fragments that should be freed by the caller.
697  *
698  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
699  * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
700  */
701 int
702 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
703     u_long if_hwassist_flags, int sw_csum)
704 {
705 	int error = 0;
706 	int hlen = ip->ip_hl << 2;
707 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
708 	int off;
709 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
710 	int firstlen;
711 	struct mbuf **mnext;
712 	int nfrags;
713 
714 	if (ip->ip_off & IP_DF) {	/* Fragmentation not allowed */
715 		IPSTAT_INC(ips_cantfrag);
716 		return EMSGSIZE;
717 	}
718 
719 	/*
720 	 * Must be able to put at least 8 bytes per fragment.
721 	 */
722 	if (len < 8)
723 		return EMSGSIZE;
724 
725 	/*
726 	 * If the interface will not calculate checksums on
727 	 * fragmented packets, then do it here.
728 	 */
729 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
730 	    (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
731 		in_delayed_cksum(m0);
732 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
733 	}
734 #ifdef SCTP
735 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP &&
736 	    (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
737 		sctp_delayed_cksum(m0, hlen);
738 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
739 	}
740 #endif
741 	if (len > PAGE_SIZE) {
742 		/*
743 		 * Fragment large datagrams such that each segment
744 		 * contains a multiple of PAGE_SIZE amount of data,
745 		 * plus headers. This enables a receiver to perform
746 		 * page-flipping zero-copy optimizations.
747 		 *
748 		 * XXX When does this help given that sender and receiver
749 		 * could have different page sizes, and also mtu could
750 		 * be less than the receiver's page size ?
751 		 */
752 		int newlen;
753 		struct mbuf *m;
754 
755 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
756 			off += m->m_len;
757 
758 		/*
759 		 * firstlen (off - hlen) must be aligned on an
760 		 * 8-byte boundary
761 		 */
762 		if (off < hlen)
763 			goto smart_frag_failure;
764 		off = ((off - hlen) & ~7) + hlen;
765 		newlen = (~PAGE_MASK) & mtu;
766 		if ((newlen + sizeof (struct ip)) > mtu) {
767 			/* we failed, go back the default */
768 smart_frag_failure:
769 			newlen = len;
770 			off = hlen + len;
771 		}
772 		len = newlen;
773 
774 	} else {
775 		off = hlen + len;
776 	}
777 
778 	firstlen = off - hlen;
779 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
780 
781 	/*
782 	 * Loop through length of segment after first fragment,
783 	 * make new header and copy data of each part and link onto chain.
784 	 * Here, m0 is the original packet, m is the fragment being created.
785 	 * The fragments are linked off the m_nextpkt of the original
786 	 * packet, which after processing serves as the first fragment.
787 	 */
788 	for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
789 		struct ip *mhip;	/* ip header on the fragment */
790 		struct mbuf *m;
791 		int mhlen = sizeof (struct ip);
792 
793 		MGETHDR(m, M_DONTWAIT, MT_DATA);
794 		if (m == NULL) {
795 			error = ENOBUFS;
796 			IPSTAT_INC(ips_odropped);
797 			goto done;
798 		}
799 		m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
800 		/*
801 		 * In the first mbuf, leave room for the link header, then
802 		 * copy the original IP header including options. The payload
803 		 * goes into an additional mbuf chain returned by m_copym().
804 		 */
805 		m->m_data += max_linkhdr;
806 		mhip = mtod(m, struct ip *);
807 		*mhip = *ip;
808 		if (hlen > sizeof (struct ip)) {
809 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
810 			mhip->ip_v = IPVERSION;
811 			mhip->ip_hl = mhlen >> 2;
812 		}
813 		m->m_len = mhlen;
814 		/* XXX do we need to add ip->ip_off below ? */
815 		mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
816 		if (off + len >= ip->ip_len) {	/* last fragment */
817 			len = ip->ip_len - off;
818 			m->m_flags |= M_LASTFRAG;
819 		} else
820 			mhip->ip_off |= IP_MF;
821 		mhip->ip_len = htons((u_short)(len + mhlen));
822 		m->m_next = m_copym(m0, off, len, M_DONTWAIT);
823 		if (m->m_next == NULL) {	/* copy failed */
824 			m_free(m);
825 			error = ENOBUFS;	/* ??? */
826 			IPSTAT_INC(ips_odropped);
827 			goto done;
828 		}
829 		m->m_pkthdr.len = mhlen + len;
830 		m->m_pkthdr.rcvif = NULL;
831 #ifdef MAC
832 		mac_netinet_fragment(m0, m);
833 #endif
834 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
835 		mhip->ip_off = htons(mhip->ip_off);
836 		mhip->ip_sum = 0;
837 		if (sw_csum & CSUM_DELAY_IP)
838 			mhip->ip_sum = in_cksum(m, mhlen);
839 		*mnext = m;
840 		mnext = &m->m_nextpkt;
841 	}
842 	IPSTAT_ADD(ips_ofragments, nfrags);
843 
844 	/* set first marker for fragment chain */
845 	m0->m_flags |= M_FIRSTFRAG | M_FRAG;
846 	m0->m_pkthdr.csum_data = nfrags;
847 
848 	/*
849 	 * Update first fragment by trimming what's been copied out
850 	 * and updating header.
851 	 */
852 	m_adj(m0, hlen + firstlen - ip->ip_len);
853 	m0->m_pkthdr.len = hlen + firstlen;
854 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
855 	ip->ip_off |= IP_MF;
856 	ip->ip_off = htons(ip->ip_off);
857 	ip->ip_sum = 0;
858 	if (sw_csum & CSUM_DELAY_IP)
859 		ip->ip_sum = in_cksum(m0, hlen);
860 
861 done:
862 	*m_frag = m0;
863 	return error;
864 }
865 
866 void
867 in_delayed_cksum(struct mbuf *m)
868 {
869 	struct ip *ip;
870 	u_short csum, offset;
871 
872 	ip = mtod(m, struct ip *);
873 	offset = ip->ip_hl << 2 ;
874 	csum = in_cksum_skip(m, ip->ip_len, offset);
875 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
876 		csum = 0xffff;
877 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
878 
879 	if (offset + sizeof(u_short) > m->m_len) {
880 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
881 		    m->m_len, offset, ip->ip_p);
882 		/*
883 		 * XXX
884 		 * this shouldn't happen, but if it does, the
885 		 * correct behavior may be to insert the checksum
886 		 * in the appropriate next mbuf in the chain.
887 		 */
888 		return;
889 	}
890 	*(u_short *)(m->m_data + offset) = csum;
891 }
892 
893 /*
894  * IP socket option processing.
895  */
896 int
897 ip_ctloutput(struct socket *so, struct sockopt *sopt)
898 {
899 	struct	inpcb *inp = sotoinpcb(so);
900 	int	error, optval;
901 
902 	error = optval = 0;
903 	if (sopt->sopt_level != IPPROTO_IP) {
904 		error = EINVAL;
905 
906 		if (sopt->sopt_level == SOL_SOCKET &&
907 		    sopt->sopt_dir == SOPT_SET) {
908 			switch (sopt->sopt_name) {
909 			case SO_REUSEADDR:
910 				INP_WLOCK(inp);
911 				if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
912 					if ((so->so_options &
913 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
914 						inp->inp_flags2 |= INP_REUSEPORT;
915 					else
916 						inp->inp_flags2 &= ~INP_REUSEPORT;
917 				}
918 				INP_WUNLOCK(inp);
919 				error = 0;
920 				break;
921 			case SO_REUSEPORT:
922 				INP_WLOCK(inp);
923 				if ((so->so_options & SO_REUSEPORT) != 0)
924 					inp->inp_flags2 |= INP_REUSEPORT;
925 				else
926 					inp->inp_flags2 &= ~INP_REUSEPORT;
927 				INP_WUNLOCK(inp);
928 				error = 0;
929 				break;
930 			case SO_SETFIB:
931 				INP_WLOCK(inp);
932 				inp->inp_inc.inc_fibnum = so->so_fibnum;
933 				INP_WUNLOCK(inp);
934 				error = 0;
935 				break;
936 			default:
937 				break;
938 			}
939 		}
940 		return (error);
941 	}
942 
943 	switch (sopt->sopt_dir) {
944 	case SOPT_SET:
945 		switch (sopt->sopt_name) {
946 		case IP_OPTIONS:
947 #ifdef notyet
948 		case IP_RETOPTS:
949 #endif
950 		{
951 			struct mbuf *m;
952 			if (sopt->sopt_valsize > MLEN) {
953 				error = EMSGSIZE;
954 				break;
955 			}
956 			MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
957 			if (m == NULL) {
958 				error = ENOBUFS;
959 				break;
960 			}
961 			m->m_len = sopt->sopt_valsize;
962 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
963 					    m->m_len);
964 			if (error) {
965 				m_free(m);
966 				break;
967 			}
968 			INP_WLOCK(inp);
969 			error = ip_pcbopts(inp, sopt->sopt_name, m);
970 			INP_WUNLOCK(inp);
971 			return (error);
972 		}
973 
974 		case IP_BINDANY:
975 			if (sopt->sopt_td != NULL) {
976 				error = priv_check(sopt->sopt_td,
977 				    PRIV_NETINET_BINDANY);
978 				if (error)
979 					break;
980 			}
981 			/* FALLTHROUGH */
982 		case IP_TOS:
983 		case IP_TTL:
984 		case IP_MINTTL:
985 		case IP_RECVOPTS:
986 		case IP_RECVRETOPTS:
987 		case IP_RECVDSTADDR:
988 		case IP_RECVTTL:
989 		case IP_RECVIF:
990 		case IP_FAITH:
991 		case IP_ONESBCAST:
992 		case IP_DONTFRAG:
993 		case IP_RECVTOS:
994 			error = sooptcopyin(sopt, &optval, sizeof optval,
995 					    sizeof optval);
996 			if (error)
997 				break;
998 
999 			switch (sopt->sopt_name) {
1000 			case IP_TOS:
1001 				inp->inp_ip_tos = optval;
1002 				break;
1003 
1004 			case IP_TTL:
1005 				inp->inp_ip_ttl = optval;
1006 				break;
1007 
1008 			case IP_MINTTL:
1009 				if (optval >= 0 && optval <= MAXTTL)
1010 					inp->inp_ip_minttl = optval;
1011 				else
1012 					error = EINVAL;
1013 				break;
1014 
1015 #define	OPTSET(bit) do {						\
1016 	INP_WLOCK(inp);							\
1017 	if (optval)							\
1018 		inp->inp_flags |= bit;					\
1019 	else								\
1020 		inp->inp_flags &= ~bit;					\
1021 	INP_WUNLOCK(inp);						\
1022 } while (0)
1023 
1024 			case IP_RECVOPTS:
1025 				OPTSET(INP_RECVOPTS);
1026 				break;
1027 
1028 			case IP_RECVRETOPTS:
1029 				OPTSET(INP_RECVRETOPTS);
1030 				break;
1031 
1032 			case IP_RECVDSTADDR:
1033 				OPTSET(INP_RECVDSTADDR);
1034 				break;
1035 
1036 			case IP_RECVTTL:
1037 				OPTSET(INP_RECVTTL);
1038 				break;
1039 
1040 			case IP_RECVIF:
1041 				OPTSET(INP_RECVIF);
1042 				break;
1043 
1044 			case IP_FAITH:
1045 				OPTSET(INP_FAITH);
1046 				break;
1047 
1048 			case IP_ONESBCAST:
1049 				OPTSET(INP_ONESBCAST);
1050 				break;
1051 			case IP_DONTFRAG:
1052 				OPTSET(INP_DONTFRAG);
1053 				break;
1054 			case IP_BINDANY:
1055 				OPTSET(INP_BINDANY);
1056 				break;
1057 			case IP_RECVTOS:
1058 				OPTSET(INP_RECVTOS);
1059 				break;
1060 			}
1061 			break;
1062 #undef OPTSET
1063 
1064 		/*
1065 		 * Multicast socket options are processed by the in_mcast
1066 		 * module.
1067 		 */
1068 		case IP_MULTICAST_IF:
1069 		case IP_MULTICAST_VIF:
1070 		case IP_MULTICAST_TTL:
1071 		case IP_MULTICAST_LOOP:
1072 		case IP_ADD_MEMBERSHIP:
1073 		case IP_DROP_MEMBERSHIP:
1074 		case IP_ADD_SOURCE_MEMBERSHIP:
1075 		case IP_DROP_SOURCE_MEMBERSHIP:
1076 		case IP_BLOCK_SOURCE:
1077 		case IP_UNBLOCK_SOURCE:
1078 		case IP_MSFILTER:
1079 		case MCAST_JOIN_GROUP:
1080 		case MCAST_LEAVE_GROUP:
1081 		case MCAST_JOIN_SOURCE_GROUP:
1082 		case MCAST_LEAVE_SOURCE_GROUP:
1083 		case MCAST_BLOCK_SOURCE:
1084 		case MCAST_UNBLOCK_SOURCE:
1085 			error = inp_setmoptions(inp, sopt);
1086 			break;
1087 
1088 		case IP_PORTRANGE:
1089 			error = sooptcopyin(sopt, &optval, sizeof optval,
1090 					    sizeof optval);
1091 			if (error)
1092 				break;
1093 
1094 			INP_WLOCK(inp);
1095 			switch (optval) {
1096 			case IP_PORTRANGE_DEFAULT:
1097 				inp->inp_flags &= ~(INP_LOWPORT);
1098 				inp->inp_flags &= ~(INP_HIGHPORT);
1099 				break;
1100 
1101 			case IP_PORTRANGE_HIGH:
1102 				inp->inp_flags &= ~(INP_LOWPORT);
1103 				inp->inp_flags |= INP_HIGHPORT;
1104 				break;
1105 
1106 			case IP_PORTRANGE_LOW:
1107 				inp->inp_flags &= ~(INP_HIGHPORT);
1108 				inp->inp_flags |= INP_LOWPORT;
1109 				break;
1110 
1111 			default:
1112 				error = EINVAL;
1113 				break;
1114 			}
1115 			INP_WUNLOCK(inp);
1116 			break;
1117 
1118 #ifdef IPSEC
1119 		case IP_IPSEC_POLICY:
1120 		{
1121 			caddr_t req;
1122 			struct mbuf *m;
1123 
1124 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1125 				break;
1126 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1127 				break;
1128 			req = mtod(m, caddr_t);
1129 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1130 			    m->m_len, (sopt->sopt_td != NULL) ?
1131 			    sopt->sopt_td->td_ucred : NULL);
1132 			m_freem(m);
1133 			break;
1134 		}
1135 #endif /* IPSEC */
1136 
1137 		default:
1138 			error = ENOPROTOOPT;
1139 			break;
1140 		}
1141 		break;
1142 
1143 	case SOPT_GET:
1144 		switch (sopt->sopt_name) {
1145 		case IP_OPTIONS:
1146 		case IP_RETOPTS:
1147 			if (inp->inp_options)
1148 				error = sooptcopyout(sopt,
1149 						     mtod(inp->inp_options,
1150 							  char *),
1151 						     inp->inp_options->m_len);
1152 			else
1153 				sopt->sopt_valsize = 0;
1154 			break;
1155 
1156 		case IP_TOS:
1157 		case IP_TTL:
1158 		case IP_MINTTL:
1159 		case IP_RECVOPTS:
1160 		case IP_RECVRETOPTS:
1161 		case IP_RECVDSTADDR:
1162 		case IP_RECVTTL:
1163 		case IP_RECVIF:
1164 		case IP_PORTRANGE:
1165 		case IP_FAITH:
1166 		case IP_ONESBCAST:
1167 		case IP_DONTFRAG:
1168 		case IP_BINDANY:
1169 		case IP_RECVTOS:
1170 			switch (sopt->sopt_name) {
1171 
1172 			case IP_TOS:
1173 				optval = inp->inp_ip_tos;
1174 				break;
1175 
1176 			case IP_TTL:
1177 				optval = inp->inp_ip_ttl;
1178 				break;
1179 
1180 			case IP_MINTTL:
1181 				optval = inp->inp_ip_minttl;
1182 				break;
1183 
1184 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1185 
1186 			case IP_RECVOPTS:
1187 				optval = OPTBIT(INP_RECVOPTS);
1188 				break;
1189 
1190 			case IP_RECVRETOPTS:
1191 				optval = OPTBIT(INP_RECVRETOPTS);
1192 				break;
1193 
1194 			case IP_RECVDSTADDR:
1195 				optval = OPTBIT(INP_RECVDSTADDR);
1196 				break;
1197 
1198 			case IP_RECVTTL:
1199 				optval = OPTBIT(INP_RECVTTL);
1200 				break;
1201 
1202 			case IP_RECVIF:
1203 				optval = OPTBIT(INP_RECVIF);
1204 				break;
1205 
1206 			case IP_PORTRANGE:
1207 				if (inp->inp_flags & INP_HIGHPORT)
1208 					optval = IP_PORTRANGE_HIGH;
1209 				else if (inp->inp_flags & INP_LOWPORT)
1210 					optval = IP_PORTRANGE_LOW;
1211 				else
1212 					optval = 0;
1213 				break;
1214 
1215 			case IP_FAITH:
1216 				optval = OPTBIT(INP_FAITH);
1217 				break;
1218 
1219 			case IP_ONESBCAST:
1220 				optval = OPTBIT(INP_ONESBCAST);
1221 				break;
1222 			case IP_DONTFRAG:
1223 				optval = OPTBIT(INP_DONTFRAG);
1224 				break;
1225 			case IP_BINDANY:
1226 				optval = OPTBIT(INP_BINDANY);
1227 				break;
1228 			case IP_RECVTOS:
1229 				optval = OPTBIT(INP_RECVTOS);
1230 				break;
1231 			}
1232 			error = sooptcopyout(sopt, &optval, sizeof optval);
1233 			break;
1234 
1235 		/*
1236 		 * Multicast socket options are processed by the in_mcast
1237 		 * module.
1238 		 */
1239 		case IP_MULTICAST_IF:
1240 		case IP_MULTICAST_VIF:
1241 		case IP_MULTICAST_TTL:
1242 		case IP_MULTICAST_LOOP:
1243 		case IP_MSFILTER:
1244 			error = inp_getmoptions(inp, sopt);
1245 			break;
1246 
1247 #ifdef IPSEC
1248 		case IP_IPSEC_POLICY:
1249 		{
1250 			struct mbuf *m = NULL;
1251 			caddr_t req = NULL;
1252 			size_t len = 0;
1253 
1254 			if (m != 0) {
1255 				req = mtod(m, caddr_t);
1256 				len = m->m_len;
1257 			}
1258 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1259 			if (error == 0)
1260 				error = soopt_mcopyout(sopt, m); /* XXX */
1261 			if (error == 0)
1262 				m_freem(m);
1263 			break;
1264 		}
1265 #endif /* IPSEC */
1266 
1267 		default:
1268 			error = ENOPROTOOPT;
1269 			break;
1270 		}
1271 		break;
1272 	}
1273 	return (error);
1274 }
1275 
1276 /*
1277  * Routine called from ip_output() to loop back a copy of an IP multicast
1278  * packet to the input queue of a specified interface.  Note that this
1279  * calls the output routine of the loopback "driver", but with an interface
1280  * pointer that might NOT be a loopback interface -- evil, but easier than
1281  * replicating that code here.
1282  */
1283 static void
1284 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1285     int hlen)
1286 {
1287 	register struct ip *ip;
1288 	struct mbuf *copym;
1289 
1290 	/*
1291 	 * Make a deep copy of the packet because we're going to
1292 	 * modify the pack in order to generate checksums.
1293 	 */
1294 	copym = m_dup(m, M_DONTWAIT);
1295 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1296 		copym = m_pullup(copym, hlen);
1297 	if (copym != NULL) {
1298 		/* If needed, compute the checksum and mark it as valid. */
1299 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1300 			in_delayed_cksum(copym);
1301 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1302 			copym->m_pkthdr.csum_flags |=
1303 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1304 			copym->m_pkthdr.csum_data = 0xffff;
1305 		}
1306 		/*
1307 		 * We don't bother to fragment if the IP length is greater
1308 		 * than the interface's MTU.  Can this possibly matter?
1309 		 */
1310 		ip = mtod(copym, struct ip *);
1311 		ip->ip_len = htons(ip->ip_len);
1312 		ip->ip_off = htons(ip->ip_off);
1313 		ip->ip_sum = 0;
1314 		ip->ip_sum = in_cksum(copym, hlen);
1315 #if 1 /* XXX */
1316 		if (dst->sin_family != AF_INET) {
1317 			printf("ip_mloopback: bad address family %d\n",
1318 						dst->sin_family);
1319 			dst->sin_family = AF_INET;
1320 		}
1321 #endif
1322 		if_simloop(ifp, copym, dst->sin_family, 0);
1323 	}
1324 }
1325