xref: /freebsd/sys/netinet/ip_output.c (revision 3b8f08459569bf0faa21473e5cec2491e95c9349)
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_inet.h"
36 #include "opt_ipfw.h"
37 #include "opt_ipsec.h"
38 #include "opt_mbuf_stress_test.h"
39 #include "opt_mpath.h"
40 #include "opt_route.h"
41 #include "opt_sctp.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/protosw.h>
51 #include <sys/sdt.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/ucred.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_llatbl.h>
60 #include <net/netisr.h>
61 #include <net/pfil.h>
62 #include <net/route.h>
63 #include <net/flowtable.h>
64 #ifdef RADIX_MPATH
65 #include <net/radix_mpath.h>
66 #endif
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_options.h>
77 #ifdef SCTP
78 #include <netinet/sctp.h>
79 #include <netinet/sctp_crc32.h>
80 #endif
81 
82 #ifdef IPSEC
83 #include <netinet/ip_ipsec.h>
84 #include <netipsec/ipsec.h>
85 #endif /* IPSEC*/
86 
87 #include <machine/in_cksum.h>
88 
89 #include <security/mac/mac_framework.h>
90 
91 VNET_DEFINE(u_short, ip_id);
92 
93 #ifdef MBUF_STRESS_TEST
94 static int mbuf_frag_size = 0;
95 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
96 	&mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
97 #endif
98 
99 static void	ip_mloopback
100 	(struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
101 
102 
103 extern int in_mcast_loop;
104 extern	struct protosw inetsw[];
105 
106 /*
107  * IP output.  The packet in mbuf chain m contains a skeletal IP
108  * header (with len, off, ttl, proto, tos, src, dst).
109  * The mbuf chain containing the packet will be freed.
110  * The mbuf opt, if present, will not be freed.
111  * If route ro is present and has ro_rt initialized, route lookup would be
112  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
113  * then result of route lookup is stored in ro->ro_rt.
114  *
115  * In the IP forwarding case, the packet will arrive with options already
116  * inserted, so must have a NULL opt pointer.
117  */
118 int
119 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
120     struct ip_moptions *imo, struct inpcb *inp)
121 {
122 	struct ip *ip;
123 	struct ifnet *ifp = NULL;	/* keep compiler happy */
124 	struct mbuf *m0;
125 	int hlen = sizeof (struct ip);
126 	int mtu;
127 	int error = 0;
128 	struct sockaddr_in *dst;
129 	const struct sockaddr_in *gw;
130 	struct in_ifaddr *ia;
131 	int isbroadcast;
132 	uint16_t ip_len, ip_off;
133 	struct route iproute;
134 	struct rtentry *rte;	/* cache for ro->ro_rt */
135 	struct in_addr odst;
136 	struct m_tag *fwd_tag = NULL;
137 #ifdef IPSEC
138 	int no_route_but_check_spd = 0;
139 #endif
140 	M_ASSERTPKTHDR(m);
141 
142 	if (inp != NULL) {
143 		INP_LOCK_ASSERT(inp);
144 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
145 		if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) {
146 			m->m_pkthdr.flowid = inp->inp_flowid;
147 			m->m_flags |= M_FLOWID;
148 		}
149 	}
150 
151 	if (ro == NULL) {
152 		ro = &iproute;
153 		bzero(ro, sizeof (*ro));
154 	}
155 
156 #ifdef FLOWTABLE
157 	if (ro->ro_rt == NULL)
158 		(void )flowtable_lookup(AF_INET, m, ro);
159 #endif
160 
161 	if (opt) {
162 		int len = 0;
163 		m = ip_insertoptions(m, opt, &len);
164 		if (len != 0)
165 			hlen = len; /* ip->ip_hl is updated above */
166 	}
167 	ip = mtod(m, struct ip *);
168 	ip_len = ntohs(ip->ip_len);
169 	ip_off = ntohs(ip->ip_off);
170 
171 	/*
172 	 * Fill in IP header.  If we are not allowing fragmentation,
173 	 * then the ip_id field is meaningless, but we don't set it
174 	 * to zero.  Doing so causes various problems when devices along
175 	 * the path (routers, load balancers, firewalls, etc.) illegally
176 	 * disable DF on our packet.  Note that a 16-bit counter
177 	 * will wrap around in less than 10 seconds at 100 Mbit/s on a
178 	 * medium with MTU 1500.  See Steven M. Bellovin, "A Technique
179 	 * for Counting NATted Hosts", Proc. IMW'02, available at
180 	 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>.
181 	 */
182 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
183 		ip->ip_v = IPVERSION;
184 		ip->ip_hl = hlen >> 2;
185 		ip->ip_id = ip_newid();
186 		IPSTAT_INC(ips_localout);
187 	} else {
188 		/* Header already set, fetch hlen from there */
189 		hlen = ip->ip_hl << 2;
190 	}
191 
192 	/*
193 	 * dst/gw handling:
194 	 *
195 	 * dst can be rewritten but always points to &ro->ro_dst.
196 	 * gw is readonly but can point either to dst OR rt_gateway,
197 	 * therefore we need restore gw if we're redoing lookup.
198 	 */
199 	gw = dst = (struct sockaddr_in *)&ro->ro_dst;
200 again:
201 	ia = NULL;
202 	/*
203 	 * If there is a cached route, check that it is to the same
204 	 * destination and is still up.  If not, free it and try again.
205 	 * The address family should also be checked in case of sharing
206 	 * the cache with IPv6.
207 	 */
208 	rte = ro->ro_rt;
209 	if (rte && ((rte->rt_flags & RTF_UP) == 0 ||
210 		    rte->rt_ifp == NULL ||
211 		    !RT_LINK_IS_UP(rte->rt_ifp) ||
212 			  dst->sin_family != AF_INET ||
213 			  dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
214 		RO_RTFREE(ro);
215 		ro->ro_lle = NULL;
216 		rte = NULL;
217 		gw = dst;
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),
234 						    RT_DEFAULT_FIB))) == NULL) {
235 			IPSTAT_INC(ips_noroute);
236 			error = ENETUNREACH;
237 			goto bad;
238 		}
239 		ip->ip_dst.s_addr = INADDR_BROADCAST;
240 		dst->sin_addr = ip->ip_dst;
241 		ifp = ia->ia_ifp;
242 		ip->ip_ttl = 1;
243 		isbroadcast = 1;
244 	} else if (flags & IP_ROUTETOIF) {
245 		if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst),
246 		    				    RT_DEFAULT_FIB))) == NULL &&
247 		    (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0,
248 		    				RT_DEFAULT_FIB))) == NULL) {
249 			IPSTAT_INC(ips_noroute);
250 			error = ENETUNREACH;
251 			goto bad;
252 		}
253 		ifp = ia->ia_ifp;
254 		ip->ip_ttl = 1;
255 		isbroadcast = in_broadcast(dst->sin_addr, ifp);
256 	} else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
257 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
258 		/*
259 		 * Bypass the normal routing lookup for multicast
260 		 * packets if the interface is specified.
261 		 */
262 		ifp = imo->imo_multicast_ifp;
263 		IFP_TO_IA(ifp, ia);
264 		isbroadcast = 0;	/* fool gcc */
265 	} else {
266 		/*
267 		 * We want to do any cloning requested by the link layer,
268 		 * as this is probably required in all cases for correct
269 		 * operation (as it is for ARP).
270 		 */
271 		if (rte == NULL) {
272 #ifdef RADIX_MPATH
273 			rtalloc_mpath_fib(ro,
274 			    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
275 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
276 #else
277 			in_rtalloc_ign(ro, 0,
278 			    inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
279 #endif
280 			rte = ro->ro_rt;
281 		}
282 		if (rte == NULL ||
283 		    rte->rt_ifp == NULL ||
284 		    !RT_LINK_IS_UP(rte->rt_ifp)) {
285 #ifdef IPSEC
286 			/*
287 			 * There is no route for this packet, but it is
288 			 * possible that a matching SPD entry exists.
289 			 */
290 			no_route_but_check_spd = 1;
291 			mtu = 0; /* Silence GCC warning. */
292 			goto sendit;
293 #endif
294 			IPSTAT_INC(ips_noroute);
295 			error = EHOSTUNREACH;
296 			goto bad;
297 		}
298 		ia = ifatoia(rte->rt_ifa);
299 		ifp = rte->rt_ifp;
300 		counter_u64_add(rte->rt_pksent, 1);
301 		if (rte->rt_flags & RTF_GATEWAY)
302 			gw = (struct sockaddr_in *)rte->rt_gateway;
303 		if (rte->rt_flags & RTF_HOST)
304 			isbroadcast = (rte->rt_flags & RTF_BROADCAST);
305 		else
306 			isbroadcast = in_broadcast(gw->sin_addr, ifp);
307 	}
308 	/*
309 	 * Calculate MTU.  If we have a route that is up, use that,
310 	 * otherwise use the interface's MTU.
311 	 */
312 	if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) {
313 		/*
314 		 * This case can happen if the user changed the MTU
315 		 * of an interface after enabling IP on it.  Because
316 		 * most netifs don't keep track of routes pointing to
317 		 * them, there is no way for one to update all its
318 		 * routes when the MTU is changed.
319 		 */
320 		if (rte->rt_mtu > ifp->if_mtu)
321 			rte->rt_mtu = ifp->if_mtu;
322 		mtu = rte->rt_mtu;
323 	} else {
324 		mtu = ifp->if_mtu;
325 	}
326 	/* Catch a possible divide by zero later. */
327 	KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p",
328 	    __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp));
329 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
330 		m->m_flags |= M_MCAST;
331 		/*
332 		 * IP destination address is multicast.  Make sure "gw"
333 		 * still points to the address in "ro".  (It may have been
334 		 * changed to point to a gateway address, above.)
335 		 */
336 		gw = dst;
337 		/*
338 		 * See if the caller provided any multicast options
339 		 */
340 		if (imo != NULL) {
341 			ip->ip_ttl = imo->imo_multicast_ttl;
342 			if (imo->imo_multicast_vif != -1)
343 				ip->ip_src.s_addr =
344 				    ip_mcast_src ?
345 				    ip_mcast_src(imo->imo_multicast_vif) :
346 				    INADDR_ANY;
347 		} else
348 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
349 		/*
350 		 * Confirm that the outgoing interface supports multicast.
351 		 */
352 		if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
353 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
354 				IPSTAT_INC(ips_noroute);
355 				error = ENETUNREACH;
356 				goto bad;
357 			}
358 		}
359 		/*
360 		 * If source address not specified yet, use address
361 		 * of outgoing interface.
362 		 */
363 		if (ip->ip_src.s_addr == INADDR_ANY) {
364 			/* Interface may have no addresses. */
365 			if (ia != NULL)
366 				ip->ip_src = IA_SIN(ia)->sin_addr;
367 		}
368 
369 		if ((imo == NULL && in_mcast_loop) ||
370 		    (imo && imo->imo_multicast_loop)) {
371 			/*
372 			 * Loop back multicast datagram if not expressly
373 			 * forbidden to do so, even if we are not a member
374 			 * of the group; ip_input() will filter it later,
375 			 * thus deferring a hash lookup and mutex acquisition
376 			 * at the expense of a cheap copy using m_copym().
377 			 */
378 			ip_mloopback(ifp, m, dst, hlen);
379 		} else {
380 			/*
381 			 * If we are acting as a multicast router, perform
382 			 * multicast forwarding as if the packet had just
383 			 * arrived on the interface to which we are about
384 			 * to send.  The multicast forwarding function
385 			 * recursively calls this function, using the
386 			 * IP_FORWARDING flag to prevent infinite recursion.
387 			 *
388 			 * Multicasts that are looped back by ip_mloopback(),
389 			 * above, will be forwarded by the ip_input() routine,
390 			 * if necessary.
391 			 */
392 			if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) {
393 				/*
394 				 * If rsvp daemon is not running, do not
395 				 * set ip_moptions. This ensures that the packet
396 				 * is multicast and not just sent down one link
397 				 * as prescribed by rsvpd.
398 				 */
399 				if (!V_rsvp_on)
400 					imo = NULL;
401 				if (ip_mforward &&
402 				    ip_mforward(ip, ifp, m, imo) != 0) {
403 					m_freem(m);
404 					goto done;
405 				}
406 			}
407 		}
408 
409 		/*
410 		 * Multicasts with a time-to-live of zero may be looped-
411 		 * back, above, but must not be transmitted on a network.
412 		 * Also, multicasts addressed to the loopback interface
413 		 * are not sent -- the above call to ip_mloopback() will
414 		 * loop back a copy. ip_input() will drop the copy if
415 		 * this host does not belong to the destination group on
416 		 * the loopback interface.
417 		 */
418 		if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
419 			m_freem(m);
420 			goto done;
421 		}
422 
423 		goto sendit;
424 	}
425 
426 	/*
427 	 * If the source address is not specified yet, use the address
428 	 * of the outoing interface.
429 	 */
430 	if (ip->ip_src.s_addr == INADDR_ANY) {
431 		/* Interface may have no addresses. */
432 		if (ia != NULL) {
433 			ip->ip_src = IA_SIN(ia)->sin_addr;
434 		}
435 	}
436 
437 	/*
438 	 * Both in the SMP world, pre-emption world if_transmit() world,
439 	 * the following code doesn't really function as intended any further.
440 	 *
441 	 * + There can and will be multiple CPUs running this code path
442 	 *   in parallel, and we do no lock holding when checking the
443 	 *   queue depth;
444 	 * + And since other threads can be running concurrently, even if
445 	 *   we do pass this check, another thread may queue some frames
446 	 *   before this thread does and it will end up partially or fully
447 	 *   failing to send anyway;
448 	 * + if_transmit() based drivers don't necessarily set ifq_len
449 	 *   at all.
450 	 *
451 	 * This should be replaced with a method of pushing an entire list
452 	 * of fragment frames to the driver and have the driver decide
453 	 * whether it can queue or not queue the entire set.
454 	 */
455 #if 0
456 	/*
457 	 * Verify that we have any chance at all of being able to queue the
458 	 * packet or packet fragments, unless ALTQ is enabled on the given
459 	 * interface in which case packetdrop should be done by queueing.
460 	 */
461 	n = ip_len / mtu + 1; /* how many fragments ? */
462 	if (
463 #ifdef ALTQ
464 	    (!ALTQ_IS_ENABLED(&ifp->if_snd)) &&
465 #endif /* ALTQ */
466 	    (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) {
467 		error = ENOBUFS;
468 		IPSTAT_INC(ips_odropped);
469 		ifp->if_snd.ifq_drops += n;
470 		goto bad;
471 	}
472 #endif
473 
474 	/*
475 	 * Look for broadcast address and
476 	 * verify user is allowed to send
477 	 * such a packet.
478 	 */
479 	if (isbroadcast) {
480 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
481 			error = EADDRNOTAVAIL;
482 			goto bad;
483 		}
484 		if ((flags & IP_ALLOWBROADCAST) == 0) {
485 			error = EACCES;
486 			goto bad;
487 		}
488 		/* don't allow broadcast messages to be fragmented */
489 		if (ip_len > mtu) {
490 			error = EMSGSIZE;
491 			goto bad;
492 		}
493 		m->m_flags |= M_BCAST;
494 	} else {
495 		m->m_flags &= ~M_BCAST;
496 	}
497 
498 sendit:
499 #ifdef IPSEC
500 	switch(ip_ipsec_output(&m, inp, &flags, &error)) {
501 	case 1:
502 		goto bad;
503 	case -1:
504 		goto done;
505 	case 0:
506 	default:
507 		break;	/* Continue with packet processing. */
508 	}
509 	/*
510 	 * Check if there was a route for this packet; return error if not.
511 	 */
512 	if (no_route_but_check_spd) {
513 		IPSTAT_INC(ips_noroute);
514 		error = EHOSTUNREACH;
515 		goto bad;
516 	}
517 	/* Update variables that are affected by ipsec4_output(). */
518 	ip = mtod(m, struct ip *);
519 	hlen = ip->ip_hl << 2;
520 #endif /* IPSEC */
521 
522 	/* Jump over all PFIL processing if hooks are not active. */
523 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
524 		goto passout;
525 
526 	/* Run through list of hooks for output packets. */
527 	odst.s_addr = ip->ip_dst.s_addr;
528 	error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
529 	if (error != 0 || m == NULL)
530 		goto done;
531 
532 	ip = mtod(m, struct ip *);
533 
534 	/* See if destination IP address was changed by packet filter. */
535 	if (odst.s_addr != ip->ip_dst.s_addr) {
536 		m->m_flags |= M_SKIP_FIREWALL;
537 		/* If destination is now ourself drop to ip_input(). */
538 		if (in_localip(ip->ip_dst)) {
539 			m->m_flags |= M_FASTFWD_OURS;
540 			if (m->m_pkthdr.rcvif == NULL)
541 				m->m_pkthdr.rcvif = V_loif;
542 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
543 				m->m_pkthdr.csum_flags |=
544 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
545 				m->m_pkthdr.csum_data = 0xffff;
546 			}
547 			m->m_pkthdr.csum_flags |=
548 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
549 #ifdef SCTP
550 			if (m->m_pkthdr.csum_flags & CSUM_SCTP)
551 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
552 #endif
553 			error = netisr_queue(NETISR_IP, m);
554 			goto done;
555 		} else
556 			goto again;	/* Redo the routing table lookup. */
557 	}
558 
559 	/* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
560 	if (m->m_flags & M_FASTFWD_OURS) {
561 		if (m->m_pkthdr.rcvif == NULL)
562 			m->m_pkthdr.rcvif = V_loif;
563 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
564 			m->m_pkthdr.csum_flags |=
565 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
566 			m->m_pkthdr.csum_data = 0xffff;
567 		}
568 #ifdef SCTP
569 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
570 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
571 #endif
572 		m->m_pkthdr.csum_flags |=
573 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
574 
575 		error = netisr_queue(NETISR_IP, m);
576 		goto done;
577 	}
578 	/* Or forward to some other address? */
579 	if ((m->m_flags & M_IP_NEXTHOP) &&
580 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
581 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
582 		m->m_flags |= M_SKIP_FIREWALL;
583 		m->m_flags &= ~M_IP_NEXTHOP;
584 		m_tag_delete(m, fwd_tag);
585 		goto again;
586 	}
587 
588 passout:
589 	/* 127/8 must not appear on wire - RFC1122. */
590 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
591 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
592 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
593 			IPSTAT_INC(ips_badaddr);
594 			error = EADDRNOTAVAIL;
595 			goto bad;
596 		}
597 	}
598 
599 	m->m_pkthdr.csum_flags |= CSUM_IP;
600 	if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) {
601 		in_delayed_cksum(m);
602 		m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
603 	}
604 #ifdef SCTP
605 	if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) {
606 		sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
607 		m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
608 	}
609 #endif
610 
611 	/*
612 	 * If small enough for interface, or the interface will take
613 	 * care of the fragmentation for us, we can just send directly.
614 	 */
615 	if (ip_len <= mtu ||
616 	    (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 ||
617 	    ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) {
618 		ip->ip_sum = 0;
619 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) {
620 			ip->ip_sum = in_cksum(m, hlen);
621 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
622 		}
623 
624 		/*
625 		 * Record statistics for this interface address.
626 		 * With CSUM_TSO the byte/packet count will be slightly
627 		 * incorrect because we count the IP+TCP headers only
628 		 * once instead of for every generated packet.
629 		 */
630 		if (!(flags & IP_FORWARDING) && ia) {
631 			if (m->m_pkthdr.csum_flags & CSUM_TSO)
632 				counter_u64_add(ia->ia_ifa.ifa_opackets,
633 				    m->m_pkthdr.len / m->m_pkthdr.tso_segsz);
634 			else
635 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
636 
637 			counter_u64_add(ia->ia_ifa.ifa_obytes, m->m_pkthdr.len);
638 		}
639 #ifdef MBUF_STRESS_TEST
640 		if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
641 			m = m_fragment(m, M_NOWAIT, mbuf_frag_size);
642 #endif
643 		/*
644 		 * Reset layer specific mbuf flags
645 		 * to avoid confusing lower layers.
646 		 */
647 		m_clrprotoflags(m);
648 		IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
649 		error = (*ifp->if_output)(ifp, m,
650 		    (const struct sockaddr *)gw, ro);
651 		goto done;
652 	}
653 
654 	/* Balk when DF bit is set or the interface didn't support TSO. */
655 	if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) {
656 		error = EMSGSIZE;
657 		IPSTAT_INC(ips_cantfrag);
658 		goto bad;
659 	}
660 
661 	/*
662 	 * Too large for interface; fragment if possible. If successful,
663 	 * on return, m will point to a list of packets to be sent.
664 	 */
665 	error = ip_fragment(ip, &m, mtu, ifp->if_hwassist);
666 	if (error)
667 		goto bad;
668 	for (; m; m = m0) {
669 		m0 = m->m_nextpkt;
670 		m->m_nextpkt = 0;
671 		if (error == 0) {
672 			/* Record statistics for this interface address. */
673 			if (ia != NULL) {
674 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
675 				counter_u64_add(ia->ia_ifa.ifa_obytes,
676 				    m->m_pkthdr.len);
677 			}
678 			/*
679 			 * Reset layer specific mbuf flags
680 			 * to avoid confusing upper layers.
681 			 */
682 			m_clrprotoflags(m);
683 
684 			IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL);
685 			error = (*ifp->if_output)(ifp, m,
686 			    (const struct sockaddr *)gw, ro);
687 		} else
688 			m_freem(m);
689 	}
690 
691 	if (error == 0)
692 		IPSTAT_INC(ips_fragmented);
693 
694 done:
695 	if (ro == &iproute)
696 		RO_RTFREE(ro);
697 	return (error);
698 bad:
699 	m_freem(m);
700 	goto done;
701 }
702 
703 /*
704  * Create a chain of fragments which fit the given mtu. m_frag points to the
705  * mbuf to be fragmented; on return it points to the chain with the fragments.
706  * Return 0 if no error. If error, m_frag may contain a partially built
707  * chain of fragments that should be freed by the caller.
708  *
709  * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
710  */
711 int
712 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
713     u_long if_hwassist_flags)
714 {
715 	int error = 0;
716 	int hlen = ip->ip_hl << 2;
717 	int len = (mtu - hlen) & ~7;	/* size of payload in each fragment */
718 	int off;
719 	struct mbuf *m0 = *m_frag;	/* the original packet		*/
720 	int firstlen;
721 	struct mbuf **mnext;
722 	int nfrags;
723 	uint16_t ip_len, ip_off;
724 
725 	ip_len = ntohs(ip->ip_len);
726 	ip_off = ntohs(ip->ip_off);
727 
728 	if (ip_off & IP_DF) {	/* Fragmentation not allowed */
729 		IPSTAT_INC(ips_cantfrag);
730 		return EMSGSIZE;
731 	}
732 
733 	/*
734 	 * Must be able to put at least 8 bytes per fragment.
735 	 */
736 	if (len < 8)
737 		return EMSGSIZE;
738 
739 	/*
740 	 * If the interface will not calculate checksums on
741 	 * fragmented packets, then do it here.
742 	 */
743 	if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
744 		in_delayed_cksum(m0);
745 		m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
746 	}
747 #ifdef SCTP
748 	if (m0->m_pkthdr.csum_flags & CSUM_SCTP) {
749 		sctp_delayed_cksum(m0, hlen);
750 		m0->m_pkthdr.csum_flags &= ~CSUM_SCTP;
751 	}
752 #endif
753 	if (len > PAGE_SIZE) {
754 		/*
755 		 * Fragment large datagrams such that each segment
756 		 * contains a multiple of PAGE_SIZE amount of data,
757 		 * plus headers. This enables a receiver to perform
758 		 * page-flipping zero-copy optimizations.
759 		 *
760 		 * XXX When does this help given that sender and receiver
761 		 * could have different page sizes, and also mtu could
762 		 * be less than the receiver's page size ?
763 		 */
764 		int newlen;
765 		struct mbuf *m;
766 
767 		for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
768 			off += m->m_len;
769 
770 		/*
771 		 * firstlen (off - hlen) must be aligned on an
772 		 * 8-byte boundary
773 		 */
774 		if (off < hlen)
775 			goto smart_frag_failure;
776 		off = ((off - hlen) & ~7) + hlen;
777 		newlen = (~PAGE_MASK) & mtu;
778 		if ((newlen + sizeof (struct ip)) > mtu) {
779 			/* we failed, go back the default */
780 smart_frag_failure:
781 			newlen = len;
782 			off = hlen + len;
783 		}
784 		len = newlen;
785 
786 	} else {
787 		off = hlen + len;
788 	}
789 
790 	firstlen = off - hlen;
791 	mnext = &m0->m_nextpkt;		/* pointer to next packet */
792 
793 	/*
794 	 * Loop through length of segment after first fragment,
795 	 * make new header and copy data of each part and link onto chain.
796 	 * Here, m0 is the original packet, m is the fragment being created.
797 	 * The fragments are linked off the m_nextpkt of the original
798 	 * packet, which after processing serves as the first fragment.
799 	 */
800 	for (nfrags = 1; off < ip_len; off += len, nfrags++) {
801 		struct ip *mhip;	/* ip header on the fragment */
802 		struct mbuf *m;
803 		int mhlen = sizeof (struct ip);
804 
805 		m = m_gethdr(M_NOWAIT, MT_DATA);
806 		if (m == NULL) {
807 			error = ENOBUFS;
808 			IPSTAT_INC(ips_odropped);
809 			goto done;
810 		}
811 		m->m_flags |= (m0->m_flags & M_MCAST);
812 		/*
813 		 * In the first mbuf, leave room for the link header, then
814 		 * copy the original IP header including options. The payload
815 		 * goes into an additional mbuf chain returned by m_copym().
816 		 */
817 		m->m_data += max_linkhdr;
818 		mhip = mtod(m, struct ip *);
819 		*mhip = *ip;
820 		if (hlen > sizeof (struct ip)) {
821 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
822 			mhip->ip_v = IPVERSION;
823 			mhip->ip_hl = mhlen >> 2;
824 		}
825 		m->m_len = mhlen;
826 		/* XXX do we need to add ip_off below ? */
827 		mhip->ip_off = ((off - hlen) >> 3) + ip_off;
828 		if (off + len >= ip_len)
829 			len = ip_len - off;
830 		else
831 			mhip->ip_off |= IP_MF;
832 		mhip->ip_len = htons((u_short)(len + mhlen));
833 		m->m_next = m_copym(m0, off, len, M_NOWAIT);
834 		if (m->m_next == NULL) {	/* copy failed */
835 			m_free(m);
836 			error = ENOBUFS;	/* ??? */
837 			IPSTAT_INC(ips_odropped);
838 			goto done;
839 		}
840 		m->m_pkthdr.len = mhlen + len;
841 		m->m_pkthdr.rcvif = NULL;
842 #ifdef MAC
843 		mac_netinet_fragment(m0, m);
844 #endif
845 		m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
846 		mhip->ip_off = htons(mhip->ip_off);
847 		mhip->ip_sum = 0;
848 		if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
849 			mhip->ip_sum = in_cksum(m, mhlen);
850 			m->m_pkthdr.csum_flags &= ~CSUM_IP;
851 		}
852 		*mnext = m;
853 		mnext = &m->m_nextpkt;
854 	}
855 	IPSTAT_ADD(ips_ofragments, nfrags);
856 
857 	/*
858 	 * Update first fragment by trimming what's been copied out
859 	 * and updating header.
860 	 */
861 	m_adj(m0, hlen + firstlen - ip_len);
862 	m0->m_pkthdr.len = hlen + firstlen;
863 	ip->ip_len = htons((u_short)m0->m_pkthdr.len);
864 	ip->ip_off = htons(ip_off | IP_MF);
865 	ip->ip_sum = 0;
866 	if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) {
867 		ip->ip_sum = in_cksum(m0, hlen);
868 		m0->m_pkthdr.csum_flags &= ~CSUM_IP;
869 	}
870 
871 done:
872 	*m_frag = m0;
873 	return error;
874 }
875 
876 void
877 in_delayed_cksum(struct mbuf *m)
878 {
879 	struct ip *ip;
880 	uint16_t csum, offset, ip_len;
881 
882 	ip = mtod(m, struct ip *);
883 	offset = ip->ip_hl << 2 ;
884 	ip_len = ntohs(ip->ip_len);
885 	csum = in_cksum_skip(m, ip_len, offset);
886 	if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
887 		csum = 0xffff;
888 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
889 
890 	if (offset + sizeof(u_short) > m->m_len) {
891 		printf("delayed m_pullup, m->len: %d  off: %d  p: %d\n",
892 		    m->m_len, offset, ip->ip_p);
893 		/*
894 		 * XXX
895 		 * this shouldn't happen, but if it does, the
896 		 * correct behavior may be to insert the checksum
897 		 * in the appropriate next mbuf in the chain.
898 		 */
899 		return;
900 	}
901 	*(u_short *)(m->m_data + offset) = csum;
902 }
903 
904 /*
905  * IP socket option processing.
906  */
907 int
908 ip_ctloutput(struct socket *so, struct sockopt *sopt)
909 {
910 	struct	inpcb *inp = sotoinpcb(so);
911 	int	error, optval;
912 
913 	error = optval = 0;
914 	if (sopt->sopt_level != IPPROTO_IP) {
915 		error = EINVAL;
916 
917 		if (sopt->sopt_level == SOL_SOCKET &&
918 		    sopt->sopt_dir == SOPT_SET) {
919 			switch (sopt->sopt_name) {
920 			case SO_REUSEADDR:
921 				INP_WLOCK(inp);
922 				if ((so->so_options & SO_REUSEADDR) != 0)
923 					inp->inp_flags2 |= INP_REUSEADDR;
924 				else
925 					inp->inp_flags2 &= ~INP_REUSEADDR;
926 				INP_WUNLOCK(inp);
927 				error = 0;
928 				break;
929 			case SO_REUSEPORT:
930 				INP_WLOCK(inp);
931 				if ((so->so_options & SO_REUSEPORT) != 0)
932 					inp->inp_flags2 |= INP_REUSEPORT;
933 				else
934 					inp->inp_flags2 &= ~INP_REUSEPORT;
935 				INP_WUNLOCK(inp);
936 				error = 0;
937 				break;
938 			case SO_SETFIB:
939 				INP_WLOCK(inp);
940 				inp->inp_inc.inc_fibnum = so->so_fibnum;
941 				INP_WUNLOCK(inp);
942 				error = 0;
943 				break;
944 			default:
945 				break;
946 			}
947 		}
948 		return (error);
949 	}
950 
951 	switch (sopt->sopt_dir) {
952 	case SOPT_SET:
953 		switch (sopt->sopt_name) {
954 		case IP_OPTIONS:
955 #ifdef notyet
956 		case IP_RETOPTS:
957 #endif
958 		{
959 			struct mbuf *m;
960 			if (sopt->sopt_valsize > MLEN) {
961 				error = EMSGSIZE;
962 				break;
963 			}
964 			m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA);
965 			if (m == NULL) {
966 				error = ENOBUFS;
967 				break;
968 			}
969 			m->m_len = sopt->sopt_valsize;
970 			error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
971 					    m->m_len);
972 			if (error) {
973 				m_free(m);
974 				break;
975 			}
976 			INP_WLOCK(inp);
977 			error = ip_pcbopts(inp, sopt->sopt_name, m);
978 			INP_WUNLOCK(inp);
979 			return (error);
980 		}
981 
982 		case IP_BINDANY:
983 			if (sopt->sopt_td != NULL) {
984 				error = priv_check(sopt->sopt_td,
985 				    PRIV_NETINET_BINDANY);
986 				if (error)
987 					break;
988 			}
989 			/* FALLTHROUGH */
990 		case IP_TOS:
991 		case IP_TTL:
992 		case IP_MINTTL:
993 		case IP_RECVOPTS:
994 		case IP_RECVRETOPTS:
995 		case IP_RECVDSTADDR:
996 		case IP_RECVTTL:
997 		case IP_RECVIF:
998 		case IP_FAITH:
999 		case IP_ONESBCAST:
1000 		case IP_DONTFRAG:
1001 		case IP_RECVTOS:
1002 			error = sooptcopyin(sopt, &optval, sizeof optval,
1003 					    sizeof optval);
1004 			if (error)
1005 				break;
1006 
1007 			switch (sopt->sopt_name) {
1008 			case IP_TOS:
1009 				inp->inp_ip_tos = optval;
1010 				break;
1011 
1012 			case IP_TTL:
1013 				inp->inp_ip_ttl = optval;
1014 				break;
1015 
1016 			case IP_MINTTL:
1017 				if (optval >= 0 && optval <= MAXTTL)
1018 					inp->inp_ip_minttl = optval;
1019 				else
1020 					error = EINVAL;
1021 				break;
1022 
1023 #define	OPTSET(bit) do {						\
1024 	INP_WLOCK(inp);							\
1025 	if (optval)							\
1026 		inp->inp_flags |= bit;					\
1027 	else								\
1028 		inp->inp_flags &= ~bit;					\
1029 	INP_WUNLOCK(inp);						\
1030 } while (0)
1031 
1032 			case IP_RECVOPTS:
1033 				OPTSET(INP_RECVOPTS);
1034 				break;
1035 
1036 			case IP_RECVRETOPTS:
1037 				OPTSET(INP_RECVRETOPTS);
1038 				break;
1039 
1040 			case IP_RECVDSTADDR:
1041 				OPTSET(INP_RECVDSTADDR);
1042 				break;
1043 
1044 			case IP_RECVTTL:
1045 				OPTSET(INP_RECVTTL);
1046 				break;
1047 
1048 			case IP_RECVIF:
1049 				OPTSET(INP_RECVIF);
1050 				break;
1051 
1052 			case IP_FAITH:
1053 				OPTSET(INP_FAITH);
1054 				break;
1055 
1056 			case IP_ONESBCAST:
1057 				OPTSET(INP_ONESBCAST);
1058 				break;
1059 			case IP_DONTFRAG:
1060 				OPTSET(INP_DONTFRAG);
1061 				break;
1062 			case IP_BINDANY:
1063 				OPTSET(INP_BINDANY);
1064 				break;
1065 			case IP_RECVTOS:
1066 				OPTSET(INP_RECVTOS);
1067 				break;
1068 			}
1069 			break;
1070 #undef OPTSET
1071 
1072 		/*
1073 		 * Multicast socket options are processed by the in_mcast
1074 		 * module.
1075 		 */
1076 		case IP_MULTICAST_IF:
1077 		case IP_MULTICAST_VIF:
1078 		case IP_MULTICAST_TTL:
1079 		case IP_MULTICAST_LOOP:
1080 		case IP_ADD_MEMBERSHIP:
1081 		case IP_DROP_MEMBERSHIP:
1082 		case IP_ADD_SOURCE_MEMBERSHIP:
1083 		case IP_DROP_SOURCE_MEMBERSHIP:
1084 		case IP_BLOCK_SOURCE:
1085 		case IP_UNBLOCK_SOURCE:
1086 		case IP_MSFILTER:
1087 		case MCAST_JOIN_GROUP:
1088 		case MCAST_LEAVE_GROUP:
1089 		case MCAST_JOIN_SOURCE_GROUP:
1090 		case MCAST_LEAVE_SOURCE_GROUP:
1091 		case MCAST_BLOCK_SOURCE:
1092 		case MCAST_UNBLOCK_SOURCE:
1093 			error = inp_setmoptions(inp, sopt);
1094 			break;
1095 
1096 		case IP_PORTRANGE:
1097 			error = sooptcopyin(sopt, &optval, sizeof optval,
1098 					    sizeof optval);
1099 			if (error)
1100 				break;
1101 
1102 			INP_WLOCK(inp);
1103 			switch (optval) {
1104 			case IP_PORTRANGE_DEFAULT:
1105 				inp->inp_flags &= ~(INP_LOWPORT);
1106 				inp->inp_flags &= ~(INP_HIGHPORT);
1107 				break;
1108 
1109 			case IP_PORTRANGE_HIGH:
1110 				inp->inp_flags &= ~(INP_LOWPORT);
1111 				inp->inp_flags |= INP_HIGHPORT;
1112 				break;
1113 
1114 			case IP_PORTRANGE_LOW:
1115 				inp->inp_flags &= ~(INP_HIGHPORT);
1116 				inp->inp_flags |= INP_LOWPORT;
1117 				break;
1118 
1119 			default:
1120 				error = EINVAL;
1121 				break;
1122 			}
1123 			INP_WUNLOCK(inp);
1124 			break;
1125 
1126 #ifdef IPSEC
1127 		case IP_IPSEC_POLICY:
1128 		{
1129 			caddr_t req;
1130 			struct mbuf *m;
1131 
1132 			if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1133 				break;
1134 			if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1135 				break;
1136 			req = mtod(m, caddr_t);
1137 			error = ipsec_set_policy(inp, sopt->sopt_name, req,
1138 			    m->m_len, (sopt->sopt_td != NULL) ?
1139 			    sopt->sopt_td->td_ucred : NULL);
1140 			m_freem(m);
1141 			break;
1142 		}
1143 #endif /* IPSEC */
1144 
1145 		default:
1146 			error = ENOPROTOOPT;
1147 			break;
1148 		}
1149 		break;
1150 
1151 	case SOPT_GET:
1152 		switch (sopt->sopt_name) {
1153 		case IP_OPTIONS:
1154 		case IP_RETOPTS:
1155 			if (inp->inp_options)
1156 				error = sooptcopyout(sopt,
1157 						     mtod(inp->inp_options,
1158 							  char *),
1159 						     inp->inp_options->m_len);
1160 			else
1161 				sopt->sopt_valsize = 0;
1162 			break;
1163 
1164 		case IP_TOS:
1165 		case IP_TTL:
1166 		case IP_MINTTL:
1167 		case IP_RECVOPTS:
1168 		case IP_RECVRETOPTS:
1169 		case IP_RECVDSTADDR:
1170 		case IP_RECVTTL:
1171 		case IP_RECVIF:
1172 		case IP_PORTRANGE:
1173 		case IP_FAITH:
1174 		case IP_ONESBCAST:
1175 		case IP_DONTFRAG:
1176 		case IP_BINDANY:
1177 		case IP_RECVTOS:
1178 			switch (sopt->sopt_name) {
1179 
1180 			case IP_TOS:
1181 				optval = inp->inp_ip_tos;
1182 				break;
1183 
1184 			case IP_TTL:
1185 				optval = inp->inp_ip_ttl;
1186 				break;
1187 
1188 			case IP_MINTTL:
1189 				optval = inp->inp_ip_minttl;
1190 				break;
1191 
1192 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
1193 
1194 			case IP_RECVOPTS:
1195 				optval = OPTBIT(INP_RECVOPTS);
1196 				break;
1197 
1198 			case IP_RECVRETOPTS:
1199 				optval = OPTBIT(INP_RECVRETOPTS);
1200 				break;
1201 
1202 			case IP_RECVDSTADDR:
1203 				optval = OPTBIT(INP_RECVDSTADDR);
1204 				break;
1205 
1206 			case IP_RECVTTL:
1207 				optval = OPTBIT(INP_RECVTTL);
1208 				break;
1209 
1210 			case IP_RECVIF:
1211 				optval = OPTBIT(INP_RECVIF);
1212 				break;
1213 
1214 			case IP_PORTRANGE:
1215 				if (inp->inp_flags & INP_HIGHPORT)
1216 					optval = IP_PORTRANGE_HIGH;
1217 				else if (inp->inp_flags & INP_LOWPORT)
1218 					optval = IP_PORTRANGE_LOW;
1219 				else
1220 					optval = 0;
1221 				break;
1222 
1223 			case IP_FAITH:
1224 				optval = OPTBIT(INP_FAITH);
1225 				break;
1226 
1227 			case IP_ONESBCAST:
1228 				optval = OPTBIT(INP_ONESBCAST);
1229 				break;
1230 			case IP_DONTFRAG:
1231 				optval = OPTBIT(INP_DONTFRAG);
1232 				break;
1233 			case IP_BINDANY:
1234 				optval = OPTBIT(INP_BINDANY);
1235 				break;
1236 			case IP_RECVTOS:
1237 				optval = OPTBIT(INP_RECVTOS);
1238 				break;
1239 			}
1240 			error = sooptcopyout(sopt, &optval, sizeof optval);
1241 			break;
1242 
1243 		/*
1244 		 * Multicast socket options are processed by the in_mcast
1245 		 * module.
1246 		 */
1247 		case IP_MULTICAST_IF:
1248 		case IP_MULTICAST_VIF:
1249 		case IP_MULTICAST_TTL:
1250 		case IP_MULTICAST_LOOP:
1251 		case IP_MSFILTER:
1252 			error = inp_getmoptions(inp, sopt);
1253 			break;
1254 
1255 #ifdef IPSEC
1256 		case IP_IPSEC_POLICY:
1257 		{
1258 			struct mbuf *m = NULL;
1259 			caddr_t req = NULL;
1260 			size_t len = 0;
1261 
1262 			if (m != 0) {
1263 				req = mtod(m, caddr_t);
1264 				len = m->m_len;
1265 			}
1266 			error = ipsec_get_policy(sotoinpcb(so), req, len, &m);
1267 			if (error == 0)
1268 				error = soopt_mcopyout(sopt, m); /* XXX */
1269 			if (error == 0)
1270 				m_freem(m);
1271 			break;
1272 		}
1273 #endif /* IPSEC */
1274 
1275 		default:
1276 			error = ENOPROTOOPT;
1277 			break;
1278 		}
1279 		break;
1280 	}
1281 	return (error);
1282 }
1283 
1284 /*
1285  * Routine called from ip_output() to loop back a copy of an IP multicast
1286  * packet to the input queue of a specified interface.  Note that this
1287  * calls the output routine of the loopback "driver", but with an interface
1288  * pointer that might NOT be a loopback interface -- evil, but easier than
1289  * replicating that code here.
1290  */
1291 static void
1292 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst,
1293     int hlen)
1294 {
1295 	register struct ip *ip;
1296 	struct mbuf *copym;
1297 
1298 	/*
1299 	 * Make a deep copy of the packet because we're going to
1300 	 * modify the pack in order to generate checksums.
1301 	 */
1302 	copym = m_dup(m, M_NOWAIT);
1303 	if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1304 		copym = m_pullup(copym, hlen);
1305 	if (copym != NULL) {
1306 		/* If needed, compute the checksum and mark it as valid. */
1307 		if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1308 			in_delayed_cksum(copym);
1309 			copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1310 			copym->m_pkthdr.csum_flags |=
1311 			    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1312 			copym->m_pkthdr.csum_data = 0xffff;
1313 		}
1314 		/*
1315 		 * We don't bother to fragment if the IP length is greater
1316 		 * than the interface's MTU.  Can this possibly matter?
1317 		 */
1318 		ip = mtod(copym, struct ip *);
1319 		ip->ip_sum = 0;
1320 		ip->ip_sum = in_cksum(copym, hlen);
1321 #if 1 /* XXX */
1322 		if (dst->sin_family != AF_INET) {
1323 			printf("ip_mloopback: bad address family %d\n",
1324 						dst->sin_family);
1325 			dst->sin_family = AF_INET;
1326 		}
1327 #endif
1328 		if_simloop(ifp, copym, dst->sin_family, 0);
1329 	}
1330 }
1331