xref: /openbsd-src/sys/netinet/ip_input.c (revision 25c4e8bd056e974b28f4a0ffd39d76c190a56013)
1 /*	$OpenBSD: ip_input.c,v 1.372 2022/06/29 09:01:48 mvs Exp $	*/
2 /*	$NetBSD: ip_input.c,v 1.30 1996/03/16 23:53:58 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
33  */
34 
35 #include "pf.h"
36 #include "carp.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/mbuf.h>
41 #include <sys/domain.h>
42 #include <sys/mutex.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/socketvar.h>
46 #include <sys/sysctl.h>
47 #include <sys/pool.h>
48 #include <sys/task.h>
49 
50 #include <net/if.h>
51 #include <net/if_var.h>
52 #include <net/if_dl.h>
53 #include <net/route.h>
54 #include <net/netisr.h>
55 
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/if_ether.h>
59 #include <netinet/ip.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_icmp.h>
64 #include <net/if_types.h>
65 
66 #ifdef INET6
67 #include <netinet6/ip6_var.h>
68 #endif
69 
70 #if NPF > 0
71 #include <net/pfvar.h>
72 #endif
73 
74 #ifdef MROUTING
75 #include <netinet/ip_mroute.h>
76 #endif
77 
78 #ifdef IPSEC
79 #include <netinet/ip_ipsp.h>
80 #endif /* IPSEC */
81 
82 #if NCARP > 0
83 #include <netinet/ip_carp.h>
84 #endif
85 
86 /* values controllable via sysctl */
87 int	ipforwarding = 0;
88 int	ipmforwarding = 0;
89 int	ipmultipath = 0;
90 int	ipsendredirects = 1;
91 int	ip_dosourceroute = 0;
92 int	ip_defttl = IPDEFTTL;
93 int	ip_mtudisc = 1;
94 int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
95 int	ip_directedbcast = 0;
96 
97 /* Protects `ipq' and `ip_frags'. */
98 struct mutex	ipq_mutex = MUTEX_INITIALIZER(IPL_SOFTNET);
99 
100 /* IP reassembly queue */
101 LIST_HEAD(, ipq) ipq;
102 
103 /* Keep track of memory used for reassembly */
104 int	ip_maxqueue = 300;
105 int	ip_frags = 0;
106 
107 const struct sysctl_bounded_args ipctl_vars[] = {
108 #ifdef MROUTING
109 	{ IPCTL_MRTPROTO, &ip_mrtproto, SYSCTL_INT_READONLY },
110 #endif
111 	{ IPCTL_FORWARDING, &ipforwarding, 0, 2 },
112 	{ IPCTL_SENDREDIRECTS, &ipsendredirects, 0, 1 },
113 	{ IPCTL_DEFTTL, &ip_defttl, 0, 255 },
114 	{ IPCTL_DIRECTEDBCAST, &ip_directedbcast, 0, 1 },
115 	{ IPCTL_IPPORT_FIRSTAUTO, &ipport_firstauto, 0, 65535 },
116 	{ IPCTL_IPPORT_LASTAUTO, &ipport_lastauto, 0, 65535 },
117 	{ IPCTL_IPPORT_HIFIRSTAUTO, &ipport_hifirstauto, 0, 65535 },
118 	{ IPCTL_IPPORT_HILASTAUTO, &ipport_hilastauto, 0, 65535 },
119 	{ IPCTL_IPPORT_MAXQUEUE, &ip_maxqueue, 0, 10000 },
120 	{ IPCTL_MFORWARDING, &ipmforwarding, 0, 1 },
121 	{ IPCTL_MULTIPATH, &ipmultipath, 0, 1 },
122 	{ IPCTL_ARPTIMEOUT, &arpt_keep, 0, INT_MAX },
123 	{ IPCTL_ARPDOWN, &arpt_down, 0, INT_MAX },
124 };
125 
126 struct niqueue ipintrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IP);
127 
128 struct pool ipqent_pool;
129 struct pool ipq_pool;
130 
131 struct cpumem *ipcounters;
132 
133 int ip_sysctl_ipstat(void *, size_t *, void *);
134 
135 static struct mbuf_queue	ipsend_mq;
136 static struct mbuf_queue	ipsendraw_mq;
137 
138 extern struct niqueue		arpinq;
139 
140 int	ip_ours(struct mbuf **, int *, int, int);
141 int	ip_local(struct mbuf **, int *, int, int);
142 int	ip_dooptions(struct mbuf *, struct ifnet *);
143 int	in_ouraddr(struct mbuf *, struct ifnet *, struct rtentry **);
144 
145 static void ip_send_dispatch(void *);
146 static void ip_sendraw_dispatch(void *);
147 static struct task ipsend_task = TASK_INITIALIZER(ip_send_dispatch, &ipsend_mq);
148 static struct task ipsendraw_task =
149 	TASK_INITIALIZER(ip_sendraw_dispatch, &ipsendraw_mq);
150 
151 /*
152  * Used to save the IP options in case a protocol wants to respond
153  * to an incoming packet over the same route if the packet got here
154  * using IP source routing.  This allows connection establishment and
155  * maintenance when the remote end is on a network that is not known
156  * to us.
157  */
158 struct ip_srcrt {
159 	int		isr_nhops;		   /* number of hops */
160 	struct in_addr	isr_dst;		   /* final destination */
161 	char		isr_nop;		   /* one NOP to align */
162 	char		isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
163 	struct in_addr	isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
164 };
165 
166 void save_rte(struct mbuf *, u_char *, struct in_addr);
167 
168 /*
169  * IP initialization: fill in IP protocol switch table.
170  * All protocols not implemented in kernel go to raw IP protocol handler.
171  */
172 void
173 ip_init(void)
174 {
175 	const struct protosw *pr;
176 	int i;
177 	const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
178 	const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
179 	const u_int16_t defrootonlyports_tcp[] = DEFROOTONLYPORTS_TCP;
180 	const u_int16_t defrootonlyports_udp[] = DEFROOTONLYPORTS_UDP;
181 
182 	ipcounters = counters_alloc(ips_ncounters);
183 
184 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0,
185 	    IPL_SOFTNET, 0, "ipqe",  NULL);
186 	pool_init(&ipq_pool, sizeof(struct ipq), 0,
187 	    IPL_SOFTNET, 0, "ipq", NULL);
188 
189 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
190 	if (pr == NULL)
191 		panic("ip_init");
192 	for (i = 0; i < IPPROTO_MAX; i++)
193 		ip_protox[i] = pr - inetsw;
194 	for (pr = inetdomain.dom_protosw;
195 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
196 		if (pr->pr_domain->dom_family == PF_INET &&
197 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
198 		    pr->pr_protocol < IPPROTO_MAX)
199 			ip_protox[pr->pr_protocol] = pr - inetsw;
200 	LIST_INIT(&ipq);
201 
202 	/* Fill in list of ports not to allocate dynamically. */
203 	memset(&baddynamicports, 0, sizeof(baddynamicports));
204 	for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
205 		DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
206 	for (i = 0; defbaddynamicports_udp[i] != 0; i++)
207 		DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
208 
209 	/* Fill in list of ports only root can bind to. */
210 	memset(&rootonlyports, 0, sizeof(rootonlyports));
211 	for (i = 0; defrootonlyports_tcp[i] != 0; i++)
212 		DP_SET(rootonlyports.tcp, defrootonlyports_tcp[i]);
213 	for (i = 0; defrootonlyports_udp[i] != 0; i++)
214 		DP_SET(rootonlyports.udp, defrootonlyports_udp[i]);
215 
216 	mq_init(&ipsend_mq, 64, IPL_SOFTNET);
217 	mq_init(&ipsendraw_mq, 64, IPL_SOFTNET);
218 
219 	arpinit();
220 #ifdef IPSEC
221 	ipsec_init();
222 #endif
223 #ifdef MROUTING
224 	rt_timer_queue_init(&ip_mrouterq, MCAST_EXPIRE_FREQUENCY,
225 	    &mfc_expire_route);
226 #endif
227 }
228 
229 /*
230  * Enqueue packet for local delivery.  Queuing is used as a boundary
231  * between the network layer (input/forward path) running with shared
232  * NET_RLOCK_IN_SOFTNET() and the transport layer needing it exclusively.
233  */
234 int
235 ip_ours(struct mbuf **mp, int *offp, int nxt, int af)
236 {
237 	/* We are already in a IPv4/IPv6 local deliver loop. */
238 	if (af != AF_UNSPEC)
239 		return ip_local(mp, offp, nxt, af);
240 
241 	niq_enqueue(&ipintrq, *mp);
242 	*mp = NULL;
243 	return IPPROTO_DONE;
244 }
245 
246 /*
247  * Dequeue and process locally delivered packets.
248  */
249 void
250 ipintr(void)
251 {
252 	struct mbuf *m;
253 	int off, nxt;
254 
255 	while ((m = niq_dequeue(&ipintrq)) != NULL) {
256 #ifdef DIAGNOSTIC
257 		if ((m->m_flags & M_PKTHDR) == 0)
258 			panic("ipintr no HDR");
259 #endif
260 		off = 0;
261 		nxt = ip_local(&m, &off, IPPROTO_IPV4, AF_UNSPEC);
262 		KASSERT(nxt == IPPROTO_DONE);
263 	}
264 }
265 
266 /*
267  * IPv4 input routine.
268  *
269  * Checksum and byte swap header.  Process options. Forward or deliver.
270  */
271 void
272 ipv4_input(struct ifnet *ifp, struct mbuf *m)
273 {
274 	int off, nxt;
275 
276 	off = 0;
277 	nxt = ip_input_if(&m, &off, IPPROTO_IPV4, AF_UNSPEC, ifp);
278 	KASSERT(nxt == IPPROTO_DONE);
279 }
280 
281 struct mbuf *
282 ipv4_check(struct ifnet *ifp, struct mbuf *m)
283 {
284 	struct ip *ip;
285 	int hlen, len;
286 
287 	if (m->m_len < sizeof(*ip)) {
288 		m = m_pullup(m, sizeof(*ip));
289 		if (m == NULL) {
290 			ipstat_inc(ips_toosmall);
291 			return (NULL);
292 		}
293 	}
294 
295 	ip = mtod(m, struct ip *);
296 	if (ip->ip_v != IPVERSION) {
297 		ipstat_inc(ips_badvers);
298 		goto bad;
299 	}
300 
301 	hlen = ip->ip_hl << 2;
302 	if (hlen < sizeof(*ip)) {	/* minimum header length */
303 		ipstat_inc(ips_badhlen);
304 		goto bad;
305 	}
306 	if (hlen > m->m_len) {
307 		m = m_pullup(m, hlen);
308 		if (m == NULL) {
309 			ipstat_inc(ips_badhlen);
310 			return (NULL);
311 		}
312 		ip = mtod(m, struct ip *);
313 	}
314 
315 	/* 127/8 must not appear on wire - RFC1122 */
316 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
317 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
318 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
319 			ipstat_inc(ips_badaddr);
320 			goto bad;
321 		}
322 	}
323 
324 	if (!ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK)) {
325 		if (ISSET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_BAD)) {
326 			ipstat_inc(ips_badsum);
327 			goto bad;
328 		}
329 
330 		ipstat_inc(ips_inswcsum);
331 		if (in_cksum(m, hlen) != 0) {
332 			ipstat_inc(ips_badsum);
333 			goto bad;
334 		}
335 
336 		SET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_IN_OK);
337 	}
338 
339 	/* Retrieve the packet length. */
340 	len = ntohs(ip->ip_len);
341 
342 	/*
343 	 * Convert fields to host representation.
344 	 */
345 	if (len < hlen) {
346 		ipstat_inc(ips_badlen);
347 		goto bad;
348 	}
349 
350 	/*
351 	 * Check that the amount of data in the buffers
352 	 * is at least as much as the IP header would have us expect.
353 	 * Trim mbufs if longer than we expect.
354 	 * Drop packet if shorter than we expect.
355 	 */
356 	if (m->m_pkthdr.len < len) {
357 		ipstat_inc(ips_tooshort);
358 		goto bad;
359 	}
360 	if (m->m_pkthdr.len > len) {
361 		if (m->m_len == m->m_pkthdr.len) {
362 			m->m_len = len;
363 			m->m_pkthdr.len = len;
364 		} else
365 			m_adj(m, len - m->m_pkthdr.len);
366 	}
367 
368 	return (m);
369 bad:
370 	m_freem(m);
371 	return (NULL);
372 }
373 
374 int
375 ip_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
376 {
377 	struct mbuf	*m;
378 	struct rtentry	*rt = NULL;
379 	struct ip	*ip;
380 	int hlen;
381 	in_addr_t pfrdr = 0;
382 
383 	KASSERT(*offp == 0);
384 
385 	ipstat_inc(ips_total);
386 	m = *mp = ipv4_check(ifp, *mp);
387 	if (m == NULL)
388 		goto bad;
389 
390 	ip = mtod(m, struct ip *);
391 
392 #if NCARP > 0
393 	if (carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
394 	    &ip->ip_dst.s_addr, (ip->ip_p == IPPROTO_ICMP ? 0 : 1)))
395 		goto bad;
396 #endif
397 
398 #if NPF > 0
399 	/*
400 	 * Packet filter
401 	 */
402 	pfrdr = ip->ip_dst.s_addr;
403 	if (pf_test(AF_INET, PF_IN, ifp, mp) != PF_PASS)
404 		goto bad;
405 	m = *mp;
406 	if (m == NULL)
407 		goto bad;
408 
409 	ip = mtod(m, struct ip *);
410 	pfrdr = (pfrdr != ip->ip_dst.s_addr);
411 #endif
412 
413 	hlen = ip->ip_hl << 2;
414 
415 	/*
416 	 * Process options and, if not destined for us,
417 	 * ship it on.  ip_dooptions returns 1 when an
418 	 * error was detected (causing an icmp message
419 	 * to be sent and the original packet to be freed).
420 	 */
421 	if (hlen > sizeof (struct ip) && ip_dooptions(m, ifp)) {
422 		m = *mp = NULL;
423 		goto bad;
424 	}
425 
426 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
427 	    ip->ip_dst.s_addr == INADDR_ANY) {
428 		nxt = ip_ours(mp, offp, nxt, af);
429 		goto out;
430 	}
431 
432 	switch(in_ouraddr(m, ifp, &rt)) {
433 	case 2:
434 		goto bad;
435 	case 1:
436 		nxt = ip_ours(mp, offp, nxt, af);
437 		goto out;
438 	}
439 
440 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
441 		/*
442 		 * Make sure M_MCAST is set.  It should theoretically
443 		 * already be there, but let's play safe because upper
444 		 * layers check for this flag.
445 		 */
446 		m->m_flags |= M_MCAST;
447 
448 #ifdef MROUTING
449 		if (ipmforwarding && ip_mrouter[ifp->if_rdomain]) {
450 			int error;
451 
452 			if (m->m_flags & M_EXT) {
453 				if ((m = *mp = m_pullup(m, hlen)) == NULL) {
454 					ipstat_inc(ips_toosmall);
455 					goto bad;
456 				}
457 				ip = mtod(m, struct ip *);
458 			}
459 			/*
460 			 * If we are acting as a multicast router, all
461 			 * incoming multicast packets are passed to the
462 			 * kernel-level multicast forwarding function.
463 			 * The packet is returned (relatively) intact; if
464 			 * ip_mforward() returns a non-zero value, the packet
465 			 * must be discarded, else it may be accepted below.
466 			 *
467 			 * (The IP ident field is put in the same byte order
468 			 * as expected when ip_mforward() is called from
469 			 * ip_output().)
470 			 */
471 			KERNEL_LOCK();
472 			error = ip_mforward(m, ifp);
473 			KERNEL_UNLOCK();
474 			if (error) {
475 				ipstat_inc(ips_cantforward);
476 				goto bad;
477 			}
478 
479 			/*
480 			 * The process-level routing daemon needs to receive
481 			 * all multicast IGMP packets, whether or not this
482 			 * host belongs to their destination groups.
483 			 */
484 			if (ip->ip_p == IPPROTO_IGMP) {
485 				nxt = ip_ours(mp, offp, nxt, af);
486 				goto out;
487 			}
488 			ipstat_inc(ips_forward);
489 		}
490 #endif
491 		/*
492 		 * See if we belong to the destination multicast group on the
493 		 * arrival interface.
494 		 */
495 		if (!in_hasmulti(&ip->ip_dst, ifp)) {
496 			ipstat_inc(ips_notmember);
497 			if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr))
498 				ipstat_inc(ips_cantforward);
499 			goto bad;
500 		}
501 		nxt = ip_ours(mp, offp, nxt, af);
502 		goto out;
503 	}
504 
505 #if NCARP > 0
506 	if (ip->ip_p == IPPROTO_ICMP &&
507 	    carp_lsdrop(ifp, m, AF_INET, &ip->ip_src.s_addr,
508 	    &ip->ip_dst.s_addr, 1))
509 		goto bad;
510 #endif
511 	/*
512 	 * Not for us; forward if possible and desirable.
513 	 */
514 	if (ipforwarding == 0) {
515 		ipstat_inc(ips_cantforward);
516 		goto bad;
517 	}
518 #ifdef IPSEC
519 	if (ipsec_in_use) {
520 		int rv;
521 
522 		rv = ipsec_forward_check(m, hlen, AF_INET);
523 		if (rv != 0) {
524 			ipstat_inc(ips_cantforward);
525 			goto bad;
526 		}
527 		/*
528 		 * Fall through, forward packet. Outbound IPsec policy
529 		 * checking will occur in ip_output().
530 		 */
531 	}
532 #endif /* IPSEC */
533 
534 	ip_forward(m, ifp, rt, pfrdr);
535 	*mp = NULL;
536 	return IPPROTO_DONE;
537  bad:
538 	nxt = IPPROTO_DONE;
539 	m_freemp(mp);
540  out:
541 	rtfree(rt);
542 	return nxt;
543 }
544 
545 /*
546  * IPv4 local-delivery routine.
547  *
548  * If fragmented try to reassemble.  Pass to next level.
549  */
550 int
551 ip_local(struct mbuf **mp, int *offp, int nxt, int af)
552 {
553 	struct mbuf *m = *mp;
554 	struct ip *ip = mtod(m, struct ip *);
555 	struct ipq *fp;
556 	struct ipqent *ipqe;
557 	int mff, hlen;
558 
559 	NET_ASSERT_WLOCKED();
560 
561 	hlen = ip->ip_hl << 2;
562 
563 	/*
564 	 * If offset or IP_MF are set, must reassemble.
565 	 * Otherwise, nothing need be done.
566 	 * (We could look in the reassembly queue to see
567 	 * if the packet was previously fragmented,
568 	 * but it's not worth the time; just let them time out.)
569 	 */
570 	if (ip->ip_off &~ htons(IP_DF | IP_RF)) {
571 		if (m->m_flags & M_EXT) {		/* XXX */
572 			if ((m = *mp = m_pullup(m, hlen)) == NULL) {
573 				ipstat_inc(ips_toosmall);
574 				return IPPROTO_DONE;
575 			}
576 			ip = mtod(m, struct ip *);
577 		}
578 
579 		mtx_enter(&ipq_mutex);
580 
581 		/*
582 		 * Look for queue of fragments
583 		 * of this datagram.
584 		 */
585 		LIST_FOREACH(fp, &ipq, ipq_q) {
586 			if (ip->ip_id == fp->ipq_id &&
587 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
588 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
589 			    ip->ip_p == fp->ipq_p)
590 				break;
591 		}
592 
593 		/*
594 		 * Adjust ip_len to not reflect header,
595 		 * set ipqe_mff if more fragments are expected,
596 		 * convert offset of this to bytes.
597 		 */
598 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
599 		mff = (ip->ip_off & htons(IP_MF)) != 0;
600 		if (mff) {
601 			/*
602 			 * Make sure that fragments have a data length
603 			 * that's a non-zero multiple of 8 bytes.
604 			 */
605 			if (ntohs(ip->ip_len) == 0 ||
606 			    (ntohs(ip->ip_len) & 0x7) != 0) {
607 				ipstat_inc(ips_badfrags);
608 				goto bad;
609 			}
610 		}
611 		ip->ip_off = htons(ntohs(ip->ip_off) << 3);
612 
613 		/*
614 		 * If datagram marked as having more fragments
615 		 * or if this is not the first fragment,
616 		 * attempt reassembly; if it succeeds, proceed.
617 		 */
618 		if (mff || ip->ip_off) {
619 			ipstat_inc(ips_fragments);
620 			if (ip_frags + 1 > ip_maxqueue) {
621 				ip_flush();
622 				ipstat_inc(ips_rcvmemdrop);
623 				goto bad;
624 			}
625 
626 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
627 			if (ipqe == NULL) {
628 				ipstat_inc(ips_rcvmemdrop);
629 				goto bad;
630 			}
631 			ip_frags++;
632 			ipqe->ipqe_mff = mff;
633 			ipqe->ipqe_m = m;
634 			ipqe->ipqe_ip = ip;
635 			m = *mp = ip_reass(ipqe, fp);
636 			if (m == NULL)
637 				goto bad;
638 			ipstat_inc(ips_reassembled);
639 			ip = mtod(m, struct ip *);
640 			hlen = ip->ip_hl << 2;
641 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
642 		} else
643 			if (fp)
644 				ip_freef(fp);
645 
646 		mtx_leave(&ipq_mutex);
647 	}
648 
649 	*offp = hlen;
650 	nxt = ip->ip_p;
651 	/* Check whether we are already in a IPv4/IPv6 local deliver loop. */
652 	if (af == AF_UNSPEC)
653 		nxt = ip_deliver(mp, offp, nxt, AF_INET);
654 	return nxt;
655  bad:
656 	mtx_leave(&ipq_mutex);
657 	m_freemp(mp);
658 	return IPPROTO_DONE;
659 }
660 
661 #ifndef INET6
662 #define IPSTAT_INC(name)	ipstat_inc(ips_##name)
663 #else
664 #define IPSTAT_INC(name)	(af == AF_INET ?	\
665     ipstat_inc(ips_##name) : ip6stat_inc(ip6s_##name))
666 #endif
667 
668 int
669 ip_deliver(struct mbuf **mp, int *offp, int nxt, int af)
670 {
671 	const struct protosw *psw;
672 	int naf = af;
673 #ifdef INET6
674 	int nest = 0;
675 #endif /* INET6 */
676 
677 	/* pf might have modified stuff, might have to chksum */
678 	switch (af) {
679 	case AF_INET:
680 		in_proto_cksum_out(*mp, NULL);
681 		break;
682 #ifdef INET6
683 	case AF_INET6:
684 		in6_proto_cksum_out(*mp, NULL);
685 		break;
686 #endif /* INET6 */
687 	}
688 
689 	/*
690 	 * Tell launch routine the next header
691 	 */
692 	IPSTAT_INC(delivered);
693 
694 	while (nxt != IPPROTO_DONE) {
695 #ifdef INET6
696 		if (af == AF_INET6 &&
697 		    ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
698 			ip6stat_inc(ip6s_toomanyhdr);
699 			goto bad;
700 		}
701 #endif /* INET6 */
702 
703 		/*
704 		 * protection against faulty packet - there should be
705 		 * more sanity checks in header chain processing.
706 		 */
707 		if ((*mp)->m_pkthdr.len < *offp) {
708 			IPSTAT_INC(tooshort);
709 			goto bad;
710 		}
711 
712 #ifdef IPSEC
713 		if (ipsec_in_use) {
714 			if (ipsec_local_check(*mp, *offp, nxt, af) != 0) {
715 				IPSTAT_INC(cantforward);
716 				goto bad;
717 			}
718 		}
719 		/* Otherwise, just fall through and deliver the packet */
720 #endif /* IPSEC */
721 
722 		switch (nxt) {
723 		case IPPROTO_IPV4:
724 			naf = AF_INET;
725 			ipstat_inc(ips_delivered);
726 			break;
727 #ifdef INET6
728 		case IPPROTO_IPV6:
729 			naf = AF_INET6;
730 			ip6stat_inc(ip6s_delivered);
731 			break;
732 #endif /* INET6 */
733 		}
734 		switch (af) {
735 		case AF_INET:
736 			psw = &inetsw[ip_protox[nxt]];
737 			break;
738 #ifdef INET6
739 		case AF_INET6:
740 			psw = &inet6sw[ip6_protox[nxt]];
741 			break;
742 #endif /* INET6 */
743 		}
744 		nxt = (*psw->pr_input)(mp, offp, nxt, af);
745 		af = naf;
746 	}
747 	return nxt;
748  bad:
749 	m_freemp(mp);
750 	return IPPROTO_DONE;
751 }
752 #undef IPSTAT_INC
753 
754 int
755 in_ouraddr(struct mbuf *m, struct ifnet *ifp, struct rtentry **prt)
756 {
757 	struct rtentry		*rt;
758 	struct ip		*ip;
759 	struct sockaddr_in	 sin;
760 	int			 match = 0;
761 
762 #if NPF > 0
763 	switch (pf_ouraddr(m)) {
764 	case 0:
765 		return (0);
766 	case 1:
767 		return (1);
768 	default:
769 		/* pf does not know it */
770 		break;
771 	}
772 #endif
773 
774 	ip = mtod(m, struct ip *);
775 
776 	memset(&sin, 0, sizeof(sin));
777 	sin.sin_len = sizeof(sin);
778 	sin.sin_family = AF_INET;
779 	sin.sin_addr = ip->ip_dst;
780 	rt = rtalloc_mpath(sintosa(&sin), &ip->ip_src.s_addr,
781 	    m->m_pkthdr.ph_rtableid);
782 	if (rtisvalid(rt)) {
783 		if (ISSET(rt->rt_flags, RTF_LOCAL))
784 			match = 1;
785 
786 		/*
787 		 * If directedbcast is enabled we only consider it local
788 		 * if it is received on the interface with that address.
789 		 */
790 		if (ISSET(rt->rt_flags, RTF_BROADCAST) &&
791 		    (!ip_directedbcast || rt->rt_ifidx == ifp->if_index)) {
792 			match = 1;
793 
794 			/* Make sure M_BCAST is set */
795 			m->m_flags |= M_BCAST;
796 		}
797 	}
798 	*prt = rt;
799 
800 	if (!match) {
801 		struct ifaddr *ifa;
802 
803 		/*
804 		 * No local address or broadcast address found, so check for
805 		 * ancient classful broadcast addresses.
806 		 * It must have been broadcast on the link layer, and for an
807 		 * address on the interface it was received on.
808 		 */
809 		if (!ISSET(m->m_flags, M_BCAST) ||
810 		    !IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, ip->ip_dst.s_addr))
811 			return (0);
812 
813 		if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid))
814 			return (0);
815 		/*
816 		 * The check in the loop assumes you only rx a packet on an UP
817 		 * interface, and that M_BCAST will only be set on a BROADCAST
818 		 * interface.
819 		 */
820 		NET_ASSERT_LOCKED();
821 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
822 			if (ifa->ifa_addr->sa_family != AF_INET)
823 				continue;
824 
825 			if (IN_CLASSFULBROADCAST(ip->ip_dst.s_addr,
826 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr)) {
827 				match = 1;
828 				break;
829 			}
830 		}
831 	} else if (ipforwarding == 0 && rt->rt_ifidx != ifp->if_index &&
832 	    !((ifp->if_flags & IFF_LOOPBACK) || (ifp->if_type == IFT_ENC) ||
833 	    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
834 		/* received on wrong interface. */
835 #if NCARP > 0
836 		struct ifnet *out_if;
837 
838 		/*
839 		 * Virtual IPs on carp interfaces need to be checked also
840 		 * against the parent interface and other carp interfaces
841 		 * sharing the same parent.
842 		 */
843 		out_if = if_get(rt->rt_ifidx);
844 		if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
845 			ipstat_inc(ips_wrongif);
846 			match = 2;
847 		}
848 		if_put(out_if);
849 #else
850 		ipstat_inc(ips_wrongif);
851 		match = 2;
852 #endif
853 	}
854 
855 	return (match);
856 }
857 
858 /*
859  * Take incoming datagram fragment and try to
860  * reassemble it into whole datagram.  If a chain for
861  * reassembly of this datagram already exists, then it
862  * is given as fp; otherwise have to make a chain.
863  */
864 struct mbuf *
865 ip_reass(struct ipqent *ipqe, struct ipq *fp)
866 {
867 	struct mbuf *m = ipqe->ipqe_m;
868 	struct ipqent *nq, *p, *q;
869 	struct ip *ip;
870 	struct mbuf *t;
871 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
872 	int i, next;
873 	u_int8_t ecn, ecn0;
874 
875 	MUTEX_ASSERT_LOCKED(&ipq_mutex);
876 
877 	/*
878 	 * Presence of header sizes in mbufs
879 	 * would confuse code below.
880 	 */
881 	m->m_data += hlen;
882 	m->m_len -= hlen;
883 
884 	/*
885 	 * If first fragment to arrive, create a reassembly queue.
886 	 */
887 	if (fp == NULL) {
888 		fp = pool_get(&ipq_pool, PR_NOWAIT);
889 		if (fp == NULL)
890 			goto dropfrag;
891 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
892 		fp->ipq_ttl = IPFRAGTTL;
893 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
894 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
895 		LIST_INIT(&fp->ipq_fragq);
896 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
897 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
898 		p = NULL;
899 		goto insert;
900 	}
901 
902 	/*
903 	 * Handle ECN by comparing this segment with the first one;
904 	 * if CE is set, do not lose CE.
905 	 * drop if CE and not-ECT are mixed for the same packet.
906 	 */
907 	ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
908 	ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
909 	if (ecn == IPTOS_ECN_CE) {
910 		if (ecn0 == IPTOS_ECN_NOTECT)
911 			goto dropfrag;
912 		if (ecn0 != IPTOS_ECN_CE)
913 			LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |=
914 			    IPTOS_ECN_CE;
915 	}
916 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
917 		goto dropfrag;
918 
919 	/*
920 	 * Find a segment which begins after this one does.
921 	 */
922 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
923 	    p = q, q = LIST_NEXT(q, ipqe_q))
924 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
925 			break;
926 
927 	/*
928 	 * If there is a preceding segment, it may provide some of
929 	 * our data already.  If so, drop the data from the incoming
930 	 * segment.  If it provides all of our data, drop us.
931 	 */
932 	if (p != NULL) {
933 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
934 		    ntohs(ipqe->ipqe_ip->ip_off);
935 		if (i > 0) {
936 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
937 				goto dropfrag;
938 			m_adj(ipqe->ipqe_m, i);
939 			ipqe->ipqe_ip->ip_off =
940 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
941 			ipqe->ipqe_ip->ip_len =
942 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
943 		}
944 	}
945 
946 	/*
947 	 * While we overlap succeeding segments trim them or,
948 	 * if they are completely covered, dequeue them.
949 	 */
950 	for (; q != NULL &&
951 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
952 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
953 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
954 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
955 		if (i < ntohs(q->ipqe_ip->ip_len)) {
956 			q->ipqe_ip->ip_len =
957 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
958 			q->ipqe_ip->ip_off =
959 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
960 			m_adj(q->ipqe_m, i);
961 			break;
962 		}
963 		nq = LIST_NEXT(q, ipqe_q);
964 		m_freem(q->ipqe_m);
965 		LIST_REMOVE(q, ipqe_q);
966 		pool_put(&ipqent_pool, q);
967 		ip_frags--;
968 	}
969 
970 insert:
971 	/*
972 	 * Stick new segment in its place;
973 	 * check for complete reassembly.
974 	 */
975 	if (p == NULL) {
976 		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
977 	} else {
978 		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
979 	}
980 	next = 0;
981 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
982 	    p = q, q = LIST_NEXT(q, ipqe_q)) {
983 		if (ntohs(q->ipqe_ip->ip_off) != next)
984 			return (0);
985 		next += ntohs(q->ipqe_ip->ip_len);
986 	}
987 	if (p->ipqe_mff)
988 		return (0);
989 
990 	/*
991 	 * Reassembly is complete.  Check for a bogus message size and
992 	 * concatenate fragments.
993 	 */
994 	q = LIST_FIRST(&fp->ipq_fragq);
995 	ip = q->ipqe_ip;
996 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
997 		ipstat_inc(ips_toolong);
998 		ip_freef(fp);
999 		return (0);
1000 	}
1001 	m = q->ipqe_m;
1002 	t = m->m_next;
1003 	m->m_next = 0;
1004 	m_cat(m, t);
1005 	nq = LIST_NEXT(q, ipqe_q);
1006 	pool_put(&ipqent_pool, q);
1007 	ip_frags--;
1008 	for (q = nq; q != NULL; q = nq) {
1009 		t = q->ipqe_m;
1010 		nq = LIST_NEXT(q, ipqe_q);
1011 		pool_put(&ipqent_pool, q);
1012 		ip_frags--;
1013 		m_removehdr(t);
1014 		m_cat(m, t);
1015 	}
1016 
1017 	/*
1018 	 * Create header for new ip packet by
1019 	 * modifying header of first packet;
1020 	 * dequeue and discard fragment reassembly header.
1021 	 * Make header visible.
1022 	 */
1023 	ip->ip_len = htons(next);
1024 	ip->ip_src = fp->ipq_src;
1025 	ip->ip_dst = fp->ipq_dst;
1026 	LIST_REMOVE(fp, ipq_q);
1027 	pool_put(&ipq_pool, fp);
1028 	m->m_len += (ip->ip_hl << 2);
1029 	m->m_data -= (ip->ip_hl << 2);
1030 	m_calchdrlen(m);
1031 	return (m);
1032 
1033 dropfrag:
1034 	ipstat_inc(ips_fragdropped);
1035 	m_freem(m);
1036 	pool_put(&ipqent_pool, ipqe);
1037 	ip_frags--;
1038 	return (NULL);
1039 }
1040 
1041 /*
1042  * Free a fragment reassembly header and all
1043  * associated datagrams.
1044  */
1045 void
1046 ip_freef(struct ipq *fp)
1047 {
1048 	struct ipqent *q;
1049 
1050 	MUTEX_ASSERT_LOCKED(&ipq_mutex);
1051 
1052 	while ((q = LIST_FIRST(&fp->ipq_fragq)) != NULL) {
1053 		LIST_REMOVE(q, ipqe_q);
1054 		m_freem(q->ipqe_m);
1055 		pool_put(&ipqent_pool, q);
1056 		ip_frags--;
1057 	}
1058 	LIST_REMOVE(fp, ipq_q);
1059 	pool_put(&ipq_pool, fp);
1060 }
1061 
1062 /*
1063  * IP timer processing;
1064  * if a timer expires on a reassembly queue, discard it.
1065  */
1066 void
1067 ip_slowtimo(void)
1068 {
1069 	struct ipq *fp, *nfp;
1070 
1071 	mtx_enter(&ipq_mutex);
1072 	LIST_FOREACH_SAFE(fp, &ipq, ipq_q, nfp) {
1073 		if (--fp->ipq_ttl == 0) {
1074 			ipstat_inc(ips_fragtimeout);
1075 			ip_freef(fp);
1076 		}
1077 	}
1078 	mtx_leave(&ipq_mutex);
1079 }
1080 
1081 /*
1082  * Flush a bunch of datagram fragments, till we are down to 75%.
1083  */
1084 void
1085 ip_flush(void)
1086 {
1087 	int max = 50;
1088 
1089 	MUTEX_ASSERT_LOCKED(&ipq_mutex);
1090 
1091 	while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) {
1092 		ipstat_inc(ips_fragdropped);
1093 		ip_freef(LIST_FIRST(&ipq));
1094 	}
1095 }
1096 
1097 /*
1098  * Do option processing on a datagram,
1099  * possibly discarding it if bad options are encountered,
1100  * or forwarding it if source-routed.
1101  * Returns 1 if packet has been forwarded/freed,
1102  * 0 if the packet should be processed further.
1103  */
1104 int
1105 ip_dooptions(struct mbuf *m, struct ifnet *ifp)
1106 {
1107 	struct ip *ip = mtod(m, struct ip *);
1108 	unsigned int rtableid = m->m_pkthdr.ph_rtableid;
1109 	struct rtentry *rt;
1110 	struct sockaddr_in ipaddr;
1111 	u_char *cp;
1112 	struct ip_timestamp ipt;
1113 	struct in_ifaddr *ia;
1114 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1115 	struct in_addr sin, dst;
1116 	u_int32_t ntime;
1117 
1118 	dst = ip->ip_dst;
1119 	cp = (u_char *)(ip + 1);
1120 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1121 
1122 	KERNEL_LOCK();
1123 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1124 		opt = cp[IPOPT_OPTVAL];
1125 		if (opt == IPOPT_EOL)
1126 			break;
1127 		if (opt == IPOPT_NOP)
1128 			optlen = 1;
1129 		else {
1130 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1131 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1132 				goto bad;
1133 			}
1134 			optlen = cp[IPOPT_OLEN];
1135 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1136 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1137 				goto bad;
1138 			}
1139 		}
1140 
1141 		switch (opt) {
1142 
1143 		default:
1144 			break;
1145 
1146 		/*
1147 		 * Source routing with record.
1148 		 * Find interface with current destination address.
1149 		 * If none on this machine then drop if strictly routed,
1150 		 * or do nothing if loosely routed.
1151 		 * Record interface address and bring up next address
1152 		 * component.  If strictly routed make sure next
1153 		 * address is on directly accessible net.
1154 		 */
1155 		case IPOPT_LSRR:
1156 		case IPOPT_SSRR:
1157 			if (!ip_dosourceroute) {
1158 				type = ICMP_UNREACH;
1159 				code = ICMP_UNREACH_SRCFAIL;
1160 				goto bad;
1161 			}
1162 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1163 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1164 				goto bad;
1165 			}
1166 			memset(&ipaddr, 0, sizeof(ipaddr));
1167 			ipaddr.sin_family = AF_INET;
1168 			ipaddr.sin_len = sizeof(ipaddr);
1169 			ipaddr.sin_addr = ip->ip_dst;
1170 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr),
1171 			    m->m_pkthdr.ph_rtableid));
1172 			if (ia == NULL) {
1173 				if (opt == IPOPT_SSRR) {
1174 					type = ICMP_UNREACH;
1175 					code = ICMP_UNREACH_SRCFAIL;
1176 					goto bad;
1177 				}
1178 				/*
1179 				 * Loose routing, and not at next destination
1180 				 * yet; nothing to do except forward.
1181 				 */
1182 				break;
1183 			}
1184 			off--;			/* 0 origin */
1185 			if ((off + sizeof(struct in_addr)) > optlen) {
1186 				/*
1187 				 * End of source route.  Should be for us.
1188 				 */
1189 				save_rte(m, cp, ip->ip_src);
1190 				break;
1191 			}
1192 
1193 			/*
1194 			 * locate outgoing interface
1195 			 */
1196 			memset(&ipaddr, 0, sizeof(ipaddr));
1197 			ipaddr.sin_family = AF_INET;
1198 			ipaddr.sin_len = sizeof(ipaddr);
1199 			memcpy(&ipaddr.sin_addr, cp + off,
1200 			    sizeof(ipaddr.sin_addr));
1201 			/* keep packet in the virtual instance */
1202 			rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1203 			if (!rtisvalid(rt) || ((opt == IPOPT_SSRR) &&
1204 			    ISSET(rt->rt_flags, RTF_GATEWAY))) {
1205 				type = ICMP_UNREACH;
1206 				code = ICMP_UNREACH_SRCFAIL;
1207 				rtfree(rt);
1208 				goto bad;
1209 			}
1210 			ia = ifatoia(rt->rt_ifa);
1211 			memcpy(cp + off, &ia->ia_addr.sin_addr,
1212 			    sizeof(struct in_addr));
1213 			rtfree(rt);
1214 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1215 			ip->ip_dst = ipaddr.sin_addr;
1216 			/*
1217 			 * Let ip_intr's mcast routing check handle mcast pkts
1218 			 */
1219 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1220 			break;
1221 
1222 		case IPOPT_RR:
1223 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1224 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1225 				goto bad;
1226 			}
1227 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1228 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1229 				goto bad;
1230 			}
1231 
1232 			/*
1233 			 * If no space remains, ignore.
1234 			 */
1235 			off--;			/* 0 origin */
1236 			if ((off + sizeof(struct in_addr)) > optlen)
1237 				break;
1238 			memset(&ipaddr, 0, sizeof(ipaddr));
1239 			ipaddr.sin_family = AF_INET;
1240 			ipaddr.sin_len = sizeof(ipaddr);
1241 			ipaddr.sin_addr = ip->ip_dst;
1242 			/*
1243 			 * locate outgoing interface; if we're the destination,
1244 			 * use the incoming interface (should be same).
1245 			 * Again keep the packet inside the virtual instance.
1246 			 */
1247 			rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1248 			if (!rtisvalid(rt)) {
1249 				type = ICMP_UNREACH;
1250 				code = ICMP_UNREACH_HOST;
1251 				rtfree(rt);
1252 				goto bad;
1253 			}
1254 			ia = ifatoia(rt->rt_ifa);
1255 			memcpy(cp + off, &ia->ia_addr.sin_addr,
1256 			    sizeof(struct in_addr));
1257 			rtfree(rt);
1258 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1259 			break;
1260 
1261 		case IPOPT_TS:
1262 			code = cp - (u_char *)ip;
1263 			if (optlen < sizeof(struct ip_timestamp))
1264 				goto bad;
1265 			memcpy(&ipt, cp, sizeof(struct ip_timestamp));
1266 			if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1267 				goto bad;
1268 			if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) > ipt.ipt_len) {
1269 				if (++ipt.ipt_oflw == 0)
1270 					goto bad;
1271 				break;
1272 			}
1273 			memcpy(&sin, cp + ipt.ipt_ptr - 1, sizeof sin);
1274 			switch (ipt.ipt_flg) {
1275 
1276 			case IPOPT_TS_TSONLY:
1277 				break;
1278 
1279 			case IPOPT_TS_TSANDADDR:
1280 				if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1281 				    sizeof(struct in_addr) > ipt.ipt_len)
1282 					goto bad;
1283 				memset(&ipaddr, 0, sizeof(ipaddr));
1284 				ipaddr.sin_family = AF_INET;
1285 				ipaddr.sin_len = sizeof(ipaddr);
1286 				ipaddr.sin_addr = dst;
1287 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1288 				    ifp));
1289 				if (ia == NULL)
1290 					continue;
1291 				memcpy(&sin, &ia->ia_addr.sin_addr,
1292 				    sizeof(struct in_addr));
1293 				ipt.ipt_ptr += sizeof(struct in_addr);
1294 				break;
1295 
1296 			case IPOPT_TS_PRESPEC:
1297 				if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1298 				    sizeof(struct in_addr) > ipt.ipt_len)
1299 					goto bad;
1300 				memset(&ipaddr, 0, sizeof(ipaddr));
1301 				ipaddr.sin_family = AF_INET;
1302 				ipaddr.sin_len = sizeof(ipaddr);
1303 				ipaddr.sin_addr = sin;
1304 				if (ifa_ifwithaddr(sintosa(&ipaddr),
1305 				    m->m_pkthdr.ph_rtableid) == NULL)
1306 					continue;
1307 				ipt.ipt_ptr += sizeof(struct in_addr);
1308 				break;
1309 
1310 			default:
1311 				/* XXX can't take &ipt->ipt_flg */
1312 				code = (u_char *)&ipt.ipt_ptr -
1313 				    (u_char *)ip + 1;
1314 				goto bad;
1315 			}
1316 			ntime = iptime();
1317 			memcpy(cp + ipt.ipt_ptr - 1, &ntime, sizeof(u_int32_t));
1318 			ipt.ipt_ptr += sizeof(u_int32_t);
1319 		}
1320 	}
1321 	KERNEL_UNLOCK();
1322 	if (forward && ipforwarding > 0) {
1323 		ip_forward(m, ifp, NULL, 1);
1324 		return (1);
1325 	}
1326 	return (0);
1327 bad:
1328 	KERNEL_UNLOCK();
1329 	icmp_error(m, type, code, 0, 0);
1330 	ipstat_inc(ips_badoptions);
1331 	return (1);
1332 }
1333 
1334 /*
1335  * Save incoming source route for use in replies,
1336  * to be picked up later by ip_srcroute if the receiver is interested.
1337  */
1338 void
1339 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1340 {
1341 	struct ip_srcrt *isr;
1342 	struct m_tag *mtag;
1343 	unsigned olen;
1344 
1345 	olen = option[IPOPT_OLEN];
1346 	if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1347 		return;
1348 
1349 	mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1350 	if (mtag == NULL)
1351 		return;
1352 	isr = (struct ip_srcrt *)(mtag + 1);
1353 
1354 	memcpy(isr->isr_hdr, option, olen);
1355 	isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1356 	isr->isr_dst = dst;
1357 	m_tag_prepend(m, mtag);
1358 }
1359 
1360 /*
1361  * Retrieve incoming source route for use in replies,
1362  * in the same form used by setsockopt.
1363  * The first hop is placed before the options, will be removed later.
1364  */
1365 struct mbuf *
1366 ip_srcroute(struct mbuf *m0)
1367 {
1368 	struct in_addr *p, *q;
1369 	struct mbuf *m;
1370 	struct ip_srcrt *isr;
1371 	struct m_tag *mtag;
1372 
1373 	if (!ip_dosourceroute)
1374 		return (NULL);
1375 
1376 	mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1377 	if (mtag == NULL)
1378 		return (NULL);
1379 	isr = (struct ip_srcrt *)(mtag + 1);
1380 
1381 	if (isr->isr_nhops == 0)
1382 		return (NULL);
1383 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1384 	if (m == NULL)
1385 		return (NULL);
1386 
1387 #define OPTSIZ	(sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1388 
1389 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1390 	m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1391 
1392 	/*
1393 	 * First save first hop for return route
1394 	 */
1395 	p = &(isr->isr_routes[isr->isr_nhops - 1]);
1396 	*(mtod(m, struct in_addr *)) = *p--;
1397 
1398 	/*
1399 	 * Copy option fields and padding (nop) to mbuf.
1400 	 */
1401 	isr->isr_nop = IPOPT_NOP;
1402 	isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1403 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &isr->isr_nop,
1404 	    OPTSIZ);
1405 	q = (struct in_addr *)(mtod(m, caddr_t) +
1406 	    sizeof(struct in_addr) + OPTSIZ);
1407 #undef OPTSIZ
1408 	/*
1409 	 * Record return path as an IP source route,
1410 	 * reversing the path (pointers are now aligned).
1411 	 */
1412 	while (p >= isr->isr_routes) {
1413 		*q++ = *p--;
1414 	}
1415 	/*
1416 	 * Last hop goes to final destination.
1417 	 */
1418 	*q = isr->isr_dst;
1419 	m_tag_delete(m0, (struct m_tag *)isr);
1420 	return (m);
1421 }
1422 
1423 /*
1424  * Strip out IP options, at higher level protocol in the kernel.
1425  */
1426 void
1427 ip_stripoptions(struct mbuf *m)
1428 {
1429 	int i;
1430 	struct ip *ip = mtod(m, struct ip *);
1431 	caddr_t opts;
1432 	int olen;
1433 
1434 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1435 	opts = (caddr_t)(ip + 1);
1436 	i = m->m_len - (sizeof (struct ip) + olen);
1437 	memmove(opts, opts  + olen, i);
1438 	m->m_len -= olen;
1439 	if (m->m_flags & M_PKTHDR)
1440 		m->m_pkthdr.len -= olen;
1441 	ip->ip_hl = sizeof(struct ip) >> 2;
1442 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1443 }
1444 
1445 const u_char inetctlerrmap[PRC_NCMDS] = {
1446 	0,		0,		0,		0,
1447 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1448 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1449 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1450 	0,		0,		0,		0,
1451 	ENOPROTOOPT
1452 };
1453 
1454 /*
1455  * Forward a packet.  If some error occurs return the sender
1456  * an icmp packet.  Note we can't always generate a meaningful
1457  * icmp message because icmp doesn't have a large enough repertoire
1458  * of codes and types.
1459  *
1460  * If not forwarding, just drop the packet.  This could be confusing
1461  * if ipforwarding was zero but some routing protocol was advancing
1462  * us as a gateway to somewhere.  However, we must let the routing
1463  * protocol deal with that.
1464  *
1465  * The srcrt parameter indicates whether the packet is being forwarded
1466  * via a source route.
1467  */
1468 void
1469 ip_forward(struct mbuf *m, struct ifnet *ifp, struct rtentry *rt, int srcrt)
1470 {
1471 	struct mbuf mfake, *mcopy = NULL;
1472 	struct ip *ip = mtod(m, struct ip *);
1473 	struct sockaddr_in *sin;
1474 	struct route ro;
1475 	int error = 0, type = 0, code = 0, destmtu = 0, fake = 0, len;
1476 	u_int32_t dest;
1477 
1478 	dest = 0;
1479 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1480 		ipstat_inc(ips_cantforward);
1481 		m_freem(m);
1482 		goto freecopy;
1483 	}
1484 	if (ip->ip_ttl <= IPTTLDEC) {
1485 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1486 		goto freecopy;
1487 	}
1488 
1489 	memset(&ro, 0, sizeof(ro));
1490 	sin = satosin(&ro.ro_dst);
1491 	sin->sin_family = AF_INET;
1492 	sin->sin_len = sizeof(*sin);
1493 	sin->sin_addr = ip->ip_dst;
1494 
1495 	if (!rtisvalid(rt)) {
1496 		rtfree(rt);
1497 		rt = rtalloc_mpath(sintosa(sin), &ip->ip_src.s_addr,
1498 		    m->m_pkthdr.ph_rtableid);
1499 		if (rt == NULL) {
1500 			ipstat_inc(ips_noroute);
1501 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1502 			return;
1503 		}
1504 	}
1505 
1506 	/*
1507 	 * Save at most 68 bytes of the packet in case
1508 	 * we need to generate an ICMP message to the src.
1509 	 * The data is saved in the mbuf on the stack that
1510 	 * acts as a temporary storage not intended to be
1511 	 * passed down the IP stack or to the mfree.
1512 	 */
1513 	memset(&mfake.m_hdr, 0, sizeof(mfake.m_hdr));
1514 	mfake.m_type = m->m_type;
1515 	if (m_dup_pkthdr(&mfake, m, M_DONTWAIT) == 0) {
1516 		mfake.m_data = mfake.m_pktdat;
1517 		len = min(ntohs(ip->ip_len), 68);
1518 		m_copydata(m, 0, len, mfake.m_pktdat);
1519 		mfake.m_pkthdr.len = mfake.m_len = len;
1520 #if NPF > 0
1521 		pf_pkt_addr_changed(&mfake);
1522 #endif	/* NPF > 0 */
1523 		fake = 1;
1524 	}
1525 
1526 	ip->ip_ttl -= IPTTLDEC;
1527 
1528 	/*
1529 	 * If forwarding packet using same interface that it came in on,
1530 	 * perhaps should send a redirect to sender to shortcut a hop.
1531 	 * Only send redirect if source is sending directly to us,
1532 	 * and if packet was not source routed (or has any options).
1533 	 * Also, don't send redirect if forwarding using a default route
1534 	 * or a route modified by a redirect.
1535 	 * Don't send redirect if we advertise destination's arp address
1536 	 * as ours (proxy arp).
1537 	 */
1538 	if ((rt->rt_ifidx == ifp->if_index) &&
1539 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1540 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1541 	    ipsendredirects && !srcrt &&
1542 	    !arpproxy(satosin(rt_key(rt))->sin_addr, m->m_pkthdr.ph_rtableid)) {
1543 		if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_netmask) ==
1544 		    ifatoia(rt->rt_ifa)->ia_net) {
1545 		    if (rt->rt_flags & RTF_GATEWAY)
1546 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1547 		    else
1548 			dest = ip->ip_dst.s_addr;
1549 		    /* Router requirements says to only send host redirects */
1550 		    type = ICMP_REDIRECT;
1551 		    code = ICMP_REDIRECT_HOST;
1552 		}
1553 	}
1554 
1555 	ro.ro_rt = rt;
1556 	ro.ro_tableid = m->m_pkthdr.ph_rtableid;
1557 	error = ip_output(m, NULL, &ro,
1558 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1559 	    NULL, NULL, 0);
1560 	rt = ro.ro_rt;
1561 	if (error)
1562 		ipstat_inc(ips_cantforward);
1563 	else {
1564 		ipstat_inc(ips_forward);
1565 		if (type)
1566 			ipstat_inc(ips_redirectsent);
1567 		else
1568 			goto freecopy;
1569 	}
1570 	if (!fake)
1571 		goto freecopy;
1572 
1573 	switch (error) {
1574 	case 0:				/* forwarded, but need redirect */
1575 		/* type, code set above */
1576 		break;
1577 
1578 	case EMSGSIZE:
1579 		type = ICMP_UNREACH;
1580 		code = ICMP_UNREACH_NEEDFRAG;
1581 		if (rt != NULL) {
1582 			if (rt->rt_mtu) {
1583 				destmtu = rt->rt_mtu;
1584 			} else {
1585 				struct ifnet *destifp;
1586 
1587 				destifp = if_get(rt->rt_ifidx);
1588 				if (destifp != NULL)
1589 					destmtu = destifp->if_mtu;
1590 				if_put(destifp);
1591 			}
1592 		}
1593 		ipstat_inc(ips_cantfrag);
1594 		if (destmtu == 0)
1595 			goto freecopy;
1596 		break;
1597 
1598 	case EACCES:
1599 		/*
1600 		 * pf(4) blocked the packet. There is no need to send an ICMP
1601 		 * packet back since pf(4) takes care of it.
1602 		 */
1603 		goto freecopy;
1604 
1605 	case ENOBUFS:
1606 		/*
1607 		 * a router should not generate ICMP_SOURCEQUENCH as
1608 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1609 		 * source quench could be a big problem under DoS attacks,
1610 		 * or the underlying interface is rate-limited.
1611 		 */
1612 		goto freecopy;
1613 
1614 	case ENETUNREACH:		/* shouldn't happen, checked above */
1615 	case EHOSTUNREACH:
1616 	case ENETDOWN:
1617 	case EHOSTDOWN:
1618 	default:
1619 		type = ICMP_UNREACH;
1620 		code = ICMP_UNREACH_HOST;
1621 		break;
1622 	}
1623 	mcopy = m_copym(&mfake, 0, len, M_DONTWAIT);
1624 	if (mcopy)
1625 		icmp_error(mcopy, type, code, dest, destmtu);
1626 
1627 freecopy:
1628 	if (fake)
1629 		m_tag_delete_chain(&mfake);
1630 	rtfree(rt);
1631 }
1632 
1633 int
1634 ip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1635     size_t newlen)
1636 {
1637 	int error;
1638 #ifdef MROUTING
1639 	extern struct mrtstat mrtstat;
1640 #endif
1641 
1642 	/* Almost all sysctl names at this level are terminal. */
1643 	if (namelen != 1 && name[0] != IPCTL_IFQUEUE &&
1644 	    name[0] != IPCTL_ARPQUEUE)
1645 		return (ENOTDIR);
1646 
1647 	switch (name[0]) {
1648 	case IPCTL_SOURCEROUTE:
1649 		NET_LOCK();
1650 		error = sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
1651 		    &ip_dosourceroute);
1652 		NET_UNLOCK();
1653 		return (error);
1654 	case IPCTL_MTUDISC:
1655 		NET_LOCK();
1656 		error = sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtudisc);
1657 		if (ip_mtudisc == 0)
1658 			rt_timer_queue_flush(&ip_mtudisc_timeout_q);
1659 		NET_UNLOCK();
1660 		return error;
1661 	case IPCTL_MTUDISCTIMEOUT:
1662 		NET_LOCK();
1663 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1664 		    &ip_mtudisc_timeout, 0, INT_MAX);
1665 		rt_timer_queue_change(&ip_mtudisc_timeout_q,
1666 		    ip_mtudisc_timeout);
1667 		NET_UNLOCK();
1668 		return (error);
1669 #ifdef IPSEC
1670 	case IPCTL_ENCDEBUG:
1671 	case IPCTL_IPSEC_STATS:
1672 	case IPCTL_IPSEC_EXPIRE_ACQUIRE:
1673 	case IPCTL_IPSEC_EMBRYONIC_SA_TIMEOUT:
1674 	case IPCTL_IPSEC_REQUIRE_PFS:
1675 	case IPCTL_IPSEC_SOFT_ALLOCATIONS:
1676 	case IPCTL_IPSEC_ALLOCATIONS:
1677 	case IPCTL_IPSEC_SOFT_BYTES:
1678 	case IPCTL_IPSEC_BYTES:
1679 	case IPCTL_IPSEC_TIMEOUT:
1680 	case IPCTL_IPSEC_SOFT_TIMEOUT:
1681 	case IPCTL_IPSEC_SOFT_FIRSTUSE:
1682 	case IPCTL_IPSEC_FIRSTUSE:
1683 	case IPCTL_IPSEC_ENC_ALGORITHM:
1684 	case IPCTL_IPSEC_AUTH_ALGORITHM:
1685 	case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1686 		return (ipsec_sysctl(name, namelen, oldp, oldlenp, newp,
1687 		    newlen));
1688 #endif
1689 	case IPCTL_IFQUEUE:
1690 		return (sysctl_niq(name + 1, namelen - 1,
1691 		    oldp, oldlenp, newp, newlen, &ipintrq));
1692 	case IPCTL_ARPQUEUE:
1693 		return (sysctl_niq(name + 1, namelen - 1,
1694 		    oldp, oldlenp, newp, newlen, &arpinq));
1695 	case IPCTL_ARPQUEUED:
1696 		return (sysctl_rdint(oldp, oldlenp, newp, la_hold_total));
1697 	case IPCTL_STATS:
1698 		return (ip_sysctl_ipstat(oldp, oldlenp, newp));
1699 #ifdef MROUTING
1700 	case IPCTL_MRTSTATS:
1701 		return (sysctl_rdstruct(oldp, oldlenp, newp,
1702 		    &mrtstat, sizeof(mrtstat)));
1703 	case IPCTL_MRTMFC:
1704 		if (newp)
1705 			return (EPERM);
1706 		NET_LOCK();
1707 		error = mrt_sysctl_mfc(oldp, oldlenp);
1708 		NET_UNLOCK();
1709 		return (error);
1710 	case IPCTL_MRTVIF:
1711 		if (newp)
1712 			return (EPERM);
1713 		NET_LOCK();
1714 		error = mrt_sysctl_vif(oldp, oldlenp);
1715 		NET_UNLOCK();
1716 		return (error);
1717 #else
1718 	case IPCTL_MRTPROTO:
1719 	case IPCTL_MRTSTATS:
1720 	case IPCTL_MRTMFC:
1721 	case IPCTL_MRTVIF:
1722 		return (EOPNOTSUPP);
1723 #endif
1724 	default:
1725 		NET_LOCK();
1726 		error = sysctl_bounded_arr(ipctl_vars, nitems(ipctl_vars),
1727 		    name, namelen, oldp, oldlenp, newp, newlen);
1728 		NET_UNLOCK();
1729 		return (error);
1730 	}
1731 	/* NOTREACHED */
1732 }
1733 
1734 int
1735 ip_sysctl_ipstat(void *oldp, size_t *oldlenp, void *newp)
1736 {
1737 	uint64_t counters[ips_ncounters];
1738 	struct ipstat ipstat;
1739 	u_long *words = (u_long *)&ipstat;
1740 	int i;
1741 
1742 	CTASSERT(sizeof(ipstat) == (nitems(counters) * sizeof(u_long)));
1743 	memset(&ipstat, 0, sizeof ipstat);
1744 	counters_read(ipcounters, counters, nitems(counters));
1745 
1746 	for (i = 0; i < nitems(counters); i++)
1747 		words[i] = (u_long)counters[i];
1748 
1749 	return (sysctl_rdstruct(oldp, oldlenp, newp, &ipstat, sizeof(ipstat)));
1750 }
1751 
1752 void
1753 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1754     struct mbuf *m)
1755 {
1756 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1757 		struct timeval tv;
1758 
1759 		m_microtime(m, &tv);
1760 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1761 		    SCM_TIMESTAMP, SOL_SOCKET);
1762 		if (*mp)
1763 			mp = &(*mp)->m_next;
1764 	}
1765 
1766 	if (inp->inp_flags & INP_RECVDSTADDR) {
1767 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1768 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1769 		if (*mp)
1770 			mp = &(*mp)->m_next;
1771 	}
1772 #ifdef notyet
1773 	/* this code is broken and will probably never be fixed. */
1774 	/* options were tossed already */
1775 	if (inp->inp_flags & INP_RECVOPTS) {
1776 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1777 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1778 		if (*mp)
1779 			mp = &(*mp)->m_next;
1780 	}
1781 	/* ip_srcroute doesn't do what we want here, need to fix */
1782 	if (inp->inp_flags & INP_RECVRETOPTS) {
1783 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1784 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1785 		if (*mp)
1786 			mp = &(*mp)->m_next;
1787 	}
1788 #endif
1789 	if (inp->inp_flags & INP_RECVIF) {
1790 		struct sockaddr_dl sdl;
1791 		struct ifnet *ifp;
1792 
1793 		ifp = if_get(m->m_pkthdr.ph_ifidx);
1794 		if (ifp == NULL || ifp->if_sadl == NULL) {
1795 			memset(&sdl, 0, sizeof(sdl));
1796 			sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1797 			sdl.sdl_family = AF_LINK;
1798 			sdl.sdl_index = ifp != NULL ? ifp->if_index : 0;
1799 			sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1800 			*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1801 			    IP_RECVIF, IPPROTO_IP);
1802 		} else {
1803 			*mp = sbcreatecontrol((caddr_t) ifp->if_sadl,
1804 			    ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP);
1805 		}
1806 		if (*mp)
1807 			mp = &(*mp)->m_next;
1808 		if_put(ifp);
1809 	}
1810 	if (inp->inp_flags & INP_RECVTTL) {
1811 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1812 		    sizeof(u_int8_t), IP_RECVTTL, IPPROTO_IP);
1813 		if (*mp)
1814 			mp = &(*mp)->m_next;
1815 	}
1816 	if (inp->inp_flags & INP_RECVRTABLE) {
1817 		u_int rtableid = inp->inp_rtableid;
1818 
1819 #if NPF > 0
1820 		if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
1821 			struct pf_divert *divert;
1822 
1823 			divert = pf_find_divert(m);
1824 			KASSERT(divert != NULL);
1825 			rtableid = divert->rdomain;
1826 		}
1827 #endif
1828 
1829 		*mp = sbcreatecontrol((caddr_t) &rtableid,
1830 		    sizeof(u_int), IP_RECVRTABLE, IPPROTO_IP);
1831 		if (*mp)
1832 			mp = &(*mp)->m_next;
1833 	}
1834 }
1835 
1836 void
1837 ip_send_do_dispatch(void *xmq, int flags)
1838 {
1839 	struct mbuf_queue *mq = xmq;
1840 	struct mbuf *m;
1841 	struct mbuf_list ml;
1842 	struct m_tag *mtag;
1843 
1844 	mq_delist(mq, &ml);
1845 	if (ml_empty(&ml))
1846 		return;
1847 
1848 	NET_LOCK();
1849 	while ((m = ml_dequeue(&ml)) != NULL) {
1850 		u_int32_t ipsecflowinfo = 0;
1851 
1852 		if ((mtag = m_tag_find(m, PACKET_TAG_IPSEC_FLOWINFO, NULL))
1853 		    != NULL) {
1854 			ipsecflowinfo = *(u_int32_t *)(mtag + 1);
1855 			m_tag_delete(m, mtag);
1856 		}
1857 		ip_output(m, NULL, NULL, flags, NULL, NULL, ipsecflowinfo);
1858 	}
1859 	NET_UNLOCK();
1860 }
1861 
1862 void
1863 ip_sendraw_dispatch(void *xmq)
1864 {
1865 	ip_send_do_dispatch(xmq, IP_RAWOUTPUT);
1866 }
1867 
1868 void
1869 ip_send_dispatch(void *xmq)
1870 {
1871 	ip_send_do_dispatch(xmq, 0);
1872 }
1873 
1874 void
1875 ip_send(struct mbuf *m)
1876 {
1877 	mq_enqueue(&ipsend_mq, m);
1878 	task_add(net_tq(0), &ipsend_task);
1879 }
1880 
1881 void
1882 ip_send_raw(struct mbuf *m)
1883 {
1884 	mq_enqueue(&ipsendraw_mq, m);
1885 	task_add(net_tq(0), &ipsendraw_task);
1886 }
1887