xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 7863ba460b0a05b553c754e5dbc29247dddec322)
1 /*	$NetBSD: udp_usrreq.c,v 1.246 2018/03/19 16:26:25 roy Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
61  */
62 
63 /*
64  * UDP protocol implementation.
65  * Per RFC 768, August, 1980.
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.246 2018/03/19 16:26:25 roy Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_inet.h"
73 #include "opt_ipsec.h"
74 #include "opt_inet_csum.h"
75 #include "opt_ipkdb.h"
76 #include "opt_mbuftrace.h"
77 #include "opt_net_mpsafe.h"
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/mbuf.h>
82 #include <sys/once.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/systm.h>
87 #include <sys/proc.h>
88 #include <sys/domain.h>
89 #include <sys/sysctl.h>
90 
91 #include <net/if.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/udp.h>
101 #include <netinet/udp_var.h>
102 #include <netinet/udp_private.h>
103 
104 #ifdef INET6
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/ip6_private.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/udp6_var.h>
111 #include <netinet6/udp6_private.h>
112 #endif
113 
114 #ifndef INET6
115 /* always need ip6.h for IP6_EXTHDR_GET */
116 #include <netinet/ip6.h>
117 #endif
118 
119 #ifdef IPSEC
120 #include <netipsec/ipsec.h>
121 #include <netipsec/ipsec_var.h>
122 #include <netipsec/esp.h>
123 #ifdef INET6
124 #include <netipsec/ipsec6.h>
125 #endif
126 #endif
127 
128 #ifdef IPKDB
129 #include <ipkdb/ipkdb.h>
130 #endif
131 
132 int udpcksum = 1;
133 int udp_do_loopback_cksum = 0;
134 
135 struct inpcbtable udbtable;
136 
137 percpu_t *udpstat_percpu;
138 
139 #ifdef INET
140 #ifdef IPSEC
141 static int udp4_espinudp(struct mbuf **, int, struct sockaddr *,
142     struct socket *);
143 #endif
144 static void udp4_sendup(struct mbuf *, int, struct sockaddr *,
145     struct socket *);
146 static int udp4_realinput(struct sockaddr_in *, struct sockaddr_in *,
147     struct mbuf **, int);
148 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
149 #endif
150 #ifdef INET
151 static void udp_notify (struct inpcb *, int);
152 #endif
153 
154 #ifndef UDBHASHSIZE
155 #define	UDBHASHSIZE	128
156 #endif
157 int udbhashsize = UDBHASHSIZE;
158 
159 /*
160  * For send - really max datagram size; for receive - 40 1K datagrams.
161  */
162 static int udp_sendspace = 9216;
163 static int udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
164 
165 #ifdef MBUFTRACE
166 struct mowner udp_mowner = MOWNER_INIT("udp", "");
167 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
168 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
169 #endif
170 
171 #ifdef UDP_CSUM_COUNTERS
172 #include <sys/device.h>
173 
174 #if defined(INET)
175 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
176     NULL, "udp", "hwcsum bad");
177 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
178     NULL, "udp", "hwcsum ok");
179 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
180     NULL, "udp", "hwcsum data");
181 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182     NULL, "udp", "swcsum");
183 
184 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
185 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
186 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
187 EVCNT_ATTACH_STATIC(udp_swcsum);
188 #endif /* defined(INET) */
189 
190 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
191 #else
192 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
193 #endif /* UDP_CSUM_COUNTERS */
194 
195 static void sysctl_net_inet_udp_setup(struct sysctllog **);
196 
197 static int
198 do_udpinit(void)
199 {
200 
201 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
202 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
203 
204 	MOWNER_ATTACH(&udp_tx_mowner);
205 	MOWNER_ATTACH(&udp_rx_mowner);
206 	MOWNER_ATTACH(&udp_mowner);
207 
208 	return 0;
209 }
210 
211 void
212 udp_init_common(void)
213 {
214 	static ONCE_DECL(doudpinit);
215 
216 	RUN_ONCE(&doudpinit, do_udpinit);
217 }
218 
219 void
220 udp_init(void)
221 {
222 
223 	sysctl_net_inet_udp_setup(NULL);
224 
225 	udp_init_common();
226 }
227 
228 /*
229  * Checksum extended UDP header and data.
230  */
231 int
232 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
233     int iphlen, int len)
234 {
235 
236 	switch (af) {
237 #ifdef INET
238 	case AF_INET:
239 		return udp4_input_checksum(m, uh, iphlen, len);
240 #endif
241 #ifdef INET6
242 	case AF_INET6:
243 		return udp6_input_checksum(m, uh, iphlen, len);
244 #endif
245 	}
246 #ifdef DIAGNOSTIC
247 	panic("udp_input_checksum: unknown af %d", af);
248 #endif
249 	/* NOTREACHED */
250 	return -1;
251 }
252 
253 #ifdef INET
254 
255 /*
256  * Checksum extended UDP header and data.
257  */
258 static int
259 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
260     int iphlen, int len)
261 {
262 
263 	/*
264 	 * XXX it's better to record and check if this mbuf is
265 	 * already checked.
266 	 */
267 
268 	if (uh->uh_sum == 0)
269 		return 0;
270 
271 	switch (m->m_pkthdr.csum_flags &
272 	    ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
273 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
274 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
275 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
276 		goto badcsum;
277 
278 	case M_CSUM_UDPv4|M_CSUM_DATA: {
279 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
280 
281 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
282 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
283 			const struct ip *ip =
284 			    mtod(m, const struct ip *);
285 
286 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
287 			    ip->ip_dst.s_addr,
288 			    htons(hw_csum + len + IPPROTO_UDP));
289 		}
290 		if ((hw_csum ^ 0xffff) != 0)
291 			goto badcsum;
292 		break;
293 	}
294 
295 	case M_CSUM_UDPv4:
296 		/* Checksum was okay. */
297 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
298 		break;
299 
300 	default:
301 		/*
302 		 * Need to compute it ourselves.  Maybe skip checksum
303 		 * on loopback interfaces.
304 		 */
305 		if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
306 				     IFF_LOOPBACK) ||
307 				   udp_do_loopback_cksum)) {
308 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
309 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
310 				goto badcsum;
311 		}
312 		break;
313 	}
314 
315 	return 0;
316 
317 badcsum:
318 	UDP_STATINC(UDP_STAT_BADSUM);
319 	return -1;
320 }
321 
322 void
323 udp_input(struct mbuf *m, ...)
324 {
325 	va_list ap;
326 	struct sockaddr_in src, dst;
327 	struct ip *ip;
328 	struct udphdr *uh;
329 	int iphlen;
330 	int len;
331 	int n;
332 	u_int16_t ip_len;
333 
334 	va_start(ap, m);
335 	iphlen = va_arg(ap, int);
336 	(void)va_arg(ap, int);		/* ignore value, advance ap */
337 	va_end(ap);
338 
339 	MCLAIM(m, &udp_rx_mowner);
340 	UDP_STATINC(UDP_STAT_IPACKETS);
341 
342 	/*
343 	 * Get IP and UDP header together in first mbuf.
344 	 */
345 	ip = mtod(m, struct ip *);
346 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
347 	if (uh == NULL) {
348 		UDP_STATINC(UDP_STAT_HDROPS);
349 		return;
350 	}
351 
352 	/*
353 	 * Enforce alignment requirements that are violated in
354 	 * some cases, see kern/50766 for details.
355 	 */
356 	if (UDP_HDR_ALIGNED_P(uh) == 0) {
357 		m = m_copyup(m, iphlen + sizeof(struct udphdr), 0);
358 		if (m == NULL) {
359 			UDP_STATINC(UDP_STAT_HDROPS);
360 			return;
361 		}
362 		ip = mtod(m, struct ip *);
363 		uh = (struct udphdr *)(mtod(m, char *) + iphlen);
364 	}
365 	KASSERT(UDP_HDR_ALIGNED_P(uh));
366 
367 	/* destination port of 0 is illegal, based on RFC768. */
368 	if (uh->uh_dport == 0)
369 		goto bad;
370 
371 	/*
372 	 * Make mbuf data length reflect UDP length.
373 	 * If not enough data to reflect UDP length, drop.
374 	 */
375 	ip_len = ntohs(ip->ip_len);
376 	len = ntohs((u_int16_t)uh->uh_ulen);
377 	if (ip_len != iphlen + len) {
378 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
379 			UDP_STATINC(UDP_STAT_BADLEN);
380 			goto bad;
381 		}
382 		m_adj(m, iphlen + len - ip_len);
383 	}
384 
385 	/*
386 	 * Checksum extended UDP header and data.
387 	 */
388 	if (udp4_input_checksum(m, uh, iphlen, len))
389 		goto badcsum;
390 
391 	/* construct source and dst sockaddrs. */
392 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
393 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
394 
395 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
396 		UDP_STATINC(UDP_STAT_HDROPS);
397 		return;
398 	}
399 	if (m == NULL) {
400 		/*
401 		 * packet has been processed by ESP stuff -
402 		 * e.g. dropped NAT-T-keep-alive-packet ...
403 		 */
404 		return;
405 	}
406 
407 	ip = mtod(m, struct ip *);
408 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
409 	if (uh == NULL) {
410 		UDP_STATINC(UDP_STAT_HDROPS);
411 		return;
412 	}
413 	/* XXX Re-enforce alignment? */
414 
415 #ifdef INET6
416 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
417 		struct sockaddr_in6 src6, dst6;
418 
419 		memset(&src6, 0, sizeof(src6));
420 		src6.sin6_family = AF_INET6;
421 		src6.sin6_len = sizeof(struct sockaddr_in6);
422 		in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
423 		src6.sin6_port = uh->uh_sport;
424 		memset(&dst6, 0, sizeof(dst6));
425 		dst6.sin6_family = AF_INET6;
426 		dst6.sin6_len = sizeof(struct sockaddr_in6);
427 		in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
428 		dst6.sin6_port = uh->uh_dport;
429 
430 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
431 	}
432 #endif
433 
434 	if (n == 0) {
435 		if (m->m_flags & (M_BCAST | M_MCAST)) {
436 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
437 			goto bad;
438 		}
439 		UDP_STATINC(UDP_STAT_NOPORT);
440 #ifdef IPKDB
441 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
442 		    m, iphlen + sizeof(struct udphdr),
443 		    m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
444 			/*
445 			 * It was a debugger connect packet,
446 			 * just drop it now
447 			 */
448 			goto bad;
449 		}
450 #endif
451 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
452 		m = NULL;
453 	}
454 
455 bad:
456 	if (m)
457 		m_freem(m);
458 	return;
459 
460 badcsum:
461 	m_freem(m);
462 }
463 #endif
464 
465 #ifdef INET
466 static void
467 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
468     struct sockaddr *src, struct socket *so)
469 {
470 	struct mbuf *opts = NULL;
471 	struct mbuf *n;
472 	struct inpcb *inp;
473 
474 	KASSERT(so != NULL);
475 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET);
476 	inp = sotoinpcb(so);
477 	KASSERT(inp != NULL);
478 
479 #if defined(IPSEC)
480 	/* check AH/ESP integrity. */
481 	if (ipsec_used && ipsec_in_reject(m, inp)) {
482 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
483 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
484 			    0, 0);
485 		return;
486 	}
487 #endif
488 
489 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
490 		if (inp->inp_flags & INP_CONTROLOPTS ||
491 		    SOOPT_TIMESTAMP(so->so_options)) {
492 			struct ip *ip = mtod(n, struct ip *);
493 			ip_savecontrol(inp, &opts, ip, n);
494 		}
495 
496 		m_adj(n, off);
497 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
498 			m_freem(n);
499 			if (opts)
500 				m_freem(opts);
501 			UDP_STATINC(UDP_STAT_FULLSOCK);
502 			soroverflow(so);
503 		} else
504 			sorwakeup(so);
505 	}
506 }
507 #endif
508 
509 #ifdef INET
510 static int
511 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
512     struct mbuf **mp, int off /* offset of udphdr */)
513 {
514 	u_int16_t *sport, *dport;
515 	int rcvcnt;
516 	struct in_addr *src4, *dst4;
517 	struct inpcb_hdr *inph;
518 	struct inpcb *inp;
519 	struct mbuf *m = *mp;
520 
521 	rcvcnt = 0;
522 	off += sizeof(struct udphdr);	/* now, offset of payload */
523 
524 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
525 		goto bad;
526 
527 	src4 = &src->sin_addr;
528 	sport = &src->sin_port;
529 	dst4 = &dst->sin_addr;
530 	dport = &dst->sin_port;
531 
532 	if (IN_MULTICAST(dst4->s_addr) ||
533 	    in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
534 		/*
535 		 * Deliver a multicast or broadcast datagram to *all* sockets
536 		 * for which the local and remote addresses and ports match
537 		 * those of the incoming datagram.  This allows more than
538 		 * one process to receive multi/broadcasts on the same port.
539 		 * (This really ought to be done for unicast datagrams as
540 		 * well, but that would cause problems with existing
541 		 * applications that open both address-specific sockets and
542 		 * a wildcard socket listening to the same port -- they would
543 		 * end up receiving duplicates of every unicast datagram.
544 		 * Those applications open the multiple sockets to overcome an
545 		 * inadequacy of the UDP socket interface, but for backwards
546 		 * compatibility we avoid the problem here rather than
547 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
548 		 */
549 
550 		/*
551 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
552 		 * we need udpiphdr for IPsec processing so we do that later.
553 		 */
554 		/*
555 		 * Locate pcb(s) for datagram.
556 		 */
557 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
558 			inp = (struct inpcb *)inph;
559 			if (inp->inp_af != AF_INET)
560 				continue;
561 
562 			if (inp->inp_lport != *dport)
563 				continue;
564 			if (!in_nullhost(inp->inp_laddr)) {
565 				if (!in_hosteq(inp->inp_laddr, *dst4))
566 					continue;
567 			}
568 			if (!in_nullhost(inp->inp_faddr)) {
569 				if (!in_hosteq(inp->inp_faddr, *src4) ||
570 				    inp->inp_fport != *sport)
571 					continue;
572 			}
573 
574 			udp4_sendup(m, off, (struct sockaddr *)src,
575 			    inp->inp_socket);
576 			rcvcnt++;
577 
578 			/*
579 			 * Don't look for additional matches if this one does
580 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
581 			 * socket options set.  This heuristic avoids searching
582 			 * through all pcbs in the common case of a non-shared
583 			 * port.  It assumes that an application will never
584 			 * clear these options after setting them.
585 			 */
586 			if ((inp->inp_socket->so_options &
587 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
588 				break;
589 		}
590 	} else {
591 		/*
592 		 * Locate pcb for datagram.
593 		 */
594 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
595 		    *dport, 0);
596 		if (inp == 0) {
597 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
598 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
599 			if (inp == 0)
600 				return rcvcnt;
601 		}
602 
603 #ifdef IPSEC
604 		/* Handle ESP over UDP */
605 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
606 			struct sockaddr *sa = (struct sockaddr *)src;
607 
608 			switch (udp4_espinudp(mp, off, sa, inp->inp_socket)) {
609 			case -1: /* Error, m was freed */
610 				rcvcnt = -1;
611 				goto bad;
612 
613 			case 1: /* ESP over UDP */
614 				rcvcnt++;
615 				goto bad;
616 
617 			case 0: /* plain UDP */
618 			default: /* Unexpected */
619 				/*
620 				 * Normal UDP processing will take place,
621 				 * m may have changed.
622 				 */
623 				m = *mp;
624 				break;
625 			}
626 		}
627 #endif
628 
629 		/*
630 		 * Check the minimum TTL for socket.
631 		 */
632 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
633 			goto bad;
634 
635 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
636 		rcvcnt++;
637 	}
638 
639 bad:
640 	return rcvcnt;
641 }
642 #endif
643 
644 #ifdef INET
645 /*
646  * Notify a udp user of an asynchronous error;
647  * just wake up so that he can collect error status.
648  */
649 static void
650 udp_notify(struct inpcb *inp, int errno)
651 {
652 	inp->inp_socket->so_error = errno;
653 	sorwakeup(inp->inp_socket);
654 	sowwakeup(inp->inp_socket);
655 }
656 
657 void *
658 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
659 {
660 	struct ip *ip = v;
661 	struct udphdr *uh;
662 	void (*notify)(struct inpcb *, int) = udp_notify;
663 	int errno;
664 
665 	if (sa->sa_family != AF_INET ||
666 	    sa->sa_len != sizeof(struct sockaddr_in))
667 		return NULL;
668 	if ((unsigned)cmd >= PRC_NCMDS)
669 		return NULL;
670 
671 	errno = inetctlerrmap[cmd];
672 	if (PRC_IS_REDIRECT(cmd)) {
673 		notify = in_rtchange;
674 		ip = NULL;
675 	} else if (cmd == PRC_HOSTDEAD) {
676 		ip = NULL;
677 	} else if (errno == 0) {
678 		return NULL;
679 	}
680 
681 	if (ip) {
682 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
683 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
684 		    ip->ip_src, uh->uh_sport, errno, notify);
685 		/* XXX mapped address case */
686 	} else {
687 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
688 		    notify);
689 	}
690 
691 	return NULL;
692 }
693 
694 int
695 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
696 {
697 	int s;
698 	int error = 0;
699 	struct inpcb *inp;
700 	int family;
701 	int optval;
702 
703 	family = so->so_proto->pr_domain->dom_family;
704 
705 	s = splsoftnet();
706 	switch (family) {
707 #ifdef INET
708 	case PF_INET:
709 		if (sopt->sopt_level != IPPROTO_UDP) {
710 			error = ip_ctloutput(op, so, sopt);
711 			goto end;
712 		}
713 		break;
714 #endif
715 #ifdef INET6
716 	case PF_INET6:
717 		if (sopt->sopt_level != IPPROTO_UDP) {
718 			error = ip6_ctloutput(op, so, sopt);
719 			goto end;
720 		}
721 		break;
722 #endif
723 	default:
724 		error = EAFNOSUPPORT;
725 		goto end;
726 	}
727 
728 
729 	switch (op) {
730 	case PRCO_SETOPT:
731 		inp = sotoinpcb(so);
732 
733 		switch (sopt->sopt_name) {
734 		case UDP_ENCAP:
735 			error = sockopt_getint(sopt, &optval);
736 			if (error)
737 				break;
738 
739 			switch(optval) {
740 			case 0:
741 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
742 				break;
743 
744 			case UDP_ENCAP_ESPINUDP:
745 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
746 				inp->inp_flags |= INP_ESPINUDP;
747 				break;
748 
749 			case UDP_ENCAP_ESPINUDP_NON_IKE:
750 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
751 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
752 				break;
753 			default:
754 				error = EINVAL;
755 				break;
756 			}
757 			break;
758 
759 		default:
760 			error = ENOPROTOOPT;
761 			break;
762 		}
763 		break;
764 
765 	default:
766 		error = EINVAL;
767 		break;
768 	}
769 
770 end:
771 	splx(s);
772 	return error;
773 }
774 
775 int
776 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control,
777     struct lwp *l)
778 {
779 	struct udpiphdr *ui;
780 	struct route *ro;
781 	struct ip_pktopts pktopts;
782 	kauth_cred_t cred;
783 	int len = m->m_pkthdr.len;
784 	int error, flags = 0;
785 
786 	MCLAIM(m, &udp_tx_mowner);
787 
788 	/*
789 	 * Calculate data length and get a mbuf
790 	 * for UDP and IP headers.
791 	 */
792 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
793 	if (m == NULL) {
794 		error = ENOBUFS;
795 		goto release;
796 	}
797 
798 	/*
799 	 * Compute the packet length of the IP header, and
800 	 * punt if the length looks bogus.
801 	 */
802 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
803 		error = EMSGSIZE;
804 		goto release;
805 	}
806 
807 	if (l == NULL)
808 		cred = NULL;
809 	else
810 		cred = l->l_cred;
811 
812 	/* Setup IP outgoing packet options */
813 	memset(&pktopts, 0, sizeof(pktopts));
814 	error = ip_setpktopts(control, &pktopts, &flags, inp, cred);
815 	if (error != 0)
816 		goto release;
817 
818 	if (control != NULL) {
819 		m_freem(control);
820 		control = NULL;
821 	}
822 
823 	/*
824 	 * Fill in mbuf with extended UDP header
825 	 * and addresses and length put into network format.
826 	 */
827 	ui = mtod(m, struct udpiphdr *);
828 	ui->ui_pr = IPPROTO_UDP;
829 	ui->ui_src = pktopts.ippo_laddr.sin_addr;
830 	ui->ui_dst = inp->inp_faddr;
831 	ui->ui_sport = inp->inp_lport;
832 	ui->ui_dport = inp->inp_fport;
833 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
834 
835 	ro = &inp->inp_route;
836 
837 	/*
838 	 * Set up checksum and output datagram.
839 	 */
840 	if (udpcksum) {
841 		/*
842 		 * XXX Cache pseudo-header checksum part for
843 		 * XXX "connected" UDP sockets.
844 		 */
845 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
846 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
847 		    sizeof(struct udphdr) + IPPROTO_UDP));
848 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
849 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
850 	} else
851 		ui->ui_sum = 0;
852 
853 	((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
854 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
855 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
856 	UDP_STATINC(UDP_STAT_OPACKETS);
857 
858 	flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST);
859 	return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp);
860 
861  release:
862 	if (control != NULL)
863 		m_freem(control);
864 	m_freem(m);
865 	return error;
866 }
867 
868 static int
869 udp_attach(struct socket *so, int proto)
870 {
871 	struct inpcb *inp;
872 	int error;
873 
874 	KASSERT(sotoinpcb(so) == NULL);
875 
876 	/* Assign the lock (must happen even if we will error out). */
877 	sosetlock(so);
878 
879 #ifdef MBUFTRACE
880 	so->so_mowner = &udp_mowner;
881 	so->so_rcv.sb_mowner = &udp_rx_mowner;
882 	so->so_snd.sb_mowner = &udp_tx_mowner;
883 #endif
884 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
885 		error = soreserve(so, udp_sendspace, udp_recvspace);
886 		if (error) {
887 			return error;
888 		}
889 	}
890 
891 	error = in_pcballoc(so, &udbtable);
892 	if (error) {
893 		return error;
894 	}
895 	inp = sotoinpcb(so);
896 	inp->inp_ip.ip_ttl = ip_defttl;
897 	KASSERT(solocked(so));
898 
899 	return error;
900 }
901 
902 static void
903 udp_detach(struct socket *so)
904 {
905 	struct inpcb *inp;
906 
907 	KASSERT(solocked(so));
908 	inp = sotoinpcb(so);
909 	KASSERT(inp != NULL);
910 	in_pcbdetach(inp);
911 }
912 
913 static int
914 udp_accept(struct socket *so, struct sockaddr *nam)
915 {
916 	KASSERT(solocked(so));
917 
918 	panic("udp_accept");
919 
920 	return EOPNOTSUPP;
921 }
922 
923 static int
924 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
925 {
926 	struct inpcb *inp = sotoinpcb(so);
927 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
928 	int error = 0;
929 	int s;
930 
931 	KASSERT(solocked(so));
932 	KASSERT(inp != NULL);
933 	KASSERT(nam != NULL);
934 
935 	s = splsoftnet();
936 	error = in_pcbbind(inp, sin, l);
937 	splx(s);
938 
939 	return error;
940 }
941 
942 static int
943 udp_listen(struct socket *so, struct lwp *l)
944 {
945 	KASSERT(solocked(so));
946 
947 	return EOPNOTSUPP;
948 }
949 
950 static int
951 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
952 {
953 	struct inpcb *inp = sotoinpcb(so);
954 	int error = 0;
955 	int s;
956 
957 	KASSERT(solocked(so));
958 	KASSERT(inp != NULL);
959 	KASSERT(nam != NULL);
960 
961 	s = splsoftnet();
962 	error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
963 	if (! error)
964 		soisconnected(so);
965 	splx(s);
966 	return error;
967 }
968 
969 static int
970 udp_connect2(struct socket *so, struct socket *so2)
971 {
972 	KASSERT(solocked(so));
973 
974 	return EOPNOTSUPP;
975 }
976 
977 static int
978 udp_disconnect(struct socket *so)
979 {
980 	struct inpcb *inp = sotoinpcb(so);
981 	int s;
982 
983 	KASSERT(solocked(so));
984 	KASSERT(inp != NULL);
985 
986 	s = splsoftnet();
987 	/*soisdisconnected(so);*/
988 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
989 	in_pcbdisconnect(inp);
990 	inp->inp_laddr = zeroin_addr;		/* XXX */
991 	in_pcbstate(inp, INP_BOUND);		/* XXX */
992 	splx(s);
993 
994 	return 0;
995 }
996 
997 static int
998 udp_shutdown(struct socket *so)
999 {
1000 	int s;
1001 
1002 	KASSERT(solocked(so));
1003 
1004 	s = splsoftnet();
1005 	socantsendmore(so);
1006 	splx(s);
1007 
1008 	return 0;
1009 }
1010 
1011 static int
1012 udp_abort(struct socket *so)
1013 {
1014 	KASSERT(solocked(so));
1015 
1016 	panic("udp_abort");
1017 
1018 	return EOPNOTSUPP;
1019 }
1020 
1021 static int
1022 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
1023 {
1024 	return in_control(so, cmd, nam, ifp);
1025 }
1026 
1027 static int
1028 udp_stat(struct socket *so, struct stat *ub)
1029 {
1030 	KASSERT(solocked(so));
1031 
1032 	/* stat: don't bother with a blocksize. */
1033 	return 0;
1034 }
1035 
1036 static int
1037 udp_peeraddr(struct socket *so, struct sockaddr *nam)
1038 {
1039 	int s;
1040 
1041 	KASSERT(solocked(so));
1042 	KASSERT(sotoinpcb(so) != NULL);
1043 	KASSERT(nam != NULL);
1044 
1045 	s = splsoftnet();
1046 	in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1047 	splx(s);
1048 
1049 	return 0;
1050 }
1051 
1052 static int
1053 udp_sockaddr(struct socket *so, struct sockaddr *nam)
1054 {
1055 	int s;
1056 
1057 	KASSERT(solocked(so));
1058 	KASSERT(sotoinpcb(so) != NULL);
1059 	KASSERT(nam != NULL);
1060 
1061 	s = splsoftnet();
1062 	in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1063 	splx(s);
1064 
1065 	return 0;
1066 }
1067 
1068 static int
1069 udp_rcvd(struct socket *so, int flags, struct lwp *l)
1070 {
1071 	KASSERT(solocked(so));
1072 
1073 	return EOPNOTSUPP;
1074 }
1075 
1076 static int
1077 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
1078 {
1079 	KASSERT(solocked(so));
1080 
1081 	return EOPNOTSUPP;
1082 }
1083 
1084 static int
1085 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1086     struct mbuf *control, struct lwp *l)
1087 {
1088 	struct inpcb *inp = sotoinpcb(so);
1089 	int error = 0;
1090 	struct in_addr laddr;			/* XXX */
1091 	int s;
1092 
1093 	KASSERT(solocked(so));
1094 	KASSERT(inp != NULL);
1095 	KASSERT(m != NULL);
1096 
1097 	memset(&laddr, 0, sizeof laddr);
1098 
1099 	s = splsoftnet();
1100 	if (nam) {
1101 		laddr = inp->inp_laddr;		/* XXX */
1102 		if ((so->so_state & SS_ISCONNECTED) != 0) {
1103 			error = EISCONN;
1104 			goto die;
1105 		}
1106 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
1107 		if (error)
1108 			goto die;
1109 	} else {
1110 		if ((so->so_state & SS_ISCONNECTED) == 0) {
1111 			error = ENOTCONN;
1112 			goto die;
1113 		}
1114 	}
1115 	error = udp_output(m, inp, control, l);
1116 	m = NULL;
1117 	control = NULL;
1118 	if (nam) {
1119 		in_pcbdisconnect(inp);
1120 		inp->inp_laddr = laddr;		/* XXX */
1121 		in_pcbstate(inp, INP_BOUND);	/* XXX */
1122 	}
1123   die:
1124 	if (m != NULL)
1125 		m_freem(m);
1126 	if (control != NULL)
1127 		m_freem(control);
1128 
1129 	splx(s);
1130 	return error;
1131 }
1132 
1133 static int
1134 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1135 {
1136 	KASSERT(solocked(so));
1137 
1138 	m_freem(m);
1139 	m_freem(control);
1140 
1141 	return EOPNOTSUPP;
1142 }
1143 
1144 static int
1145 udp_purgeif(struct socket *so, struct ifnet *ifp)
1146 {
1147 	int s;
1148 
1149 	s = splsoftnet();
1150 	mutex_enter(softnet_lock);
1151 	in_pcbpurgeif0(&udbtable, ifp);
1152 #ifdef NET_MPSAFE
1153 	mutex_exit(softnet_lock);
1154 #endif
1155 	in_purgeif(ifp);
1156 #ifdef NET_MPSAFE
1157 	mutex_enter(softnet_lock);
1158 #endif
1159 	in_pcbpurgeif(&udbtable, ifp);
1160 	mutex_exit(softnet_lock);
1161 	splx(s);
1162 
1163 	return 0;
1164 }
1165 
1166 static int
1167 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1168 {
1169 
1170 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1171 }
1172 
1173 /*
1174  * Sysctl for udp variables.
1175  */
1176 static void
1177 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1178 {
1179 
1180 	sysctl_createv(clog, 0, NULL, NULL,
1181 		       CTLFLAG_PERMANENT,
1182 		       CTLTYPE_NODE, "inet", NULL,
1183 		       NULL, 0, NULL, 0,
1184 		       CTL_NET, PF_INET, CTL_EOL);
1185 	sysctl_createv(clog, 0, NULL, NULL,
1186 		       CTLFLAG_PERMANENT,
1187 		       CTLTYPE_NODE, "udp",
1188 		       SYSCTL_DESCR("UDPv4 related settings"),
1189 		       NULL, 0, NULL, 0,
1190 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1191 
1192 	sysctl_createv(clog, 0, NULL, NULL,
1193 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1194 		       CTLTYPE_INT, "checksum",
1195 		       SYSCTL_DESCR("Compute UDP checksums"),
1196 		       NULL, 0, &udpcksum, 0,
1197 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1198 		       CTL_EOL);
1199 	sysctl_createv(clog, 0, NULL, NULL,
1200 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1201 		       CTLTYPE_INT, "sendspace",
1202 		       SYSCTL_DESCR("Default UDP send buffer size"),
1203 		       NULL, 0, &udp_sendspace, 0,
1204 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1205 		       CTL_EOL);
1206 	sysctl_createv(clog, 0, NULL, NULL,
1207 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1208 		       CTLTYPE_INT, "recvspace",
1209 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1210 		       NULL, 0, &udp_recvspace, 0,
1211 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1212 		       CTL_EOL);
1213 	sysctl_createv(clog, 0, NULL, NULL,
1214 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1215 		       CTLTYPE_INT, "do_loopback_cksum",
1216 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1217 		       NULL, 0, &udp_do_loopback_cksum, 0,
1218 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1219 		       CTL_EOL);
1220 	sysctl_createv(clog, 0, NULL, NULL,
1221 		       CTLFLAG_PERMANENT,
1222 		       CTLTYPE_STRUCT, "pcblist",
1223 		       SYSCTL_DESCR("UDP protocol control block list"),
1224 		       sysctl_inpcblist, 0, &udbtable, 0,
1225 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1226 		       CTL_EOL);
1227 	sysctl_createv(clog, 0, NULL, NULL,
1228 		       CTLFLAG_PERMANENT,
1229 		       CTLTYPE_STRUCT, "stats",
1230 		       SYSCTL_DESCR("UDP statistics"),
1231 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1232 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1233 		       CTL_EOL);
1234 }
1235 #endif
1236 
1237 void
1238 udp_statinc(u_int stat)
1239 {
1240 
1241 	KASSERT(stat < UDP_NSTATS);
1242 	UDP_STATINC(stat);
1243 }
1244 
1245 #if defined(INET) && defined(IPSEC)
1246 /*
1247  * Returns:
1248  *     1 if the packet was processed
1249  *     0 if normal UDP processing should take place
1250  *    -1 if an error occurred and m was freed
1251  */
1252 static int
1253 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1254     struct socket *so)
1255 {
1256 	size_t len;
1257 	void *data;
1258 	struct inpcb *inp;
1259 	size_t skip = 0;
1260 	size_t minlen;
1261 	size_t iphdrlen;
1262 	struct ip *ip;
1263 	struct m_tag *tag;
1264 	struct udphdr *udphdr;
1265 	u_int16_t sport, dport;
1266 	struct mbuf *m = *mp;
1267 
1268 	/*
1269 	 * Collapse the mbuf chain if the first mbuf is too short
1270 	 * The longest case is: UDP + non ESP marker + ESP.
1271 	 */
1272 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1273 	if (minlen > m->m_pkthdr.len)
1274 		minlen = m->m_pkthdr.len;
1275 
1276 	if (m->m_len < minlen) {
1277 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1278 			return -1;
1279 		}
1280 		m = *mp;
1281 	}
1282 
1283 	len = m->m_len - off;
1284 	data = mtod(m, char *) + off;
1285 	inp = sotoinpcb(so);
1286 
1287 	/* Ignore keepalive packets */
1288 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1289 		m_freem(m);
1290 		*mp = NULL; /* avoid any further processing by caller ... */
1291 		return 1;
1292 	}
1293 
1294 	/*
1295 	 * Check that the payload is long enough to hold
1296 	 * an ESP header and compute the length of encapsulation
1297 	 * header to remove
1298 	 */
1299 	if (inp->inp_flags & INP_ESPINUDP) {
1300 		u_int32_t *st = (u_int32_t *)data;
1301 
1302 		if ((len <= sizeof(struct esp)) || (*st == 0))
1303 			return 0; /* Normal UDP processing */
1304 
1305 		skip = sizeof(struct udphdr);
1306 	}
1307 
1308 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1309 		u_int32_t *st = (u_int32_t *)data;
1310 
1311 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) ||
1312 		    ((st[0] | st[1]) != 0))
1313 			return 0; /* Normal UDP processing */
1314 
1315 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1316 	}
1317 
1318 	/*
1319 	 * Get the UDP ports. They are handled in network
1320 	 * order everywhere in IPSEC_NAT_T code.
1321 	 */
1322 	udphdr = (struct udphdr *)((char *)data - skip);
1323 	sport = udphdr->uh_sport;
1324 	dport = udphdr->uh_dport;
1325 
1326 	/*
1327 	 * Remove the UDP header (and possibly the non ESP marker)
1328 	 * IP header length is iphdrlen
1329 	 * Before:
1330 	 *   <--- off --->
1331 	 *   +----+------+-----+
1332 	 *   | IP |  UDP | ESP |
1333 	 *   +----+------+-----+
1334 	 *        <-skip->
1335 	 * After:
1336 	 *          +----+-----+
1337 	 *          | IP | ESP |
1338 	 *          +----+-----+
1339 	 *   <-skip->
1340 	 */
1341 	iphdrlen = off - sizeof(struct udphdr);
1342 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1343 	m_adj(m, skip);
1344 
1345 	ip = mtod(m, struct ip *);
1346 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1347 	ip->ip_p = IPPROTO_ESP;
1348 
1349 	/*
1350 	 * We have modified the packet - it is now ESP, so we should not
1351 	 * return to UDP processing ...
1352 	 *
1353 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1354 	 * the source UDP port. This is required if we want
1355 	 * to select the right SPD for multiple hosts behind
1356 	 * same NAT
1357 	 */
1358 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1359 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1360 		m_freem(m);
1361 		return -1;
1362 	}
1363 	((u_int16_t *)(tag + 1))[0] = sport;
1364 	((u_int16_t *)(tag + 1))[1] = dport;
1365 	m_tag_prepend(m, tag);
1366 
1367 	if (ipsec_used)
1368 		ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1369 	else
1370 		m_freem(m);
1371 
1372 	/* We handled it, it shouldn't be handled by UDP */
1373 	*mp = NULL; /* avoid free by caller ... */
1374 	return 1;
1375 }
1376 #endif
1377 
1378 PR_WRAP_USRREQS(udp)
1379 #define	udp_attach	udp_attach_wrapper
1380 #define	udp_detach	udp_detach_wrapper
1381 #define	udp_accept	udp_accept_wrapper
1382 #define	udp_bind	udp_bind_wrapper
1383 #define	udp_listen	udp_listen_wrapper
1384 #define	udp_connect	udp_connect_wrapper
1385 #define	udp_connect2	udp_connect2_wrapper
1386 #define	udp_disconnect	udp_disconnect_wrapper
1387 #define	udp_shutdown	udp_shutdown_wrapper
1388 #define	udp_abort	udp_abort_wrapper
1389 #define	udp_ioctl	udp_ioctl_wrapper
1390 #define	udp_stat	udp_stat_wrapper
1391 #define	udp_peeraddr	udp_peeraddr_wrapper
1392 #define	udp_sockaddr	udp_sockaddr_wrapper
1393 #define	udp_rcvd	udp_rcvd_wrapper
1394 #define	udp_recvoob	udp_recvoob_wrapper
1395 #define	udp_send	udp_send_wrapper
1396 #define	udp_sendoob	udp_sendoob_wrapper
1397 #define	udp_purgeif	udp_purgeif_wrapper
1398 
1399 const struct pr_usrreqs udp_usrreqs = {
1400 	.pr_attach	= udp_attach,
1401 	.pr_detach	= udp_detach,
1402 	.pr_accept	= udp_accept,
1403 	.pr_bind	= udp_bind,
1404 	.pr_listen	= udp_listen,
1405 	.pr_connect	= udp_connect,
1406 	.pr_connect2	= udp_connect2,
1407 	.pr_disconnect	= udp_disconnect,
1408 	.pr_shutdown	= udp_shutdown,
1409 	.pr_abort	= udp_abort,
1410 	.pr_ioctl	= udp_ioctl,
1411 	.pr_stat	= udp_stat,
1412 	.pr_peeraddr	= udp_peeraddr,
1413 	.pr_sockaddr	= udp_sockaddr,
1414 	.pr_rcvd	= udp_rcvd,
1415 	.pr_recvoob	= udp_recvoob,
1416 	.pr_send	= udp_send,
1417 	.pr_sendoob	= udp_sendoob,
1418 	.pr_purgeif	= udp_purgeif,
1419 };
1420