xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision fc4f42693f9b1c31f39f9cf50af1bf2010325808)
1 /*	$NetBSD: udp_usrreq.c,v 1.248 2018/04/13 09:29:04 maxv 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 1.248 2018/04/13 09:29:04 maxv 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 (len < sizeof(struct udphdr)) {
378 		UDP_STATINC(UDP_STAT_BADLEN);
379 		goto bad;
380 	}
381 	if (ip_len != iphlen + len) {
382 		if (ip_len < iphlen + len) {
383 			UDP_STATINC(UDP_STAT_BADLEN);
384 			goto bad;
385 		}
386 		m_adj(m, iphlen + len - ip_len);
387 	}
388 
389 	/*
390 	 * Checksum extended UDP header and data.
391 	 */
392 	if (udp4_input_checksum(m, uh, iphlen, len))
393 		goto badcsum;
394 
395 	/* construct source and dst sockaddrs. */
396 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
397 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
398 
399 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
400 		UDP_STATINC(UDP_STAT_HDROPS);
401 		return;
402 	}
403 	if (m == NULL) {
404 		/*
405 		 * packet has been processed by ESP stuff -
406 		 * e.g. dropped NAT-T-keep-alive-packet ...
407 		 */
408 		return;
409 	}
410 
411 	ip = mtod(m, struct ip *);
412 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
413 	if (uh == NULL) {
414 		UDP_STATINC(UDP_STAT_HDROPS);
415 		return;
416 	}
417 	/* XXX Re-enforce alignment? */
418 
419 #ifdef INET6
420 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
421 		struct sockaddr_in6 src6, dst6;
422 
423 		memset(&src6, 0, sizeof(src6));
424 		src6.sin6_family = AF_INET6;
425 		src6.sin6_len = sizeof(struct sockaddr_in6);
426 		in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
427 		src6.sin6_port = uh->uh_sport;
428 		memset(&dst6, 0, sizeof(dst6));
429 		dst6.sin6_family = AF_INET6;
430 		dst6.sin6_len = sizeof(struct sockaddr_in6);
431 		in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
432 		dst6.sin6_port = uh->uh_dport;
433 
434 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
435 	}
436 #endif
437 
438 	if (n == 0) {
439 		if (m->m_flags & (M_BCAST | M_MCAST)) {
440 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
441 			goto bad;
442 		}
443 		UDP_STATINC(UDP_STAT_NOPORT);
444 #ifdef IPKDB
445 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
446 		    m, iphlen + sizeof(struct udphdr),
447 		    m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
448 			/*
449 			 * It was a debugger connect packet,
450 			 * just drop it now
451 			 */
452 			goto bad;
453 		}
454 #endif
455 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
456 		m = NULL;
457 	}
458 
459 bad:
460 	if (m)
461 		m_freem(m);
462 	return;
463 
464 badcsum:
465 	m_freem(m);
466 }
467 #endif
468 
469 #ifdef INET
470 static void
471 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
472     struct sockaddr *src, struct socket *so)
473 {
474 	struct mbuf *opts = NULL;
475 	struct mbuf *n;
476 	struct inpcb *inp;
477 
478 	KASSERT(so != NULL);
479 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET);
480 	inp = sotoinpcb(so);
481 	KASSERT(inp != NULL);
482 
483 #if defined(IPSEC)
484 	if (ipsec_used && ipsec_in_reject(m, inp)) {
485 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
486 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
487 			    0, 0);
488 		return;
489 	}
490 #endif
491 
492 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
493 		if (inp->inp_flags & INP_CONTROLOPTS ||
494 		    SOOPT_TIMESTAMP(so->so_options)) {
495 			struct ip *ip = mtod(n, struct ip *);
496 			ip_savecontrol(inp, &opts, ip, n);
497 		}
498 
499 		m_adj(n, off);
500 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
501 			m_freem(n);
502 			if (opts)
503 				m_freem(opts);
504 			UDP_STATINC(UDP_STAT_FULLSOCK);
505 			soroverflow(so);
506 		} else
507 			sorwakeup(so);
508 	}
509 }
510 #endif
511 
512 #ifdef INET
513 static int
514 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
515     struct mbuf **mp, int off /* offset of udphdr */)
516 {
517 	u_int16_t *sport, *dport;
518 	int rcvcnt;
519 	struct in_addr *src4, *dst4;
520 	struct inpcb_hdr *inph;
521 	struct inpcb *inp;
522 	struct mbuf *m = *mp;
523 
524 	rcvcnt = 0;
525 	off += sizeof(struct udphdr);	/* now, offset of payload */
526 
527 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
528 		goto bad;
529 
530 	src4 = &src->sin_addr;
531 	sport = &src->sin_port;
532 	dst4 = &dst->sin_addr;
533 	dport = &dst->sin_port;
534 
535 	if (IN_MULTICAST(dst4->s_addr) ||
536 	    in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
537 		/*
538 		 * Deliver a multicast or broadcast datagram to *all* sockets
539 		 * for which the local and remote addresses and ports match
540 		 * those of the incoming datagram.  This allows more than
541 		 * one process to receive multi/broadcasts on the same port.
542 		 * (This really ought to be done for unicast datagrams as
543 		 * well, but that would cause problems with existing
544 		 * applications that open both address-specific sockets and
545 		 * a wildcard socket listening to the same port -- they would
546 		 * end up receiving duplicates of every unicast datagram.
547 		 * Those applications open the multiple sockets to overcome an
548 		 * inadequacy of the UDP socket interface, but for backwards
549 		 * compatibility we avoid the problem here rather than
550 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
551 		 */
552 
553 		/*
554 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
555 		 * we need udpiphdr for IPsec processing so we do that later.
556 		 */
557 		/*
558 		 * Locate pcb(s) for datagram.
559 		 */
560 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
561 			inp = (struct inpcb *)inph;
562 			if (inp->inp_af != AF_INET)
563 				continue;
564 
565 			if (inp->inp_lport != *dport)
566 				continue;
567 			if (!in_nullhost(inp->inp_laddr)) {
568 				if (!in_hosteq(inp->inp_laddr, *dst4))
569 					continue;
570 			}
571 			if (!in_nullhost(inp->inp_faddr)) {
572 				if (!in_hosteq(inp->inp_faddr, *src4) ||
573 				    inp->inp_fport != *sport)
574 					continue;
575 			}
576 
577 			udp4_sendup(m, off, (struct sockaddr *)src,
578 			    inp->inp_socket);
579 			rcvcnt++;
580 
581 			/*
582 			 * Don't look for additional matches if this one does
583 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
584 			 * socket options set.  This heuristic avoids searching
585 			 * through all pcbs in the common case of a non-shared
586 			 * port.  It assumes that an application will never
587 			 * clear these options after setting them.
588 			 */
589 			if ((inp->inp_socket->so_options &
590 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
591 				break;
592 		}
593 	} else {
594 		/*
595 		 * Locate pcb for datagram.
596 		 */
597 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
598 		    *dport, 0);
599 		if (inp == 0) {
600 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
601 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
602 			if (inp == 0)
603 				return rcvcnt;
604 		}
605 
606 #ifdef IPSEC
607 		/* Handle ESP over UDP */
608 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
609 			struct sockaddr *sa = (struct sockaddr *)src;
610 
611 			switch (udp4_espinudp(mp, off, sa, inp->inp_socket)) {
612 			case -1: /* Error, m was freed */
613 				rcvcnt = -1;
614 				goto bad;
615 
616 			case 1: /* ESP over UDP */
617 				rcvcnt++;
618 				goto bad;
619 
620 			case 0: /* plain UDP */
621 			default: /* Unexpected */
622 				/*
623 				 * Normal UDP processing will take place,
624 				 * m may have changed.
625 				 */
626 				m = *mp;
627 				break;
628 			}
629 		}
630 #endif
631 
632 		/*
633 		 * Check the minimum TTL for socket.
634 		 */
635 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
636 			goto bad;
637 
638 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
639 		rcvcnt++;
640 	}
641 
642 bad:
643 	return rcvcnt;
644 }
645 #endif
646 
647 #ifdef INET
648 /*
649  * Notify a udp user of an asynchronous error;
650  * just wake up so that he can collect error status.
651  */
652 static void
653 udp_notify(struct inpcb *inp, int errno)
654 {
655 	inp->inp_socket->so_error = errno;
656 	sorwakeup(inp->inp_socket);
657 	sowwakeup(inp->inp_socket);
658 }
659 
660 void *
661 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
662 {
663 	struct ip *ip = v;
664 	struct udphdr *uh;
665 	void (*notify)(struct inpcb *, int) = udp_notify;
666 	int errno;
667 
668 	if (sa->sa_family != AF_INET ||
669 	    sa->sa_len != sizeof(struct sockaddr_in))
670 		return NULL;
671 	if ((unsigned)cmd >= PRC_NCMDS)
672 		return NULL;
673 
674 	errno = inetctlerrmap[cmd];
675 	if (PRC_IS_REDIRECT(cmd)) {
676 		notify = in_rtchange;
677 		ip = NULL;
678 	} else if (cmd == PRC_HOSTDEAD) {
679 		ip = NULL;
680 	} else if (errno == 0) {
681 		return NULL;
682 	}
683 
684 	if (ip) {
685 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
686 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
687 		    ip->ip_src, uh->uh_sport, errno, notify);
688 		/* XXX mapped address case */
689 	} else {
690 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
691 		    notify);
692 	}
693 
694 	return NULL;
695 }
696 
697 int
698 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
699 {
700 	int s;
701 	int error = 0;
702 	struct inpcb *inp;
703 	int family;
704 	int optval;
705 
706 	family = so->so_proto->pr_domain->dom_family;
707 
708 	s = splsoftnet();
709 	switch (family) {
710 #ifdef INET
711 	case PF_INET:
712 		if (sopt->sopt_level != IPPROTO_UDP) {
713 			error = ip_ctloutput(op, so, sopt);
714 			goto end;
715 		}
716 		break;
717 #endif
718 #ifdef INET6
719 	case PF_INET6:
720 		if (sopt->sopt_level != IPPROTO_UDP) {
721 			error = ip6_ctloutput(op, so, sopt);
722 			goto end;
723 		}
724 		break;
725 #endif
726 	default:
727 		error = EAFNOSUPPORT;
728 		goto end;
729 	}
730 
731 
732 	switch (op) {
733 	case PRCO_SETOPT:
734 		inp = sotoinpcb(so);
735 
736 		switch (sopt->sopt_name) {
737 		case UDP_ENCAP:
738 			error = sockopt_getint(sopt, &optval);
739 			if (error)
740 				break;
741 
742 			switch(optval) {
743 			case 0:
744 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
745 				break;
746 
747 			case UDP_ENCAP_ESPINUDP:
748 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
749 				inp->inp_flags |= INP_ESPINUDP;
750 				break;
751 
752 			case UDP_ENCAP_ESPINUDP_NON_IKE:
753 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
754 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
755 				break;
756 			default:
757 				error = EINVAL;
758 				break;
759 			}
760 			break;
761 
762 		default:
763 			error = ENOPROTOOPT;
764 			break;
765 		}
766 		break;
767 
768 	default:
769 		error = EINVAL;
770 		break;
771 	}
772 
773 end:
774 	splx(s);
775 	return error;
776 }
777 
778 int
779 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control,
780     struct lwp *l)
781 {
782 	struct udpiphdr *ui;
783 	struct route *ro;
784 	struct ip_pktopts pktopts;
785 	kauth_cred_t cred;
786 	int len = m->m_pkthdr.len;
787 	int error, flags = 0;
788 
789 	MCLAIM(m, &udp_tx_mowner);
790 
791 	/*
792 	 * Calculate data length and get a mbuf
793 	 * for UDP and IP headers.
794 	 */
795 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
796 	if (m == NULL) {
797 		error = ENOBUFS;
798 		goto release;
799 	}
800 
801 	/*
802 	 * Compute the packet length of the IP header, and
803 	 * punt if the length looks bogus.
804 	 */
805 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
806 		error = EMSGSIZE;
807 		goto release;
808 	}
809 
810 	if (l == NULL)
811 		cred = NULL;
812 	else
813 		cred = l->l_cred;
814 
815 	/* Setup IP outgoing packet options */
816 	memset(&pktopts, 0, sizeof(pktopts));
817 	error = ip_setpktopts(control, &pktopts, &flags, inp, cred);
818 	if (error != 0)
819 		goto release;
820 
821 	if (control != NULL) {
822 		m_freem(control);
823 		control = NULL;
824 	}
825 
826 	/*
827 	 * Fill in mbuf with extended UDP header
828 	 * and addresses and length put into network format.
829 	 */
830 	ui = mtod(m, struct udpiphdr *);
831 	ui->ui_pr = IPPROTO_UDP;
832 	ui->ui_src = pktopts.ippo_laddr.sin_addr;
833 	ui->ui_dst = inp->inp_faddr;
834 	ui->ui_sport = inp->inp_lport;
835 	ui->ui_dport = inp->inp_fport;
836 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
837 
838 	ro = &inp->inp_route;
839 
840 	/*
841 	 * Set up checksum and output datagram.
842 	 */
843 	if (udpcksum) {
844 		/*
845 		 * XXX Cache pseudo-header checksum part for
846 		 * XXX "connected" UDP sockets.
847 		 */
848 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
849 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
850 		    sizeof(struct udphdr) + IPPROTO_UDP));
851 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
852 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
853 	} else
854 		ui->ui_sum = 0;
855 
856 	((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
857 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
858 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
859 	UDP_STATINC(UDP_STAT_OPACKETS);
860 
861 	flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST);
862 	return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp);
863 
864  release:
865 	if (control != NULL)
866 		m_freem(control);
867 	m_freem(m);
868 	return error;
869 }
870 
871 static int
872 udp_attach(struct socket *so, int proto)
873 {
874 	struct inpcb *inp;
875 	int error;
876 
877 	KASSERT(sotoinpcb(so) == NULL);
878 
879 	/* Assign the lock (must happen even if we will error out). */
880 	sosetlock(so);
881 
882 #ifdef MBUFTRACE
883 	so->so_mowner = &udp_mowner;
884 	so->so_rcv.sb_mowner = &udp_rx_mowner;
885 	so->so_snd.sb_mowner = &udp_tx_mowner;
886 #endif
887 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
888 		error = soreserve(so, udp_sendspace, udp_recvspace);
889 		if (error) {
890 			return error;
891 		}
892 	}
893 
894 	error = in_pcballoc(so, &udbtable);
895 	if (error) {
896 		return error;
897 	}
898 	inp = sotoinpcb(so);
899 	inp->inp_ip.ip_ttl = ip_defttl;
900 	KASSERT(solocked(so));
901 
902 	return error;
903 }
904 
905 static void
906 udp_detach(struct socket *so)
907 {
908 	struct inpcb *inp;
909 
910 	KASSERT(solocked(so));
911 	inp = sotoinpcb(so);
912 	KASSERT(inp != NULL);
913 	in_pcbdetach(inp);
914 }
915 
916 static int
917 udp_accept(struct socket *so, struct sockaddr *nam)
918 {
919 	KASSERT(solocked(so));
920 
921 	panic("udp_accept");
922 
923 	return EOPNOTSUPP;
924 }
925 
926 static int
927 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
928 {
929 	struct inpcb *inp = sotoinpcb(so);
930 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
931 	int error = 0;
932 	int s;
933 
934 	KASSERT(solocked(so));
935 	KASSERT(inp != NULL);
936 	KASSERT(nam != NULL);
937 
938 	s = splsoftnet();
939 	error = in_pcbbind(inp, sin, l);
940 	splx(s);
941 
942 	return error;
943 }
944 
945 static int
946 udp_listen(struct socket *so, struct lwp *l)
947 {
948 	KASSERT(solocked(so));
949 
950 	return EOPNOTSUPP;
951 }
952 
953 static int
954 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
955 {
956 	struct inpcb *inp = sotoinpcb(so);
957 	int error = 0;
958 	int s;
959 
960 	KASSERT(solocked(so));
961 	KASSERT(inp != NULL);
962 	KASSERT(nam != NULL);
963 
964 	s = splsoftnet();
965 	error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
966 	if (! error)
967 		soisconnected(so);
968 	splx(s);
969 	return error;
970 }
971 
972 static int
973 udp_connect2(struct socket *so, struct socket *so2)
974 {
975 	KASSERT(solocked(so));
976 
977 	return EOPNOTSUPP;
978 }
979 
980 static int
981 udp_disconnect(struct socket *so)
982 {
983 	struct inpcb *inp = sotoinpcb(so);
984 	int s;
985 
986 	KASSERT(solocked(so));
987 	KASSERT(inp != NULL);
988 
989 	s = splsoftnet();
990 	/*soisdisconnected(so);*/
991 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
992 	in_pcbdisconnect(inp);
993 	inp->inp_laddr = zeroin_addr;		/* XXX */
994 	in_pcbstate(inp, INP_BOUND);		/* XXX */
995 	splx(s);
996 
997 	return 0;
998 }
999 
1000 static int
1001 udp_shutdown(struct socket *so)
1002 {
1003 	int s;
1004 
1005 	KASSERT(solocked(so));
1006 
1007 	s = splsoftnet();
1008 	socantsendmore(so);
1009 	splx(s);
1010 
1011 	return 0;
1012 }
1013 
1014 static int
1015 udp_abort(struct socket *so)
1016 {
1017 	KASSERT(solocked(so));
1018 
1019 	panic("udp_abort");
1020 
1021 	return EOPNOTSUPP;
1022 }
1023 
1024 static int
1025 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
1026 {
1027 	return in_control(so, cmd, nam, ifp);
1028 }
1029 
1030 static int
1031 udp_stat(struct socket *so, struct stat *ub)
1032 {
1033 	KASSERT(solocked(so));
1034 
1035 	/* stat: don't bother with a blocksize. */
1036 	return 0;
1037 }
1038 
1039 static int
1040 udp_peeraddr(struct socket *so, struct sockaddr *nam)
1041 {
1042 	int s;
1043 
1044 	KASSERT(solocked(so));
1045 	KASSERT(sotoinpcb(so) != NULL);
1046 	KASSERT(nam != NULL);
1047 
1048 	s = splsoftnet();
1049 	in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1050 	splx(s);
1051 
1052 	return 0;
1053 }
1054 
1055 static int
1056 udp_sockaddr(struct socket *so, struct sockaddr *nam)
1057 {
1058 	int s;
1059 
1060 	KASSERT(solocked(so));
1061 	KASSERT(sotoinpcb(so) != NULL);
1062 	KASSERT(nam != NULL);
1063 
1064 	s = splsoftnet();
1065 	in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1066 	splx(s);
1067 
1068 	return 0;
1069 }
1070 
1071 static int
1072 udp_rcvd(struct socket *so, int flags, struct lwp *l)
1073 {
1074 	KASSERT(solocked(so));
1075 
1076 	return EOPNOTSUPP;
1077 }
1078 
1079 static int
1080 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
1081 {
1082 	KASSERT(solocked(so));
1083 
1084 	return EOPNOTSUPP;
1085 }
1086 
1087 static int
1088 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1089     struct mbuf *control, struct lwp *l)
1090 {
1091 	struct inpcb *inp = sotoinpcb(so);
1092 	int error = 0;
1093 	struct in_addr laddr;			/* XXX */
1094 	int s;
1095 
1096 	KASSERT(solocked(so));
1097 	KASSERT(inp != NULL);
1098 	KASSERT(m != NULL);
1099 
1100 	memset(&laddr, 0, sizeof laddr);
1101 
1102 	s = splsoftnet();
1103 	if (nam) {
1104 		laddr = inp->inp_laddr;		/* XXX */
1105 		if ((so->so_state & SS_ISCONNECTED) != 0) {
1106 			error = EISCONN;
1107 			goto die;
1108 		}
1109 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
1110 		if (error)
1111 			goto die;
1112 	} else {
1113 		if ((so->so_state & SS_ISCONNECTED) == 0) {
1114 			error = ENOTCONN;
1115 			goto die;
1116 		}
1117 	}
1118 	error = udp_output(m, inp, control, l);
1119 	m = NULL;
1120 	control = NULL;
1121 	if (nam) {
1122 		in_pcbdisconnect(inp);
1123 		inp->inp_laddr = laddr;		/* XXX */
1124 		in_pcbstate(inp, INP_BOUND);	/* XXX */
1125 	}
1126   die:
1127 	if (m != NULL)
1128 		m_freem(m);
1129 	if (control != NULL)
1130 		m_freem(control);
1131 
1132 	splx(s);
1133 	return error;
1134 }
1135 
1136 static int
1137 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1138 {
1139 	KASSERT(solocked(so));
1140 
1141 	m_freem(m);
1142 	m_freem(control);
1143 
1144 	return EOPNOTSUPP;
1145 }
1146 
1147 static int
1148 udp_purgeif(struct socket *so, struct ifnet *ifp)
1149 {
1150 	int s;
1151 
1152 	s = splsoftnet();
1153 	mutex_enter(softnet_lock);
1154 	in_pcbpurgeif0(&udbtable, ifp);
1155 #ifdef NET_MPSAFE
1156 	mutex_exit(softnet_lock);
1157 #endif
1158 	in_purgeif(ifp);
1159 #ifdef NET_MPSAFE
1160 	mutex_enter(softnet_lock);
1161 #endif
1162 	in_pcbpurgeif(&udbtable, ifp);
1163 	mutex_exit(softnet_lock);
1164 	splx(s);
1165 
1166 	return 0;
1167 }
1168 
1169 static int
1170 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1171 {
1172 
1173 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1174 }
1175 
1176 /*
1177  * Sysctl for udp variables.
1178  */
1179 static void
1180 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1181 {
1182 
1183 	sysctl_createv(clog, 0, NULL, NULL,
1184 		       CTLFLAG_PERMANENT,
1185 		       CTLTYPE_NODE, "inet", NULL,
1186 		       NULL, 0, NULL, 0,
1187 		       CTL_NET, PF_INET, CTL_EOL);
1188 	sysctl_createv(clog, 0, NULL, NULL,
1189 		       CTLFLAG_PERMANENT,
1190 		       CTLTYPE_NODE, "udp",
1191 		       SYSCTL_DESCR("UDPv4 related settings"),
1192 		       NULL, 0, NULL, 0,
1193 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1194 
1195 	sysctl_createv(clog, 0, NULL, NULL,
1196 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1197 		       CTLTYPE_INT, "checksum",
1198 		       SYSCTL_DESCR("Compute UDP checksums"),
1199 		       NULL, 0, &udpcksum, 0,
1200 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1201 		       CTL_EOL);
1202 	sysctl_createv(clog, 0, NULL, NULL,
1203 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1204 		       CTLTYPE_INT, "sendspace",
1205 		       SYSCTL_DESCR("Default UDP send buffer size"),
1206 		       NULL, 0, &udp_sendspace, 0,
1207 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1208 		       CTL_EOL);
1209 	sysctl_createv(clog, 0, NULL, NULL,
1210 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1211 		       CTLTYPE_INT, "recvspace",
1212 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1213 		       NULL, 0, &udp_recvspace, 0,
1214 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1215 		       CTL_EOL);
1216 	sysctl_createv(clog, 0, NULL, NULL,
1217 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1218 		       CTLTYPE_INT, "do_loopback_cksum",
1219 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1220 		       NULL, 0, &udp_do_loopback_cksum, 0,
1221 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1222 		       CTL_EOL);
1223 	sysctl_createv(clog, 0, NULL, NULL,
1224 		       CTLFLAG_PERMANENT,
1225 		       CTLTYPE_STRUCT, "pcblist",
1226 		       SYSCTL_DESCR("UDP protocol control block list"),
1227 		       sysctl_inpcblist, 0, &udbtable, 0,
1228 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1229 		       CTL_EOL);
1230 	sysctl_createv(clog, 0, NULL, NULL,
1231 		       CTLFLAG_PERMANENT,
1232 		       CTLTYPE_STRUCT, "stats",
1233 		       SYSCTL_DESCR("UDP statistics"),
1234 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1235 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1236 		       CTL_EOL);
1237 }
1238 #endif
1239 
1240 void
1241 udp_statinc(u_int stat)
1242 {
1243 
1244 	KASSERT(stat < UDP_NSTATS);
1245 	UDP_STATINC(stat);
1246 }
1247 
1248 #if defined(INET) && defined(IPSEC)
1249 /*
1250  * Returns:
1251  *     1 if the packet was processed
1252  *     0 if normal UDP processing should take place
1253  *    -1 if an error occurred and m was freed
1254  */
1255 static int
1256 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1257     struct socket *so)
1258 {
1259 	size_t len;
1260 	void *data;
1261 	struct inpcb *inp;
1262 	size_t skip = 0;
1263 	size_t minlen;
1264 	size_t iphdrlen;
1265 	struct ip *ip;
1266 	struct m_tag *tag;
1267 	struct udphdr *udphdr;
1268 	u_int16_t sport, dport;
1269 	struct mbuf *m = *mp;
1270 
1271 	/*
1272 	 * Collapse the mbuf chain if the first mbuf is too short
1273 	 * The longest case is: UDP + non ESP marker + ESP.
1274 	 */
1275 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1276 	if (minlen > m->m_pkthdr.len)
1277 		minlen = m->m_pkthdr.len;
1278 
1279 	if (m->m_len < minlen) {
1280 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1281 			return -1;
1282 		}
1283 		m = *mp;
1284 	}
1285 
1286 	len = m->m_len - off;
1287 	data = mtod(m, char *) + off;
1288 	inp = sotoinpcb(so);
1289 
1290 	/* Ignore keepalive packets */
1291 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1292 		m_freem(m);
1293 		*mp = NULL; /* avoid any further processing by caller ... */
1294 		return 1;
1295 	}
1296 
1297 	/*
1298 	 * Check that the payload is long enough to hold
1299 	 * an ESP header and compute the length of encapsulation
1300 	 * header to remove
1301 	 */
1302 	if (inp->inp_flags & INP_ESPINUDP) {
1303 		u_int32_t *st = (u_int32_t *)data;
1304 
1305 		if ((len <= sizeof(struct esp)) || (*st == 0))
1306 			return 0; /* Normal UDP processing */
1307 
1308 		skip = sizeof(struct udphdr);
1309 	}
1310 
1311 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1312 		u_int32_t *st = (u_int32_t *)data;
1313 
1314 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp)) ||
1315 		    ((st[0] | st[1]) != 0))
1316 			return 0; /* Normal UDP processing */
1317 
1318 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1319 	}
1320 
1321 	/*
1322 	 * Get the UDP ports. They are handled in network
1323 	 * order everywhere in IPSEC_NAT_T code.
1324 	 */
1325 	udphdr = (struct udphdr *)((char *)data - skip);
1326 	sport = udphdr->uh_sport;
1327 	dport = udphdr->uh_dport;
1328 
1329 	/*
1330 	 * Remove the UDP header (and possibly the non ESP marker)
1331 	 * IP header length is iphdrlen
1332 	 * Before:
1333 	 *   <--- off --->
1334 	 *   +----+------+-----+
1335 	 *   | IP |  UDP | ESP |
1336 	 *   +----+------+-----+
1337 	 *        <-skip->
1338 	 * After:
1339 	 *          +----+-----+
1340 	 *          | IP | ESP |
1341 	 *          +----+-----+
1342 	 *   <-skip->
1343 	 */
1344 	iphdrlen = off - sizeof(struct udphdr);
1345 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1346 	m_adj(m, skip);
1347 
1348 	ip = mtod(m, struct ip *);
1349 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1350 	ip->ip_p = IPPROTO_ESP;
1351 
1352 	/*
1353 	 * We have modified the packet - it is now ESP, so we should not
1354 	 * return to UDP processing ...
1355 	 *
1356 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1357 	 * the source UDP port. This is required if we want
1358 	 * to select the right SPD for multiple hosts behind
1359 	 * same NAT
1360 	 */
1361 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1362 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1363 		m_freem(m);
1364 		return -1;
1365 	}
1366 	((u_int16_t *)(tag + 1))[0] = sport;
1367 	((u_int16_t *)(tag + 1))[1] = dport;
1368 	m_tag_prepend(m, tag);
1369 
1370 	if (ipsec_used)
1371 		ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1372 	else
1373 		m_freem(m);
1374 
1375 	/* We handled it, it shouldn't be handled by UDP */
1376 	*mp = NULL; /* avoid free by caller ... */
1377 	return 1;
1378 }
1379 #endif
1380 
1381 PR_WRAP_USRREQS(udp)
1382 #define	udp_attach	udp_attach_wrapper
1383 #define	udp_detach	udp_detach_wrapper
1384 #define	udp_accept	udp_accept_wrapper
1385 #define	udp_bind	udp_bind_wrapper
1386 #define	udp_listen	udp_listen_wrapper
1387 #define	udp_connect	udp_connect_wrapper
1388 #define	udp_connect2	udp_connect2_wrapper
1389 #define	udp_disconnect	udp_disconnect_wrapper
1390 #define	udp_shutdown	udp_shutdown_wrapper
1391 #define	udp_abort	udp_abort_wrapper
1392 #define	udp_ioctl	udp_ioctl_wrapper
1393 #define	udp_stat	udp_stat_wrapper
1394 #define	udp_peeraddr	udp_peeraddr_wrapper
1395 #define	udp_sockaddr	udp_sockaddr_wrapper
1396 #define	udp_rcvd	udp_rcvd_wrapper
1397 #define	udp_recvoob	udp_recvoob_wrapper
1398 #define	udp_send	udp_send_wrapper
1399 #define	udp_sendoob	udp_sendoob_wrapper
1400 #define	udp_purgeif	udp_purgeif_wrapper
1401 
1402 const struct pr_usrreqs udp_usrreqs = {
1403 	.pr_attach	= udp_attach,
1404 	.pr_detach	= udp_detach,
1405 	.pr_accept	= udp_accept,
1406 	.pr_bind	= udp_bind,
1407 	.pr_listen	= udp_listen,
1408 	.pr_connect	= udp_connect,
1409 	.pr_connect2	= udp_connect2,
1410 	.pr_disconnect	= udp_disconnect,
1411 	.pr_shutdown	= udp_shutdown,
1412 	.pr_abort	= udp_abort,
1413 	.pr_ioctl	= udp_ioctl,
1414 	.pr_stat	= udp_stat,
1415 	.pr_peeraddr	= udp_peeraddr,
1416 	.pr_sockaddr	= udp_sockaddr,
1417 	.pr_rcvd	= udp_rcvd,
1418 	.pr_recvoob	= udp_recvoob,
1419 	.pr_send	= udp_send,
1420 	.pr_sendoob	= udp_sendoob,
1421 	.pr_purgeif	= udp_purgeif,
1422 };
1423