xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 27527e67bbdf8d9ec84fd58803048ed6d181ece2)
1 /*	$NetBSD: udp_usrreq.c,v 1.146 2006/01/21 00:15:36 rpaulo 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 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.146 2006/01/21 00:15:36 rpaulo Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_ipsec.h"
68 #include "opt_inet_csum.h"
69 #include "opt_ipkdb.h"
70 #include "opt_mbuftrace.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/stat.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_icmp.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 
98 #ifdef IPSEC_NAT_T
99 #include <netinet6/ipsec.h>
100 #include <netinet6/esp.h>
101 #endif
102 
103 #ifdef INET6
104 #include <netinet/ip6.h>
105 #include <netinet/icmp6.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/udp6_var.h>
109 #include <netinet6/scope6_var.h>
110 #endif
111 
112 #ifndef INET6
113 /* always need ip6.h for IP6_EXTHDR_GET */
114 #include <netinet/ip6.h>
115 #endif
116 
117 #include "faith.h"
118 #if defined(NFAITH) && NFAITH > 0
119 #include <net/if_faith.h>
120 #endif
121 
122 #include <machine/stdarg.h>
123 
124 #ifdef FAST_IPSEC
125 #include <netipsec/ipsec.h>
126 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
127 #ifdef INET6
128 #include <netipsec/ipsec6.h>
129 #endif
130 #endif	/* FAST_IPSEC*/
131 
132 #ifdef IPSEC
133 #include <netinet6/ipsec.h>
134 #include <netkey/key.h>
135 #endif /*IPSEC*/
136 
137 #ifdef IPKDB
138 #include <ipkdb/ipkdb.h>
139 #endif
140 
141 /*
142  * UDP protocol implementation.
143  * Per RFC 768, August, 1980.
144  */
145 #ifndef	COMPAT_42
146 int	udpcksum = 1;
147 #else
148 int	udpcksum = 0;		/* XXX */
149 #endif
150 int	udp_do_loopback_cksum = 0;
151 
152 struct	inpcbtable udbtable;
153 struct	udpstat udpstat;
154 
155 #ifdef INET
156 #ifdef IPSEC_NAT_T
157 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
158 	struct socket *);
159 #endif
160 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
161 	struct socket *);
162 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
163 	struct mbuf **, int);
164 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
165 #endif
166 #ifdef INET6
167 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
168 	struct socket *);
169 static int udp6_realinput (int, struct sockaddr_in6 *,
170 	struct sockaddr_in6 *, struct mbuf *, int);
171 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
172 #endif
173 #ifdef INET
174 static	void udp_notify (struct inpcb *, int);
175 #endif
176 
177 #ifndef UDBHASHSIZE
178 #define	UDBHASHSIZE	128
179 #endif
180 int	udbhashsize = UDBHASHSIZE;
181 
182 #ifdef MBUFTRACE
183 struct mowner udp_mowner = { "udp" };
184 struct mowner udp_rx_mowner = { "udp", "rx" };
185 struct mowner udp_tx_mowner = { "udp", "tx" };
186 #endif
187 
188 #ifdef UDP_CSUM_COUNTERS
189 #include <sys/device.h>
190 
191 #if defined(INET)
192 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
193     NULL, "udp", "hwcsum bad");
194 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
195     NULL, "udp", "hwcsum ok");
196 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
197     NULL, "udp", "hwcsum data");
198 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
199     NULL, "udp", "swcsum");
200 
201 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
202 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
203 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
204 EVCNT_ATTACH_STATIC(udp_swcsum);
205 #endif /* defined(INET) */
206 
207 #if defined(INET6)
208 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
209     NULL, "udp6", "hwcsum bad");
210 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
211     NULL, "udp6", "hwcsum ok");
212 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
213     NULL, "udp6", "hwcsum data");
214 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
215     NULL, "udp6", "swcsum");
216 
217 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
218 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
219 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
220 EVCNT_ATTACH_STATIC(udp6_swcsum);
221 #endif /* defined(INET6) */
222 
223 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
224 
225 #else
226 
227 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
228 
229 #endif /* UDP_CSUM_COUNTERS */
230 
231 void
232 udp_init(void)
233 {
234 
235 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
236 
237 	MOWNER_ATTACH(&udp_tx_mowner);
238 	MOWNER_ATTACH(&udp_rx_mowner);
239 	MOWNER_ATTACH(&udp_mowner);
240 }
241 
242 /*
243  * Checksum extended UDP header and data.
244  */
245 
246 int
247 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
248     int iphlen, int len)
249 {
250 
251 	switch (af) {
252 #ifdef INET
253 	case AF_INET:
254 		return udp4_input_checksum(m, uh, iphlen, len);
255 #endif
256 #ifdef INET6
257 	case AF_INET6:
258 		return udp6_input_checksum(m, uh, iphlen, len);
259 #endif
260 	}
261 #ifdef DIAGNOSTIC
262 	panic("udp_input_checksum: unknown af %d", af);
263 #endif
264 	/* NOTREACHED */
265 	return -1;
266 }
267 
268 #ifdef INET
269 
270 /*
271  * Checksum extended UDP header and data.
272  */
273 
274 static int
275 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
276     int iphlen, int len)
277 {
278 
279 	/*
280 	 * XXX it's better to record and check if this mbuf is
281 	 * already checked.
282 	 */
283 
284 	if (uh->uh_sum == 0)
285 		return 0;
286 
287 	switch (m->m_pkthdr.csum_flags &
288 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
289 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
290 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
291 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
292 		goto badcsum;
293 
294 	case M_CSUM_UDPv4|M_CSUM_DATA: {
295 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
296 
297 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
298 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
299 			const struct ip *ip =
300 			    mtod(m, const struct ip *);
301 
302 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
303 			    ip->ip_dst.s_addr,
304 			    htons(hw_csum + len + IPPROTO_UDP));
305 		}
306 		if ((hw_csum ^ 0xffff) != 0)
307 			goto badcsum;
308 		break;
309 	}
310 
311 	case M_CSUM_UDPv4:
312 		/* Checksum was okay. */
313 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
314 		break;
315 
316 	default:
317 		/*
318 		 * Need to compute it ourselves.  Maybe skip checksum
319 		 * on loopback interfaces.
320 		 */
321 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
322 				     IFF_LOOPBACK) ||
323 				   udp_do_loopback_cksum)) {
324 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
325 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
326 				goto badcsum;
327 		}
328 		break;
329 	}
330 
331 	return 0;
332 
333 badcsum:
334 	udpstat.udps_badsum++;
335 	return -1;
336 }
337 
338 void
339 udp_input(struct mbuf *m, ...)
340 {
341 	va_list ap;
342 	struct sockaddr_in src, dst;
343 	struct ip *ip;
344 	struct udphdr *uh;
345 	int iphlen;
346 	int len;
347 	int n;
348 	u_int16_t ip_len;
349 
350 	va_start(ap, m);
351 	iphlen = va_arg(ap, int);
352 	(void)va_arg(ap, int);		/* ignore value, advance ap */
353 	va_end(ap);
354 
355 	MCLAIM(m, &udp_rx_mowner);
356 	udpstat.udps_ipackets++;
357 
358 	/*
359 	 * Get IP and UDP header together in first mbuf.
360 	 */
361 	ip = mtod(m, struct ip *);
362 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
363 	if (uh == NULL) {
364 		udpstat.udps_hdrops++;
365 		return;
366 	}
367 	KASSERT(UDP_HDR_ALIGNED_P(uh));
368 
369 	/* destination port of 0 is illegal, based on RFC768. */
370 	if (uh->uh_dport == 0)
371 		goto bad;
372 
373 	/*
374 	 * Make mbuf data length reflect UDP length.
375 	 * If not enough data to reflect UDP length, drop.
376 	 */
377 	ip_len = ntohs(ip->ip_len);
378 	len = ntohs((u_int16_t)uh->uh_ulen);
379 	if (ip_len != iphlen + len) {
380 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
381 			udpstat.udps_badlen++;
382 			goto bad;
383 		}
384 		m_adj(m, iphlen + len - ip_len);
385 	}
386 
387 	/*
388 	 * Checksum extended UDP header and data.
389 	 */
390 	if (udp4_input_checksum(m, uh, iphlen, len))
391 		goto badcsum;
392 
393 	/* construct source and dst sockaddrs. */
394 	bzero(&src, sizeof(src));
395 	src.sin_family = AF_INET;
396 	src.sin_len = sizeof(struct sockaddr_in);
397 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
398 	src.sin_port = uh->uh_sport;
399 	bzero(&dst, sizeof(dst));
400 	dst.sin_family = AF_INET;
401 	dst.sin_len = sizeof(struct sockaddr_in);
402 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
403 	dst.sin_port = uh->uh_dport;
404 
405 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
406 		udpstat.udps_hdrops++;
407 		return;
408 	}
409 #ifdef INET6
410 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
411 		struct sockaddr_in6 src6, dst6;
412 
413 		bzero(&src6, sizeof(src6));
414 		src6.sin6_family = AF_INET6;
415 		src6.sin6_len = sizeof(struct sockaddr_in6);
416 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
417 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
418 			sizeof(ip->ip_src));
419 		src6.sin6_port = uh->uh_sport;
420 		bzero(&dst6, sizeof(dst6));
421 		dst6.sin6_family = AF_INET6;
422 		dst6.sin6_len = sizeof(struct sockaddr_in6);
423 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
424 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
425 			sizeof(ip->ip_dst));
426 		dst6.sin6_port = uh->uh_dport;
427 
428 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
429 	}
430 #endif
431 
432 	if (n == 0) {
433 		if (m->m_flags & (M_BCAST | M_MCAST)) {
434 			udpstat.udps_noportbcast++;
435 			goto bad;
436 		}
437 		udpstat.udps_noport++;
438 #ifdef IPKDB
439 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
440 				m, iphlen + sizeof(struct udphdr),
441 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
442 			/*
443 			 * It was a debugger connect packet,
444 			 * just drop it now
445 			 */
446 			goto bad;
447 		}
448 #endif
449 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
450 		m = NULL;
451 	}
452 
453 bad:
454 	if (m)
455 		m_freem(m);
456 	return;
457 
458 badcsum:
459 	m_freem(m);
460 }
461 #endif
462 
463 #ifdef INET6
464 static int
465 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
466 {
467 
468 	/*
469 	 * XXX it's better to record and check if this mbuf is
470 	 * already checked.
471 	 */
472 
473 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
474 		goto good;
475 	}
476 	if (uh->uh_sum == 0) {
477 		udp6stat.udp6s_nosum++;
478 		goto bad;
479 	}
480 
481 	switch (m->m_pkthdr.csum_flags &
482 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
483 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
484 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
485 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
486 		udp6stat.udp6s_badsum++;
487 		goto bad;
488 
489 #if 0 /* notyet */
490 	case M_CSUM_UDPv6|M_CSUM_DATA:
491 #endif
492 
493 	case M_CSUM_UDPv6:
494 		/* Checksum was okay. */
495 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
496 		break;
497 
498 	default:
499 		/*
500 		 * Need to compute it ourselves.  Maybe skip checksum
501 		 * on loopback interfaces.
502 		 */
503 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
504 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
505 			udp6stat.udp6s_badsum++;
506 			goto bad;
507 		}
508 	}
509 
510 good:
511 	return 0;
512 bad:
513 	return -1;
514 }
515 
516 int
517 udp6_input(struct mbuf **mp, int *offp, int proto)
518 {
519 	struct mbuf *m = *mp;
520 	int off = *offp;
521 	struct sockaddr_in6 src, dst;
522 	struct ip6_hdr *ip6;
523 	struct udphdr *uh;
524 	u_int32_t plen, ulen;
525 
526 	ip6 = mtod(m, struct ip6_hdr *);
527 
528 #if defined(NFAITH) && 0 < NFAITH
529 	if (faithprefix(&ip6->ip6_dst)) {
530 		/* send icmp6 host unreach? */
531 		m_freem(m);
532 		return IPPROTO_DONE;
533 	}
534 #endif
535 
536 	udp6stat.udp6s_ipackets++;
537 
538 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
539 	plen = m->m_pkthdr.len - off;
540 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
541 	if (uh == NULL) {
542 		ip6stat.ip6s_tooshort++;
543 		return IPPROTO_DONE;
544 	}
545 	KASSERT(UDP_HDR_ALIGNED_P(uh));
546 	ulen = ntohs((u_short)uh->uh_ulen);
547 	/*
548 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
549 	 * iff payload length > 0xffff.
550 	 */
551 	if (ulen == 0 && plen > 0xffff)
552 		ulen = plen;
553 
554 	if (plen != ulen) {
555 		udp6stat.udp6s_badlen++;
556 		goto bad;
557 	}
558 
559 	/* destination port of 0 is illegal, based on RFC768. */
560 	if (uh->uh_dport == 0)
561 		goto bad;
562 
563 	/* Be proactive about malicious use of IPv4 mapped address */
564 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
565 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
566 		/* XXX stat */
567 		goto bad;
568 	}
569 
570 	/*
571 	 * Checksum extended UDP header and data.  Maybe skip checksum
572 	 * on loopback interfaces.
573 	 */
574 	if (udp6_input_checksum(m, uh, off, ulen))
575 		goto bad;
576 
577 	/*
578 	 * Construct source and dst sockaddrs.
579 	 */
580 	bzero(&src, sizeof(src));
581 	src.sin6_family = AF_INET6;
582 	src.sin6_len = sizeof(struct sockaddr_in6);
583 	src.sin6_addr = ip6->ip6_src;
584 	src.sin6_port = uh->uh_sport;
585 	bzero(&dst, sizeof(dst));
586 	dst.sin6_family = AF_INET6;
587 	dst.sin6_len = sizeof(struct sockaddr_in6);
588 	dst.sin6_addr = ip6->ip6_dst;
589 	dst.sin6_port = uh->uh_dport;
590 
591 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
592 		if (m->m_flags & M_MCAST) {
593 			udp6stat.udp6s_noportmcast++;
594 			goto bad;
595 		}
596 		udp6stat.udp6s_noport++;
597 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
598 		m = NULL;
599 	}
600 
601 bad:
602 	if (m)
603 		m_freem(m);
604 	return IPPROTO_DONE;
605 }
606 #endif
607 
608 #ifdef INET
609 static void
610 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
611 	struct sockaddr *src, struct socket *so)
612 {
613 	struct mbuf *opts = NULL;
614 	struct mbuf *n;
615 	struct inpcb *inp = NULL;
616 
617 	if (!so)
618 		return;
619 	switch (so->so_proto->pr_domain->dom_family) {
620 	case AF_INET:
621 		inp = sotoinpcb(so);
622 		break;
623 #ifdef INET6
624 	case AF_INET6:
625 		break;
626 #endif
627 	default:
628 		return;
629 	}
630 
631 #if defined(IPSEC) || defined(FAST_IPSEC)
632 	/* check AH/ESP integrity. */
633 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
634 		ipsecstat.in_polvio++;
635 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
636 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
637 			    0, 0);
638 		return;
639 	}
640 #endif /*IPSEC*/
641 
642 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
643 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
644 			 || so->so_options & SO_TIMESTAMP)) {
645 			struct ip *ip = mtod(n, struct ip *);
646 			ip_savecontrol(inp, &opts, ip, n);
647 		}
648 
649 		m_adj(n, off);
650 		if (sbappendaddr(&so->so_rcv, src, n,
651 				opts) == 0) {
652 			m_freem(n);
653 			if (opts)
654 				m_freem(opts);
655 			so->so_rcv.sb_overflowed++;
656 			udpstat.udps_fullsock++;
657 		} else
658 			sorwakeup(so);
659 	}
660 }
661 #endif
662 
663 #ifdef INET6
664 static void
665 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
666 	struct sockaddr *src, struct socket *so)
667 {
668 	struct mbuf *opts = NULL;
669 	struct mbuf *n;
670 	struct in6pcb *in6p = NULL;
671 
672 	if (!so)
673 		return;
674 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
675 		return;
676 	in6p = sotoin6pcb(so);
677 
678 #if defined(IPSEC) || defined(FAST_IPSEC)
679 	/* check AH/ESP integrity. */
680 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
681 		ipsec6stat.in_polvio++;
682 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
683 			icmp6_error(n, ICMP6_DST_UNREACH,
684 			    ICMP6_DST_UNREACH_ADMIN, 0);
685 		return;
686 	}
687 #endif /*IPSEC*/
688 
689 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
690 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
691 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
692 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
693 			ip6_savecontrol(in6p, &opts, ip6, n);
694 		}
695 
696 		m_adj(n, off);
697 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
698 			m_freem(n);
699 			if (opts)
700 				m_freem(opts);
701 			so->so_rcv.sb_overflowed++;
702 			udp6stat.udp6s_fullsock++;
703 		} else
704 			sorwakeup(so);
705 	}
706 }
707 #endif
708 
709 #ifdef INET
710 static int
711 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
712 	struct mbuf **mp, int off /* offset of udphdr */)
713 {
714 	u_int16_t *sport, *dport;
715 	int rcvcnt;
716 	struct in_addr *src4, *dst4;
717 	struct inpcb_hdr *inph;
718 	struct inpcb *inp;
719 	struct mbuf *m = *mp;
720 
721 	rcvcnt = 0;
722 	off += sizeof(struct udphdr);	/* now, offset of payload */
723 
724 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
725 		goto bad;
726 
727 	src4 = &src->sin_addr;
728 	sport = &src->sin_port;
729 	dst4 = &dst->sin_addr;
730 	dport = &dst->sin_port;
731 
732 	if (IN_MULTICAST(dst4->s_addr) ||
733 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
734 		/*
735 		 * Deliver a multicast or broadcast datagram to *all* sockets
736 		 * for which the local and remote addresses and ports match
737 		 * those of the incoming datagram.  This allows more than
738 		 * one process to receive multi/broadcasts on the same port.
739 		 * (This really ought to be done for unicast datagrams as
740 		 * well, but that would cause problems with existing
741 		 * applications that open both address-specific sockets and
742 		 * a wildcard socket listening to the same port -- they would
743 		 * end up receiving duplicates of every unicast datagram.
744 		 * Those applications open the multiple sockets to overcome an
745 		 * inadequacy of the UDP socket interface, but for backwards
746 		 * compatibility we avoid the problem here rather than
747 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
748 		 */
749 
750 		/*
751 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
752 		 * we need udpiphdr for IPsec processing so we do that later.
753 		 */
754 		/*
755 		 * Locate pcb(s) for datagram.
756 		 */
757 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
758 			inp = (struct inpcb *)inph;
759 			if (inp->inp_af != AF_INET)
760 				continue;
761 
762 			if (inp->inp_lport != *dport)
763 				continue;
764 			if (!in_nullhost(inp->inp_laddr)) {
765 				if (!in_hosteq(inp->inp_laddr, *dst4))
766 					continue;
767 			}
768 			if (!in_nullhost(inp->inp_faddr)) {
769 				if (!in_hosteq(inp->inp_faddr, *src4) ||
770 				    inp->inp_fport != *sport)
771 					continue;
772 			}
773 
774 			udp4_sendup(m, off, (struct sockaddr *)src,
775 				inp->inp_socket);
776 			rcvcnt++;
777 
778 			/*
779 			 * Don't look for additional matches if this one does
780 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
781 			 * socket options set.  This heuristic avoids searching
782 			 * through all pcbs in the common case of a non-shared
783 			 * port.  It assumes that an application will never
784 			 * clear these options after setting them.
785 			 */
786 			if ((inp->inp_socket->so_options &
787 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
788 				break;
789 		}
790 	} else {
791 		/*
792 		 * Locate pcb for datagram.
793 		 */
794 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
795 		if (inp == 0) {
796 			++udpstat.udps_pcbhashmiss;
797 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
798 			if (inp == 0)
799 				return rcvcnt;
800 		}
801 
802 #ifdef IPSEC_NAT_T
803 		/* Handle ESP over UDP */
804 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
805 			struct sockaddr *sa = (struct sockaddr *)src;
806 
807 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
808 			case -1: 	/* Error, m was freeed */
809 				rcvcnt = -1;
810 				goto bad;
811 				break;
812 
813 			case 1:		/* ESP over UDP */
814 				rcvcnt++;
815 				goto bad;
816 				break;
817 
818 			case 0: 	/* plain UDP */
819 			default: 	/* Unexpected */
820 				/*
821 				 * Normal UDP processing will take place
822 				 * m may have changed.
823 				 */
824 				m = *mp;
825 				break;
826 			}
827 		}
828 #endif
829 
830 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
831 		rcvcnt++;
832 	}
833 
834 bad:
835 	return rcvcnt;
836 }
837 #endif
838 
839 #ifdef INET6
840 static int
841 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
842 	struct mbuf *m, int off)
843 {
844 	u_int16_t sport, dport;
845 	int rcvcnt;
846 	struct in6_addr src6, *dst6;
847 	const struct in_addr *dst4;
848 	struct inpcb_hdr *inph;
849 	struct in6pcb *in6p;
850 
851 	rcvcnt = 0;
852 	off += sizeof(struct udphdr);	/* now, offset of payload */
853 
854 	if (af != AF_INET && af != AF_INET6)
855 		goto bad;
856 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
857 		goto bad;
858 
859 	src6 = src->sin6_addr;
860 	if (sa6_recoverscope(src) != 0) {
861 		/* XXX: should be impossible. */
862 		goto bad;
863 	}
864 	sport = src->sin6_port;
865 
866 	dport = dst->sin6_port;
867 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
868 	dst6 = &dst->sin6_addr;
869 
870 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
871 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
872 		/*
873 		 * Deliver a multicast or broadcast datagram to *all* sockets
874 		 * for which the local and remote addresses and ports match
875 		 * those of the incoming datagram.  This allows more than
876 		 * one process to receive multi/broadcasts on the same port.
877 		 * (This really ought to be done for unicast datagrams as
878 		 * well, but that would cause problems with existing
879 		 * applications that open both address-specific sockets and
880 		 * a wildcard socket listening to the same port -- they would
881 		 * end up receiving duplicates of every unicast datagram.
882 		 * Those applications open the multiple sockets to overcome an
883 		 * inadequacy of the UDP socket interface, but for backwards
884 		 * compatibility we avoid the problem here rather than
885 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
886 		 */
887 
888 		/*
889 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
890 		 * we need udpiphdr for IPsec processing so we do that later.
891 		 */
892 		/*
893 		 * Locate pcb(s) for datagram.
894 		 */
895 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
896 			in6p = (struct in6pcb *)inph;
897 			if (in6p->in6p_af != AF_INET6)
898 				continue;
899 
900 			if (in6p->in6p_lport != dport)
901 				continue;
902 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
903 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
904 				    dst6))
905 					continue;
906 			} else {
907 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
908 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
909 					continue;
910 			}
911 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
912 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
913 				    &src6) || in6p->in6p_fport != sport)
914 					continue;
915 			} else {
916 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
917 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
918 					continue;
919 			}
920 
921 			udp6_sendup(m, off, (struct sockaddr *)src,
922 				in6p->in6p_socket);
923 			rcvcnt++;
924 
925 			/*
926 			 * Don't look for additional matches if this one does
927 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
928 			 * socket options set.  This heuristic avoids searching
929 			 * through all pcbs in the common case of a non-shared
930 			 * port.  It assumes that an application will never
931 			 * clear these options after setting them.
932 			 */
933 			if ((in6p->in6p_socket->so_options &
934 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
935 				break;
936 		}
937 	} else {
938 		/*
939 		 * Locate pcb for datagram.
940 		 */
941 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
942 		    dport, 0);
943 		if (in6p == 0) {
944 			++udpstat.udps_pcbhashmiss;
945 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
946 			if (in6p == 0)
947 				return rcvcnt;
948 		}
949 
950 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
951 		rcvcnt++;
952 	}
953 
954 bad:
955 	return rcvcnt;
956 }
957 #endif
958 
959 #ifdef INET
960 /*
961  * Notify a udp user of an asynchronous error;
962  * just wake up so that he can collect error status.
963  */
964 static void
965 udp_notify(struct inpcb *inp, int errno)
966 {
967 	inp->inp_socket->so_error = errno;
968 	sorwakeup(inp->inp_socket);
969 	sowwakeup(inp->inp_socket);
970 }
971 
972 void *
973 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
974 {
975 	struct ip *ip = v;
976 	struct udphdr *uh;
977 	void (*notify)(struct inpcb *, int) = udp_notify;
978 	int errno;
979 
980 	if (sa->sa_family != AF_INET
981 	 || sa->sa_len != sizeof(struct sockaddr_in))
982 		return NULL;
983 	if ((unsigned)cmd >= PRC_NCMDS)
984 		return NULL;
985 	errno = inetctlerrmap[cmd];
986 	if (PRC_IS_REDIRECT(cmd))
987 		notify = in_rtchange, ip = 0;
988 	else if (cmd == PRC_HOSTDEAD)
989 		ip = 0;
990 	else if (errno == 0)
991 		return NULL;
992 	if (ip) {
993 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
994 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
995 		    ip->ip_src, uh->uh_sport, errno, notify);
996 
997 		/* XXX mapped address case */
998 	} else
999 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
1000 		    notify);
1001 	return NULL;
1002 }
1003 
1004 int
1005 udp_ctloutput(op, so, level, optname, mp)
1006 	int op;
1007 	struct socket *so;
1008 	int level, optname;
1009 	struct mbuf **mp;
1010 {
1011 	int s;
1012 	int error = 0;
1013 	struct mbuf *m;
1014 	struct inpcb *inp;
1015 	int family;
1016 
1017 	family = so->so_proto->pr_domain->dom_family;
1018 
1019 	s = splsoftnet();
1020 	switch (family) {
1021 #ifdef INET
1022 	case PF_INET:
1023 		if (level != IPPROTO_UDP) {
1024 			error = ip_ctloutput(op, so, level, optname, mp);
1025 			goto end;
1026 		}
1027 		break;
1028 #endif
1029 #ifdef INET6
1030 	case PF_INET6:
1031 		if (level != IPPROTO_UDP) {
1032 			error = ip6_ctloutput(op, so, level, optname, mp);
1033 			goto end;
1034 		}
1035 		break;
1036 #endif
1037 	default:
1038 		error = EAFNOSUPPORT;
1039 		goto end;
1040 		break;
1041 	}
1042 
1043 
1044 	switch (op) {
1045 	case PRCO_SETOPT:
1046 		m = *mp;
1047 		inp = sotoinpcb(so);
1048 
1049 		switch (optname) {
1050 		case UDP_ENCAP:
1051 			if (m == NULL || m->m_len < sizeof (int)) {
1052 				error = EINVAL;
1053 				goto end;
1054 			}
1055 
1056 			switch(*mtod(m, int *)) {
1057 #ifdef IPSEC_NAT_T
1058 			case 0:
1059 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1060 				break;
1061 
1062 			case UDP_ENCAP_ESPINUDP:
1063 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1064 				inp->inp_flags |= INP_ESPINUDP;
1065 				break;
1066 
1067 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1068 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1069 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1070 				break;
1071 #endif
1072 			default:
1073 				error = EINVAL;
1074 				goto end;
1075 				break;
1076 			}
1077 			break;
1078 
1079 		default:
1080 			error = ENOPROTOOPT;
1081 			goto end;
1082 			break;
1083 		}
1084 		break;
1085 
1086 	default:
1087 		error = EINVAL;
1088 		goto end;
1089 		break;
1090 	}
1091 
1092 end:
1093 	splx(s);
1094 	return error;
1095 }
1096 
1097 
1098 int
1099 udp_output(struct mbuf *m, ...)
1100 {
1101 	struct inpcb *inp;
1102 	struct udpiphdr *ui;
1103 	struct route *ro;
1104 	int len = m->m_pkthdr.len;
1105 	int error = 0;
1106 	va_list ap;
1107 
1108 	MCLAIM(m, &udp_tx_mowner);
1109 	va_start(ap, m);
1110 	inp = va_arg(ap, struct inpcb *);
1111 	va_end(ap);
1112 
1113 	/*
1114 	 * Calculate data length and get a mbuf
1115 	 * for UDP and IP headers.
1116 	 */
1117 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1118 	if (m == 0) {
1119 		error = ENOBUFS;
1120 		goto release;
1121 	}
1122 
1123 	/*
1124 	 * Compute the packet length of the IP header, and
1125 	 * punt if the length looks bogus.
1126 	 */
1127 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1128 		error = EMSGSIZE;
1129 		goto release;
1130 	}
1131 
1132 	/*
1133 	 * Fill in mbuf with extended UDP header
1134 	 * and addresses and length put into network format.
1135 	 */
1136 	ui = mtod(m, struct udpiphdr *);
1137 	ui->ui_pr = IPPROTO_UDP;
1138 	ui->ui_src = inp->inp_laddr;
1139 	ui->ui_dst = inp->inp_faddr;
1140 	ui->ui_sport = inp->inp_lport;
1141 	ui->ui_dport = inp->inp_fport;
1142 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1143 
1144 	ro = &inp->inp_route;
1145 
1146 	/*
1147 	 * Set up checksum and output datagram.
1148 	 */
1149 	if (udpcksum) {
1150 		/*
1151 		 * XXX Cache pseudo-header checksum part for
1152 		 * XXX "connected" UDP sockets.
1153 		 */
1154 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1155 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1156 		    sizeof(struct udphdr) + IPPROTO_UDP));
1157 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1158 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1159 	} else
1160 		ui->ui_sum = 0;
1161 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1162 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1163 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1164 	udpstat.udps_opackets++;
1165 
1166 	return (ip_output(m, inp->inp_options, ro,
1167 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1168 	    inp->inp_moptions, inp->inp_socket));
1169 
1170 release:
1171 	m_freem(m);
1172 	return (error);
1173 }
1174 
1175 int	udp_sendspace = 9216;		/* really max datagram size */
1176 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1177 					/* 40 1K datagrams */
1178 
1179 /*ARGSUSED*/
1180 int
1181 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1182 	struct mbuf *control, struct lwp *l)
1183 {
1184 	struct inpcb *inp;
1185 	struct proc *p;
1186 	int s;
1187 	int error = 0;
1188 
1189 	p = l ? l->l_proc : NULL;
1190 	if (req == PRU_CONTROL)
1191 		return (in_control(so, (long)m, (caddr_t)nam,
1192 		    (struct ifnet *)control, p));
1193 
1194 	if (req == PRU_PURGEIF) {
1195 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1196 		in_purgeif((struct ifnet *)control);
1197 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1198 		return (0);
1199 	}
1200 
1201 	s = splsoftnet();
1202 	inp = sotoinpcb(so);
1203 #ifdef DIAGNOSTIC
1204 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1205 		panic("udp_usrreq: unexpected control mbuf");
1206 #endif
1207 	if (inp == 0 && req != PRU_ATTACH) {
1208 		error = EINVAL;
1209 		goto release;
1210 	}
1211 
1212 	/*
1213 	 * Note: need to block udp_input while changing
1214 	 * the udp pcb queue and/or pcb addresses.
1215 	 */
1216 	switch (req) {
1217 
1218 	case PRU_ATTACH:
1219 		if (inp != 0) {
1220 			error = EISCONN;
1221 			break;
1222 		}
1223 #ifdef MBUFTRACE
1224 		so->so_mowner = &udp_mowner;
1225 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1226 		so->so_snd.sb_mowner = &udp_tx_mowner;
1227 #endif
1228 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1229 			error = soreserve(so, udp_sendspace, udp_recvspace);
1230 			if (error)
1231 				break;
1232 		}
1233 		error = in_pcballoc(so, &udbtable);
1234 		if (error)
1235 			break;
1236 		inp = sotoinpcb(so);
1237 		inp->inp_ip.ip_ttl = ip_defttl;
1238 		break;
1239 
1240 	case PRU_DETACH:
1241 		in_pcbdetach(inp);
1242 		break;
1243 
1244 	case PRU_BIND:
1245 		error = in_pcbbind(inp, nam, p);
1246 		break;
1247 
1248 	case PRU_LISTEN:
1249 		error = EOPNOTSUPP;
1250 		break;
1251 
1252 	case PRU_CONNECT:
1253 		error = in_pcbconnect(inp, nam, p);
1254 		if (error)
1255 			break;
1256 		soisconnected(so);
1257 		break;
1258 
1259 	case PRU_CONNECT2:
1260 		error = EOPNOTSUPP;
1261 		break;
1262 
1263 	case PRU_DISCONNECT:
1264 		/*soisdisconnected(so);*/
1265 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1266 		in_pcbdisconnect(inp);
1267 		inp->inp_laddr = zeroin_addr;		/* XXX */
1268 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1269 		break;
1270 
1271 	case PRU_SHUTDOWN:
1272 		socantsendmore(so);
1273 		break;
1274 
1275 	case PRU_RCVD:
1276 		error = EOPNOTSUPP;
1277 		break;
1278 
1279 	case PRU_SEND:
1280 		if (control && control->m_len) {
1281 			m_freem(control);
1282 			m_freem(m);
1283 			error = EINVAL;
1284 			break;
1285 		}
1286 	{
1287 		struct in_addr laddr;			/* XXX */
1288 
1289 		if (nam) {
1290 			laddr = inp->inp_laddr;		/* XXX */
1291 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1292 				error = EISCONN;
1293 				goto die;
1294 			}
1295 			error = in_pcbconnect(inp, nam, p);
1296 			if (error)
1297 				goto die;
1298 		} else {
1299 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1300 				error = ENOTCONN;
1301 				goto die;
1302 			}
1303 		}
1304 		error = udp_output(m, inp);
1305 		m = NULL;
1306 		if (nam) {
1307 			in_pcbdisconnect(inp);
1308 			inp->inp_laddr = laddr;		/* XXX */
1309 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1310 		}
1311 	  die:
1312 		if (m)
1313 			m_freem(m);
1314 	}
1315 		break;
1316 
1317 	case PRU_SENSE:
1318 		/*
1319 		 * stat: don't bother with a blocksize.
1320 		 */
1321 		splx(s);
1322 		return (0);
1323 
1324 	case PRU_RCVOOB:
1325 		error =  EOPNOTSUPP;
1326 		break;
1327 
1328 	case PRU_SENDOOB:
1329 		m_freem(control);
1330 		m_freem(m);
1331 		error =  EOPNOTSUPP;
1332 		break;
1333 
1334 	case PRU_SOCKADDR:
1335 		in_setsockaddr(inp, nam);
1336 		break;
1337 
1338 	case PRU_PEERADDR:
1339 		in_setpeeraddr(inp, nam);
1340 		break;
1341 
1342 	default:
1343 		panic("udp_usrreq");
1344 	}
1345 
1346 release:
1347 	splx(s);
1348 	return (error);
1349 }
1350 
1351 /*
1352  * Sysctl for udp variables.
1353  */
1354 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1355 {
1356 
1357 	sysctl_createv(clog, 0, NULL, NULL,
1358 		       CTLFLAG_PERMANENT,
1359 		       CTLTYPE_NODE, "net", NULL,
1360 		       NULL, 0, NULL, 0,
1361 		       CTL_NET, CTL_EOL);
1362 	sysctl_createv(clog, 0, NULL, NULL,
1363 		       CTLFLAG_PERMANENT,
1364 		       CTLTYPE_NODE, "inet", NULL,
1365 		       NULL, 0, NULL, 0,
1366 		       CTL_NET, PF_INET, CTL_EOL);
1367 	sysctl_createv(clog, 0, NULL, NULL,
1368 		       CTLFLAG_PERMANENT,
1369 		       CTLTYPE_NODE, "udp",
1370 		       SYSCTL_DESCR("UDPv4 related settings"),
1371 		       NULL, 0, NULL, 0,
1372 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1373 
1374 	sysctl_createv(clog, 0, NULL, NULL,
1375 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1376 		       CTLTYPE_INT, "checksum",
1377 		       SYSCTL_DESCR("Compute UDP checksums"),
1378 		       NULL, 0, &udpcksum, 0,
1379 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1380 		       CTL_EOL);
1381 	sysctl_createv(clog, 0, NULL, NULL,
1382 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1383 		       CTLTYPE_INT, "sendspace",
1384 		       SYSCTL_DESCR("Default UDP send buffer size"),
1385 		       NULL, 0, &udp_sendspace, 0,
1386 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1387 		       CTL_EOL);
1388 	sysctl_createv(clog, 0, NULL, NULL,
1389 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1390 		       CTLTYPE_INT, "recvspace",
1391 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1392 		       NULL, 0, &udp_recvspace, 0,
1393 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1394 		       CTL_EOL);
1395 	sysctl_createv(clog, 0, NULL, NULL,
1396 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1397 		       CTLTYPE_INT, "do_loopback_cksum",
1398 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1399 		       NULL, 0, &udp_do_loopback_cksum, 0,
1400 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1401 		       CTL_EOL);
1402 	sysctl_createv(clog, 0, NULL, NULL,
1403 		       CTLFLAG_PERMANENT,
1404 		       CTLTYPE_STRUCT, "pcblist",
1405 		       SYSCTL_DESCR("UDP protocol control block list"),
1406 		       sysctl_inpcblist, 0, &udbtable, 0,
1407 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1408 		       CTL_EOL);
1409 	sysctl_createv(clog, 0, NULL, NULL,
1410 		       CTLFLAG_PERMANENT,
1411 		       CTLTYPE_STRUCT, "stats",
1412 		       SYSCTL_DESCR("UDP statistics"),
1413 		       NULL, 0, &udpstat, sizeof(udpstat),
1414 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1415 		       CTL_EOL);
1416 }
1417 #endif
1418 
1419 #if (defined INET && defined IPSEC_NAT_T)
1420 /*
1421  * Returns:
1422  * 1 if the packet was processed
1423  * 0 if normal UDP processing should take place
1424  * -1 if an error occurent and m was freed
1425  */
1426 static int
1427 udp4_espinudp(mp, off, src, so)
1428 	struct mbuf **mp;
1429 	int off;
1430 	struct sockaddr *src;
1431 	struct socket *so;
1432 {
1433 	size_t len;
1434 	caddr_t data;
1435 	struct inpcb *inp;
1436 	size_t skip = 0;
1437 	size_t minlen;
1438 	size_t iphdrlen;
1439 	struct ip *ip;
1440 	struct mbuf *n;
1441 	struct m_tag *tag;
1442 	struct udphdr *udphdr;
1443 	u_int16_t sport, dport;
1444 	struct mbuf *m = *mp;
1445 
1446 	/*
1447 	 * Collapse the mbuf chain if the first mbuf is too short
1448 	 * The longest case is: UDP + non ESP marker + ESP
1449 	 */
1450 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1451 	if (minlen > m->m_pkthdr.len)
1452 		minlen = m->m_pkthdr.len;
1453 
1454 	if (m->m_len < minlen) {
1455 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1456 			printf("udp4_espinudp: m_pullup failed\n");
1457 			return -1;
1458 		}
1459 		m = *mp;
1460 	}
1461 
1462 	len = m->m_len - off;
1463 	data = mtod(m, caddr_t) + off;
1464 	inp = sotoinpcb(so);
1465 
1466 	/* Ignore keepalive packets */
1467 	if ((len == 1) && (data[0] == '\xff')) {
1468 		return 1;
1469 	}
1470 
1471 	/*
1472 	 * Check that the payload is long enough to hold
1473 	 * an ESP header and compute the length of encapsulation
1474 	 * header to remove
1475 	 */
1476 	if (inp->inp_flags & INP_ESPINUDP) {
1477 		u_int32_t *st = (u_int32_t *)data;
1478 
1479 		if ((len <= sizeof(struct esp)) || (*st == 0))
1480 			return 0; /* Normal UDP processing */
1481 
1482 		skip = sizeof(struct udphdr);
1483 	}
1484 
1485 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1486 		u_int32_t *st = (u_int32_t *)data;
1487 
1488 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1489 		    || ((st[0] | st[1]) != 0))
1490 			return 0; /* Normal UDP processing */
1491 
1492 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1493 	}
1494 
1495 	/*
1496 	 * Get the UDP ports. They are handled in network
1497 	 * order everywhere in IPSEC_NAT_T code.
1498 	 */
1499 	udphdr = (struct udphdr *)(data - skip);
1500 	sport = udphdr->uh_sport;
1501 	dport = udphdr->uh_dport;
1502 
1503 	/*
1504 	 * Remove the UDP header (and possibly the non ESP marker)
1505 	 * IP header lendth is iphdrlen
1506 	 * Before:
1507 	 *   <--- off --->
1508 	 *   +----+------+-----+
1509 	 *   | IP |  UDP | ESP |
1510 	 *   +----+------+-----+
1511 	 *        <-skip->
1512 	 * After:
1513 	 *          +----+-----+
1514 	 *          | IP | ESP |
1515 	 *          +----+-----+
1516 	 *   <-skip->
1517 	 */
1518 	iphdrlen = off - sizeof(struct udphdr);
1519 	memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1520 	m_adj(m, skip);
1521 
1522 	ip = mtod(m, struct ip *);
1523 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1524 	ip->ip_p = IPPROTO_ESP;
1525 
1526 	/*
1527 	 * Copy the mbuf to avoid multiple free, as both
1528 	 * esp4_input (which we call) and udp_input (which
1529 	 * called us) free the mbuf.
1530 	 */
1531 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1532 		printf("udp4_espinudp: m_dup failed\n");
1533 		return 0;
1534 	}
1535 
1536 	/*
1537 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1538 	 * the source UDP port. This is required if we want
1539 	 * to select the right SPD for multiple hosts behind
1540 	 * same NAT
1541 	 */
1542 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1543 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1544 		printf("udp4_espinudp: m_tag_get failed\n");
1545 		m_freem(n);
1546 		return 0;
1547 	}
1548 	((u_int16_t *)(tag + 1))[0] = sport;
1549 	((u_int16_t *)(tag + 1))[1] = dport;
1550 	m_tag_prepend(n, tag);
1551 
1552 	esp4_input(n, iphdrlen);
1553 
1554 	/* We handled it, it shoudln't be handled by UDP */
1555 	return 1;
1556 }
1557 #endif
1558