xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 9c1da17e908379b8a470f1117a6395bd6a0ca559)
1 /*	$NetBSD: udp_usrreq.c,v 1.141 2005/08/10 13:06:49 yamt 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 1.141 2005/08/10 13:06:49 yamt 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 #endif
110 
111 #ifndef INET6
112 /* always need ip6.h for IP6_EXTHDR_GET */
113 #include <netinet/ip6.h>
114 #endif
115 
116 #include "faith.h"
117 #if defined(NFAITH) && NFAITH > 0
118 #include <net/if_faith.h>
119 #endif
120 
121 #include <machine/stdarg.h>
122 
123 #ifdef FAST_IPSEC
124 #include <netipsec/ipsec.h>
125 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
126 #ifdef INET6
127 #include <netipsec/ipsec6.h>
128 #endif
129 #endif	/* FAST_IPSEC*/
130 
131 #ifdef IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netkey/key.h>
134 #endif /*IPSEC*/
135 
136 #ifdef IPKDB
137 #include <ipkdb/ipkdb.h>
138 #endif
139 
140 /*
141  * UDP protocol implementation.
142  * Per RFC 768, August, 1980.
143  */
144 #ifndef	COMPAT_42
145 int	udpcksum = 1;
146 #else
147 int	udpcksum = 0;		/* XXX */
148 #endif
149 int	udp_do_loopback_cksum = 0;
150 
151 struct	inpcbtable udbtable;
152 struct	udpstat udpstat;
153 
154 #ifdef INET
155 #ifdef IPSEC_NAT_T
156 static int udp4_espinudp (struct mbuf *, int, struct sockaddr *,
157 	struct socket *);
158 #endif
159 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
160 	struct socket *);
161 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
162 	struct mbuf *, int);
163 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
164 #endif
165 #ifdef INET6
166 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
167 	struct socket *);
168 static int udp6_realinput (int, struct sockaddr_in6 *,
169 	struct sockaddr_in6 *, struct mbuf *, int);
170 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
171 #endif
172 #ifdef INET
173 static	void udp_notify (struct inpcb *, int);
174 #endif
175 
176 #ifndef UDBHASHSIZE
177 #define	UDBHASHSIZE	128
178 #endif
179 int	udbhashsize = UDBHASHSIZE;
180 
181 #ifdef MBUFTRACE
182 struct mowner udp_mowner = { "udp" };
183 struct mowner udp_rx_mowner = { "udp", "rx" };
184 struct mowner udp_tx_mowner = { "udp", "tx" };
185 #endif
186 
187 #ifdef UDP_CSUM_COUNTERS
188 #include <sys/device.h>
189 
190 #if defined(INET)
191 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192     NULL, "udp", "hwcsum bad");
193 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
194     NULL, "udp", "hwcsum ok");
195 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
196     NULL, "udp", "hwcsum data");
197 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
198     NULL, "udp", "swcsum");
199 
200 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
201 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
202 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
203 EVCNT_ATTACH_STATIC(udp_swcsum);
204 #endif /* defined(INET) */
205 
206 #if defined(INET6)
207 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208     NULL, "udp6", "hwcsum bad");
209 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
210     NULL, "udp6", "hwcsum ok");
211 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
212     NULL, "udp6", "hwcsum data");
213 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
214     NULL, "udp6", "swcsum");
215 
216 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
217 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
218 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
219 EVCNT_ATTACH_STATIC(udp6_swcsum);
220 #endif /* defined(INET6) */
221 
222 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
223 
224 #else
225 
226 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
227 
228 #endif /* UDP_CSUM_COUNTERS */
229 
230 void
231 udp_init(void)
232 {
233 
234 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
235 
236 	MOWNER_ATTACH(&udp_tx_mowner);
237 	MOWNER_ATTACH(&udp_rx_mowner);
238 	MOWNER_ATTACH(&udp_mowner);
239 }
240 
241 /*
242  * Checksum extended UDP header and data.
243  */
244 
245 int
246 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
247     int iphlen, int len)
248 {
249 
250 	switch (af) {
251 #ifdef INET
252 	case AF_INET:
253 		return udp4_input_checksum(m, uh, iphlen, len);
254 #endif
255 #ifdef INET6
256 	case AF_INET6:
257 		return udp6_input_checksum(m, uh, iphlen, len);
258 #endif
259 	}
260 #ifdef DIAGNOSTIC
261 	panic("udp_input_checksum: unknown af %d", af);
262 #endif
263 	/* NOTREACHED */
264 	return -1;
265 }
266 
267 #ifdef INET
268 
269 /*
270  * Checksum extended UDP header and data.
271  */
272 
273 static int
274 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
275     int iphlen, int len)
276 {
277 
278 	/*
279 	 * XXX it's better to record and check if this mbuf is
280 	 * already checked.
281 	 */
282 
283 	if (uh->uh_sum == 0)
284 		return 0;
285 
286 	switch (m->m_pkthdr.csum_flags &
287 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
288 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
289 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
290 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
291 		goto badcsum;
292 
293 	case M_CSUM_UDPv4|M_CSUM_DATA: {
294 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
295 
296 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
297 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
298 			const struct ip *ip =
299 			    mtod(m, const struct ip *);
300 
301 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
302 			    ip->ip_dst.s_addr,
303 			    htons(hw_csum + len + IPPROTO_UDP));
304 		}
305 		if ((hw_csum ^ 0xffff) != 0)
306 			goto badcsum;
307 		break;
308 	}
309 
310 	case M_CSUM_UDPv4:
311 		/* Checksum was okay. */
312 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
313 		break;
314 
315 	default:
316 		/*
317 		 * Need to compute it ourselves.  Maybe skip checksum
318 		 * on loopback interfaces.
319 		 */
320 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
321 				     IFF_LOOPBACK) ||
322 				   udp_do_loopback_cksum)) {
323 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
324 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
325 				goto badcsum;
326 		}
327 		break;
328 	}
329 
330 	return 0;
331 
332 badcsum:
333 	udpstat.udps_badsum++;
334 	return -1;
335 }
336 
337 void
338 udp_input(struct mbuf *m, ...)
339 {
340 	va_list ap;
341 	struct sockaddr_in src, dst;
342 	struct ip *ip;
343 	struct udphdr *uh;
344 	int iphlen;
345 	int len;
346 	int n;
347 	u_int16_t ip_len;
348 
349 	va_start(ap, m);
350 	iphlen = va_arg(ap, int);
351 	(void)va_arg(ap, int);		/* ignore value, advance ap */
352 	va_end(ap);
353 
354 	MCLAIM(m, &udp_rx_mowner);
355 	udpstat.udps_ipackets++;
356 
357 	/*
358 	 * Get IP and UDP header together in first mbuf.
359 	 */
360 	ip = mtod(m, struct ip *);
361 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
362 	if (uh == NULL) {
363 		udpstat.udps_hdrops++;
364 		return;
365 	}
366 	KASSERT(UDP_HDR_ALIGNED_P(uh));
367 
368 	/* destination port of 0 is illegal, based on RFC768. */
369 	if (uh->uh_dport == 0)
370 		goto bad;
371 
372 	/*
373 	 * Make mbuf data length reflect UDP length.
374 	 * If not enough data to reflect UDP length, drop.
375 	 */
376 	ip_len = ntohs(ip->ip_len);
377 	len = ntohs((u_int16_t)uh->uh_ulen);
378 	if (ip_len != iphlen + len) {
379 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
380 			udpstat.udps_badlen++;
381 			goto bad;
382 		}
383 		m_adj(m, iphlen + len - ip_len);
384 	}
385 
386 	/*
387 	 * Checksum extended UDP header and data.
388 	 */
389 	if (udp4_input_checksum(m, uh, iphlen, len))
390 		goto badcsum;
391 
392 	/* construct source and dst sockaddrs. */
393 	bzero(&src, sizeof(src));
394 	src.sin_family = AF_INET;
395 	src.sin_len = sizeof(struct sockaddr_in);
396 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
397 	src.sin_port = uh->uh_sport;
398 	bzero(&dst, sizeof(dst));
399 	dst.sin_family = AF_INET;
400 	dst.sin_len = sizeof(struct sockaddr_in);
401 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
402 	dst.sin_port = uh->uh_dport;
403 
404 	n = udp4_realinput(&src, &dst, m, iphlen);
405 #ifdef INET6
406 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
407 		struct sockaddr_in6 src6, dst6;
408 
409 		bzero(&src6, sizeof(src6));
410 		src6.sin6_family = AF_INET6;
411 		src6.sin6_len = sizeof(struct sockaddr_in6);
412 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
413 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
414 			sizeof(ip->ip_src));
415 		src6.sin6_port = uh->uh_sport;
416 		bzero(&dst6, sizeof(dst6));
417 		dst6.sin6_family = AF_INET6;
418 		dst6.sin6_len = sizeof(struct sockaddr_in6);
419 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
420 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
421 			sizeof(ip->ip_dst));
422 		dst6.sin6_port = uh->uh_dport;
423 
424 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
425 	}
426 #endif
427 
428 	if (n == 0) {
429 		if (m->m_flags & (M_BCAST | M_MCAST)) {
430 			udpstat.udps_noportbcast++;
431 			goto bad;
432 		}
433 		udpstat.udps_noport++;
434 #ifdef IPKDB
435 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
436 				m, iphlen + sizeof(struct udphdr),
437 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
438 			/*
439 			 * It was a debugger connect packet,
440 			 * just drop it now
441 			 */
442 			goto bad;
443 		}
444 #endif
445 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
446 		m = NULL;
447 	}
448 
449 bad:
450 	if (m)
451 		m_freem(m);
452 	return;
453 
454 badcsum:
455 	m_freem(m);
456 }
457 #endif
458 
459 #ifdef INET6
460 static int
461 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
462 {
463 
464 	/*
465 	 * XXX it's better to record and check if this mbuf is
466 	 * already checked.
467 	 */
468 
469 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
470 		goto good;
471 	}
472 	if (uh->uh_sum == 0) {
473 		udp6stat.udp6s_nosum++;
474 		goto bad;
475 	}
476 
477 	switch (m->m_pkthdr.csum_flags &
478 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
479 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
480 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
481 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
482 		udp6stat.udp6s_badsum++;
483 		goto bad;
484 
485 #if 0 /* notyet */
486 	case M_CSUM_UDPv6|M_CSUM_DATA:
487 #endif
488 
489 	case M_CSUM_UDPv6:
490 		/* Checksum was okay. */
491 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
492 		break;
493 
494 	default:
495 		/*
496 		 * Need to compute it ourselves.  Maybe skip checksum
497 		 * on loopback interfaces.
498 		 */
499 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
500 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
501 			udp6stat.udp6s_badsum++;
502 			goto bad;
503 		}
504 	}
505 
506 good:
507 	return 0;
508 bad:
509 	return -1;
510 }
511 
512 int
513 udp6_input(struct mbuf **mp, int *offp, int proto)
514 {
515 	struct mbuf *m = *mp;
516 	int off = *offp;
517 	struct sockaddr_in6 src, dst;
518 	struct ip6_hdr *ip6;
519 	struct udphdr *uh;
520 	u_int32_t plen, ulen;
521 
522 	ip6 = mtod(m, struct ip6_hdr *);
523 
524 #if defined(NFAITH) && 0 < NFAITH
525 	if (faithprefix(&ip6->ip6_dst)) {
526 		/* send icmp6 host unreach? */
527 		m_freem(m);
528 		return IPPROTO_DONE;
529 	}
530 #endif
531 
532 	udp6stat.udp6s_ipackets++;
533 
534 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
535 	plen = m->m_pkthdr.len - off;
536 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
537 	if (uh == NULL) {
538 		ip6stat.ip6s_tooshort++;
539 		return IPPROTO_DONE;
540 	}
541 	KASSERT(UDP_HDR_ALIGNED_P(uh));
542 	ulen = ntohs((u_short)uh->uh_ulen);
543 	/*
544 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
545 	 * iff payload length > 0xffff.
546 	 */
547 	if (ulen == 0 && plen > 0xffff)
548 		ulen = plen;
549 
550 	if (plen != ulen) {
551 		udp6stat.udp6s_badlen++;
552 		goto bad;
553 	}
554 
555 	/* destination port of 0 is illegal, based on RFC768. */
556 	if (uh->uh_dport == 0)
557 		goto bad;
558 
559 	/* Be proactive about malicious use of IPv4 mapped address */
560 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
561 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
562 		/* XXX stat */
563 		goto bad;
564 	}
565 
566 	/*
567 	 * Checksum extended UDP header and data.  Maybe skip checksum
568 	 * on loopback interfaces.
569 	 */
570 	if (udp6_input_checksum(m, uh, off, ulen))
571 		goto bad;
572 
573 	/*
574 	 * Construct source and dst sockaddrs.
575 	 * Note that ifindex (s6_addr16[1]) is already filled.
576 	 */
577 	bzero(&src, sizeof(src));
578 	src.sin6_family = AF_INET6;
579 	src.sin6_len = sizeof(struct sockaddr_in6);
580 	/* KAME hack: recover scopeid */
581 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
582 	src.sin6_port = uh->uh_sport;
583 	bzero(&dst, sizeof(dst));
584 	dst.sin6_family = AF_INET6;
585 	dst.sin6_len = sizeof(struct sockaddr_in6);
586 	/* KAME hack: recover scopeid */
587 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
588 	dst.sin6_port = uh->uh_dport;
589 
590 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
591 		if (m->m_flags & M_MCAST) {
592 			udp6stat.udp6s_noportmcast++;
593 			goto bad;
594 		}
595 		udp6stat.udp6s_noport++;
596 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
597 		m = NULL;
598 	}
599 
600 bad:
601 	if (m)
602 		m_freem(m);
603 	return IPPROTO_DONE;
604 }
605 #endif
606 
607 #ifdef INET
608 static void
609 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
610 	struct sockaddr *src, struct socket *so)
611 {
612 	struct mbuf *opts = NULL;
613 	struct mbuf *n;
614 	struct inpcb *inp = NULL;
615 
616 	if (!so)
617 		return;
618 	switch (so->so_proto->pr_domain->dom_family) {
619 	case AF_INET:
620 		inp = sotoinpcb(so);
621 		break;
622 #ifdef INET6
623 	case AF_INET6:
624 		break;
625 #endif
626 	default:
627 		return;
628 	}
629 
630 #if defined(IPSEC) || defined(FAST_IPSEC)
631 	/* check AH/ESP integrity. */
632 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
633 		ipsecstat.in_polvio++;
634 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
635 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
636 			    0, 0);
637 		return;
638 	}
639 #endif /*IPSEC*/
640 
641 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
642 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
643 			 || so->so_options & SO_TIMESTAMP)) {
644 			struct ip *ip = mtod(n, struct ip *);
645 			ip_savecontrol(inp, &opts, ip, n);
646 		}
647 
648 		m_adj(n, off);
649 		if (sbappendaddr(&so->so_rcv, src, n,
650 				opts) == 0) {
651 			m_freem(n);
652 			if (opts)
653 				m_freem(opts);
654 			so->so_rcv.sb_overflowed++;
655 			udpstat.udps_fullsock++;
656 		} else
657 			sorwakeup(so);
658 	}
659 }
660 #endif
661 
662 #ifdef INET6
663 static void
664 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
665 	struct sockaddr *src, struct socket *so)
666 {
667 	struct mbuf *opts = NULL;
668 	struct mbuf *n;
669 	struct in6pcb *in6p = NULL;
670 
671 	if (!so)
672 		return;
673 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
674 		return;
675 	in6p = sotoin6pcb(so);
676 
677 #if defined(IPSEC) || defined(FAST_IPSEC)
678 	/* check AH/ESP integrity. */
679 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
680 		ipsec6stat.in_polvio++;
681 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
682 			icmp6_error(n, ICMP6_DST_UNREACH,
683 			    ICMP6_DST_UNREACH_ADMIN, 0);
684 		return;
685 	}
686 #endif /*IPSEC*/
687 
688 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
689 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
690 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
691 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
692 			ip6_savecontrol(in6p, &opts, ip6, n);
693 		}
694 
695 		m_adj(n, off);
696 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
697 			m_freem(n);
698 			if (opts)
699 				m_freem(opts);
700 			so->so_rcv.sb_overflowed++;
701 			udp6stat.udp6s_fullsock++;
702 		} else
703 			sorwakeup(so);
704 	}
705 }
706 #endif
707 
708 #ifdef INET
709 static int
710 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
711 	struct mbuf *m, int off /* offset of udphdr */)
712 {
713 	u_int16_t *sport, *dport;
714 	int rcvcnt;
715 	struct in_addr *src4, *dst4;
716 	struct inpcb_hdr *inph;
717 	struct inpcb *inp;
718 
719 	rcvcnt = 0;
720 	off += sizeof(struct udphdr);	/* now, offset of payload */
721 
722 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
723 		goto bad;
724 
725 	src4 = &src->sin_addr;
726 	sport = &src->sin_port;
727 	dst4 = &dst->sin_addr;
728 	dport = &dst->sin_port;
729 
730 	if (IN_MULTICAST(dst4->s_addr) ||
731 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
732 		/*
733 		 * Deliver a multicast or broadcast datagram to *all* sockets
734 		 * for which the local and remote addresses and ports match
735 		 * those of the incoming datagram.  This allows more than
736 		 * one process to receive multi/broadcasts on the same port.
737 		 * (This really ought to be done for unicast datagrams as
738 		 * well, but that would cause problems with existing
739 		 * applications that open both address-specific sockets and
740 		 * a wildcard socket listening to the same port -- they would
741 		 * end up receiving duplicates of every unicast datagram.
742 		 * Those applications open the multiple sockets to overcome an
743 		 * inadequacy of the UDP socket interface, but for backwards
744 		 * compatibility we avoid the problem here rather than
745 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
746 		 */
747 
748 		/*
749 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
750 		 * we need udpiphdr for IPsec processing so we do that later.
751 		 */
752 		/*
753 		 * Locate pcb(s) for datagram.
754 		 */
755 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
756 			inp = (struct inpcb *)inph;
757 			if (inp->inp_af != AF_INET)
758 				continue;
759 
760 			if (inp->inp_lport != *dport)
761 				continue;
762 			if (!in_nullhost(inp->inp_laddr)) {
763 				if (!in_hosteq(inp->inp_laddr, *dst4))
764 					continue;
765 			}
766 			if (!in_nullhost(inp->inp_faddr)) {
767 				if (!in_hosteq(inp->inp_faddr, *src4) ||
768 				    inp->inp_fport != *sport)
769 					continue;
770 			}
771 
772 			udp4_sendup(m, off, (struct sockaddr *)src,
773 				inp->inp_socket);
774 			rcvcnt++;
775 
776 			/*
777 			 * Don't look for additional matches if this one does
778 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
779 			 * socket options set.  This heuristic avoids searching
780 			 * through all pcbs in the common case of a non-shared
781 			 * port.  It assumes that an application will never
782 			 * clear these options after setting them.
783 			 */
784 			if ((inp->inp_socket->so_options &
785 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
786 				break;
787 		}
788 	} else {
789 		/*
790 		 * Locate pcb for datagram.
791 		 */
792 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
793 		if (inp == 0) {
794 			++udpstat.udps_pcbhashmiss;
795 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
796 			if (inp == 0)
797 				return rcvcnt;
798 		}
799 
800 #ifdef IPSEC_NAT_T
801 		/* Handle ESP over UDP */
802 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
803 			struct sockaddr *sa = (struct sockaddr *)src;
804 
805 			if (udp4_espinudp(m, off, sa, inp->inp_socket) != 0) {
806 				rcvcnt++;
807 				goto bad;
808 			}
809 
810 			/* Normal UDP processing will take place */
811 		}
812 #endif
813 
814 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
815 		rcvcnt++;
816 	}
817 
818 bad:
819 	return rcvcnt;
820 }
821 #endif
822 
823 #ifdef INET6
824 static int
825 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
826 	struct mbuf *m, int off)
827 {
828 	u_int16_t sport, dport;
829 	int rcvcnt;
830 	struct in6_addr src6, dst6;
831 	const struct in_addr *dst4;
832 	struct inpcb_hdr *inph;
833 	struct in6pcb *in6p;
834 
835 	rcvcnt = 0;
836 	off += sizeof(struct udphdr);	/* now, offset of payload */
837 
838 	if (af != AF_INET && af != AF_INET6)
839 		goto bad;
840 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
841 		goto bad;
842 
843 	in6_embedscope(&src6, src, NULL, NULL);
844 	sport = src->sin6_port;
845 	in6_embedscope(&dst6, dst, NULL, NULL);
846 	dport = dst->sin6_port;
847 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
848 
849 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
850 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
851 		/*
852 		 * Deliver a multicast or broadcast datagram to *all* sockets
853 		 * for which the local and remote addresses and ports match
854 		 * those of the incoming datagram.  This allows more than
855 		 * one process to receive multi/broadcasts on the same port.
856 		 * (This really ought to be done for unicast datagrams as
857 		 * well, but that would cause problems with existing
858 		 * applications that open both address-specific sockets and
859 		 * a wildcard socket listening to the same port -- they would
860 		 * end up receiving duplicates of every unicast datagram.
861 		 * Those applications open the multiple sockets to overcome an
862 		 * inadequacy of the UDP socket interface, but for backwards
863 		 * compatibility we avoid the problem here rather than
864 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
865 		 */
866 
867 		/*
868 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
869 		 * we need udpiphdr for IPsec processing so we do that later.
870 		 */
871 		/*
872 		 * Locate pcb(s) for datagram.
873 		 */
874 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
875 			in6p = (struct in6pcb *)inph;
876 			if (in6p->in6p_af != AF_INET6)
877 				continue;
878 
879 			if (in6p->in6p_lport != dport)
880 				continue;
881 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
882 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
883 					continue;
884 			} else {
885 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
886 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
887 					continue;
888 			}
889 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
890 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
891 				    &src6) || in6p->in6p_fport != sport)
892 					continue;
893 			} else {
894 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
895 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
896 					continue;
897 			}
898 
899 			udp6_sendup(m, off, (struct sockaddr *)src,
900 				in6p->in6p_socket);
901 			rcvcnt++;
902 
903 			/*
904 			 * Don't look for additional matches if this one does
905 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
906 			 * socket options set.  This heuristic avoids searching
907 			 * through all pcbs in the common case of a non-shared
908 			 * port.  It assumes that an application will never
909 			 * clear these options after setting them.
910 			 */
911 			if ((in6p->in6p_socket->so_options &
912 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
913 				break;
914 		}
915 	} else {
916 		/*
917 		 * Locate pcb for datagram.
918 		 */
919 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
920 		    &dst6, dport, 0);
921 		if (in6p == 0) {
922 			++udpstat.udps_pcbhashmiss;
923 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
924 			if (in6p == 0)
925 				return rcvcnt;
926 		}
927 
928 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
929 		rcvcnt++;
930 	}
931 
932 bad:
933 	return rcvcnt;
934 }
935 #endif
936 
937 #ifdef INET
938 /*
939  * Notify a udp user of an asynchronous error;
940  * just wake up so that he can collect error status.
941  */
942 static void
943 udp_notify(struct inpcb *inp, int errno)
944 {
945 	inp->inp_socket->so_error = errno;
946 	sorwakeup(inp->inp_socket);
947 	sowwakeup(inp->inp_socket);
948 }
949 
950 void *
951 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
952 {
953 	struct ip *ip = v;
954 	struct udphdr *uh;
955 	void (*notify)(struct inpcb *, int) = udp_notify;
956 	int errno;
957 
958 	if (sa->sa_family != AF_INET
959 	 || sa->sa_len != sizeof(struct sockaddr_in))
960 		return NULL;
961 	if ((unsigned)cmd >= PRC_NCMDS)
962 		return NULL;
963 	errno = inetctlerrmap[cmd];
964 	if (PRC_IS_REDIRECT(cmd))
965 		notify = in_rtchange, ip = 0;
966 	else if (cmd == PRC_HOSTDEAD)
967 		ip = 0;
968 	else if (errno == 0)
969 		return NULL;
970 	if (ip) {
971 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
972 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
973 		    ip->ip_src, uh->uh_sport, errno, notify);
974 
975 		/* XXX mapped address case */
976 	} else
977 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
978 		    notify);
979 	return NULL;
980 }
981 
982 int
983 udp_ctloutput(op, so, level, optname, mp)
984 	int op;
985 	struct socket *so;
986 	int level, optname;
987 	struct mbuf **mp;
988 {
989 	int s;
990 	int error = 0;
991 	struct mbuf *m;
992 	struct inpcb *inp;
993 	int family;
994 
995 	family = so->so_proto->pr_domain->dom_family;
996 
997 	s = splsoftnet();
998 	switch (family) {
999 #ifdef INET
1000 	case PF_INET:
1001 		if (level != IPPROTO_UDP) {
1002 			error = ip_ctloutput(op, so, level, optname, mp);
1003 			goto end;
1004 		}
1005 		break;
1006 #endif
1007 #ifdef INET6
1008 	case PF_INET6:
1009 		if (level != IPPROTO_UDP) {
1010 			error = ip6_ctloutput(op, so, level, optname, mp);
1011 			goto end;
1012 		}
1013 		break;
1014 #endif
1015 	default:
1016 		error = EAFNOSUPPORT;
1017 		goto end;
1018 		break;
1019 	}
1020 
1021 
1022 	switch (op) {
1023 	case PRCO_SETOPT:
1024 		m = *mp;
1025 		inp = sotoinpcb(so);
1026 
1027 		switch (optname) {
1028 		case UDP_ENCAP:
1029 			if (m == NULL || m->m_len < sizeof (int)) {
1030 				error = EINVAL;
1031 				goto end;
1032 			}
1033 
1034 			switch(*mtod(m, int *)) {
1035 #ifdef IPSEC_NAT_T
1036 			case 0:
1037 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1038 				break;
1039 
1040 			case UDP_ENCAP_ESPINUDP:
1041 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1042 				inp->inp_flags |= INP_ESPINUDP;
1043 				break;
1044 
1045 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1046 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1047 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1048 				break;
1049 #endif
1050 			default:
1051 				error = EINVAL;
1052 				goto end;
1053 				break;
1054 			}
1055 			break;
1056 
1057 		default:
1058 			error = ENOPROTOOPT;
1059 			goto end;
1060 			break;
1061 		}
1062 		break;
1063 
1064 	default:
1065 		error = EINVAL;
1066 		goto end;
1067 		break;
1068 	}
1069 
1070 end:
1071 	splx(s);
1072 	return error;
1073 }
1074 
1075 
1076 int
1077 udp_output(struct mbuf *m, ...)
1078 {
1079 	struct inpcb *inp;
1080 	struct udpiphdr *ui;
1081 	struct route *ro;
1082 	int len = m->m_pkthdr.len;
1083 	int error = 0;
1084 	va_list ap;
1085 
1086 	MCLAIM(m, &udp_tx_mowner);
1087 	va_start(ap, m);
1088 	inp = va_arg(ap, struct inpcb *);
1089 	va_end(ap);
1090 
1091 	/*
1092 	 * Calculate data length and get a mbuf
1093 	 * for UDP and IP headers.
1094 	 */
1095 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1096 	if (m == 0) {
1097 		error = ENOBUFS;
1098 		goto release;
1099 	}
1100 
1101 	/*
1102 	 * Compute the packet length of the IP header, and
1103 	 * punt if the length looks bogus.
1104 	 */
1105 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1106 		error = EMSGSIZE;
1107 		goto release;
1108 	}
1109 
1110 	/*
1111 	 * Fill in mbuf with extended UDP header
1112 	 * and addresses and length put into network format.
1113 	 */
1114 	ui = mtod(m, struct udpiphdr *);
1115 	ui->ui_pr = IPPROTO_UDP;
1116 	ui->ui_src = inp->inp_laddr;
1117 	ui->ui_dst = inp->inp_faddr;
1118 	ui->ui_sport = inp->inp_lport;
1119 	ui->ui_dport = inp->inp_fport;
1120 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1121 
1122 	ro = &inp->inp_route;
1123 
1124 	/*
1125 	 * Set up checksum and output datagram.
1126 	 */
1127 	if (udpcksum) {
1128 		/*
1129 		 * XXX Cache pseudo-header checksum part for
1130 		 * XXX "connected" UDP sockets.
1131 		 */
1132 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1133 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1134 		    sizeof(struct udphdr) + IPPROTO_UDP));
1135 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1136 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1137 	} else
1138 		ui->ui_sum = 0;
1139 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1140 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1141 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1142 	udpstat.udps_opackets++;
1143 
1144 	return (ip_output(m, inp->inp_options, ro,
1145 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1146 	    inp->inp_moptions, inp->inp_socket));
1147 
1148 release:
1149 	m_freem(m);
1150 	return (error);
1151 }
1152 
1153 int	udp_sendspace = 9216;		/* really max datagram size */
1154 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1155 					/* 40 1K datagrams */
1156 
1157 /*ARGSUSED*/
1158 int
1159 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1160 	struct mbuf *control, struct proc *p)
1161 {
1162 	struct inpcb *inp;
1163 	int s;
1164 	int error = 0;
1165 
1166 	if (req == PRU_CONTROL)
1167 		return (in_control(so, (long)m, (caddr_t)nam,
1168 		    (struct ifnet *)control, p));
1169 
1170 	if (req == PRU_PURGEIF) {
1171 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1172 		in_purgeif((struct ifnet *)control);
1173 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1174 		return (0);
1175 	}
1176 
1177 	s = splsoftnet();
1178 	inp = sotoinpcb(so);
1179 #ifdef DIAGNOSTIC
1180 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1181 		panic("udp_usrreq: unexpected control mbuf");
1182 #endif
1183 	if (inp == 0 && req != PRU_ATTACH) {
1184 		error = EINVAL;
1185 		goto release;
1186 	}
1187 
1188 	/*
1189 	 * Note: need to block udp_input while changing
1190 	 * the udp pcb queue and/or pcb addresses.
1191 	 */
1192 	switch (req) {
1193 
1194 	case PRU_ATTACH:
1195 		if (inp != 0) {
1196 			error = EISCONN;
1197 			break;
1198 		}
1199 #ifdef MBUFTRACE
1200 		so->so_mowner = &udp_mowner;
1201 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1202 		so->so_snd.sb_mowner = &udp_tx_mowner;
1203 #endif
1204 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1205 			error = soreserve(so, udp_sendspace, udp_recvspace);
1206 			if (error)
1207 				break;
1208 		}
1209 		error = in_pcballoc(so, &udbtable);
1210 		if (error)
1211 			break;
1212 		inp = sotoinpcb(so);
1213 		inp->inp_ip.ip_ttl = ip_defttl;
1214 		break;
1215 
1216 	case PRU_DETACH:
1217 		in_pcbdetach(inp);
1218 		break;
1219 
1220 	case PRU_BIND:
1221 		error = in_pcbbind(inp, nam, p);
1222 		break;
1223 
1224 	case PRU_LISTEN:
1225 		error = EOPNOTSUPP;
1226 		break;
1227 
1228 	case PRU_CONNECT:
1229 		error = in_pcbconnect(inp, nam);
1230 		if (error)
1231 			break;
1232 		soisconnected(so);
1233 		break;
1234 
1235 	case PRU_CONNECT2:
1236 		error = EOPNOTSUPP;
1237 		break;
1238 
1239 	case PRU_DISCONNECT:
1240 		/*soisdisconnected(so);*/
1241 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1242 		in_pcbdisconnect(inp);
1243 		inp->inp_laddr = zeroin_addr;		/* XXX */
1244 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1245 		break;
1246 
1247 	case PRU_SHUTDOWN:
1248 		socantsendmore(so);
1249 		break;
1250 
1251 	case PRU_RCVD:
1252 		error = EOPNOTSUPP;
1253 		break;
1254 
1255 	case PRU_SEND:
1256 		if (control && control->m_len) {
1257 			m_freem(control);
1258 			m_freem(m);
1259 			error = EINVAL;
1260 			break;
1261 		}
1262 	{
1263 		struct in_addr laddr;			/* XXX */
1264 
1265 		if (nam) {
1266 			laddr = inp->inp_laddr;		/* XXX */
1267 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1268 				error = EISCONN;
1269 				goto die;
1270 			}
1271 			error = in_pcbconnect(inp, nam);
1272 			if (error)
1273 				goto die;
1274 		} else {
1275 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1276 				error = ENOTCONN;
1277 				goto die;
1278 			}
1279 		}
1280 		error = udp_output(m, inp);
1281 		m = NULL;
1282 		if (nam) {
1283 			in_pcbdisconnect(inp);
1284 			inp->inp_laddr = laddr;		/* XXX */
1285 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1286 		}
1287 	  die:
1288 		if (m)
1289 			m_freem(m);
1290 	}
1291 		break;
1292 
1293 	case PRU_SENSE:
1294 		/*
1295 		 * stat: don't bother with a blocksize.
1296 		 */
1297 		splx(s);
1298 		return (0);
1299 
1300 	case PRU_RCVOOB:
1301 		error =  EOPNOTSUPP;
1302 		break;
1303 
1304 	case PRU_SENDOOB:
1305 		m_freem(control);
1306 		m_freem(m);
1307 		error =  EOPNOTSUPP;
1308 		break;
1309 
1310 	case PRU_SOCKADDR:
1311 		in_setsockaddr(inp, nam);
1312 		break;
1313 
1314 	case PRU_PEERADDR:
1315 		in_setpeeraddr(inp, nam);
1316 		break;
1317 
1318 	default:
1319 		panic("udp_usrreq");
1320 	}
1321 
1322 release:
1323 	splx(s);
1324 	return (error);
1325 }
1326 
1327 /*
1328  * Sysctl for udp variables.
1329  */
1330 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1331 {
1332 
1333 	sysctl_createv(clog, 0, NULL, NULL,
1334 		       CTLFLAG_PERMANENT,
1335 		       CTLTYPE_NODE, "net", NULL,
1336 		       NULL, 0, NULL, 0,
1337 		       CTL_NET, CTL_EOL);
1338 	sysctl_createv(clog, 0, NULL, NULL,
1339 		       CTLFLAG_PERMANENT,
1340 		       CTLTYPE_NODE, "inet", NULL,
1341 		       NULL, 0, NULL, 0,
1342 		       CTL_NET, PF_INET, CTL_EOL);
1343 	sysctl_createv(clog, 0, NULL, NULL,
1344 		       CTLFLAG_PERMANENT,
1345 		       CTLTYPE_NODE, "udp",
1346 		       SYSCTL_DESCR("UDPv4 related settings"),
1347 		       NULL, 0, NULL, 0,
1348 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1349 
1350 	sysctl_createv(clog, 0, NULL, NULL,
1351 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1352 		       CTLTYPE_INT, "checksum",
1353 		       SYSCTL_DESCR("Compute UDP checksums"),
1354 		       NULL, 0, &udpcksum, 0,
1355 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1356 		       CTL_EOL);
1357 	sysctl_createv(clog, 0, NULL, NULL,
1358 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1359 		       CTLTYPE_INT, "sendspace",
1360 		       SYSCTL_DESCR("Default UDP send buffer size"),
1361 		       NULL, 0, &udp_sendspace, 0,
1362 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1363 		       CTL_EOL);
1364 	sysctl_createv(clog, 0, NULL, NULL,
1365 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1366 		       CTLTYPE_INT, "recvspace",
1367 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1368 		       NULL, 0, &udp_recvspace, 0,
1369 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1370 		       CTL_EOL);
1371 	sysctl_createv(clog, 0, NULL, NULL,
1372 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1373 		       CTLTYPE_INT, "do_loopback_cksum",
1374 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1375 		       NULL, 0, &udp_do_loopback_cksum, 0,
1376 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1377 		       CTL_EOL);
1378 	sysctl_createv(clog, 0, NULL, NULL,
1379 		       CTLFLAG_PERMANENT,
1380 		       CTLTYPE_STRUCT, "pcblist",
1381 		       SYSCTL_DESCR("UDP protocol control block list"),
1382 		       sysctl_inpcblist, 0, &udbtable, 0,
1383 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1384 		       CTL_EOL);
1385 	sysctl_createv(clog, 0, NULL, NULL,
1386 		       CTLFLAG_PERMANENT,
1387 		       CTLTYPE_STRUCT, "stats",
1388 		       SYSCTL_DESCR("UDP statistics"),
1389 		       NULL, 0, &udpstat, sizeof(udpstat),
1390 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1391 		       CTL_EOL);
1392 }
1393 #endif
1394 
1395 #if (defined INET && defined IPSEC_NAT_T)
1396 /*
1397  * Returns:
1398  * 1 if the packet was processed
1399  * 0 if normal UDP processing should take place
1400  */
1401 static int
1402 udp4_espinudp(m, off, src, so)
1403 	struct mbuf *m;
1404 	int off;
1405 	struct sockaddr *src;
1406 	struct socket *so;
1407 {
1408 	size_t len;
1409 	caddr_t data;
1410 	struct inpcb *inp;
1411 	size_t skip = 0;
1412 	size_t minlen;
1413 	size_t iphdrlen;
1414 	struct ip *ip;
1415 	struct mbuf *n;
1416 	struct m_tag *tag;
1417 	struct udphdr *udphdr;
1418 	u_int16_t sport, dport;
1419 
1420 	/*
1421 	 * Collapse the mbuf chain if the first mbuf is too short
1422 	 * The longest case is: UDP + non ESP marker + ESP
1423 	 */
1424 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1425 	if (minlen > m->m_pkthdr.len)
1426 		minlen = m->m_pkthdr.len;
1427 
1428 	if (m->m_len < minlen) {
1429 		if ((m = m_pullup(m, minlen)) == NULL) {
1430 			printf("udp4_espinudp: m_pullup failed\n");
1431 			return 0;
1432 		}
1433 	}
1434 
1435 	len = m->m_len - off;
1436 	data = mtod(m, caddr_t) + off;
1437 	inp = sotoinpcb(so);
1438 
1439 	/* Ignore keepalive packets */
1440 	if ((len == 1) && (data[0] == '\xff')) {
1441 		return 1;
1442 	}
1443 
1444 	/*
1445 	 * Check that the payload is long enough to hold
1446 	 * an ESP header and compute the length of encapsulation
1447 	 * header to remove
1448 	 */
1449 	if (inp->inp_flags & INP_ESPINUDP) {
1450 		u_int32_t *st = (u_int32_t *)data;
1451 
1452 		if ((len <= sizeof(struct esp)) || (*st == 0))
1453 			return 0; /* Normal UDP processing */
1454 
1455 		skip = sizeof(struct udphdr);
1456 	}
1457 
1458 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1459 		u_int64_t *st = (u_int64_t *)data;
1460 
1461 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1462 		    || (*st != 0))
1463 			return 0; /* Normal UDP processing */
1464 
1465 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1466 	}
1467 
1468 	/*
1469 	 * Get the UDP ports. They are handled in network
1470 	 * order everywhere in IPSEC_NAT_T code.
1471 	 */
1472 	udphdr = (struct udphdr *)(data - skip);
1473 	sport = udphdr->uh_sport;
1474 	dport = udphdr->uh_dport;
1475 
1476 	/*
1477 	 * Remove the UDP header (and possibly the non ESP marker)
1478 	 * IP header lendth is iphdrlen
1479 	 * Before:
1480 	 *   <--- off --->
1481 	 *   +----+------+-----+
1482 	 *   | IP |  UDP | ESP |
1483 	 *   +----+------+-----+
1484 	 *        <-skip->
1485 	 * After:
1486 	 *          +----+-----+
1487 	 *          | IP | ESP |
1488 	 *          +----+-----+
1489 	 *   <-skip->
1490 	 */
1491 	iphdrlen = off - sizeof(struct udphdr);
1492 	memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1493 	m_adj(m, skip);
1494 
1495 	ip = mtod(m, struct ip *);
1496 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1497 	ip->ip_p = IPPROTO_ESP;
1498 
1499 	/*
1500 	 * Copy the mbuf to avoid multiple free, as both
1501 	 * esp4_input (which we call) and udp_input (which
1502 	 * called us) free the mbuf.
1503 	 */
1504 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1505 		printf("udp4_espinudp: m_dup failed\n");
1506 		return 0;
1507 	}
1508 
1509 	/*
1510 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1511 	 * the source UDP port. This is required if we want
1512 	 * to select the right SPD for multiple hosts behind
1513 	 * same NAT
1514 	 */
1515 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1516 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1517 		printf("udp4_espinudp: m_tag_get failed\n");
1518 		m_freem(n);
1519 		return 0;
1520 	}
1521 	((u_int16_t *)(tag + 1))[0] = sport;
1522 	((u_int16_t *)(tag + 1))[1] = dport;
1523 	m_tag_prepend(n, tag);
1524 
1525 	esp4_input(n, iphdrlen);
1526 
1527 	/* We handled it, it shoudln't be handled by UDP */
1528 	return 1;
1529 }
1530 #endif
1531