xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision ce099b40997c43048fb78bd578195f81d2456523)
1 /*	$NetBSD: udp_usrreq.c,v 1.171 2008/04/26 08:13:59 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 1.171 2008/04/26 08:13:59 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 #include <netinet/udp_private.h>
98 
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet/icmp6.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/ip6_private.h>
104 #include <netinet6/in6_pcb.h>
105 #include <netinet6/udp6_var.h>
106 #include <netinet6/udp6_private.h>
107 #include <netinet6/scope6_var.h>
108 #endif
109 
110 #ifndef INET6
111 /* always need ip6.h for IP6_EXTHDR_GET */
112 #include <netinet/ip6.h>
113 #endif
114 
115 #include "faith.h"
116 #if defined(NFAITH) && NFAITH > 0
117 #include <net/if_faith.h>
118 #endif
119 
120 #include <machine/stdarg.h>
121 
122 #ifdef FAST_IPSEC
123 #include <netipsec/ipsec.h>
124 #include <netipsec/ipsec_var.h>
125 #include <netipsec/ipsec_private.h>
126 #include <netipsec/esp.h>
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 <netinet6/ipsec_private.h>
135 #include <netinet6/esp.h>
136 #include <netkey/key.h>
137 #endif /* IPSEC */
138 
139 #ifdef IPKDB
140 #include <ipkdb/ipkdb.h>
141 #endif
142 
143 /*
144  * UDP protocol implementation.
145  * Per RFC 768, August, 1980.
146  */
147 int	udpcksum = 1;
148 int	udp_do_loopback_cksum = 0;
149 
150 struct	inpcbtable udbtable;
151 
152 percpu_t *udpstat_percpu;
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 = MOWNER_INIT("udp", "");
183 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
184 struct mowner udp_tx_mowner = MOWNER_INIT("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 #ifdef INET
241 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
242 #endif
243 #ifdef INET6
244 	udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS);
245 #endif
246 }
247 
248 /*
249  * Checksum extended UDP header and data.
250  */
251 
252 int
253 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
254     int iphlen, int len)
255 {
256 
257 	switch (af) {
258 #ifdef INET
259 	case AF_INET:
260 		return udp4_input_checksum(m, uh, iphlen, len);
261 #endif
262 #ifdef INET6
263 	case AF_INET6:
264 		return udp6_input_checksum(m, uh, iphlen, len);
265 #endif
266 	}
267 #ifdef DIAGNOSTIC
268 	panic("udp_input_checksum: unknown af %d", af);
269 #endif
270 	/* NOTREACHED */
271 	return -1;
272 }
273 
274 #ifdef INET
275 
276 /*
277  * Checksum extended UDP header and data.
278  */
279 
280 static int
281 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
282     int iphlen, int len)
283 {
284 
285 	/*
286 	 * XXX it's better to record and check if this mbuf is
287 	 * already checked.
288 	 */
289 
290 	if (uh->uh_sum == 0)
291 		return 0;
292 
293 	switch (m->m_pkthdr.csum_flags &
294 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
295 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
296 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
297 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
298 		goto badcsum;
299 
300 	case M_CSUM_UDPv4|M_CSUM_DATA: {
301 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
302 
303 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
304 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
305 			const struct ip *ip =
306 			    mtod(m, const struct ip *);
307 
308 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
309 			    ip->ip_dst.s_addr,
310 			    htons(hw_csum + len + IPPROTO_UDP));
311 		}
312 		if ((hw_csum ^ 0xffff) != 0)
313 			goto badcsum;
314 		break;
315 	}
316 
317 	case M_CSUM_UDPv4:
318 		/* Checksum was okay. */
319 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
320 		break;
321 
322 	default:
323 		/*
324 		 * Need to compute it ourselves.  Maybe skip checksum
325 		 * on loopback interfaces.
326 		 */
327 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
328 				     IFF_LOOPBACK) ||
329 				   udp_do_loopback_cksum)) {
330 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
331 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
332 				goto badcsum;
333 		}
334 		break;
335 	}
336 
337 	return 0;
338 
339 badcsum:
340 	UDP_STATINC(UDP_STAT_BADSUM);
341 	return -1;
342 }
343 
344 void
345 udp_input(struct mbuf *m, ...)
346 {
347 	va_list ap;
348 	struct sockaddr_in src, dst;
349 	struct ip *ip;
350 	struct udphdr *uh;
351 	int iphlen;
352 	int len;
353 	int n;
354 	u_int16_t ip_len;
355 
356 	va_start(ap, m);
357 	iphlen = va_arg(ap, int);
358 	(void)va_arg(ap, int);		/* ignore value, advance ap */
359 	va_end(ap);
360 
361 	MCLAIM(m, &udp_rx_mowner);
362 	UDP_STATINC(UDP_STAT_IPACKETS);
363 
364 	/*
365 	 * Get IP and UDP header together in first mbuf.
366 	 */
367 	ip = mtod(m, struct ip *);
368 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
369 	if (uh == NULL) {
370 		UDP_STATINC(UDP_STAT_HDROPS);
371 		return;
372 	}
373 	KASSERT(UDP_HDR_ALIGNED_P(uh));
374 
375 	/* destination port of 0 is illegal, based on RFC768. */
376 	if (uh->uh_dport == 0)
377 		goto bad;
378 
379 	/*
380 	 * Make mbuf data length reflect UDP length.
381 	 * If not enough data to reflect UDP length, drop.
382 	 */
383 	ip_len = ntohs(ip->ip_len);
384 	len = ntohs((u_int16_t)uh->uh_ulen);
385 	if (ip_len != iphlen + len) {
386 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
387 			UDP_STATINC(UDP_STAT_BADLEN);
388 			goto bad;
389 		}
390 		m_adj(m, iphlen + len - ip_len);
391 	}
392 
393 	/*
394 	 * Checksum extended UDP header and data.
395 	 */
396 	if (udp4_input_checksum(m, uh, iphlen, len))
397 		goto badcsum;
398 
399 	/* construct source and dst sockaddrs. */
400 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
401 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
402 
403 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
404 		UDP_STATINC(UDP_STAT_HDROPS);
405 		return;
406 	}
407 #ifdef INET6
408 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
409 		struct sockaddr_in6 src6, dst6;
410 
411 		bzero(&src6, sizeof(src6));
412 		src6.sin6_family = AF_INET6;
413 		src6.sin6_len = sizeof(struct sockaddr_in6);
414 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
415 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
416 			sizeof(ip->ip_src));
417 		src6.sin6_port = uh->uh_sport;
418 		bzero(&dst6, sizeof(dst6));
419 		dst6.sin6_family = AF_INET6;
420 		dst6.sin6_len = sizeof(struct sockaddr_in6);
421 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
422 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
423 			sizeof(ip->ip_dst));
424 		dst6.sin6_port = uh->uh_dport;
425 
426 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
427 	}
428 #endif
429 
430 	if (n == 0) {
431 		if (m->m_flags & (M_BCAST | M_MCAST)) {
432 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
433 			goto bad;
434 		}
435 		UDP_STATINC(UDP_STAT_NOPORT);
436 #ifdef IPKDB
437 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
438 				m, iphlen + sizeof(struct udphdr),
439 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
440 			/*
441 			 * It was a debugger connect packet,
442 			 * just drop it now
443 			 */
444 			goto bad;
445 		}
446 #endif
447 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
448 		m = NULL;
449 	}
450 
451 bad:
452 	if (m)
453 		m_freem(m);
454 	return;
455 
456 badcsum:
457 	m_freem(m);
458 }
459 #endif
460 
461 #ifdef INET6
462 static int
463 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
464 {
465 
466 	/*
467 	 * XXX it's better to record and check if this mbuf is
468 	 * already checked.
469 	 */
470 
471 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
472 		goto good;
473 	}
474 	if (uh->uh_sum == 0) {
475 		UDP6_STATINC(UDP6_STAT_NOSUM);
476 		goto bad;
477 	}
478 
479 	switch (m->m_pkthdr.csum_flags &
480 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
481 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
482 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
483 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
484 		UDP6_STATINC(UDP6_STAT_BADSUM);
485 		goto bad;
486 
487 #if 0 /* notyet */
488 	case M_CSUM_UDPv6|M_CSUM_DATA:
489 #endif
490 
491 	case M_CSUM_UDPv6:
492 		/* Checksum was okay. */
493 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
494 		break;
495 
496 	default:
497 		/*
498 		 * Need to compute it ourselves.  Maybe skip checksum
499 		 * on loopback interfaces.
500 		 */
501 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
502 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
503 			UDP6_STATINC(UDP6_STAT_BADSUM);
504 			goto bad;
505 		}
506 	}
507 
508 good:
509 	return 0;
510 bad:
511 	return -1;
512 }
513 
514 int
515 udp6_input(struct mbuf **mp, int *offp, int proto)
516 {
517 	struct mbuf *m = *mp;
518 	int off = *offp;
519 	struct sockaddr_in6 src, dst;
520 	struct ip6_hdr *ip6;
521 	struct udphdr *uh;
522 	u_int32_t plen, ulen;
523 
524 	ip6 = mtod(m, struct ip6_hdr *);
525 
526 #if defined(NFAITH) && 0 < NFAITH
527 	if (faithprefix(&ip6->ip6_dst)) {
528 		/* send icmp6 host unreach? */
529 		m_freem(m);
530 		return IPPROTO_DONE;
531 	}
532 #endif
533 
534 	UDP6_STATINC(UDP6_STAT_IPACKETS);
535 
536 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
537 	plen = m->m_pkthdr.len - off;
538 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
539 	if (uh == NULL) {
540 		IP6_STATINC(IP6_STAT_TOOSHORT);
541 		return IPPROTO_DONE;
542 	}
543 	KASSERT(UDP_HDR_ALIGNED_P(uh));
544 	ulen = ntohs((u_short)uh->uh_ulen);
545 	/*
546 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
547 	 * iff payload length > 0xffff.
548 	 */
549 	if (ulen == 0 && plen > 0xffff)
550 		ulen = plen;
551 
552 	if (plen != ulen) {
553 		UDP6_STATINC(UDP6_STAT_BADLEN);
554 		goto bad;
555 	}
556 
557 	/* destination port of 0 is illegal, based on RFC768. */
558 	if (uh->uh_dport == 0)
559 		goto bad;
560 
561 	/* Be proactive about malicious use of IPv4 mapped address */
562 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
563 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
564 		/* XXX stat */
565 		goto bad;
566 	}
567 
568 	/*
569 	 * Checksum extended UDP header and data.  Maybe skip checksum
570 	 * on loopback interfaces.
571 	 */
572 	if (udp6_input_checksum(m, uh, off, ulen))
573 		goto bad;
574 
575 	/*
576 	 * Construct source and dst sockaddrs.
577 	 */
578 	bzero(&src, sizeof(src));
579 	src.sin6_family = AF_INET6;
580 	src.sin6_len = sizeof(struct sockaddr_in6);
581 	src.sin6_addr = ip6->ip6_src;
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 	dst.sin6_addr = ip6->ip6_dst;
587 	dst.sin6_port = uh->uh_dport;
588 
589 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
590 		if (m->m_flags & M_MCAST) {
591 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
592 			goto bad;
593 		}
594 		UDP6_STATINC(UDP6_STAT_NOPORT);
595 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
596 		m = NULL;
597 	}
598 
599 bad:
600 	if (m)
601 		m_freem(m);
602 	return IPPROTO_DONE;
603 }
604 #endif
605 
606 #ifdef INET
607 static void
608 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
609 	struct sockaddr *src, struct socket *so)
610 {
611 	struct mbuf *opts = NULL;
612 	struct mbuf *n;
613 	struct inpcb *inp = NULL;
614 
615 	if (!so)
616 		return;
617 	switch (so->so_proto->pr_domain->dom_family) {
618 	case AF_INET:
619 		inp = sotoinpcb(so);
620 		break;
621 #ifdef INET6
622 	case AF_INET6:
623 		break;
624 #endif
625 	default:
626 		return;
627 	}
628 
629 #if defined(IPSEC) || defined(FAST_IPSEC)
630 	/* check AH/ESP integrity. */
631 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
632 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
633 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
634 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
635 			    0, 0);
636 		return;
637 	}
638 #endif /*IPSEC*/
639 
640 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
641 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
642 			 || so->so_options & SO_TIMESTAMP)) {
643 			struct ip *ip = mtod(n, struct ip *);
644 			ip_savecontrol(inp, &opts, ip, n);
645 		}
646 
647 		m_adj(n, off);
648 		if (sbappendaddr(&so->so_rcv, src, n,
649 				opts) == 0) {
650 			m_freem(n);
651 			if (opts)
652 				m_freem(opts);
653 			so->so_rcv.sb_overflowed++;
654 			UDP_STATINC(UDP_STAT_FULLSOCK);
655 		} else
656 			sorwakeup(so);
657 	}
658 }
659 #endif
660 
661 #ifdef INET6
662 static void
663 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
664 	struct sockaddr *src, struct socket *so)
665 {
666 	struct mbuf *opts = NULL;
667 	struct mbuf *n;
668 	struct in6pcb *in6p = NULL;
669 
670 	if (!so)
671 		return;
672 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
673 		return;
674 	in6p = sotoin6pcb(so);
675 
676 #if defined(IPSEC) || defined(FAST_IPSEC)
677 	/* check AH/ESP integrity. */
678 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
679 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
680 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
681 			icmp6_error(n, ICMP6_DST_UNREACH,
682 			    ICMP6_DST_UNREACH_ADMIN, 0);
683 		return;
684 	}
685 #endif /*IPSEC*/
686 
687 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
688 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
689 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
690 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
691 			ip6_savecontrol(in6p, &opts, ip6, n);
692 		}
693 
694 		m_adj(n, off);
695 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
696 			m_freem(n);
697 			if (opts)
698 				m_freem(opts);
699 			so->so_rcv.sb_overflowed++;
700 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
701 		} else
702 			sorwakeup(so);
703 	}
704 }
705 #endif
706 
707 #ifdef INET
708 static int
709 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
710 	struct mbuf **mp, int off /* offset of udphdr */)
711 {
712 	u_int16_t *sport, *dport;
713 	int rcvcnt;
714 	struct in_addr *src4, *dst4;
715 	struct inpcb_hdr *inph;
716 	struct inpcb *inp;
717 	struct mbuf *m = *mp;
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 			UDP_STATINC(UDP_STAT_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 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
806 			case -1: 	/* Error, m was freeed */
807 				rcvcnt = -1;
808 				goto bad;
809 				break;
810 
811 			case 1:		/* ESP over UDP */
812 				rcvcnt++;
813 				goto bad;
814 				break;
815 
816 			case 0: 	/* plain UDP */
817 			default: 	/* Unexpected */
818 				/*
819 				 * Normal UDP processing will take place
820 				 * m may have changed.
821 				 */
822 				m = *mp;
823 				break;
824 			}
825 		}
826 #endif
827 
828 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
829 		rcvcnt++;
830 	}
831 
832 bad:
833 	return rcvcnt;
834 }
835 #endif
836 
837 #ifdef INET6
838 static int
839 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
840 	struct mbuf *m, int off)
841 {
842 	u_int16_t sport, dport;
843 	int rcvcnt;
844 	struct in6_addr src6, *dst6;
845 	const struct in_addr *dst4;
846 	struct inpcb_hdr *inph;
847 	struct in6pcb *in6p;
848 
849 	rcvcnt = 0;
850 	off += sizeof(struct udphdr);	/* now, offset of payload */
851 
852 	if (af != AF_INET && af != AF_INET6)
853 		goto bad;
854 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
855 		goto bad;
856 
857 	src6 = src->sin6_addr;
858 	if (sa6_recoverscope(src) != 0) {
859 		/* XXX: should be impossible. */
860 		goto bad;
861 	}
862 	sport = src->sin6_port;
863 
864 	dport = dst->sin6_port;
865 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
866 	dst6 = &dst->sin6_addr;
867 
868 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
869 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
870 		/*
871 		 * Deliver a multicast or broadcast datagram to *all* sockets
872 		 * for which the local and remote addresses and ports match
873 		 * those of the incoming datagram.  This allows more than
874 		 * one process to receive multi/broadcasts on the same port.
875 		 * (This really ought to be done for unicast datagrams as
876 		 * well, but that would cause problems with existing
877 		 * applications that open both address-specific sockets and
878 		 * a wildcard socket listening to the same port -- they would
879 		 * end up receiving duplicates of every unicast datagram.
880 		 * Those applications open the multiple sockets to overcome an
881 		 * inadequacy of the UDP socket interface, but for backwards
882 		 * compatibility we avoid the problem here rather than
883 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
884 		 */
885 
886 		/*
887 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
888 		 * we need udpiphdr for IPsec processing so we do that later.
889 		 */
890 		/*
891 		 * Locate pcb(s) for datagram.
892 		 */
893 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
894 			in6p = (struct in6pcb *)inph;
895 			if (in6p->in6p_af != AF_INET6)
896 				continue;
897 
898 			if (in6p->in6p_lport != dport)
899 				continue;
900 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
901 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
902 				    dst6))
903 					continue;
904 			} else {
905 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
906 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
907 					continue;
908 			}
909 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
910 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
911 				    &src6) || in6p->in6p_fport != sport)
912 					continue;
913 			} else {
914 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
915 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
916 					continue;
917 			}
918 
919 			udp6_sendup(m, off, (struct sockaddr *)src,
920 				in6p->in6p_socket);
921 			rcvcnt++;
922 
923 			/*
924 			 * Don't look for additional matches if this one does
925 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
926 			 * socket options set.  This heuristic avoids searching
927 			 * through all pcbs in the common case of a non-shared
928 			 * port.  It assumes that an application will never
929 			 * clear these options after setting them.
930 			 */
931 			if ((in6p->in6p_socket->so_options &
932 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
933 				break;
934 		}
935 	} else {
936 		/*
937 		 * Locate pcb for datagram.
938 		 */
939 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
940 		    dport, 0);
941 		if (in6p == 0) {
942 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
943 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
944 			if (in6p == 0)
945 				return rcvcnt;
946 		}
947 
948 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
949 		rcvcnt++;
950 	}
951 
952 bad:
953 	return rcvcnt;
954 }
955 #endif
956 
957 #ifdef INET
958 /*
959  * Notify a udp user of an asynchronous error;
960  * just wake up so that he can collect error status.
961  */
962 static void
963 udp_notify(struct inpcb *inp, int errno)
964 {
965 	inp->inp_socket->so_error = errno;
966 	sorwakeup(inp->inp_socket);
967 	sowwakeup(inp->inp_socket);
968 }
969 
970 void *
971 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
972 {
973 	struct ip *ip = v;
974 	struct udphdr *uh;
975 	void (*notify)(struct inpcb *, int) = udp_notify;
976 	int errno;
977 
978 	if (sa->sa_family != AF_INET
979 	 || sa->sa_len != sizeof(struct sockaddr_in))
980 		return NULL;
981 	if ((unsigned)cmd >= PRC_NCMDS)
982 		return NULL;
983 	errno = inetctlerrmap[cmd];
984 	if (PRC_IS_REDIRECT(cmd))
985 		notify = in_rtchange, ip = 0;
986 	else if (cmd == PRC_HOSTDEAD)
987 		ip = 0;
988 	else if (errno == 0)
989 		return NULL;
990 	if (ip) {
991 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
992 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
993 		    ip->ip_src, uh->uh_sport, errno, notify);
994 
995 		/* XXX mapped address case */
996 	} else
997 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
998 		    notify);
999 	return NULL;
1000 }
1001 
1002 int
1003 udp_ctloutput(int op, struct socket *so, int level, int optname,
1004     struct mbuf **mp)
1005 {
1006 	int s;
1007 	int error = 0;
1008 	struct mbuf *m;
1009 	struct inpcb *inp;
1010 	int family;
1011 
1012 	family = so->so_proto->pr_domain->dom_family;
1013 
1014 	s = splsoftnet();
1015 	switch (family) {
1016 #ifdef INET
1017 	case PF_INET:
1018 		if (level != IPPROTO_UDP) {
1019 			error = ip_ctloutput(op, so, level, optname, mp);
1020 			goto end;
1021 		}
1022 		break;
1023 #endif
1024 #ifdef INET6
1025 	case PF_INET6:
1026 		if (level != IPPROTO_UDP) {
1027 			error = ip6_ctloutput(op, so, level, optname, mp);
1028 			goto end;
1029 		}
1030 		break;
1031 #endif
1032 	default:
1033 		error = EAFNOSUPPORT;
1034 		goto end;
1035 	}
1036 
1037 
1038 	switch (op) {
1039 	case PRCO_SETOPT:
1040 		m = *mp;
1041 		inp = sotoinpcb(so);
1042 
1043 		switch (optname) {
1044 		case UDP_ENCAP:
1045 			if (m == NULL || m->m_len != sizeof(int)) {
1046 				error = EINVAL;
1047 				break;
1048 			}
1049 
1050 			switch(*mtod(m, int *)) {
1051 #ifdef IPSEC_NAT_T
1052 			case 0:
1053 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1054 				break;
1055 
1056 			case UDP_ENCAP_ESPINUDP:
1057 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1058 				inp->inp_flags |= INP_ESPINUDP;
1059 				break;
1060 
1061 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1062 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1063 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1064 				break;
1065 #endif
1066 			default:
1067 				error = EINVAL;
1068 				break;
1069 			}
1070 			break;
1071 
1072 		default:
1073 			error = ENOPROTOOPT;
1074 			break;
1075 		}
1076 		if (m != NULL) {
1077 			m_free(m);
1078 		}
1079 		break;
1080 
1081 	default:
1082 		error = EINVAL;
1083 		break;
1084 	}
1085 
1086 end:
1087 	splx(s);
1088 	return error;
1089 }
1090 
1091 
1092 int
1093 udp_output(struct mbuf *m, ...)
1094 {
1095 	struct inpcb *inp;
1096 	struct udpiphdr *ui;
1097 	struct route *ro;
1098 	int len = m->m_pkthdr.len;
1099 	int error = 0;
1100 	va_list ap;
1101 
1102 	MCLAIM(m, &udp_tx_mowner);
1103 	va_start(ap, m);
1104 	inp = va_arg(ap, struct inpcb *);
1105 	va_end(ap);
1106 
1107 	/*
1108 	 * Calculate data length and get a mbuf
1109 	 * for UDP and IP headers.
1110 	 */
1111 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1112 	if (m == 0) {
1113 		error = ENOBUFS;
1114 		goto release;
1115 	}
1116 
1117 	/*
1118 	 * Compute the packet length of the IP header, and
1119 	 * punt if the length looks bogus.
1120 	 */
1121 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1122 		error = EMSGSIZE;
1123 		goto release;
1124 	}
1125 
1126 	/*
1127 	 * Fill in mbuf with extended UDP header
1128 	 * and addresses and length put into network format.
1129 	 */
1130 	ui = mtod(m, struct udpiphdr *);
1131 	ui->ui_pr = IPPROTO_UDP;
1132 	ui->ui_src = inp->inp_laddr;
1133 	ui->ui_dst = inp->inp_faddr;
1134 	ui->ui_sport = inp->inp_lport;
1135 	ui->ui_dport = inp->inp_fport;
1136 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1137 
1138 	ro = &inp->inp_route;
1139 
1140 	/*
1141 	 * Set up checksum and output datagram.
1142 	 */
1143 	if (udpcksum) {
1144 		/*
1145 		 * XXX Cache pseudo-header checksum part for
1146 		 * XXX "connected" UDP sockets.
1147 		 */
1148 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1149 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1150 		    sizeof(struct udphdr) + IPPROTO_UDP));
1151 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1152 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1153 	} else
1154 		ui->ui_sum = 0;
1155 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1156 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1157 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1158 	UDP_STATINC(UDP_STAT_OPACKETS);
1159 
1160 	return (ip_output(m, inp->inp_options, ro,
1161 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1162 	    inp->inp_moptions, inp->inp_socket));
1163 
1164 release:
1165 	m_freem(m);
1166 	return (error);
1167 }
1168 
1169 int	udp_sendspace = 9216;		/* really max datagram size */
1170 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1171 					/* 40 1K datagrams */
1172 
1173 /*ARGSUSED*/
1174 int
1175 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1176 	struct mbuf *control, struct lwp *l)
1177 {
1178 	struct inpcb *inp;
1179 	int s;
1180 	int error = 0;
1181 
1182 	if (req == PRU_CONTROL)
1183 		return (in_control(so, (long)m, (void *)nam,
1184 		    (struct ifnet *)control, l));
1185 
1186 	s = splsoftnet();
1187 
1188 	if (req == PRU_PURGEIF) {
1189 		mutex_enter(softnet_lock);
1190 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1191 		in_purgeif((struct ifnet *)control);
1192 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1193 		mutex_exit(softnet_lock);
1194 		splx(s);
1195 		return (0);
1196 	}
1197 
1198 	inp = sotoinpcb(so);
1199 #ifdef DIAGNOSTIC
1200 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1201 		panic("udp_usrreq: unexpected control mbuf");
1202 #endif
1203 	if (req == PRU_ATTACH) {
1204 		sosetlock(so);
1205 	} else if (inp == 0) {
1206 		error = EINVAL;
1207 		goto release;
1208 	}
1209 
1210 	/*
1211 	 * Note: need to block udp_input while changing
1212 	 * the udp pcb queue and/or pcb addresses.
1213 	 */
1214 	switch (req) {
1215 
1216 	case PRU_ATTACH:
1217 		if (inp != 0) {
1218 			error = EISCONN;
1219 			break;
1220 		}
1221 #ifdef MBUFTRACE
1222 		so->so_mowner = &udp_mowner;
1223 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1224 		so->so_snd.sb_mowner = &udp_tx_mowner;
1225 #endif
1226 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1227 			error = soreserve(so, udp_sendspace, udp_recvspace);
1228 			if (error)
1229 				break;
1230 		}
1231 		error = in_pcballoc(so, &udbtable);
1232 		if (error)
1233 			break;
1234 		inp = sotoinpcb(so);
1235 		inp->inp_ip.ip_ttl = ip_defttl;
1236 		break;
1237 
1238 	case PRU_DETACH:
1239 		in_pcbdetach(inp);
1240 		break;
1241 
1242 	case PRU_BIND:
1243 		error = in_pcbbind(inp, nam, l);
1244 		break;
1245 
1246 	case PRU_LISTEN:
1247 		error = EOPNOTSUPP;
1248 		break;
1249 
1250 	case PRU_CONNECT:
1251 		error = in_pcbconnect(inp, nam, l);
1252 		if (error)
1253 			break;
1254 		soisconnected(so);
1255 		break;
1256 
1257 	case PRU_CONNECT2:
1258 		error = EOPNOTSUPP;
1259 		break;
1260 
1261 	case PRU_DISCONNECT:
1262 		/*soisdisconnected(so);*/
1263 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1264 		in_pcbdisconnect(inp);
1265 		inp->inp_laddr = zeroin_addr;		/* XXX */
1266 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1267 		break;
1268 
1269 	case PRU_SHUTDOWN:
1270 		socantsendmore(so);
1271 		break;
1272 
1273 	case PRU_RCVD:
1274 		error = EOPNOTSUPP;
1275 		break;
1276 
1277 	case PRU_SEND:
1278 		if (control && control->m_len) {
1279 			m_freem(control);
1280 			m_freem(m);
1281 			error = EINVAL;
1282 			break;
1283 		}
1284 	{
1285 		struct in_addr laddr;			/* XXX */
1286 
1287 		if (nam) {
1288 			laddr = inp->inp_laddr;		/* XXX */
1289 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1290 				error = EISCONN;
1291 				goto die;
1292 			}
1293 			error = in_pcbconnect(inp, nam, l);
1294 			if (error)
1295 				goto die;
1296 		} else {
1297 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1298 				error = ENOTCONN;
1299 				goto die;
1300 			}
1301 		}
1302 		error = udp_output(m, inp);
1303 		m = NULL;
1304 		if (nam) {
1305 			in_pcbdisconnect(inp);
1306 			inp->inp_laddr = laddr;		/* XXX */
1307 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1308 		}
1309 	  die:
1310 		if (m)
1311 			m_freem(m);
1312 	}
1313 		break;
1314 
1315 	case PRU_SENSE:
1316 		/*
1317 		 * stat: don't bother with a blocksize.
1318 		 */
1319 		splx(s);
1320 		return (0);
1321 
1322 	case PRU_RCVOOB:
1323 		error =  EOPNOTSUPP;
1324 		break;
1325 
1326 	case PRU_SENDOOB:
1327 		m_freem(control);
1328 		m_freem(m);
1329 		error =  EOPNOTSUPP;
1330 		break;
1331 
1332 	case PRU_SOCKADDR:
1333 		in_setsockaddr(inp, nam);
1334 		break;
1335 
1336 	case PRU_PEERADDR:
1337 		in_setpeeraddr(inp, nam);
1338 		break;
1339 
1340 	default:
1341 		panic("udp_usrreq");
1342 	}
1343 
1344 release:
1345 	splx(s);
1346 	return (error);
1347 }
1348 
1349 static int
1350 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1351 {
1352 	netstat_sysctl_context ctx;
1353 	uint64_t udps[UDP_NSTATS];
1354 
1355 	ctx.ctx_stat = udpstat_percpu;
1356 	ctx.ctx_counters = udps;
1357 	ctx.ctx_ncounters = UDP_NSTATS;
1358 	return (NETSTAT_SYSCTL(&ctx));
1359 }
1360 
1361 /*
1362  * Sysctl for udp variables.
1363  */
1364 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1365 {
1366 
1367 	sysctl_createv(clog, 0, NULL, NULL,
1368 		       CTLFLAG_PERMANENT,
1369 		       CTLTYPE_NODE, "net", NULL,
1370 		       NULL, 0, NULL, 0,
1371 		       CTL_NET, CTL_EOL);
1372 	sysctl_createv(clog, 0, NULL, NULL,
1373 		       CTLFLAG_PERMANENT,
1374 		       CTLTYPE_NODE, "inet", NULL,
1375 		       NULL, 0, NULL, 0,
1376 		       CTL_NET, PF_INET, CTL_EOL);
1377 	sysctl_createv(clog, 0, NULL, NULL,
1378 		       CTLFLAG_PERMANENT,
1379 		       CTLTYPE_NODE, "udp",
1380 		       SYSCTL_DESCR("UDPv4 related settings"),
1381 		       NULL, 0, NULL, 0,
1382 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1383 
1384 	sysctl_createv(clog, 0, NULL, NULL,
1385 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1386 		       CTLTYPE_INT, "checksum",
1387 		       SYSCTL_DESCR("Compute UDP checksums"),
1388 		       NULL, 0, &udpcksum, 0,
1389 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1390 		       CTL_EOL);
1391 	sysctl_createv(clog, 0, NULL, NULL,
1392 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1393 		       CTLTYPE_INT, "sendspace",
1394 		       SYSCTL_DESCR("Default UDP send buffer size"),
1395 		       NULL, 0, &udp_sendspace, 0,
1396 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1397 		       CTL_EOL);
1398 	sysctl_createv(clog, 0, NULL, NULL,
1399 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1400 		       CTLTYPE_INT, "recvspace",
1401 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1402 		       NULL, 0, &udp_recvspace, 0,
1403 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1404 		       CTL_EOL);
1405 	sysctl_createv(clog, 0, NULL, NULL,
1406 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1407 		       CTLTYPE_INT, "do_loopback_cksum",
1408 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1409 		       NULL, 0, &udp_do_loopback_cksum, 0,
1410 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1411 		       CTL_EOL);
1412 	sysctl_createv(clog, 0, NULL, NULL,
1413 		       CTLFLAG_PERMANENT,
1414 		       CTLTYPE_STRUCT, "pcblist",
1415 		       SYSCTL_DESCR("UDP protocol control block list"),
1416 		       sysctl_inpcblist, 0, &udbtable, 0,
1417 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1418 		       CTL_EOL);
1419 	sysctl_createv(clog, 0, NULL, NULL,
1420 		       CTLFLAG_PERMANENT,
1421 		       CTLTYPE_STRUCT, "stats",
1422 		       SYSCTL_DESCR("UDP statistics"),
1423 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1424 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1425 		       CTL_EOL);
1426 }
1427 #endif
1428 
1429 void
1430 udp_statinc(u_int stat)
1431 {
1432 
1433 	KASSERT(stat < UDP_NSTATS);
1434 	UDP_STATINC(stat);
1435 }
1436 
1437 #if (defined INET && defined IPSEC_NAT_T)
1438 /*
1439  * Returns:
1440  * 1 if the packet was processed
1441  * 0 if normal UDP processing should take place
1442  * -1 if an error occurent and m was freed
1443  */
1444 static int
1445 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1446     struct socket *so)
1447 {
1448 	size_t len;
1449 	void *data;
1450 	struct inpcb *inp;
1451 	size_t skip = 0;
1452 	size_t minlen;
1453 	size_t iphdrlen;
1454 	struct ip *ip;
1455 	struct mbuf *n;
1456 	struct m_tag *tag;
1457 	struct udphdr *udphdr;
1458 	u_int16_t sport, dport;
1459 	struct mbuf *m = *mp;
1460 
1461 	/*
1462 	 * Collapse the mbuf chain if the first mbuf is too short
1463 	 * The longest case is: UDP + non ESP marker + ESP
1464 	 */
1465 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1466 	if (minlen > m->m_pkthdr.len)
1467 		minlen = m->m_pkthdr.len;
1468 
1469 	if (m->m_len < minlen) {
1470 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1471 			printf("udp4_espinudp: m_pullup failed\n");
1472 			return -1;
1473 		}
1474 		m = *mp;
1475 	}
1476 
1477 	len = m->m_len - off;
1478 	data = mtod(m, char *) + off;
1479 	inp = sotoinpcb(so);
1480 
1481 	/* Ignore keepalive packets */
1482 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1483 		return 1;
1484 	}
1485 
1486 	/*
1487 	 * Check that the payload is long enough to hold
1488 	 * an ESP header and compute the length of encapsulation
1489 	 * header to remove
1490 	 */
1491 	if (inp->inp_flags & INP_ESPINUDP) {
1492 		u_int32_t *st = (u_int32_t *)data;
1493 
1494 		if ((len <= sizeof(struct esp)) || (*st == 0))
1495 			return 0; /* Normal UDP processing */
1496 
1497 		skip = sizeof(struct udphdr);
1498 	}
1499 
1500 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1501 		u_int32_t *st = (u_int32_t *)data;
1502 
1503 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1504 		    || ((st[0] | st[1]) != 0))
1505 			return 0; /* Normal UDP processing */
1506 
1507 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1508 	}
1509 
1510 	/*
1511 	 * Get the UDP ports. They are handled in network
1512 	 * order everywhere in IPSEC_NAT_T code.
1513 	 */
1514 	udphdr = (struct udphdr *)((char *)data - skip);
1515 	sport = udphdr->uh_sport;
1516 	dport = udphdr->uh_dport;
1517 
1518 	/*
1519 	 * Remove the UDP header (and possibly the non ESP marker)
1520 	 * IP header lendth is iphdrlen
1521 	 * Before:
1522 	 *   <--- off --->
1523 	 *   +----+------+-----+
1524 	 *   | IP |  UDP | ESP |
1525 	 *   +----+------+-----+
1526 	 *        <-skip->
1527 	 * After:
1528 	 *          +----+-----+
1529 	 *          | IP | ESP |
1530 	 *          +----+-----+
1531 	 *   <-skip->
1532 	 */
1533 	iphdrlen = off - sizeof(struct udphdr);
1534 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1535 	m_adj(m, skip);
1536 
1537 	ip = mtod(m, struct ip *);
1538 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1539 	ip->ip_p = IPPROTO_ESP;
1540 
1541 	/*
1542 	 * Copy the mbuf to avoid multiple free, as both
1543 	 * esp4_input (which we call) and udp_input (which
1544 	 * called us) free the mbuf.
1545 	 */
1546 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1547 		printf("udp4_espinudp: m_dup failed\n");
1548 		return 0;
1549 	}
1550 
1551 	/*
1552 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1553 	 * the source UDP port. This is required if we want
1554 	 * to select the right SPD for multiple hosts behind
1555 	 * same NAT
1556 	 */
1557 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1558 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1559 		printf("udp4_espinudp: m_tag_get failed\n");
1560 		m_freem(n);
1561 		return 0;
1562 	}
1563 	((u_int16_t *)(tag + 1))[0] = sport;
1564 	((u_int16_t *)(tag + 1))[1] = dport;
1565 	m_tag_prepend(n, tag);
1566 
1567 #ifdef FAST_IPSEC
1568 	ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1569 #else
1570 	esp4_input(n, iphdrlen);
1571 #endif
1572 
1573 	/* We handled it, it shoudln't be handled by UDP */
1574 	return 1;
1575 }
1576 #endif
1577