xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision ca453df649ce9db45b64d73678ba06cbccf9aa11)
1 /*	$NetBSD: udp_usrreq.c,v 1.182 2011/07/17 20:54:53 joerg 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 1.182 2011/07/17 20:54:53 joerg Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_compat_netbsd.h"
68 #include "opt_ipsec.h"
69 #include "opt_inet_csum.h"
70 #include "opt_ipkdb.h"
71 #include "opt_mbuftrace.h"
72 
73 #include <sys/param.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/errno.h>
80 #include <sys/stat.h>
81 #include <sys/systm.h>
82 #include <sys/proc.h>
83 #include <sys/domain.h>
84 #include <sys/sysctl.h>
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_systm.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/ip_var.h>
95 #include <netinet/ip_icmp.h>
96 #include <netinet/udp.h>
97 #include <netinet/udp_var.h>
98 #include <netinet/udp_private.h>
99 
100 #ifdef INET6
101 #include <netinet/ip6.h>
102 #include <netinet/icmp6.h>
103 #include <netinet6/ip6_var.h>
104 #include <netinet6/ip6_private.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/udp6_var.h>
107 #include <netinet6/udp6_private.h>
108 #include <netinet6/scope6_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 #ifdef FAST_IPSEC
122 #include <netipsec/ipsec.h>
123 #include <netipsec/ipsec_var.h>
124 #include <netipsec/ipsec_private.h>
125 #include <netipsec/esp.h>
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 <netinet6/ipsec_private.h>
134 #include <netinet6/esp.h>
135 #include <netkey/key.h>
136 #endif /* IPSEC */
137 
138 #ifdef COMPAT_50
139 #include <compat/sys/socket.h>
140 #endif
141 
142 #ifdef IPKDB
143 #include <ipkdb/ipkdb.h>
144 #endif
145 
146 /*
147  * UDP protocol implementation.
148  * Per RFC 768, August, 1980.
149  */
150 int	udpcksum = 1;
151 int	udp_do_loopback_cksum = 0;
152 
153 struct	inpcbtable udbtable;
154 
155 percpu_t *udpstat_percpu;
156 
157 #ifdef INET
158 #ifdef IPSEC_NAT_T
159 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
160 	struct socket *);
161 #endif
162 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
163 	struct socket *);
164 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
165 	struct mbuf **, int);
166 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
167 #endif
168 #ifdef INET6
169 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
170 	struct socket *);
171 static int udp6_realinput (int, struct sockaddr_in6 *,
172 	struct sockaddr_in6 *, struct mbuf *, int);
173 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
174 #endif
175 #ifdef INET
176 static	void udp_notify (struct inpcb *, int);
177 #endif
178 
179 #ifndef UDBHASHSIZE
180 #define	UDBHASHSIZE	128
181 #endif
182 int	udbhashsize = UDBHASHSIZE;
183 
184 #ifdef MBUFTRACE
185 struct mowner udp_mowner = MOWNER_INIT("udp", "");
186 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
187 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
188 #endif
189 
190 #ifdef UDP_CSUM_COUNTERS
191 #include <sys/device.h>
192 
193 #if defined(INET)
194 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
195     NULL, "udp", "hwcsum bad");
196 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
197     NULL, "udp", "hwcsum ok");
198 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
199     NULL, "udp", "hwcsum data");
200 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
201     NULL, "udp", "swcsum");
202 
203 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
204 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
205 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
206 EVCNT_ATTACH_STATIC(udp_swcsum);
207 #endif /* defined(INET) */
208 
209 #if defined(INET6)
210 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
211     NULL, "udp6", "hwcsum bad");
212 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
213     NULL, "udp6", "hwcsum ok");
214 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
215     NULL, "udp6", "hwcsum data");
216 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
217     NULL, "udp6", "swcsum");
218 
219 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
220 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
221 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
222 EVCNT_ATTACH_STATIC(udp6_swcsum);
223 #endif /* defined(INET6) */
224 
225 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
226 
227 #else
228 
229 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
230 
231 #endif /* UDP_CSUM_COUNTERS */
232 
233 static void sysctl_net_inet_udp_setup(struct sysctllog **);
234 
235 void
236 udp_init(void)
237 {
238 
239 	sysctl_net_inet_udp_setup(NULL);
240 
241 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
242 
243 	MOWNER_ATTACH(&udp_tx_mowner);
244 	MOWNER_ATTACH(&udp_rx_mowner);
245 	MOWNER_ATTACH(&udp_mowner);
246 
247 #ifdef INET
248 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
249 #endif
250 #ifdef INET6
251 	udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS);
252 #endif
253 }
254 
255 /*
256  * Checksum extended UDP header and data.
257  */
258 
259 int
260 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
261     int iphlen, int len)
262 {
263 
264 	switch (af) {
265 #ifdef INET
266 	case AF_INET:
267 		return udp4_input_checksum(m, uh, iphlen, len);
268 #endif
269 #ifdef INET6
270 	case AF_INET6:
271 		return udp6_input_checksum(m, uh, iphlen, len);
272 #endif
273 	}
274 #ifdef DIAGNOSTIC
275 	panic("udp_input_checksum: unknown af %d", af);
276 #endif
277 	/* NOTREACHED */
278 	return -1;
279 }
280 
281 #ifdef INET
282 
283 /*
284  * Checksum extended UDP header and data.
285  */
286 
287 static int
288 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
289     int iphlen, int len)
290 {
291 
292 	/*
293 	 * XXX it's better to record and check if this mbuf is
294 	 * already checked.
295 	 */
296 
297 	if (uh->uh_sum == 0)
298 		return 0;
299 
300 	switch (m->m_pkthdr.csum_flags &
301 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
302 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
303 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
304 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
305 		goto badcsum;
306 
307 	case M_CSUM_UDPv4|M_CSUM_DATA: {
308 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
309 
310 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
311 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
312 			const struct ip *ip =
313 			    mtod(m, const struct ip *);
314 
315 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
316 			    ip->ip_dst.s_addr,
317 			    htons(hw_csum + len + IPPROTO_UDP));
318 		}
319 		if ((hw_csum ^ 0xffff) != 0)
320 			goto badcsum;
321 		break;
322 	}
323 
324 	case M_CSUM_UDPv4:
325 		/* Checksum was okay. */
326 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
327 		break;
328 
329 	default:
330 		/*
331 		 * Need to compute it ourselves.  Maybe skip checksum
332 		 * on loopback interfaces.
333 		 */
334 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
335 				     IFF_LOOPBACK) ||
336 				   udp_do_loopback_cksum)) {
337 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
338 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
339 				goto badcsum;
340 		}
341 		break;
342 	}
343 
344 	return 0;
345 
346 badcsum:
347 	UDP_STATINC(UDP_STAT_BADSUM);
348 	return -1;
349 }
350 
351 void
352 udp_input(struct mbuf *m, ...)
353 {
354 	va_list ap;
355 	struct sockaddr_in src, dst;
356 	struct ip *ip;
357 	struct udphdr *uh;
358 	int iphlen;
359 	int len;
360 	int n;
361 	u_int16_t ip_len;
362 
363 	va_start(ap, m);
364 	iphlen = va_arg(ap, int);
365 	(void)va_arg(ap, int);		/* ignore value, advance ap */
366 	va_end(ap);
367 
368 	MCLAIM(m, &udp_rx_mowner);
369 	UDP_STATINC(UDP_STAT_IPACKETS);
370 
371 	/*
372 	 * Get IP and UDP header together in first mbuf.
373 	 */
374 	ip = mtod(m, struct ip *);
375 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
376 	if (uh == NULL) {
377 		UDP_STATINC(UDP_STAT_HDROPS);
378 		return;
379 	}
380 	KASSERT(UDP_HDR_ALIGNED_P(uh));
381 
382 	/* destination port of 0 is illegal, based on RFC768. */
383 	if (uh->uh_dport == 0)
384 		goto bad;
385 
386 	/*
387 	 * Make mbuf data length reflect UDP length.
388 	 * If not enough data to reflect UDP length, drop.
389 	 */
390 	ip_len = ntohs(ip->ip_len);
391 	len = ntohs((u_int16_t)uh->uh_ulen);
392 	if (ip_len != iphlen + len) {
393 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
394 			UDP_STATINC(UDP_STAT_BADLEN);
395 			goto bad;
396 		}
397 		m_adj(m, iphlen + len - ip_len);
398 	}
399 
400 	/*
401 	 * Checksum extended UDP header and data.
402 	 */
403 	if (udp4_input_checksum(m, uh, iphlen, len))
404 		goto badcsum;
405 
406 	/* construct source and dst sockaddrs. */
407 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
408 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
409 
410 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
411 		UDP_STATINC(UDP_STAT_HDROPS);
412 		return;
413 	}
414 #ifdef INET6
415 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
416 		struct sockaddr_in6 src6, dst6;
417 
418 		memset(&src6, 0, sizeof(src6));
419 		src6.sin6_family = AF_INET6;
420 		src6.sin6_len = sizeof(struct sockaddr_in6);
421 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
422 		memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src,
423 			sizeof(ip->ip_src));
424 		src6.sin6_port = uh->uh_sport;
425 		memset(&dst6, 0, sizeof(dst6));
426 		dst6.sin6_family = AF_INET6;
427 		dst6.sin6_len = sizeof(struct sockaddr_in6);
428 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
429 		memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst,
430 			sizeof(ip->ip_dst));
431 		dst6.sin6_port = uh->uh_dport;
432 
433 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
434 	}
435 #endif
436 
437 	if (n == 0) {
438 		if (m->m_flags & (M_BCAST | M_MCAST)) {
439 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
440 			goto bad;
441 		}
442 		UDP_STATINC(UDP_STAT_NOPORT);
443 #ifdef IPKDB
444 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
445 				m, iphlen + sizeof(struct udphdr),
446 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
447 			/*
448 			 * It was a debugger connect packet,
449 			 * just drop it now
450 			 */
451 			goto bad;
452 		}
453 #endif
454 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
455 		m = NULL;
456 	}
457 
458 bad:
459 	if (m)
460 		m_freem(m);
461 	return;
462 
463 badcsum:
464 	m_freem(m);
465 }
466 #endif
467 
468 #ifdef INET6
469 static int
470 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
471 {
472 
473 	/*
474 	 * XXX it's better to record and check if this mbuf is
475 	 * already checked.
476 	 */
477 
478 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
479 		goto good;
480 	}
481 	if (uh->uh_sum == 0) {
482 		UDP6_STATINC(UDP6_STAT_NOSUM);
483 		goto bad;
484 	}
485 
486 	switch (m->m_pkthdr.csum_flags &
487 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
488 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
489 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
490 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
491 		UDP6_STATINC(UDP6_STAT_BADSUM);
492 		goto bad;
493 
494 #if 0 /* notyet */
495 	case M_CSUM_UDPv6|M_CSUM_DATA:
496 #endif
497 
498 	case M_CSUM_UDPv6:
499 		/* Checksum was okay. */
500 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
501 		break;
502 
503 	default:
504 		/*
505 		 * Need to compute it ourselves.  Maybe skip checksum
506 		 * on loopback interfaces.
507 		 */
508 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
509 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
510 			UDP6_STATINC(UDP6_STAT_BADSUM);
511 			goto bad;
512 		}
513 	}
514 
515 good:
516 	return 0;
517 bad:
518 	return -1;
519 }
520 
521 int
522 udp6_input(struct mbuf **mp, int *offp, int proto)
523 {
524 	struct mbuf *m = *mp;
525 	int off = *offp;
526 	struct sockaddr_in6 src, dst;
527 	struct ip6_hdr *ip6;
528 	struct udphdr *uh;
529 	u_int32_t plen, ulen;
530 
531 	ip6 = mtod(m, struct ip6_hdr *);
532 
533 #if defined(NFAITH) && 0 < NFAITH
534 	if (faithprefix(&ip6->ip6_dst)) {
535 		/* send icmp6 host unreach? */
536 		m_freem(m);
537 		return IPPROTO_DONE;
538 	}
539 #endif
540 
541 	UDP6_STATINC(UDP6_STAT_IPACKETS);
542 
543 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
544 	plen = m->m_pkthdr.len - off;
545 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
546 	if (uh == NULL) {
547 		IP6_STATINC(IP6_STAT_TOOSHORT);
548 		return IPPROTO_DONE;
549 	}
550 	KASSERT(UDP_HDR_ALIGNED_P(uh));
551 	ulen = ntohs((u_short)uh->uh_ulen);
552 	/*
553 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
554 	 * iff payload length > 0xffff.
555 	 */
556 	if (ulen == 0 && plen > 0xffff)
557 		ulen = plen;
558 
559 	if (plen != ulen) {
560 		UDP6_STATINC(UDP6_STAT_BADLEN);
561 		goto bad;
562 	}
563 
564 	/* destination port of 0 is illegal, based on RFC768. */
565 	if (uh->uh_dport == 0)
566 		goto bad;
567 
568 	/* Be proactive about malicious use of IPv4 mapped address */
569 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
570 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
571 		/* XXX stat */
572 		goto bad;
573 	}
574 
575 	/*
576 	 * Checksum extended UDP header and data.  Maybe skip checksum
577 	 * on loopback interfaces.
578 	 */
579 	if (udp6_input_checksum(m, uh, off, ulen))
580 		goto bad;
581 
582 	/*
583 	 * Construct source and dst sockaddrs.
584 	 */
585 	memset(&src, 0, sizeof(src));
586 	src.sin6_family = AF_INET6;
587 	src.sin6_len = sizeof(struct sockaddr_in6);
588 	src.sin6_addr = ip6->ip6_src;
589 	src.sin6_port = uh->uh_sport;
590 	memset(&dst, 0, sizeof(dst));
591 	dst.sin6_family = AF_INET6;
592 	dst.sin6_len = sizeof(struct sockaddr_in6);
593 	dst.sin6_addr = ip6->ip6_dst;
594 	dst.sin6_port = uh->uh_dport;
595 
596 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
597 		if (m->m_flags & M_MCAST) {
598 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
599 			goto bad;
600 		}
601 		UDP6_STATINC(UDP6_STAT_NOPORT);
602 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
603 		m = NULL;
604 	}
605 
606 bad:
607 	if (m)
608 		m_freem(m);
609 	return IPPROTO_DONE;
610 }
611 #endif
612 
613 #ifdef INET
614 static void
615 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
616 	struct sockaddr *src, struct socket *so)
617 {
618 	struct mbuf *opts = NULL;
619 	struct mbuf *n;
620 	struct inpcb *inp = NULL;
621 
622 	if (!so)
623 		return;
624 	switch (so->so_proto->pr_domain->dom_family) {
625 	case AF_INET:
626 		inp = sotoinpcb(so);
627 		break;
628 #ifdef INET6
629 	case AF_INET6:
630 		break;
631 #endif
632 	default:
633 		return;
634 	}
635 
636 #if defined(IPSEC) || defined(FAST_IPSEC)
637 	/* check AH/ESP integrity. */
638 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
639 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
640 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
641 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
642 			    0, 0);
643 		return;
644 	}
645 #endif /*IPSEC*/
646 
647 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
648 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
649 #ifdef SO_OTIMESTAMP
650 			 || so->so_options & SO_OTIMESTAMP
651 #endif
652 			 || so->so_options & SO_TIMESTAMP)) {
653 			struct ip *ip = mtod(n, struct ip *);
654 			ip_savecontrol(inp, &opts, ip, n);
655 		}
656 
657 		m_adj(n, off);
658 		if (sbappendaddr(&so->so_rcv, src, n,
659 				opts) == 0) {
660 			m_freem(n);
661 			if (opts)
662 				m_freem(opts);
663 			so->so_rcv.sb_overflowed++;
664 			UDP_STATINC(UDP_STAT_FULLSOCK);
665 		} else
666 			sorwakeup(so);
667 	}
668 }
669 #endif
670 
671 #ifdef INET6
672 static void
673 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
674 	struct sockaddr *src, struct socket *so)
675 {
676 	struct mbuf *opts = NULL;
677 	struct mbuf *n;
678 	struct in6pcb *in6p = NULL;
679 
680 	if (!so)
681 		return;
682 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
683 		return;
684 	in6p = sotoin6pcb(so);
685 
686 #if defined(IPSEC) || defined(FAST_IPSEC)
687 	/* check AH/ESP integrity. */
688 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
689 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
690 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
691 			icmp6_error(n, ICMP6_DST_UNREACH,
692 			    ICMP6_DST_UNREACH_ADMIN, 0);
693 		return;
694 	}
695 #endif /*IPSEC*/
696 
697 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
698 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
699 #ifdef SO_OTIMESTAMP
700 		    || in6p->in6p_socket->so_options & SO_OTIMESTAMP
701 #endif
702 		    || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
703 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
704 			ip6_savecontrol(in6p, &opts, ip6, n);
705 		}
706 
707 		m_adj(n, off);
708 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
709 			m_freem(n);
710 			if (opts)
711 				m_freem(opts);
712 			so->so_rcv.sb_overflowed++;
713 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
714 		} else
715 			sorwakeup(so);
716 	}
717 }
718 #endif
719 
720 #ifdef INET
721 static int
722 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
723 	struct mbuf **mp, int off /* offset of udphdr */)
724 {
725 	u_int16_t *sport, *dport;
726 	int rcvcnt;
727 	struct in_addr *src4, *dst4;
728 	struct inpcb_hdr *inph;
729 	struct inpcb *inp;
730 	struct mbuf *m = *mp;
731 
732 	rcvcnt = 0;
733 	off += sizeof(struct udphdr);	/* now, offset of payload */
734 
735 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
736 		goto bad;
737 
738 	src4 = &src->sin_addr;
739 	sport = &src->sin_port;
740 	dst4 = &dst->sin_addr;
741 	dport = &dst->sin_port;
742 
743 	if (IN_MULTICAST(dst4->s_addr) ||
744 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
745 		/*
746 		 * Deliver a multicast or broadcast datagram to *all* sockets
747 		 * for which the local and remote addresses and ports match
748 		 * those of the incoming datagram.  This allows more than
749 		 * one process to receive multi/broadcasts on the same port.
750 		 * (This really ought to be done for unicast datagrams as
751 		 * well, but that would cause problems with existing
752 		 * applications that open both address-specific sockets and
753 		 * a wildcard socket listening to the same port -- they would
754 		 * end up receiving duplicates of every unicast datagram.
755 		 * Those applications open the multiple sockets to overcome an
756 		 * inadequacy of the UDP socket interface, but for backwards
757 		 * compatibility we avoid the problem here rather than
758 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
759 		 */
760 
761 		/*
762 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
763 		 * we need udpiphdr for IPsec processing so we do that later.
764 		 */
765 		/*
766 		 * Locate pcb(s) for datagram.
767 		 */
768 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
769 			inp = (struct inpcb *)inph;
770 			if (inp->inp_af != AF_INET)
771 				continue;
772 
773 			if (inp->inp_lport != *dport)
774 				continue;
775 			if (!in_nullhost(inp->inp_laddr)) {
776 				if (!in_hosteq(inp->inp_laddr, *dst4))
777 					continue;
778 			}
779 			if (!in_nullhost(inp->inp_faddr)) {
780 				if (!in_hosteq(inp->inp_faddr, *src4) ||
781 				    inp->inp_fport != *sport)
782 					continue;
783 			}
784 
785 			udp4_sendup(m, off, (struct sockaddr *)src,
786 				inp->inp_socket);
787 			rcvcnt++;
788 
789 			/*
790 			 * Don't look for additional matches if this one does
791 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
792 			 * socket options set.  This heuristic avoids searching
793 			 * through all pcbs in the common case of a non-shared
794 			 * port.  It assumes that an application will never
795 			 * clear these options after setting them.
796 			 */
797 			if ((inp->inp_socket->so_options &
798 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
799 				break;
800 		}
801 	} else {
802 		/*
803 		 * Locate pcb for datagram.
804 		 */
805 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
806 		    *dport, 0);
807 		if (inp == 0) {
808 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
809 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
810 			if (inp == 0)
811 				return rcvcnt;
812 		}
813 
814 #ifdef IPSEC_NAT_T
815 		/* Handle ESP over UDP */
816 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
817 			struct sockaddr *sa = (struct sockaddr *)src;
818 
819 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
820 			case -1: 	/* Error, m was freeed */
821 				rcvcnt = -1;
822 				goto bad;
823 				break;
824 
825 			case 1:		/* ESP over UDP */
826 				rcvcnt++;
827 				goto bad;
828 				break;
829 
830 			case 0: 	/* plain UDP */
831 			default: 	/* Unexpected */
832 				/*
833 				 * Normal UDP processing will take place
834 				 * m may have changed.
835 				 */
836 				m = *mp;
837 				break;
838 			}
839 		}
840 #endif
841 
842 		/*
843 		 * Check the minimum TTL for socket.
844 		 */
845 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
846 			goto bad;
847 
848 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
849 		rcvcnt++;
850 	}
851 
852 bad:
853 	return rcvcnt;
854 }
855 #endif
856 
857 #ifdef INET6
858 static int
859 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
860 	struct mbuf *m, int off)
861 {
862 	u_int16_t sport, dport;
863 	int rcvcnt;
864 	struct in6_addr src6, *dst6;
865 	const struct in_addr *dst4;
866 	struct inpcb_hdr *inph;
867 	struct in6pcb *in6p;
868 
869 	rcvcnt = 0;
870 	off += sizeof(struct udphdr);	/* now, offset of payload */
871 
872 	if (af != AF_INET && af != AF_INET6)
873 		goto bad;
874 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
875 		goto bad;
876 
877 	src6 = src->sin6_addr;
878 	if (sa6_recoverscope(src) != 0) {
879 		/* XXX: should be impossible. */
880 		goto bad;
881 	}
882 	sport = src->sin6_port;
883 
884 	dport = dst->sin6_port;
885 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
886 	dst6 = &dst->sin6_addr;
887 
888 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
889 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
890 		/*
891 		 * Deliver a multicast or broadcast datagram to *all* sockets
892 		 * for which the local and remote addresses and ports match
893 		 * those of the incoming datagram.  This allows more than
894 		 * one process to receive multi/broadcasts on the same port.
895 		 * (This really ought to be done for unicast datagrams as
896 		 * well, but that would cause problems with existing
897 		 * applications that open both address-specific sockets and
898 		 * a wildcard socket listening to the same port -- they would
899 		 * end up receiving duplicates of every unicast datagram.
900 		 * Those applications open the multiple sockets to overcome an
901 		 * inadequacy of the UDP socket interface, but for backwards
902 		 * compatibility we avoid the problem here rather than
903 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
904 		 */
905 
906 		/*
907 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
908 		 * we need udpiphdr for IPsec processing so we do that later.
909 		 */
910 		/*
911 		 * Locate pcb(s) for datagram.
912 		 */
913 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
914 			in6p = (struct in6pcb *)inph;
915 			if (in6p->in6p_af != AF_INET6)
916 				continue;
917 
918 			if (in6p->in6p_lport != dport)
919 				continue;
920 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
921 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
922 				    dst6))
923 					continue;
924 			} else {
925 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
926 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
927 					continue;
928 			}
929 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
930 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
931 				    &src6) || in6p->in6p_fport != sport)
932 					continue;
933 			} else {
934 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
935 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
936 					continue;
937 			}
938 
939 			udp6_sendup(m, off, (struct sockaddr *)src,
940 				in6p->in6p_socket);
941 			rcvcnt++;
942 
943 			/*
944 			 * Don't look for additional matches if this one does
945 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
946 			 * socket options set.  This heuristic avoids searching
947 			 * through all pcbs in the common case of a non-shared
948 			 * port.  It assumes that an application will never
949 			 * clear these options after setting them.
950 			 */
951 			if ((in6p->in6p_socket->so_options &
952 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
953 				break;
954 		}
955 	} else {
956 		/*
957 		 * Locate pcb for datagram.
958 		 */
959 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
960 					     dport, 0, 0);
961 		if (in6p == 0) {
962 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
963 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
964 			if (in6p == 0)
965 				return rcvcnt;
966 		}
967 
968 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
969 		rcvcnt++;
970 	}
971 
972 bad:
973 	return rcvcnt;
974 }
975 #endif
976 
977 #ifdef INET
978 /*
979  * Notify a udp user of an asynchronous error;
980  * just wake up so that he can collect error status.
981  */
982 static void
983 udp_notify(struct inpcb *inp, int errno)
984 {
985 	inp->inp_socket->so_error = errno;
986 	sorwakeup(inp->inp_socket);
987 	sowwakeup(inp->inp_socket);
988 }
989 
990 void *
991 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
992 {
993 	struct ip *ip = v;
994 	struct udphdr *uh;
995 	void (*notify)(struct inpcb *, int) = udp_notify;
996 	int errno;
997 
998 	if (sa->sa_family != AF_INET
999 	 || sa->sa_len != sizeof(struct sockaddr_in))
1000 		return NULL;
1001 	if ((unsigned)cmd >= PRC_NCMDS)
1002 		return NULL;
1003 	errno = inetctlerrmap[cmd];
1004 	if (PRC_IS_REDIRECT(cmd))
1005 		notify = in_rtchange, ip = 0;
1006 	else if (cmd == PRC_HOSTDEAD)
1007 		ip = 0;
1008 	else if (errno == 0)
1009 		return NULL;
1010 	if (ip) {
1011 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1012 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1013 		    ip->ip_src, uh->uh_sport, errno, notify);
1014 
1015 		/* XXX mapped address case */
1016 	} else
1017 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1018 		    notify);
1019 	return NULL;
1020 }
1021 
1022 int
1023 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1024 {
1025 	int s;
1026 	int error = 0;
1027 	struct inpcb *inp;
1028 	int family;
1029 	int optval;
1030 
1031 	family = so->so_proto->pr_domain->dom_family;
1032 
1033 	s = splsoftnet();
1034 	switch (family) {
1035 #ifdef INET
1036 	case PF_INET:
1037 		if (sopt->sopt_level != IPPROTO_UDP) {
1038 			error = ip_ctloutput(op, so, sopt);
1039 			goto end;
1040 		}
1041 		break;
1042 #endif
1043 #ifdef INET6
1044 	case PF_INET6:
1045 		if (sopt->sopt_level != IPPROTO_UDP) {
1046 			error = ip6_ctloutput(op, so, sopt);
1047 			goto end;
1048 		}
1049 		break;
1050 #endif
1051 	default:
1052 		error = EAFNOSUPPORT;
1053 		goto end;
1054 	}
1055 
1056 
1057 	switch (op) {
1058 	case PRCO_SETOPT:
1059 		inp = sotoinpcb(so);
1060 
1061 		switch (sopt->sopt_name) {
1062 		case UDP_ENCAP:
1063 			error = sockopt_getint(sopt, &optval);
1064 			if (error)
1065 				break;
1066 
1067 			switch(optval) {
1068 #ifdef IPSEC_NAT_T
1069 			case 0:
1070 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1071 				break;
1072 
1073 			case UDP_ENCAP_ESPINUDP:
1074 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1075 				inp->inp_flags |= INP_ESPINUDP;
1076 				break;
1077 
1078 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1079 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1080 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1081 				break;
1082 #endif
1083 			default:
1084 				error = EINVAL;
1085 				break;
1086 			}
1087 			break;
1088 
1089 		default:
1090 			error = ENOPROTOOPT;
1091 			break;
1092 		}
1093 		break;
1094 
1095 	default:
1096 		error = EINVAL;
1097 		break;
1098 	}
1099 
1100 end:
1101 	splx(s);
1102 	return error;
1103 }
1104 
1105 
1106 int
1107 udp_output(struct mbuf *m, ...)
1108 {
1109 	struct inpcb *inp;
1110 	struct udpiphdr *ui;
1111 	struct route *ro;
1112 	int len = m->m_pkthdr.len;
1113 	int error = 0;
1114 	va_list ap;
1115 
1116 	MCLAIM(m, &udp_tx_mowner);
1117 	va_start(ap, m);
1118 	inp = va_arg(ap, struct inpcb *);
1119 	va_end(ap);
1120 
1121 	/*
1122 	 * Calculate data length and get a mbuf
1123 	 * for UDP and IP headers.
1124 	 */
1125 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1126 	if (m == 0) {
1127 		error = ENOBUFS;
1128 		goto release;
1129 	}
1130 
1131 	/*
1132 	 * Compute the packet length of the IP header, and
1133 	 * punt if the length looks bogus.
1134 	 */
1135 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1136 		error = EMSGSIZE;
1137 		goto release;
1138 	}
1139 
1140 	/*
1141 	 * Fill in mbuf with extended UDP header
1142 	 * and addresses and length put into network format.
1143 	 */
1144 	ui = mtod(m, struct udpiphdr *);
1145 	ui->ui_pr = IPPROTO_UDP;
1146 	ui->ui_src = inp->inp_laddr;
1147 	ui->ui_dst = inp->inp_faddr;
1148 	ui->ui_sport = inp->inp_lport;
1149 	ui->ui_dport = inp->inp_fport;
1150 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1151 
1152 	ro = &inp->inp_route;
1153 
1154 	/*
1155 	 * Set up checksum and output datagram.
1156 	 */
1157 	if (udpcksum) {
1158 		/*
1159 		 * XXX Cache pseudo-header checksum part for
1160 		 * XXX "connected" UDP sockets.
1161 		 */
1162 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1163 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1164 		    sizeof(struct udphdr) + IPPROTO_UDP));
1165 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1166 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1167 	} else
1168 		ui->ui_sum = 0;
1169 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1170 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1171 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1172 	UDP_STATINC(UDP_STAT_OPACKETS);
1173 
1174 	return (ip_output(m, inp->inp_options, ro,
1175 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1176 	    inp->inp_moptions, inp->inp_socket));
1177 
1178 release:
1179 	m_freem(m);
1180 	return (error);
1181 }
1182 
1183 int	udp_sendspace = 9216;		/* really max datagram size */
1184 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1185 					/* 40 1K datagrams */
1186 
1187 /*ARGSUSED*/
1188 int
1189 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1190 	struct mbuf *control, struct lwp *l)
1191 {
1192 	struct inpcb *inp;
1193 	int s;
1194 	int error = 0;
1195 
1196 	if (req == PRU_CONTROL)
1197 		return (in_control(so, (long)m, (void *)nam,
1198 		    (struct ifnet *)control, l));
1199 
1200 	s = splsoftnet();
1201 
1202 	if (req == PRU_PURGEIF) {
1203 		mutex_enter(softnet_lock);
1204 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1205 		in_purgeif((struct ifnet *)control);
1206 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1207 		mutex_exit(softnet_lock);
1208 		splx(s);
1209 		return (0);
1210 	}
1211 
1212 	inp = sotoinpcb(so);
1213 #ifdef DIAGNOSTIC
1214 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1215 		panic("udp_usrreq: unexpected control mbuf");
1216 #endif
1217 	if (req == PRU_ATTACH) {
1218 		sosetlock(so);
1219 	} else if (inp == 0) {
1220 		error = EINVAL;
1221 		goto release;
1222 	}
1223 
1224 	/*
1225 	 * Note: need to block udp_input while changing
1226 	 * the udp pcb queue and/or pcb addresses.
1227 	 */
1228 	switch (req) {
1229 
1230 	case PRU_ATTACH:
1231 		if (inp != 0) {
1232 			error = EISCONN;
1233 			break;
1234 		}
1235 #ifdef MBUFTRACE
1236 		so->so_mowner = &udp_mowner;
1237 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1238 		so->so_snd.sb_mowner = &udp_tx_mowner;
1239 #endif
1240 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1241 			error = soreserve(so, udp_sendspace, udp_recvspace);
1242 			if (error)
1243 				break;
1244 		}
1245 		error = in_pcballoc(so, &udbtable);
1246 		if (error)
1247 			break;
1248 		inp = sotoinpcb(so);
1249 		inp->inp_ip.ip_ttl = ip_defttl;
1250 		break;
1251 
1252 	case PRU_DETACH:
1253 		in_pcbdetach(inp);
1254 		break;
1255 
1256 	case PRU_BIND:
1257 		error = in_pcbbind(inp, nam, l);
1258 		break;
1259 
1260 	case PRU_LISTEN:
1261 		error = EOPNOTSUPP;
1262 		break;
1263 
1264 	case PRU_CONNECT:
1265 		error = in_pcbconnect(inp, nam, l);
1266 		if (error)
1267 			break;
1268 		soisconnected(so);
1269 		break;
1270 
1271 	case PRU_CONNECT2:
1272 		error = EOPNOTSUPP;
1273 		break;
1274 
1275 	case PRU_DISCONNECT:
1276 		/*soisdisconnected(so);*/
1277 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1278 		in_pcbdisconnect(inp);
1279 		inp->inp_laddr = zeroin_addr;		/* XXX */
1280 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1281 		break;
1282 
1283 	case PRU_SHUTDOWN:
1284 		socantsendmore(so);
1285 		break;
1286 
1287 	case PRU_RCVD:
1288 		error = EOPNOTSUPP;
1289 		break;
1290 
1291 	case PRU_SEND:
1292 		if (control && control->m_len) {
1293 			m_freem(control);
1294 			m_freem(m);
1295 			error = EINVAL;
1296 			break;
1297 		}
1298 	{
1299 		struct in_addr laddr;			/* XXX */
1300 
1301 		memset(&laddr, 0, sizeof laddr);
1302 		if (nam) {
1303 			laddr = inp->inp_laddr;		/* XXX */
1304 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1305 				error = EISCONN;
1306 				goto die;
1307 			}
1308 			error = in_pcbconnect(inp, nam, l);
1309 			if (error)
1310 				goto die;
1311 		} else {
1312 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1313 				error = ENOTCONN;
1314 				goto die;
1315 			}
1316 		}
1317 		error = udp_output(m, inp);
1318 		m = NULL;
1319 		if (nam) {
1320 			in_pcbdisconnect(inp);
1321 			inp->inp_laddr = laddr;		/* XXX */
1322 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1323 		}
1324 	  die:
1325 		if (m)
1326 			m_freem(m);
1327 	}
1328 		break;
1329 
1330 	case PRU_SENSE:
1331 		/*
1332 		 * stat: don't bother with a blocksize.
1333 		 */
1334 		splx(s);
1335 		return (0);
1336 
1337 	case PRU_RCVOOB:
1338 		error =  EOPNOTSUPP;
1339 		break;
1340 
1341 	case PRU_SENDOOB:
1342 		m_freem(control);
1343 		m_freem(m);
1344 		error =  EOPNOTSUPP;
1345 		break;
1346 
1347 	case PRU_SOCKADDR:
1348 		in_setsockaddr(inp, nam);
1349 		break;
1350 
1351 	case PRU_PEERADDR:
1352 		in_setpeeraddr(inp, nam);
1353 		break;
1354 
1355 	default:
1356 		panic("udp_usrreq");
1357 	}
1358 
1359 release:
1360 	splx(s);
1361 	return (error);
1362 }
1363 
1364 static int
1365 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1366 {
1367 
1368 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1369 }
1370 
1371 /*
1372  * Sysctl for udp variables.
1373  */
1374 static void
1375 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1376 {
1377 
1378 	sysctl_createv(clog, 0, NULL, NULL,
1379 		       CTLFLAG_PERMANENT,
1380 		       CTLTYPE_NODE, "net", NULL,
1381 		       NULL, 0, NULL, 0,
1382 		       CTL_NET, CTL_EOL);
1383 	sysctl_createv(clog, 0, NULL, NULL,
1384 		       CTLFLAG_PERMANENT,
1385 		       CTLTYPE_NODE, "inet", NULL,
1386 		       NULL, 0, NULL, 0,
1387 		       CTL_NET, PF_INET, CTL_EOL);
1388 	sysctl_createv(clog, 0, NULL, NULL,
1389 		       CTLFLAG_PERMANENT,
1390 		       CTLTYPE_NODE, "udp",
1391 		       SYSCTL_DESCR("UDPv4 related settings"),
1392 		       NULL, 0, NULL, 0,
1393 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1394 
1395 	sysctl_createv(clog, 0, NULL, NULL,
1396 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1397 		       CTLTYPE_INT, "checksum",
1398 		       SYSCTL_DESCR("Compute UDP checksums"),
1399 		       NULL, 0, &udpcksum, 0,
1400 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1401 		       CTL_EOL);
1402 	sysctl_createv(clog, 0, NULL, NULL,
1403 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1404 		       CTLTYPE_INT, "sendspace",
1405 		       SYSCTL_DESCR("Default UDP send buffer size"),
1406 		       NULL, 0, &udp_sendspace, 0,
1407 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1408 		       CTL_EOL);
1409 	sysctl_createv(clog, 0, NULL, NULL,
1410 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1411 		       CTLTYPE_INT, "recvspace",
1412 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1413 		       NULL, 0, &udp_recvspace, 0,
1414 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1415 		       CTL_EOL);
1416 	sysctl_createv(clog, 0, NULL, NULL,
1417 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1418 		       CTLTYPE_INT, "do_loopback_cksum",
1419 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1420 		       NULL, 0, &udp_do_loopback_cksum, 0,
1421 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1422 		       CTL_EOL);
1423 	sysctl_createv(clog, 0, NULL, NULL,
1424 		       CTLFLAG_PERMANENT,
1425 		       CTLTYPE_STRUCT, "pcblist",
1426 		       SYSCTL_DESCR("UDP protocol control block list"),
1427 		       sysctl_inpcblist, 0, &udbtable, 0,
1428 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1429 		       CTL_EOL);
1430 	sysctl_createv(clog, 0, NULL, NULL,
1431 		       CTLFLAG_PERMANENT,
1432 		       CTLTYPE_STRUCT, "stats",
1433 		       SYSCTL_DESCR("UDP statistics"),
1434 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1435 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1436 		       CTL_EOL);
1437 }
1438 #endif
1439 
1440 void
1441 udp_statinc(u_int stat)
1442 {
1443 
1444 	KASSERT(stat < UDP_NSTATS);
1445 	UDP_STATINC(stat);
1446 }
1447 
1448 #if (defined INET && defined IPSEC_NAT_T)
1449 /*
1450  * Returns:
1451  * 1 if the packet was processed
1452  * 0 if normal UDP processing should take place
1453  * -1 if an error occurent and m was freed
1454  */
1455 static int
1456 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1457     struct socket *so)
1458 {
1459 	size_t len;
1460 	void *data;
1461 	struct inpcb *inp;
1462 	size_t skip = 0;
1463 	size_t minlen;
1464 	size_t iphdrlen;
1465 	struct ip *ip;
1466 	struct mbuf *n;
1467 	struct m_tag *tag;
1468 	struct udphdr *udphdr;
1469 	u_int16_t sport, dport;
1470 	struct mbuf *m = *mp;
1471 
1472 	/*
1473 	 * Collapse the mbuf chain if the first mbuf is too short
1474 	 * The longest case is: UDP + non ESP marker + ESP
1475 	 */
1476 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1477 	if (minlen > m->m_pkthdr.len)
1478 		minlen = m->m_pkthdr.len;
1479 
1480 	if (m->m_len < minlen) {
1481 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1482 			printf("udp4_espinudp: m_pullup failed\n");
1483 			return -1;
1484 		}
1485 		m = *mp;
1486 	}
1487 
1488 	len = m->m_len - off;
1489 	data = mtod(m, char *) + off;
1490 	inp = sotoinpcb(so);
1491 
1492 	/* Ignore keepalive packets */
1493 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1494 		return 1;
1495 	}
1496 
1497 	/*
1498 	 * Check that the payload is long enough to hold
1499 	 * an ESP header and compute the length of encapsulation
1500 	 * header to remove
1501 	 */
1502 	if (inp->inp_flags & INP_ESPINUDP) {
1503 		u_int32_t *st = (u_int32_t *)data;
1504 
1505 		if ((len <= sizeof(struct esp)) || (*st == 0))
1506 			return 0; /* Normal UDP processing */
1507 
1508 		skip = sizeof(struct udphdr);
1509 	}
1510 
1511 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1512 		u_int32_t *st = (u_int32_t *)data;
1513 
1514 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1515 		    || ((st[0] | st[1]) != 0))
1516 			return 0; /* Normal UDP processing */
1517 
1518 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1519 	}
1520 
1521 	/*
1522 	 * Get the UDP ports. They are handled in network
1523 	 * order everywhere in IPSEC_NAT_T code.
1524 	 */
1525 	udphdr = (struct udphdr *)((char *)data - skip);
1526 	sport = udphdr->uh_sport;
1527 	dport = udphdr->uh_dport;
1528 
1529 	/*
1530 	 * Remove the UDP header (and possibly the non ESP marker)
1531 	 * IP header lendth is iphdrlen
1532 	 * Before:
1533 	 *   <--- off --->
1534 	 *   +----+------+-----+
1535 	 *   | IP |  UDP | ESP |
1536 	 *   +----+------+-----+
1537 	 *        <-skip->
1538 	 * After:
1539 	 *          +----+-----+
1540 	 *          | IP | ESP |
1541 	 *          +----+-----+
1542 	 *   <-skip->
1543 	 */
1544 	iphdrlen = off - sizeof(struct udphdr);
1545 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1546 	m_adj(m, skip);
1547 
1548 	ip = mtod(m, struct ip *);
1549 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1550 	ip->ip_p = IPPROTO_ESP;
1551 
1552 	/*
1553 	 * Copy the mbuf to avoid multiple free, as both
1554 	 * esp4_input (which we call) and udp_input (which
1555 	 * called us) free the mbuf.
1556 	 */
1557 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1558 		printf("udp4_espinudp: m_dup failed\n");
1559 		return 0;
1560 	}
1561 
1562 	/*
1563 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1564 	 * the source UDP port. This is required if we want
1565 	 * to select the right SPD for multiple hosts behind
1566 	 * same NAT
1567 	 */
1568 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1569 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1570 		printf("udp4_espinudp: m_tag_get failed\n");
1571 		m_freem(n);
1572 		return 0;
1573 	}
1574 	((u_int16_t *)(tag + 1))[0] = sport;
1575 	((u_int16_t *)(tag + 1))[1] = dport;
1576 	m_tag_prepend(n, tag);
1577 
1578 #ifdef FAST_IPSEC
1579 	ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1580 #else
1581 	esp4_input(n, iphdrlen);
1582 #endif
1583 
1584 	/* We handled it, it shoudln't be handled by UDP */
1585 	return 1;
1586 }
1587 #endif
1588