xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 4d12bfcd155352508213ace5ccc59ce930ea2974)
1 /*	$NetBSD: udp_usrreq.c,v 1.190 2013/06/05 19:01:26 christos 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 1.190 2013/06/05 19:01:26 christos 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 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	/* IPSEC */
130 
131 #ifdef COMPAT_50
132 #include <compat/sys/socket.h>
133 #endif
134 
135 #ifdef IPKDB
136 #include <ipkdb/ipkdb.h>
137 #endif
138 
139 /*
140  * UDP protocol implementation.
141  * Per RFC 768, August, 1980.
142  */
143 int	udpcksum = 1;
144 int	udp_do_loopback_cksum = 0;
145 
146 struct	inpcbtable udbtable;
147 
148 percpu_t *udpstat_percpu;
149 
150 #ifdef INET
151 #ifdef IPSEC
152 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
153 	struct socket *);
154 #endif
155 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
156 	struct socket *);
157 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
158 	struct mbuf **, int);
159 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
160 #endif
161 #ifdef INET6
162 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
163 	struct socket *);
164 static int udp6_realinput (int, struct sockaddr_in6 *,
165 	struct sockaddr_in6 *, struct mbuf *, int);
166 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
167 #endif
168 #ifdef INET
169 static	void udp_notify (struct inpcb *, int);
170 #endif
171 
172 #ifndef UDBHASHSIZE
173 #define	UDBHASHSIZE	128
174 #endif
175 int	udbhashsize = UDBHASHSIZE;
176 
177 #ifdef MBUFTRACE
178 struct mowner udp_mowner = MOWNER_INIT("udp", "");
179 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
180 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
181 #endif
182 
183 #ifdef UDP_CSUM_COUNTERS
184 #include <sys/device.h>
185 
186 #if defined(INET)
187 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
188     NULL, "udp", "hwcsum bad");
189 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
190     NULL, "udp", "hwcsum ok");
191 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192     NULL, "udp", "hwcsum data");
193 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
194     NULL, "udp", "swcsum");
195 
196 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
197 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
198 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
199 EVCNT_ATTACH_STATIC(udp_swcsum);
200 #endif /* defined(INET) */
201 
202 #if defined(INET6)
203 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
204     NULL, "udp6", "hwcsum bad");
205 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
206     NULL, "udp6", "hwcsum ok");
207 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208     NULL, "udp6", "hwcsum data");
209 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
210     NULL, "udp6", "swcsum");
211 
212 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
213 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
214 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
215 EVCNT_ATTACH_STATIC(udp6_swcsum);
216 #endif /* defined(INET6) */
217 
218 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
219 
220 #else
221 
222 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
223 
224 #endif /* UDP_CSUM_COUNTERS */
225 
226 static void sysctl_net_inet_udp_setup(struct sysctllog **);
227 
228 void
229 udp_init(void)
230 {
231 
232 	sysctl_net_inet_udp_setup(NULL);
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 	if (m == NULL) {
408 		/*
409 		 * packet has been processed by ESP stuff -
410 		 * e.g. dropped NAT-T-keep-alive-packet ...
411 		 */
412 		return;
413 	}
414 	ip = mtod(m, struct ip *);
415 #ifdef INET6
416 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
417 		struct sockaddr_in6 src6, dst6;
418 
419 		memset(&src6, 0, sizeof(src6));
420 		src6.sin6_family = AF_INET6;
421 		src6.sin6_len = sizeof(struct sockaddr_in6);
422 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
423 		memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src,
424 			sizeof(ip->ip_src));
425 		src6.sin6_port = uh->uh_sport;
426 		memset(&dst6, 0, sizeof(dst6));
427 		dst6.sin6_family = AF_INET6;
428 		dst6.sin6_len = sizeof(struct sockaddr_in6);
429 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
430 		memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst,
431 			sizeof(ip->ip_dst));
432 		dst6.sin6_port = uh->uh_dport;
433 
434 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
435 	}
436 #endif
437 
438 	if (n == 0) {
439 		if (m->m_flags & (M_BCAST | M_MCAST)) {
440 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
441 			goto bad;
442 		}
443 		UDP_STATINC(UDP_STAT_NOPORT);
444 #ifdef IPKDB
445 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
446 				m, iphlen + sizeof(struct udphdr),
447 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
448 			/*
449 			 * It was a debugger connect packet,
450 			 * just drop it now
451 			 */
452 			goto bad;
453 		}
454 #endif
455 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
456 		m = NULL;
457 	}
458 
459 bad:
460 	if (m)
461 		m_freem(m);
462 	return;
463 
464 badcsum:
465 	m_freem(m);
466 }
467 #endif
468 
469 #ifdef INET6
470 static int
471 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
472 {
473 
474 	/*
475 	 * XXX it's better to record and check if this mbuf is
476 	 * already checked.
477 	 */
478 
479 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
480 		goto good;
481 	}
482 	if (uh->uh_sum == 0) {
483 		UDP6_STATINC(UDP6_STAT_NOSUM);
484 		goto bad;
485 	}
486 
487 	switch (m->m_pkthdr.csum_flags &
488 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
489 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
490 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
491 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
492 		UDP6_STATINC(UDP6_STAT_BADSUM);
493 		goto bad;
494 
495 #if 0 /* notyet */
496 	case M_CSUM_UDPv6|M_CSUM_DATA:
497 #endif
498 
499 	case M_CSUM_UDPv6:
500 		/* Checksum was okay. */
501 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
502 		break;
503 
504 	default:
505 		/*
506 		 * Need to compute it ourselves.  Maybe skip checksum
507 		 * on loopback interfaces.
508 		 */
509 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
510 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
511 			UDP6_STATINC(UDP6_STAT_BADSUM);
512 			goto bad;
513 		}
514 	}
515 
516 good:
517 	return 0;
518 bad:
519 	return -1;
520 }
521 
522 int
523 udp6_input(struct mbuf **mp, int *offp, int proto)
524 {
525 	struct mbuf *m = *mp;
526 	int off = *offp;
527 	struct sockaddr_in6 src, dst;
528 	struct ip6_hdr *ip6;
529 	struct udphdr *uh;
530 	u_int32_t plen, ulen;
531 
532 	ip6 = mtod(m, struct ip6_hdr *);
533 
534 #if defined(NFAITH) && 0 < NFAITH
535 	if (faithprefix(&ip6->ip6_dst)) {
536 		/* send icmp6 host unreach? */
537 		m_freem(m);
538 		return IPPROTO_DONE;
539 	}
540 #endif
541 
542 	UDP6_STATINC(UDP6_STAT_IPACKETS);
543 
544 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
545 	plen = m->m_pkthdr.len - off;
546 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
547 	if (uh == NULL) {
548 		IP6_STATINC(IP6_STAT_TOOSHORT);
549 		return IPPROTO_DONE;
550 	}
551 	KASSERT(UDP_HDR_ALIGNED_P(uh));
552 	ulen = ntohs((u_short)uh->uh_ulen);
553 	/*
554 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
555 	 * iff payload length > 0xffff.
556 	 */
557 	if (ulen == 0 && plen > 0xffff)
558 		ulen = plen;
559 
560 	if (plen != ulen) {
561 		UDP6_STATINC(UDP6_STAT_BADLEN);
562 		goto bad;
563 	}
564 
565 	/* destination port of 0 is illegal, based on RFC768. */
566 	if (uh->uh_dport == 0)
567 		goto bad;
568 
569 	/* Be proactive about malicious use of IPv4 mapped address */
570 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
571 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
572 		/* XXX stat */
573 		goto bad;
574 	}
575 
576 	/*
577 	 * Checksum extended UDP header and data.  Maybe skip checksum
578 	 * on loopback interfaces.
579 	 */
580 	if (udp6_input_checksum(m, uh, off, ulen))
581 		goto bad;
582 
583 	/*
584 	 * Construct source and dst sockaddrs.
585 	 */
586 	memset(&src, 0, sizeof(src));
587 	src.sin6_family = AF_INET6;
588 	src.sin6_len = sizeof(struct sockaddr_in6);
589 	src.sin6_addr = ip6->ip6_src;
590 	src.sin6_port = uh->uh_sport;
591 	memset(&dst, 0, sizeof(dst));
592 	dst.sin6_family = AF_INET6;
593 	dst.sin6_len = sizeof(struct sockaddr_in6);
594 	dst.sin6_addr = ip6->ip6_dst;
595 	dst.sin6_port = uh->uh_dport;
596 
597 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
598 		if (m->m_flags & M_MCAST) {
599 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
600 			goto bad;
601 		}
602 		UDP6_STATINC(UDP6_STAT_NOPORT);
603 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
604 		m = NULL;
605 	}
606 
607 bad:
608 	if (m)
609 		m_freem(m);
610 	return IPPROTO_DONE;
611 }
612 #endif
613 
614 #ifdef INET
615 static void
616 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
617 	struct sockaddr *src, struct socket *so)
618 {
619 	struct mbuf *opts = NULL;
620 	struct mbuf *n;
621 	struct inpcb *inp = NULL;
622 
623 	if (!so)
624 		return;
625 	switch (so->so_proto->pr_domain->dom_family) {
626 	case AF_INET:
627 		inp = sotoinpcb(so);
628 		break;
629 #ifdef INET6
630 	case AF_INET6:
631 		break;
632 #endif
633 	default:
634 		return;
635 	}
636 
637 #if defined(IPSEC)
638 	/* check AH/ESP integrity. */
639 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
640 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
641 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
642 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
643 			    0, 0);
644 		return;
645 	}
646 #endif /*IPSEC*/
647 
648 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
649 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
650 #ifdef SO_OTIMESTAMP
651 			 || so->so_options & SO_OTIMESTAMP
652 #endif
653 			 || so->so_options & SO_TIMESTAMP)) {
654 			struct ip *ip = mtod(n, struct ip *);
655 			ip_savecontrol(inp, &opts, ip, n);
656 		}
657 
658 		m_adj(n, off);
659 		if (sbappendaddr(&so->so_rcv, src, n,
660 				opts) == 0) {
661 			m_freem(n);
662 			if (opts)
663 				m_freem(opts);
664 			so->so_rcv.sb_overflowed++;
665 			UDP_STATINC(UDP_STAT_FULLSOCK);
666 		} else
667 			sorwakeup(so);
668 	}
669 }
670 #endif
671 
672 #ifdef INET6
673 static void
674 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
675 	struct sockaddr *src, struct socket *so)
676 {
677 	struct mbuf *opts = NULL;
678 	struct mbuf *n;
679 	struct in6pcb *in6p = NULL;
680 
681 	if (!so)
682 		return;
683 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
684 		return;
685 	in6p = sotoin6pcb(so);
686 
687 #if defined(IPSEC)
688 	/* check AH/ESP integrity. */
689 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
690 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
691 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
692 			icmp6_error(n, ICMP6_DST_UNREACH,
693 			    ICMP6_DST_UNREACH_ADMIN, 0);
694 		return;
695 	}
696 #endif /*IPSEC*/
697 
698 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
699 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
700 #ifdef SO_OTIMESTAMP
701 		    || in6p->in6p_socket->so_options & SO_OTIMESTAMP
702 #endif
703 		    || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
704 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
705 			ip6_savecontrol(in6p, &opts, ip6, n);
706 		}
707 
708 		m_adj(n, off);
709 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
710 			m_freem(n);
711 			if (opts)
712 				m_freem(opts);
713 			so->so_rcv.sb_overflowed++;
714 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
715 		} else
716 			sorwakeup(so);
717 	}
718 }
719 #endif
720 
721 #ifdef INET
722 static int
723 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
724 	struct mbuf **mp, int off /* offset of udphdr */)
725 {
726 	u_int16_t *sport, *dport;
727 	int rcvcnt;
728 	struct in_addr *src4, *dst4;
729 	struct inpcb_hdr *inph;
730 	struct inpcb *inp;
731 	struct mbuf *m = *mp;
732 
733 	rcvcnt = 0;
734 	off += sizeof(struct udphdr);	/* now, offset of payload */
735 
736 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
737 		goto bad;
738 
739 	src4 = &src->sin_addr;
740 	sport = &src->sin_port;
741 	dst4 = &dst->sin_addr;
742 	dport = &dst->sin_port;
743 
744 	if (IN_MULTICAST(dst4->s_addr) ||
745 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
746 		/*
747 		 * Deliver a multicast or broadcast datagram to *all* sockets
748 		 * for which the local and remote addresses and ports match
749 		 * those of the incoming datagram.  This allows more than
750 		 * one process to receive multi/broadcasts on the same port.
751 		 * (This really ought to be done for unicast datagrams as
752 		 * well, but that would cause problems with existing
753 		 * applications that open both address-specific sockets and
754 		 * a wildcard socket listening to the same port -- they would
755 		 * end up receiving duplicates of every unicast datagram.
756 		 * Those applications open the multiple sockets to overcome an
757 		 * inadequacy of the UDP socket interface, but for backwards
758 		 * compatibility we avoid the problem here rather than
759 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
760 		 */
761 
762 		/*
763 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
764 		 * we need udpiphdr for IPsec processing so we do that later.
765 		 */
766 		/*
767 		 * Locate pcb(s) for datagram.
768 		 */
769 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
770 			inp = (struct inpcb *)inph;
771 			if (inp->inp_af != AF_INET)
772 				continue;
773 
774 			if (inp->inp_lport != *dport)
775 				continue;
776 			if (!in_nullhost(inp->inp_laddr)) {
777 				if (!in_hosteq(inp->inp_laddr, *dst4))
778 					continue;
779 			}
780 			if (!in_nullhost(inp->inp_faddr)) {
781 				if (!in_hosteq(inp->inp_faddr, *src4) ||
782 				    inp->inp_fport != *sport)
783 					continue;
784 			}
785 
786 			udp4_sendup(m, off, (struct sockaddr *)src,
787 				inp->inp_socket);
788 			rcvcnt++;
789 
790 			/*
791 			 * Don't look for additional matches if this one does
792 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
793 			 * socket options set.  This heuristic avoids searching
794 			 * through all pcbs in the common case of a non-shared
795 			 * port.  It assumes that an application will never
796 			 * clear these options after setting them.
797 			 */
798 			if ((inp->inp_socket->so_options &
799 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
800 				break;
801 		}
802 	} else {
803 		/*
804 		 * Locate pcb for datagram.
805 		 */
806 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
807 		    *dport, 0);
808 		if (inp == 0) {
809 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
810 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
811 			if (inp == 0)
812 				return rcvcnt;
813 		}
814 
815 #ifdef IPSEC
816 		/* Handle ESP over UDP */
817 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
818 			struct sockaddr *sa = (struct sockaddr *)src;
819 
820 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
821 			case -1: 	/* Error, m was freeed */
822 				rcvcnt = -1;
823 				goto bad;
824 				break;
825 
826 			case 1:		/* ESP over UDP */
827 				rcvcnt++;
828 				goto bad;
829 				break;
830 
831 			case 0: 	/* plain UDP */
832 			default: 	/* Unexpected */
833 				/*
834 				 * Normal UDP processing will take place
835 				 * m may have changed.
836 				 */
837 				m = *mp;
838 				break;
839 			}
840 		}
841 #endif
842 
843 		/*
844 		 * Check the minimum TTL for socket.
845 		 */
846 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
847 			goto bad;
848 
849 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
850 		rcvcnt++;
851 	}
852 
853 bad:
854 	return rcvcnt;
855 }
856 #endif
857 
858 #ifdef INET6
859 static int
860 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
861 	struct mbuf *m, int off)
862 {
863 	u_int16_t sport, dport;
864 	int rcvcnt;
865 	struct in6_addr src6, *dst6;
866 	const struct in_addr *dst4;
867 	struct inpcb_hdr *inph;
868 	struct in6pcb *in6p;
869 
870 	rcvcnt = 0;
871 	off += sizeof(struct udphdr);	/* now, offset of payload */
872 
873 	if (af != AF_INET && af != AF_INET6)
874 		goto bad;
875 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
876 		goto bad;
877 
878 	src6 = src->sin6_addr;
879 	if (sa6_recoverscope(src) != 0) {
880 		/* XXX: should be impossible. */
881 		goto bad;
882 	}
883 	sport = src->sin6_port;
884 
885 	dport = dst->sin6_port;
886 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
887 	dst6 = &dst->sin6_addr;
888 
889 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
890 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
891 		/*
892 		 * Deliver a multicast or broadcast datagram to *all* sockets
893 		 * for which the local and remote addresses and ports match
894 		 * those of the incoming datagram.  This allows more than
895 		 * one process to receive multi/broadcasts on the same port.
896 		 * (This really ought to be done for unicast datagrams as
897 		 * well, but that would cause problems with existing
898 		 * applications that open both address-specific sockets and
899 		 * a wildcard socket listening to the same port -- they would
900 		 * end up receiving duplicates of every unicast datagram.
901 		 * Those applications open the multiple sockets to overcome an
902 		 * inadequacy of the UDP socket interface, but for backwards
903 		 * compatibility we avoid the problem here rather than
904 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
905 		 */
906 
907 		/*
908 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
909 		 * we need udpiphdr for IPsec processing so we do that later.
910 		 */
911 		/*
912 		 * Locate pcb(s) for datagram.
913 		 */
914 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
915 			in6p = (struct in6pcb *)inph;
916 			if (in6p->in6p_af != AF_INET6)
917 				continue;
918 
919 			if (in6p->in6p_lport != dport)
920 				continue;
921 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
922 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
923 				    dst6))
924 					continue;
925 			} else {
926 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
927 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
928 					continue;
929 			}
930 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
931 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
932 				    &src6) || in6p->in6p_fport != sport)
933 					continue;
934 			} else {
935 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
936 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
937 					continue;
938 			}
939 
940 			udp6_sendup(m, off, (struct sockaddr *)src,
941 				in6p->in6p_socket);
942 			rcvcnt++;
943 
944 			/*
945 			 * Don't look for additional matches if this one does
946 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
947 			 * socket options set.  This heuristic avoids searching
948 			 * through all pcbs in the common case of a non-shared
949 			 * port.  It assumes that an application will never
950 			 * clear these options after setting them.
951 			 */
952 			if ((in6p->in6p_socket->so_options &
953 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
954 				break;
955 		}
956 	} else {
957 		/*
958 		 * Locate pcb for datagram.
959 		 */
960 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
961 					     dport, 0, 0);
962 		if (in6p == 0) {
963 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
964 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
965 			if (in6p == 0)
966 				return rcvcnt;
967 		}
968 
969 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
970 		rcvcnt++;
971 	}
972 
973 bad:
974 	return rcvcnt;
975 }
976 #endif
977 
978 #ifdef INET
979 /*
980  * Notify a udp user of an asynchronous error;
981  * just wake up so that he can collect error status.
982  */
983 static void
984 udp_notify(struct inpcb *inp, int errno)
985 {
986 	inp->inp_socket->so_error = errno;
987 	sorwakeup(inp->inp_socket);
988 	sowwakeup(inp->inp_socket);
989 }
990 
991 void *
992 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
993 {
994 	struct ip *ip = v;
995 	struct udphdr *uh;
996 	void (*notify)(struct inpcb *, int) = udp_notify;
997 	int errno;
998 
999 	if (sa->sa_family != AF_INET
1000 	 || sa->sa_len != sizeof(struct sockaddr_in))
1001 		return NULL;
1002 	if ((unsigned)cmd >= PRC_NCMDS)
1003 		return NULL;
1004 	errno = inetctlerrmap[cmd];
1005 	if (PRC_IS_REDIRECT(cmd))
1006 		notify = in_rtchange, ip = 0;
1007 	else if (cmd == PRC_HOSTDEAD)
1008 		ip = 0;
1009 	else if (errno == 0)
1010 		return NULL;
1011 	if (ip) {
1012 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1013 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1014 		    ip->ip_src, uh->uh_sport, errno, notify);
1015 
1016 		/* XXX mapped address case */
1017 	} else
1018 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1019 		    notify);
1020 	return NULL;
1021 }
1022 
1023 int
1024 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1025 {
1026 	int s;
1027 	int error = 0;
1028 	struct inpcb *inp;
1029 	int family;
1030 	int optval;
1031 
1032 	family = so->so_proto->pr_domain->dom_family;
1033 
1034 	s = splsoftnet();
1035 	switch (family) {
1036 #ifdef INET
1037 	case PF_INET:
1038 		if (sopt->sopt_level != IPPROTO_UDP) {
1039 			error = ip_ctloutput(op, so, sopt);
1040 			goto end;
1041 		}
1042 		break;
1043 #endif
1044 #ifdef INET6
1045 	case PF_INET6:
1046 		if (sopt->sopt_level != IPPROTO_UDP) {
1047 			error = ip6_ctloutput(op, so, sopt);
1048 			goto end;
1049 		}
1050 		break;
1051 #endif
1052 	default:
1053 		error = EAFNOSUPPORT;
1054 		goto end;
1055 	}
1056 
1057 
1058 	switch (op) {
1059 	case PRCO_SETOPT:
1060 		inp = sotoinpcb(so);
1061 
1062 		switch (sopt->sopt_name) {
1063 		case UDP_ENCAP:
1064 			error = sockopt_getint(sopt, &optval);
1065 			if (error)
1066 				break;
1067 
1068 			switch(optval) {
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 			default:
1083 				error = EINVAL;
1084 				break;
1085 			}
1086 			break;
1087 
1088 		default:
1089 			error = ENOPROTOOPT;
1090 			break;
1091 		}
1092 		break;
1093 
1094 	default:
1095 		error = EINVAL;
1096 		break;
1097 	}
1098 
1099 end:
1100 	splx(s);
1101 	return error;
1102 }
1103 
1104 
1105 int
1106 udp_output(struct mbuf *m, ...)
1107 {
1108 	struct inpcb *inp;
1109 	struct udpiphdr *ui;
1110 	struct route *ro;
1111 	int len = m->m_pkthdr.len;
1112 	int error = 0;
1113 	va_list ap;
1114 
1115 	MCLAIM(m, &udp_tx_mowner);
1116 	va_start(ap, m);
1117 	inp = va_arg(ap, struct inpcb *);
1118 	va_end(ap);
1119 
1120 	/*
1121 	 * Calculate data length and get a mbuf
1122 	 * for UDP and IP headers.
1123 	 */
1124 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1125 	if (m == 0) {
1126 		error = ENOBUFS;
1127 		goto release;
1128 	}
1129 
1130 	/*
1131 	 * Compute the packet length of the IP header, and
1132 	 * punt if the length looks bogus.
1133 	 */
1134 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1135 		error = EMSGSIZE;
1136 		goto release;
1137 	}
1138 
1139 	/*
1140 	 * Fill in mbuf with extended UDP header
1141 	 * and addresses and length put into network format.
1142 	 */
1143 	ui = mtod(m, struct udpiphdr *);
1144 	ui->ui_pr = IPPROTO_UDP;
1145 	ui->ui_src = inp->inp_laddr;
1146 	ui->ui_dst = inp->inp_faddr;
1147 	ui->ui_sport = inp->inp_lport;
1148 	ui->ui_dport = inp->inp_fport;
1149 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1150 
1151 	ro = &inp->inp_route;
1152 
1153 	/*
1154 	 * Set up checksum and output datagram.
1155 	 */
1156 	if (udpcksum) {
1157 		/*
1158 		 * XXX Cache pseudo-header checksum part for
1159 		 * XXX "connected" UDP sockets.
1160 		 */
1161 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1162 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1163 		    sizeof(struct udphdr) + IPPROTO_UDP));
1164 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1165 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1166 	} else
1167 		ui->ui_sum = 0;
1168 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1169 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1170 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1171 	UDP_STATINC(UDP_STAT_OPACKETS);
1172 
1173 	return (ip_output(m, inp->inp_options, ro,
1174 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1175 	    inp->inp_moptions, inp->inp_socket));
1176 
1177 release:
1178 	m_freem(m);
1179 	return (error);
1180 }
1181 
1182 int	udp_sendspace = 9216;		/* really max datagram size */
1183 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1184 					/* 40 1K datagrams */
1185 
1186 /*ARGSUSED*/
1187 int
1188 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1189 	struct mbuf *control, struct lwp *l)
1190 {
1191 	struct inpcb *inp;
1192 	int s;
1193 	int error = 0;
1194 
1195 	if (req == PRU_CONTROL)
1196 		return (in_control(so, (long)m, (void *)nam,
1197 		    (struct ifnet *)control, l));
1198 
1199 	s = splsoftnet();
1200 
1201 	if (req == PRU_PURGEIF) {
1202 		mutex_enter(softnet_lock);
1203 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1204 		in_purgeif((struct ifnet *)control);
1205 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1206 		mutex_exit(softnet_lock);
1207 		splx(s);
1208 		return (0);
1209 	}
1210 
1211 	inp = sotoinpcb(so);
1212 #ifdef DIAGNOSTIC
1213 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1214 		panic("udp_usrreq: unexpected control mbuf");
1215 #endif
1216 	if (req == PRU_ATTACH) {
1217 		sosetlock(so);
1218 	} else if (inp == 0) {
1219 		error = EINVAL;
1220 		goto release;
1221 	}
1222 
1223 	/*
1224 	 * Note: need to block udp_input while changing
1225 	 * the udp pcb queue and/or pcb addresses.
1226 	 */
1227 	switch (req) {
1228 
1229 	case PRU_ATTACH:
1230 		if (inp != 0) {
1231 			error = EISCONN;
1232 			break;
1233 		}
1234 #ifdef MBUFTRACE
1235 		so->so_mowner = &udp_mowner;
1236 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1237 		so->so_snd.sb_mowner = &udp_tx_mowner;
1238 #endif
1239 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1240 			error = soreserve(so, udp_sendspace, udp_recvspace);
1241 			if (error)
1242 				break;
1243 		}
1244 		error = in_pcballoc(so, &udbtable);
1245 		if (error)
1246 			break;
1247 		inp = sotoinpcb(so);
1248 		inp->inp_ip.ip_ttl = ip_defttl;
1249 		break;
1250 
1251 	case PRU_DETACH:
1252 		in_pcbdetach(inp);
1253 		break;
1254 
1255 	case PRU_BIND:
1256 		error = in_pcbbind(inp, nam, l);
1257 		break;
1258 
1259 	case PRU_LISTEN:
1260 		error = EOPNOTSUPP;
1261 		break;
1262 
1263 	case PRU_CONNECT:
1264 		error = in_pcbconnect(inp, nam, l);
1265 		if (error)
1266 			break;
1267 		soisconnected(so);
1268 		break;
1269 
1270 	case PRU_CONNECT2:
1271 		error = EOPNOTSUPP;
1272 		break;
1273 
1274 	case PRU_DISCONNECT:
1275 		/*soisdisconnected(so);*/
1276 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1277 		in_pcbdisconnect(inp);
1278 		inp->inp_laddr = zeroin_addr;		/* XXX */
1279 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1280 		break;
1281 
1282 	case PRU_SHUTDOWN:
1283 		socantsendmore(so);
1284 		break;
1285 
1286 	case PRU_RCVD:
1287 		error = EOPNOTSUPP;
1288 		break;
1289 
1290 	case PRU_SEND:
1291 		if (control && control->m_len) {
1292 			m_freem(control);
1293 			m_freem(m);
1294 			error = EINVAL;
1295 			break;
1296 		}
1297 	{
1298 		struct in_addr laddr;			/* XXX */
1299 
1300 		memset(&laddr, 0, sizeof laddr);
1301 		if (nam) {
1302 			laddr = inp->inp_laddr;		/* XXX */
1303 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1304 				error = EISCONN;
1305 				goto die;
1306 			}
1307 			error = in_pcbconnect(inp, nam, l);
1308 			if (error)
1309 				goto die;
1310 		} else {
1311 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1312 				error = ENOTCONN;
1313 				goto die;
1314 			}
1315 		}
1316 		error = udp_output(m, inp);
1317 		m = NULL;
1318 		if (nam) {
1319 			in_pcbdisconnect(inp);
1320 			inp->inp_laddr = laddr;		/* XXX */
1321 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1322 		}
1323 	  die:
1324 		if (m)
1325 			m_freem(m);
1326 	}
1327 		break;
1328 
1329 	case PRU_SENSE:
1330 		/*
1331 		 * stat: don't bother with a blocksize.
1332 		 */
1333 		splx(s);
1334 		return (0);
1335 
1336 	case PRU_RCVOOB:
1337 		error =  EOPNOTSUPP;
1338 		break;
1339 
1340 	case PRU_SENDOOB:
1341 		m_freem(control);
1342 		m_freem(m);
1343 		error =  EOPNOTSUPP;
1344 		break;
1345 
1346 	case PRU_SOCKADDR:
1347 		in_setsockaddr(inp, nam);
1348 		break;
1349 
1350 	case PRU_PEERADDR:
1351 		in_setpeeraddr(inp, nam);
1352 		break;
1353 
1354 	default:
1355 		panic("udp_usrreq");
1356 	}
1357 
1358 release:
1359 	splx(s);
1360 	return (error);
1361 }
1362 
1363 static int
1364 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1365 {
1366 
1367 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1368 }
1369 
1370 /*
1371  * Sysctl for udp variables.
1372  */
1373 static void
1374 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1375 {
1376 	sysctl_createv(clog, 0, NULL, NULL,
1377 		       CTLFLAG_PERMANENT,
1378 		       CTLTYPE_NODE, "net", NULL,
1379 		       NULL, 0, NULL, 0,
1380 		       CTL_NET, CTL_EOL);
1381 	sysctl_createv(clog, 0, NULL, NULL,
1382 		       CTLFLAG_PERMANENT,
1383 		       CTLTYPE_NODE, "inet", NULL,
1384 		       NULL, 0, NULL, 0,
1385 		       CTL_NET, PF_INET, CTL_EOL);
1386 	sysctl_createv(clog, 0, NULL, NULL,
1387 		       CTLFLAG_PERMANENT,
1388 		       CTLTYPE_NODE, "udp",
1389 		       SYSCTL_DESCR("UDPv4 related settings"),
1390 		       NULL, 0, NULL, 0,
1391 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1392 
1393 	sysctl_createv(clog, 0, NULL, NULL,
1394 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1395 		       CTLTYPE_INT, "checksum",
1396 		       SYSCTL_DESCR("Compute UDP checksums"),
1397 		       NULL, 0, &udpcksum, 0,
1398 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1399 		       CTL_EOL);
1400 	sysctl_createv(clog, 0, NULL, NULL,
1401 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1402 		       CTLTYPE_INT, "sendspace",
1403 		       SYSCTL_DESCR("Default UDP send buffer size"),
1404 		       NULL, 0, &udp_sendspace, 0,
1405 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1406 		       CTL_EOL);
1407 	sysctl_createv(clog, 0, NULL, NULL,
1408 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1409 		       CTLTYPE_INT, "recvspace",
1410 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1411 		       NULL, 0, &udp_recvspace, 0,
1412 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1413 		       CTL_EOL);
1414 	sysctl_createv(clog, 0, NULL, NULL,
1415 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1416 		       CTLTYPE_INT, "do_loopback_cksum",
1417 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1418 		       NULL, 0, &udp_do_loopback_cksum, 0,
1419 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1420 		       CTL_EOL);
1421 	sysctl_createv(clog, 0, NULL, NULL,
1422 		       CTLFLAG_PERMANENT,
1423 		       CTLTYPE_STRUCT, "pcblist",
1424 		       SYSCTL_DESCR("UDP protocol control block list"),
1425 		       sysctl_inpcblist, 0, &udbtable, 0,
1426 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1427 		       CTL_EOL);
1428 	sysctl_createv(clog, 0, NULL, NULL,
1429 		       CTLFLAG_PERMANENT,
1430 		       CTLTYPE_STRUCT, "stats",
1431 		       SYSCTL_DESCR("UDP statistics"),
1432 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1433 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1434 		       CTL_EOL);
1435 }
1436 #endif
1437 
1438 void
1439 udp_statinc(u_int stat)
1440 {
1441 
1442 	KASSERT(stat < UDP_NSTATS);
1443 	UDP_STATINC(stat);
1444 }
1445 
1446 #if defined(INET) && defined(IPSEC)
1447 /*
1448  * Returns:
1449  * 1 if the packet was processed
1450  * 0 if normal UDP processing should take place
1451  * -1 if an error occurent and m was freed
1452  */
1453 static int
1454 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1455     struct socket *so)
1456 {
1457 	size_t len;
1458 	void *data;
1459 	struct inpcb *inp;
1460 	size_t skip = 0;
1461 	size_t minlen;
1462 	size_t iphdrlen;
1463 	struct ip *ip;
1464 	struct m_tag *tag;
1465 	struct udphdr *udphdr;
1466 	u_int16_t sport, dport;
1467 	struct mbuf *m = *mp;
1468 
1469 	/*
1470 	 * Collapse the mbuf chain if the first mbuf is too short
1471 	 * The longest case is: UDP + non ESP marker + ESP
1472 	 */
1473 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1474 	if (minlen > m->m_pkthdr.len)
1475 		minlen = m->m_pkthdr.len;
1476 
1477 	if (m->m_len < minlen) {
1478 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1479 			printf("udp4_espinudp: m_pullup failed\n");
1480 			return -1;
1481 		}
1482 		m = *mp;
1483 	}
1484 
1485 	len = m->m_len - off;
1486 	data = mtod(m, char *) + off;
1487 	inp = sotoinpcb(so);
1488 
1489 	/* Ignore keepalive packets */
1490 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1491 		m_free(m);
1492 		*mp = NULL; /* avoid any further processiong by caller ... */
1493 		return 1;
1494 	}
1495 
1496 	/*
1497 	 * Check that the payload is long enough to hold
1498 	 * an ESP header and compute the length of encapsulation
1499 	 * header to remove
1500 	 */
1501 	if (inp->inp_flags & INP_ESPINUDP) {
1502 		u_int32_t *st = (u_int32_t *)data;
1503 
1504 		if ((len <= sizeof(struct esp)) || (*st == 0))
1505 			return 0; /* Normal UDP processing */
1506 
1507 		skip = sizeof(struct udphdr);
1508 	}
1509 
1510 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1511 		u_int32_t *st = (u_int32_t *)data;
1512 
1513 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1514 		    || ((st[0] | st[1]) != 0))
1515 			return 0; /* Normal UDP processing */
1516 
1517 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1518 	}
1519 
1520 	/*
1521 	 * Get the UDP ports. They are handled in network
1522 	 * order everywhere in IPSEC_NAT_T code.
1523 	 */
1524 	udphdr = (struct udphdr *)((char *)data - skip);
1525 	sport = udphdr->uh_sport;
1526 	dport = udphdr->uh_dport;
1527 
1528 	/*
1529 	 * Remove the UDP header (and possibly the non ESP marker)
1530 	 * IP header lendth is iphdrlen
1531 	 * Before:
1532 	 *   <--- off --->
1533 	 *   +----+------+-----+
1534 	 *   | IP |  UDP | ESP |
1535 	 *   +----+------+-----+
1536 	 *        <-skip->
1537 	 * After:
1538 	 *          +----+-----+
1539 	 *          | IP | ESP |
1540 	 *          +----+-----+
1541 	 *   <-skip->
1542 	 */
1543 	iphdrlen = off - sizeof(struct udphdr);
1544 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1545 	m_adj(m, skip);
1546 
1547 	ip = mtod(m, struct ip *);
1548 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1549 	ip->ip_p = IPPROTO_ESP;
1550 
1551 	/*
1552 	 * We have modified the packet - it is now ESP, so we should not
1553 	 * return to UDP processing ...
1554 	 *
1555 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1556 	 * the source UDP port. This is required if we want
1557 	 * to select the right SPD for multiple hosts behind
1558 	 * same NAT
1559 	 */
1560 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1561 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1562 		printf("udp4_espinudp: m_tag_get failed\n");
1563 		m_freem(m);
1564 		return -1;
1565 	}
1566 	((u_int16_t *)(tag + 1))[0] = sport;
1567 	((u_int16_t *)(tag + 1))[1] = dport;
1568 	m_tag_prepend(m, tag);
1569 
1570 #ifdef IPSEC
1571 	ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1572 #else
1573 	esp4_input(m, iphdrlen);
1574 #endif
1575 
1576 	/* We handled it, it shouldn't be handled by UDP */
1577 	*mp = NULL; /* avoid free by caller ... */
1578 	return 1;
1579 }
1580 #endif
1581