xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision a536ee5124e62c9a0051a252f7833dc8f50f44c9)
1 /*	$NetBSD: udp_usrreq.c,v 1.187 2012/06/22 14:54:35 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 1.187 2012/06/22 14:54:35 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 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 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_NAT_T
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 #ifdef INET6
408 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
409 		struct sockaddr_in6 src6, dst6;
410 
411 		memset(&src6, 0, 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 		memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src,
416 			sizeof(ip->ip_src));
417 		src6.sin6_port = uh->uh_sport;
418 		memset(&dst6, 0, 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 		memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst,
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 	memset(&src, 0, 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 	memset(&dst, 0, 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(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 #ifdef SO_OTIMESTAMP
643 			 || so->so_options & SO_OTIMESTAMP
644 #endif
645 			 || so->so_options & SO_TIMESTAMP)) {
646 			struct ip *ip = mtod(n, struct ip *);
647 			ip_savecontrol(inp, &opts, ip, n);
648 		}
649 
650 		m_adj(n, off);
651 		if (sbappendaddr(&so->so_rcv, src, n,
652 				opts) == 0) {
653 			m_freem(n);
654 			if (opts)
655 				m_freem(opts);
656 			so->so_rcv.sb_overflowed++;
657 			UDP_STATINC(UDP_STAT_FULLSOCK);
658 		} else
659 			sorwakeup(so);
660 	}
661 }
662 #endif
663 
664 #ifdef INET6
665 static void
666 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
667 	struct sockaddr *src, struct socket *so)
668 {
669 	struct mbuf *opts = NULL;
670 	struct mbuf *n;
671 	struct in6pcb *in6p = NULL;
672 
673 	if (!so)
674 		return;
675 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
676 		return;
677 	in6p = sotoin6pcb(so);
678 
679 #if defined(FAST_IPSEC)
680 	/* check AH/ESP integrity. */
681 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
682 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
683 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
684 			icmp6_error(n, ICMP6_DST_UNREACH,
685 			    ICMP6_DST_UNREACH_ADMIN, 0);
686 		return;
687 	}
688 #endif /*IPSEC*/
689 
690 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
691 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
692 #ifdef SO_OTIMESTAMP
693 		    || in6p->in6p_socket->so_options & SO_OTIMESTAMP
694 #endif
695 		    || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
696 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
697 			ip6_savecontrol(in6p, &opts, ip6, n);
698 		}
699 
700 		m_adj(n, off);
701 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
702 			m_freem(n);
703 			if (opts)
704 				m_freem(opts);
705 			so->so_rcv.sb_overflowed++;
706 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
707 		} else
708 			sorwakeup(so);
709 	}
710 }
711 #endif
712 
713 #ifdef INET
714 static int
715 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
716 	struct mbuf **mp, int off /* offset of udphdr */)
717 {
718 	u_int16_t *sport, *dport;
719 	int rcvcnt;
720 	struct in_addr *src4, *dst4;
721 	struct inpcb_hdr *inph;
722 	struct inpcb *inp;
723 	struct mbuf *m = *mp;
724 
725 	rcvcnt = 0;
726 	off += sizeof(struct udphdr);	/* now, offset of payload */
727 
728 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
729 		goto bad;
730 
731 	src4 = &src->sin_addr;
732 	sport = &src->sin_port;
733 	dst4 = &dst->sin_addr;
734 	dport = &dst->sin_port;
735 
736 	if (IN_MULTICAST(dst4->s_addr) ||
737 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
738 		/*
739 		 * Deliver a multicast or broadcast datagram to *all* sockets
740 		 * for which the local and remote addresses and ports match
741 		 * those of the incoming datagram.  This allows more than
742 		 * one process to receive multi/broadcasts on the same port.
743 		 * (This really ought to be done for unicast datagrams as
744 		 * well, but that would cause problems with existing
745 		 * applications that open both address-specific sockets and
746 		 * a wildcard socket listening to the same port -- they would
747 		 * end up receiving duplicates of every unicast datagram.
748 		 * Those applications open the multiple sockets to overcome an
749 		 * inadequacy of the UDP socket interface, but for backwards
750 		 * compatibility we avoid the problem here rather than
751 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
752 		 */
753 
754 		/*
755 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
756 		 * we need udpiphdr for IPsec processing so we do that later.
757 		 */
758 		/*
759 		 * Locate pcb(s) for datagram.
760 		 */
761 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
762 			inp = (struct inpcb *)inph;
763 			if (inp->inp_af != AF_INET)
764 				continue;
765 
766 			if (inp->inp_lport != *dport)
767 				continue;
768 			if (!in_nullhost(inp->inp_laddr)) {
769 				if (!in_hosteq(inp->inp_laddr, *dst4))
770 					continue;
771 			}
772 			if (!in_nullhost(inp->inp_faddr)) {
773 				if (!in_hosteq(inp->inp_faddr, *src4) ||
774 				    inp->inp_fport != *sport)
775 					continue;
776 			}
777 
778 			udp4_sendup(m, off, (struct sockaddr *)src,
779 				inp->inp_socket);
780 			rcvcnt++;
781 
782 			/*
783 			 * Don't look for additional matches if this one does
784 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
785 			 * socket options set.  This heuristic avoids searching
786 			 * through all pcbs in the common case of a non-shared
787 			 * port.  It assumes that an application will never
788 			 * clear these options after setting them.
789 			 */
790 			if ((inp->inp_socket->so_options &
791 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
792 				break;
793 		}
794 	} else {
795 		/*
796 		 * Locate pcb for datagram.
797 		 */
798 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
799 		    *dport, 0);
800 		if (inp == 0) {
801 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
802 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
803 			if (inp == 0)
804 				return rcvcnt;
805 		}
806 
807 #ifdef IPSEC_NAT_T
808 		/* Handle ESP over UDP */
809 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
810 			struct sockaddr *sa = (struct sockaddr *)src;
811 
812 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
813 			case -1: 	/* Error, m was freeed */
814 				rcvcnt = -1;
815 				goto bad;
816 				break;
817 
818 			case 1:		/* ESP over UDP */
819 				rcvcnt++;
820 				goto bad;
821 				break;
822 
823 			case 0: 	/* plain UDP */
824 			default: 	/* Unexpected */
825 				/*
826 				 * Normal UDP processing will take place
827 				 * m may have changed.
828 				 */
829 				m = *mp;
830 				break;
831 			}
832 		}
833 #endif
834 
835 		/*
836 		 * Check the minimum TTL for socket.
837 		 */
838 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
839 			goto bad;
840 
841 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
842 		rcvcnt++;
843 	}
844 
845 bad:
846 	return rcvcnt;
847 }
848 #endif
849 
850 #ifdef INET6
851 static int
852 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
853 	struct mbuf *m, int off)
854 {
855 	u_int16_t sport, dport;
856 	int rcvcnt;
857 	struct in6_addr src6, *dst6;
858 	const struct in_addr *dst4;
859 	struct inpcb_hdr *inph;
860 	struct in6pcb *in6p;
861 
862 	rcvcnt = 0;
863 	off += sizeof(struct udphdr);	/* now, offset of payload */
864 
865 	if (af != AF_INET && af != AF_INET6)
866 		goto bad;
867 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
868 		goto bad;
869 
870 	src6 = src->sin6_addr;
871 	if (sa6_recoverscope(src) != 0) {
872 		/* XXX: should be impossible. */
873 		goto bad;
874 	}
875 	sport = src->sin6_port;
876 
877 	dport = dst->sin6_port;
878 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
879 	dst6 = &dst->sin6_addr;
880 
881 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
882 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
883 		/*
884 		 * Deliver a multicast or broadcast datagram to *all* sockets
885 		 * for which the local and remote addresses and ports match
886 		 * those of the incoming datagram.  This allows more than
887 		 * one process to receive multi/broadcasts on the same port.
888 		 * (This really ought to be done for unicast datagrams as
889 		 * well, but that would cause problems with existing
890 		 * applications that open both address-specific sockets and
891 		 * a wildcard socket listening to the same port -- they would
892 		 * end up receiving duplicates of every unicast datagram.
893 		 * Those applications open the multiple sockets to overcome an
894 		 * inadequacy of the UDP socket interface, but for backwards
895 		 * compatibility we avoid the problem here rather than
896 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
897 		 */
898 
899 		/*
900 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
901 		 * we need udpiphdr for IPsec processing so we do that later.
902 		 */
903 		/*
904 		 * Locate pcb(s) for datagram.
905 		 */
906 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
907 			in6p = (struct in6pcb *)inph;
908 			if (in6p->in6p_af != AF_INET6)
909 				continue;
910 
911 			if (in6p->in6p_lport != dport)
912 				continue;
913 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
914 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
915 				    dst6))
916 					continue;
917 			} else {
918 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
919 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
920 					continue;
921 			}
922 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
923 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
924 				    &src6) || in6p->in6p_fport != sport)
925 					continue;
926 			} else {
927 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
928 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
929 					continue;
930 			}
931 
932 			udp6_sendup(m, off, (struct sockaddr *)src,
933 				in6p->in6p_socket);
934 			rcvcnt++;
935 
936 			/*
937 			 * Don't look for additional matches if this one does
938 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
939 			 * socket options set.  This heuristic avoids searching
940 			 * through all pcbs in the common case of a non-shared
941 			 * port.  It assumes that an application will never
942 			 * clear these options after setting them.
943 			 */
944 			if ((in6p->in6p_socket->so_options &
945 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
946 				break;
947 		}
948 	} else {
949 		/*
950 		 * Locate pcb for datagram.
951 		 */
952 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
953 					     dport, 0, 0);
954 		if (in6p == 0) {
955 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
956 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
957 			if (in6p == 0)
958 				return rcvcnt;
959 		}
960 
961 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
962 		rcvcnt++;
963 	}
964 
965 bad:
966 	return rcvcnt;
967 }
968 #endif
969 
970 #ifdef INET
971 /*
972  * Notify a udp user of an asynchronous error;
973  * just wake up so that he can collect error status.
974  */
975 static void
976 udp_notify(struct inpcb *inp, int errno)
977 {
978 	inp->inp_socket->so_error = errno;
979 	sorwakeup(inp->inp_socket);
980 	sowwakeup(inp->inp_socket);
981 }
982 
983 void *
984 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
985 {
986 	struct ip *ip = v;
987 	struct udphdr *uh;
988 	void (*notify)(struct inpcb *, int) = udp_notify;
989 	int errno;
990 
991 	if (sa->sa_family != AF_INET
992 	 || sa->sa_len != sizeof(struct sockaddr_in))
993 		return NULL;
994 	if ((unsigned)cmd >= PRC_NCMDS)
995 		return NULL;
996 	errno = inetctlerrmap[cmd];
997 	if (PRC_IS_REDIRECT(cmd))
998 		notify = in_rtchange, ip = 0;
999 	else if (cmd == PRC_HOSTDEAD)
1000 		ip = 0;
1001 	else if (errno == 0)
1002 		return NULL;
1003 	if (ip) {
1004 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1005 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1006 		    ip->ip_src, uh->uh_sport, errno, notify);
1007 
1008 		/* XXX mapped address case */
1009 	} else
1010 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1011 		    notify);
1012 	return NULL;
1013 }
1014 
1015 int
1016 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1017 {
1018 	int s;
1019 	int error = 0;
1020 	struct inpcb *inp;
1021 	int family;
1022 	int optval;
1023 
1024 	family = so->so_proto->pr_domain->dom_family;
1025 
1026 	s = splsoftnet();
1027 	switch (family) {
1028 #ifdef INET
1029 	case PF_INET:
1030 		if (sopt->sopt_level != IPPROTO_UDP) {
1031 			error = ip_ctloutput(op, so, sopt);
1032 			goto end;
1033 		}
1034 		break;
1035 #endif
1036 #ifdef INET6
1037 	case PF_INET6:
1038 		if (sopt->sopt_level != IPPROTO_UDP) {
1039 			error = ip6_ctloutput(op, so, sopt);
1040 			goto end;
1041 		}
1042 		break;
1043 #endif
1044 	default:
1045 		error = EAFNOSUPPORT;
1046 		goto end;
1047 	}
1048 
1049 
1050 	switch (op) {
1051 	case PRCO_SETOPT:
1052 		inp = sotoinpcb(so);
1053 
1054 		switch (sopt->sopt_name) {
1055 		case UDP_ENCAP:
1056 			error = sockopt_getint(sopt, &optval);
1057 			if (error)
1058 				break;
1059 
1060 			switch(optval) {
1061 #ifdef IPSEC_NAT_T
1062 			case 0:
1063 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1064 				break;
1065 
1066 			case UDP_ENCAP_ESPINUDP:
1067 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1068 				inp->inp_flags |= INP_ESPINUDP;
1069 				break;
1070 
1071 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1072 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1073 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1074 				break;
1075 #endif
1076 			default:
1077 				error = EINVAL;
1078 				break;
1079 			}
1080 			break;
1081 
1082 		default:
1083 			error = ENOPROTOOPT;
1084 			break;
1085 		}
1086 		break;
1087 
1088 	default:
1089 		error = EINVAL;
1090 		break;
1091 	}
1092 
1093 end:
1094 	splx(s);
1095 	return error;
1096 }
1097 
1098 
1099 int
1100 udp_output(struct mbuf *m, ...)
1101 {
1102 	struct inpcb *inp;
1103 	struct udpiphdr *ui;
1104 	struct route *ro;
1105 	int len = m->m_pkthdr.len;
1106 	int error = 0;
1107 	va_list ap;
1108 
1109 	MCLAIM(m, &udp_tx_mowner);
1110 	va_start(ap, m);
1111 	inp = va_arg(ap, struct inpcb *);
1112 	va_end(ap);
1113 
1114 	/*
1115 	 * Calculate data length and get a mbuf
1116 	 * for UDP and IP headers.
1117 	 */
1118 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1119 	if (m == 0) {
1120 		error = ENOBUFS;
1121 		goto release;
1122 	}
1123 
1124 	/*
1125 	 * Compute the packet length of the IP header, and
1126 	 * punt if the length looks bogus.
1127 	 */
1128 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1129 		error = EMSGSIZE;
1130 		goto release;
1131 	}
1132 
1133 	/*
1134 	 * Fill in mbuf with extended UDP header
1135 	 * and addresses and length put into network format.
1136 	 */
1137 	ui = mtod(m, struct udpiphdr *);
1138 	ui->ui_pr = IPPROTO_UDP;
1139 	ui->ui_src = inp->inp_laddr;
1140 	ui->ui_dst = inp->inp_faddr;
1141 	ui->ui_sport = inp->inp_lport;
1142 	ui->ui_dport = inp->inp_fport;
1143 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1144 
1145 	ro = &inp->inp_route;
1146 
1147 	/*
1148 	 * Set up checksum and output datagram.
1149 	 */
1150 	if (udpcksum) {
1151 		/*
1152 		 * XXX Cache pseudo-header checksum part for
1153 		 * XXX "connected" UDP sockets.
1154 		 */
1155 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1156 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1157 		    sizeof(struct udphdr) + IPPROTO_UDP));
1158 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1159 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1160 	} else
1161 		ui->ui_sum = 0;
1162 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1163 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1164 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1165 	UDP_STATINC(UDP_STAT_OPACKETS);
1166 
1167 	return (ip_output(m, inp->inp_options, ro,
1168 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1169 	    inp->inp_moptions, inp->inp_socket));
1170 
1171 release:
1172 	m_freem(m);
1173 	return (error);
1174 }
1175 
1176 int	udp_sendspace = 9216;		/* really max datagram size */
1177 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1178 					/* 40 1K datagrams */
1179 
1180 /*ARGSUSED*/
1181 int
1182 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1183 	struct mbuf *control, struct lwp *l)
1184 {
1185 	struct inpcb *inp;
1186 	int s;
1187 	int error = 0;
1188 
1189 	if (req == PRU_CONTROL)
1190 		return (in_control(so, (long)m, (void *)nam,
1191 		    (struct ifnet *)control, l));
1192 
1193 	s = splsoftnet();
1194 
1195 	if (req == PRU_PURGEIF) {
1196 		mutex_enter(softnet_lock);
1197 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1198 		in_purgeif((struct ifnet *)control);
1199 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1200 		mutex_exit(softnet_lock);
1201 		splx(s);
1202 		return (0);
1203 	}
1204 
1205 	inp = sotoinpcb(so);
1206 #ifdef DIAGNOSTIC
1207 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1208 		panic("udp_usrreq: unexpected control mbuf");
1209 #endif
1210 	if (req == PRU_ATTACH) {
1211 		sosetlock(so);
1212 	} else if (inp == 0) {
1213 		error = EINVAL;
1214 		goto release;
1215 	}
1216 
1217 	/*
1218 	 * Note: need to block udp_input while changing
1219 	 * the udp pcb queue and/or pcb addresses.
1220 	 */
1221 	switch (req) {
1222 
1223 	case PRU_ATTACH:
1224 		if (inp != 0) {
1225 			error = EISCONN;
1226 			break;
1227 		}
1228 #ifdef MBUFTRACE
1229 		so->so_mowner = &udp_mowner;
1230 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1231 		so->so_snd.sb_mowner = &udp_tx_mowner;
1232 #endif
1233 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1234 			error = soreserve(so, udp_sendspace, udp_recvspace);
1235 			if (error)
1236 				break;
1237 		}
1238 		error = in_pcballoc(so, &udbtable);
1239 		if (error)
1240 			break;
1241 		inp = sotoinpcb(so);
1242 		inp->inp_ip.ip_ttl = ip_defttl;
1243 		break;
1244 
1245 	case PRU_DETACH:
1246 		in_pcbdetach(inp);
1247 		break;
1248 
1249 	case PRU_BIND:
1250 		error = in_pcbbind(inp, nam, l);
1251 		break;
1252 
1253 	case PRU_LISTEN:
1254 		error = EOPNOTSUPP;
1255 		break;
1256 
1257 	case PRU_CONNECT:
1258 		error = in_pcbconnect(inp, nam, l);
1259 		if (error)
1260 			break;
1261 		soisconnected(so);
1262 		break;
1263 
1264 	case PRU_CONNECT2:
1265 		error = EOPNOTSUPP;
1266 		break;
1267 
1268 	case PRU_DISCONNECT:
1269 		/*soisdisconnected(so);*/
1270 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1271 		in_pcbdisconnect(inp);
1272 		inp->inp_laddr = zeroin_addr;		/* XXX */
1273 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1274 		break;
1275 
1276 	case PRU_SHUTDOWN:
1277 		socantsendmore(so);
1278 		break;
1279 
1280 	case PRU_RCVD:
1281 		error = EOPNOTSUPP;
1282 		break;
1283 
1284 	case PRU_SEND:
1285 		if (control && control->m_len) {
1286 			m_freem(control);
1287 			m_freem(m);
1288 			error = EINVAL;
1289 			break;
1290 		}
1291 	{
1292 		struct in_addr laddr;			/* XXX */
1293 
1294 		memset(&laddr, 0, sizeof laddr);
1295 		if (nam) {
1296 			laddr = inp->inp_laddr;		/* XXX */
1297 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1298 				error = EISCONN;
1299 				goto die;
1300 			}
1301 			error = in_pcbconnect(inp, nam, l);
1302 			if (error)
1303 				goto die;
1304 		} else {
1305 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1306 				error = ENOTCONN;
1307 				goto die;
1308 			}
1309 		}
1310 		error = udp_output(m, inp);
1311 		m = NULL;
1312 		if (nam) {
1313 			in_pcbdisconnect(inp);
1314 			inp->inp_laddr = laddr;		/* XXX */
1315 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1316 		}
1317 	  die:
1318 		if (m)
1319 			m_freem(m);
1320 	}
1321 		break;
1322 
1323 	case PRU_SENSE:
1324 		/*
1325 		 * stat: don't bother with a blocksize.
1326 		 */
1327 		splx(s);
1328 		return (0);
1329 
1330 	case PRU_RCVOOB:
1331 		error =  EOPNOTSUPP;
1332 		break;
1333 
1334 	case PRU_SENDOOB:
1335 		m_freem(control);
1336 		m_freem(m);
1337 		error =  EOPNOTSUPP;
1338 		break;
1339 
1340 	case PRU_SOCKADDR:
1341 		in_setsockaddr(inp, nam);
1342 		break;
1343 
1344 	case PRU_PEERADDR:
1345 		in_setpeeraddr(inp, nam);
1346 		break;
1347 
1348 	default:
1349 		panic("udp_usrreq");
1350 	}
1351 
1352 release:
1353 	splx(s);
1354 	return (error);
1355 }
1356 
1357 static int
1358 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1359 {
1360 
1361 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1362 }
1363 
1364 /*
1365  * Sysctl for udp variables.
1366  */
1367 static void
1368 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1369 {
1370 	sysctl_createv(clog, 0, NULL, NULL,
1371 		       CTLFLAG_PERMANENT,
1372 		       CTLTYPE_NODE, "net", NULL,
1373 		       NULL, 0, NULL, 0,
1374 		       CTL_NET, CTL_EOL);
1375 	sysctl_createv(clog, 0, NULL, NULL,
1376 		       CTLFLAG_PERMANENT,
1377 		       CTLTYPE_NODE, "inet", NULL,
1378 		       NULL, 0, NULL, 0,
1379 		       CTL_NET, PF_INET, CTL_EOL);
1380 	sysctl_createv(clog, 0, NULL, NULL,
1381 		       CTLFLAG_PERMANENT,
1382 		       CTLTYPE_NODE, "udp",
1383 		       SYSCTL_DESCR("UDPv4 related settings"),
1384 		       NULL, 0, NULL, 0,
1385 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1386 
1387 	sysctl_createv(clog, 0, NULL, NULL,
1388 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1389 		       CTLTYPE_INT, "checksum",
1390 		       SYSCTL_DESCR("Compute UDP checksums"),
1391 		       NULL, 0, &udpcksum, 0,
1392 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1393 		       CTL_EOL);
1394 	sysctl_createv(clog, 0, NULL, NULL,
1395 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1396 		       CTLTYPE_INT, "sendspace",
1397 		       SYSCTL_DESCR("Default UDP send buffer size"),
1398 		       NULL, 0, &udp_sendspace, 0,
1399 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1400 		       CTL_EOL);
1401 	sysctl_createv(clog, 0, NULL, NULL,
1402 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1403 		       CTLTYPE_INT, "recvspace",
1404 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1405 		       NULL, 0, &udp_recvspace, 0,
1406 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1407 		       CTL_EOL);
1408 	sysctl_createv(clog, 0, NULL, NULL,
1409 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1410 		       CTLTYPE_INT, "do_loopback_cksum",
1411 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1412 		       NULL, 0, &udp_do_loopback_cksum, 0,
1413 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1414 		       CTL_EOL);
1415 	sysctl_createv(clog, 0, NULL, NULL,
1416 		       CTLFLAG_PERMANENT,
1417 		       CTLTYPE_STRUCT, "pcblist",
1418 		       SYSCTL_DESCR("UDP protocol control block list"),
1419 		       sysctl_inpcblist, 0, &udbtable, 0,
1420 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1421 		       CTL_EOL);
1422 	sysctl_createv(clog, 0, NULL, NULL,
1423 		       CTLFLAG_PERMANENT,
1424 		       CTLTYPE_STRUCT, "stats",
1425 		       SYSCTL_DESCR("UDP statistics"),
1426 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1427 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1428 		       CTL_EOL);
1429 }
1430 #endif
1431 
1432 void
1433 udp_statinc(u_int stat)
1434 {
1435 
1436 	KASSERT(stat < UDP_NSTATS);
1437 	UDP_STATINC(stat);
1438 }
1439 
1440 #if (defined INET && defined IPSEC_NAT_T)
1441 /*
1442  * Returns:
1443  * 1 if the packet was processed
1444  * 0 if normal UDP processing should take place
1445  * -1 if an error occurent and m was freed
1446  */
1447 static int
1448 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1449     struct socket *so)
1450 {
1451 	size_t len;
1452 	void *data;
1453 	struct inpcb *inp;
1454 	size_t skip = 0;
1455 	size_t minlen;
1456 	size_t iphdrlen;
1457 	struct ip *ip;
1458 	struct mbuf *n;
1459 	struct m_tag *tag;
1460 	struct udphdr *udphdr;
1461 	u_int16_t sport, dport;
1462 	struct mbuf *m = *mp;
1463 
1464 	/*
1465 	 * Collapse the mbuf chain if the first mbuf is too short
1466 	 * The longest case is: UDP + non ESP marker + ESP
1467 	 */
1468 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1469 	if (minlen > m->m_pkthdr.len)
1470 		minlen = m->m_pkthdr.len;
1471 
1472 	if (m->m_len < minlen) {
1473 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1474 			printf("udp4_espinudp: m_pullup failed\n");
1475 			return -1;
1476 		}
1477 		m = *mp;
1478 	}
1479 
1480 	len = m->m_len - off;
1481 	data = mtod(m, char *) + off;
1482 	inp = sotoinpcb(so);
1483 
1484 	/* Ignore keepalive packets */
1485 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1486 		return 1;
1487 	}
1488 
1489 	/*
1490 	 * Check that the payload is long enough to hold
1491 	 * an ESP header and compute the length of encapsulation
1492 	 * header to remove
1493 	 */
1494 	if (inp->inp_flags & INP_ESPINUDP) {
1495 		u_int32_t *st = (u_int32_t *)data;
1496 
1497 		if ((len <= sizeof(struct esp)) || (*st == 0))
1498 			return 0; /* Normal UDP processing */
1499 
1500 		skip = sizeof(struct udphdr);
1501 	}
1502 
1503 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1504 		u_int32_t *st = (u_int32_t *)data;
1505 
1506 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1507 		    || ((st[0] | st[1]) != 0))
1508 			return 0; /* Normal UDP processing */
1509 
1510 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1511 	}
1512 
1513 	/*
1514 	 * Get the UDP ports. They are handled in network
1515 	 * order everywhere in IPSEC_NAT_T code.
1516 	 */
1517 	udphdr = (struct udphdr *)((char *)data - skip);
1518 	sport = udphdr->uh_sport;
1519 	dport = udphdr->uh_dport;
1520 
1521 	/*
1522 	 * Remove the UDP header (and possibly the non ESP marker)
1523 	 * IP header lendth is iphdrlen
1524 	 * Before:
1525 	 *   <--- off --->
1526 	 *   +----+------+-----+
1527 	 *   | IP |  UDP | ESP |
1528 	 *   +----+------+-----+
1529 	 *        <-skip->
1530 	 * After:
1531 	 *          +----+-----+
1532 	 *          | IP | ESP |
1533 	 *          +----+-----+
1534 	 *   <-skip->
1535 	 */
1536 	iphdrlen = off - sizeof(struct udphdr);
1537 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1538 	m_adj(m, skip);
1539 
1540 	ip = mtod(m, struct ip *);
1541 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1542 	ip->ip_p = IPPROTO_ESP;
1543 
1544 	/*
1545 	 * Copy the mbuf to avoid multiple free, as both
1546 	 * esp4_input (which we call) and udp_input (which
1547 	 * called us) free the mbuf.
1548 	 */
1549 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1550 		printf("udp4_espinudp: m_dup failed\n");
1551 		return 0;
1552 	}
1553 
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(n);
1564 		return 0;
1565 	}
1566 	((u_int16_t *)(tag + 1))[0] = sport;
1567 	((u_int16_t *)(tag + 1))[1] = dport;
1568 	m_tag_prepend(n, tag);
1569 
1570 #ifdef FAST_IPSEC
1571 	ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1572 #else
1573 	esp4_input(n, iphdrlen);
1574 #endif
1575 
1576 	/* We handled it, it shouldn't be handled by UDP */
1577 	return 1;
1578 }
1579 #endif
1580