xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision b7ae68fde0d8ef1c03714e8bbb1ee7c6118ea93b)
1 /*	$NetBSD: udp_usrreq.c,v 1.148 2006/07/23 22:06:13 ad 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 1.148 2006/07/23 22:06:13 ad Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_ipsec.h"
68 #include "opt_inet_csum.h"
69 #include "opt_ipkdb.h"
70 #include "opt_mbuftrace.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/stat.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_icmp.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/udp6_var.h>
104 #include <netinet6/scope6_var.h>
105 #endif
106 
107 #ifndef INET6
108 /* always need ip6.h for IP6_EXTHDR_GET */
109 #include <netinet/ip6.h>
110 #endif
111 
112 #include "faith.h"
113 #if defined(NFAITH) && NFAITH > 0
114 #include <net/if_faith.h>
115 #endif
116 
117 #include <machine/stdarg.h>
118 
119 #ifdef FAST_IPSEC
120 #include <netipsec/ipsec.h>
121 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
122 #include <netipsec/esp.h>
123 #ifdef INET6
124 #include <netipsec/ipsec6.h>
125 #endif
126 #endif	/* FAST_IPSEC*/
127 
128 #ifdef IPSEC
129 #include <netinet6/ipsec.h>
130 #include <netinet6/esp.h>
131 #include <netkey/key.h>
132 #endif /*IPSEC*/
133 
134 #ifdef IPKDB
135 #include <ipkdb/ipkdb.h>
136 #endif
137 
138 /*
139  * UDP protocol implementation.
140  * Per RFC 768, August, 1980.
141  */
142 #ifndef	COMPAT_42
143 int	udpcksum = 1;
144 #else
145 int	udpcksum = 0;		/* XXX */
146 #endif
147 int	udp_do_loopback_cksum = 0;
148 
149 struct	inpcbtable udbtable;
150 struct	udpstat udpstat;
151 
152 #ifdef INET
153 #ifdef IPSEC_NAT_T
154 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
155 	struct socket *);
156 #endif
157 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
158 	struct socket *);
159 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
160 	struct mbuf **, int);
161 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
162 #endif
163 #ifdef INET6
164 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
165 	struct socket *);
166 static int udp6_realinput (int, struct sockaddr_in6 *,
167 	struct sockaddr_in6 *, struct mbuf *, int);
168 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
169 #endif
170 #ifdef INET
171 static	void udp_notify (struct inpcb *, int);
172 #endif
173 
174 #ifndef UDBHASHSIZE
175 #define	UDBHASHSIZE	128
176 #endif
177 int	udbhashsize = UDBHASHSIZE;
178 
179 #ifdef MBUFTRACE
180 struct mowner udp_mowner = { "udp" };
181 struct mowner udp_rx_mowner = { "udp", "rx" };
182 struct mowner udp_tx_mowner = { "udp", "tx" };
183 #endif
184 
185 #ifdef UDP_CSUM_COUNTERS
186 #include <sys/device.h>
187 
188 #if defined(INET)
189 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
190     NULL, "udp", "hwcsum bad");
191 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192     NULL, "udp", "hwcsum ok");
193 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
194     NULL, "udp", "hwcsum data");
195 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
196     NULL, "udp", "swcsum");
197 
198 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
199 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
200 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
201 EVCNT_ATTACH_STATIC(udp_swcsum);
202 #endif /* defined(INET) */
203 
204 #if defined(INET6)
205 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
206     NULL, "udp6", "hwcsum bad");
207 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208     NULL, "udp6", "hwcsum ok");
209 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
210     NULL, "udp6", "hwcsum data");
211 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
212     NULL, "udp6", "swcsum");
213 
214 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
215 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
216 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
217 EVCNT_ATTACH_STATIC(udp6_swcsum);
218 #endif /* defined(INET6) */
219 
220 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
221 
222 #else
223 
224 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
225 
226 #endif /* UDP_CSUM_COUNTERS */
227 
228 void
229 udp_init(void)
230 {
231 
232 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
233 
234 	MOWNER_ATTACH(&udp_tx_mowner);
235 	MOWNER_ATTACH(&udp_rx_mowner);
236 	MOWNER_ATTACH(&udp_mowner);
237 }
238 
239 /*
240  * Checksum extended UDP header and data.
241  */
242 
243 int
244 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
245     int iphlen, int len)
246 {
247 
248 	switch (af) {
249 #ifdef INET
250 	case AF_INET:
251 		return udp4_input_checksum(m, uh, iphlen, len);
252 #endif
253 #ifdef INET6
254 	case AF_INET6:
255 		return udp6_input_checksum(m, uh, iphlen, len);
256 #endif
257 	}
258 #ifdef DIAGNOSTIC
259 	panic("udp_input_checksum: unknown af %d", af);
260 #endif
261 	/* NOTREACHED */
262 	return -1;
263 }
264 
265 #ifdef INET
266 
267 /*
268  * Checksum extended UDP header and data.
269  */
270 
271 static int
272 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
273     int iphlen, int len)
274 {
275 
276 	/*
277 	 * XXX it's better to record and check if this mbuf is
278 	 * already checked.
279 	 */
280 
281 	if (uh->uh_sum == 0)
282 		return 0;
283 
284 	switch (m->m_pkthdr.csum_flags &
285 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
286 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
287 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
288 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
289 		goto badcsum;
290 
291 	case M_CSUM_UDPv4|M_CSUM_DATA: {
292 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
293 
294 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
295 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
296 			const struct ip *ip =
297 			    mtod(m, const struct ip *);
298 
299 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
300 			    ip->ip_dst.s_addr,
301 			    htons(hw_csum + len + IPPROTO_UDP));
302 		}
303 		if ((hw_csum ^ 0xffff) != 0)
304 			goto badcsum;
305 		break;
306 	}
307 
308 	case M_CSUM_UDPv4:
309 		/* Checksum was okay. */
310 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
311 		break;
312 
313 	default:
314 		/*
315 		 * Need to compute it ourselves.  Maybe skip checksum
316 		 * on loopback interfaces.
317 		 */
318 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
319 				     IFF_LOOPBACK) ||
320 				   udp_do_loopback_cksum)) {
321 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
322 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
323 				goto badcsum;
324 		}
325 		break;
326 	}
327 
328 	return 0;
329 
330 badcsum:
331 	udpstat.udps_badsum++;
332 	return -1;
333 }
334 
335 void
336 udp_input(struct mbuf *m, ...)
337 {
338 	va_list ap;
339 	struct sockaddr_in src, dst;
340 	struct ip *ip;
341 	struct udphdr *uh;
342 	int iphlen;
343 	int len;
344 	int n;
345 	u_int16_t ip_len;
346 
347 	va_start(ap, m);
348 	iphlen = va_arg(ap, int);
349 	(void)va_arg(ap, int);		/* ignore value, advance ap */
350 	va_end(ap);
351 
352 	MCLAIM(m, &udp_rx_mowner);
353 	udpstat.udps_ipackets++;
354 
355 	/*
356 	 * Get IP and UDP header together in first mbuf.
357 	 */
358 	ip = mtod(m, struct ip *);
359 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
360 	if (uh == NULL) {
361 		udpstat.udps_hdrops++;
362 		return;
363 	}
364 	KASSERT(UDP_HDR_ALIGNED_P(uh));
365 
366 	/* destination port of 0 is illegal, based on RFC768. */
367 	if (uh->uh_dport == 0)
368 		goto bad;
369 
370 	/*
371 	 * Make mbuf data length reflect UDP length.
372 	 * If not enough data to reflect UDP length, drop.
373 	 */
374 	ip_len = ntohs(ip->ip_len);
375 	len = ntohs((u_int16_t)uh->uh_ulen);
376 	if (ip_len != iphlen + len) {
377 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
378 			udpstat.udps_badlen++;
379 			goto bad;
380 		}
381 		m_adj(m, iphlen + len - ip_len);
382 	}
383 
384 	/*
385 	 * Checksum extended UDP header and data.
386 	 */
387 	if (udp4_input_checksum(m, uh, iphlen, len))
388 		goto badcsum;
389 
390 	/* construct source and dst sockaddrs. */
391 	bzero(&src, sizeof(src));
392 	src.sin_family = AF_INET;
393 	src.sin_len = sizeof(struct sockaddr_in);
394 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
395 	src.sin_port = uh->uh_sport;
396 	bzero(&dst, sizeof(dst));
397 	dst.sin_family = AF_INET;
398 	dst.sin_len = sizeof(struct sockaddr_in);
399 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
400 	dst.sin_port = uh->uh_dport;
401 
402 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
403 		udpstat.udps_hdrops++;
404 		return;
405 	}
406 #ifdef INET6
407 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
408 		struct sockaddr_in6 src6, dst6;
409 
410 		bzero(&src6, sizeof(src6));
411 		src6.sin6_family = AF_INET6;
412 		src6.sin6_len = sizeof(struct sockaddr_in6);
413 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
414 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
415 			sizeof(ip->ip_src));
416 		src6.sin6_port = uh->uh_sport;
417 		bzero(&dst6, sizeof(dst6));
418 		dst6.sin6_family = AF_INET6;
419 		dst6.sin6_len = sizeof(struct sockaddr_in6);
420 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
421 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
422 			sizeof(ip->ip_dst));
423 		dst6.sin6_port = uh->uh_dport;
424 
425 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
426 	}
427 #endif
428 
429 	if (n == 0) {
430 		if (m->m_flags & (M_BCAST | M_MCAST)) {
431 			udpstat.udps_noportbcast++;
432 			goto bad;
433 		}
434 		udpstat.udps_noport++;
435 #ifdef IPKDB
436 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
437 				m, iphlen + sizeof(struct udphdr),
438 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
439 			/*
440 			 * It was a debugger connect packet,
441 			 * just drop it now
442 			 */
443 			goto bad;
444 		}
445 #endif
446 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
447 		m = NULL;
448 	}
449 
450 bad:
451 	if (m)
452 		m_freem(m);
453 	return;
454 
455 badcsum:
456 	m_freem(m);
457 }
458 #endif
459 
460 #ifdef INET6
461 static int
462 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
463 {
464 
465 	/*
466 	 * XXX it's better to record and check if this mbuf is
467 	 * already checked.
468 	 */
469 
470 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
471 		goto good;
472 	}
473 	if (uh->uh_sum == 0) {
474 		udp6stat.udp6s_nosum++;
475 		goto bad;
476 	}
477 
478 	switch (m->m_pkthdr.csum_flags &
479 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
480 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
481 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
482 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
483 		udp6stat.udp6s_badsum++;
484 		goto bad;
485 
486 #if 0 /* notyet */
487 	case M_CSUM_UDPv6|M_CSUM_DATA:
488 #endif
489 
490 	case M_CSUM_UDPv6:
491 		/* Checksum was okay. */
492 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
493 		break;
494 
495 	default:
496 		/*
497 		 * Need to compute it ourselves.  Maybe skip checksum
498 		 * on loopback interfaces.
499 		 */
500 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
501 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
502 			udp6stat.udp6s_badsum++;
503 			goto bad;
504 		}
505 	}
506 
507 good:
508 	return 0;
509 bad:
510 	return -1;
511 }
512 
513 int
514 udp6_input(struct mbuf **mp, int *offp, int proto)
515 {
516 	struct mbuf *m = *mp;
517 	int off = *offp;
518 	struct sockaddr_in6 src, dst;
519 	struct ip6_hdr *ip6;
520 	struct udphdr *uh;
521 	u_int32_t plen, ulen;
522 
523 	ip6 = mtod(m, struct ip6_hdr *);
524 
525 #if defined(NFAITH) && 0 < NFAITH
526 	if (faithprefix(&ip6->ip6_dst)) {
527 		/* send icmp6 host unreach? */
528 		m_freem(m);
529 		return IPPROTO_DONE;
530 	}
531 #endif
532 
533 	udp6stat.udp6s_ipackets++;
534 
535 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
536 	plen = m->m_pkthdr.len - off;
537 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
538 	if (uh == NULL) {
539 		ip6stat.ip6s_tooshort++;
540 		return IPPROTO_DONE;
541 	}
542 	KASSERT(UDP_HDR_ALIGNED_P(uh));
543 	ulen = ntohs((u_short)uh->uh_ulen);
544 	/*
545 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
546 	 * iff payload length > 0xffff.
547 	 */
548 	if (ulen == 0 && plen > 0xffff)
549 		ulen = plen;
550 
551 	if (plen != ulen) {
552 		udp6stat.udp6s_badlen++;
553 		goto bad;
554 	}
555 
556 	/* destination port of 0 is illegal, based on RFC768. */
557 	if (uh->uh_dport == 0)
558 		goto bad;
559 
560 	/* Be proactive about malicious use of IPv4 mapped address */
561 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
562 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
563 		/* XXX stat */
564 		goto bad;
565 	}
566 
567 	/*
568 	 * Checksum extended UDP header and data.  Maybe skip checksum
569 	 * on loopback interfaces.
570 	 */
571 	if (udp6_input_checksum(m, uh, off, ulen))
572 		goto bad;
573 
574 	/*
575 	 * Construct source and dst sockaddrs.
576 	 */
577 	bzero(&src, sizeof(src));
578 	src.sin6_family = AF_INET6;
579 	src.sin6_len = sizeof(struct sockaddr_in6);
580 	src.sin6_addr = ip6->ip6_src;
581 	src.sin6_port = uh->uh_sport;
582 	bzero(&dst, sizeof(dst));
583 	dst.sin6_family = AF_INET6;
584 	dst.sin6_len = sizeof(struct sockaddr_in6);
585 	dst.sin6_addr = ip6->ip6_dst;
586 	dst.sin6_port = uh->uh_dport;
587 
588 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
589 		if (m->m_flags & M_MCAST) {
590 			udp6stat.udp6s_noportmcast++;
591 			goto bad;
592 		}
593 		udp6stat.udp6s_noport++;
594 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
595 		m = NULL;
596 	}
597 
598 bad:
599 	if (m)
600 		m_freem(m);
601 	return IPPROTO_DONE;
602 }
603 #endif
604 
605 #ifdef INET
606 static void
607 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
608 	struct sockaddr *src, struct socket *so)
609 {
610 	struct mbuf *opts = NULL;
611 	struct mbuf *n;
612 	struct inpcb *inp = NULL;
613 
614 	if (!so)
615 		return;
616 	switch (so->so_proto->pr_domain->dom_family) {
617 	case AF_INET:
618 		inp = sotoinpcb(so);
619 		break;
620 #ifdef INET6
621 	case AF_INET6:
622 		break;
623 #endif
624 	default:
625 		return;
626 	}
627 
628 #if defined(IPSEC) || defined(FAST_IPSEC)
629 	/* check AH/ESP integrity. */
630 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
631 		ipsecstat.in_polvio++;
632 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
633 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
634 			    0, 0);
635 		return;
636 	}
637 #endif /*IPSEC*/
638 
639 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
640 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
641 			 || so->so_options & SO_TIMESTAMP)) {
642 			struct ip *ip = mtod(n, struct ip *);
643 			ip_savecontrol(inp, &opts, ip, n);
644 		}
645 
646 		m_adj(n, off);
647 		if (sbappendaddr(&so->so_rcv, src, n,
648 				opts) == 0) {
649 			m_freem(n);
650 			if (opts)
651 				m_freem(opts);
652 			so->so_rcv.sb_overflowed++;
653 			udpstat.udps_fullsock++;
654 		} else
655 			sorwakeup(so);
656 	}
657 }
658 #endif
659 
660 #ifdef INET6
661 static void
662 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
663 	struct sockaddr *src, struct socket *so)
664 {
665 	struct mbuf *opts = NULL;
666 	struct mbuf *n;
667 	struct in6pcb *in6p = NULL;
668 
669 	if (!so)
670 		return;
671 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
672 		return;
673 	in6p = sotoin6pcb(so);
674 
675 #if defined(IPSEC) || defined(FAST_IPSEC)
676 	/* check AH/ESP integrity. */
677 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
678 		ipsec6stat.in_polvio++;
679 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
680 			icmp6_error(n, ICMP6_DST_UNREACH,
681 			    ICMP6_DST_UNREACH_ADMIN, 0);
682 		return;
683 	}
684 #endif /*IPSEC*/
685 
686 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
687 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
688 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
689 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
690 			ip6_savecontrol(in6p, &opts, ip6, n);
691 		}
692 
693 		m_adj(n, off);
694 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
695 			m_freem(n);
696 			if (opts)
697 				m_freem(opts);
698 			so->so_rcv.sb_overflowed++;
699 			udp6stat.udp6s_fullsock++;
700 		} else
701 			sorwakeup(so);
702 	}
703 }
704 #endif
705 
706 #ifdef INET
707 static int
708 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
709 	struct mbuf **mp, int off /* offset of udphdr */)
710 {
711 	u_int16_t *sport, *dport;
712 	int rcvcnt;
713 	struct in_addr *src4, *dst4;
714 	struct inpcb_hdr *inph;
715 	struct inpcb *inp;
716 	struct mbuf *m = *mp;
717 
718 	rcvcnt = 0;
719 	off += sizeof(struct udphdr);	/* now, offset of payload */
720 
721 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
722 		goto bad;
723 
724 	src4 = &src->sin_addr;
725 	sport = &src->sin_port;
726 	dst4 = &dst->sin_addr;
727 	dport = &dst->sin_port;
728 
729 	if (IN_MULTICAST(dst4->s_addr) ||
730 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
731 		/*
732 		 * Deliver a multicast or broadcast datagram to *all* sockets
733 		 * for which the local and remote addresses and ports match
734 		 * those of the incoming datagram.  This allows more than
735 		 * one process to receive multi/broadcasts on the same port.
736 		 * (This really ought to be done for unicast datagrams as
737 		 * well, but that would cause problems with existing
738 		 * applications that open both address-specific sockets and
739 		 * a wildcard socket listening to the same port -- they would
740 		 * end up receiving duplicates of every unicast datagram.
741 		 * Those applications open the multiple sockets to overcome an
742 		 * inadequacy of the UDP socket interface, but for backwards
743 		 * compatibility we avoid the problem here rather than
744 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
745 		 */
746 
747 		/*
748 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
749 		 * we need udpiphdr for IPsec processing so we do that later.
750 		 */
751 		/*
752 		 * Locate pcb(s) for datagram.
753 		 */
754 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
755 			inp = (struct inpcb *)inph;
756 			if (inp->inp_af != AF_INET)
757 				continue;
758 
759 			if (inp->inp_lport != *dport)
760 				continue;
761 			if (!in_nullhost(inp->inp_laddr)) {
762 				if (!in_hosteq(inp->inp_laddr, *dst4))
763 					continue;
764 			}
765 			if (!in_nullhost(inp->inp_faddr)) {
766 				if (!in_hosteq(inp->inp_faddr, *src4) ||
767 				    inp->inp_fport != *sport)
768 					continue;
769 			}
770 
771 			udp4_sendup(m, off, (struct sockaddr *)src,
772 				inp->inp_socket);
773 			rcvcnt++;
774 
775 			/*
776 			 * Don't look for additional matches if this one does
777 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
778 			 * socket options set.  This heuristic avoids searching
779 			 * through all pcbs in the common case of a non-shared
780 			 * port.  It assumes that an application will never
781 			 * clear these options after setting them.
782 			 */
783 			if ((inp->inp_socket->so_options &
784 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
785 				break;
786 		}
787 	} else {
788 		/*
789 		 * Locate pcb for datagram.
790 		 */
791 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
792 		if (inp == 0) {
793 			++udpstat.udps_pcbhashmiss;
794 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
795 			if (inp == 0)
796 				return rcvcnt;
797 		}
798 
799 #ifdef IPSEC_NAT_T
800 		/* Handle ESP over UDP */
801 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
802 			struct sockaddr *sa = (struct sockaddr *)src;
803 
804 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
805 			case -1: 	/* Error, m was freeed */
806 				rcvcnt = -1;
807 				goto bad;
808 				break;
809 
810 			case 1:		/* ESP over UDP */
811 				rcvcnt++;
812 				goto bad;
813 				break;
814 
815 			case 0: 	/* plain UDP */
816 			default: 	/* Unexpected */
817 				/*
818 				 * Normal UDP processing will take place
819 				 * m may have changed.
820 				 */
821 				m = *mp;
822 				break;
823 			}
824 		}
825 #endif
826 
827 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
828 		rcvcnt++;
829 	}
830 
831 bad:
832 	return rcvcnt;
833 }
834 #endif
835 
836 #ifdef INET6
837 static int
838 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
839 	struct mbuf *m, int off)
840 {
841 	u_int16_t sport, dport;
842 	int rcvcnt;
843 	struct in6_addr src6, *dst6;
844 	const struct in_addr *dst4;
845 	struct inpcb_hdr *inph;
846 	struct in6pcb *in6p;
847 
848 	rcvcnt = 0;
849 	off += sizeof(struct udphdr);	/* now, offset of payload */
850 
851 	if (af != AF_INET && af != AF_INET6)
852 		goto bad;
853 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
854 		goto bad;
855 
856 	src6 = src->sin6_addr;
857 	if (sa6_recoverscope(src) != 0) {
858 		/* XXX: should be impossible. */
859 		goto bad;
860 	}
861 	sport = src->sin6_port;
862 
863 	dport = dst->sin6_port;
864 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
865 	dst6 = &dst->sin6_addr;
866 
867 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
868 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
869 		/*
870 		 * Deliver a multicast or broadcast datagram to *all* sockets
871 		 * for which the local and remote addresses and ports match
872 		 * those of the incoming datagram.  This allows more than
873 		 * one process to receive multi/broadcasts on the same port.
874 		 * (This really ought to be done for unicast datagrams as
875 		 * well, but that would cause problems with existing
876 		 * applications that open both address-specific sockets and
877 		 * a wildcard socket listening to the same port -- they would
878 		 * end up receiving duplicates of every unicast datagram.
879 		 * Those applications open the multiple sockets to overcome an
880 		 * inadequacy of the UDP socket interface, but for backwards
881 		 * compatibility we avoid the problem here rather than
882 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
883 		 */
884 
885 		/*
886 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
887 		 * we need udpiphdr for IPsec processing so we do that later.
888 		 */
889 		/*
890 		 * Locate pcb(s) for datagram.
891 		 */
892 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
893 			in6p = (struct in6pcb *)inph;
894 			if (in6p->in6p_af != AF_INET6)
895 				continue;
896 
897 			if (in6p->in6p_lport != dport)
898 				continue;
899 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
900 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
901 				    dst6))
902 					continue;
903 			} else {
904 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
905 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
906 					continue;
907 			}
908 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
909 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
910 				    &src6) || in6p->in6p_fport != sport)
911 					continue;
912 			} else {
913 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
914 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
915 					continue;
916 			}
917 
918 			udp6_sendup(m, off, (struct sockaddr *)src,
919 				in6p->in6p_socket);
920 			rcvcnt++;
921 
922 			/*
923 			 * Don't look for additional matches if this one does
924 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
925 			 * socket options set.  This heuristic avoids searching
926 			 * through all pcbs in the common case of a non-shared
927 			 * port.  It assumes that an application will never
928 			 * clear these options after setting them.
929 			 */
930 			if ((in6p->in6p_socket->so_options &
931 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
932 				break;
933 		}
934 	} else {
935 		/*
936 		 * Locate pcb for datagram.
937 		 */
938 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
939 		    dport, 0);
940 		if (in6p == 0) {
941 			++udpstat.udps_pcbhashmiss;
942 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
943 			if (in6p == 0)
944 				return rcvcnt;
945 		}
946 
947 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
948 		rcvcnt++;
949 	}
950 
951 bad:
952 	return rcvcnt;
953 }
954 #endif
955 
956 #ifdef INET
957 /*
958  * Notify a udp user of an asynchronous error;
959  * just wake up so that he can collect error status.
960  */
961 static void
962 udp_notify(struct inpcb *inp, int errno)
963 {
964 	inp->inp_socket->so_error = errno;
965 	sorwakeup(inp->inp_socket);
966 	sowwakeup(inp->inp_socket);
967 }
968 
969 void *
970 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
971 {
972 	struct ip *ip = v;
973 	struct udphdr *uh;
974 	void (*notify)(struct inpcb *, int) = udp_notify;
975 	int errno;
976 
977 	if (sa->sa_family != AF_INET
978 	 || sa->sa_len != sizeof(struct sockaddr_in))
979 		return NULL;
980 	if ((unsigned)cmd >= PRC_NCMDS)
981 		return NULL;
982 	errno = inetctlerrmap[cmd];
983 	if (PRC_IS_REDIRECT(cmd))
984 		notify = in_rtchange, ip = 0;
985 	else if (cmd == PRC_HOSTDEAD)
986 		ip = 0;
987 	else if (errno == 0)
988 		return NULL;
989 	if (ip) {
990 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
991 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
992 		    ip->ip_src, uh->uh_sport, errno, notify);
993 
994 		/* XXX mapped address case */
995 	} else
996 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
997 		    notify);
998 	return NULL;
999 }
1000 
1001 int
1002 udp_ctloutput(op, so, level, optname, mp)
1003 	int op;
1004 	struct socket *so;
1005 	int level, optname;
1006 	struct mbuf **mp;
1007 {
1008 	int s;
1009 	int error = 0;
1010 	struct mbuf *m;
1011 	struct inpcb *inp;
1012 	int family;
1013 
1014 	family = so->so_proto->pr_domain->dom_family;
1015 
1016 	s = splsoftnet();
1017 	switch (family) {
1018 #ifdef INET
1019 	case PF_INET:
1020 		if (level != IPPROTO_UDP) {
1021 			error = ip_ctloutput(op, so, level, optname, mp);
1022 			goto end;
1023 		}
1024 		break;
1025 #endif
1026 #ifdef INET6
1027 	case PF_INET6:
1028 		if (level != IPPROTO_UDP) {
1029 			error = ip6_ctloutput(op, so, level, optname, mp);
1030 			goto end;
1031 		}
1032 		break;
1033 #endif
1034 	default:
1035 		error = EAFNOSUPPORT;
1036 		goto end;
1037 		break;
1038 	}
1039 
1040 
1041 	switch (op) {
1042 	case PRCO_SETOPT:
1043 		m = *mp;
1044 		inp = sotoinpcb(so);
1045 
1046 		switch (optname) {
1047 		case UDP_ENCAP:
1048 			if (m == NULL || m->m_len < sizeof (int)) {
1049 				error = EINVAL;
1050 				goto end;
1051 			}
1052 
1053 			switch(*mtod(m, int *)) {
1054 #ifdef IPSEC_NAT_T
1055 			case 0:
1056 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1057 				break;
1058 
1059 			case UDP_ENCAP_ESPINUDP:
1060 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1061 				inp->inp_flags |= INP_ESPINUDP;
1062 				break;
1063 
1064 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1065 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1066 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1067 				break;
1068 #endif
1069 			default:
1070 				error = EINVAL;
1071 				goto end;
1072 				break;
1073 			}
1074 			break;
1075 
1076 		default:
1077 			error = ENOPROTOOPT;
1078 			goto end;
1079 			break;
1080 		}
1081 		break;
1082 
1083 	default:
1084 		error = EINVAL;
1085 		goto end;
1086 		break;
1087 	}
1088 
1089 end:
1090 	splx(s);
1091 	return error;
1092 }
1093 
1094 
1095 int
1096 udp_output(struct mbuf *m, ...)
1097 {
1098 	struct inpcb *inp;
1099 	struct udpiphdr *ui;
1100 	struct route *ro;
1101 	int len = m->m_pkthdr.len;
1102 	int error = 0;
1103 	va_list ap;
1104 
1105 	MCLAIM(m, &udp_tx_mowner);
1106 	va_start(ap, m);
1107 	inp = va_arg(ap, struct inpcb *);
1108 	va_end(ap);
1109 
1110 	/*
1111 	 * Calculate data length and get a mbuf
1112 	 * for UDP and IP headers.
1113 	 */
1114 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1115 	if (m == 0) {
1116 		error = ENOBUFS;
1117 		goto release;
1118 	}
1119 
1120 	/*
1121 	 * Compute the packet length of the IP header, and
1122 	 * punt if the length looks bogus.
1123 	 */
1124 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1125 		error = EMSGSIZE;
1126 		goto release;
1127 	}
1128 
1129 	/*
1130 	 * Fill in mbuf with extended UDP header
1131 	 * and addresses and length put into network format.
1132 	 */
1133 	ui = mtod(m, struct udpiphdr *);
1134 	ui->ui_pr = IPPROTO_UDP;
1135 	ui->ui_src = inp->inp_laddr;
1136 	ui->ui_dst = inp->inp_faddr;
1137 	ui->ui_sport = inp->inp_lport;
1138 	ui->ui_dport = inp->inp_fport;
1139 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1140 
1141 	ro = &inp->inp_route;
1142 
1143 	/*
1144 	 * Set up checksum and output datagram.
1145 	 */
1146 	if (udpcksum) {
1147 		/*
1148 		 * XXX Cache pseudo-header checksum part for
1149 		 * XXX "connected" UDP sockets.
1150 		 */
1151 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1152 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1153 		    sizeof(struct udphdr) + IPPROTO_UDP));
1154 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1155 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1156 	} else
1157 		ui->ui_sum = 0;
1158 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1159 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1160 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1161 	udpstat.udps_opackets++;
1162 
1163 	return (ip_output(m, inp->inp_options, ro,
1164 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1165 	    inp->inp_moptions, inp->inp_socket));
1166 
1167 release:
1168 	m_freem(m);
1169 	return (error);
1170 }
1171 
1172 int	udp_sendspace = 9216;		/* really max datagram size */
1173 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1174 					/* 40 1K datagrams */
1175 
1176 /*ARGSUSED*/
1177 int
1178 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1179 	struct mbuf *control, struct lwp *l)
1180 {
1181 	struct inpcb *inp;
1182 	int s;
1183 	int error = 0;
1184 
1185 	if (req == PRU_CONTROL)
1186 		return (in_control(so, (long)m, (caddr_t)nam,
1187 		    (struct ifnet *)control, l));
1188 
1189 	if (req == PRU_PURGEIF) {
1190 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1191 		in_purgeif((struct ifnet *)control);
1192 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1193 		return (0);
1194 	}
1195 
1196 	s = splsoftnet();
1197 	inp = sotoinpcb(so);
1198 #ifdef DIAGNOSTIC
1199 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1200 		panic("udp_usrreq: unexpected control mbuf");
1201 #endif
1202 	if (inp == 0 && req != PRU_ATTACH) {
1203 		error = EINVAL;
1204 		goto release;
1205 	}
1206 
1207 	/*
1208 	 * Note: need to block udp_input while changing
1209 	 * the udp pcb queue and/or pcb addresses.
1210 	 */
1211 	switch (req) {
1212 
1213 	case PRU_ATTACH:
1214 		if (inp != 0) {
1215 			error = EISCONN;
1216 			break;
1217 		}
1218 #ifdef MBUFTRACE
1219 		so->so_mowner = &udp_mowner;
1220 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1221 		so->so_snd.sb_mowner = &udp_tx_mowner;
1222 #endif
1223 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1224 			error = soreserve(so, udp_sendspace, udp_recvspace);
1225 			if (error)
1226 				break;
1227 		}
1228 		error = in_pcballoc(so, &udbtable);
1229 		if (error)
1230 			break;
1231 		inp = sotoinpcb(so);
1232 		inp->inp_ip.ip_ttl = ip_defttl;
1233 		break;
1234 
1235 	case PRU_DETACH:
1236 		in_pcbdetach(inp);
1237 		break;
1238 
1239 	case PRU_BIND:
1240 		error = in_pcbbind(inp, nam, l);
1241 		break;
1242 
1243 	case PRU_LISTEN:
1244 		error = EOPNOTSUPP;
1245 		break;
1246 
1247 	case PRU_CONNECT:
1248 		error = in_pcbconnect(inp, nam, l);
1249 		if (error)
1250 			break;
1251 		soisconnected(so);
1252 		break;
1253 
1254 	case PRU_CONNECT2:
1255 		error = EOPNOTSUPP;
1256 		break;
1257 
1258 	case PRU_DISCONNECT:
1259 		/*soisdisconnected(so);*/
1260 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1261 		in_pcbdisconnect(inp);
1262 		inp->inp_laddr = zeroin_addr;		/* XXX */
1263 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1264 		break;
1265 
1266 	case PRU_SHUTDOWN:
1267 		socantsendmore(so);
1268 		break;
1269 
1270 	case PRU_RCVD:
1271 		error = EOPNOTSUPP;
1272 		break;
1273 
1274 	case PRU_SEND:
1275 		if (control && control->m_len) {
1276 			m_freem(control);
1277 			m_freem(m);
1278 			error = EINVAL;
1279 			break;
1280 		}
1281 	{
1282 		struct in_addr laddr;			/* XXX */
1283 
1284 		if (nam) {
1285 			laddr = inp->inp_laddr;		/* XXX */
1286 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1287 				error = EISCONN;
1288 				goto die;
1289 			}
1290 			error = in_pcbconnect(inp, nam, l);
1291 			if (error)
1292 				goto die;
1293 		} else {
1294 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1295 				error = ENOTCONN;
1296 				goto die;
1297 			}
1298 		}
1299 		error = udp_output(m, inp);
1300 		m = NULL;
1301 		if (nam) {
1302 			in_pcbdisconnect(inp);
1303 			inp->inp_laddr = laddr;		/* XXX */
1304 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1305 		}
1306 	  die:
1307 		if (m)
1308 			m_freem(m);
1309 	}
1310 		break;
1311 
1312 	case PRU_SENSE:
1313 		/*
1314 		 * stat: don't bother with a blocksize.
1315 		 */
1316 		splx(s);
1317 		return (0);
1318 
1319 	case PRU_RCVOOB:
1320 		error =  EOPNOTSUPP;
1321 		break;
1322 
1323 	case PRU_SENDOOB:
1324 		m_freem(control);
1325 		m_freem(m);
1326 		error =  EOPNOTSUPP;
1327 		break;
1328 
1329 	case PRU_SOCKADDR:
1330 		in_setsockaddr(inp, nam);
1331 		break;
1332 
1333 	case PRU_PEERADDR:
1334 		in_setpeeraddr(inp, nam);
1335 		break;
1336 
1337 	default:
1338 		panic("udp_usrreq");
1339 	}
1340 
1341 release:
1342 	splx(s);
1343 	return (error);
1344 }
1345 
1346 /*
1347  * Sysctl for udp variables.
1348  */
1349 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1350 {
1351 
1352 	sysctl_createv(clog, 0, NULL, NULL,
1353 		       CTLFLAG_PERMANENT,
1354 		       CTLTYPE_NODE, "net", NULL,
1355 		       NULL, 0, NULL, 0,
1356 		       CTL_NET, CTL_EOL);
1357 	sysctl_createv(clog, 0, NULL, NULL,
1358 		       CTLFLAG_PERMANENT,
1359 		       CTLTYPE_NODE, "inet", NULL,
1360 		       NULL, 0, NULL, 0,
1361 		       CTL_NET, PF_INET, CTL_EOL);
1362 	sysctl_createv(clog, 0, NULL, NULL,
1363 		       CTLFLAG_PERMANENT,
1364 		       CTLTYPE_NODE, "udp",
1365 		       SYSCTL_DESCR("UDPv4 related settings"),
1366 		       NULL, 0, NULL, 0,
1367 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1368 
1369 	sysctl_createv(clog, 0, NULL, NULL,
1370 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1371 		       CTLTYPE_INT, "checksum",
1372 		       SYSCTL_DESCR("Compute UDP checksums"),
1373 		       NULL, 0, &udpcksum, 0,
1374 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1375 		       CTL_EOL);
1376 	sysctl_createv(clog, 0, NULL, NULL,
1377 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1378 		       CTLTYPE_INT, "sendspace",
1379 		       SYSCTL_DESCR("Default UDP send buffer size"),
1380 		       NULL, 0, &udp_sendspace, 0,
1381 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1382 		       CTL_EOL);
1383 	sysctl_createv(clog, 0, NULL, NULL,
1384 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1385 		       CTLTYPE_INT, "recvspace",
1386 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1387 		       NULL, 0, &udp_recvspace, 0,
1388 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1389 		       CTL_EOL);
1390 	sysctl_createv(clog, 0, NULL, NULL,
1391 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1392 		       CTLTYPE_INT, "do_loopback_cksum",
1393 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1394 		       NULL, 0, &udp_do_loopback_cksum, 0,
1395 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1396 		       CTL_EOL);
1397 	sysctl_createv(clog, 0, NULL, NULL,
1398 		       CTLFLAG_PERMANENT,
1399 		       CTLTYPE_STRUCT, "pcblist",
1400 		       SYSCTL_DESCR("UDP protocol control block list"),
1401 		       sysctl_inpcblist, 0, &udbtable, 0,
1402 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1403 		       CTL_EOL);
1404 	sysctl_createv(clog, 0, NULL, NULL,
1405 		       CTLFLAG_PERMANENT,
1406 		       CTLTYPE_STRUCT, "stats",
1407 		       SYSCTL_DESCR("UDP statistics"),
1408 		       NULL, 0, &udpstat, sizeof(udpstat),
1409 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1410 		       CTL_EOL);
1411 }
1412 #endif
1413 
1414 #if (defined INET && defined IPSEC_NAT_T)
1415 /*
1416  * Returns:
1417  * 1 if the packet was processed
1418  * 0 if normal UDP processing should take place
1419  * -1 if an error occurent and m was freed
1420  */
1421 static int
1422 udp4_espinudp(mp, off, src, so)
1423 	struct mbuf **mp;
1424 	int off;
1425 	struct sockaddr *src;
1426 	struct socket *so;
1427 {
1428 	size_t len;
1429 	caddr_t data;
1430 	struct inpcb *inp;
1431 	size_t skip = 0;
1432 	size_t minlen;
1433 	size_t iphdrlen;
1434 	struct ip *ip;
1435 	struct mbuf *n;
1436 	struct m_tag *tag;
1437 	struct udphdr *udphdr;
1438 	u_int16_t sport, dport;
1439 	struct mbuf *m = *mp;
1440 
1441 	/*
1442 	 * Collapse the mbuf chain if the first mbuf is too short
1443 	 * The longest case is: UDP + non ESP marker + ESP
1444 	 */
1445 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1446 	if (minlen > m->m_pkthdr.len)
1447 		minlen = m->m_pkthdr.len;
1448 
1449 	if (m->m_len < minlen) {
1450 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1451 			printf("udp4_espinudp: m_pullup failed\n");
1452 			return -1;
1453 		}
1454 		m = *mp;
1455 	}
1456 
1457 	len = m->m_len - off;
1458 	data = mtod(m, caddr_t) + off;
1459 	inp = sotoinpcb(so);
1460 
1461 	/* Ignore keepalive packets */
1462 	if ((len == 1) && (data[0] == '\xff')) {
1463 		return 1;
1464 	}
1465 
1466 	/*
1467 	 * Check that the payload is long enough to hold
1468 	 * an ESP header and compute the length of encapsulation
1469 	 * header to remove
1470 	 */
1471 	if (inp->inp_flags & INP_ESPINUDP) {
1472 		u_int32_t *st = (u_int32_t *)data;
1473 
1474 		if ((len <= sizeof(struct esp)) || (*st == 0))
1475 			return 0; /* Normal UDP processing */
1476 
1477 		skip = sizeof(struct udphdr);
1478 	}
1479 
1480 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1481 		u_int32_t *st = (u_int32_t *)data;
1482 
1483 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1484 		    || ((st[0] | st[1]) != 0))
1485 			return 0; /* Normal UDP processing */
1486 
1487 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1488 	}
1489 
1490 	/*
1491 	 * Get the UDP ports. They are handled in network
1492 	 * order everywhere in IPSEC_NAT_T code.
1493 	 */
1494 	udphdr = (struct udphdr *)(data - skip);
1495 	sport = udphdr->uh_sport;
1496 	dport = udphdr->uh_dport;
1497 
1498 	/*
1499 	 * Remove the UDP header (and possibly the non ESP marker)
1500 	 * IP header lendth is iphdrlen
1501 	 * Before:
1502 	 *   <--- off --->
1503 	 *   +----+------+-----+
1504 	 *   | IP |  UDP | ESP |
1505 	 *   +----+------+-----+
1506 	 *        <-skip->
1507 	 * After:
1508 	 *          +----+-----+
1509 	 *          | IP | ESP |
1510 	 *          +----+-----+
1511 	 *   <-skip->
1512 	 */
1513 	iphdrlen = off - sizeof(struct udphdr);
1514 	memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1515 	m_adj(m, skip);
1516 
1517 	ip = mtod(m, struct ip *);
1518 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1519 	ip->ip_p = IPPROTO_ESP;
1520 
1521 	/*
1522 	 * Copy the mbuf to avoid multiple free, as both
1523 	 * esp4_input (which we call) and udp_input (which
1524 	 * called us) free the mbuf.
1525 	 */
1526 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1527 		printf("udp4_espinudp: m_dup failed\n");
1528 		return 0;
1529 	}
1530 
1531 	/*
1532 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1533 	 * the source UDP port. This is required if we want
1534 	 * to select the right SPD for multiple hosts behind
1535 	 * same NAT
1536 	 */
1537 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1538 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1539 		printf("udp4_espinudp: m_tag_get failed\n");
1540 		m_freem(n);
1541 		return 0;
1542 	}
1543 	((u_int16_t *)(tag + 1))[0] = sport;
1544 	((u_int16_t *)(tag + 1))[1] = dport;
1545 	m_tag_prepend(n, tag);
1546 
1547 #ifdef FAST_IPSEC
1548 	ipsec4_common_input(n, iphdrlen);
1549 #else
1550 	esp4_input(n, iphdrlen);
1551 #endif
1552 
1553 	/* We handled it, it shoudln't be handled by UDP */
1554 	return 1;
1555 }
1556 #endif
1557