xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: udp_usrreq.c,v 1.138 2005/04/29 13:06:23 manu 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 1.138 2005/04/29 13:06:23 manu 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 IPSEC_NAT_T
99 #include <netinet6/ipsec.h>
100 #include <netinet6/esp.h>
101 #endif
102 
103 #ifdef INET6
104 #include <netinet/ip6.h>
105 #include <netinet/icmp6.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/udp6_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 #include <machine/stdarg.h>
122 
123 #ifdef FAST_IPSEC
124 #include <netipsec/ipsec.h>
125 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
126 #ifdef INET6
127 #include <netipsec/ipsec6.h>
128 #endif
129 #endif	/* FAST_IPSEC*/
130 
131 #ifdef IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netkey/key.h>
134 #endif /*IPSEC*/
135 
136 #ifdef IPKDB
137 #include <ipkdb/ipkdb.h>
138 #endif
139 
140 /*
141  * UDP protocol implementation.
142  * Per RFC 768, August, 1980.
143  */
144 #ifndef	COMPAT_42
145 int	udpcksum = 1;
146 #else
147 int	udpcksum = 0;		/* XXX */
148 #endif
149 
150 struct	inpcbtable udbtable;
151 struct	udpstat udpstat;
152 
153 #ifdef INET
154 #ifdef IPSEC_NAT_T
155 static int udp4_espinudp (struct mbuf *, int, struct sockaddr *,
156 	struct socket *);
157 #endif
158 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
159 	struct socket *);
160 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
161 	struct mbuf *, int);
162 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
163 #endif
164 #ifdef INET6
165 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
166 	struct socket *);
167 static int udp6_realinput (int, struct sockaddr_in6 *,
168 	struct sockaddr_in6 *, struct mbuf *, int);
169 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
170 #endif
171 #ifdef INET
172 static	void udp_notify (struct inpcb *, int);
173 #endif
174 
175 #ifndef UDBHASHSIZE
176 #define	UDBHASHSIZE	128
177 #endif
178 int	udbhashsize = UDBHASHSIZE;
179 
180 #ifdef MBUFTRACE
181 struct mowner udp_mowner = { "udp" };
182 struct mowner udp_rx_mowner = { "udp", "rx" };
183 struct mowner udp_tx_mowner = { "udp", "tx" };
184 #endif
185 
186 #ifdef UDP_CSUM_COUNTERS
187 #include <sys/device.h>
188 
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 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
199 
200 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
201 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
202 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
203 EVCNT_ATTACH_STATIC(udp_swcsum);
204 
205 #else
206 
207 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
208 
209 #endif /* UDP_CSUM_COUNTERS */
210 
211 void
212 udp_init(void)
213 {
214 
215 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
216 
217 	MOWNER_ATTACH(&udp_tx_mowner);
218 	MOWNER_ATTACH(&udp_rx_mowner);
219 	MOWNER_ATTACH(&udp_mowner);
220 }
221 
222 /*
223  * Checksum extended UDP header and data.
224  */
225 
226 int
227 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
228     int iphlen, int len)
229 {
230 
231 	switch (af) {
232 #ifdef INET
233 	case AF_INET:
234 		return udp4_input_checksum(m, uh, iphlen, len);
235 #endif
236 #ifdef INET6
237 	case AF_INET6:
238 		return udp6_input_checksum(m, uh, iphlen, len);
239 #endif
240 	}
241 #ifdef DIAGNOSTIC
242 	panic("udp_input_checksum: unknown af %d", af);
243 #endif
244 	/* NOTREACHED */
245 	return -1;
246 }
247 
248 #ifdef INET
249 
250 /*
251  * Checksum extended UDP header and data.
252  */
253 
254 static int
255 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
256     int iphlen, int len)
257 {
258 
259 	/*
260 	 * XXX it's better to record and check if this mbuf is
261 	 * already checked.
262 	 */
263 
264 	if (uh->uh_sum == 0)
265 		return 0;
266 
267 	switch (m->m_pkthdr.csum_flags &
268 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
269 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
270 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
271 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
272 		goto badcsum;
273 
274 	case M_CSUM_UDPv4|M_CSUM_DATA: {
275 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
276 
277 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
278 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
279 			const struct ip *ip =
280 			    mtod(m, const struct ip *);
281 
282 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
283 			    ip->ip_dst.s_addr,
284 			    htons(hw_csum + len + IPPROTO_UDP));
285 		}
286 		if ((hw_csum ^ 0xffff) != 0)
287 			goto badcsum;
288 		break;
289 	}
290 
291 	case M_CSUM_UDPv4:
292 		/* Checksum was okay. */
293 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
294 		break;
295 
296 	default:
297 		/*
298 		 * Need to compute it ourselves.  Maybe skip checksum
299 		 * on loopback interfaces.
300 		 */
301 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
302 				     IFF_LOOPBACK) ||
303 				   udp_do_loopback_cksum)) {
304 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
305 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
306 				goto badcsum;
307 		}
308 		break;
309 	}
310 
311 	return 0;
312 
313 badcsum:
314 	udpstat.udps_badsum++;
315 	return -1;
316 }
317 
318 void
319 udp_input(struct mbuf *m, ...)
320 {
321 	va_list ap;
322 	struct sockaddr_in src, dst;
323 	struct ip *ip;
324 	struct udphdr *uh;
325 	int iphlen;
326 	int len;
327 	int n;
328 	u_int16_t ip_len;
329 
330 	va_start(ap, m);
331 	iphlen = va_arg(ap, int);
332 	(void)va_arg(ap, int);		/* ignore value, advance ap */
333 	va_end(ap);
334 
335 	MCLAIM(m, &udp_rx_mowner);
336 	udpstat.udps_ipackets++;
337 
338 	/*
339 	 * Get IP and UDP header together in first mbuf.
340 	 */
341 	ip = mtod(m, struct ip *);
342 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
343 	if (uh == NULL) {
344 		udpstat.udps_hdrops++;
345 		return;
346 	}
347 	KASSERT(UDP_HDR_ALIGNED_P(uh));
348 
349 	/* destination port of 0 is illegal, based on RFC768. */
350 	if (uh->uh_dport == 0)
351 		goto bad;
352 
353 	/*
354 	 * Make mbuf data length reflect UDP length.
355 	 * If not enough data to reflect UDP length, drop.
356 	 */
357 	ip_len = ntohs(ip->ip_len);
358 	len = ntohs((u_int16_t)uh->uh_ulen);
359 	if (ip_len != iphlen + len) {
360 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
361 			udpstat.udps_badlen++;
362 			goto bad;
363 		}
364 		m_adj(m, iphlen + len - ip_len);
365 	}
366 
367 	/*
368 	 * Checksum extended UDP header and data.
369 	 */
370 	if (udp4_input_checksum(m, uh, iphlen, len))
371 		goto badcsum;
372 
373 	/* construct source and dst sockaddrs. */
374 	bzero(&src, sizeof(src));
375 	src.sin_family = AF_INET;
376 	src.sin_len = sizeof(struct sockaddr_in);
377 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
378 	src.sin_port = uh->uh_sport;
379 	bzero(&dst, sizeof(dst));
380 	dst.sin_family = AF_INET;
381 	dst.sin_len = sizeof(struct sockaddr_in);
382 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
383 	dst.sin_port = uh->uh_dport;
384 
385 	n = udp4_realinput(&src, &dst, m, iphlen);
386 #ifdef INET6
387 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
388 		struct sockaddr_in6 src6, dst6;
389 
390 		bzero(&src6, sizeof(src6));
391 		src6.sin6_family = AF_INET6;
392 		src6.sin6_len = sizeof(struct sockaddr_in6);
393 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
394 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
395 			sizeof(ip->ip_src));
396 		src6.sin6_port = uh->uh_sport;
397 		bzero(&dst6, sizeof(dst6));
398 		dst6.sin6_family = AF_INET6;
399 		dst6.sin6_len = sizeof(struct sockaddr_in6);
400 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
401 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
402 			sizeof(ip->ip_dst));
403 		dst6.sin6_port = uh->uh_dport;
404 
405 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
406 	}
407 #endif
408 
409 	if (n == 0) {
410 		if (m->m_flags & (M_BCAST | M_MCAST)) {
411 			udpstat.udps_noportbcast++;
412 			goto bad;
413 		}
414 		udpstat.udps_noport++;
415 #ifdef IPKDB
416 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
417 				m, iphlen + sizeof(struct udphdr),
418 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
419 			/*
420 			 * It was a debugger connect packet,
421 			 * just drop it now
422 			 */
423 			goto bad;
424 		}
425 #endif
426 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
427 		m = NULL;
428 	}
429 
430 bad:
431 	if (m)
432 		m_freem(m);
433 	return;
434 
435 badcsum:
436 	m_freem(m);
437 }
438 #endif
439 
440 #ifdef INET6
441 static int
442 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
443 {
444 
445 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
446 		goto good;
447 	}
448 	if (uh->uh_sum == 0) {
449 		udp6stat.udp6s_nosum++;
450 		goto bad;
451 	}
452 	if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
453 		udp6stat.udp6s_badsum++;
454 		goto bad;
455 	}
456 
457 good:
458 	return 0;
459 bad:
460 	return -1;
461 }
462 
463 int
464 udp6_input(struct mbuf **mp, int *offp, int proto)
465 {
466 	struct mbuf *m = *mp;
467 	int off = *offp;
468 	struct sockaddr_in6 src, dst;
469 	struct ip6_hdr *ip6;
470 	struct udphdr *uh;
471 	u_int32_t plen, ulen;
472 
473 	ip6 = mtod(m, struct ip6_hdr *);
474 
475 #if defined(NFAITH) && 0 < NFAITH
476 	if (faithprefix(&ip6->ip6_dst)) {
477 		/* send icmp6 host unreach? */
478 		m_freem(m);
479 		return IPPROTO_DONE;
480 	}
481 #endif
482 
483 	udp6stat.udp6s_ipackets++;
484 
485 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
486 	plen = m->m_pkthdr.len - off;
487 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
488 	if (uh == NULL) {
489 		ip6stat.ip6s_tooshort++;
490 		return IPPROTO_DONE;
491 	}
492 	KASSERT(UDP_HDR_ALIGNED_P(uh));
493 	ulen = ntohs((u_short)uh->uh_ulen);
494 	/*
495 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
496 	 * iff payload length > 0xffff.
497 	 */
498 	if (ulen == 0 && plen > 0xffff)
499 		ulen = plen;
500 
501 	if (plen != ulen) {
502 		udp6stat.udp6s_badlen++;
503 		goto bad;
504 	}
505 
506 	/* destination port of 0 is illegal, based on RFC768. */
507 	if (uh->uh_dport == 0)
508 		goto bad;
509 
510 	/* Be proactive about malicious use of IPv4 mapped address */
511 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
512 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
513 		/* XXX stat */
514 		goto bad;
515 	}
516 
517 	/*
518 	 * Checksum extended UDP header and data.  Maybe skip checksum
519 	 * on loopback interfaces.
520 	 */
521 	if (udp6_input_checksum(m, uh, off, ulen))
522 		goto bad;
523 
524 	/*
525 	 * Construct source and dst sockaddrs.
526 	 * Note that ifindex (s6_addr16[1]) is already filled.
527 	 */
528 	bzero(&src, sizeof(src));
529 	src.sin6_family = AF_INET6;
530 	src.sin6_len = sizeof(struct sockaddr_in6);
531 	/* KAME hack: recover scopeid */
532 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
533 	src.sin6_port = uh->uh_sport;
534 	bzero(&dst, sizeof(dst));
535 	dst.sin6_family = AF_INET6;
536 	dst.sin6_len = sizeof(struct sockaddr_in6);
537 	/* KAME hack: recover scopeid */
538 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
539 	dst.sin6_port = uh->uh_dport;
540 
541 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
542 		if (m->m_flags & M_MCAST) {
543 			udp6stat.udp6s_noportmcast++;
544 			goto bad;
545 		}
546 		udp6stat.udp6s_noport++;
547 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
548 		m = NULL;
549 	}
550 
551 bad:
552 	if (m)
553 		m_freem(m);
554 	return IPPROTO_DONE;
555 }
556 #endif
557 
558 #ifdef INET
559 static void
560 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
561 	struct sockaddr *src, struct socket *so)
562 {
563 	struct mbuf *opts = NULL;
564 	struct mbuf *n;
565 	struct inpcb *inp = NULL;
566 
567 	if (!so)
568 		return;
569 	switch (so->so_proto->pr_domain->dom_family) {
570 	case AF_INET:
571 		inp = sotoinpcb(so);
572 		break;
573 #ifdef INET6
574 	case AF_INET6:
575 		break;
576 #endif
577 	default:
578 		return;
579 	}
580 
581 #if defined(IPSEC) || defined(FAST_IPSEC)
582 	/* check AH/ESP integrity. */
583 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
584 		ipsecstat.in_polvio++;
585 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
586 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
587 			    0, 0);
588 		return;
589 	}
590 #endif /*IPSEC*/
591 
592 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
593 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
594 			 || so->so_options & SO_TIMESTAMP)) {
595 			struct ip *ip = mtod(n, struct ip *);
596 			ip_savecontrol(inp, &opts, ip, n);
597 		}
598 
599 		m_adj(n, off);
600 		if (sbappendaddr(&so->so_rcv, src, n,
601 				opts) == 0) {
602 			m_freem(n);
603 			if (opts)
604 				m_freem(opts);
605 			so->so_rcv.sb_overflowed++;
606 			udpstat.udps_fullsock++;
607 		} else
608 			sorwakeup(so);
609 	}
610 }
611 #endif
612 
613 #ifdef INET6
614 static void
615 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
616 	struct sockaddr *src, struct socket *so)
617 {
618 	struct mbuf *opts = NULL;
619 	struct mbuf *n;
620 	struct in6pcb *in6p = NULL;
621 
622 	if (!so)
623 		return;
624 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
625 		return;
626 	in6p = sotoin6pcb(so);
627 
628 #if defined(IPSEC) || defined(FAST_IPSEC)
629 	/* check AH/ESP integrity. */
630 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
631 		ipsec6stat.in_polvio++;
632 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
633 			icmp6_error(n, ICMP6_DST_UNREACH,
634 			    ICMP6_DST_UNREACH_ADMIN, 0);
635 		return;
636 	}
637 #endif /*IPSEC*/
638 
639 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
640 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
641 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
642 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
643 			ip6_savecontrol(in6p, &opts, ip6, n);
644 		}
645 
646 		m_adj(n, off);
647 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
648 			m_freem(n);
649 			if (opts)
650 				m_freem(opts);
651 			so->so_rcv.sb_overflowed++;
652 			udp6stat.udp6s_fullsock++;
653 		} else
654 			sorwakeup(so);
655 	}
656 }
657 #endif
658 
659 #ifdef INET
660 static int
661 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
662 	struct mbuf *m, int off /* offset of udphdr */)
663 {
664 	u_int16_t *sport, *dport;
665 	int rcvcnt;
666 	struct in_addr *src4, *dst4;
667 	struct inpcb_hdr *inph;
668 	struct inpcb *inp;
669 
670 	rcvcnt = 0;
671 	off += sizeof(struct udphdr);	/* now, offset of payload */
672 
673 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
674 		goto bad;
675 
676 	src4 = &src->sin_addr;
677 	sport = &src->sin_port;
678 	dst4 = &dst->sin_addr;
679 	dport = &dst->sin_port;
680 
681 	if (IN_MULTICAST(dst4->s_addr) ||
682 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
683 		/*
684 		 * Deliver a multicast or broadcast datagram to *all* sockets
685 		 * for which the local and remote addresses and ports match
686 		 * those of the incoming datagram.  This allows more than
687 		 * one process to receive multi/broadcasts on the same port.
688 		 * (This really ought to be done for unicast datagrams as
689 		 * well, but that would cause problems with existing
690 		 * applications that open both address-specific sockets and
691 		 * a wildcard socket listening to the same port -- they would
692 		 * end up receiving duplicates of every unicast datagram.
693 		 * Those applications open the multiple sockets to overcome an
694 		 * inadequacy of the UDP socket interface, but for backwards
695 		 * compatibility we avoid the problem here rather than
696 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
697 		 */
698 
699 		/*
700 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
701 		 * we need udpiphdr for IPsec processing so we do that later.
702 		 */
703 		/*
704 		 * Locate pcb(s) for datagram.
705 		 */
706 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
707 			inp = (struct inpcb *)inph;
708 			if (inp->inp_af != AF_INET)
709 				continue;
710 
711 			if (inp->inp_lport != *dport)
712 				continue;
713 			if (!in_nullhost(inp->inp_laddr)) {
714 				if (!in_hosteq(inp->inp_laddr, *dst4))
715 					continue;
716 			}
717 			if (!in_nullhost(inp->inp_faddr)) {
718 				if (!in_hosteq(inp->inp_faddr, *src4) ||
719 				    inp->inp_fport != *sport)
720 					continue;
721 			}
722 
723 			udp4_sendup(m, off, (struct sockaddr *)src,
724 				inp->inp_socket);
725 			rcvcnt++;
726 
727 			/*
728 			 * Don't look for additional matches if this one does
729 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
730 			 * socket options set.  This heuristic avoids searching
731 			 * through all pcbs in the common case of a non-shared
732 			 * port.  It assumes that an application will never
733 			 * clear these options after setting them.
734 			 */
735 			if ((inp->inp_socket->so_options &
736 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
737 				break;
738 		}
739 	} else {
740 		/*
741 		 * Locate pcb for datagram.
742 		 */
743 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
744 		if (inp == 0) {
745 			++udpstat.udps_pcbhashmiss;
746 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
747 			if (inp == 0)
748 				return rcvcnt;
749 		}
750 
751 #ifdef IPSEC_NAT_T
752 		/* Handle ESP over UDP */
753 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
754 			struct sockaddr *sa = (struct sockaddr *)src;
755 
756 			if (udp4_espinudp(m, off, sa, inp->inp_socket) != 0) {
757 				rcvcnt++;
758 				goto bad;
759 			}
760 
761 			/* Normal UDP processing will take place */
762 		}
763 #endif
764 
765 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
766 		rcvcnt++;
767 	}
768 
769 bad:
770 	return rcvcnt;
771 }
772 #endif
773 
774 #ifdef INET6
775 static int
776 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
777 	struct mbuf *m, int off)
778 {
779 	u_int16_t sport, dport;
780 	int rcvcnt;
781 	struct in6_addr src6, dst6;
782 	const struct in_addr *dst4;
783 	struct inpcb_hdr *inph;
784 	struct in6pcb *in6p;
785 
786 	rcvcnt = 0;
787 	off += sizeof(struct udphdr);	/* now, offset of payload */
788 
789 	if (af != AF_INET && af != AF_INET6)
790 		goto bad;
791 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
792 		goto bad;
793 
794 	in6_embedscope(&src6, src, NULL, NULL);
795 	sport = src->sin6_port;
796 	in6_embedscope(&dst6, dst, NULL, NULL);
797 	dport = dst->sin6_port;
798 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
799 
800 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
801 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
802 		/*
803 		 * Deliver a multicast or broadcast datagram to *all* sockets
804 		 * for which the local and remote addresses and ports match
805 		 * those of the incoming datagram.  This allows more than
806 		 * one process to receive multi/broadcasts on the same port.
807 		 * (This really ought to be done for unicast datagrams as
808 		 * well, but that would cause problems with existing
809 		 * applications that open both address-specific sockets and
810 		 * a wildcard socket listening to the same port -- they would
811 		 * end up receiving duplicates of every unicast datagram.
812 		 * Those applications open the multiple sockets to overcome an
813 		 * inadequacy of the UDP socket interface, but for backwards
814 		 * compatibility we avoid the problem here rather than
815 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
816 		 */
817 
818 		/*
819 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
820 		 * we need udpiphdr for IPsec processing so we do that later.
821 		 */
822 		/*
823 		 * Locate pcb(s) for datagram.
824 		 */
825 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
826 			in6p = (struct in6pcb *)inph;
827 			if (in6p->in6p_af != AF_INET6)
828 				continue;
829 
830 			if (in6p->in6p_lport != dport)
831 				continue;
832 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
833 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
834 					continue;
835 			} else {
836 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
837 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
838 					continue;
839 			}
840 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
841 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
842 				    &src6) || in6p->in6p_fport != sport)
843 					continue;
844 			} else {
845 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
846 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
847 					continue;
848 			}
849 
850 			udp6_sendup(m, off, (struct sockaddr *)src,
851 				in6p->in6p_socket);
852 			rcvcnt++;
853 
854 			/*
855 			 * Don't look for additional matches if this one does
856 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
857 			 * socket options set.  This heuristic avoids searching
858 			 * through all pcbs in the common case of a non-shared
859 			 * port.  It assumes that an application will never
860 			 * clear these options after setting them.
861 			 */
862 			if ((in6p->in6p_socket->so_options &
863 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
864 				break;
865 		}
866 	} else {
867 		/*
868 		 * Locate pcb for datagram.
869 		 */
870 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
871 		    &dst6, dport, 0);
872 		if (in6p == 0) {
873 			++udpstat.udps_pcbhashmiss;
874 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
875 			if (in6p == 0)
876 				return rcvcnt;
877 		}
878 
879 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
880 		rcvcnt++;
881 	}
882 
883 bad:
884 	return rcvcnt;
885 }
886 #endif
887 
888 #ifdef INET
889 /*
890  * Notify a udp user of an asynchronous error;
891  * just wake up so that he can collect error status.
892  */
893 static void
894 udp_notify(struct inpcb *inp, int errno)
895 {
896 	inp->inp_socket->so_error = errno;
897 	sorwakeup(inp->inp_socket);
898 	sowwakeup(inp->inp_socket);
899 }
900 
901 void *
902 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
903 {
904 	struct ip *ip = v;
905 	struct udphdr *uh;
906 	void (*notify)(struct inpcb *, int) = udp_notify;
907 	int errno;
908 
909 	if (sa->sa_family != AF_INET
910 	 || sa->sa_len != sizeof(struct sockaddr_in))
911 		return NULL;
912 	if ((unsigned)cmd >= PRC_NCMDS)
913 		return NULL;
914 	errno = inetctlerrmap[cmd];
915 	if (PRC_IS_REDIRECT(cmd))
916 		notify = in_rtchange, ip = 0;
917 	else if (cmd == PRC_HOSTDEAD)
918 		ip = 0;
919 	else if (errno == 0)
920 		return NULL;
921 	if (ip) {
922 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
923 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
924 		    ip->ip_src, uh->uh_sport, errno, notify);
925 
926 		/* XXX mapped address case */
927 	} else
928 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
929 		    notify);
930 	return NULL;
931 }
932 
933 int
934 udp_ctloutput(op, so, level, optname, mp)
935 	int op;
936 	struct socket *so;
937 	int level, optname;
938 	struct mbuf **mp;
939 {
940 	int s;
941 	int error = 0;
942 	struct mbuf *m;
943 	struct inpcb *inp;
944 	int family;
945 
946 	family = so->so_proto->pr_domain->dom_family;
947 
948 	s = splsoftnet();
949 	switch (family) {
950 #ifdef INET
951 	case PF_INET:
952 		if (level != IPPROTO_UDP) {
953 			error = ip_ctloutput(op, so, level, optname, mp);
954 			goto end;
955 		}
956 		break;
957 #endif
958 #ifdef INET6
959 	case PF_INET6:
960 		if (level != IPPROTO_UDP) {
961 			error = ip6_ctloutput(op, so, level, optname, mp);
962 			goto end;
963 		}
964 		break;
965 #endif
966 	default:
967 		error = EAFNOSUPPORT;
968 		goto end;
969 		break;
970 	}
971 
972 
973 	switch (op) {
974 	case PRCO_SETOPT:
975 		m = *mp;
976 		inp = sotoinpcb(so);
977 
978 		switch (optname) {
979 		case UDP_ENCAP:
980 			if (m == NULL || m->m_len < sizeof (int)) {
981 				error = EINVAL;
982 				goto end;
983 			}
984 
985 			switch(*mtod(m, int *)) {
986 #ifdef IPSEC_NAT_T
987 			case 0:
988 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
989 				break;
990 
991 			case UDP_ENCAP_ESPINUDP:
992 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
993 				inp->inp_flags |= INP_ESPINUDP;
994 				break;
995 
996 			case UDP_ENCAP_ESPINUDP_NON_IKE:
997 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
998 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
999 				break;
1000 #endif
1001 			default:
1002 				error = EINVAL;
1003 				goto end;
1004 				break;
1005 			}
1006 			break;
1007 
1008 		default:
1009 			error = ENOPROTOOPT;
1010 			goto end;
1011 			break;
1012 		}
1013 		break;
1014 
1015 	default:
1016 		error = EINVAL;
1017 		goto end;
1018 		break;
1019 	}
1020 
1021 end:
1022 	splx(s);
1023 	return error;
1024 }
1025 
1026 
1027 int
1028 udp_output(struct mbuf *m, ...)
1029 {
1030 	struct inpcb *inp;
1031 	struct udpiphdr *ui;
1032 	struct route *ro;
1033 	int len = m->m_pkthdr.len;
1034 	int error = 0;
1035 	va_list ap;
1036 
1037 	MCLAIM(m, &udp_tx_mowner);
1038 	va_start(ap, m);
1039 	inp = va_arg(ap, struct inpcb *);
1040 	va_end(ap);
1041 
1042 	/*
1043 	 * Calculate data length and get a mbuf
1044 	 * for UDP and IP headers.
1045 	 */
1046 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1047 	if (m == 0) {
1048 		error = ENOBUFS;
1049 		goto release;
1050 	}
1051 
1052 	/*
1053 	 * Compute the packet length of the IP header, and
1054 	 * punt if the length looks bogus.
1055 	 */
1056 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1057 		error = EMSGSIZE;
1058 		goto release;
1059 	}
1060 
1061 	/*
1062 	 * Fill in mbuf with extended UDP header
1063 	 * and addresses and length put into network format.
1064 	 */
1065 	ui = mtod(m, struct udpiphdr *);
1066 	ui->ui_pr = IPPROTO_UDP;
1067 	ui->ui_src = inp->inp_laddr;
1068 	ui->ui_dst = inp->inp_faddr;
1069 	ui->ui_sport = inp->inp_lport;
1070 	ui->ui_dport = inp->inp_fport;
1071 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1072 
1073 	ro = &inp->inp_route;
1074 
1075 	/*
1076 	 * Set up checksum and output datagram.
1077 	 */
1078 	if (udpcksum) {
1079 		/*
1080 		 * XXX Cache pseudo-header checksum part for
1081 		 * XXX "connected" UDP sockets.
1082 		 */
1083 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1084 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1085 		    sizeof(struct udphdr) + IPPROTO_UDP));
1086 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1087 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1088 	} else
1089 		ui->ui_sum = 0;
1090 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1091 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1092 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1093 	udpstat.udps_opackets++;
1094 
1095 	return (ip_output(m, inp->inp_options, ro,
1096 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1097 	    inp->inp_moptions, inp->inp_socket));
1098 
1099 release:
1100 	m_freem(m);
1101 	return (error);
1102 }
1103 
1104 int	udp_sendspace = 9216;		/* really max datagram size */
1105 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1106 					/* 40 1K datagrams */
1107 
1108 /*ARGSUSED*/
1109 int
1110 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1111 	struct mbuf *control, struct proc *p)
1112 {
1113 	struct inpcb *inp;
1114 	int s;
1115 	int error = 0;
1116 
1117 	if (req == PRU_CONTROL)
1118 		return (in_control(so, (long)m, (caddr_t)nam,
1119 		    (struct ifnet *)control, p));
1120 
1121 	if (req == PRU_PURGEIF) {
1122 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1123 		in_purgeif((struct ifnet *)control);
1124 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1125 		return (0);
1126 	}
1127 
1128 	s = splsoftnet();
1129 	inp = sotoinpcb(so);
1130 #ifdef DIAGNOSTIC
1131 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1132 		panic("udp_usrreq: unexpected control mbuf");
1133 #endif
1134 	if (inp == 0 && req != PRU_ATTACH) {
1135 		error = EINVAL;
1136 		goto release;
1137 	}
1138 
1139 	/*
1140 	 * Note: need to block udp_input while changing
1141 	 * the udp pcb queue and/or pcb addresses.
1142 	 */
1143 	switch (req) {
1144 
1145 	case PRU_ATTACH:
1146 		if (inp != 0) {
1147 			error = EISCONN;
1148 			break;
1149 		}
1150 #ifdef MBUFTRACE
1151 		so->so_mowner = &udp_mowner;
1152 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1153 		so->so_snd.sb_mowner = &udp_tx_mowner;
1154 #endif
1155 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1156 			error = soreserve(so, udp_sendspace, udp_recvspace);
1157 			if (error)
1158 				break;
1159 		}
1160 		error = in_pcballoc(so, &udbtable);
1161 		if (error)
1162 			break;
1163 		inp = sotoinpcb(so);
1164 		inp->inp_ip.ip_ttl = ip_defttl;
1165 		break;
1166 
1167 	case PRU_DETACH:
1168 		in_pcbdetach(inp);
1169 		break;
1170 
1171 	case PRU_BIND:
1172 		error = in_pcbbind(inp, nam, p);
1173 		break;
1174 
1175 	case PRU_LISTEN:
1176 		error = EOPNOTSUPP;
1177 		break;
1178 
1179 	case PRU_CONNECT:
1180 		error = in_pcbconnect(inp, nam);
1181 		if (error)
1182 			break;
1183 		soisconnected(so);
1184 		break;
1185 
1186 	case PRU_CONNECT2:
1187 		error = EOPNOTSUPP;
1188 		break;
1189 
1190 	case PRU_DISCONNECT:
1191 		/*soisdisconnected(so);*/
1192 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1193 		in_pcbdisconnect(inp);
1194 		inp->inp_laddr = zeroin_addr;		/* XXX */
1195 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1196 		break;
1197 
1198 	case PRU_SHUTDOWN:
1199 		socantsendmore(so);
1200 		break;
1201 
1202 	case PRU_RCVD:
1203 		error = EOPNOTSUPP;
1204 		break;
1205 
1206 	case PRU_SEND:
1207 		if (control && control->m_len) {
1208 			m_freem(control);
1209 			m_freem(m);
1210 			error = EINVAL;
1211 			break;
1212 		}
1213 	{
1214 		struct in_addr laddr;			/* XXX */
1215 
1216 		if (nam) {
1217 			laddr = inp->inp_laddr;		/* XXX */
1218 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1219 				error = EISCONN;
1220 				goto die;
1221 			}
1222 			error = in_pcbconnect(inp, nam);
1223 			if (error)
1224 				goto die;
1225 		} else {
1226 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1227 				error = ENOTCONN;
1228 				goto die;
1229 			}
1230 		}
1231 		error = udp_output(m, inp);
1232 		m = NULL;
1233 		if (nam) {
1234 			in_pcbdisconnect(inp);
1235 			inp->inp_laddr = laddr;		/* XXX */
1236 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1237 		}
1238 	  die:
1239 		if (m)
1240 			m_freem(m);
1241 	}
1242 		break;
1243 
1244 	case PRU_SENSE:
1245 		/*
1246 		 * stat: don't bother with a blocksize.
1247 		 */
1248 		splx(s);
1249 		return (0);
1250 
1251 	case PRU_RCVOOB:
1252 		error =  EOPNOTSUPP;
1253 		break;
1254 
1255 	case PRU_SENDOOB:
1256 		m_freem(control);
1257 		m_freem(m);
1258 		error =  EOPNOTSUPP;
1259 		break;
1260 
1261 	case PRU_SOCKADDR:
1262 		in_setsockaddr(inp, nam);
1263 		break;
1264 
1265 	case PRU_PEERADDR:
1266 		in_setpeeraddr(inp, nam);
1267 		break;
1268 
1269 	default:
1270 		panic("udp_usrreq");
1271 	}
1272 
1273 release:
1274 	splx(s);
1275 	return (error);
1276 }
1277 
1278 /*
1279  * Sysctl for udp variables.
1280  */
1281 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1282 {
1283 
1284 	sysctl_createv(clog, 0, NULL, NULL,
1285 		       CTLFLAG_PERMANENT,
1286 		       CTLTYPE_NODE, "net", NULL,
1287 		       NULL, 0, NULL, 0,
1288 		       CTL_NET, CTL_EOL);
1289 	sysctl_createv(clog, 0, NULL, NULL,
1290 		       CTLFLAG_PERMANENT,
1291 		       CTLTYPE_NODE, "inet", NULL,
1292 		       NULL, 0, NULL, 0,
1293 		       CTL_NET, PF_INET, CTL_EOL);
1294 	sysctl_createv(clog, 0, NULL, NULL,
1295 		       CTLFLAG_PERMANENT,
1296 		       CTLTYPE_NODE, "udp",
1297 		       SYSCTL_DESCR("UDPv4 related settings"),
1298 		       NULL, 0, NULL, 0,
1299 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1300 
1301 	sysctl_createv(clog, 0, NULL, NULL,
1302 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1303 		       CTLTYPE_INT, "checksum",
1304 		       SYSCTL_DESCR("Compute UDP checksums"),
1305 		       NULL, 0, &udpcksum, 0,
1306 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1307 		       CTL_EOL);
1308 	sysctl_createv(clog, 0, NULL, NULL,
1309 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1310 		       CTLTYPE_INT, "sendspace",
1311 		       SYSCTL_DESCR("Default UDP send buffer size"),
1312 		       NULL, 0, &udp_sendspace, 0,
1313 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1314 		       CTL_EOL);
1315 	sysctl_createv(clog, 0, NULL, NULL,
1316 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1317 		       CTLTYPE_INT, "recvspace",
1318 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1319 		       NULL, 0, &udp_recvspace, 0,
1320 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1321 		       CTL_EOL);
1322 	sysctl_createv(clog, 0, NULL, NULL,
1323 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1324 		       CTLTYPE_INT, "do_loopback_cksum",
1325 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1326 		       NULL, 0, &udp_do_loopback_cksum, 0,
1327 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1328 		       CTL_EOL);
1329 	sysctl_createv(clog, 0, NULL, NULL,
1330 		       CTLFLAG_PERMANENT,
1331 		       CTLTYPE_STRUCT, "pcblist",
1332 		       SYSCTL_DESCR("UDP protocol control block list"),
1333 		       sysctl_inpcblist, 0, &udbtable, 0,
1334 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1335 		       CTL_EOL);
1336 }
1337 #endif
1338 
1339 #if (defined INET && defined IPSEC_NAT_T)
1340 /*
1341  * Returns:
1342  * 1 if the packet was processed
1343  * 0 if normal UDP processing should take place
1344  */
1345 static int
1346 udp4_espinudp(m, off, src, so)
1347 	struct mbuf *m;
1348 	int off;
1349 	struct sockaddr *src;
1350 	struct socket *so;
1351 {
1352 	size_t len;
1353 	caddr_t data;
1354 	struct inpcb *inp;
1355 	size_t skip = 0;
1356 	size_t minlen;
1357 	size_t iphdrlen;
1358 	struct ip *ip;
1359 	struct mbuf *n;
1360 	struct m_tag *tag;
1361 	struct udphdr *udphdr;
1362 	u_int16_t sport, dport;
1363 
1364 	/*
1365 	 * Collapse the mbuf chain if the first mbuf is too short
1366 	 * The longest case is: UDP + non ESP marker + ESP
1367 	 */
1368 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1369 	if (minlen > m->m_pkthdr.len)
1370 		minlen = m->m_pkthdr.len;
1371 
1372 	if (m->m_len < minlen) {
1373 		if ((m = m_pullup(m, minlen)) == NULL) {
1374 			printf("udp4_espinudp: m_pullup failed\n");
1375 			return 0;
1376 		}
1377 	}
1378 
1379 	len = m->m_len - off;
1380 	data = mtod(m, caddr_t) + off;
1381 	inp = sotoinpcb(so);
1382 
1383 	/* Ignore keepalive packets */
1384 	if ((len == 1) && (data[0] == '\xff')) {
1385 		return 1;
1386 	}
1387 
1388 	/*
1389 	 * Check that the payload is long enough to hold
1390 	 * an ESP header and compute the length of encapsulation
1391 	 * header to remove
1392 	 */
1393 	if (inp->inp_flags & INP_ESPINUDP) {
1394 		u_int32_t *st = (u_int32_t *)data;
1395 
1396 		if ((len <= sizeof(struct esp)) || (*st == 0))
1397 			return 0; /* Normal UDP processing */
1398 
1399 		skip = sizeof(struct udphdr);
1400 	}
1401 
1402 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1403 		u_int64_t *st = (u_int64_t *)data;
1404 
1405 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1406 		    || (*st != 0))
1407 			return 0; /* Normal UDP processing */
1408 
1409 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1410 	}
1411 
1412 	/*
1413 	 * Get the UDP ports. They are handled in network
1414 	 * order everywhere in IPSEC_NAT_T code.
1415 	 */
1416 	udphdr = (struct udphdr *)(data - skip);
1417 	sport = udphdr->uh_sport;
1418 	dport = udphdr->uh_dport;
1419 
1420 	/*
1421 	 * Remove the UDP header (and possibly the non ESP marker)
1422 	 * IP header lendth is iphdrlen
1423 	 * Before:
1424 	 *   <--- off --->
1425 	 *   +----+------+-----+
1426 	 *   | IP |  UDP | ESP |
1427 	 *   +----+------+-----+
1428 	 *        <-skip->
1429 	 * After:
1430 	 *          +----+-----+
1431 	 *          | IP | ESP |
1432 	 *          +----+-----+
1433 	 *   <-skip->
1434 	 */
1435 	iphdrlen = off - sizeof(struct udphdr);
1436 	memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
1437 	m_adj(m, skip);
1438 
1439 	ip = mtod(m, struct ip *);
1440 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1441 	ip->ip_p = IPPROTO_ESP;
1442 
1443 	/*
1444 	 * Copy the mbuf to avoid multiple free, as both
1445 	 * esp4_input (which we call) and udp_input (which
1446 	 * called us) free the mbuf.
1447 	 */
1448 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1449 		printf("udp4_espinudp: m_dup failed\n");
1450 		return 0;
1451 	}
1452 
1453 	/*
1454 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1455 	 * the source UDP port. This is required if we want
1456 	 * to select the right SPD for multiple hosts behind
1457 	 * same NAT
1458 	 */
1459 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1460 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1461 		printf("udp4_espinudp: m_tag_get failed\n");
1462 		m_freem(n);
1463 		return 0;
1464 	}
1465 	((u_int16_t *)(tag + 1))[0] = sport;
1466 	((u_int16_t *)(tag + 1))[1] = dport;
1467 	m_tag_prepend(n, tag);
1468 
1469 	esp4_input(n, iphdrlen);
1470 
1471 	/* We handled it, it shoudln't be handled by UDP */
1472 	return 1;
1473 }
1474 #endif
1475