xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: udp_usrreq.c,v 1.163 2007/11/27 22:45:30 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 1.163 2007/11/27 22:45:30 christos 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 int	udpcksum = 1;
143 int	udp_do_loopback_cksum = 0;
144 
145 struct	inpcbtable udbtable;
146 struct	udpstat udpstat;
147 
148 #ifdef INET
149 #ifdef IPSEC_NAT_T
150 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
151 	struct socket *);
152 #endif
153 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
154 	struct socket *);
155 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
156 	struct mbuf **, int);
157 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
158 #endif
159 #ifdef INET6
160 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
161 	struct socket *);
162 static int udp6_realinput (int, struct sockaddr_in6 *,
163 	struct sockaddr_in6 *, struct mbuf *, int);
164 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
165 #endif
166 #ifdef INET
167 static	void udp_notify (struct inpcb *, int);
168 #endif
169 
170 #ifndef UDBHASHSIZE
171 #define	UDBHASHSIZE	128
172 #endif
173 int	udbhashsize = UDBHASHSIZE;
174 
175 #ifdef MBUFTRACE
176 struct mowner udp_mowner = MOWNER_INIT("udp", "");
177 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
178 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
179 #endif
180 
181 #ifdef UDP_CSUM_COUNTERS
182 #include <sys/device.h>
183 
184 #if defined(INET)
185 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
186     NULL, "udp", "hwcsum bad");
187 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
188     NULL, "udp", "hwcsum ok");
189 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
190     NULL, "udp", "hwcsum data");
191 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192     NULL, "udp", "swcsum");
193 
194 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
195 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
196 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
197 EVCNT_ATTACH_STATIC(udp_swcsum);
198 #endif /* defined(INET) */
199 
200 #if defined(INET6)
201 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
202     NULL, "udp6", "hwcsum bad");
203 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
204     NULL, "udp6", "hwcsum ok");
205 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
206     NULL, "udp6", "hwcsum data");
207 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208     NULL, "udp6", "swcsum");
209 
210 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
211 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
212 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
213 EVCNT_ATTACH_STATIC(udp6_swcsum);
214 #endif /* defined(INET6) */
215 
216 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
217 
218 #else
219 
220 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
221 
222 #endif /* UDP_CSUM_COUNTERS */
223 
224 void
225 udp_init(void)
226 {
227 
228 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
229 
230 	MOWNER_ATTACH(&udp_tx_mowner);
231 	MOWNER_ATTACH(&udp_rx_mowner);
232 	MOWNER_ATTACH(&udp_mowner);
233 }
234 
235 /*
236  * Checksum extended UDP header and data.
237  */
238 
239 int
240 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
241     int iphlen, int len)
242 {
243 
244 	switch (af) {
245 #ifdef INET
246 	case AF_INET:
247 		return udp4_input_checksum(m, uh, iphlen, len);
248 #endif
249 #ifdef INET6
250 	case AF_INET6:
251 		return udp6_input_checksum(m, uh, iphlen, len);
252 #endif
253 	}
254 #ifdef DIAGNOSTIC
255 	panic("udp_input_checksum: unknown af %d", af);
256 #endif
257 	/* NOTREACHED */
258 	return -1;
259 }
260 
261 #ifdef INET
262 
263 /*
264  * Checksum extended UDP header and data.
265  */
266 
267 static int
268 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
269     int iphlen, int len)
270 {
271 
272 	/*
273 	 * XXX it's better to record and check if this mbuf is
274 	 * already checked.
275 	 */
276 
277 	if (uh->uh_sum == 0)
278 		return 0;
279 
280 	switch (m->m_pkthdr.csum_flags &
281 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
282 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
283 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
284 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
285 		goto badcsum;
286 
287 	case M_CSUM_UDPv4|M_CSUM_DATA: {
288 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
289 
290 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
291 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
292 			const struct ip *ip =
293 			    mtod(m, const struct ip *);
294 
295 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
296 			    ip->ip_dst.s_addr,
297 			    htons(hw_csum + len + IPPROTO_UDP));
298 		}
299 		if ((hw_csum ^ 0xffff) != 0)
300 			goto badcsum;
301 		break;
302 	}
303 
304 	case M_CSUM_UDPv4:
305 		/* Checksum was okay. */
306 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
307 		break;
308 
309 	default:
310 		/*
311 		 * Need to compute it ourselves.  Maybe skip checksum
312 		 * on loopback interfaces.
313 		 */
314 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
315 				     IFF_LOOPBACK) ||
316 				   udp_do_loopback_cksum)) {
317 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
318 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
319 				goto badcsum;
320 		}
321 		break;
322 	}
323 
324 	return 0;
325 
326 badcsum:
327 	udpstat.udps_badsum++;
328 	return -1;
329 }
330 
331 void
332 udp_input(struct mbuf *m, ...)
333 {
334 	va_list ap;
335 	struct sockaddr_in src, dst;
336 	struct ip *ip;
337 	struct udphdr *uh;
338 	int iphlen;
339 	int len;
340 	int n;
341 	u_int16_t ip_len;
342 
343 	va_start(ap, m);
344 	iphlen = va_arg(ap, int);
345 	(void)va_arg(ap, int);		/* ignore value, advance ap */
346 	va_end(ap);
347 
348 	MCLAIM(m, &udp_rx_mowner);
349 	udpstat.udps_ipackets++;
350 
351 	/*
352 	 * Get IP and UDP header together in first mbuf.
353 	 */
354 	ip = mtod(m, struct ip *);
355 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
356 	if (uh == NULL) {
357 		udpstat.udps_hdrops++;
358 		return;
359 	}
360 	KASSERT(UDP_HDR_ALIGNED_P(uh));
361 
362 	/* destination port of 0 is illegal, based on RFC768. */
363 	if (uh->uh_dport == 0)
364 		goto bad;
365 
366 	/*
367 	 * Make mbuf data length reflect UDP length.
368 	 * If not enough data to reflect UDP length, drop.
369 	 */
370 	ip_len = ntohs(ip->ip_len);
371 	len = ntohs((u_int16_t)uh->uh_ulen);
372 	if (ip_len != iphlen + len) {
373 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
374 			udpstat.udps_badlen++;
375 			goto bad;
376 		}
377 		m_adj(m, iphlen + len - ip_len);
378 	}
379 
380 	/*
381 	 * Checksum extended UDP header and data.
382 	 */
383 	if (udp4_input_checksum(m, uh, iphlen, len))
384 		goto badcsum;
385 
386 	/* construct source and dst sockaddrs. */
387 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
388 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
389 
390 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
391 		udpstat.udps_hdrops++;
392 		return;
393 	}
394 #ifdef INET6
395 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
396 		struct sockaddr_in6 src6, dst6;
397 
398 		bzero(&src6, sizeof(src6));
399 		src6.sin6_family = AF_INET6;
400 		src6.sin6_len = sizeof(struct sockaddr_in6);
401 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
402 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
403 			sizeof(ip->ip_src));
404 		src6.sin6_port = uh->uh_sport;
405 		bzero(&dst6, sizeof(dst6));
406 		dst6.sin6_family = AF_INET6;
407 		dst6.sin6_len = sizeof(struct sockaddr_in6);
408 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
409 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
410 			sizeof(ip->ip_dst));
411 		dst6.sin6_port = uh->uh_dport;
412 
413 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
414 	}
415 #endif
416 
417 	if (n == 0) {
418 		if (m->m_flags & (M_BCAST | M_MCAST)) {
419 			udpstat.udps_noportbcast++;
420 			goto bad;
421 		}
422 		udpstat.udps_noport++;
423 #ifdef IPKDB
424 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
425 				m, iphlen + sizeof(struct udphdr),
426 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
427 			/*
428 			 * It was a debugger connect packet,
429 			 * just drop it now
430 			 */
431 			goto bad;
432 		}
433 #endif
434 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
435 		m = NULL;
436 	}
437 
438 bad:
439 	if (m)
440 		m_freem(m);
441 	return;
442 
443 badcsum:
444 	m_freem(m);
445 }
446 #endif
447 
448 #ifdef INET6
449 static int
450 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
451 {
452 
453 	/*
454 	 * XXX it's better to record and check if this mbuf is
455 	 * already checked.
456 	 */
457 
458 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
459 		goto good;
460 	}
461 	if (uh->uh_sum == 0) {
462 		udp6stat.udp6s_nosum++;
463 		goto bad;
464 	}
465 
466 	switch (m->m_pkthdr.csum_flags &
467 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
468 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
469 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
470 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
471 		udp6stat.udp6s_badsum++;
472 		goto bad;
473 
474 #if 0 /* notyet */
475 	case M_CSUM_UDPv6|M_CSUM_DATA:
476 #endif
477 
478 	case M_CSUM_UDPv6:
479 		/* Checksum was okay. */
480 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
481 		break;
482 
483 	default:
484 		/*
485 		 * Need to compute it ourselves.  Maybe skip checksum
486 		 * on loopback interfaces.
487 		 */
488 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
489 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
490 			udp6stat.udp6s_badsum++;
491 			goto bad;
492 		}
493 	}
494 
495 good:
496 	return 0;
497 bad:
498 	return -1;
499 }
500 
501 int
502 udp6_input(struct mbuf **mp, int *offp, int proto)
503 {
504 	struct mbuf *m = *mp;
505 	int off = *offp;
506 	struct sockaddr_in6 src, dst;
507 	struct ip6_hdr *ip6;
508 	struct udphdr *uh;
509 	u_int32_t plen, ulen;
510 
511 	ip6 = mtod(m, struct ip6_hdr *);
512 
513 #if defined(NFAITH) && 0 < NFAITH
514 	if (faithprefix(&ip6->ip6_dst)) {
515 		/* send icmp6 host unreach? */
516 		m_freem(m);
517 		return IPPROTO_DONE;
518 	}
519 #endif
520 
521 	udp6stat.udp6s_ipackets++;
522 
523 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
524 	plen = m->m_pkthdr.len - off;
525 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
526 	if (uh == NULL) {
527 		ip6stat.ip6s_tooshort++;
528 		return IPPROTO_DONE;
529 	}
530 	KASSERT(UDP_HDR_ALIGNED_P(uh));
531 	ulen = ntohs((u_short)uh->uh_ulen);
532 	/*
533 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
534 	 * iff payload length > 0xffff.
535 	 */
536 	if (ulen == 0 && plen > 0xffff)
537 		ulen = plen;
538 
539 	if (plen != ulen) {
540 		udp6stat.udp6s_badlen++;
541 		goto bad;
542 	}
543 
544 	/* destination port of 0 is illegal, based on RFC768. */
545 	if (uh->uh_dport == 0)
546 		goto bad;
547 
548 	/* Be proactive about malicious use of IPv4 mapped address */
549 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
550 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
551 		/* XXX stat */
552 		goto bad;
553 	}
554 
555 	/*
556 	 * Checksum extended UDP header and data.  Maybe skip checksum
557 	 * on loopback interfaces.
558 	 */
559 	if (udp6_input_checksum(m, uh, off, ulen))
560 		goto bad;
561 
562 	/*
563 	 * Construct source and dst sockaddrs.
564 	 */
565 	bzero(&src, sizeof(src));
566 	src.sin6_family = AF_INET6;
567 	src.sin6_len = sizeof(struct sockaddr_in6);
568 	src.sin6_addr = ip6->ip6_src;
569 	src.sin6_port = uh->uh_sport;
570 	bzero(&dst, sizeof(dst));
571 	dst.sin6_family = AF_INET6;
572 	dst.sin6_len = sizeof(struct sockaddr_in6);
573 	dst.sin6_addr = ip6->ip6_dst;
574 	dst.sin6_port = uh->uh_dport;
575 
576 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
577 		if (m->m_flags & M_MCAST) {
578 			udp6stat.udp6s_noportmcast++;
579 			goto bad;
580 		}
581 		udp6stat.udp6s_noport++;
582 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
583 		m = NULL;
584 	}
585 
586 bad:
587 	if (m)
588 		m_freem(m);
589 	return IPPROTO_DONE;
590 }
591 #endif
592 
593 #ifdef INET
594 static void
595 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
596 	struct sockaddr *src, struct socket *so)
597 {
598 	struct mbuf *opts = NULL;
599 	struct mbuf *n;
600 	struct inpcb *inp = NULL;
601 
602 	if (!so)
603 		return;
604 	switch (so->so_proto->pr_domain->dom_family) {
605 	case AF_INET:
606 		inp = sotoinpcb(so);
607 		break;
608 #ifdef INET6
609 	case AF_INET6:
610 		break;
611 #endif
612 	default:
613 		return;
614 	}
615 
616 #if defined(IPSEC) || defined(FAST_IPSEC)
617 	/* check AH/ESP integrity. */
618 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
619 		ipsecstat.in_polvio++;
620 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
621 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
622 			    0, 0);
623 		return;
624 	}
625 #endif /*IPSEC*/
626 
627 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
628 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
629 			 || so->so_options & SO_TIMESTAMP)) {
630 			struct ip *ip = mtod(n, struct ip *);
631 			ip_savecontrol(inp, &opts, ip, n);
632 		}
633 
634 		m_adj(n, off);
635 		if (sbappendaddr(&so->so_rcv, src, n,
636 				opts) == 0) {
637 			m_freem(n);
638 			if (opts)
639 				m_freem(opts);
640 			so->so_rcv.sb_overflowed++;
641 			udpstat.udps_fullsock++;
642 		} else
643 			sorwakeup(so);
644 	}
645 }
646 #endif
647 
648 #ifdef INET6
649 static void
650 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
651 	struct sockaddr *src, struct socket *so)
652 {
653 	struct mbuf *opts = NULL;
654 	struct mbuf *n;
655 	struct in6pcb *in6p = NULL;
656 
657 	if (!so)
658 		return;
659 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
660 		return;
661 	in6p = sotoin6pcb(so);
662 
663 #if defined(IPSEC) || defined(FAST_IPSEC)
664 	/* check AH/ESP integrity. */
665 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
666 		ipsec6stat.in_polvio++;
667 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
668 			icmp6_error(n, ICMP6_DST_UNREACH,
669 			    ICMP6_DST_UNREACH_ADMIN, 0);
670 		return;
671 	}
672 #endif /*IPSEC*/
673 
674 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
675 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
676 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
677 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
678 			ip6_savecontrol(in6p, &opts, ip6, n);
679 		}
680 
681 		m_adj(n, off);
682 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
683 			m_freem(n);
684 			if (opts)
685 				m_freem(opts);
686 			so->so_rcv.sb_overflowed++;
687 			udp6stat.udp6s_fullsock++;
688 		} else
689 			sorwakeup(so);
690 	}
691 }
692 #endif
693 
694 #ifdef INET
695 static int
696 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
697 	struct mbuf **mp, int off /* offset of udphdr */)
698 {
699 	u_int16_t *sport, *dport;
700 	int rcvcnt;
701 	struct in_addr *src4, *dst4;
702 	struct inpcb_hdr *inph;
703 	struct inpcb *inp;
704 	struct mbuf *m = *mp;
705 
706 	rcvcnt = 0;
707 	off += sizeof(struct udphdr);	/* now, offset of payload */
708 
709 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
710 		goto bad;
711 
712 	src4 = &src->sin_addr;
713 	sport = &src->sin_port;
714 	dst4 = &dst->sin_addr;
715 	dport = &dst->sin_port;
716 
717 	if (IN_MULTICAST(dst4->s_addr) ||
718 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
719 		/*
720 		 * Deliver a multicast or broadcast datagram to *all* sockets
721 		 * for which the local and remote addresses and ports match
722 		 * those of the incoming datagram.  This allows more than
723 		 * one process to receive multi/broadcasts on the same port.
724 		 * (This really ought to be done for unicast datagrams as
725 		 * well, but that would cause problems with existing
726 		 * applications that open both address-specific sockets and
727 		 * a wildcard socket listening to the same port -- they would
728 		 * end up receiving duplicates of every unicast datagram.
729 		 * Those applications open the multiple sockets to overcome an
730 		 * inadequacy of the UDP socket interface, but for backwards
731 		 * compatibility we avoid the problem here rather than
732 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
733 		 */
734 
735 		/*
736 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
737 		 * we need udpiphdr for IPsec processing so we do that later.
738 		 */
739 		/*
740 		 * Locate pcb(s) for datagram.
741 		 */
742 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
743 			inp = (struct inpcb *)inph;
744 			if (inp->inp_af != AF_INET)
745 				continue;
746 
747 			if (inp->inp_lport != *dport)
748 				continue;
749 			if (!in_nullhost(inp->inp_laddr)) {
750 				if (!in_hosteq(inp->inp_laddr, *dst4))
751 					continue;
752 			}
753 			if (!in_nullhost(inp->inp_faddr)) {
754 				if (!in_hosteq(inp->inp_faddr, *src4) ||
755 				    inp->inp_fport != *sport)
756 					continue;
757 			}
758 
759 			udp4_sendup(m, off, (struct sockaddr *)src,
760 				inp->inp_socket);
761 			rcvcnt++;
762 
763 			/*
764 			 * Don't look for additional matches if this one does
765 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
766 			 * socket options set.  This heuristic avoids searching
767 			 * through all pcbs in the common case of a non-shared
768 			 * port.  It assumes that an application will never
769 			 * clear these options after setting them.
770 			 */
771 			if ((inp->inp_socket->so_options &
772 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
773 				break;
774 		}
775 	} else {
776 		/*
777 		 * Locate pcb for datagram.
778 		 */
779 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
780 		if (inp == 0) {
781 			++udpstat.udps_pcbhashmiss;
782 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
783 			if (inp == 0)
784 				return rcvcnt;
785 		}
786 
787 #ifdef IPSEC_NAT_T
788 		/* Handle ESP over UDP */
789 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
790 			struct sockaddr *sa = (struct sockaddr *)src;
791 
792 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
793 			case -1: 	/* Error, m was freeed */
794 				rcvcnt = -1;
795 				goto bad;
796 				break;
797 
798 			case 1:		/* ESP over UDP */
799 				rcvcnt++;
800 				goto bad;
801 				break;
802 
803 			case 0: 	/* plain UDP */
804 			default: 	/* Unexpected */
805 				/*
806 				 * Normal UDP processing will take place
807 				 * m may have changed.
808 				 */
809 				m = *mp;
810 				break;
811 			}
812 		}
813 #endif
814 
815 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
816 		rcvcnt++;
817 	}
818 
819 bad:
820 	return rcvcnt;
821 }
822 #endif
823 
824 #ifdef INET6
825 static int
826 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
827 	struct mbuf *m, int off)
828 {
829 	u_int16_t sport, dport;
830 	int rcvcnt;
831 	struct in6_addr src6, *dst6;
832 	const struct in_addr *dst4;
833 	struct inpcb_hdr *inph;
834 	struct in6pcb *in6p;
835 
836 	rcvcnt = 0;
837 	off += sizeof(struct udphdr);	/* now, offset of payload */
838 
839 	if (af != AF_INET && af != AF_INET6)
840 		goto bad;
841 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
842 		goto bad;
843 
844 	src6 = src->sin6_addr;
845 	if (sa6_recoverscope(src) != 0) {
846 		/* XXX: should be impossible. */
847 		goto bad;
848 	}
849 	sport = src->sin6_port;
850 
851 	dport = dst->sin6_port;
852 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
853 	dst6 = &dst->sin6_addr;
854 
855 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
856 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
857 		/*
858 		 * Deliver a multicast or broadcast datagram to *all* sockets
859 		 * for which the local and remote addresses and ports match
860 		 * those of the incoming datagram.  This allows more than
861 		 * one process to receive multi/broadcasts on the same port.
862 		 * (This really ought to be done for unicast datagrams as
863 		 * well, but that would cause problems with existing
864 		 * applications that open both address-specific sockets and
865 		 * a wildcard socket listening to the same port -- they would
866 		 * end up receiving duplicates of every unicast datagram.
867 		 * Those applications open the multiple sockets to overcome an
868 		 * inadequacy of the UDP socket interface, but for backwards
869 		 * compatibility we avoid the problem here rather than
870 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
871 		 */
872 
873 		/*
874 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
875 		 * we need udpiphdr for IPsec processing so we do that later.
876 		 */
877 		/*
878 		 * Locate pcb(s) for datagram.
879 		 */
880 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
881 			in6p = (struct in6pcb *)inph;
882 			if (in6p->in6p_af != AF_INET6)
883 				continue;
884 
885 			if (in6p->in6p_lport != dport)
886 				continue;
887 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
888 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
889 				    dst6))
890 					continue;
891 			} else {
892 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
893 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
894 					continue;
895 			}
896 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
897 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
898 				    &src6) || in6p->in6p_fport != sport)
899 					continue;
900 			} else {
901 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
902 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
903 					continue;
904 			}
905 
906 			udp6_sendup(m, off, (struct sockaddr *)src,
907 				in6p->in6p_socket);
908 			rcvcnt++;
909 
910 			/*
911 			 * Don't look for additional matches if this one does
912 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
913 			 * socket options set.  This heuristic avoids searching
914 			 * through all pcbs in the common case of a non-shared
915 			 * port.  It assumes that an application will never
916 			 * clear these options after setting them.
917 			 */
918 			if ((in6p->in6p_socket->so_options &
919 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
920 				break;
921 		}
922 	} else {
923 		/*
924 		 * Locate pcb for datagram.
925 		 */
926 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
927 		    dport, 0);
928 		if (in6p == 0) {
929 			++udpstat.udps_pcbhashmiss;
930 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
931 			if (in6p == 0)
932 				return rcvcnt;
933 		}
934 
935 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
936 		rcvcnt++;
937 	}
938 
939 bad:
940 	return rcvcnt;
941 }
942 #endif
943 
944 #ifdef INET
945 /*
946  * Notify a udp user of an asynchronous error;
947  * just wake up so that he can collect error status.
948  */
949 static void
950 udp_notify(struct inpcb *inp, int errno)
951 {
952 	inp->inp_socket->so_error = errno;
953 	sorwakeup(inp->inp_socket);
954 	sowwakeup(inp->inp_socket);
955 }
956 
957 void *
958 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
959 {
960 	struct ip *ip = v;
961 	struct udphdr *uh;
962 	void (*notify)(struct inpcb *, int) = udp_notify;
963 	int errno;
964 
965 	if (sa->sa_family != AF_INET
966 	 || sa->sa_len != sizeof(struct sockaddr_in))
967 		return NULL;
968 	if ((unsigned)cmd >= PRC_NCMDS)
969 		return NULL;
970 	errno = inetctlerrmap[cmd];
971 	if (PRC_IS_REDIRECT(cmd))
972 		notify = in_rtchange, ip = 0;
973 	else if (cmd == PRC_HOSTDEAD)
974 		ip = 0;
975 	else if (errno == 0)
976 		return NULL;
977 	if (ip) {
978 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
979 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
980 		    ip->ip_src, uh->uh_sport, errno, notify);
981 
982 		/* XXX mapped address case */
983 	} else
984 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
985 		    notify);
986 	return NULL;
987 }
988 
989 int
990 udp_ctloutput(int op, struct socket *so, int level, int optname,
991     struct mbuf **mp)
992 {
993 	int s;
994 	int error = 0;
995 	struct mbuf *m;
996 	struct inpcb *inp;
997 	int family;
998 
999 	family = so->so_proto->pr_domain->dom_family;
1000 
1001 	s = splsoftnet();
1002 	switch (family) {
1003 #ifdef INET
1004 	case PF_INET:
1005 		if (level != IPPROTO_UDP) {
1006 			error = ip_ctloutput(op, so, level, optname, mp);
1007 			goto end;
1008 		}
1009 		break;
1010 #endif
1011 #ifdef INET6
1012 	case PF_INET6:
1013 		if (level != IPPROTO_UDP) {
1014 			error = ip6_ctloutput(op, so, level, optname, mp);
1015 			goto end;
1016 		}
1017 		break;
1018 #endif
1019 	default:
1020 		error = EAFNOSUPPORT;
1021 		goto end;
1022 	}
1023 
1024 
1025 	switch (op) {
1026 	case PRCO_SETOPT:
1027 		m = *mp;
1028 		inp = sotoinpcb(so);
1029 
1030 		switch (optname) {
1031 		case UDP_ENCAP:
1032 			if (m == NULL || m->m_len != sizeof(int)) {
1033 				error = EINVAL;
1034 				break;
1035 			}
1036 
1037 			switch(*mtod(m, int *)) {
1038 #ifdef IPSEC_NAT_T
1039 			case 0:
1040 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1041 				break;
1042 
1043 			case UDP_ENCAP_ESPINUDP:
1044 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1045 				inp->inp_flags |= INP_ESPINUDP;
1046 				break;
1047 
1048 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1049 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1050 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1051 				break;
1052 #endif
1053 			default:
1054 				error = EINVAL;
1055 				break;
1056 			}
1057 			break;
1058 
1059 		default:
1060 			error = ENOPROTOOPT;
1061 			break;
1062 		}
1063 		if (m != NULL) {
1064 			m_free(m);
1065 		}
1066 		break;
1067 
1068 	default:
1069 		error = EINVAL;
1070 		break;
1071 	}
1072 
1073 end:
1074 	splx(s);
1075 	return error;
1076 }
1077 
1078 
1079 int
1080 udp_output(struct mbuf *m, ...)
1081 {
1082 	struct inpcb *inp;
1083 	struct udpiphdr *ui;
1084 	struct route *ro;
1085 	int len = m->m_pkthdr.len;
1086 	int error = 0;
1087 	va_list ap;
1088 
1089 	MCLAIM(m, &udp_tx_mowner);
1090 	va_start(ap, m);
1091 	inp = va_arg(ap, struct inpcb *);
1092 	va_end(ap);
1093 
1094 	/*
1095 	 * Calculate data length and get a mbuf
1096 	 * for UDP and IP headers.
1097 	 */
1098 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1099 	if (m == 0) {
1100 		error = ENOBUFS;
1101 		goto release;
1102 	}
1103 
1104 	/*
1105 	 * Compute the packet length of the IP header, and
1106 	 * punt if the length looks bogus.
1107 	 */
1108 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1109 		error = EMSGSIZE;
1110 		goto release;
1111 	}
1112 
1113 	/*
1114 	 * Fill in mbuf with extended UDP header
1115 	 * and addresses and length put into network format.
1116 	 */
1117 	ui = mtod(m, struct udpiphdr *);
1118 	ui->ui_pr = IPPROTO_UDP;
1119 	ui->ui_src = inp->inp_laddr;
1120 	ui->ui_dst = inp->inp_faddr;
1121 	ui->ui_sport = inp->inp_lport;
1122 	ui->ui_dport = inp->inp_fport;
1123 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1124 
1125 	ro = &inp->inp_route;
1126 
1127 	/*
1128 	 * Set up checksum and output datagram.
1129 	 */
1130 	if (udpcksum) {
1131 		/*
1132 		 * XXX Cache pseudo-header checksum part for
1133 		 * XXX "connected" UDP sockets.
1134 		 */
1135 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1136 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1137 		    sizeof(struct udphdr) + IPPROTO_UDP));
1138 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1139 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1140 	} else
1141 		ui->ui_sum = 0;
1142 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1143 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1144 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1145 	udpstat.udps_opackets++;
1146 
1147 	return (ip_output(m, inp->inp_options, ro,
1148 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1149 	    inp->inp_moptions, inp->inp_socket));
1150 
1151 release:
1152 	m_freem(m);
1153 	return (error);
1154 }
1155 
1156 int	udp_sendspace = 9216;		/* really max datagram size */
1157 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1158 					/* 40 1K datagrams */
1159 
1160 /*ARGSUSED*/
1161 int
1162 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1163 	struct mbuf *control, struct lwp *l)
1164 {
1165 	struct inpcb *inp;
1166 	int s;
1167 	int error = 0;
1168 
1169 	if (req == PRU_CONTROL)
1170 		return (in_control(so, (long)m, (void *)nam,
1171 		    (struct ifnet *)control, l));
1172 
1173 	s = splsoftnet();
1174 
1175 	if (req == PRU_PURGEIF) {
1176 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1177 		in_purgeif((struct ifnet *)control);
1178 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1179 		splx(s);
1180 		return (0);
1181 	}
1182 
1183 	inp = sotoinpcb(so);
1184 #ifdef DIAGNOSTIC
1185 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1186 		panic("udp_usrreq: unexpected control mbuf");
1187 #endif
1188 	if (inp == 0 && req != PRU_ATTACH) {
1189 		error = EINVAL;
1190 		goto release;
1191 	}
1192 
1193 	/*
1194 	 * Note: need to block udp_input while changing
1195 	 * the udp pcb queue and/or pcb addresses.
1196 	 */
1197 	switch (req) {
1198 
1199 	case PRU_ATTACH:
1200 		if (inp != 0) {
1201 			error = EISCONN;
1202 			break;
1203 		}
1204 #ifdef MBUFTRACE
1205 		so->so_mowner = &udp_mowner;
1206 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1207 		so->so_snd.sb_mowner = &udp_tx_mowner;
1208 #endif
1209 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1210 			error = soreserve(so, udp_sendspace, udp_recvspace);
1211 			if (error)
1212 				break;
1213 		}
1214 		error = in_pcballoc(so, &udbtable);
1215 		if (error)
1216 			break;
1217 		inp = sotoinpcb(so);
1218 		inp->inp_ip.ip_ttl = ip_defttl;
1219 		break;
1220 
1221 	case PRU_DETACH:
1222 		in_pcbdetach(inp);
1223 		break;
1224 
1225 	case PRU_BIND:
1226 		error = in_pcbbind(inp, nam, l);
1227 		break;
1228 
1229 	case PRU_LISTEN:
1230 		error = EOPNOTSUPP;
1231 		break;
1232 
1233 	case PRU_CONNECT:
1234 		error = in_pcbconnect(inp, nam, l);
1235 		if (error)
1236 			break;
1237 		soisconnected(so);
1238 		break;
1239 
1240 	case PRU_CONNECT2:
1241 		error = EOPNOTSUPP;
1242 		break;
1243 
1244 	case PRU_DISCONNECT:
1245 		/*soisdisconnected(so);*/
1246 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1247 		in_pcbdisconnect(inp);
1248 		inp->inp_laddr = zeroin_addr;		/* XXX */
1249 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1250 		break;
1251 
1252 	case PRU_SHUTDOWN:
1253 		socantsendmore(so);
1254 		break;
1255 
1256 	case PRU_RCVD:
1257 		error = EOPNOTSUPP;
1258 		break;
1259 
1260 	case PRU_SEND:
1261 		if (control && control->m_len) {
1262 			m_freem(control);
1263 			m_freem(m);
1264 			error = EINVAL;
1265 			break;
1266 		}
1267 	{
1268 		struct in_addr laddr;			/* XXX */
1269 
1270 		if (nam) {
1271 			laddr = inp->inp_laddr;		/* XXX */
1272 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1273 				error = EISCONN;
1274 				goto die;
1275 			}
1276 			error = in_pcbconnect(inp, nam, l);
1277 			if (error)
1278 				goto die;
1279 		} else {
1280 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1281 				error = ENOTCONN;
1282 				goto die;
1283 			}
1284 		}
1285 		error = udp_output(m, inp);
1286 		m = NULL;
1287 		if (nam) {
1288 			in_pcbdisconnect(inp);
1289 			inp->inp_laddr = laddr;		/* XXX */
1290 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1291 		}
1292 	  die:
1293 		if (m)
1294 			m_freem(m);
1295 	}
1296 		break;
1297 
1298 	case PRU_SENSE:
1299 		/*
1300 		 * stat: don't bother with a blocksize.
1301 		 */
1302 		splx(s);
1303 		return (0);
1304 
1305 	case PRU_RCVOOB:
1306 		error =  EOPNOTSUPP;
1307 		break;
1308 
1309 	case PRU_SENDOOB:
1310 		m_freem(control);
1311 		m_freem(m);
1312 		error =  EOPNOTSUPP;
1313 		break;
1314 
1315 	case PRU_SOCKADDR:
1316 		in_setsockaddr(inp, nam);
1317 		break;
1318 
1319 	case PRU_PEERADDR:
1320 		in_setpeeraddr(inp, nam);
1321 		break;
1322 
1323 	default:
1324 		panic("udp_usrreq");
1325 	}
1326 
1327 release:
1328 	splx(s);
1329 	return (error);
1330 }
1331 
1332 /*
1333  * Sysctl for udp variables.
1334  */
1335 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1336 {
1337 
1338 	sysctl_createv(clog, 0, NULL, NULL,
1339 		       CTLFLAG_PERMANENT,
1340 		       CTLTYPE_NODE, "net", NULL,
1341 		       NULL, 0, NULL, 0,
1342 		       CTL_NET, CTL_EOL);
1343 	sysctl_createv(clog, 0, NULL, NULL,
1344 		       CTLFLAG_PERMANENT,
1345 		       CTLTYPE_NODE, "inet", NULL,
1346 		       NULL, 0, NULL, 0,
1347 		       CTL_NET, PF_INET, CTL_EOL);
1348 	sysctl_createv(clog, 0, NULL, NULL,
1349 		       CTLFLAG_PERMANENT,
1350 		       CTLTYPE_NODE, "udp",
1351 		       SYSCTL_DESCR("UDPv4 related settings"),
1352 		       NULL, 0, NULL, 0,
1353 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1354 
1355 	sysctl_createv(clog, 0, NULL, NULL,
1356 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1357 		       CTLTYPE_INT, "checksum",
1358 		       SYSCTL_DESCR("Compute UDP checksums"),
1359 		       NULL, 0, &udpcksum, 0,
1360 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1361 		       CTL_EOL);
1362 	sysctl_createv(clog, 0, NULL, NULL,
1363 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1364 		       CTLTYPE_INT, "sendspace",
1365 		       SYSCTL_DESCR("Default UDP send buffer size"),
1366 		       NULL, 0, &udp_sendspace, 0,
1367 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1368 		       CTL_EOL);
1369 	sysctl_createv(clog, 0, NULL, NULL,
1370 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1371 		       CTLTYPE_INT, "recvspace",
1372 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1373 		       NULL, 0, &udp_recvspace, 0,
1374 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1375 		       CTL_EOL);
1376 	sysctl_createv(clog, 0, NULL, NULL,
1377 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1378 		       CTLTYPE_INT, "do_loopback_cksum",
1379 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1380 		       NULL, 0, &udp_do_loopback_cksum, 0,
1381 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1382 		       CTL_EOL);
1383 	sysctl_createv(clog, 0, NULL, NULL,
1384 		       CTLFLAG_PERMANENT,
1385 		       CTLTYPE_STRUCT, "pcblist",
1386 		       SYSCTL_DESCR("UDP protocol control block list"),
1387 		       sysctl_inpcblist, 0, &udbtable, 0,
1388 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1389 		       CTL_EOL);
1390 	sysctl_createv(clog, 0, NULL, NULL,
1391 		       CTLFLAG_PERMANENT,
1392 		       CTLTYPE_STRUCT, "stats",
1393 		       SYSCTL_DESCR("UDP statistics"),
1394 		       NULL, 0, &udpstat, sizeof(udpstat),
1395 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1396 		       CTL_EOL);
1397 }
1398 #endif
1399 
1400 #if (defined INET && defined IPSEC_NAT_T)
1401 /*
1402  * Returns:
1403  * 1 if the packet was processed
1404  * 0 if normal UDP processing should take place
1405  * -1 if an error occurent and m was freed
1406  */
1407 static int
1408 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1409     struct socket *so)
1410 {
1411 	size_t len;
1412 	void *data;
1413 	struct inpcb *inp;
1414 	size_t skip = 0;
1415 	size_t minlen;
1416 	size_t iphdrlen;
1417 	struct ip *ip;
1418 	struct mbuf *n;
1419 	struct m_tag *tag;
1420 	struct udphdr *udphdr;
1421 	u_int16_t sport, dport;
1422 	struct mbuf *m = *mp;
1423 
1424 	/*
1425 	 * Collapse the mbuf chain if the first mbuf is too short
1426 	 * The longest case is: UDP + non ESP marker + ESP
1427 	 */
1428 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1429 	if (minlen > m->m_pkthdr.len)
1430 		minlen = m->m_pkthdr.len;
1431 
1432 	if (m->m_len < minlen) {
1433 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1434 			printf("udp4_espinudp: m_pullup failed\n");
1435 			return -1;
1436 		}
1437 		m = *mp;
1438 	}
1439 
1440 	len = m->m_len - off;
1441 	data = mtod(m, char *) + off;
1442 	inp = sotoinpcb(so);
1443 
1444 	/* Ignore keepalive packets */
1445 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1446 		return 1;
1447 	}
1448 
1449 	/*
1450 	 * Check that the payload is long enough to hold
1451 	 * an ESP header and compute the length of encapsulation
1452 	 * header to remove
1453 	 */
1454 	if (inp->inp_flags & INP_ESPINUDP) {
1455 		u_int32_t *st = (u_int32_t *)data;
1456 
1457 		if ((len <= sizeof(struct esp)) || (*st == 0))
1458 			return 0; /* Normal UDP processing */
1459 
1460 		skip = sizeof(struct udphdr);
1461 	}
1462 
1463 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1464 		u_int32_t *st = (u_int32_t *)data;
1465 
1466 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1467 		    || ((st[0] | st[1]) != 0))
1468 			return 0; /* Normal UDP processing */
1469 
1470 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1471 	}
1472 
1473 	/*
1474 	 * Get the UDP ports. They are handled in network
1475 	 * order everywhere in IPSEC_NAT_T code.
1476 	 */
1477 	udphdr = (struct udphdr *)((char *)data - skip);
1478 	sport = udphdr->uh_sport;
1479 	dport = udphdr->uh_dport;
1480 
1481 	/*
1482 	 * Remove the UDP header (and possibly the non ESP marker)
1483 	 * IP header lendth is iphdrlen
1484 	 * Before:
1485 	 *   <--- off --->
1486 	 *   +----+------+-----+
1487 	 *   | IP |  UDP | ESP |
1488 	 *   +----+------+-----+
1489 	 *        <-skip->
1490 	 * After:
1491 	 *          +----+-----+
1492 	 *          | IP | ESP |
1493 	 *          +----+-----+
1494 	 *   <-skip->
1495 	 */
1496 	iphdrlen = off - sizeof(struct udphdr);
1497 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1498 	m_adj(m, skip);
1499 
1500 	ip = mtod(m, struct ip *);
1501 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1502 	ip->ip_p = IPPROTO_ESP;
1503 
1504 	/*
1505 	 * Copy the mbuf to avoid multiple free, as both
1506 	 * esp4_input (which we call) and udp_input (which
1507 	 * called us) free the mbuf.
1508 	 */
1509 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1510 		printf("udp4_espinudp: m_dup failed\n");
1511 		return 0;
1512 	}
1513 
1514 	/*
1515 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1516 	 * the source UDP port. This is required if we want
1517 	 * to select the right SPD for multiple hosts behind
1518 	 * same NAT
1519 	 */
1520 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1521 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1522 		printf("udp4_espinudp: m_tag_get failed\n");
1523 		m_freem(n);
1524 		return 0;
1525 	}
1526 	((u_int16_t *)(tag + 1))[0] = sport;
1527 	((u_int16_t *)(tag + 1))[1] = dport;
1528 	m_tag_prepend(n, tag);
1529 
1530 #ifdef FAST_IPSEC
1531 	ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1532 #else
1533 	esp4_input(n, iphdrlen);
1534 #endif
1535 
1536 	/* We handled it, it shoudln't be handled by UDP */
1537 	return 1;
1538 }
1539 #endif
1540