xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 4b896b232495b7a9b8b94a1cf1e21873296d53b8)
1 /*	$NetBSD: udp_usrreq.c,v 1.122 2004/05/25 04:34:00 atatat 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 1.122 2004/05/25 04:34:00 atatat 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 #endif
105 
106 #ifndef INET6
107 /* always need ip6.h for IP6_EXTHDR_GET */
108 #include <netinet/ip6.h>
109 #endif
110 
111 #include "faith.h"
112 #if defined(NFAITH) && NFAITH > 0
113 #include <net/if_faith.h>
114 #endif
115 
116 #include <machine/stdarg.h>
117 
118 #ifdef FAST_IPSEC
119 #include <netipsec/ipsec.h>
120 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
121 #ifdef INET6
122 #include <netipsec/ipsec6.h>
123 #endif
124 #endif	/* FAST_IPSEC*/
125 
126 #ifdef IPSEC
127 #include <netinet6/ipsec.h>
128 #include <netkey/key.h>
129 #endif /*IPSEC*/
130 
131 #ifdef IPKDB
132 #include <ipkdb/ipkdb.h>
133 #endif
134 
135 /*
136  * UDP protocol implementation.
137  * Per RFC 768, August, 1980.
138  */
139 #ifndef	COMPAT_42
140 int	udpcksum = 1;
141 #else
142 int	udpcksum = 0;		/* XXX */
143 #endif
144 
145 struct	inpcbtable udbtable;
146 struct	udpstat udpstat;
147 
148 #ifdef INET
149 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
150 	struct socket *);
151 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
152 	struct mbuf *, int);
153 #endif
154 #ifdef INET6
155 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
156 	struct socket *);
157 static int udp6_realinput (int, struct sockaddr_in6 *,
158 	struct sockaddr_in6 *, struct mbuf *, int);
159 #endif
160 #ifdef INET
161 static	void udp_notify (struct inpcb *, int);
162 #endif
163 
164 #ifndef UDBHASHSIZE
165 #define	UDBHASHSIZE	128
166 #endif
167 int	udbhashsize = UDBHASHSIZE;
168 
169 #ifdef MBUFTRACE
170 struct mowner udp_mowner = { "udp" };
171 struct mowner udp_rx_mowner = { "udp", "rx" };
172 struct mowner udp_tx_mowner = { "udp", "tx" };
173 #endif
174 
175 #ifdef UDP_CSUM_COUNTERS
176 #include <sys/device.h>
177 
178 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
179     NULL, "udp", "hwcsum bad");
180 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
181     NULL, "udp", "hwcsum ok");
182 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
183     NULL, "udp", "hwcsum data");
184 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
185     NULL, "udp", "swcsum");
186 
187 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
188 
189 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
190 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
191 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
192 EVCNT_ATTACH_STATIC(udp_swcsum);
193 
194 #else
195 
196 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
197 
198 #endif /* UDP_CSUM_COUNTERS */
199 
200 void
201 udp_init(void)
202 {
203 
204 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
205 
206 	MOWNER_ATTACH(&udp_tx_mowner);
207 	MOWNER_ATTACH(&udp_rx_mowner);
208 	MOWNER_ATTACH(&udp_mowner);
209 }
210 
211 #ifdef INET
212 void
213 udp_input(struct mbuf *m, ...)
214 {
215 	va_list ap;
216 	struct sockaddr_in src, dst;
217 	struct ip *ip;
218 	struct udphdr *uh;
219 	int iphlen;
220 	int len;
221 	int n;
222 	u_int16_t ip_len;
223 
224 	va_start(ap, m);
225 	iphlen = va_arg(ap, int);
226 	(void)va_arg(ap, int);		/* ignore value, advance ap */
227 	va_end(ap);
228 
229 	MCLAIM(m, &udp_rx_mowner);
230 	udpstat.udps_ipackets++;
231 
232 	/*
233 	 * Get IP and UDP header together in first mbuf.
234 	 */
235 	ip = mtod(m, struct ip *);
236 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
237 	if (uh == NULL) {
238 		udpstat.udps_hdrops++;
239 		return;
240 	}
241 	KASSERT(UDP_HDR_ALIGNED_P(uh));
242 
243 	/* destination port of 0 is illegal, based on RFC768. */
244 	if (uh->uh_dport == 0)
245 		goto bad;
246 
247 	/*
248 	 * Make mbuf data length reflect UDP length.
249 	 * If not enough data to reflect UDP length, drop.
250 	 */
251 	ip_len = ntohs(ip->ip_len);
252 	len = ntohs((u_int16_t)uh->uh_ulen);
253 	if (ip_len != iphlen + len) {
254 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
255 			udpstat.udps_badlen++;
256 			goto bad;
257 		}
258 		m_adj(m, iphlen + len - ip_len);
259 	}
260 
261 	/*
262 	 * Checksum extended UDP header and data.
263 	 */
264 	if (uh->uh_sum) {
265 		switch (m->m_pkthdr.csum_flags &
266 		    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
267 		    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
268 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
269 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
270 			goto badcsum;
271 
272 		case M_CSUM_UDPv4|M_CSUM_DATA: {
273 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
274 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
275 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR)
276 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
277 				    ip->ip_dst.s_addr,
278 				    htons(hw_csum + len + IPPROTO_UDP));
279 			if ((hw_csum ^ 0xffff) != 0)
280 				goto badcsum;
281 			break;
282 		}
283 
284 		case M_CSUM_UDPv4:
285 			/* Checksum was okay. */
286 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
287 			break;
288 
289 		default:
290 			/* Need to compute it ourselves. */
291 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
292 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
293 				goto badcsum;
294 			break;
295 		}
296 	}
297 
298 	/* construct source and dst sockaddrs. */
299 	bzero(&src, sizeof(src));
300 	src.sin_family = AF_INET;
301 	src.sin_len = sizeof(struct sockaddr_in);
302 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
303 	src.sin_port = uh->uh_sport;
304 	bzero(&dst, sizeof(dst));
305 	dst.sin_family = AF_INET;
306 	dst.sin_len = sizeof(struct sockaddr_in);
307 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
308 	dst.sin_port = uh->uh_dport;
309 
310 	n = udp4_realinput(&src, &dst, m, iphlen);
311 #ifdef INET6
312 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
313 		struct sockaddr_in6 src6, dst6;
314 
315 		bzero(&src6, sizeof(src6));
316 		src6.sin6_family = AF_INET6;
317 		src6.sin6_len = sizeof(struct sockaddr_in6);
318 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
319 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
320 			sizeof(ip->ip_src));
321 		src6.sin6_port = uh->uh_sport;
322 		bzero(&dst6, sizeof(dst6));
323 		dst6.sin6_family = AF_INET6;
324 		dst6.sin6_len = sizeof(struct sockaddr_in6);
325 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
326 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
327 			sizeof(ip->ip_dst));
328 		dst6.sin6_port = uh->uh_dport;
329 
330 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
331 	}
332 #endif
333 
334 	if (n == 0) {
335 		if (m->m_flags & (M_BCAST | M_MCAST)) {
336 			udpstat.udps_noportbcast++;
337 			goto bad;
338 		}
339 		udpstat.udps_noport++;
340 #ifdef IPKDB
341 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
342 				m, iphlen + sizeof(struct udphdr),
343 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
344 			/*
345 			 * It was a debugger connect packet,
346 			 * just drop it now
347 			 */
348 			goto bad;
349 		}
350 #endif
351 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
352 		m = NULL;
353 	}
354 
355 bad:
356 	if (m)
357 		m_freem(m);
358 	return;
359 
360 badcsum:
361 	m_freem(m);
362 	udpstat.udps_badsum++;
363 }
364 #endif
365 
366 #ifdef INET6
367 int
368 udp6_input(struct mbuf **mp, int *offp, int proto)
369 {
370 	struct mbuf *m = *mp;
371 	int off = *offp;
372 	struct sockaddr_in6 src, dst;
373 	struct ip6_hdr *ip6;
374 	struct udphdr *uh;
375 	u_int32_t plen, ulen;
376 
377 	ip6 = mtod(m, struct ip6_hdr *);
378 
379 #if defined(NFAITH) && 0 < NFAITH
380 	if (faithprefix(&ip6->ip6_dst)) {
381 		/* send icmp6 host unreach? */
382 		m_freem(m);
383 		return IPPROTO_DONE;
384 	}
385 #endif
386 
387 	udp6stat.udp6s_ipackets++;
388 
389 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
390 	plen = m->m_pkthdr.len - off;
391 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
392 	if (uh == NULL) {
393 		ip6stat.ip6s_tooshort++;
394 		return IPPROTO_DONE;
395 	}
396 	KASSERT(UDP_HDR_ALIGNED_P(uh));
397 	ulen = ntohs((u_short)uh->uh_ulen);
398 	/*
399 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
400 	 * iff payload length > 0xffff.
401 	 */
402 	if (ulen == 0 && plen > 0xffff)
403 		ulen = plen;
404 
405 	if (plen != ulen) {
406 		udp6stat.udp6s_badlen++;
407 		goto bad;
408 	}
409 
410 	/* destination port of 0 is illegal, based on RFC768. */
411 	if (uh->uh_dport == 0)
412 		goto bad;
413 
414 	/* Be proactive about malicious use of IPv4 mapped address */
415 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
416 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
417 		/* XXX stat */
418 		goto bad;
419 	}
420 
421 	/*
422 	 * Checksum extended UDP header and data.
423 	 */
424 	if (uh->uh_sum == 0) {
425 		udp6stat.udp6s_nosum++;
426 		goto bad;
427 	}
428 	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
429 		udp6stat.udp6s_badsum++;
430 		goto bad;
431 	}
432 
433 	/*
434 	 * Construct source and dst sockaddrs.
435 	 * Note that ifindex (s6_addr16[1]) is already filled.
436 	 */
437 	bzero(&src, sizeof(src));
438 	src.sin6_family = AF_INET6;
439 	src.sin6_len = sizeof(struct sockaddr_in6);
440 	/* KAME hack: recover scopeid */
441 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
442 	src.sin6_port = uh->uh_sport;
443 	bzero(&dst, sizeof(dst));
444 	dst.sin6_family = AF_INET6;
445 	dst.sin6_len = sizeof(struct sockaddr_in6);
446 	/* KAME hack: recover scopeid */
447 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
448 	dst.sin6_port = uh->uh_dport;
449 
450 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
451 		if (m->m_flags & M_MCAST) {
452 			udp6stat.udp6s_noportmcast++;
453 			goto bad;
454 		}
455 		udp6stat.udp6s_noport++;
456 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
457 		m = NULL;
458 	}
459 
460 bad:
461 	if (m)
462 		m_freem(m);
463 	return IPPROTO_DONE;
464 }
465 #endif
466 
467 #ifdef INET
468 static void
469 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
470 	struct sockaddr *src, struct socket *so)
471 {
472 	struct mbuf *opts = NULL;
473 	struct mbuf *n;
474 	struct inpcb *inp = NULL;
475 
476 	if (!so)
477 		return;
478 	switch (so->so_proto->pr_domain->dom_family) {
479 	case AF_INET:
480 		inp = sotoinpcb(so);
481 		break;
482 #ifdef INET6
483 	case AF_INET6:
484 		break;
485 #endif
486 	default:
487 		return;
488 	}
489 
490 #if defined(IPSEC) || defined(FAST_IPSEC)
491 	/* check AH/ESP integrity. */
492 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
493 		ipsecstat.in_polvio++;
494 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
495 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
496 			    0, 0);
497 		return;
498 	}
499 #endif /*IPSEC*/
500 
501 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
502 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
503 			 || so->so_options & SO_TIMESTAMP)) {
504 			struct ip *ip = mtod(n, struct ip *);
505 			ip_savecontrol(inp, &opts, ip, n);
506 		}
507 
508 		m_adj(n, off);
509 		if (sbappendaddr(&so->so_rcv, src, n,
510 				opts) == 0) {
511 			m_freem(n);
512 			if (opts)
513 				m_freem(opts);
514 			udpstat.udps_fullsock++;
515 		} else
516 			sorwakeup(so);
517 	}
518 }
519 #endif
520 
521 #ifdef INET6
522 static void
523 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
524 	struct sockaddr *src, struct socket *so)
525 {
526 	struct mbuf *opts = NULL;
527 	struct mbuf *n;
528 	struct in6pcb *in6p = NULL;
529 
530 	if (!so)
531 		return;
532 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
533 		return;
534 	in6p = sotoin6pcb(so);
535 
536 #if defined(IPSEC) || defined(FAST_IPSEC)
537 	/* check AH/ESP integrity. */
538 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
539 		ipsec6stat.in_polvio++;
540 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
541 			icmp6_error(n, ICMP6_DST_UNREACH,
542 			    ICMP6_DST_UNREACH_ADMIN, 0);
543 		return;
544 	}
545 #endif /*IPSEC*/
546 
547 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
548 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
549 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
550 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
551 			ip6_savecontrol(in6p, &opts, ip6, n);
552 		}
553 
554 		m_adj(n, off);
555 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
556 			m_freem(n);
557 			if (opts)
558 				m_freem(opts);
559 			udp6stat.udp6s_fullsock++;
560 		} else
561 			sorwakeup(so);
562 	}
563 }
564 #endif
565 
566 #ifdef INET
567 static int
568 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
569 	struct mbuf *m, int off /* offset of udphdr */)
570 {
571 	u_int16_t *sport, *dport;
572 	int rcvcnt;
573 	struct in_addr *src4, *dst4;
574 	struct inpcb_hdr *inph;
575 	struct inpcb *inp;
576 
577 	rcvcnt = 0;
578 	off += sizeof(struct udphdr);	/* now, offset of payload */
579 
580 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
581 		goto bad;
582 
583 	src4 = &src->sin_addr;
584 	sport = &src->sin_port;
585 	dst4 = &dst->sin_addr;
586 	dport = &dst->sin_port;
587 
588 	if (IN_MULTICAST(dst4->s_addr) ||
589 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
590 		/*
591 		 * Deliver a multicast or broadcast datagram to *all* sockets
592 		 * for which the local and remote addresses and ports match
593 		 * those of the incoming datagram.  This allows more than
594 		 * one process to receive multi/broadcasts on the same port.
595 		 * (This really ought to be done for unicast datagrams as
596 		 * well, but that would cause problems with existing
597 		 * applications that open both address-specific sockets and
598 		 * a wildcard socket listening to the same port -- they would
599 		 * end up receiving duplicates of every unicast datagram.
600 		 * Those applications open the multiple sockets to overcome an
601 		 * inadequacy of the UDP socket interface, but for backwards
602 		 * compatibility we avoid the problem here rather than
603 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
604 		 */
605 
606 		/*
607 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
608 		 * we need udpiphdr for IPsec processing so we do that later.
609 		 */
610 		/*
611 		 * Locate pcb(s) for datagram.
612 		 */
613 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
614 			inp = (struct inpcb *)inph;
615 			if (inp->inp_af != AF_INET)
616 				continue;
617 
618 			if (inp->inp_lport != *dport)
619 				continue;
620 			if (!in_nullhost(inp->inp_laddr)) {
621 				if (!in_hosteq(inp->inp_laddr, *dst4))
622 					continue;
623 			}
624 			if (!in_nullhost(inp->inp_faddr)) {
625 				if (!in_hosteq(inp->inp_faddr, *src4) ||
626 				    inp->inp_fport != *sport)
627 					continue;
628 			}
629 
630 			udp4_sendup(m, off, (struct sockaddr *)src,
631 				inp->inp_socket);
632 			rcvcnt++;
633 
634 			/*
635 			 * Don't look for additional matches if this one does
636 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
637 			 * socket options set.  This heuristic avoids searching
638 			 * through all pcbs in the common case of a non-shared
639 			 * port.  It assumes that an application will never
640 			 * clear these options after setting them.
641 			 */
642 			if ((inp->inp_socket->so_options &
643 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
644 				break;
645 		}
646 	} else {
647 		/*
648 		 * Locate pcb for datagram.
649 		 */
650 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
651 		if (inp == 0) {
652 			++udpstat.udps_pcbhashmiss;
653 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
654 			if (inp == 0)
655 				return rcvcnt;
656 		}
657 
658 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
659 		rcvcnt++;
660 	}
661 
662 bad:
663 	return rcvcnt;
664 }
665 #endif
666 
667 #ifdef INET6
668 static int
669 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
670 	struct mbuf *m, int off)
671 {
672 	u_int16_t sport, dport;
673 	int rcvcnt;
674 	struct in6_addr src6, dst6;
675 	const struct in_addr *dst4;
676 	struct inpcb_hdr *inph;
677 	struct in6pcb *in6p;
678 
679 	rcvcnt = 0;
680 	off += sizeof(struct udphdr);	/* now, offset of payload */
681 
682 	if (af != AF_INET && af != AF_INET6)
683 		goto bad;
684 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
685 		goto bad;
686 
687 	in6_embedscope(&src6, src, NULL, NULL);
688 	sport = src->sin6_port;
689 	in6_embedscope(&dst6, dst, NULL, NULL);
690 	dport = dst->sin6_port;
691 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
692 
693 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
694 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
695 		/*
696 		 * Deliver a multicast or broadcast datagram to *all* sockets
697 		 * for which the local and remote addresses and ports match
698 		 * those of the incoming datagram.  This allows more than
699 		 * one process to receive multi/broadcasts on the same port.
700 		 * (This really ought to be done for unicast datagrams as
701 		 * well, but that would cause problems with existing
702 		 * applications that open both address-specific sockets and
703 		 * a wildcard socket listening to the same port -- they would
704 		 * end up receiving duplicates of every unicast datagram.
705 		 * Those applications open the multiple sockets to overcome an
706 		 * inadequacy of the UDP socket interface, but for backwards
707 		 * compatibility we avoid the problem here rather than
708 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
709 		 */
710 
711 		/*
712 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
713 		 * we need udpiphdr for IPsec processing so we do that later.
714 		 */
715 		/*
716 		 * Locate pcb(s) for datagram.
717 		 */
718 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
719 			in6p = (struct in6pcb *)inph;
720 			if (in6p->in6p_af != AF_INET6)
721 				continue;
722 
723 			if (in6p->in6p_lport != dport)
724 				continue;
725 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
726 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
727 					continue;
728 			} else {
729 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
730 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
731 					continue;
732 			}
733 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
734 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
735 				    &src6) || in6p->in6p_fport != sport)
736 					continue;
737 			} else {
738 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
739 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
740 					continue;
741 			}
742 
743 			udp6_sendup(m, off, (struct sockaddr *)src,
744 				in6p->in6p_socket);
745 			rcvcnt++;
746 
747 			/*
748 			 * Don't look for additional matches if this one does
749 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
750 			 * socket options set.  This heuristic avoids searching
751 			 * through all pcbs in the common case of a non-shared
752 			 * port.  It assumes that an application will never
753 			 * clear these options after setting them.
754 			 */
755 			if ((in6p->in6p_socket->so_options &
756 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
757 				break;
758 		}
759 	} else {
760 		/*
761 		 * Locate pcb for datagram.
762 		 */
763 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
764 		    &dst6, dport, 0);
765 		if (in6p == 0) {
766 			++udpstat.udps_pcbhashmiss;
767 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
768 			if (in6p == 0)
769 				return rcvcnt;
770 		}
771 
772 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
773 		rcvcnt++;
774 	}
775 
776 bad:
777 	return rcvcnt;
778 }
779 #endif
780 
781 #ifdef INET
782 /*
783  * Notify a udp user of an asynchronous error;
784  * just wake up so that he can collect error status.
785  */
786 static void
787 udp_notify(struct inpcb *inp, int errno)
788 {
789 	inp->inp_socket->so_error = errno;
790 	sorwakeup(inp->inp_socket);
791 	sowwakeup(inp->inp_socket);
792 }
793 
794 void *
795 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
796 {
797 	struct ip *ip = v;
798 	struct udphdr *uh;
799 	void (*notify)(struct inpcb *, int) = udp_notify;
800 	int errno;
801 
802 	if (sa->sa_family != AF_INET
803 	 || sa->sa_len != sizeof(struct sockaddr_in))
804 		return NULL;
805 	if ((unsigned)cmd >= PRC_NCMDS)
806 		return NULL;
807 	errno = inetctlerrmap[cmd];
808 	if (PRC_IS_REDIRECT(cmd))
809 		notify = in_rtchange, ip = 0;
810 	else if (cmd == PRC_HOSTDEAD)
811 		ip = 0;
812 	else if (errno == 0)
813 		return NULL;
814 	if (ip) {
815 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
816 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
817 		    ip->ip_src, uh->uh_sport, errno, notify);
818 
819 		/* XXX mapped address case */
820 	} else
821 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
822 		    notify);
823 	return NULL;
824 }
825 
826 int
827 udp_output(struct mbuf *m, ...)
828 {
829 	struct inpcb *inp;
830 	struct udpiphdr *ui;
831 	int len = m->m_pkthdr.len;
832 	int error = 0;
833 	va_list ap;
834 
835 	MCLAIM(m, &udp_tx_mowner);
836 	va_start(ap, m);
837 	inp = va_arg(ap, struct inpcb *);
838 	va_end(ap);
839 
840 	/*
841 	 * Calculate data length and get a mbuf
842 	 * for UDP and IP headers.
843 	 */
844 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
845 	if (m == 0) {
846 		error = ENOBUFS;
847 		goto release;
848 	}
849 
850 	/*
851 	 * Compute the packet length of the IP header, and
852 	 * punt if the length looks bogus.
853 	 */
854 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
855 		error = EMSGSIZE;
856 		goto release;
857 	}
858 
859 	/*
860 	 * Fill in mbuf with extended UDP header
861 	 * and addresses and length put into network format.
862 	 */
863 	ui = mtod(m, struct udpiphdr *);
864 	ui->ui_pr = IPPROTO_UDP;
865 	ui->ui_src = inp->inp_laddr;
866 	ui->ui_dst = inp->inp_faddr;
867 	ui->ui_sport = inp->inp_lport;
868 	ui->ui_dport = inp->inp_fport;
869 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
870 
871 	/*
872 	 * Set up checksum and output datagram.
873 	 */
874 	if (udpcksum) {
875 		/*
876 		 * XXX Cache pseudo-header checksum part for
877 		 * XXX "connected" UDP sockets.
878 		 */
879 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
880 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
881 		    sizeof(struct udphdr) + IPPROTO_UDP));
882 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
883 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
884 	} else
885 		ui->ui_sum = 0;
886 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
887 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
888 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
889 	udpstat.udps_opackets++;
890 
891 	return (ip_output(m, inp->inp_options, &inp->inp_route,
892 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
893 	    inp->inp_moptions, inp->inp_socket));
894 
895 release:
896 	m_freem(m);
897 	return (error);
898 }
899 
900 int	udp_sendspace = 9216;		/* really max datagram size */
901 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
902 					/* 40 1K datagrams */
903 
904 /*ARGSUSED*/
905 int
906 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
907 	struct mbuf *control, struct proc *p)
908 {
909 	struct inpcb *inp;
910 	int s;
911 	int error = 0;
912 
913 	if (req == PRU_CONTROL)
914 		return (in_control(so, (long)m, (caddr_t)nam,
915 		    (struct ifnet *)control, p));
916 
917 	if (req == PRU_PURGEIF) {
918 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
919 		in_purgeif((struct ifnet *)control);
920 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
921 		return (0);
922 	}
923 
924 	s = splsoftnet();
925 	inp = sotoinpcb(so);
926 #ifdef DIAGNOSTIC
927 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
928 		panic("udp_usrreq: unexpected control mbuf");
929 #endif
930 	if (inp == 0 && req != PRU_ATTACH) {
931 		error = EINVAL;
932 		goto release;
933 	}
934 
935 	/*
936 	 * Note: need to block udp_input while changing
937 	 * the udp pcb queue and/or pcb addresses.
938 	 */
939 	switch (req) {
940 
941 	case PRU_ATTACH:
942 		if (inp != 0) {
943 			error = EISCONN;
944 			break;
945 		}
946 #ifdef MBUFTRACE
947 		so->so_mowner = &udp_mowner;
948 		so->so_rcv.sb_mowner = &udp_rx_mowner;
949 		so->so_snd.sb_mowner = &udp_tx_mowner;
950 #endif
951 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
952 			error = soreserve(so, udp_sendspace, udp_recvspace);
953 			if (error)
954 				break;
955 		}
956 		error = in_pcballoc(so, &udbtable);
957 		if (error)
958 			break;
959 		inp = sotoinpcb(so);
960 		inp->inp_ip.ip_ttl = ip_defttl;
961 		break;
962 
963 	case PRU_DETACH:
964 		in_pcbdetach(inp);
965 		break;
966 
967 	case PRU_BIND:
968 		error = in_pcbbind(inp, nam, p);
969 		break;
970 
971 	case PRU_LISTEN:
972 		error = EOPNOTSUPP;
973 		break;
974 
975 	case PRU_CONNECT:
976 		error = in_pcbconnect(inp, nam);
977 		if (error)
978 			break;
979 		soisconnected(so);
980 		break;
981 
982 	case PRU_CONNECT2:
983 		error = EOPNOTSUPP;
984 		break;
985 
986 	case PRU_DISCONNECT:
987 		/*soisdisconnected(so);*/
988 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
989 		in_pcbdisconnect(inp);
990 		inp->inp_laddr = zeroin_addr;		/* XXX */
991 		in_pcbstate(inp, INP_BOUND);		/* XXX */
992 		break;
993 
994 	case PRU_SHUTDOWN:
995 		socantsendmore(so);
996 		break;
997 
998 	case PRU_RCVD:
999 		error = EOPNOTSUPP;
1000 		break;
1001 
1002 	case PRU_SEND:
1003 		if (control && control->m_len) {
1004 			m_freem(control);
1005 			m_freem(m);
1006 			error = EINVAL;
1007 			break;
1008 		}
1009 	{
1010 		struct in_addr laddr;			/* XXX */
1011 
1012 		if (nam) {
1013 			laddr = inp->inp_laddr;		/* XXX */
1014 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1015 				error = EISCONN;
1016 				goto die;
1017 			}
1018 			error = in_pcbconnect(inp, nam);
1019 			if (error)
1020 				goto die;
1021 		} else {
1022 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1023 				error = ENOTCONN;
1024 				goto die;
1025 			}
1026 		}
1027 		error = udp_output(m, inp);
1028 		m = NULL;
1029 		if (nam) {
1030 			in_pcbdisconnect(inp);
1031 			inp->inp_laddr = laddr;		/* XXX */
1032 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1033 		}
1034 	  die:
1035 		if (m)
1036 			m_freem(m);
1037 	}
1038 		break;
1039 
1040 	case PRU_SENSE:
1041 		/*
1042 		 * stat: don't bother with a blocksize.
1043 		 */
1044 		splx(s);
1045 		return (0);
1046 
1047 	case PRU_RCVOOB:
1048 		error =  EOPNOTSUPP;
1049 		break;
1050 
1051 	case PRU_SENDOOB:
1052 		m_freem(control);
1053 		m_freem(m);
1054 		error =  EOPNOTSUPP;
1055 		break;
1056 
1057 	case PRU_SOCKADDR:
1058 		in_setsockaddr(inp, nam);
1059 		break;
1060 
1061 	case PRU_PEERADDR:
1062 		in_setpeeraddr(inp, nam);
1063 		break;
1064 
1065 	default:
1066 		panic("udp_usrreq");
1067 	}
1068 
1069 release:
1070 	splx(s);
1071 	return (error);
1072 }
1073 
1074 /*
1075  * Sysctl for udp variables.
1076  */
1077 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1078 {
1079 
1080 	sysctl_createv(clog, 0, NULL, NULL,
1081 		       CTLFLAG_PERMANENT,
1082 		       CTLTYPE_NODE, "net", NULL,
1083 		       NULL, 0, NULL, 0,
1084 		       CTL_NET, CTL_EOL);
1085 	sysctl_createv(clog, 0, NULL, NULL,
1086 		       CTLFLAG_PERMANENT,
1087 		       CTLTYPE_NODE, "inet", NULL,
1088 		       NULL, 0, NULL, 0,
1089 		       CTL_NET, PF_INET, CTL_EOL);
1090 	sysctl_createv(clog, 0, NULL, NULL,
1091 		       CTLFLAG_PERMANENT,
1092 		       CTLTYPE_NODE, "udp",
1093 		       SYSCTL_DESCR("UDPv4 related settings"),
1094 		       NULL, 0, NULL, 0,
1095 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1096 
1097 	sysctl_createv(clog, 0, NULL, NULL,
1098 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1099 		       CTLTYPE_INT, "checksum",
1100 		       SYSCTL_DESCR("Compute and check UDP checksums"),
1101 		       NULL, 0, &udpcksum, 0,
1102 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1103 		       CTL_EOL);
1104 	sysctl_createv(clog, 0, NULL, NULL,
1105 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1106 		       CTLTYPE_INT, "sendspace",
1107 		       SYSCTL_DESCR("Default UDP send buffer size"),
1108 		       NULL, 0, &udp_sendspace, 0,
1109 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1110 		       CTL_EOL);
1111 	sysctl_createv(clog, 0, NULL, NULL,
1112 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1113 		       CTLTYPE_INT, "recvspace",
1114 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1115 		       NULL, 0, &udp_recvspace, 0,
1116 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1117 		       CTL_EOL);
1118 }
1119 #endif
1120