xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 8e6f7afb5bba2c07fd084c08249718961dfbc1d1)
1 /*	$NetBSD: udp_usrreq.c,v 1.91 2001/11/13 00:32:42 lukem 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. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgement:
46  *	This product includes software developed by the University of
47  *	California, Berkeley and its contributors.
48  * 4. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
65  */
66 
67 #include <sys/cdefs.h>
68 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.91 2001/11/13 00:32:42 lukem Exp $");
69 
70 #include "opt_inet.h"
71 #include "opt_ipsec.h"
72 #include "opt_inet_csum.h"
73 #include "opt_ipkdb.h"
74 
75 #include <sys/param.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/stat.h>
83 #include <sys/systm.h>
84 #include <sys/proc.h>
85 #include <sys/domain.h>
86 #include <sys/sysctl.h>
87 
88 #include <net/if.h>
89 #include <net/route.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_systm.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/ip_var.h>
97 #include <netinet/ip_icmp.h>
98 #include <netinet/udp.h>
99 #include <netinet/udp_var.h>
100 
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/udp6_var.h>
107 #endif
108 
109 #ifdef PULLDOWN_TEST
110 #ifndef INET6
111 /* always need ip6.h for IP6_EXTHDR_GET */
112 #include <netinet/ip6.h>
113 #endif
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 IPSEC
124 #include <netinet6/ipsec.h>
125 #include <netkey/key.h>
126 #endif /*IPSEC*/
127 
128 #ifdef IPKDB
129 #include <ipkdb/ipkdb.h>
130 #endif
131 
132 /*
133  * UDP protocol implementation.
134  * Per RFC 768, August, 1980.
135  */
136 #ifndef	COMPAT_42
137 int	udpcksum = 1;
138 #else
139 int	udpcksum = 0;		/* XXX */
140 #endif
141 
142 #ifdef INET
143 static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *,
144 	struct socket *));
145 static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *,
146 	struct mbuf *, int));
147 #endif
148 #ifdef INET6
149 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *,
150 	struct socket *));
151 static int udp6_realinput __P((int, struct sockaddr_in6 *,
152 	struct sockaddr_in6 *, struct mbuf *, int));
153 #endif
154 #ifdef INET
155 static	void udp_notify __P((struct inpcb *, int));
156 #endif
157 
158 #ifndef UDBHASHSIZE
159 #define	UDBHASHSIZE	128
160 #endif
161 int	udbhashsize = UDBHASHSIZE;
162 
163 #ifdef UDP_CSUM_COUNTERS
164 #include <sys/device.h>
165 
166 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
167     NULL, "udp", "hwcsum bad");
168 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
169     NULL, "udp", "hwcsum ok");
170 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
171     NULL, "udp", "hwcsum data");
172 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
173     NULL, "udp", "swcsum");
174 
175 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
176 
177 #else
178 
179 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
180 
181 #endif /* UDP_CSUM_COUNTERS */
182 
183 void
184 udp_init()
185 {
186 
187 #ifdef INET
188 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
189 #endif
190 
191 #ifdef UDP_CSUM_COUNTERS
192 	evcnt_attach_static(&udp_hwcsum_bad);
193 	evcnt_attach_static(&udp_hwcsum_ok);
194 	evcnt_attach_static(&udp_hwcsum_data);
195 	evcnt_attach_static(&udp_swcsum);
196 #endif /* UDP_CSUM_COUNTERS */
197 }
198 
199 #ifdef INET
200 void
201 #if __STDC__
202 udp_input(struct mbuf *m, ...)
203 #else
204 udp_input(m, va_alist)
205 	struct mbuf *m;
206 	va_dcl
207 #endif
208 {
209 	va_list ap;
210 	struct sockaddr_in src, dst;
211 	struct ip *ip;
212 	struct udphdr *uh;
213 	int iphlen, proto;
214 	int len;
215 	int n;
216 
217 	va_start(ap, m);
218 	iphlen = va_arg(ap, int);
219 	proto = va_arg(ap, int);
220 	va_end(ap);
221 
222 	udpstat.udps_ipackets++;
223 
224 #ifndef PULLDOWN_TEST
225 	/*
226 	 * Strip IP options, if any; should skip this,
227 	 * make available to user, and use on returned packets,
228 	 * but we don't yet have a way to check the checksum
229 	 * with options still present.
230 	 */
231 	if (iphlen > sizeof (struct ip)) {
232 		ip_stripoptions(m, (struct mbuf *)0);
233 		iphlen = sizeof(struct ip);
234 	}
235 #else
236 	/*
237 	 * we may enable the above code if we save and pass IPv4 options
238 	 * to the userland.
239 	 */
240 #endif
241 
242 	/*
243 	 * Get IP and UDP header together in first mbuf.
244 	 */
245 	ip = mtod(m, struct ip *);
246 #ifndef PULLDOWN_TEST
247 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
248 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
249 			udpstat.udps_hdrops++;
250 			return;
251 		}
252 		ip = mtod(m, struct ip *);
253 	}
254 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
255 #else
256 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
257 	if (uh == NULL) {
258 		udpstat.udps_hdrops++;
259 		return;
260 	}
261 #endif
262 
263 	/* destination port of 0 is illegal, based on RFC768. */
264 	if (uh->uh_dport == 0)
265 		goto bad;
266 
267 	/*
268 	 * Make mbuf data length reflect UDP length.
269 	 * If not enough data to reflect UDP length, drop.
270 	 */
271 	len = ntohs((u_int16_t)uh->uh_ulen);
272 	if (ip->ip_len != iphlen + len) {
273 		if (ip->ip_len < iphlen + len || len < sizeof(struct udphdr)) {
274 			udpstat.udps_badlen++;
275 			goto bad;
276 		}
277 		m_adj(m, iphlen + len - ip->ip_len);
278 	}
279 
280 	/*
281 	 * Checksum extended UDP header and data.
282 	 */
283 	if (uh->uh_sum) {
284 		switch (m->m_pkthdr.csum_flags &
285 			((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
286 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
287 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
288 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
289 			goto badcsum;
290 
291 		case M_CSUM_UDPv4|M_CSUM_DATA:
292 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
293 			if ((m->m_pkthdr.csum_data ^ 0xffff) != 0)
294 				goto badcsum;
295 			break;
296 
297 		case M_CSUM_UDPv4:
298 			/* Checksum was okay. */
299 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
300 			break;
301 
302 		default:
303 			/* Need to compute it ourselves. */
304 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
305 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
306 				goto badcsum;
307 			break;
308 		}
309 	}
310 
311 	/* construct source and dst sockaddrs. */
312 	bzero(&src, sizeof(src));
313 	src.sin_family = AF_INET;
314 	src.sin_len = sizeof(struct sockaddr_in);
315 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
316 	src.sin_port = uh->uh_sport;
317 	bzero(&dst, sizeof(dst));
318 	dst.sin_family = AF_INET;
319 	dst.sin_len = sizeof(struct sockaddr_in);
320 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
321 	dst.sin_port = uh->uh_dport;
322 
323 	n = udp4_realinput(&src, &dst, m, iphlen);
324 #ifdef INET6
325 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
326 		struct sockaddr_in6 src6, dst6;
327 
328 		bzero(&src6, sizeof(src6));
329 		src6.sin6_family = AF_INET6;
330 		src6.sin6_len = sizeof(struct sockaddr_in6);
331 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
332 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
333 			sizeof(ip->ip_src));
334 		src6.sin6_port = uh->uh_sport;
335 		bzero(&dst6, sizeof(dst6));
336 		dst6.sin6_family = AF_INET6;
337 		dst6.sin6_len = sizeof(struct sockaddr_in6);
338 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
339 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
340 			sizeof(ip->ip_dst));
341 		dst6.sin6_port = uh->uh_dport;
342 
343 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
344 	}
345 #endif
346 
347 	if (n == 0) {
348 		if (m->m_flags & (M_BCAST | M_MCAST)) {
349 			udpstat.udps_noportbcast++;
350 			goto bad;
351 		}
352 		udpstat.udps_noport++;
353 #ifdef IPKDB
354 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
355 				m, iphlen + sizeof(struct udphdr),
356 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
357 			/*
358 			 * It was a debugger connect packet,
359 			 * just drop it now
360 			 */
361 			goto bad;
362 		}
363 #endif
364 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
365 		m = NULL;
366 	}
367 
368 bad:
369 	if (m)
370 		m_freem(m);
371 	return;
372 
373 badcsum:
374 	m_freem(m);
375 	udpstat.udps_badsum++;
376 }
377 #endif
378 
379 #ifdef INET6
380 int
381 udp6_input(mp, offp, proto)
382 	struct mbuf **mp;
383 	int *offp, proto;
384 {
385 	struct mbuf *m = *mp;
386 	int off = *offp;
387 	struct sockaddr_in6 src, dst;
388 	struct ip6_hdr *ip6;
389 	struct udphdr *uh;
390 	u_int32_t plen, ulen;
391 
392 #ifndef PULLDOWN_TEST
393 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
394 #endif
395 	ip6 = mtod(m, struct ip6_hdr *);
396 
397 #if defined(NFAITH) && 0 < NFAITH
398 	if (faithprefix(&ip6->ip6_dst)) {
399 		/* send icmp6 host unreach? */
400 		m_freem(m);
401 		return IPPROTO_DONE;
402 	}
403 #endif
404 
405 	udp6stat.udp6s_ipackets++;
406 
407 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
408 	plen = m->m_pkthdr.len - off;
409 #ifndef PULLDOWN_TEST
410 	uh = (struct udphdr *)((caddr_t)ip6 + off);
411 #else
412 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
413 	if (uh == NULL) {
414 		ip6stat.ip6s_tooshort++;
415 		return IPPROTO_DONE;
416 	}
417 #endif
418 	ulen = ntohs((u_short)uh->uh_ulen);
419 	/*
420 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
421 	 * iff payload length > 0xffff.
422 	 */
423 	if (ulen == 0 && plen > 0xffff)
424 		ulen = plen;
425 
426 	if (plen != ulen) {
427 		udp6stat.udp6s_badlen++;
428 		goto bad;
429 	}
430 
431 	/* destination port of 0 is illegal, based on RFC768. */
432 	if (uh->uh_dport == 0)
433 		goto bad;
434 
435 	/* Be proactive about malicious use of IPv4 mapped address */
436 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
437 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
438 		/* XXX stat */
439 		goto bad;
440 	}
441 
442 	/*
443 	 * Checksum extended UDP header and data.
444 	 */
445 	if (uh->uh_sum == 0)
446 		udp6stat.udp6s_nosum++;
447 	else if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
448 		udp6stat.udp6s_badsum++;
449 		goto bad;
450 	}
451 
452 	/*
453 	 * Construct source and dst sockaddrs.
454 	 * Note that ifindex (s6_addr16[1]) is already filled.
455 	 */
456 	bzero(&src, sizeof(src));
457 	src.sin6_family = AF_INET6;
458 	src.sin6_len = sizeof(struct sockaddr_in6);
459 	/* KAME hack: recover scopeid */
460 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
461 	src.sin6_port = uh->uh_sport;
462 	bzero(&dst, sizeof(dst));
463 	dst.sin6_family = AF_INET6;
464 	dst.sin6_len = sizeof(struct sockaddr_in6);
465 	/* KAME hack: recover scopeid */
466 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
467 	dst.sin6_port = uh->uh_dport;
468 
469 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
470 		if (m->m_flags & M_MCAST) {
471 			udp6stat.udp6s_noportmcast++;
472 			goto bad;
473 		}
474 		udp6stat.udp6s_noport++;
475 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
476 		m = NULL;
477 	}
478 
479 bad:
480 	if (m)
481 		m_freem(m);
482 	return IPPROTO_DONE;
483 }
484 #endif
485 
486 #ifdef INET
487 static void
488 udp4_sendup(m, off, src, so)
489 	struct mbuf *m;
490 	int off;	/* offset of data portion */
491 	struct sockaddr *src;
492 	struct socket *so;
493 {
494 	struct mbuf *opts = NULL;
495 	struct mbuf *n;
496 	struct inpcb *inp = NULL;
497 #ifdef INET6
498 	struct in6pcb *in6p = NULL;
499 #endif
500 
501 	if (!so)
502 		return;
503 	switch (so->so_proto->pr_domain->dom_family) {
504 	case AF_INET:
505 		inp = sotoinpcb(so);
506 		break;
507 #ifdef INET6
508 	case AF_INET6:
509 		in6p = sotoin6pcb(so);
510 		break;
511 #endif
512 	default:
513 		return;
514 	}
515 
516 #ifdef IPSEC
517 	/* check AH/ESP integrity. */
518 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
519 		ipsecstat.in_polvio++;
520 		return;
521 	}
522 #endif /*IPSEC*/
523 
524 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
525 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
526 			 || so->so_options & SO_TIMESTAMP)) {
527 			struct ip *ip = mtod(n, struct ip *);
528 			ip_savecontrol(inp, &opts, ip, n);
529 		}
530 
531 		m_adj(n, off);
532 		if (sbappendaddr(&so->so_rcv, src, n,
533 				opts) == 0) {
534 			m_freem(n);
535 			if (opts)
536 				m_freem(opts);
537 			udpstat.udps_fullsock++;
538 		} else
539 			sorwakeup(so);
540 	}
541 }
542 #endif
543 
544 #ifdef INET6
545 static void
546 udp6_sendup(m, off, src, so)
547 	struct mbuf *m;
548 	int off;	/* offset of data portion */
549 	struct sockaddr *src;
550 	struct socket *so;
551 {
552 	struct mbuf *opts = NULL;
553 	struct mbuf *n;
554 	struct in6pcb *in6p = NULL;
555 
556 	if (!so)
557 		return;
558 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
559 		return;
560 	in6p = sotoin6pcb(so);
561 
562 #ifdef IPSEC
563 	/* check AH/ESP integrity. */
564 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
565 		ipsec6stat.in_polvio++;
566 		return;
567 	}
568 #endif /*IPSEC*/
569 
570 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
571 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
572 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
573 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
574 			ip6_savecontrol(in6p, &opts, ip6, n);
575 		}
576 
577 		m_adj(n, off);
578 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
579 			m_freem(n);
580 			if (opts)
581 				m_freem(opts);
582 			udp6stat.udp6s_fullsock++;
583 		} else
584 			sorwakeup(so);
585 	}
586 }
587 #endif
588 
589 #ifdef INET
590 static int
591 udp4_realinput(src, dst, m, off)
592 	struct sockaddr_in *src;
593 	struct sockaddr_in *dst;
594 	struct mbuf *m;
595 	int off;	/* offset of udphdr */
596 {
597 	u_int16_t *sport, *dport;
598 	int rcvcnt;
599 	struct in_addr *src4, *dst4;
600 	struct inpcb *inp;
601 
602 	rcvcnt = 0;
603 	off += sizeof(struct udphdr);	/* now, offset of payload */
604 
605 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
606 		goto bad;
607 
608 	src4 = &src->sin_addr;
609 	sport = &src->sin_port;
610 	dst4 = &dst->sin_addr;
611 	dport = &dst->sin_port;
612 
613 	if (IN_MULTICAST(dst4->s_addr) ||
614 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
615 		struct inpcb *last;
616 		/*
617 		 * Deliver a multicast or broadcast datagram to *all* sockets
618 		 * for which the local and remote addresses and ports match
619 		 * those of the incoming datagram.  This allows more than
620 		 * one process to receive multi/broadcasts on the same port.
621 		 * (This really ought to be done for unicast datagrams as
622 		 * well, but that would cause problems with existing
623 		 * applications that open both address-specific sockets and
624 		 * a wildcard socket listening to the same port -- they would
625 		 * end up receiving duplicates of every unicast datagram.
626 		 * Those applications open the multiple sockets to overcome an
627 		 * inadequacy of the UDP socket interface, but for backwards
628 		 * compatibility we avoid the problem here rather than
629 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
630 		 */
631 
632 		/*
633 		 * KAME note: usually we drop udpiphdr from mbuf here.
634 		 * we need udpiphdr for IPsec processing so we do that later.
635 		 */
636 		/*
637 		 * Locate pcb(s) for datagram.
638 		 */
639 		CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
640 			if (inp->inp_lport != *dport)
641 				continue;
642 			if (!in_nullhost(inp->inp_laddr)) {
643 				if (!in_hosteq(inp->inp_laddr, *dst4))
644 					continue;
645 			}
646 			if (!in_nullhost(inp->inp_faddr)) {
647 				if (!in_hosteq(inp->inp_faddr, *src4) ||
648 				    inp->inp_fport != *sport)
649 					continue;
650 			}
651 
652 			last = inp;
653 			udp4_sendup(m, off, (struct sockaddr *)src,
654 				inp->inp_socket);
655 			rcvcnt++;
656 
657 			/*
658 			 * Don't look for additional matches if this one does
659 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
660 			 * socket options set.  This heuristic avoids searching
661 			 * through all pcbs in the common case of a non-shared
662 			 * port.  It assumes that an application will never
663 			 * clear these options after setting them.
664 			 */
665 			if ((inp->inp_socket->so_options &
666 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
667 				break;
668 		}
669 	} else {
670 		/*
671 		 * Locate pcb for datagram.
672 		 */
673 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
674 		if (inp == 0) {
675 			++udpstat.udps_pcbhashmiss;
676 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
677 			if (inp == 0)
678 				return rcvcnt;
679 		}
680 
681 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
682 		rcvcnt++;
683 	}
684 
685 bad:
686 	return rcvcnt;
687 }
688 #endif
689 
690 #ifdef INET6
691 static int
692 udp6_realinput(af, src, dst, m, off)
693 	int af;		/* af on packet */
694 	struct sockaddr_in6 *src;
695 	struct sockaddr_in6 *dst;
696 	struct mbuf *m;
697 	int off;	/* offset of udphdr */
698 {
699 	u_int16_t sport, dport;
700 	int rcvcnt;
701 	struct in6_addr src6, dst6;
702 	const struct in_addr *dst4;
703 	struct in6pcb *in6p;
704 
705 	rcvcnt = 0;
706 	off += sizeof(struct udphdr);	/* now, offset of payload */
707 
708 	if (af != AF_INET && af != AF_INET6)
709 		goto bad;
710 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
711 		goto bad;
712 
713 	in6_embedscope(&src6, src, NULL, NULL);
714 	sport = src->sin6_port;
715 	in6_embedscope(&dst6, dst, NULL, NULL);
716 	dport = dst->sin6_port;
717 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
718 
719 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
720 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
721 		struct in6pcb *last;
722 		/*
723 		 * Deliver a multicast or broadcast datagram to *all* sockets
724 		 * for which the local and remote addresses and ports match
725 		 * those of the incoming datagram.  This allows more than
726 		 * one process to receive multi/broadcasts on the same port.
727 		 * (This really ought to be done for unicast datagrams as
728 		 * well, but that would cause problems with existing
729 		 * applications that open both address-specific sockets and
730 		 * a wildcard socket listening to the same port -- they would
731 		 * end up receiving duplicates of every unicast datagram.
732 		 * Those applications open the multiple sockets to overcome an
733 		 * inadequacy of the UDP socket interface, but for backwards
734 		 * compatibility we avoid the problem here rather than
735 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
736 		 */
737 
738 		/*
739 		 * KAME note: usually we drop udpiphdr from mbuf here.
740 		 * we need udpiphdr for IPsec processing so we do that later.
741 		 */
742 		/*
743 		 * Locate pcb(s) for datagram.
744 		 */
745 		for (in6p = udb6.in6p_next; in6p != &udb6;
746 		     in6p = in6p->in6p_next) {
747 			if (in6p->in6p_lport != dport)
748 				continue;
749 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
750 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
751 					continue;
752 			}
753 			else {
754 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
755 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
756 					continue;
757 			}
758 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
759 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
760 				    &src6) || in6p->in6p_fport != sport)
761 					continue;
762 			}
763 			else {
764 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
765 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
766 					continue;
767 			}
768 
769 			last = in6p;
770 			udp6_sendup(m, off, (struct sockaddr *)src,
771 				in6p->in6p_socket);
772 			rcvcnt++;
773 
774 			/*
775 			 * Don't look for additional matches if this one does
776 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
777 			 * socket options set.  This heuristic avoids searching
778 			 * through all pcbs in the common case of a non-shared
779 			 * port.  It assumes that an application will never
780 			 * clear these options after setting them.
781 			 */
782 			if ((in6p->in6p_socket->so_options &
783 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
784 				break;
785 		}
786 	} else {
787 		/*
788 		 * Locate pcb for datagram.
789 		 */
790 		in6p = in6_pcblookup_connect(&udb6, &src6, sport,
791 		    &dst6, dport, 0);
792 		if (in6p == 0) {
793 			++udpstat.udps_pcbhashmiss;
794 			in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0);
795 			if (in6p == 0)
796 				return rcvcnt;
797 		}
798 
799 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
800 		rcvcnt++;
801 	}
802 
803 bad:
804 	return rcvcnt;
805 }
806 #endif
807 
808 #ifdef INET
809 /*
810  * Notify a udp user of an asynchronous error;
811  * just wake up so that he can collect error status.
812  */
813 static void
814 udp_notify(inp, errno)
815 	struct inpcb *inp;
816 	int errno;
817 {
818 
819 	inp->inp_socket->so_error = errno;
820 	sorwakeup(inp->inp_socket);
821 	sowwakeup(inp->inp_socket);
822 }
823 
824 void *
825 udp_ctlinput(cmd, sa, v)
826 	int cmd;
827 	struct sockaddr *sa;
828 	void *v;
829 {
830 	struct ip *ip = v;
831 	struct udphdr *uh;
832 	void (*notify) __P((struct inpcb *, int)) = udp_notify;
833 	int errno;
834 
835 	if (sa->sa_family != AF_INET
836 	 || sa->sa_len != sizeof(struct sockaddr_in))
837 		return NULL;
838 	if ((unsigned)cmd >= PRC_NCMDS)
839 		return NULL;
840 	errno = inetctlerrmap[cmd];
841 	if (PRC_IS_REDIRECT(cmd))
842 		notify = in_rtchange, ip = 0;
843 	else if (cmd == PRC_HOSTDEAD)
844 		ip = 0;
845 	else if (errno == 0)
846 		return NULL;
847 	if (ip) {
848 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
849 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
850 		    ip->ip_src, uh->uh_sport, errno, notify);
851 
852 		/* XXX mapped address case */
853 	} else
854 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
855 		    notify);
856 	return NULL;
857 }
858 
859 int
860 #if __STDC__
861 udp_output(struct mbuf *m, ...)
862 #else
863 udp_output(m, va_alist)
864 	struct mbuf *m;
865 	va_dcl
866 #endif
867 {
868 	struct inpcb *inp;
869 	struct udpiphdr *ui;
870 	int len = m->m_pkthdr.len;
871 	int error = 0;
872 	va_list ap;
873 
874 	va_start(ap, m);
875 	inp = va_arg(ap, struct inpcb *);
876 	va_end(ap);
877 
878 	/*
879 	 * Calculate data length and get a mbuf
880 	 * for UDP and IP headers.
881 	 */
882 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
883 	if (m == 0) {
884 		error = ENOBUFS;
885 		goto release;
886 	}
887 
888 	/*
889 	 * Compute the packet length of the IP header, and
890 	 * punt if the length looks bogus.
891 	 */
892 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
893 		error = EMSGSIZE;
894 		goto release;
895 	}
896 
897 	/*
898 	 * Fill in mbuf with extended UDP header
899 	 * and addresses and length put into network format.
900 	 */
901 	ui = mtod(m, struct udpiphdr *);
902 	ui->ui_pr = IPPROTO_UDP;
903 	ui->ui_src = inp->inp_laddr;
904 	ui->ui_dst = inp->inp_faddr;
905 	ui->ui_sport = inp->inp_lport;
906 	ui->ui_dport = inp->inp_fport;
907 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
908 
909 	/*
910 	 * Set up checksum and output datagram.
911 	 */
912 	if (udpcksum) {
913 		/*
914 		 * XXX Cache pseudo-header checksum part for
915 		 * XXX "connected" UDP sockets.
916 		 */
917 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
918 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
919 		    sizeof(struct udphdr) + IPPROTO_UDP));
920 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
921 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
922 	} else
923 		ui->ui_sum = 0;
924 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
925 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
926 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
927 	udpstat.udps_opackets++;
928 
929 #ifdef IPSEC
930 	if (ipsec_setsocket(m, inp->inp_socket) != 0) {
931 		error = ENOBUFS;
932 		goto release;
933 	}
934 #endif /*IPSEC*/
935 
936 	return (ip_output(m, inp->inp_options, &inp->inp_route,
937 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
938 	    inp->inp_moptions));
939 
940 release:
941 	m_freem(m);
942 	return (error);
943 }
944 
945 int	udp_sendspace = 9216;		/* really max datagram size */
946 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
947 					/* 40 1K datagrams */
948 
949 /*ARGSUSED*/
950 int
951 udp_usrreq(so, req, m, nam, control, p)
952 	struct socket *so;
953 	int req;
954 	struct mbuf *m, *nam, *control;
955 	struct proc *p;
956 {
957 	struct inpcb *inp;
958 	int s;
959 	int error = 0;
960 
961 	if (req == PRU_CONTROL)
962 		return (in_control(so, (long)m, (caddr_t)nam,
963 		    (struct ifnet *)control, p));
964 
965 	if (req == PRU_PURGEIF) {
966 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
967 		in_purgeif((struct ifnet *)control);
968 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
969 		return (0);
970 	}
971 
972 	s = splsoftnet();
973 	inp = sotoinpcb(so);
974 #ifdef DIAGNOSTIC
975 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
976 		panic("udp_usrreq: unexpected control mbuf");
977 #endif
978 	if (inp == 0 && req != PRU_ATTACH) {
979 		error = EINVAL;
980 		goto release;
981 	}
982 
983 	/*
984 	 * Note: need to block udp_input while changing
985 	 * the udp pcb queue and/or pcb addresses.
986 	 */
987 	switch (req) {
988 
989 	case PRU_ATTACH:
990 		if (inp != 0) {
991 			error = EISCONN;
992 			break;
993 		}
994 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
995 			error = soreserve(so, udp_sendspace, udp_recvspace);
996 			if (error)
997 				break;
998 		}
999 		error = in_pcballoc(so, &udbtable);
1000 		if (error)
1001 			break;
1002 		inp = sotoinpcb(so);
1003 		inp->inp_ip.ip_ttl = ip_defttl;
1004 		break;
1005 
1006 	case PRU_DETACH:
1007 		in_pcbdetach(inp);
1008 		break;
1009 
1010 	case PRU_BIND:
1011 		error = in_pcbbind(inp, nam, p);
1012 		break;
1013 
1014 	case PRU_LISTEN:
1015 		error = EOPNOTSUPP;
1016 		break;
1017 
1018 	case PRU_CONNECT:
1019 		error = in_pcbconnect(inp, nam);
1020 		if (error)
1021 			break;
1022 		soisconnected(so);
1023 		break;
1024 
1025 	case PRU_CONNECT2:
1026 		error = EOPNOTSUPP;
1027 		break;
1028 
1029 	case PRU_DISCONNECT:
1030 		/*soisdisconnected(so);*/
1031 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1032 		in_pcbdisconnect(inp);
1033 		inp->inp_laddr = zeroin_addr;		/* XXX */
1034 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1035 		break;
1036 
1037 	case PRU_SHUTDOWN:
1038 		socantsendmore(so);
1039 		break;
1040 
1041 	case PRU_RCVD:
1042 		error = EOPNOTSUPP;
1043 		break;
1044 
1045 	case PRU_SEND:
1046 		if (control && control->m_len) {
1047 			m_freem(control);
1048 			m_freem(m);
1049 			error = EINVAL;
1050 			break;
1051 		}
1052 	{
1053 		struct in_addr laddr;			/* XXX */
1054 
1055 		if (nam) {
1056 			laddr = inp->inp_laddr;		/* XXX */
1057 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1058 				error = EISCONN;
1059 				goto die;
1060 			}
1061 			error = in_pcbconnect(inp, nam);
1062 			if (error) {
1063 			die:
1064 				m_freem(m);
1065 				break;
1066 			}
1067 		} else {
1068 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1069 				error = ENOTCONN;
1070 				goto die;
1071 			}
1072 		}
1073 		error = udp_output(m, inp);
1074 		if (nam) {
1075 			in_pcbdisconnect(inp);
1076 			inp->inp_laddr = laddr;		/* XXX */
1077 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1078 		}
1079 	}
1080 		break;
1081 
1082 	case PRU_SENSE:
1083 		/*
1084 		 * stat: don't bother with a blocksize.
1085 		 */
1086 		splx(s);
1087 		return (0);
1088 
1089 	case PRU_RCVOOB:
1090 		error =  EOPNOTSUPP;
1091 		break;
1092 
1093 	case PRU_SENDOOB:
1094 		m_freem(control);
1095 		m_freem(m);
1096 		error =  EOPNOTSUPP;
1097 		break;
1098 
1099 	case PRU_SOCKADDR:
1100 		in_setsockaddr(inp, nam);
1101 		break;
1102 
1103 	case PRU_PEERADDR:
1104 		in_setpeeraddr(inp, nam);
1105 		break;
1106 
1107 	default:
1108 		panic("udp_usrreq");
1109 	}
1110 
1111 release:
1112 	splx(s);
1113 	return (error);
1114 }
1115 
1116 /*
1117  * Sysctl for udp variables.
1118  */
1119 int
1120 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1121 	int *name;
1122 	u_int namelen;
1123 	void *oldp;
1124 	size_t *oldlenp;
1125 	void *newp;
1126 	size_t newlen;
1127 {
1128 	/* All sysctl names at this level are terminal. */
1129 	if (namelen != 1)
1130 		return (ENOTDIR);
1131 
1132 	switch (name[0]) {
1133 	case UDPCTL_CHECKSUM:
1134 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1135 	case UDPCTL_SENDSPACE:
1136 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1137 		    &udp_sendspace));
1138 	case UDPCTL_RECVSPACE:
1139 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1140 		    &udp_recvspace));
1141 	default:
1142 		return (ENOPROTOOPT);
1143 	}
1144 	/* NOTREACHED */
1145 }
1146 #endif
1147