xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 33cd1faa348fe1cb7947ea59b0556b9bad76cee9)
1 /*	$NetBSD: udp_usrreq.c,v 1.118 2004/03/31 07:57:06 itojun 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 1.118 2004/03/31 07:57:06 itojun 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 #ifdef INET6
121 #include <netipsec/ipsec6.h>
122 #endif
123 #endif	/* FAST_IPSEC*/
124 
125 #ifdef IPSEC
126 #include <netinet6/ipsec.h>
127 #include <netkey/key.h>
128 #endif /*IPSEC*/
129 
130 #ifdef IPKDB
131 #include <ipkdb/ipkdb.h>
132 #endif
133 
134 /*
135  * UDP protocol implementation.
136  * Per RFC 768, August, 1980.
137  */
138 #ifndef	COMPAT_42
139 int	udpcksum = 1;
140 #else
141 int	udpcksum = 0;		/* XXX */
142 #endif
143 
144 struct	inpcbtable udbtable;
145 struct	udpstat udpstat;
146 
147 #ifdef INET
148 static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *,
149 	struct socket *));
150 static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *,
151 	struct mbuf *, int));
152 #endif
153 #ifdef INET6
154 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *,
155 	struct socket *));
156 static int udp6_realinput __P((int, struct sockaddr_in6 *,
157 	struct sockaddr_in6 *, struct mbuf *, int));
158 #endif
159 #ifdef INET
160 static	void udp_notify __P((struct inpcb *, int));
161 #endif
162 
163 #ifndef UDBHASHSIZE
164 #define	UDBHASHSIZE	128
165 #endif
166 int	udbhashsize = UDBHASHSIZE;
167 
168 #ifdef MBUFTRACE
169 struct mowner udp_mowner = { "udp" };
170 struct mowner udp_rx_mowner = { "udp", "rx" };
171 struct mowner udp_tx_mowner = { "udp", "tx" };
172 #endif
173 
174 #ifdef UDP_CSUM_COUNTERS
175 #include <sys/device.h>
176 
177 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
178     NULL, "udp", "hwcsum bad");
179 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
180     NULL, "udp", "hwcsum ok");
181 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182     NULL, "udp", "hwcsum data");
183 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
184     NULL, "udp", "swcsum");
185 
186 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
187 
188 #else
189 
190 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
191 
192 #endif /* UDP_CSUM_COUNTERS */
193 
194 void
195 udp_init()
196 {
197 
198 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
199 
200 #ifdef UDP_CSUM_COUNTERS
201 	evcnt_attach_static(&udp_hwcsum_bad);
202 	evcnt_attach_static(&udp_hwcsum_ok);
203 	evcnt_attach_static(&udp_hwcsum_data);
204 	evcnt_attach_static(&udp_swcsum);
205 #endif /* UDP_CSUM_COUNTERS */
206 
207 	MOWNER_ATTACH(&udp_tx_mowner);
208 	MOWNER_ATTACH(&udp_rx_mowner);
209 	MOWNER_ATTACH(&udp_mowner);
210 }
211 
212 #ifdef INET
213 void
214 #if __STDC__
215 udp_input(struct mbuf *m, ...)
216 #else
217 udp_input(m, va_alist)
218 	struct mbuf *m;
219 	va_dcl
220 #endif
221 {
222 	va_list ap;
223 	struct sockaddr_in src, dst;
224 	struct ip *ip;
225 	struct udphdr *uh;
226 	int iphlen;
227 	int len;
228 	int n;
229 	u_int16_t ip_len;
230 
231 	va_start(ap, m);
232 	iphlen = va_arg(ap, int);
233 	(void)va_arg(ap, int);		/* ignore value, advance ap */
234 	va_end(ap);
235 
236 	MCLAIM(m, &udp_rx_mowner);
237 	udpstat.udps_ipackets++;
238 
239 	/*
240 	 * Get IP and UDP header together in first mbuf.
241 	 */
242 	ip = mtod(m, struct ip *);
243 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
244 	if (uh == NULL) {
245 		udpstat.udps_hdrops++;
246 		return;
247 	}
248 	KASSERT(UDP_HDR_ALIGNED_P(uh));
249 
250 	/* destination port of 0 is illegal, based on RFC768. */
251 	if (uh->uh_dport == 0)
252 		goto bad;
253 
254 	/*
255 	 * Make mbuf data length reflect UDP length.
256 	 * If not enough data to reflect UDP length, drop.
257 	 */
258 	ip_len = ntohs(ip->ip_len);
259 	len = ntohs((u_int16_t)uh->uh_ulen);
260 	if (ip_len != iphlen + len) {
261 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
262 			udpstat.udps_badlen++;
263 			goto bad;
264 		}
265 		m_adj(m, iphlen + len - ip_len);
266 	}
267 
268 	/*
269 	 * Checksum extended UDP header and data.
270 	 */
271 	if (uh->uh_sum) {
272 		switch (m->m_pkthdr.csum_flags &
273 		    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
274 		    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
275 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
276 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
277 			goto badcsum;
278 
279 		case M_CSUM_UDPv4|M_CSUM_DATA: {
280 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
281 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
282 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR)
283 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
284 				    ip->ip_dst.s_addr,
285 				    htons(hw_csum + len + IPPROTO_UDP));
286 			if ((hw_csum ^ 0xffff) != 0)
287 				goto badcsum;
288 			break;
289 		}
290 
291 		case M_CSUM_UDPv4:
292 			/* Checksum was okay. */
293 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
294 			break;
295 
296 		default:
297 			/* Need to compute it ourselves. */
298 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
299 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
300 				goto badcsum;
301 			break;
302 		}
303 	}
304 
305 	/* construct source and dst sockaddrs. */
306 	bzero(&src, sizeof(src));
307 	src.sin_family = AF_INET;
308 	src.sin_len = sizeof(struct sockaddr_in);
309 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
310 	src.sin_port = uh->uh_sport;
311 	bzero(&dst, sizeof(dst));
312 	dst.sin_family = AF_INET;
313 	dst.sin_len = sizeof(struct sockaddr_in);
314 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
315 	dst.sin_port = uh->uh_dport;
316 
317 	n = udp4_realinput(&src, &dst, m, iphlen);
318 #ifdef INET6
319 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
320 		struct sockaddr_in6 src6, dst6;
321 
322 		bzero(&src6, sizeof(src6));
323 		src6.sin6_family = AF_INET6;
324 		src6.sin6_len = sizeof(struct sockaddr_in6);
325 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
326 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
327 			sizeof(ip->ip_src));
328 		src6.sin6_port = uh->uh_sport;
329 		bzero(&dst6, sizeof(dst6));
330 		dst6.sin6_family = AF_INET6;
331 		dst6.sin6_len = sizeof(struct sockaddr_in6);
332 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
333 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
334 			sizeof(ip->ip_dst));
335 		dst6.sin6_port = uh->uh_dport;
336 
337 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
338 	}
339 #endif
340 
341 	if (n == 0) {
342 		if (m->m_flags & (M_BCAST | M_MCAST)) {
343 			udpstat.udps_noportbcast++;
344 			goto bad;
345 		}
346 		udpstat.udps_noport++;
347 #ifdef IPKDB
348 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
349 				m, iphlen + sizeof(struct udphdr),
350 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
351 			/*
352 			 * It was a debugger connect packet,
353 			 * just drop it now
354 			 */
355 			goto bad;
356 		}
357 #endif
358 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
359 		m = NULL;
360 	}
361 
362 bad:
363 	if (m)
364 		m_freem(m);
365 	return;
366 
367 badcsum:
368 	m_freem(m);
369 	udpstat.udps_badsum++;
370 }
371 #endif
372 
373 #ifdef INET6
374 int
375 udp6_input(mp, offp, proto)
376 	struct mbuf **mp;
377 	int *offp, proto;
378 {
379 	struct mbuf *m = *mp;
380 	int off = *offp;
381 	struct sockaddr_in6 src, dst;
382 	struct ip6_hdr *ip6;
383 	struct udphdr *uh;
384 	u_int32_t plen, ulen;
385 
386 	ip6 = mtod(m, struct ip6_hdr *);
387 
388 #if defined(NFAITH) && 0 < NFAITH
389 	if (faithprefix(&ip6->ip6_dst)) {
390 		/* send icmp6 host unreach? */
391 		m_freem(m);
392 		return IPPROTO_DONE;
393 	}
394 #endif
395 
396 	udp6stat.udp6s_ipackets++;
397 
398 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
399 	plen = m->m_pkthdr.len - off;
400 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
401 	if (uh == NULL) {
402 		ip6stat.ip6s_tooshort++;
403 		return IPPROTO_DONE;
404 	}
405 	KASSERT(UDP_HDR_ALIGNED_P(uh));
406 	ulen = ntohs((u_short)uh->uh_ulen);
407 	/*
408 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
409 	 * iff payload length > 0xffff.
410 	 */
411 	if (ulen == 0 && plen > 0xffff)
412 		ulen = plen;
413 
414 	if (plen != ulen) {
415 		udp6stat.udp6s_badlen++;
416 		goto bad;
417 	}
418 
419 	/* destination port of 0 is illegal, based on RFC768. */
420 	if (uh->uh_dport == 0)
421 		goto bad;
422 
423 	/* Be proactive about malicious use of IPv4 mapped address */
424 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
425 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
426 		/* XXX stat */
427 		goto bad;
428 	}
429 
430 	/*
431 	 * Checksum extended UDP header and data.
432 	 */
433 	if (uh->uh_sum == 0) {
434 		udp6stat.udp6s_nosum++;
435 		goto bad;
436 	}
437 	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
438 		udp6stat.udp6s_badsum++;
439 		goto bad;
440 	}
441 
442 	/*
443 	 * Construct source and dst sockaddrs.
444 	 * Note that ifindex (s6_addr16[1]) is already filled.
445 	 */
446 	bzero(&src, sizeof(src));
447 	src.sin6_family = AF_INET6;
448 	src.sin6_len = sizeof(struct sockaddr_in6);
449 	/* KAME hack: recover scopeid */
450 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
451 	src.sin6_port = uh->uh_sport;
452 	bzero(&dst, sizeof(dst));
453 	dst.sin6_family = AF_INET6;
454 	dst.sin6_len = sizeof(struct sockaddr_in6);
455 	/* KAME hack: recover scopeid */
456 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
457 	dst.sin6_port = uh->uh_dport;
458 
459 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
460 		if (m->m_flags & M_MCAST) {
461 			udp6stat.udp6s_noportmcast++;
462 			goto bad;
463 		}
464 		udp6stat.udp6s_noport++;
465 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
466 		m = NULL;
467 	}
468 
469 bad:
470 	if (m)
471 		m_freem(m);
472 	return IPPROTO_DONE;
473 }
474 #endif
475 
476 #ifdef INET
477 static void
478 udp4_sendup(m, off, src, so)
479 	struct mbuf *m;
480 	int off;	/* offset of data portion */
481 	struct sockaddr *src;
482 	struct socket *so;
483 {
484 	struct mbuf *opts = NULL;
485 	struct mbuf *n;
486 	struct inpcb *inp = NULL;
487 
488 	if (!so)
489 		return;
490 	switch (so->so_proto->pr_domain->dom_family) {
491 	case AF_INET:
492 		inp = sotoinpcb(so);
493 		break;
494 #ifdef INET6
495 	case AF_INET6:
496 		break;
497 #endif
498 	default:
499 		return;
500 	}
501 
502 #if defined(IPSEC) || defined(FAST_IPSEC)
503 	/* check AH/ESP integrity. */
504 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
505 		ipsecstat.in_polvio++;
506 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
507 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
508 			    0, 0);
509 		return;
510 	}
511 #endif /*IPSEC*/
512 
513 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
514 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
515 			 || so->so_options & SO_TIMESTAMP)) {
516 			struct ip *ip = mtod(n, struct ip *);
517 			ip_savecontrol(inp, &opts, ip, n);
518 		}
519 
520 		m_adj(n, off);
521 		if (sbappendaddr(&so->so_rcv, src, n,
522 				opts) == 0) {
523 			m_freem(n);
524 			if (opts)
525 				m_freem(opts);
526 			udpstat.udps_fullsock++;
527 		} else
528 			sorwakeup(so);
529 	}
530 }
531 #endif
532 
533 #ifdef INET6
534 static void
535 udp6_sendup(m, off, src, so)
536 	struct mbuf *m;
537 	int off;	/* offset of data portion */
538 	struct sockaddr *src;
539 	struct socket *so;
540 {
541 	struct mbuf *opts = NULL;
542 	struct mbuf *n;
543 	struct in6pcb *in6p = NULL;
544 
545 	if (!so)
546 		return;
547 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
548 		return;
549 	in6p = sotoin6pcb(so);
550 
551 #if defined(IPSEC) || defined(FAST_IPSEC)
552 	/* check AH/ESP integrity. */
553 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
554 		ipsec6stat.in_polvio++;
555 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
556 			icmp6_error(n, ICMP6_DST_UNREACH,
557 			    ICMP6_DST_UNREACH_ADMIN, 0);
558 		return;
559 	}
560 #endif /*IPSEC*/
561 
562 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
563 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
564 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
565 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
566 			ip6_savecontrol(in6p, &opts, ip6, n);
567 		}
568 
569 		m_adj(n, off);
570 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
571 			m_freem(n);
572 			if (opts)
573 				m_freem(opts);
574 			udp6stat.udp6s_fullsock++;
575 		} else
576 			sorwakeup(so);
577 	}
578 }
579 #endif
580 
581 #ifdef INET
582 static int
583 udp4_realinput(src, dst, m, off)
584 	struct sockaddr_in *src;
585 	struct sockaddr_in *dst;
586 	struct mbuf *m;
587 	int off;	/* offset of udphdr */
588 {
589 	u_int16_t *sport, *dport;
590 	int rcvcnt;
591 	struct in_addr *src4, *dst4;
592 	struct inpcb_hdr *inph;
593 	struct inpcb *inp;
594 
595 	rcvcnt = 0;
596 	off += sizeof(struct udphdr);	/* now, offset of payload */
597 
598 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
599 		goto bad;
600 
601 	src4 = &src->sin_addr;
602 	sport = &src->sin_port;
603 	dst4 = &dst->sin_addr;
604 	dport = &dst->sin_port;
605 
606 	if (IN_MULTICAST(dst4->s_addr) ||
607 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
608 		/*
609 		 * Deliver a multicast or broadcast datagram to *all* sockets
610 		 * for which the local and remote addresses and ports match
611 		 * those of the incoming datagram.  This allows more than
612 		 * one process to receive multi/broadcasts on the same port.
613 		 * (This really ought to be done for unicast datagrams as
614 		 * well, but that would cause problems with existing
615 		 * applications that open both address-specific sockets and
616 		 * a wildcard socket listening to the same port -- they would
617 		 * end up receiving duplicates of every unicast datagram.
618 		 * Those applications open the multiple sockets to overcome an
619 		 * inadequacy of the UDP socket interface, but for backwards
620 		 * compatibility we avoid the problem here rather than
621 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
622 		 */
623 
624 		/*
625 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
626 		 * we need udpiphdr for IPsec processing so we do that later.
627 		 */
628 		/*
629 		 * Locate pcb(s) for datagram.
630 		 */
631 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
632 			inp = (struct inpcb *)inph;
633 			if (inp->inp_af != AF_INET)
634 				continue;
635 
636 			if (inp->inp_lport != *dport)
637 				continue;
638 			if (!in_nullhost(inp->inp_laddr)) {
639 				if (!in_hosteq(inp->inp_laddr, *dst4))
640 					continue;
641 			}
642 			if (!in_nullhost(inp->inp_faddr)) {
643 				if (!in_hosteq(inp->inp_faddr, *src4) ||
644 				    inp->inp_fport != *sport)
645 					continue;
646 			}
647 
648 			udp4_sendup(m, off, (struct sockaddr *)src,
649 				inp->inp_socket);
650 			rcvcnt++;
651 
652 			/*
653 			 * Don't look for additional matches if this one does
654 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
655 			 * socket options set.  This heuristic avoids searching
656 			 * through all pcbs in the common case of a non-shared
657 			 * port.  It assumes that an application will never
658 			 * clear these options after setting them.
659 			 */
660 			if ((inp->inp_socket->so_options &
661 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
662 				break;
663 		}
664 	} else {
665 		/*
666 		 * Locate pcb for datagram.
667 		 */
668 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
669 		if (inp == 0) {
670 			++udpstat.udps_pcbhashmiss;
671 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
672 			if (inp == 0)
673 				return rcvcnt;
674 		}
675 
676 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
677 		rcvcnt++;
678 	}
679 
680 bad:
681 	return rcvcnt;
682 }
683 #endif
684 
685 #ifdef INET6
686 static int
687 udp6_realinput(af, src, dst, m, off)
688 	int af;		/* af on packet */
689 	struct sockaddr_in6 *src;
690 	struct sockaddr_in6 *dst;
691 	struct mbuf *m;
692 	int off;	/* offset of udphdr */
693 {
694 	u_int16_t sport, dport;
695 	int rcvcnt;
696 	struct in6_addr src6, dst6;
697 	const struct in_addr *dst4;
698 	struct inpcb_hdr *inph;
699 	struct in6pcb *in6p;
700 
701 	rcvcnt = 0;
702 	off += sizeof(struct udphdr);	/* now, offset of payload */
703 
704 	if (af != AF_INET && af != AF_INET6)
705 		goto bad;
706 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
707 		goto bad;
708 
709 	in6_embedscope(&src6, src, NULL, NULL);
710 	sport = src->sin6_port;
711 	in6_embedscope(&dst6, dst, NULL, NULL);
712 	dport = dst->sin6_port;
713 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
714 
715 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
716 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
717 		/*
718 		 * Deliver a multicast or broadcast datagram to *all* sockets
719 		 * for which the local and remote addresses and ports match
720 		 * those of the incoming datagram.  This allows more than
721 		 * one process to receive multi/broadcasts on the same port.
722 		 * (This really ought to be done for unicast datagrams as
723 		 * well, but that would cause problems with existing
724 		 * applications that open both address-specific sockets and
725 		 * a wildcard socket listening to the same port -- they would
726 		 * end up receiving duplicates of every unicast datagram.
727 		 * Those applications open the multiple sockets to overcome an
728 		 * inadequacy of the UDP socket interface, but for backwards
729 		 * compatibility we avoid the problem here rather than
730 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
731 		 */
732 
733 		/*
734 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
735 		 * we need udpiphdr for IPsec processing so we do that later.
736 		 */
737 		/*
738 		 * Locate pcb(s) for datagram.
739 		 */
740 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
741 			in6p = (struct in6pcb *)inph;
742 			if (in6p->in6p_af != AF_INET6)
743 				continue;
744 
745 			if (in6p->in6p_lport != dport)
746 				continue;
747 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
748 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
749 					continue;
750 			} else {
751 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
752 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
753 					continue;
754 			}
755 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
756 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
757 				    &src6) || in6p->in6p_fport != sport)
758 					continue;
759 			} else {
760 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
761 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
762 					continue;
763 			}
764 
765 			udp6_sendup(m, off, (struct sockaddr *)src,
766 				in6p->in6p_socket);
767 			rcvcnt++;
768 
769 			/*
770 			 * Don't look for additional matches if this one does
771 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
772 			 * socket options set.  This heuristic avoids searching
773 			 * through all pcbs in the common case of a non-shared
774 			 * port.  It assumes that an application will never
775 			 * clear these options after setting them.
776 			 */
777 			if ((in6p->in6p_socket->so_options &
778 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
779 				break;
780 		}
781 	} else {
782 		/*
783 		 * Locate pcb for datagram.
784 		 */
785 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
786 		    &dst6, dport, 0);
787 		if (in6p == 0) {
788 			++udpstat.udps_pcbhashmiss;
789 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
790 			if (in6p == 0)
791 				return rcvcnt;
792 		}
793 
794 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
795 		rcvcnt++;
796 	}
797 
798 bad:
799 	return rcvcnt;
800 }
801 #endif
802 
803 #ifdef INET
804 /*
805  * Notify a udp user of an asynchronous error;
806  * just wake up so that he can collect error status.
807  */
808 static void
809 udp_notify(inp, errno)
810 	struct inpcb *inp;
811 	int errno;
812 {
813 
814 	inp->inp_socket->so_error = errno;
815 	sorwakeup(inp->inp_socket);
816 	sowwakeup(inp->inp_socket);
817 }
818 
819 void *
820 udp_ctlinput(cmd, sa, v)
821 	int cmd;
822 	struct sockaddr *sa;
823 	void *v;
824 {
825 	struct ip *ip = v;
826 	struct udphdr *uh;
827 	void (*notify) __P((struct inpcb *, int)) = udp_notify;
828 	int errno;
829 
830 	if (sa->sa_family != AF_INET
831 	 || sa->sa_len != sizeof(struct sockaddr_in))
832 		return NULL;
833 	if ((unsigned)cmd >= PRC_NCMDS)
834 		return NULL;
835 	errno = inetctlerrmap[cmd];
836 	if (PRC_IS_REDIRECT(cmd))
837 		notify = in_rtchange, ip = 0;
838 	else if (cmd == PRC_HOSTDEAD)
839 		ip = 0;
840 	else if (errno == 0)
841 		return NULL;
842 	if (ip) {
843 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
844 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
845 		    ip->ip_src, uh->uh_sport, errno, notify);
846 
847 		/* XXX mapped address case */
848 	} else
849 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
850 		    notify);
851 	return NULL;
852 }
853 
854 int
855 #if __STDC__
856 udp_output(struct mbuf *m, ...)
857 #else
858 udp_output(m, va_alist)
859 	struct mbuf *m;
860 	va_dcl
861 #endif
862 {
863 	struct inpcb *inp;
864 	struct udpiphdr *ui;
865 	int len = m->m_pkthdr.len;
866 	int error = 0;
867 	va_list ap;
868 
869 	MCLAIM(m, &udp_tx_mowner);
870 	va_start(ap, m);
871 	inp = va_arg(ap, struct inpcb *);
872 	va_end(ap);
873 
874 	/*
875 	 * Calculate data length and get a mbuf
876 	 * for UDP and IP headers.
877 	 */
878 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
879 	if (m == 0) {
880 		error = ENOBUFS;
881 		goto release;
882 	}
883 
884 	/*
885 	 * Compute the packet length of the IP header, and
886 	 * punt if the length looks bogus.
887 	 */
888 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
889 		error = EMSGSIZE;
890 		goto release;
891 	}
892 
893 	/*
894 	 * Fill in mbuf with extended UDP header
895 	 * and addresses and length put into network format.
896 	 */
897 	ui = mtod(m, struct udpiphdr *);
898 	ui->ui_pr = IPPROTO_UDP;
899 	ui->ui_src = inp->inp_laddr;
900 	ui->ui_dst = inp->inp_faddr;
901 	ui->ui_sport = inp->inp_lport;
902 	ui->ui_dport = inp->inp_fport;
903 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
904 
905 	/*
906 	 * Set up checksum and output datagram.
907 	 */
908 	if (udpcksum) {
909 		/*
910 		 * XXX Cache pseudo-header checksum part for
911 		 * XXX "connected" UDP sockets.
912 		 */
913 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
914 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
915 		    sizeof(struct udphdr) + IPPROTO_UDP));
916 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
917 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
918 	} else
919 		ui->ui_sum = 0;
920 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
921 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
922 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
923 	udpstat.udps_opackets++;
924 
925 	return (ip_output(m, inp->inp_options, &inp->inp_route,
926 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
927 	    inp->inp_moptions, inp->inp_socket));
928 
929 release:
930 	m_freem(m);
931 	return (error);
932 }
933 
934 int	udp_sendspace = 9216;		/* really max datagram size */
935 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
936 					/* 40 1K datagrams */
937 
938 /*ARGSUSED*/
939 int
940 udp_usrreq(so, req, m, nam, control, p)
941 	struct socket *so;
942 	int req;
943 	struct mbuf *m, *nam, *control;
944 	struct proc *p;
945 {
946 	struct inpcb *inp;
947 	int s;
948 	int error = 0;
949 
950 	if (req == PRU_CONTROL)
951 		return (in_control(so, (long)m, (caddr_t)nam,
952 		    (struct ifnet *)control, p));
953 
954 	if (req == PRU_PURGEIF) {
955 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
956 		in_purgeif((struct ifnet *)control);
957 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
958 		return (0);
959 	}
960 
961 	s = splsoftnet();
962 	inp = sotoinpcb(so);
963 #ifdef DIAGNOSTIC
964 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
965 		panic("udp_usrreq: unexpected control mbuf");
966 #endif
967 	if (inp == 0 && req != PRU_ATTACH) {
968 		error = EINVAL;
969 		goto release;
970 	}
971 
972 	/*
973 	 * Note: need to block udp_input while changing
974 	 * the udp pcb queue and/or pcb addresses.
975 	 */
976 	switch (req) {
977 
978 	case PRU_ATTACH:
979 		if (inp != 0) {
980 			error = EISCONN;
981 			break;
982 		}
983 #ifdef MBUFTRACE
984 		so->so_mowner = &udp_mowner;
985 		so->so_rcv.sb_mowner = &udp_rx_mowner;
986 		so->so_snd.sb_mowner = &udp_tx_mowner;
987 #endif
988 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
989 			error = soreserve(so, udp_sendspace, udp_recvspace);
990 			if (error)
991 				break;
992 		}
993 		error = in_pcballoc(so, &udbtable);
994 		if (error)
995 			break;
996 		inp = sotoinpcb(so);
997 		inp->inp_ip.ip_ttl = ip_defttl;
998 		break;
999 
1000 	case PRU_DETACH:
1001 		in_pcbdetach(inp);
1002 		break;
1003 
1004 	case PRU_BIND:
1005 		error = in_pcbbind(inp, nam, p);
1006 		break;
1007 
1008 	case PRU_LISTEN:
1009 		error = EOPNOTSUPP;
1010 		break;
1011 
1012 	case PRU_CONNECT:
1013 		error = in_pcbconnect(inp, nam);
1014 		if (error)
1015 			break;
1016 		soisconnected(so);
1017 		break;
1018 
1019 	case PRU_CONNECT2:
1020 		error = EOPNOTSUPP;
1021 		break;
1022 
1023 	case PRU_DISCONNECT:
1024 		/*soisdisconnected(so);*/
1025 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1026 		in_pcbdisconnect(inp);
1027 		inp->inp_laddr = zeroin_addr;		/* XXX */
1028 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1029 		break;
1030 
1031 	case PRU_SHUTDOWN:
1032 		socantsendmore(so);
1033 		break;
1034 
1035 	case PRU_RCVD:
1036 		error = EOPNOTSUPP;
1037 		break;
1038 
1039 	case PRU_SEND:
1040 		if (control && control->m_len) {
1041 			m_freem(control);
1042 			m_freem(m);
1043 			error = EINVAL;
1044 			break;
1045 		}
1046 	{
1047 		struct in_addr laddr;			/* XXX */
1048 
1049 		if (nam) {
1050 			laddr = inp->inp_laddr;		/* XXX */
1051 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1052 				error = EISCONN;
1053 				goto die;
1054 			}
1055 			error = in_pcbconnect(inp, nam);
1056 			if (error)
1057 				goto die;
1058 		} else {
1059 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1060 				error = ENOTCONN;
1061 				goto die;
1062 			}
1063 		}
1064 		error = udp_output(m, inp);
1065 		m = NULL;
1066 		if (nam) {
1067 			in_pcbdisconnect(inp);
1068 			inp->inp_laddr = laddr;		/* XXX */
1069 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1070 		}
1071 	  die:
1072 		if (m)
1073 			m_freem(m);
1074 	}
1075 		break;
1076 
1077 	case PRU_SENSE:
1078 		/*
1079 		 * stat: don't bother with a blocksize.
1080 		 */
1081 		splx(s);
1082 		return (0);
1083 
1084 	case PRU_RCVOOB:
1085 		error =  EOPNOTSUPP;
1086 		break;
1087 
1088 	case PRU_SENDOOB:
1089 		m_freem(control);
1090 		m_freem(m);
1091 		error =  EOPNOTSUPP;
1092 		break;
1093 
1094 	case PRU_SOCKADDR:
1095 		in_setsockaddr(inp, nam);
1096 		break;
1097 
1098 	case PRU_PEERADDR:
1099 		in_setpeeraddr(inp, nam);
1100 		break;
1101 
1102 	default:
1103 		panic("udp_usrreq");
1104 	}
1105 
1106 release:
1107 	splx(s);
1108 	return (error);
1109 }
1110 
1111 /*
1112  * Sysctl for udp variables.
1113  */
1114 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1115 {
1116 
1117 	sysctl_createv(clog, 0, NULL, NULL,
1118 		       CTLFLAG_PERMANENT,
1119 		       CTLTYPE_NODE, "net", NULL,
1120 		       NULL, 0, NULL, 0,
1121 		       CTL_NET, CTL_EOL);
1122 	sysctl_createv(clog, 0, NULL, NULL,
1123 		       CTLFLAG_PERMANENT,
1124 		       CTLTYPE_NODE, "inet", NULL,
1125 		       NULL, 0, NULL, 0,
1126 		       CTL_NET, PF_INET, CTL_EOL);
1127 	sysctl_createv(clog, 0, NULL, NULL,
1128 		       CTLFLAG_PERMANENT,
1129 		       CTLTYPE_NODE, "udp", NULL,
1130 		       NULL, 0, NULL, 0,
1131 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1132 
1133 	sysctl_createv(clog, 0, NULL, NULL,
1134 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1135 		       CTLTYPE_INT, "checksum", NULL,
1136 		       NULL, 0, &udpcksum, 0,
1137 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1138 		       CTL_EOL);
1139 	sysctl_createv(clog, 0, NULL, NULL,
1140 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1141 		       CTLTYPE_INT, "sendspace", NULL,
1142 		       NULL, 0, &udp_sendspace, 0,
1143 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1144 		       CTL_EOL);
1145 	sysctl_createv(clog, 0, NULL, NULL,
1146 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1147 		       CTLTYPE_INT, "recvspace", NULL,
1148 		       NULL, 0, &udp_recvspace, 0,
1149 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1150 		       CTL_EOL);
1151 }
1152 #endif
1153