xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision da5f4674a3fc214be3572d358b66af40ab9401e7)
1 /*	$NetBSD: udp_usrreq.c,v 1.108 2003/08/22 22:00:38 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 1.108 2003/08/22 22:00:38 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 #ifdef INET
199 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
200 #endif
201 
202 #ifdef UDP_CSUM_COUNTERS
203 	evcnt_attach_static(&udp_hwcsum_bad);
204 	evcnt_attach_static(&udp_hwcsum_ok);
205 	evcnt_attach_static(&udp_hwcsum_data);
206 	evcnt_attach_static(&udp_swcsum);
207 #endif /* UDP_CSUM_COUNTERS */
208 
209 	MOWNER_ATTACH(&udp_tx_mowner);
210 	MOWNER_ATTACH(&udp_rx_mowner);
211 	MOWNER_ATTACH(&udp_mowner);
212 }
213 
214 #ifdef INET
215 void
216 #if __STDC__
217 udp_input(struct mbuf *m, ...)
218 #else
219 udp_input(m, va_alist)
220 	struct mbuf *m;
221 	va_dcl
222 #endif
223 {
224 	va_list ap;
225 	struct sockaddr_in src, dst;
226 	struct ip *ip;
227 	struct udphdr *uh;
228 	int iphlen;
229 	int len;
230 	int n;
231 	u_int16_t ip_len;
232 
233 	va_start(ap, m);
234 	iphlen = va_arg(ap, int);
235 	(void)va_arg(ap, int);		/* ignore value, advance ap */
236 	va_end(ap);
237 
238 	MCLAIM(m, &udp_rx_mowner);
239 	udpstat.udps_ipackets++;
240 
241 	/*
242 	 * Get IP and UDP header together in first mbuf.
243 	 */
244 	ip = mtod(m, struct ip *);
245 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
246 	if (uh == NULL) {
247 		udpstat.udps_hdrops++;
248 		return;
249 	}
250 	KASSERT(UDP_HDR_ALIGNED_P(uh));
251 
252 	/* destination port of 0 is illegal, based on RFC768. */
253 	if (uh->uh_dport == 0)
254 		goto bad;
255 
256 	/*
257 	 * Make mbuf data length reflect UDP length.
258 	 * If not enough data to reflect UDP length, drop.
259 	 */
260 	ip_len = ntohs(ip->ip_len);
261 	len = ntohs((u_int16_t)uh->uh_ulen);
262 	if (ip_len != iphlen + len) {
263 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
264 			udpstat.udps_badlen++;
265 			goto bad;
266 		}
267 		m_adj(m, iphlen + len - ip_len);
268 	}
269 
270 	/*
271 	 * Checksum extended UDP header and data.
272 	 */
273 	if (uh->uh_sum) {
274 		switch (m->m_pkthdr.csum_flags &
275 			((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
276 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
277 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
278 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
279 			goto badcsum;
280 
281 		case M_CSUM_UDPv4|M_CSUM_DATA: {
282 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
283 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
284 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR)
285 			  hw_csum = in_cksum_phdr(ip->ip_src.s_addr, ip->ip_dst.s_addr,
286 				htonl(hw_csum + ntohs(ip->ip_len) + IPPROTO_UDP));
287 			if ((hw_csum ^ 0xffff) != 0)
288 				goto badcsum;
289 			break;
290 		}
291 
292 		case M_CSUM_UDPv4:
293 			/* Checksum was okay. */
294 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
295 			break;
296 
297 		default:
298 			/* Need to compute it ourselves. */
299 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
300 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
301 				goto badcsum;
302 			break;
303 		}
304 	}
305 
306 	/* construct source and dst sockaddrs. */
307 	bzero(&src, sizeof(src));
308 	src.sin_family = AF_INET;
309 	src.sin_len = sizeof(struct sockaddr_in);
310 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
311 	src.sin_port = uh->uh_sport;
312 	bzero(&dst, sizeof(dst));
313 	dst.sin_family = AF_INET;
314 	dst.sin_len = sizeof(struct sockaddr_in);
315 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
316 	dst.sin_port = uh->uh_dport;
317 
318 	n = udp4_realinput(&src, &dst, m, iphlen);
319 #ifdef INET6
320 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
321 		struct sockaddr_in6 src6, dst6;
322 
323 		bzero(&src6, sizeof(src6));
324 		src6.sin6_family = AF_INET6;
325 		src6.sin6_len = sizeof(struct sockaddr_in6);
326 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
327 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
328 			sizeof(ip->ip_src));
329 		src6.sin6_port = uh->uh_sport;
330 		bzero(&dst6, sizeof(dst6));
331 		dst6.sin6_family = AF_INET6;
332 		dst6.sin6_len = sizeof(struct sockaddr_in6);
333 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
334 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
335 			sizeof(ip->ip_dst));
336 		dst6.sin6_port = uh->uh_dport;
337 
338 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
339 	}
340 #endif
341 
342 	if (n == 0) {
343 		if (m->m_flags & (M_BCAST | M_MCAST)) {
344 			udpstat.udps_noportbcast++;
345 			goto bad;
346 		}
347 		udpstat.udps_noport++;
348 #ifdef IPKDB
349 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
350 				m, iphlen + sizeof(struct udphdr),
351 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
352 			/*
353 			 * It was a debugger connect packet,
354 			 * just drop it now
355 			 */
356 			goto bad;
357 		}
358 #endif
359 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
360 		m = NULL;
361 	}
362 
363 bad:
364 	if (m)
365 		m_freem(m);
366 	return;
367 
368 badcsum:
369 	m_freem(m);
370 	udpstat.udps_badsum++;
371 }
372 #endif
373 
374 #ifdef INET6
375 int
376 udp6_input(mp, offp, proto)
377 	struct mbuf **mp;
378 	int *offp, proto;
379 {
380 	struct mbuf *m = *mp;
381 	int off = *offp;
382 	struct sockaddr_in6 src, dst;
383 	struct ip6_hdr *ip6;
384 	struct udphdr *uh;
385 	u_int32_t plen, ulen;
386 
387 	ip6 = mtod(m, struct ip6_hdr *);
388 
389 #if defined(NFAITH) && 0 < NFAITH
390 	if (faithprefix(&ip6->ip6_dst)) {
391 		/* send icmp6 host unreach? */
392 		m_freem(m);
393 		return IPPROTO_DONE;
394 	}
395 #endif
396 
397 	udp6stat.udp6s_ipackets++;
398 
399 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
400 	plen = m->m_pkthdr.len - off;
401 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
402 	if (uh == NULL) {
403 		ip6stat.ip6s_tooshort++;
404 		return IPPROTO_DONE;
405 	}
406 	KASSERT(UDP_HDR_ALIGNED_P(uh));
407 	ulen = ntohs((u_short)uh->uh_ulen);
408 	/*
409 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
410 	 * iff payload length > 0xffff.
411 	 */
412 	if (ulen == 0 && plen > 0xffff)
413 		ulen = plen;
414 
415 	if (plen != ulen) {
416 		udp6stat.udp6s_badlen++;
417 		goto bad;
418 	}
419 
420 	/* destination port of 0 is illegal, based on RFC768. */
421 	if (uh->uh_dport == 0)
422 		goto bad;
423 
424 	/* Be proactive about malicious use of IPv4 mapped address */
425 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
426 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
427 		/* XXX stat */
428 		goto bad;
429 	}
430 
431 	/*
432 	 * Checksum extended UDP header and data.
433 	 */
434 	if (uh->uh_sum == 0)
435 		udp6stat.udp6s_nosum++;
436 	else if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
437 		udp6stat.udp6s_badsum++;
438 		goto bad;
439 	}
440 
441 	/*
442 	 * Construct source and dst sockaddrs.
443 	 * Note that ifindex (s6_addr16[1]) is already filled.
444 	 */
445 	bzero(&src, sizeof(src));
446 	src.sin6_family = AF_INET6;
447 	src.sin6_len = sizeof(struct sockaddr_in6);
448 	/* KAME hack: recover scopeid */
449 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
450 	src.sin6_port = uh->uh_sport;
451 	bzero(&dst, sizeof(dst));
452 	dst.sin6_family = AF_INET6;
453 	dst.sin6_len = sizeof(struct sockaddr_in6);
454 	/* KAME hack: recover scopeid */
455 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
456 	dst.sin6_port = uh->uh_dport;
457 
458 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
459 		if (m->m_flags & M_MCAST) {
460 			udp6stat.udp6s_noportmcast++;
461 			goto bad;
462 		}
463 		udp6stat.udp6s_noport++;
464 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
465 		m = NULL;
466 	}
467 
468 bad:
469 	if (m)
470 		m_freem(m);
471 	return IPPROTO_DONE;
472 }
473 #endif
474 
475 #ifdef INET
476 static void
477 udp4_sendup(m, off, src, so)
478 	struct mbuf *m;
479 	int off;	/* offset of data portion */
480 	struct sockaddr *src;
481 	struct socket *so;
482 {
483 	struct mbuf *opts = NULL;
484 	struct mbuf *n;
485 	struct inpcb *inp = NULL;
486 
487 	if (!so)
488 		return;
489 	switch (so->so_proto->pr_domain->dom_family) {
490 	case AF_INET:
491 		inp = sotoinpcb(so);
492 		break;
493 #ifdef INET6
494 	case AF_INET6:
495 		break;
496 #endif
497 	default:
498 		return;
499 	}
500 
501 #if defined(IPSEC) || defined(FAST_IPSEC)
502 	/* check AH/ESP integrity. */
503 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
504 		ipsecstat.in_polvio++;
505 		return;
506 	}
507 #endif /*IPSEC*/
508 
509 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
510 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
511 			 || so->so_options & SO_TIMESTAMP)) {
512 			struct ip *ip = mtod(n, struct ip *);
513 			ip_savecontrol(inp, &opts, ip, n);
514 		}
515 
516 		m_adj(n, off);
517 		if (sbappendaddr(&so->so_rcv, src, n,
518 				opts) == 0) {
519 			m_freem(n);
520 			if (opts)
521 				m_freem(opts);
522 			udpstat.udps_fullsock++;
523 		} else
524 			sorwakeup(so);
525 	}
526 }
527 #endif
528 
529 #ifdef INET6
530 static void
531 udp6_sendup(m, off, src, so)
532 	struct mbuf *m;
533 	int off;	/* offset of data portion */
534 	struct sockaddr *src;
535 	struct socket *so;
536 {
537 	struct mbuf *opts = NULL;
538 	struct mbuf *n;
539 	struct in6pcb *in6p = NULL;
540 
541 	if (!so)
542 		return;
543 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
544 		return;
545 	in6p = sotoin6pcb(so);
546 
547 #if defined(IPSEC) || defined(FAST_IPSEC)
548 	/* check AH/ESP integrity. */
549 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
550 		ipsec6stat.in_polvio++;
551 		return;
552 	}
553 #endif /*IPSEC*/
554 
555 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
556 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
557 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
558 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
559 			ip6_savecontrol(in6p, &opts, ip6, n);
560 		}
561 
562 		m_adj(n, off);
563 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
564 			m_freem(n);
565 			if (opts)
566 				m_freem(opts);
567 			udp6stat.udp6s_fullsock++;
568 		} else
569 			sorwakeup(so);
570 	}
571 }
572 #endif
573 
574 #ifdef INET
575 static int
576 udp4_realinput(src, dst, m, off)
577 	struct sockaddr_in *src;
578 	struct sockaddr_in *dst;
579 	struct mbuf *m;
580 	int off;	/* offset of udphdr */
581 {
582 	u_int16_t *sport, *dport;
583 	int rcvcnt;
584 	struct in_addr *src4, *dst4;
585 	struct inpcb *inp;
586 
587 	rcvcnt = 0;
588 	off += sizeof(struct udphdr);	/* now, offset of payload */
589 
590 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
591 		goto bad;
592 
593 	src4 = &src->sin_addr;
594 	sport = &src->sin_port;
595 	dst4 = &dst->sin_addr;
596 	dport = &dst->sin_port;
597 
598 	if (IN_MULTICAST(dst4->s_addr) ||
599 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
600 		/*
601 		 * Deliver a multicast or broadcast datagram to *all* sockets
602 		 * for which the local and remote addresses and ports match
603 		 * those of the incoming datagram.  This allows more than
604 		 * one process to receive multi/broadcasts on the same port.
605 		 * (This really ought to be done for unicast datagrams as
606 		 * well, but that would cause problems with existing
607 		 * applications that open both address-specific sockets and
608 		 * a wildcard socket listening to the same port -- they would
609 		 * end up receiving duplicates of every unicast datagram.
610 		 * Those applications open the multiple sockets to overcome an
611 		 * inadequacy of the UDP socket interface, but for backwards
612 		 * compatibility we avoid the problem here rather than
613 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
614 		 */
615 
616 		/*
617 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
618 		 * we need udpiphdr for IPsec processing so we do that later.
619 		 */
620 		/*
621 		 * Locate pcb(s) for datagram.
622 		 */
623 		CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
624 			if (inp->inp_lport != *dport)
625 				continue;
626 			if (!in_nullhost(inp->inp_laddr)) {
627 				if (!in_hosteq(inp->inp_laddr, *dst4))
628 					continue;
629 			}
630 			if (!in_nullhost(inp->inp_faddr)) {
631 				if (!in_hosteq(inp->inp_faddr, *src4) ||
632 				    inp->inp_fport != *sport)
633 					continue;
634 			}
635 
636 			udp4_sendup(m, off, (struct sockaddr *)src,
637 				inp->inp_socket);
638 			rcvcnt++;
639 
640 			/*
641 			 * Don't look for additional matches if this one does
642 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
643 			 * socket options set.  This heuristic avoids searching
644 			 * through all pcbs in the common case of a non-shared
645 			 * port.  It assumes that an application will never
646 			 * clear these options after setting them.
647 			 */
648 			if ((inp->inp_socket->so_options &
649 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
650 				break;
651 		}
652 	} else {
653 		/*
654 		 * Locate pcb for datagram.
655 		 */
656 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
657 		if (inp == 0) {
658 			++udpstat.udps_pcbhashmiss;
659 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
660 			if (inp == 0)
661 				return rcvcnt;
662 		}
663 
664 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
665 		rcvcnt++;
666 	}
667 
668 bad:
669 	return rcvcnt;
670 }
671 #endif
672 
673 #ifdef INET6
674 static int
675 udp6_realinput(af, src, dst, m, off)
676 	int af;		/* af on packet */
677 	struct sockaddr_in6 *src;
678 	struct sockaddr_in6 *dst;
679 	struct mbuf *m;
680 	int off;	/* offset of udphdr */
681 {
682 	u_int16_t sport, dport;
683 	int rcvcnt;
684 	struct in6_addr src6, dst6;
685 	const struct in_addr *dst4;
686 	struct in6pcb *in6p;
687 
688 	rcvcnt = 0;
689 	off += sizeof(struct udphdr);	/* now, offset of payload */
690 
691 	if (af != AF_INET && af != AF_INET6)
692 		goto bad;
693 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
694 		goto bad;
695 
696 	in6_embedscope(&src6, src, NULL, NULL);
697 	sport = src->sin6_port;
698 	in6_embedscope(&dst6, dst, NULL, NULL);
699 	dport = dst->sin6_port;
700 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
701 
702 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
703 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
704 		/*
705 		 * Deliver a multicast or broadcast datagram to *all* sockets
706 		 * for which the local and remote addresses and ports match
707 		 * those of the incoming datagram.  This allows more than
708 		 * one process to receive multi/broadcasts on the same port.
709 		 * (This really ought to be done for unicast datagrams as
710 		 * well, but that would cause problems with existing
711 		 * applications that open both address-specific sockets and
712 		 * a wildcard socket listening to the same port -- they would
713 		 * end up receiving duplicates of every unicast datagram.
714 		 * Those applications open the multiple sockets to overcome an
715 		 * inadequacy of the UDP socket interface, but for backwards
716 		 * compatibility we avoid the problem here rather than
717 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
718 		 */
719 
720 		/*
721 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
722 		 * we need udpiphdr for IPsec processing so we do that later.
723 		 */
724 		/*
725 		 * Locate pcb(s) for datagram.
726 		 */
727 		for (in6p = udb6.in6p_next; in6p != &udb6;
728 		     in6p = in6p->in6p_next) {
729 			if (in6p->in6p_lport != dport)
730 				continue;
731 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
732 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
733 					continue;
734 			} else {
735 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
736 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
737 					continue;
738 			}
739 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
740 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
741 				    &src6) || in6p->in6p_fport != sport)
742 					continue;
743 			} else {
744 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
745 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
746 					continue;
747 			}
748 
749 			udp6_sendup(m, off, (struct sockaddr *)src,
750 				in6p->in6p_socket);
751 			rcvcnt++;
752 
753 			/*
754 			 * Don't look for additional matches if this one does
755 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
756 			 * socket options set.  This heuristic avoids searching
757 			 * through all pcbs in the common case of a non-shared
758 			 * port.  It assumes that an application will never
759 			 * clear these options after setting them.
760 			 */
761 			if ((in6p->in6p_socket->so_options &
762 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
763 				break;
764 		}
765 	} else {
766 		/*
767 		 * Locate pcb for datagram.
768 		 */
769 		in6p = in6_pcblookup_connect(&udb6, &src6, sport,
770 		    &dst6, dport, 0);
771 		if (in6p == 0) {
772 			++udpstat.udps_pcbhashmiss;
773 			in6p = in6_pcblookup_bind(&udb6, &dst6, dport, 0);
774 			if (in6p == 0)
775 				return rcvcnt;
776 		}
777 
778 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
779 		rcvcnt++;
780 	}
781 
782 bad:
783 	return rcvcnt;
784 }
785 #endif
786 
787 #ifdef INET
788 /*
789  * Notify a udp user of an asynchronous error;
790  * just wake up so that he can collect error status.
791  */
792 static void
793 udp_notify(inp, errno)
794 	struct inpcb *inp;
795 	int errno;
796 {
797 
798 	inp->inp_socket->so_error = errno;
799 	sorwakeup(inp->inp_socket);
800 	sowwakeup(inp->inp_socket);
801 }
802 
803 void *
804 udp_ctlinput(cmd, sa, v)
805 	int cmd;
806 	struct sockaddr *sa;
807 	void *v;
808 {
809 	struct ip *ip = v;
810 	struct udphdr *uh;
811 	void (*notify) __P((struct inpcb *, int)) = udp_notify;
812 	int errno;
813 
814 	if (sa->sa_family != AF_INET
815 	 || sa->sa_len != sizeof(struct sockaddr_in))
816 		return NULL;
817 	if ((unsigned)cmd >= PRC_NCMDS)
818 		return NULL;
819 	errno = inetctlerrmap[cmd];
820 	if (PRC_IS_REDIRECT(cmd))
821 		notify = in_rtchange, ip = 0;
822 	else if (cmd == PRC_HOSTDEAD)
823 		ip = 0;
824 	else if (errno == 0)
825 		return NULL;
826 	if (ip) {
827 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
828 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
829 		    ip->ip_src, uh->uh_sport, errno, notify);
830 
831 		/* XXX mapped address case */
832 	} else
833 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
834 		    notify);
835 	return NULL;
836 }
837 
838 int
839 #if __STDC__
840 udp_output(struct mbuf *m, ...)
841 #else
842 udp_output(m, va_alist)
843 	struct mbuf *m;
844 	va_dcl
845 #endif
846 {
847 	struct inpcb *inp;
848 	struct udpiphdr *ui;
849 	int len = m->m_pkthdr.len;
850 	int error = 0;
851 	va_list ap;
852 
853 	MCLAIM(m, &udp_tx_mowner);
854 	va_start(ap, m);
855 	inp = va_arg(ap, struct inpcb *);
856 	va_end(ap);
857 
858 	/*
859 	 * Calculate data length and get a mbuf
860 	 * for UDP and IP headers.
861 	 */
862 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
863 	if (m == 0) {
864 		error = ENOBUFS;
865 		goto release;
866 	}
867 
868 	/*
869 	 * Compute the packet length of the IP header, and
870 	 * punt if the length looks bogus.
871 	 */
872 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
873 		error = EMSGSIZE;
874 		goto release;
875 	}
876 
877 	/*
878 	 * Fill in mbuf with extended UDP header
879 	 * and addresses and length put into network format.
880 	 */
881 	ui = mtod(m, struct udpiphdr *);
882 	ui->ui_pr = IPPROTO_UDP;
883 	ui->ui_src = inp->inp_laddr;
884 	ui->ui_dst = inp->inp_faddr;
885 	ui->ui_sport = inp->inp_lport;
886 	ui->ui_dport = inp->inp_fport;
887 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
888 
889 	/*
890 	 * Set up checksum and output datagram.
891 	 */
892 	if (udpcksum) {
893 		/*
894 		 * XXX Cache pseudo-header checksum part for
895 		 * XXX "connected" UDP sockets.
896 		 */
897 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
898 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
899 		    sizeof(struct udphdr) + IPPROTO_UDP));
900 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
901 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
902 	} else
903 		ui->ui_sum = 0;
904 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
905 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
906 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
907 	udpstat.udps_opackets++;
908 
909 	return (ip_output(m, inp->inp_options, &inp->inp_route,
910 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
911 	    inp->inp_moptions, inp->inp_socket));
912 
913 release:
914 	m_freem(m);
915 	return (error);
916 }
917 
918 int	udp_sendspace = 9216;		/* really max datagram size */
919 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
920 					/* 40 1K datagrams */
921 
922 /*ARGSUSED*/
923 int
924 udp_usrreq(so, req, m, nam, control, p)
925 	struct socket *so;
926 	int req;
927 	struct mbuf *m, *nam, *control;
928 	struct proc *p;
929 {
930 	struct inpcb *inp;
931 	int s;
932 	int error = 0;
933 
934 	if (req == PRU_CONTROL)
935 		return (in_control(so, (long)m, (caddr_t)nam,
936 		    (struct ifnet *)control, p));
937 
938 	if (req == PRU_PURGEIF) {
939 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
940 		in_purgeif((struct ifnet *)control);
941 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
942 		return (0);
943 	}
944 
945 	s = splsoftnet();
946 	inp = sotoinpcb(so);
947 #ifdef DIAGNOSTIC
948 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
949 		panic("udp_usrreq: unexpected control mbuf");
950 #endif
951 	if (inp == 0 && req != PRU_ATTACH) {
952 		error = EINVAL;
953 		goto release;
954 	}
955 
956 	/*
957 	 * Note: need to block udp_input while changing
958 	 * the udp pcb queue and/or pcb addresses.
959 	 */
960 	switch (req) {
961 
962 	case PRU_ATTACH:
963 		if (inp != 0) {
964 			error = EISCONN;
965 			break;
966 		}
967 #ifdef MBUFTRACE
968 		so->so_mowner = &udp_mowner;
969 		so->so_rcv.sb_mowner = &udp_rx_mowner;
970 		so->so_snd.sb_mowner = &udp_tx_mowner;
971 #endif
972 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
973 			error = soreserve(so, udp_sendspace, udp_recvspace);
974 			if (error)
975 				break;
976 		}
977 		error = in_pcballoc(so, &udbtable);
978 		if (error)
979 			break;
980 		inp = sotoinpcb(so);
981 		inp->inp_ip.ip_ttl = ip_defttl;
982 		break;
983 
984 	case PRU_DETACH:
985 		in_pcbdetach(inp);
986 		break;
987 
988 	case PRU_BIND:
989 		error = in_pcbbind(inp, nam, p);
990 		break;
991 
992 	case PRU_LISTEN:
993 		error = EOPNOTSUPP;
994 		break;
995 
996 	case PRU_CONNECT:
997 		error = in_pcbconnect(inp, nam);
998 		if (error)
999 			break;
1000 		soisconnected(so);
1001 		break;
1002 
1003 	case PRU_CONNECT2:
1004 		error = EOPNOTSUPP;
1005 		break;
1006 
1007 	case PRU_DISCONNECT:
1008 		/*soisdisconnected(so);*/
1009 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1010 		in_pcbdisconnect(inp);
1011 		inp->inp_laddr = zeroin_addr;		/* XXX */
1012 		if (inp->inp_ia != NULL) {
1013 			LIST_REMOVE(inp, inp_ialink);
1014 			IFAFREE(&inp->inp_ia->ia_ifa);
1015 			inp->inp_ia = NULL;
1016 		}
1017 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1018 		break;
1019 
1020 	case PRU_SHUTDOWN:
1021 		socantsendmore(so);
1022 		break;
1023 
1024 	case PRU_RCVD:
1025 		error = EOPNOTSUPP;
1026 		break;
1027 
1028 	case PRU_SEND:
1029 		if (control && control->m_len) {
1030 			m_freem(control);
1031 			m_freem(m);
1032 			error = EINVAL;
1033 			break;
1034 		}
1035 	{
1036 		struct in_addr laddr;			/* XXX */
1037 
1038 		if (nam) {
1039 			laddr = inp->inp_laddr;		/* XXX */
1040 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1041 				error = EISCONN;
1042 				goto die;
1043 			}
1044 			error = in_pcbconnect(inp, nam);
1045 			if (error)
1046 				goto die;
1047 		} else {
1048 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1049 				error = ENOTCONN;
1050 				goto die;
1051 			}
1052 		}
1053 		error = udp_output(m, inp);
1054 		m = NULL;
1055 		if (nam) {
1056 			in_pcbdisconnect(inp);
1057 			inp->inp_laddr = laddr;		/* XXX */
1058 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1059 		}
1060 	  die:
1061 		if (inp->inp_ia != NULL && in_nullhost(inp->inp_laddr)) {
1062 			LIST_REMOVE(inp, inp_ialink);
1063 			IFAFREE(&inp->inp_ia->ia_ifa);
1064 			inp->inp_ia = NULL;
1065 		}
1066 		if (m)
1067 			m_freem(m);
1068 	}
1069 		break;
1070 
1071 	case PRU_SENSE:
1072 		/*
1073 		 * stat: don't bother with a blocksize.
1074 		 */
1075 		splx(s);
1076 		return (0);
1077 
1078 	case PRU_RCVOOB:
1079 		error =  EOPNOTSUPP;
1080 		break;
1081 
1082 	case PRU_SENDOOB:
1083 		m_freem(control);
1084 		m_freem(m);
1085 		error =  EOPNOTSUPP;
1086 		break;
1087 
1088 	case PRU_SOCKADDR:
1089 		in_setsockaddr(inp, nam);
1090 		break;
1091 
1092 	case PRU_PEERADDR:
1093 		in_setpeeraddr(inp, nam);
1094 		break;
1095 
1096 	default:
1097 		panic("udp_usrreq");
1098 	}
1099 
1100 release:
1101 	splx(s);
1102 	return (error);
1103 }
1104 
1105 /*
1106  * Sysctl for udp variables.
1107  */
1108 int
1109 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1110 	int *name;
1111 	u_int namelen;
1112 	void *oldp;
1113 	size_t *oldlenp;
1114 	void *newp;
1115 	size_t newlen;
1116 {
1117 	/* All sysctl names at this level are terminal. */
1118 	if (namelen != 1)
1119 		return (ENOTDIR);
1120 
1121 	switch (name[0]) {
1122 	case UDPCTL_CHECKSUM:
1123 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1124 	case UDPCTL_SENDSPACE:
1125 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1126 		    &udp_sendspace));
1127 	case UDPCTL_RECVSPACE:
1128 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1129 		    &udp_recvspace));
1130 	default:
1131 		return (ENOPROTOOPT);
1132 	}
1133 	/* NOTREACHED */
1134 }
1135 #endif
1136