xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision db6316d1518382eecd2fdbe55a1205e0620a1b35)
1 /*	$NetBSD: udp_usrreq.c,v 1.125 2004/12/15 04:25:20 thorpej 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 1.125 2004/12/15 04:25:20 thorpej Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_ipsec.h"
68 #include "opt_inet_csum.h"
69 #include "opt_ipkdb.h"
70 #include "opt_mbuftrace.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/stat.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_icmp.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/udp6_var.h>
104 #endif
105 
106 #ifndef INET6
107 /* always need ip6.h for IP6_EXTHDR_GET */
108 #include <netinet/ip6.h>
109 #endif
110 
111 #include "faith.h"
112 #if defined(NFAITH) && NFAITH > 0
113 #include <net/if_faith.h>
114 #endif
115 
116 #include <machine/stdarg.h>
117 
118 #ifdef FAST_IPSEC
119 #include <netipsec/ipsec.h>
120 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
121 #ifdef INET6
122 #include <netipsec/ipsec6.h>
123 #endif
124 #endif	/* FAST_IPSEC*/
125 
126 #ifdef IPSEC
127 #include <netinet6/ipsec.h>
128 #include <netkey/key.h>
129 #endif /*IPSEC*/
130 
131 #ifdef IPKDB
132 #include <ipkdb/ipkdb.h>
133 #endif
134 
135 /*
136  * UDP protocol implementation.
137  * Per RFC 768, August, 1980.
138  */
139 #ifndef	COMPAT_42
140 int	udpcksum = 1;
141 #else
142 int	udpcksum = 0;		/* XXX */
143 #endif
144 int	udp_do_loopback_cksum = 0;
145 
146 struct	inpcbtable udbtable;
147 struct	udpstat udpstat;
148 
149 #ifdef INET
150 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
151 	struct socket *);
152 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
153 	struct mbuf *, int);
154 #endif
155 #ifdef INET6
156 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
157 	struct socket *);
158 static int udp6_realinput (int, struct sockaddr_in6 *,
159 	struct sockaddr_in6 *, struct mbuf *, int);
160 #endif
161 #ifdef INET
162 static	void udp_notify (struct inpcb *, int);
163 #endif
164 
165 #ifndef UDBHASHSIZE
166 #define	UDBHASHSIZE	128
167 #endif
168 int	udbhashsize = UDBHASHSIZE;
169 
170 #ifdef MBUFTRACE
171 struct mowner udp_mowner = { "udp" };
172 struct mowner udp_rx_mowner = { "udp", "rx" };
173 struct mowner udp_tx_mowner = { "udp", "tx" };
174 #endif
175 
176 #ifdef UDP_CSUM_COUNTERS
177 #include <sys/device.h>
178 
179 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
180     NULL, "udp", "hwcsum bad");
181 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182     NULL, "udp", "hwcsum ok");
183 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
184     NULL, "udp", "hwcsum data");
185 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
186     NULL, "udp", "swcsum");
187 
188 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
189 
190 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
191 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
192 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
193 EVCNT_ATTACH_STATIC(udp_swcsum);
194 
195 #else
196 
197 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
198 
199 #endif /* UDP_CSUM_COUNTERS */
200 
201 void
202 udp_init(void)
203 {
204 
205 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
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 udp_input(struct mbuf *m, ...)
215 {
216 	va_list ap;
217 	struct sockaddr_in src, dst;
218 	struct ip *ip;
219 	struct udphdr *uh;
220 	int iphlen;
221 	int len;
222 	int n;
223 	u_int16_t ip_len;
224 
225 	va_start(ap, m);
226 	iphlen = va_arg(ap, int);
227 	(void)va_arg(ap, int);		/* ignore value, advance ap */
228 	va_end(ap);
229 
230 	MCLAIM(m, &udp_rx_mowner);
231 	udpstat.udps_ipackets++;
232 
233 	/*
234 	 * Get IP and UDP header together in first mbuf.
235 	 */
236 	ip = mtod(m, struct ip *);
237 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
238 	if (uh == NULL) {
239 		udpstat.udps_hdrops++;
240 		return;
241 	}
242 	KASSERT(UDP_HDR_ALIGNED_P(uh));
243 
244 	/* destination port of 0 is illegal, based on RFC768. */
245 	if (uh->uh_dport == 0)
246 		goto bad;
247 
248 	/*
249 	 * Make mbuf data length reflect UDP length.
250 	 * If not enough data to reflect UDP length, drop.
251 	 */
252 	ip_len = ntohs(ip->ip_len);
253 	len = ntohs((u_int16_t)uh->uh_ulen);
254 	if (ip_len != iphlen + len) {
255 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
256 			udpstat.udps_badlen++;
257 			goto bad;
258 		}
259 		m_adj(m, iphlen + len - ip_len);
260 	}
261 
262 	/*
263 	 * Checksum extended UDP header and data.
264 	 */
265 	if (uh->uh_sum) {
266 		switch (m->m_pkthdr.csum_flags &
267 		    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
268 		    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
269 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
270 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
271 			goto badcsum;
272 
273 		case M_CSUM_UDPv4|M_CSUM_DATA: {
274 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
275 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
276 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR)
277 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
278 				    ip->ip_dst.s_addr,
279 				    htons(hw_csum + len + IPPROTO_UDP));
280 			if ((hw_csum ^ 0xffff) != 0)
281 				goto badcsum;
282 			break;
283 		}
284 
285 		case M_CSUM_UDPv4:
286 			/* Checksum was okay. */
287 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
288 			break;
289 
290 		default:
291 			/*
292 			 * Need to compute it ourselves.  Maybe skip checksum
293 			 * on loopback interfaces.
294 			 */
295 			if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
296 					     IFF_LOOPBACK) ||
297 					   udp_do_loopback_cksum)) {
298 				UDP_CSUM_COUNTER_INCR(&udp_swcsum);
299 				if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
300 					goto badcsum;
301 			}
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(struct mbuf **mp, int *offp, int proto)
377 {
378 	struct mbuf *m = *mp;
379 	int off = *offp;
380 	struct sockaddr_in6 src, dst;
381 	struct ip6_hdr *ip6;
382 	struct udphdr *uh;
383 	u_int32_t plen, ulen;
384 
385 	ip6 = mtod(m, struct ip6_hdr *);
386 
387 #if defined(NFAITH) && 0 < NFAITH
388 	if (faithprefix(&ip6->ip6_dst)) {
389 		/* send icmp6 host unreach? */
390 		m_freem(m);
391 		return IPPROTO_DONE;
392 	}
393 #endif
394 
395 	udp6stat.udp6s_ipackets++;
396 
397 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
398 	plen = m->m_pkthdr.len - off;
399 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
400 	if (uh == NULL) {
401 		ip6stat.ip6s_tooshort++;
402 		return IPPROTO_DONE;
403 	}
404 	KASSERT(UDP_HDR_ALIGNED_P(uh));
405 	ulen = ntohs((u_short)uh->uh_ulen);
406 	/*
407 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
408 	 * iff payload length > 0xffff.
409 	 */
410 	if (ulen == 0 && plen > 0xffff)
411 		ulen = plen;
412 
413 	if (plen != ulen) {
414 		udp6stat.udp6s_badlen++;
415 		goto bad;
416 	}
417 
418 	/* destination port of 0 is illegal, based on RFC768. */
419 	if (uh->uh_dport == 0)
420 		goto bad;
421 
422 	/* Be proactive about malicious use of IPv4 mapped address */
423 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
424 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
425 		/* XXX stat */
426 		goto bad;
427 	}
428 
429 	/*
430 	 * Checksum extended UDP header and data.  Maybe skip checksum
431 	 * on loopback interfaces.
432 	 */
433 	if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
434 			     IFF_LOOPBACK) ||
435 			   udp_do_loopback_cksum)) {
436 		if (uh->uh_sum == 0) {
437 			udp6stat.udp6s_nosum++;
438 			goto bad;
439 		}
440 		if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
441 			udp6stat.udp6s_badsum++;
442 			goto bad;
443 		}
444 	}
445 
446 	/*
447 	 * Construct source and dst sockaddrs.
448 	 * Note that ifindex (s6_addr16[1]) is already filled.
449 	 */
450 	bzero(&src, sizeof(src));
451 	src.sin6_family = AF_INET6;
452 	src.sin6_len = sizeof(struct sockaddr_in6);
453 	/* KAME hack: recover scopeid */
454 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
455 	src.sin6_port = uh->uh_sport;
456 	bzero(&dst, sizeof(dst));
457 	dst.sin6_family = AF_INET6;
458 	dst.sin6_len = sizeof(struct sockaddr_in6);
459 	/* KAME hack: recover scopeid */
460 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
461 	dst.sin6_port = uh->uh_dport;
462 
463 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
464 		if (m->m_flags & M_MCAST) {
465 			udp6stat.udp6s_noportmcast++;
466 			goto bad;
467 		}
468 		udp6stat.udp6s_noport++;
469 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
470 		m = NULL;
471 	}
472 
473 bad:
474 	if (m)
475 		m_freem(m);
476 	return IPPROTO_DONE;
477 }
478 #endif
479 
480 #ifdef INET
481 static void
482 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
483 	struct sockaddr *src, struct socket *so)
484 {
485 	struct mbuf *opts = NULL;
486 	struct mbuf *n;
487 	struct inpcb *inp = NULL;
488 
489 	if (!so)
490 		return;
491 	switch (so->so_proto->pr_domain->dom_family) {
492 	case AF_INET:
493 		inp = sotoinpcb(so);
494 		break;
495 #ifdef INET6
496 	case AF_INET6:
497 		break;
498 #endif
499 	default:
500 		return;
501 	}
502 
503 #if defined(IPSEC) || defined(FAST_IPSEC)
504 	/* check AH/ESP integrity. */
505 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
506 		ipsecstat.in_polvio++;
507 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
508 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
509 			    0, 0);
510 		return;
511 	}
512 #endif /*IPSEC*/
513 
514 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
515 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
516 			 || so->so_options & SO_TIMESTAMP)) {
517 			struct ip *ip = mtod(n, struct ip *);
518 			ip_savecontrol(inp, &opts, ip, n);
519 		}
520 
521 		m_adj(n, off);
522 		if (sbappendaddr(&so->so_rcv, src, n,
523 				opts) == 0) {
524 			m_freem(n);
525 			if (opts)
526 				m_freem(opts);
527 			so->so_rcv.sb_overflowed++;
528 			udpstat.udps_fullsock++;
529 		} else
530 			sorwakeup(so);
531 	}
532 }
533 #endif
534 
535 #ifdef INET6
536 static void
537 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
538 	struct sockaddr *src, struct socket *so)
539 {
540 	struct mbuf *opts = NULL;
541 	struct mbuf *n;
542 	struct in6pcb *in6p = NULL;
543 
544 	if (!so)
545 		return;
546 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
547 		return;
548 	in6p = sotoin6pcb(so);
549 
550 #if defined(IPSEC) || defined(FAST_IPSEC)
551 	/* check AH/ESP integrity. */
552 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
553 		ipsec6stat.in_polvio++;
554 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
555 			icmp6_error(n, ICMP6_DST_UNREACH,
556 			    ICMP6_DST_UNREACH_ADMIN, 0);
557 		return;
558 	}
559 #endif /*IPSEC*/
560 
561 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
562 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
563 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
564 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
565 			ip6_savecontrol(in6p, &opts, ip6, n);
566 		}
567 
568 		m_adj(n, off);
569 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
570 			m_freem(n);
571 			if (opts)
572 				m_freem(opts);
573 			so->so_rcv.sb_overflowed++;
574 			udp6stat.udp6s_fullsock++;
575 		} else
576 			sorwakeup(so);
577 	}
578 }
579 #endif
580 
581 #ifdef INET
582 static int
583 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
584 	struct mbuf *m, int off /* offset of udphdr */)
585 {
586 	u_int16_t *sport, *dport;
587 	int rcvcnt;
588 	struct in_addr *src4, *dst4;
589 	struct inpcb_hdr *inph;
590 	struct inpcb *inp;
591 
592 	rcvcnt = 0;
593 	off += sizeof(struct udphdr);	/* now, offset of payload */
594 
595 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
596 		goto bad;
597 
598 	src4 = &src->sin_addr;
599 	sport = &src->sin_port;
600 	dst4 = &dst->sin_addr;
601 	dport = &dst->sin_port;
602 
603 	if (IN_MULTICAST(dst4->s_addr) ||
604 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
605 		/*
606 		 * Deliver a multicast or broadcast datagram to *all* sockets
607 		 * for which the local and remote addresses and ports match
608 		 * those of the incoming datagram.  This allows more than
609 		 * one process to receive multi/broadcasts on the same port.
610 		 * (This really ought to be done for unicast datagrams as
611 		 * well, but that would cause problems with existing
612 		 * applications that open both address-specific sockets and
613 		 * a wildcard socket listening to the same port -- they would
614 		 * end up receiving duplicates of every unicast datagram.
615 		 * Those applications open the multiple sockets to overcome an
616 		 * inadequacy of the UDP socket interface, but for backwards
617 		 * compatibility we avoid the problem here rather than
618 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
619 		 */
620 
621 		/*
622 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
623 		 * we need udpiphdr for IPsec processing so we do that later.
624 		 */
625 		/*
626 		 * Locate pcb(s) for datagram.
627 		 */
628 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
629 			inp = (struct inpcb *)inph;
630 			if (inp->inp_af != AF_INET)
631 				continue;
632 
633 			if (inp->inp_lport != *dport)
634 				continue;
635 			if (!in_nullhost(inp->inp_laddr)) {
636 				if (!in_hosteq(inp->inp_laddr, *dst4))
637 					continue;
638 			}
639 			if (!in_nullhost(inp->inp_faddr)) {
640 				if (!in_hosteq(inp->inp_faddr, *src4) ||
641 				    inp->inp_fport != *sport)
642 					continue;
643 			}
644 
645 			udp4_sendup(m, off, (struct sockaddr *)src,
646 				inp->inp_socket);
647 			rcvcnt++;
648 
649 			/*
650 			 * Don't look for additional matches if this one does
651 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
652 			 * socket options set.  This heuristic avoids searching
653 			 * through all pcbs in the common case of a non-shared
654 			 * port.  It assumes that an application will never
655 			 * clear these options after setting them.
656 			 */
657 			if ((inp->inp_socket->so_options &
658 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
659 				break;
660 		}
661 	} else {
662 		/*
663 		 * Locate pcb for datagram.
664 		 */
665 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
666 		if (inp == 0) {
667 			++udpstat.udps_pcbhashmiss;
668 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
669 			if (inp == 0)
670 				return rcvcnt;
671 		}
672 
673 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
674 		rcvcnt++;
675 	}
676 
677 bad:
678 	return rcvcnt;
679 }
680 #endif
681 
682 #ifdef INET6
683 static int
684 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
685 	struct mbuf *m, int off)
686 {
687 	u_int16_t sport, dport;
688 	int rcvcnt;
689 	struct in6_addr src6, dst6;
690 	const struct in_addr *dst4;
691 	struct inpcb_hdr *inph;
692 	struct in6pcb *in6p;
693 
694 	rcvcnt = 0;
695 	off += sizeof(struct udphdr);	/* now, offset of payload */
696 
697 	if (af != AF_INET && af != AF_INET6)
698 		goto bad;
699 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
700 		goto bad;
701 
702 	in6_embedscope(&src6, src, NULL, NULL);
703 	sport = src->sin6_port;
704 	in6_embedscope(&dst6, dst, NULL, NULL);
705 	dport = dst->sin6_port;
706 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
707 
708 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
709 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
710 		/*
711 		 * Deliver a multicast or broadcast datagram to *all* sockets
712 		 * for which the local and remote addresses and ports match
713 		 * those of the incoming datagram.  This allows more than
714 		 * one process to receive multi/broadcasts on the same port.
715 		 * (This really ought to be done for unicast datagrams as
716 		 * well, but that would cause problems with existing
717 		 * applications that open both address-specific sockets and
718 		 * a wildcard socket listening to the same port -- they would
719 		 * end up receiving duplicates of every unicast datagram.
720 		 * Those applications open the multiple sockets to overcome an
721 		 * inadequacy of the UDP socket interface, but for backwards
722 		 * compatibility we avoid the problem here rather than
723 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
724 		 */
725 
726 		/*
727 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
728 		 * we need udpiphdr for IPsec processing so we do that later.
729 		 */
730 		/*
731 		 * Locate pcb(s) for datagram.
732 		 */
733 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
734 			in6p = (struct in6pcb *)inph;
735 			if (in6p->in6p_af != AF_INET6)
736 				continue;
737 
738 			if (in6p->in6p_lport != dport)
739 				continue;
740 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
741 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
742 					continue;
743 			} else {
744 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
745 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
746 					continue;
747 			}
748 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
749 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
750 				    &src6) || in6p->in6p_fport != sport)
751 					continue;
752 			} else {
753 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
754 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
755 					continue;
756 			}
757 
758 			udp6_sendup(m, off, (struct sockaddr *)src,
759 				in6p->in6p_socket);
760 			rcvcnt++;
761 
762 			/*
763 			 * Don't look for additional matches if this one does
764 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
765 			 * socket options set.  This heuristic avoids searching
766 			 * through all pcbs in the common case of a non-shared
767 			 * port.  It assumes that an application will never
768 			 * clear these options after setting them.
769 			 */
770 			if ((in6p->in6p_socket->so_options &
771 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
772 				break;
773 		}
774 	} else {
775 		/*
776 		 * Locate pcb for datagram.
777 		 */
778 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
779 		    &dst6, dport, 0);
780 		if (in6p == 0) {
781 			++udpstat.udps_pcbhashmiss;
782 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
783 			if (in6p == 0)
784 				return rcvcnt;
785 		}
786 
787 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
788 		rcvcnt++;
789 	}
790 
791 bad:
792 	return rcvcnt;
793 }
794 #endif
795 
796 #ifdef INET
797 /*
798  * Notify a udp user of an asynchronous error;
799  * just wake up so that he can collect error status.
800  */
801 static void
802 udp_notify(struct inpcb *inp, int errno)
803 {
804 	inp->inp_socket->so_error = errno;
805 	sorwakeup(inp->inp_socket);
806 	sowwakeup(inp->inp_socket);
807 }
808 
809 void *
810 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
811 {
812 	struct ip *ip = v;
813 	struct udphdr *uh;
814 	void (*notify)(struct inpcb *, int) = udp_notify;
815 	int errno;
816 
817 	if (sa->sa_family != AF_INET
818 	 || sa->sa_len != sizeof(struct sockaddr_in))
819 		return NULL;
820 	if ((unsigned)cmd >= PRC_NCMDS)
821 		return NULL;
822 	errno = inetctlerrmap[cmd];
823 	if (PRC_IS_REDIRECT(cmd))
824 		notify = in_rtchange, ip = 0;
825 	else if (cmd == PRC_HOSTDEAD)
826 		ip = 0;
827 	else if (errno == 0)
828 		return NULL;
829 	if (ip) {
830 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
831 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
832 		    ip->ip_src, uh->uh_sport, errno, notify);
833 
834 		/* XXX mapped address case */
835 	} else
836 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
837 		    notify);
838 	return NULL;
839 }
840 
841 int
842 udp_output(struct mbuf *m, ...)
843 {
844 	struct inpcb *inp;
845 	struct udpiphdr *ui;
846 	struct route *ro;
847 	int len = m->m_pkthdr.len;
848 	int error = 0;
849 	va_list ap;
850 
851 	MCLAIM(m, &udp_tx_mowner);
852 	va_start(ap, m);
853 	inp = va_arg(ap, struct inpcb *);
854 	va_end(ap);
855 
856 	/*
857 	 * Calculate data length and get a mbuf
858 	 * for UDP and IP headers.
859 	 */
860 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
861 	if (m == 0) {
862 		error = ENOBUFS;
863 		goto release;
864 	}
865 
866 	/*
867 	 * Compute the packet length of the IP header, and
868 	 * punt if the length looks bogus.
869 	 */
870 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
871 		error = EMSGSIZE;
872 		goto release;
873 	}
874 
875 	/*
876 	 * Fill in mbuf with extended UDP header
877 	 * and addresses and length put into network format.
878 	 */
879 	ui = mtod(m, struct udpiphdr *);
880 	ui->ui_pr = IPPROTO_UDP;
881 	ui->ui_src = inp->inp_laddr;
882 	ui->ui_dst = inp->inp_faddr;
883 	ui->ui_sport = inp->inp_lport;
884 	ui->ui_dport = inp->inp_fport;
885 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
886 
887 	ro = &inp->inp_route;
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 		 * Maybe skip checksums on loopback interfaces.
897 		 */
898 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
899 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
900 		    sizeof(struct udphdr) + IPPROTO_UDP));
901 		if (__predict_true(ro->ro_rt == NULL ||
902 				   !(ro->ro_rt->rt_ifp->if_flags &
903 				     IFF_LOOPBACK) ||
904 				   udp_do_loopback_cksum))
905 			m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
906 		else
907 			m->m_pkthdr.csum_flags = 0;
908 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
909 	} else
910 		ui->ui_sum = 0;
911 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
912 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
913 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
914 	udpstat.udps_opackets++;
915 
916 	return (ip_output(m, inp->inp_options, ro,
917 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
918 	    inp->inp_moptions, inp->inp_socket));
919 
920 release:
921 	m_freem(m);
922 	return (error);
923 }
924 
925 int	udp_sendspace = 9216;		/* really max datagram size */
926 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
927 					/* 40 1K datagrams */
928 
929 /*ARGSUSED*/
930 int
931 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
932 	struct mbuf *control, struct proc *p)
933 {
934 	struct inpcb *inp;
935 	int s;
936 	int error = 0;
937 
938 	if (req == PRU_CONTROL)
939 		return (in_control(so, (long)m, (caddr_t)nam,
940 		    (struct ifnet *)control, p));
941 
942 	if (req == PRU_PURGEIF) {
943 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
944 		in_purgeif((struct ifnet *)control);
945 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
946 		return (0);
947 	}
948 
949 	s = splsoftnet();
950 	inp = sotoinpcb(so);
951 #ifdef DIAGNOSTIC
952 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
953 		panic("udp_usrreq: unexpected control mbuf");
954 #endif
955 	if (inp == 0 && req != PRU_ATTACH) {
956 		error = EINVAL;
957 		goto release;
958 	}
959 
960 	/*
961 	 * Note: need to block udp_input while changing
962 	 * the udp pcb queue and/or pcb addresses.
963 	 */
964 	switch (req) {
965 
966 	case PRU_ATTACH:
967 		if (inp != 0) {
968 			error = EISCONN;
969 			break;
970 		}
971 #ifdef MBUFTRACE
972 		so->so_mowner = &udp_mowner;
973 		so->so_rcv.sb_mowner = &udp_rx_mowner;
974 		so->so_snd.sb_mowner = &udp_tx_mowner;
975 #endif
976 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
977 			error = soreserve(so, udp_sendspace, udp_recvspace);
978 			if (error)
979 				break;
980 		}
981 		error = in_pcballoc(so, &udbtable);
982 		if (error)
983 			break;
984 		inp = sotoinpcb(so);
985 		inp->inp_ip.ip_ttl = ip_defttl;
986 		break;
987 
988 	case PRU_DETACH:
989 		in_pcbdetach(inp);
990 		break;
991 
992 	case PRU_BIND:
993 		error = in_pcbbind(inp, nam, p);
994 		break;
995 
996 	case PRU_LISTEN:
997 		error = EOPNOTSUPP;
998 		break;
999 
1000 	case PRU_CONNECT:
1001 		error = in_pcbconnect(inp, nam);
1002 		if (error)
1003 			break;
1004 		soisconnected(so);
1005 		break;
1006 
1007 	case PRU_CONNECT2:
1008 		error = EOPNOTSUPP;
1009 		break;
1010 
1011 	case PRU_DISCONNECT:
1012 		/*soisdisconnected(so);*/
1013 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1014 		in_pcbdisconnect(inp);
1015 		inp->inp_laddr = zeroin_addr;		/* XXX */
1016 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1017 		break;
1018 
1019 	case PRU_SHUTDOWN:
1020 		socantsendmore(so);
1021 		break;
1022 
1023 	case PRU_RCVD:
1024 		error = EOPNOTSUPP;
1025 		break;
1026 
1027 	case PRU_SEND:
1028 		if (control && control->m_len) {
1029 			m_freem(control);
1030 			m_freem(m);
1031 			error = EINVAL;
1032 			break;
1033 		}
1034 	{
1035 		struct in_addr laddr;			/* XXX */
1036 
1037 		if (nam) {
1038 			laddr = inp->inp_laddr;		/* XXX */
1039 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1040 				error = EISCONN;
1041 				goto die;
1042 			}
1043 			error = in_pcbconnect(inp, nam);
1044 			if (error)
1045 				goto die;
1046 		} else {
1047 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1048 				error = ENOTCONN;
1049 				goto die;
1050 			}
1051 		}
1052 		error = udp_output(m, inp);
1053 		m = NULL;
1054 		if (nam) {
1055 			in_pcbdisconnect(inp);
1056 			inp->inp_laddr = laddr;		/* XXX */
1057 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1058 		}
1059 	  die:
1060 		if (m)
1061 			m_freem(m);
1062 	}
1063 		break;
1064 
1065 	case PRU_SENSE:
1066 		/*
1067 		 * stat: don't bother with a blocksize.
1068 		 */
1069 		splx(s);
1070 		return (0);
1071 
1072 	case PRU_RCVOOB:
1073 		error =  EOPNOTSUPP;
1074 		break;
1075 
1076 	case PRU_SENDOOB:
1077 		m_freem(control);
1078 		m_freem(m);
1079 		error =  EOPNOTSUPP;
1080 		break;
1081 
1082 	case PRU_SOCKADDR:
1083 		in_setsockaddr(inp, nam);
1084 		break;
1085 
1086 	case PRU_PEERADDR:
1087 		in_setpeeraddr(inp, nam);
1088 		break;
1089 
1090 	default:
1091 		panic("udp_usrreq");
1092 	}
1093 
1094 release:
1095 	splx(s);
1096 	return (error);
1097 }
1098 
1099 /*
1100  * Sysctl for udp variables.
1101  */
1102 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1103 {
1104 
1105 	sysctl_createv(clog, 0, NULL, NULL,
1106 		       CTLFLAG_PERMANENT,
1107 		       CTLTYPE_NODE, "net", NULL,
1108 		       NULL, 0, NULL, 0,
1109 		       CTL_NET, CTL_EOL);
1110 	sysctl_createv(clog, 0, NULL, NULL,
1111 		       CTLFLAG_PERMANENT,
1112 		       CTLTYPE_NODE, "inet", NULL,
1113 		       NULL, 0, NULL, 0,
1114 		       CTL_NET, PF_INET, CTL_EOL);
1115 	sysctl_createv(clog, 0, NULL, NULL,
1116 		       CTLFLAG_PERMANENT,
1117 		       CTLTYPE_NODE, "udp",
1118 		       SYSCTL_DESCR("UDPv4 related settings"),
1119 		       NULL, 0, NULL, 0,
1120 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1121 
1122 	sysctl_createv(clog, 0, NULL, NULL,
1123 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1124 		       CTLTYPE_INT, "checksum",
1125 		       SYSCTL_DESCR("Compute UDP checksums"),
1126 		       NULL, 0, &udpcksum, 0,
1127 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1128 		       CTL_EOL);
1129 	sysctl_createv(clog, 0, NULL, NULL,
1130 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1131 		       CTLTYPE_INT, "sendspace",
1132 		       SYSCTL_DESCR("Default UDP send buffer size"),
1133 		       NULL, 0, &udp_sendspace, 0,
1134 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1135 		       CTL_EOL);
1136 	sysctl_createv(clog, 0, NULL, NULL,
1137 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1138 		       CTLTYPE_INT, "recvspace",
1139 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1140 		       NULL, 0, &udp_recvspace, 0,
1141 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1142 		       CTL_EOL);
1143 	sysctl_createv(clog, 0, NULL, NULL,
1144 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1145 		       CTLTYPE_INT, "do_loopback_cksum",
1146 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1147 		       NULL, 0, &udp_do_loopback_cksum, 0,
1148 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1149 		       CTL_EOL);
1150 }
1151 #endif
1152