xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision fd5cb0acea84d278e04e640d37ca2398f894991f)
1 /*	$NetBSD: udp_usrreq.c,v 1.129 2004/12/21 05:51:32 yamt 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 1.129 2004/12/21 05:51:32 yamt 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 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
155 #endif
156 #ifdef INET6
157 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
158 	struct socket *);
159 static int udp6_realinput (int, struct sockaddr_in6 *,
160 	struct sockaddr_in6 *, struct mbuf *, int);
161 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
162 #endif
163 #ifdef INET
164 static	void udp_notify (struct inpcb *, int);
165 #endif
166 
167 #ifndef UDBHASHSIZE
168 #define	UDBHASHSIZE	128
169 #endif
170 int	udbhashsize = UDBHASHSIZE;
171 
172 #ifdef MBUFTRACE
173 struct mowner udp_mowner = { "udp" };
174 struct mowner udp_rx_mowner = { "udp", "rx" };
175 struct mowner udp_tx_mowner = { "udp", "tx" };
176 #endif
177 
178 #ifdef UDP_CSUM_COUNTERS
179 #include <sys/device.h>
180 
181 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182     NULL, "udp", "hwcsum bad");
183 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
184     NULL, "udp", "hwcsum ok");
185 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
186     NULL, "udp", "hwcsum data");
187 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
188     NULL, "udp", "swcsum");
189 
190 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
191 
192 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
193 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
194 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
195 EVCNT_ATTACH_STATIC(udp_swcsum);
196 
197 #else
198 
199 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
200 
201 #endif /* UDP_CSUM_COUNTERS */
202 
203 void
204 udp_init(void)
205 {
206 
207 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
208 
209 	MOWNER_ATTACH(&udp_tx_mowner);
210 	MOWNER_ATTACH(&udp_rx_mowner);
211 	MOWNER_ATTACH(&udp_mowner);
212 }
213 
214 /*
215  * Checksum extended UDP header and data.
216  */
217 
218 int
219 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
220     int iphlen, int len)
221 {
222 
223 	switch (af) {
224 #ifdef INET
225 	case AF_INET:
226 		return udp4_input_checksum(m, uh, iphlen, len);
227 #endif
228 #ifdef INET6
229 	case AF_INET6:
230 		return udp6_input_checksum(m, uh, iphlen, len);
231 #endif
232 	}
233 #ifdef DIAGNOSTIC
234 	panic("udp_input_checksum: unknown af %d", af);
235 #endif
236 	/* NOTREACHED */
237 	return -1;
238 }
239 
240 #ifdef INET
241 
242 /*
243  * Checksum extended UDP header and data.
244  */
245 
246 static int
247 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
248     int iphlen, int len)
249 {
250 
251 	/*
252 	 * XXX it's better to record and check if this mbuf is
253 	 * already checked.
254 	 */
255 
256 	if (uh->uh_sum == 0)
257 		return 0;
258 
259 	switch (m->m_pkthdr.csum_flags &
260 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
261 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
262 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
263 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
264 		goto badcsum;
265 
266 	case M_CSUM_UDPv4|M_CSUM_DATA: {
267 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
268 
269 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
270 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
271 			const struct ip *ip =
272 			    mtod(m, const struct ip *);
273 
274 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
275 			    ip->ip_dst.s_addr,
276 			    htons(hw_csum + len + IPPROTO_UDP));
277 		}
278 		if ((hw_csum ^ 0xffff) != 0)
279 			goto badcsum;
280 		break;
281 	}
282 
283 	case M_CSUM_UDPv4:
284 		/* Checksum was okay. */
285 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
286 		break;
287 
288 	default:
289 		/*
290 		 * Need to compute it ourselves.  Maybe skip checksum
291 		 * on loopback interfaces.
292 		 */
293 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
294 				     IFF_LOOPBACK) ||
295 				   udp_do_loopback_cksum)) {
296 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
297 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
298 				goto badcsum;
299 		}
300 		break;
301 	}
302 
303 	return 0;
304 
305 badcsum:
306 	udpstat.udps_badsum++;
307 	return -1;
308 }
309 
310 void
311 udp_input(struct mbuf *m, ...)
312 {
313 	va_list ap;
314 	struct sockaddr_in src, dst;
315 	struct ip *ip;
316 	struct udphdr *uh;
317 	int iphlen;
318 	int len;
319 	int n;
320 	u_int16_t ip_len;
321 
322 	va_start(ap, m);
323 	iphlen = va_arg(ap, int);
324 	(void)va_arg(ap, int);		/* ignore value, advance ap */
325 	va_end(ap);
326 
327 	MCLAIM(m, &udp_rx_mowner);
328 	udpstat.udps_ipackets++;
329 
330 	/*
331 	 * Get IP and UDP header together in first mbuf.
332 	 */
333 	ip = mtod(m, struct ip *);
334 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
335 	if (uh == NULL) {
336 		udpstat.udps_hdrops++;
337 		return;
338 	}
339 	KASSERT(UDP_HDR_ALIGNED_P(uh));
340 
341 	/* destination port of 0 is illegal, based on RFC768. */
342 	if (uh->uh_dport == 0)
343 		goto bad;
344 
345 	/*
346 	 * Make mbuf data length reflect UDP length.
347 	 * If not enough data to reflect UDP length, drop.
348 	 */
349 	ip_len = ntohs(ip->ip_len);
350 	len = ntohs((u_int16_t)uh->uh_ulen);
351 	if (ip_len != iphlen + len) {
352 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
353 			udpstat.udps_badlen++;
354 			goto bad;
355 		}
356 		m_adj(m, iphlen + len - ip_len);
357 	}
358 
359 	/*
360 	 * Checksum extended UDP header and data.
361 	 */
362 	if (udp4_input_checksum(m, uh, iphlen, len))
363 		goto badcsum;
364 
365 	/* construct source and dst sockaddrs. */
366 	bzero(&src, sizeof(src));
367 	src.sin_family = AF_INET;
368 	src.sin_len = sizeof(struct sockaddr_in);
369 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
370 	src.sin_port = uh->uh_sport;
371 	bzero(&dst, sizeof(dst));
372 	dst.sin_family = AF_INET;
373 	dst.sin_len = sizeof(struct sockaddr_in);
374 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
375 	dst.sin_port = uh->uh_dport;
376 
377 	n = udp4_realinput(&src, &dst, m, iphlen);
378 #ifdef INET6
379 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
380 		struct sockaddr_in6 src6, dst6;
381 
382 		bzero(&src6, sizeof(src6));
383 		src6.sin6_family = AF_INET6;
384 		src6.sin6_len = sizeof(struct sockaddr_in6);
385 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
386 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
387 			sizeof(ip->ip_src));
388 		src6.sin6_port = uh->uh_sport;
389 		bzero(&dst6, sizeof(dst6));
390 		dst6.sin6_family = AF_INET6;
391 		dst6.sin6_len = sizeof(struct sockaddr_in6);
392 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
393 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
394 			sizeof(ip->ip_dst));
395 		dst6.sin6_port = uh->uh_dport;
396 
397 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
398 	}
399 #endif
400 
401 	if (n == 0) {
402 		if (m->m_flags & (M_BCAST | M_MCAST)) {
403 			udpstat.udps_noportbcast++;
404 			goto bad;
405 		}
406 		udpstat.udps_noport++;
407 #ifdef IPKDB
408 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
409 				m, iphlen + sizeof(struct udphdr),
410 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
411 			/*
412 			 * It was a debugger connect packet,
413 			 * just drop it now
414 			 */
415 			goto bad;
416 		}
417 #endif
418 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
419 		m = NULL;
420 	}
421 
422 bad:
423 	if (m)
424 		m_freem(m);
425 	return;
426 
427 badcsum:
428 	m_freem(m);
429 }
430 #endif
431 
432 #ifdef INET6
433 static int
434 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
435 {
436 
437 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
438 		goto good;
439 	}
440 	if (uh->uh_sum == 0) {
441 		udp6stat.udp6s_nosum++;
442 		goto bad;
443 	}
444 	if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
445 		udp6stat.udp6s_badsum++;
446 		goto bad;
447 	}
448 
449 good:
450 	return 0;
451 bad:
452 	return -1;
453 }
454 
455 int
456 udp6_input(struct mbuf **mp, int *offp, int proto)
457 {
458 	struct mbuf *m = *mp;
459 	int off = *offp;
460 	struct sockaddr_in6 src, dst;
461 	struct ip6_hdr *ip6;
462 	struct udphdr *uh;
463 	u_int32_t plen, ulen;
464 
465 	ip6 = mtod(m, struct ip6_hdr *);
466 
467 #if defined(NFAITH) && 0 < NFAITH
468 	if (faithprefix(&ip6->ip6_dst)) {
469 		/* send icmp6 host unreach? */
470 		m_freem(m);
471 		return IPPROTO_DONE;
472 	}
473 #endif
474 
475 	udp6stat.udp6s_ipackets++;
476 
477 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
478 	plen = m->m_pkthdr.len - off;
479 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
480 	if (uh == NULL) {
481 		ip6stat.ip6s_tooshort++;
482 		return IPPROTO_DONE;
483 	}
484 	KASSERT(UDP_HDR_ALIGNED_P(uh));
485 	ulen = ntohs((u_short)uh->uh_ulen);
486 	/*
487 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
488 	 * iff payload length > 0xffff.
489 	 */
490 	if (ulen == 0 && plen > 0xffff)
491 		ulen = plen;
492 
493 	if (plen != ulen) {
494 		udp6stat.udp6s_badlen++;
495 		goto bad;
496 	}
497 
498 	/* destination port of 0 is illegal, based on RFC768. */
499 	if (uh->uh_dport == 0)
500 		goto bad;
501 
502 	/* Be proactive about malicious use of IPv4 mapped address */
503 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
504 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
505 		/* XXX stat */
506 		goto bad;
507 	}
508 
509 	/*
510 	 * Checksum extended UDP header and data.  Maybe skip checksum
511 	 * on loopback interfaces.
512 	 */
513 	if (udp6_input_checksum(m, uh, off, ulen))
514 		goto bad;
515 
516 	/*
517 	 * Construct source and dst sockaddrs.
518 	 * Note that ifindex (s6_addr16[1]) is already filled.
519 	 */
520 	bzero(&src, sizeof(src));
521 	src.sin6_family = AF_INET6;
522 	src.sin6_len = sizeof(struct sockaddr_in6);
523 	/* KAME hack: recover scopeid */
524 	(void)in6_recoverscope(&src, &ip6->ip6_src, m->m_pkthdr.rcvif);
525 	src.sin6_port = uh->uh_sport;
526 	bzero(&dst, sizeof(dst));
527 	dst.sin6_family = AF_INET6;
528 	dst.sin6_len = sizeof(struct sockaddr_in6);
529 	/* KAME hack: recover scopeid */
530 	(void)in6_recoverscope(&dst, &ip6->ip6_dst, m->m_pkthdr.rcvif);
531 	dst.sin6_port = uh->uh_dport;
532 
533 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
534 		if (m->m_flags & M_MCAST) {
535 			udp6stat.udp6s_noportmcast++;
536 			goto bad;
537 		}
538 		udp6stat.udp6s_noport++;
539 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
540 		m = NULL;
541 	}
542 
543 bad:
544 	if (m)
545 		m_freem(m);
546 	return IPPROTO_DONE;
547 }
548 #endif
549 
550 #ifdef INET
551 static void
552 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
553 	struct sockaddr *src, struct socket *so)
554 {
555 	struct mbuf *opts = NULL;
556 	struct mbuf *n;
557 	struct inpcb *inp = NULL;
558 
559 	if (!so)
560 		return;
561 	switch (so->so_proto->pr_domain->dom_family) {
562 	case AF_INET:
563 		inp = sotoinpcb(so);
564 		break;
565 #ifdef INET6
566 	case AF_INET6:
567 		break;
568 #endif
569 	default:
570 		return;
571 	}
572 
573 #if defined(IPSEC) || defined(FAST_IPSEC)
574 	/* check AH/ESP integrity. */
575 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
576 		ipsecstat.in_polvio++;
577 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
578 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
579 			    0, 0);
580 		return;
581 	}
582 #endif /*IPSEC*/
583 
584 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
585 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
586 			 || so->so_options & SO_TIMESTAMP)) {
587 			struct ip *ip = mtod(n, struct ip *);
588 			ip_savecontrol(inp, &opts, ip, n);
589 		}
590 
591 		m_adj(n, off);
592 		if (sbappendaddr(&so->so_rcv, src, n,
593 				opts) == 0) {
594 			m_freem(n);
595 			if (opts)
596 				m_freem(opts);
597 			so->so_rcv.sb_overflowed++;
598 			udpstat.udps_fullsock++;
599 		} else
600 			sorwakeup(so);
601 	}
602 }
603 #endif
604 
605 #ifdef INET6
606 static void
607 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
608 	struct sockaddr *src, struct socket *so)
609 {
610 	struct mbuf *opts = NULL;
611 	struct mbuf *n;
612 	struct in6pcb *in6p = NULL;
613 
614 	if (!so)
615 		return;
616 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
617 		return;
618 	in6p = sotoin6pcb(so);
619 
620 #if defined(IPSEC) || defined(FAST_IPSEC)
621 	/* check AH/ESP integrity. */
622 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
623 		ipsec6stat.in_polvio++;
624 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
625 			icmp6_error(n, ICMP6_DST_UNREACH,
626 			    ICMP6_DST_UNREACH_ADMIN, 0);
627 		return;
628 	}
629 #endif /*IPSEC*/
630 
631 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
632 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
633 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
634 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
635 			ip6_savecontrol(in6p, &opts, ip6, n);
636 		}
637 
638 		m_adj(n, off);
639 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
640 			m_freem(n);
641 			if (opts)
642 				m_freem(opts);
643 			so->so_rcv.sb_overflowed++;
644 			udp6stat.udp6s_fullsock++;
645 		} else
646 			sorwakeup(so);
647 	}
648 }
649 #endif
650 
651 #ifdef INET
652 static int
653 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
654 	struct mbuf *m, int off /* offset of udphdr */)
655 {
656 	u_int16_t *sport, *dport;
657 	int rcvcnt;
658 	struct in_addr *src4, *dst4;
659 	struct inpcb_hdr *inph;
660 	struct inpcb *inp;
661 
662 	rcvcnt = 0;
663 	off += sizeof(struct udphdr);	/* now, offset of payload */
664 
665 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
666 		goto bad;
667 
668 	src4 = &src->sin_addr;
669 	sport = &src->sin_port;
670 	dst4 = &dst->sin_addr;
671 	dport = &dst->sin_port;
672 
673 	if (IN_MULTICAST(dst4->s_addr) ||
674 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
675 		/*
676 		 * Deliver a multicast or broadcast datagram to *all* sockets
677 		 * for which the local and remote addresses and ports match
678 		 * those of the incoming datagram.  This allows more than
679 		 * one process to receive multi/broadcasts on the same port.
680 		 * (This really ought to be done for unicast datagrams as
681 		 * well, but that would cause problems with existing
682 		 * applications that open both address-specific sockets and
683 		 * a wildcard socket listening to the same port -- they would
684 		 * end up receiving duplicates of every unicast datagram.
685 		 * Those applications open the multiple sockets to overcome an
686 		 * inadequacy of the UDP socket interface, but for backwards
687 		 * compatibility we avoid the problem here rather than
688 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
689 		 */
690 
691 		/*
692 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
693 		 * we need udpiphdr for IPsec processing so we do that later.
694 		 */
695 		/*
696 		 * Locate pcb(s) for datagram.
697 		 */
698 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
699 			inp = (struct inpcb *)inph;
700 			if (inp->inp_af != AF_INET)
701 				continue;
702 
703 			if (inp->inp_lport != *dport)
704 				continue;
705 			if (!in_nullhost(inp->inp_laddr)) {
706 				if (!in_hosteq(inp->inp_laddr, *dst4))
707 					continue;
708 			}
709 			if (!in_nullhost(inp->inp_faddr)) {
710 				if (!in_hosteq(inp->inp_faddr, *src4) ||
711 				    inp->inp_fport != *sport)
712 					continue;
713 			}
714 
715 			udp4_sendup(m, off, (struct sockaddr *)src,
716 				inp->inp_socket);
717 			rcvcnt++;
718 
719 			/*
720 			 * Don't look for additional matches if this one does
721 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
722 			 * socket options set.  This heuristic avoids searching
723 			 * through all pcbs in the common case of a non-shared
724 			 * port.  It assumes that an application will never
725 			 * clear these options after setting them.
726 			 */
727 			if ((inp->inp_socket->so_options &
728 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
729 				break;
730 		}
731 	} else {
732 		/*
733 		 * Locate pcb for datagram.
734 		 */
735 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
736 		if (inp == 0) {
737 			++udpstat.udps_pcbhashmiss;
738 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
739 			if (inp == 0)
740 				return rcvcnt;
741 		}
742 
743 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
744 		rcvcnt++;
745 	}
746 
747 bad:
748 	return rcvcnt;
749 }
750 #endif
751 
752 #ifdef INET6
753 static int
754 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
755 	struct mbuf *m, int off)
756 {
757 	u_int16_t sport, dport;
758 	int rcvcnt;
759 	struct in6_addr src6, dst6;
760 	const struct in_addr *dst4;
761 	struct inpcb_hdr *inph;
762 	struct in6pcb *in6p;
763 
764 	rcvcnt = 0;
765 	off += sizeof(struct udphdr);	/* now, offset of payload */
766 
767 	if (af != AF_INET && af != AF_INET6)
768 		goto bad;
769 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
770 		goto bad;
771 
772 	in6_embedscope(&src6, src, NULL, NULL);
773 	sport = src->sin6_port;
774 	in6_embedscope(&dst6, dst, NULL, NULL);
775 	dport = dst->sin6_port;
776 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
777 
778 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
779 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
780 		/*
781 		 * Deliver a multicast or broadcast datagram to *all* sockets
782 		 * for which the local and remote addresses and ports match
783 		 * those of the incoming datagram.  This allows more than
784 		 * one process to receive multi/broadcasts on the same port.
785 		 * (This really ought to be done for unicast datagrams as
786 		 * well, but that would cause problems with existing
787 		 * applications that open both address-specific sockets and
788 		 * a wildcard socket listening to the same port -- they would
789 		 * end up receiving duplicates of every unicast datagram.
790 		 * Those applications open the multiple sockets to overcome an
791 		 * inadequacy of the UDP socket interface, but for backwards
792 		 * compatibility we avoid the problem here rather than
793 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
794 		 */
795 
796 		/*
797 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
798 		 * we need udpiphdr for IPsec processing so we do that later.
799 		 */
800 		/*
801 		 * Locate pcb(s) for datagram.
802 		 */
803 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
804 			in6p = (struct in6pcb *)inph;
805 			if (in6p->in6p_af != AF_INET6)
806 				continue;
807 
808 			if (in6p->in6p_lport != dport)
809 				continue;
810 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
811 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
812 					continue;
813 			} else {
814 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
815 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
816 					continue;
817 			}
818 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
819 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
820 				    &src6) || in6p->in6p_fport != sport)
821 					continue;
822 			} else {
823 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
824 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
825 					continue;
826 			}
827 
828 			udp6_sendup(m, off, (struct sockaddr *)src,
829 				in6p->in6p_socket);
830 			rcvcnt++;
831 
832 			/*
833 			 * Don't look for additional matches if this one does
834 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
835 			 * socket options set.  This heuristic avoids searching
836 			 * through all pcbs in the common case of a non-shared
837 			 * port.  It assumes that an application will never
838 			 * clear these options after setting them.
839 			 */
840 			if ((in6p->in6p_socket->so_options &
841 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
842 				break;
843 		}
844 	} else {
845 		/*
846 		 * Locate pcb for datagram.
847 		 */
848 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
849 		    &dst6, dport, 0);
850 		if (in6p == 0) {
851 			++udpstat.udps_pcbhashmiss;
852 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
853 			if (in6p == 0)
854 				return rcvcnt;
855 		}
856 
857 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
858 		rcvcnt++;
859 	}
860 
861 bad:
862 	return rcvcnt;
863 }
864 #endif
865 
866 #ifdef INET
867 /*
868  * Notify a udp user of an asynchronous error;
869  * just wake up so that he can collect error status.
870  */
871 static void
872 udp_notify(struct inpcb *inp, int errno)
873 {
874 	inp->inp_socket->so_error = errno;
875 	sorwakeup(inp->inp_socket);
876 	sowwakeup(inp->inp_socket);
877 }
878 
879 void *
880 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
881 {
882 	struct ip *ip = v;
883 	struct udphdr *uh;
884 	void (*notify)(struct inpcb *, int) = udp_notify;
885 	int errno;
886 
887 	if (sa->sa_family != AF_INET
888 	 || sa->sa_len != sizeof(struct sockaddr_in))
889 		return NULL;
890 	if ((unsigned)cmd >= PRC_NCMDS)
891 		return NULL;
892 	errno = inetctlerrmap[cmd];
893 	if (PRC_IS_REDIRECT(cmd))
894 		notify = in_rtchange, ip = 0;
895 	else if (cmd == PRC_HOSTDEAD)
896 		ip = 0;
897 	else if (errno == 0)
898 		return NULL;
899 	if (ip) {
900 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
901 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
902 		    ip->ip_src, uh->uh_sport, errno, notify);
903 
904 		/* XXX mapped address case */
905 	} else
906 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
907 		    notify);
908 	return NULL;
909 }
910 
911 int
912 udp_output(struct mbuf *m, ...)
913 {
914 	struct inpcb *inp;
915 	struct udpiphdr *ui;
916 	struct route *ro;
917 	int len = m->m_pkthdr.len;
918 	int error = 0;
919 	va_list ap;
920 
921 	MCLAIM(m, &udp_tx_mowner);
922 	va_start(ap, m);
923 	inp = va_arg(ap, struct inpcb *);
924 	va_end(ap);
925 
926 	/*
927 	 * Calculate data length and get a mbuf
928 	 * for UDP and IP headers.
929 	 */
930 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
931 	if (m == 0) {
932 		error = ENOBUFS;
933 		goto release;
934 	}
935 
936 	/*
937 	 * Compute the packet length of the IP header, and
938 	 * punt if the length looks bogus.
939 	 */
940 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
941 		error = EMSGSIZE;
942 		goto release;
943 	}
944 
945 	/*
946 	 * Fill in mbuf with extended UDP header
947 	 * and addresses and length put into network format.
948 	 */
949 	ui = mtod(m, struct udpiphdr *);
950 	ui->ui_pr = IPPROTO_UDP;
951 	ui->ui_src = inp->inp_laddr;
952 	ui->ui_dst = inp->inp_faddr;
953 	ui->ui_sport = inp->inp_lport;
954 	ui->ui_dport = inp->inp_fport;
955 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
956 
957 	ro = &inp->inp_route;
958 
959 	/*
960 	 * Set up checksum and output datagram.
961 	 */
962 	if (udpcksum) {
963 		/*
964 		 * XXX Cache pseudo-header checksum part for
965 		 * XXX "connected" UDP sockets.
966 		 * Maybe skip checksums on loopback interfaces.
967 		 */
968 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
969 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
970 		    sizeof(struct udphdr) + IPPROTO_UDP));
971 		if (__predict_true(ro->ro_rt == NULL ||
972 				   !(ro->ro_rt->rt_ifp->if_flags &
973 				     IFF_LOOPBACK) ||
974 				   udp_do_loopback_cksum))
975 			m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
976 		else
977 			m->m_pkthdr.csum_flags = 0;
978 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
979 	} else
980 		ui->ui_sum = 0;
981 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
982 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
983 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
984 	udpstat.udps_opackets++;
985 
986 	return (ip_output(m, inp->inp_options, ro,
987 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
988 	    inp->inp_moptions, inp->inp_socket));
989 
990 release:
991 	m_freem(m);
992 	return (error);
993 }
994 
995 int	udp_sendspace = 9216;		/* really max datagram size */
996 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
997 					/* 40 1K datagrams */
998 
999 /*ARGSUSED*/
1000 int
1001 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1002 	struct mbuf *control, struct proc *p)
1003 {
1004 	struct inpcb *inp;
1005 	int s;
1006 	int error = 0;
1007 
1008 	if (req == PRU_CONTROL)
1009 		return (in_control(so, (long)m, (caddr_t)nam,
1010 		    (struct ifnet *)control, p));
1011 
1012 	if (req == PRU_PURGEIF) {
1013 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1014 		in_purgeif((struct ifnet *)control);
1015 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1016 		return (0);
1017 	}
1018 
1019 	s = splsoftnet();
1020 	inp = sotoinpcb(so);
1021 #ifdef DIAGNOSTIC
1022 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1023 		panic("udp_usrreq: unexpected control mbuf");
1024 #endif
1025 	if (inp == 0 && req != PRU_ATTACH) {
1026 		error = EINVAL;
1027 		goto release;
1028 	}
1029 
1030 	/*
1031 	 * Note: need to block udp_input while changing
1032 	 * the udp pcb queue and/or pcb addresses.
1033 	 */
1034 	switch (req) {
1035 
1036 	case PRU_ATTACH:
1037 		if (inp != 0) {
1038 			error = EISCONN;
1039 			break;
1040 		}
1041 #ifdef MBUFTRACE
1042 		so->so_mowner = &udp_mowner;
1043 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1044 		so->so_snd.sb_mowner = &udp_tx_mowner;
1045 #endif
1046 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1047 			error = soreserve(so, udp_sendspace, udp_recvspace);
1048 			if (error)
1049 				break;
1050 		}
1051 		error = in_pcballoc(so, &udbtable);
1052 		if (error)
1053 			break;
1054 		inp = sotoinpcb(so);
1055 		inp->inp_ip.ip_ttl = ip_defttl;
1056 		break;
1057 
1058 	case PRU_DETACH:
1059 		in_pcbdetach(inp);
1060 		break;
1061 
1062 	case PRU_BIND:
1063 		error = in_pcbbind(inp, nam, p);
1064 		break;
1065 
1066 	case PRU_LISTEN:
1067 		error = EOPNOTSUPP;
1068 		break;
1069 
1070 	case PRU_CONNECT:
1071 		error = in_pcbconnect(inp, nam);
1072 		if (error)
1073 			break;
1074 		soisconnected(so);
1075 		break;
1076 
1077 	case PRU_CONNECT2:
1078 		error = EOPNOTSUPP;
1079 		break;
1080 
1081 	case PRU_DISCONNECT:
1082 		/*soisdisconnected(so);*/
1083 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1084 		in_pcbdisconnect(inp);
1085 		inp->inp_laddr = zeroin_addr;		/* XXX */
1086 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1087 		break;
1088 
1089 	case PRU_SHUTDOWN:
1090 		socantsendmore(so);
1091 		break;
1092 
1093 	case PRU_RCVD:
1094 		error = EOPNOTSUPP;
1095 		break;
1096 
1097 	case PRU_SEND:
1098 		if (control && control->m_len) {
1099 			m_freem(control);
1100 			m_freem(m);
1101 			error = EINVAL;
1102 			break;
1103 		}
1104 	{
1105 		struct in_addr laddr;			/* XXX */
1106 
1107 		if (nam) {
1108 			laddr = inp->inp_laddr;		/* XXX */
1109 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1110 				error = EISCONN;
1111 				goto die;
1112 			}
1113 			error = in_pcbconnect(inp, nam);
1114 			if (error)
1115 				goto die;
1116 		} else {
1117 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1118 				error = ENOTCONN;
1119 				goto die;
1120 			}
1121 		}
1122 		error = udp_output(m, inp);
1123 		m = NULL;
1124 		if (nam) {
1125 			in_pcbdisconnect(inp);
1126 			inp->inp_laddr = laddr;		/* XXX */
1127 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1128 		}
1129 	  die:
1130 		if (m)
1131 			m_freem(m);
1132 	}
1133 		break;
1134 
1135 	case PRU_SENSE:
1136 		/*
1137 		 * stat: don't bother with a blocksize.
1138 		 */
1139 		splx(s);
1140 		return (0);
1141 
1142 	case PRU_RCVOOB:
1143 		error =  EOPNOTSUPP;
1144 		break;
1145 
1146 	case PRU_SENDOOB:
1147 		m_freem(control);
1148 		m_freem(m);
1149 		error =  EOPNOTSUPP;
1150 		break;
1151 
1152 	case PRU_SOCKADDR:
1153 		in_setsockaddr(inp, nam);
1154 		break;
1155 
1156 	case PRU_PEERADDR:
1157 		in_setpeeraddr(inp, nam);
1158 		break;
1159 
1160 	default:
1161 		panic("udp_usrreq");
1162 	}
1163 
1164 release:
1165 	splx(s);
1166 	return (error);
1167 }
1168 
1169 /*
1170  * Sysctl for udp variables.
1171  */
1172 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1173 {
1174 
1175 	sysctl_createv(clog, 0, NULL, NULL,
1176 		       CTLFLAG_PERMANENT,
1177 		       CTLTYPE_NODE, "net", NULL,
1178 		       NULL, 0, NULL, 0,
1179 		       CTL_NET, CTL_EOL);
1180 	sysctl_createv(clog, 0, NULL, NULL,
1181 		       CTLFLAG_PERMANENT,
1182 		       CTLTYPE_NODE, "inet", NULL,
1183 		       NULL, 0, NULL, 0,
1184 		       CTL_NET, PF_INET, CTL_EOL);
1185 	sysctl_createv(clog, 0, NULL, NULL,
1186 		       CTLFLAG_PERMANENT,
1187 		       CTLTYPE_NODE, "udp",
1188 		       SYSCTL_DESCR("UDPv4 related settings"),
1189 		       NULL, 0, NULL, 0,
1190 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1191 
1192 	sysctl_createv(clog, 0, NULL, NULL,
1193 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1194 		       CTLTYPE_INT, "checksum",
1195 		       SYSCTL_DESCR("Compute UDP checksums"),
1196 		       NULL, 0, &udpcksum, 0,
1197 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1198 		       CTL_EOL);
1199 	sysctl_createv(clog, 0, NULL, NULL,
1200 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1201 		       CTLTYPE_INT, "sendspace",
1202 		       SYSCTL_DESCR("Default UDP send buffer size"),
1203 		       NULL, 0, &udp_sendspace, 0,
1204 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1205 		       CTL_EOL);
1206 	sysctl_createv(clog, 0, NULL, NULL,
1207 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1208 		       CTLTYPE_INT, "recvspace",
1209 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1210 		       NULL, 0, &udp_recvspace, 0,
1211 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1212 		       CTL_EOL);
1213 	sysctl_createv(clog, 0, NULL, NULL,
1214 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1215 		       CTLTYPE_INT, "do_loopback_cksum",
1216 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1217 		       NULL, 0, &udp_do_loopback_cksum, 0,
1218 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1219 		       CTL_EOL);
1220 }
1221 #endif
1222