xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision d20841bb642898112fe68f0ad3f7b26dddf56f07)
1 /*	$NetBSD: udp_usrreq.c,v 1.114 2003/12/04 19:38:24 atatat 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 1.114 2003/12/04 19:38:24 atatat Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_ipsec.h"
68 #include "opt_inet_csum.h"
69 #include "opt_ipkdb.h"
70 #include "opt_mbuftrace.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/stat.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_icmp.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/udp6_var.h>
104 #endif
105 
106 #ifndef INET6
107 /* always need ip6.h for IP6_EXTHDR_GET */
108 #include <netinet/ip6.h>
109 #endif
110 
111 #include "faith.h"
112 #if defined(NFAITH) && NFAITH > 0
113 #include <net/if_faith.h>
114 #endif
115 
116 #include <machine/stdarg.h>
117 
118 #ifdef FAST_IPSEC
119 #include <netipsec/ipsec.h>
120 #ifdef INET6
121 #include <netipsec/ipsec6.h>
122 #endif
123 #endif	/* FAST_IPSEC*/
124 
125 #ifdef IPSEC
126 #include <netinet6/ipsec.h>
127 #include <netkey/key.h>
128 #endif /*IPSEC*/
129 
130 #ifdef IPKDB
131 #include <ipkdb/ipkdb.h>
132 #endif
133 
134 /*
135  * UDP protocol implementation.
136  * Per RFC 768, August, 1980.
137  */
138 #ifndef	COMPAT_42
139 int	udpcksum = 1;
140 #else
141 int	udpcksum = 0;		/* XXX */
142 #endif
143 
144 struct	inpcbtable udbtable;
145 struct	udpstat udpstat;
146 
147 #ifdef INET
148 static void udp4_sendup __P((struct mbuf *, int, struct sockaddr *,
149 	struct socket *));
150 static int udp4_realinput __P((struct sockaddr_in *, struct sockaddr_in *,
151 	struct mbuf *, int));
152 #endif
153 #ifdef INET6
154 static void udp6_sendup __P((struct mbuf *, int, struct sockaddr *,
155 	struct socket *));
156 static int udp6_realinput __P((int, struct sockaddr_in6 *,
157 	struct sockaddr_in6 *, struct mbuf *, int));
158 #endif
159 #ifdef INET
160 static	void udp_notify __P((struct inpcb *, int));
161 #endif
162 
163 #ifndef UDBHASHSIZE
164 #define	UDBHASHSIZE	128
165 #endif
166 int	udbhashsize = UDBHASHSIZE;
167 
168 #ifdef MBUFTRACE
169 struct mowner udp_mowner = { "udp" };
170 struct mowner udp_rx_mowner = { "udp", "rx" };
171 struct mowner udp_tx_mowner = { "udp", "tx" };
172 #endif
173 
174 #ifdef UDP_CSUM_COUNTERS
175 #include <sys/device.h>
176 
177 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
178     NULL, "udp", "hwcsum bad");
179 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
180     NULL, "udp", "hwcsum ok");
181 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
182     NULL, "udp", "hwcsum data");
183 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
184     NULL, "udp", "swcsum");
185 
186 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
187 
188 #else
189 
190 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
191 
192 #endif /* UDP_CSUM_COUNTERS */
193 
194 void
195 udp_init()
196 {
197 
198 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
199 
200 #ifdef UDP_CSUM_COUNTERS
201 	evcnt_attach_static(&udp_hwcsum_bad);
202 	evcnt_attach_static(&udp_hwcsum_ok);
203 	evcnt_attach_static(&udp_hwcsum_data);
204 	evcnt_attach_static(&udp_swcsum);
205 #endif /* UDP_CSUM_COUNTERS */
206 
207 	MOWNER_ATTACH(&udp_tx_mowner);
208 	MOWNER_ATTACH(&udp_rx_mowner);
209 	MOWNER_ATTACH(&udp_mowner);
210 }
211 
212 #ifdef INET
213 void
214 #if __STDC__
215 udp_input(struct mbuf *m, ...)
216 #else
217 udp_input(m, va_alist)
218 	struct mbuf *m;
219 	va_dcl
220 #endif
221 {
222 	va_list ap;
223 	struct sockaddr_in src, dst;
224 	struct ip *ip;
225 	struct udphdr *uh;
226 	int iphlen;
227 	int len;
228 	int n;
229 	u_int16_t ip_len;
230 
231 	va_start(ap, m);
232 	iphlen = va_arg(ap, int);
233 	(void)va_arg(ap, int);		/* ignore value, advance ap */
234 	va_end(ap);
235 
236 	MCLAIM(m, &udp_rx_mowner);
237 	udpstat.udps_ipackets++;
238 
239 	/*
240 	 * Get IP and UDP header together in first mbuf.
241 	 */
242 	ip = mtod(m, struct ip *);
243 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
244 	if (uh == NULL) {
245 		udpstat.udps_hdrops++;
246 		return;
247 	}
248 	KASSERT(UDP_HDR_ALIGNED_P(uh));
249 
250 	/* destination port of 0 is illegal, based on RFC768. */
251 	if (uh->uh_dport == 0)
252 		goto bad;
253 
254 	/*
255 	 * Make mbuf data length reflect UDP length.
256 	 * If not enough data to reflect UDP length, drop.
257 	 */
258 	ip_len = ntohs(ip->ip_len);
259 	len = ntohs((u_int16_t)uh->uh_ulen);
260 	if (ip_len != iphlen + len) {
261 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
262 			udpstat.udps_badlen++;
263 			goto bad;
264 		}
265 		m_adj(m, iphlen + len - ip_len);
266 	}
267 
268 	/*
269 	 * Checksum extended UDP header and data.
270 	 */
271 	if (uh->uh_sum) {
272 		switch (m->m_pkthdr.csum_flags &
273 		    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
274 		    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
275 		case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
276 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
277 			goto badcsum;
278 
279 		case M_CSUM_UDPv4|M_CSUM_DATA: {
280 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
281 			UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
282 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR)
283 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
284 				    ip->ip_dst.s_addr,
285 				    htonl(hw_csum + ntohs(ip->ip_len) +
286 				    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 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
506 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
507 			    0, 0);
508 		return;
509 	}
510 #endif /*IPSEC*/
511 
512 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
513 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
514 			 || so->so_options & SO_TIMESTAMP)) {
515 			struct ip *ip = mtod(n, struct ip *);
516 			ip_savecontrol(inp, &opts, ip, n);
517 		}
518 
519 		m_adj(n, off);
520 		if (sbappendaddr(&so->so_rcv, src, n,
521 				opts) == 0) {
522 			m_freem(n);
523 			if (opts)
524 				m_freem(opts);
525 			udpstat.udps_fullsock++;
526 		} else
527 			sorwakeup(so);
528 	}
529 }
530 #endif
531 
532 #ifdef INET6
533 static void
534 udp6_sendup(m, off, src, so)
535 	struct mbuf *m;
536 	int off;	/* offset of data portion */
537 	struct sockaddr *src;
538 	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 			udp6stat.udp6s_fullsock++;
574 		} else
575 			sorwakeup(so);
576 	}
577 }
578 #endif
579 
580 #ifdef INET
581 static int
582 udp4_realinput(src, dst, m, off)
583 	struct sockaddr_in *src;
584 	struct sockaddr_in *dst;
585 	struct mbuf *m;
586 	int off;	/* offset of udphdr */
587 {
588 	u_int16_t *sport, *dport;
589 	int rcvcnt;
590 	struct in_addr *src4, *dst4;
591 	struct inpcb_hdr *inph;
592 	struct inpcb *inp;
593 
594 	rcvcnt = 0;
595 	off += sizeof(struct udphdr);	/* now, offset of payload */
596 
597 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
598 		goto bad;
599 
600 	src4 = &src->sin_addr;
601 	sport = &src->sin_port;
602 	dst4 = &dst->sin_addr;
603 	dport = &dst->sin_port;
604 
605 	if (IN_MULTICAST(dst4->s_addr) ||
606 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
607 		/*
608 		 * Deliver a multicast or broadcast datagram to *all* sockets
609 		 * for which the local and remote addresses and ports match
610 		 * those of the incoming datagram.  This allows more than
611 		 * one process to receive multi/broadcasts on the same port.
612 		 * (This really ought to be done for unicast datagrams as
613 		 * well, but that would cause problems with existing
614 		 * applications that open both address-specific sockets and
615 		 * a wildcard socket listening to the same port -- they would
616 		 * end up receiving duplicates of every unicast datagram.
617 		 * Those applications open the multiple sockets to overcome an
618 		 * inadequacy of the UDP socket interface, but for backwards
619 		 * compatibility we avoid the problem here rather than
620 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
621 		 */
622 
623 		/*
624 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
625 		 * we need udpiphdr for IPsec processing so we do that later.
626 		 */
627 		/*
628 		 * Locate pcb(s) for datagram.
629 		 */
630 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
631 			inp = (struct inpcb *)inph;
632 			if (inp->inp_af != AF_INET)
633 				continue;
634 
635 			if (inp->inp_lport != *dport)
636 				continue;
637 			if (!in_nullhost(inp->inp_laddr)) {
638 				if (!in_hosteq(inp->inp_laddr, *dst4))
639 					continue;
640 			}
641 			if (!in_nullhost(inp->inp_faddr)) {
642 				if (!in_hosteq(inp->inp_faddr, *src4) ||
643 				    inp->inp_fport != *sport)
644 					continue;
645 			}
646 
647 			udp4_sendup(m, off, (struct sockaddr *)src,
648 				inp->inp_socket);
649 			rcvcnt++;
650 
651 			/*
652 			 * Don't look for additional matches if this one does
653 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
654 			 * socket options set.  This heuristic avoids searching
655 			 * through all pcbs in the common case of a non-shared
656 			 * port.  It assumes that an application will never
657 			 * clear these options after setting them.
658 			 */
659 			if ((inp->inp_socket->so_options &
660 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
661 				break;
662 		}
663 	} else {
664 		/*
665 		 * Locate pcb for datagram.
666 		 */
667 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
668 		if (inp == 0) {
669 			++udpstat.udps_pcbhashmiss;
670 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
671 			if (inp == 0)
672 				return rcvcnt;
673 		}
674 
675 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
676 		rcvcnt++;
677 	}
678 
679 bad:
680 	return rcvcnt;
681 }
682 #endif
683 
684 #ifdef INET6
685 static int
686 udp6_realinput(af, src, dst, m, off)
687 	int af;		/* af on packet */
688 	struct sockaddr_in6 *src;
689 	struct sockaddr_in6 *dst;
690 	struct mbuf *m;
691 	int off;	/* offset of udphdr */
692 {
693 	u_int16_t sport, dport;
694 	int rcvcnt;
695 	struct in6_addr src6, dst6;
696 	const struct in_addr *dst4;
697 	struct inpcb_hdr *inph;
698 	struct in6pcb *in6p;
699 
700 	rcvcnt = 0;
701 	off += sizeof(struct udphdr);	/* now, offset of payload */
702 
703 	if (af != AF_INET && af != AF_INET6)
704 		goto bad;
705 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
706 		goto bad;
707 
708 	in6_embedscope(&src6, src, NULL, NULL);
709 	sport = src->sin6_port;
710 	in6_embedscope(&dst6, dst, NULL, NULL);
711 	dport = dst->sin6_port;
712 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
713 
714 	if (IN6_IS_ADDR_MULTICAST(&dst6) ||
715 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
716 		/*
717 		 * Deliver a multicast or broadcast datagram to *all* sockets
718 		 * for which the local and remote addresses and ports match
719 		 * those of the incoming datagram.  This allows more than
720 		 * one process to receive multi/broadcasts on the same port.
721 		 * (This really ought to be done for unicast datagrams as
722 		 * well, but that would cause problems with existing
723 		 * applications that open both address-specific sockets and
724 		 * a wildcard socket listening to the same port -- they would
725 		 * end up receiving duplicates of every unicast datagram.
726 		 * Those applications open the multiple sockets to overcome an
727 		 * inadequacy of the UDP socket interface, but for backwards
728 		 * compatibility we avoid the problem here rather than
729 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
730 		 */
731 
732 		/*
733 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
734 		 * we need udpiphdr for IPsec processing so we do that later.
735 		 */
736 		/*
737 		 * Locate pcb(s) for datagram.
738 		 */
739 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
740 			in6p = (struct in6pcb *)inph;
741 			if (in6p->in6p_af != AF_INET6)
742 				continue;
743 
744 			if (in6p->in6p_lport != dport)
745 				continue;
746 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
747 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &dst6))
748 					continue;
749 			} else {
750 				if (IN6_IS_ADDR_V4MAPPED(&dst6) &&
751 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
752 					continue;
753 			}
754 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
755 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
756 				    &src6) || in6p->in6p_fport != sport)
757 					continue;
758 			} else {
759 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
760 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
761 					continue;
762 			}
763 
764 			udp6_sendup(m, off, (struct sockaddr *)src,
765 				in6p->in6p_socket);
766 			rcvcnt++;
767 
768 			/*
769 			 * Don't look for additional matches if this one does
770 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
771 			 * socket options set.  This heuristic avoids searching
772 			 * through all pcbs in the common case of a non-shared
773 			 * port.  It assumes that an application will never
774 			 * clear these options after setting them.
775 			 */
776 			if ((in6p->in6p_socket->so_options &
777 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
778 				break;
779 		}
780 	} else {
781 		/*
782 		 * Locate pcb for datagram.
783 		 */
784 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport,
785 		    &dst6, dport, 0);
786 		if (in6p == 0) {
787 			++udpstat.udps_pcbhashmiss;
788 			in6p = in6_pcblookup_bind(&udbtable, &dst6, dport, 0);
789 			if (in6p == 0)
790 				return rcvcnt;
791 		}
792 
793 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
794 		rcvcnt++;
795 	}
796 
797 bad:
798 	return rcvcnt;
799 }
800 #endif
801 
802 #ifdef INET
803 /*
804  * Notify a udp user of an asynchronous error;
805  * just wake up so that he can collect error status.
806  */
807 static void
808 udp_notify(inp, errno)
809 	struct inpcb *inp;
810 	int errno;
811 {
812 
813 	inp->inp_socket->so_error = errno;
814 	sorwakeup(inp->inp_socket);
815 	sowwakeup(inp->inp_socket);
816 }
817 
818 void *
819 udp_ctlinput(cmd, sa, v)
820 	int cmd;
821 	struct sockaddr *sa;
822 	void *v;
823 {
824 	struct ip *ip = v;
825 	struct udphdr *uh;
826 	void (*notify) __P((struct inpcb *, int)) = udp_notify;
827 	int errno;
828 
829 	if (sa->sa_family != AF_INET
830 	 || sa->sa_len != sizeof(struct sockaddr_in))
831 		return NULL;
832 	if ((unsigned)cmd >= PRC_NCMDS)
833 		return NULL;
834 	errno = inetctlerrmap[cmd];
835 	if (PRC_IS_REDIRECT(cmd))
836 		notify = in_rtchange, ip = 0;
837 	else if (cmd == PRC_HOSTDEAD)
838 		ip = 0;
839 	else if (errno == 0)
840 		return NULL;
841 	if (ip) {
842 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
843 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
844 		    ip->ip_src, uh->uh_sport, errno, notify);
845 
846 		/* XXX mapped address case */
847 	} else
848 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
849 		    notify);
850 	return NULL;
851 }
852 
853 int
854 #if __STDC__
855 udp_output(struct mbuf *m, ...)
856 #else
857 udp_output(m, va_alist)
858 	struct mbuf *m;
859 	va_dcl
860 #endif
861 {
862 	struct inpcb *inp;
863 	struct udpiphdr *ui;
864 	int len = m->m_pkthdr.len;
865 	int error = 0;
866 	va_list ap;
867 
868 	MCLAIM(m, &udp_tx_mowner);
869 	va_start(ap, m);
870 	inp = va_arg(ap, struct inpcb *);
871 	va_end(ap);
872 
873 	/*
874 	 * Calculate data length and get a mbuf
875 	 * for UDP and IP headers.
876 	 */
877 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
878 	if (m == 0) {
879 		error = ENOBUFS;
880 		goto release;
881 	}
882 
883 	/*
884 	 * Compute the packet length of the IP header, and
885 	 * punt if the length looks bogus.
886 	 */
887 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
888 		error = EMSGSIZE;
889 		goto release;
890 	}
891 
892 	/*
893 	 * Fill in mbuf with extended UDP header
894 	 * and addresses and length put into network format.
895 	 */
896 	ui = mtod(m, struct udpiphdr *);
897 	ui->ui_pr = IPPROTO_UDP;
898 	ui->ui_src = inp->inp_laddr;
899 	ui->ui_dst = inp->inp_faddr;
900 	ui->ui_sport = inp->inp_lport;
901 	ui->ui_dport = inp->inp_fport;
902 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
903 
904 	/*
905 	 * Set up checksum and output datagram.
906 	 */
907 	if (udpcksum) {
908 		/*
909 		 * XXX Cache pseudo-header checksum part for
910 		 * XXX "connected" UDP sockets.
911 		 */
912 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
913 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
914 		    sizeof(struct udphdr) + IPPROTO_UDP));
915 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
916 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
917 	} else
918 		ui->ui_sum = 0;
919 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
920 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
921 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
922 	udpstat.udps_opackets++;
923 
924 	return (ip_output(m, inp->inp_options, &inp->inp_route,
925 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
926 	    inp->inp_moptions, inp->inp_socket));
927 
928 release:
929 	m_freem(m);
930 	return (error);
931 }
932 
933 int	udp_sendspace = 9216;		/* really max datagram size */
934 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
935 					/* 40 1K datagrams */
936 
937 /*ARGSUSED*/
938 int
939 udp_usrreq(so, req, m, nam, control, p)
940 	struct socket *so;
941 	int req;
942 	struct mbuf *m, *nam, *control;
943 	struct proc *p;
944 {
945 	struct inpcb *inp;
946 	int s;
947 	int error = 0;
948 
949 	if (req == PRU_CONTROL)
950 		return (in_control(so, (long)m, (caddr_t)nam,
951 		    (struct ifnet *)control, p));
952 
953 	if (req == PRU_PURGEIF) {
954 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
955 		in_purgeif((struct ifnet *)control);
956 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
957 		return (0);
958 	}
959 
960 	s = splsoftnet();
961 	inp = sotoinpcb(so);
962 #ifdef DIAGNOSTIC
963 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
964 		panic("udp_usrreq: unexpected control mbuf");
965 #endif
966 	if (inp == 0 && req != PRU_ATTACH) {
967 		error = EINVAL;
968 		goto release;
969 	}
970 
971 	/*
972 	 * Note: need to block udp_input while changing
973 	 * the udp pcb queue and/or pcb addresses.
974 	 */
975 	switch (req) {
976 
977 	case PRU_ATTACH:
978 		if (inp != 0) {
979 			error = EISCONN;
980 			break;
981 		}
982 #ifdef MBUFTRACE
983 		so->so_mowner = &udp_mowner;
984 		so->so_rcv.sb_mowner = &udp_rx_mowner;
985 		so->so_snd.sb_mowner = &udp_tx_mowner;
986 #endif
987 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
988 			error = soreserve(so, udp_sendspace, udp_recvspace);
989 			if (error)
990 				break;
991 		}
992 		error = in_pcballoc(so, &udbtable);
993 		if (error)
994 			break;
995 		inp = sotoinpcb(so);
996 		inp->inp_ip.ip_ttl = ip_defttl;
997 		break;
998 
999 	case PRU_DETACH:
1000 		in_pcbdetach(inp);
1001 		break;
1002 
1003 	case PRU_BIND:
1004 		error = in_pcbbind(inp, nam, p);
1005 		break;
1006 
1007 	case PRU_LISTEN:
1008 		error = EOPNOTSUPP;
1009 		break;
1010 
1011 	case PRU_CONNECT:
1012 		error = in_pcbconnect(inp, nam);
1013 		if (error)
1014 			break;
1015 		soisconnected(so);
1016 		break;
1017 
1018 	case PRU_CONNECT2:
1019 		error = EOPNOTSUPP;
1020 		break;
1021 
1022 	case PRU_DISCONNECT:
1023 		/*soisdisconnected(so);*/
1024 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1025 		in_pcbdisconnect(inp);
1026 		inp->inp_laddr = zeroin_addr;		/* XXX */
1027 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1028 		break;
1029 
1030 	case PRU_SHUTDOWN:
1031 		socantsendmore(so);
1032 		break;
1033 
1034 	case PRU_RCVD:
1035 		error = EOPNOTSUPP;
1036 		break;
1037 
1038 	case PRU_SEND:
1039 		if (control && control->m_len) {
1040 			m_freem(control);
1041 			m_freem(m);
1042 			error = EINVAL;
1043 			break;
1044 		}
1045 	{
1046 		struct in_addr laddr;			/* XXX */
1047 
1048 		if (nam) {
1049 			laddr = inp->inp_laddr;		/* XXX */
1050 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1051 				error = EISCONN;
1052 				goto die;
1053 			}
1054 			error = in_pcbconnect(inp, nam);
1055 			if (error)
1056 				goto die;
1057 		} else {
1058 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1059 				error = ENOTCONN;
1060 				goto die;
1061 			}
1062 		}
1063 		error = udp_output(m, inp);
1064 		m = NULL;
1065 		if (nam) {
1066 			in_pcbdisconnect(inp);
1067 			inp->inp_laddr = laddr;		/* XXX */
1068 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1069 		}
1070 	  die:
1071 		if (m)
1072 			m_freem(m);
1073 	}
1074 		break;
1075 
1076 	case PRU_SENSE:
1077 		/*
1078 		 * stat: don't bother with a blocksize.
1079 		 */
1080 		splx(s);
1081 		return (0);
1082 
1083 	case PRU_RCVOOB:
1084 		error =  EOPNOTSUPP;
1085 		break;
1086 
1087 	case PRU_SENDOOB:
1088 		m_freem(control);
1089 		m_freem(m);
1090 		error =  EOPNOTSUPP;
1091 		break;
1092 
1093 	case PRU_SOCKADDR:
1094 		in_setsockaddr(inp, nam);
1095 		break;
1096 
1097 	case PRU_PEERADDR:
1098 		in_setpeeraddr(inp, nam);
1099 		break;
1100 
1101 	default:
1102 		panic("udp_usrreq");
1103 	}
1104 
1105 release:
1106 	splx(s);
1107 	return (error);
1108 }
1109 
1110 /*
1111  * Sysctl for udp variables.
1112  */
1113 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1114 {
1115 
1116 	sysctl_createv(SYSCTL_PERMANENT,
1117 		       CTLTYPE_NODE, "net", NULL,
1118 		       NULL, 0, NULL, 0,
1119 		       CTL_NET, CTL_EOL);
1120 	sysctl_createv(SYSCTL_PERMANENT,
1121 		       CTLTYPE_NODE, "inet", NULL,
1122 		       NULL, 0, NULL, 0,
1123 		       CTL_NET, PF_INET, CTL_EOL);
1124 	sysctl_createv(SYSCTL_PERMANENT,
1125 		       CTLTYPE_NODE, "udp", NULL,
1126 		       NULL, 0, NULL, 0,
1127 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1128 
1129 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE,
1130 		       CTLTYPE_INT, "checksum", NULL,
1131 		       NULL, 0, &udpcksum, 0,
1132 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1133 		       CTL_EOL);
1134 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE,
1135 		       CTLTYPE_INT, "sendspace", NULL,
1136 		       NULL, 0, &udp_sendspace, 0,
1137 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1138 		       CTL_EOL);
1139 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE,
1140 		       CTLTYPE_INT, "recvspace", NULL,
1141 		       NULL, 0, &udp_recvspace, 0,
1142 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1143 		       CTL_EOL);
1144 }
1145 #endif
1146