xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 6d322f2f4598f0d8a138f10ea648ec4fabe41f8b)
1 /*	$NetBSD: udp_usrreq.c,v 1.193 2014/01/04 14:18:12 pooka 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 1.193 2014/01/04 14:18:12 pooka Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_compat_netbsd.h"
68 #include "opt_ipsec.h"
69 #include "opt_inet_csum.h"
70 #include "opt_ipkdb.h"
71 #include "opt_mbuftrace.h"
72 
73 #include <sys/param.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/once.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/errno.h>
81 #include <sys/stat.h>
82 #include <sys/systm.h>
83 #include <sys/proc.h>
84 #include <sys/domain.h>
85 #include <sys/sysctl.h>
86 
87 #include <net/if.h>
88 #include <net/route.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/in_var.h>
93 #include <netinet/ip.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/ip_icmp.h>
97 #include <netinet/udp.h>
98 #include <netinet/udp_var.h>
99 #include <netinet/udp_private.h>
100 
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/ip6_private.h>
106 #include <netinet6/in6_pcb.h>
107 #include <netinet6/udp6_var.h>
108 #include <netinet6/udp6_private.h>
109 #include <netinet6/scope6_var.h>
110 #endif
111 
112 #ifndef INET6
113 /* always need ip6.h for IP6_EXTHDR_GET */
114 #include <netinet/ip6.h>
115 #endif
116 
117 #include "faith.h"
118 #if defined(NFAITH) && NFAITH > 0
119 #include <net/if_faith.h>
120 #endif
121 
122 #ifdef IPSEC
123 #include <netipsec/ipsec.h>
124 #include <netipsec/ipsec_var.h>
125 #include <netipsec/ipsec_private.h>
126 #include <netipsec/esp.h>
127 #ifdef INET6
128 #include <netipsec/ipsec6.h>
129 #endif
130 #endif	/* IPSEC */
131 
132 #ifdef COMPAT_50
133 #include <compat/sys/socket.h>
134 #endif
135 
136 #ifdef IPKDB
137 #include <ipkdb/ipkdb.h>
138 #endif
139 
140 /*
141  * UDP protocol implementation.
142  * Per RFC 768, August, 1980.
143  */
144 int	udpcksum = 1;
145 int	udp_do_loopback_cksum = 0;
146 
147 struct	inpcbtable udbtable;
148 
149 percpu_t *udpstat_percpu;
150 
151 #ifdef INET
152 #ifdef IPSEC
153 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
154 	struct socket *);
155 #endif
156 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
157 	struct socket *);
158 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
159 	struct mbuf **, int);
160 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
161 #endif
162 #ifdef INET6
163 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
164 	struct socket *);
165 static int udp6_realinput (int, struct sockaddr_in6 *,
166 	struct sockaddr_in6 *, struct mbuf *, int);
167 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
168 #endif
169 #ifdef INET
170 static	void udp_notify (struct inpcb *, int);
171 #endif
172 
173 #ifndef UDBHASHSIZE
174 #define	UDBHASHSIZE	128
175 #endif
176 int	udbhashsize = UDBHASHSIZE;
177 
178 #ifdef MBUFTRACE
179 struct mowner udp_mowner = MOWNER_INIT("udp", "");
180 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
181 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
182 #endif
183 
184 #ifdef UDP_CSUM_COUNTERS
185 #include <sys/device.h>
186 
187 #if defined(INET)
188 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
189     NULL, "udp", "hwcsum bad");
190 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
191     NULL, "udp", "hwcsum ok");
192 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
193     NULL, "udp", "hwcsum data");
194 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
195     NULL, "udp", "swcsum");
196 
197 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
198 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
199 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
200 EVCNT_ATTACH_STATIC(udp_swcsum);
201 #endif /* defined(INET) */
202 
203 #if defined(INET6)
204 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
205     NULL, "udp6", "hwcsum bad");
206 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
207     NULL, "udp6", "hwcsum ok");
208 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
209     NULL, "udp6", "hwcsum data");
210 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
211     NULL, "udp6", "swcsum");
212 
213 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
214 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
215 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
216 EVCNT_ATTACH_STATIC(udp6_swcsum);
217 #endif /* defined(INET6) */
218 
219 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
220 
221 #else
222 
223 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
224 
225 #endif /* UDP_CSUM_COUNTERS */
226 
227 static void sysctl_net_inet_udp_setup(struct sysctllog **);
228 
229 static int
230 do_udpinit(void)
231 {
232 
233 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
234 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
235 
236 	MOWNER_ATTACH(&udp_tx_mowner);
237 	MOWNER_ATTACH(&udp_rx_mowner);
238 	MOWNER_ATTACH(&udp_mowner);
239 
240 	return 0;
241 }
242 
243 void
244 udp_init_common(void)
245 {
246 	static ONCE_DECL(doudpinit);
247 
248 	RUN_ONCE(&doudpinit, do_udpinit);
249 }
250 
251 void
252 udp_init(void)
253 {
254 
255 	sysctl_net_inet_udp_setup(NULL);
256 
257 	udp_init_common();
258 }
259 
260 /*
261  * Checksum extended UDP header and data.
262  */
263 
264 int
265 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
266     int iphlen, int len)
267 {
268 
269 	switch (af) {
270 #ifdef INET
271 	case AF_INET:
272 		return udp4_input_checksum(m, uh, iphlen, len);
273 #endif
274 #ifdef INET6
275 	case AF_INET6:
276 		return udp6_input_checksum(m, uh, iphlen, len);
277 #endif
278 	}
279 #ifdef DIAGNOSTIC
280 	panic("udp_input_checksum: unknown af %d", af);
281 #endif
282 	/* NOTREACHED */
283 	return -1;
284 }
285 
286 #ifdef INET
287 
288 /*
289  * Checksum extended UDP header and data.
290  */
291 
292 static int
293 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
294     int iphlen, int len)
295 {
296 
297 	/*
298 	 * XXX it's better to record and check if this mbuf is
299 	 * already checked.
300 	 */
301 
302 	if (uh->uh_sum == 0)
303 		return 0;
304 
305 	switch (m->m_pkthdr.csum_flags &
306 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
307 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
308 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
309 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
310 		goto badcsum;
311 
312 	case M_CSUM_UDPv4|M_CSUM_DATA: {
313 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
314 
315 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
316 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
317 			const struct ip *ip =
318 			    mtod(m, const struct ip *);
319 
320 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
321 			    ip->ip_dst.s_addr,
322 			    htons(hw_csum + len + IPPROTO_UDP));
323 		}
324 		if ((hw_csum ^ 0xffff) != 0)
325 			goto badcsum;
326 		break;
327 	}
328 
329 	case M_CSUM_UDPv4:
330 		/* Checksum was okay. */
331 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
332 		break;
333 
334 	default:
335 		/*
336 		 * Need to compute it ourselves.  Maybe skip checksum
337 		 * on loopback interfaces.
338 		 */
339 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
340 				     IFF_LOOPBACK) ||
341 				   udp_do_loopback_cksum)) {
342 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
343 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
344 				goto badcsum;
345 		}
346 		break;
347 	}
348 
349 	return 0;
350 
351 badcsum:
352 	UDP_STATINC(UDP_STAT_BADSUM);
353 	return -1;
354 }
355 
356 void
357 udp_input(struct mbuf *m, ...)
358 {
359 	va_list ap;
360 	struct sockaddr_in src, dst;
361 	struct ip *ip;
362 	struct udphdr *uh;
363 	int iphlen;
364 	int len;
365 	int n;
366 	u_int16_t ip_len;
367 
368 	va_start(ap, m);
369 	iphlen = va_arg(ap, int);
370 	(void)va_arg(ap, int);		/* ignore value, advance ap */
371 	va_end(ap);
372 
373 	MCLAIM(m, &udp_rx_mowner);
374 	UDP_STATINC(UDP_STAT_IPACKETS);
375 
376 	/*
377 	 * Get IP and UDP header together in first mbuf.
378 	 */
379 	ip = mtod(m, struct ip *);
380 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
381 	if (uh == NULL) {
382 		UDP_STATINC(UDP_STAT_HDROPS);
383 		return;
384 	}
385 	KASSERT(UDP_HDR_ALIGNED_P(uh));
386 
387 	/* destination port of 0 is illegal, based on RFC768. */
388 	if (uh->uh_dport == 0)
389 		goto bad;
390 
391 	/*
392 	 * Make mbuf data length reflect UDP length.
393 	 * If not enough data to reflect UDP length, drop.
394 	 */
395 	ip_len = ntohs(ip->ip_len);
396 	len = ntohs((u_int16_t)uh->uh_ulen);
397 	if (ip_len != iphlen + len) {
398 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
399 			UDP_STATINC(UDP_STAT_BADLEN);
400 			goto bad;
401 		}
402 		m_adj(m, iphlen + len - ip_len);
403 	}
404 
405 	/*
406 	 * Checksum extended UDP header and data.
407 	 */
408 	if (udp4_input_checksum(m, uh, iphlen, len))
409 		goto badcsum;
410 
411 	/* construct source and dst sockaddrs. */
412 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
413 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
414 
415 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
416 		UDP_STATINC(UDP_STAT_HDROPS);
417 		return;
418 	}
419 	if (m == NULL) {
420 		/*
421 		 * packet has been processed by ESP stuff -
422 		 * e.g. dropped NAT-T-keep-alive-packet ...
423 		 */
424 		return;
425 	}
426 	ip = mtod(m, struct ip *);
427 #ifdef INET6
428 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
429 		struct sockaddr_in6 src6, dst6;
430 
431 		memset(&src6, 0, sizeof(src6));
432 		src6.sin6_family = AF_INET6;
433 		src6.sin6_len = sizeof(struct sockaddr_in6);
434 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
435 		memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src,
436 			sizeof(ip->ip_src));
437 		src6.sin6_port = uh->uh_sport;
438 		memset(&dst6, 0, sizeof(dst6));
439 		dst6.sin6_family = AF_INET6;
440 		dst6.sin6_len = sizeof(struct sockaddr_in6);
441 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
442 		memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst,
443 			sizeof(ip->ip_dst));
444 		dst6.sin6_port = uh->uh_dport;
445 
446 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
447 	}
448 #endif
449 
450 	if (n == 0) {
451 		if (m->m_flags & (M_BCAST | M_MCAST)) {
452 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
453 			goto bad;
454 		}
455 		UDP_STATINC(UDP_STAT_NOPORT);
456 #ifdef IPKDB
457 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
458 				m, iphlen + sizeof(struct udphdr),
459 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
460 			/*
461 			 * It was a debugger connect packet,
462 			 * just drop it now
463 			 */
464 			goto bad;
465 		}
466 #endif
467 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
468 		m = NULL;
469 	}
470 
471 bad:
472 	if (m)
473 		m_freem(m);
474 	return;
475 
476 badcsum:
477 	m_freem(m);
478 }
479 #endif
480 
481 #ifdef INET6
482 static int
483 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
484 {
485 
486 	/*
487 	 * XXX it's better to record and check if this mbuf is
488 	 * already checked.
489 	 */
490 
491 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
492 		goto good;
493 	}
494 	if (uh->uh_sum == 0) {
495 		UDP6_STATINC(UDP6_STAT_NOSUM);
496 		goto bad;
497 	}
498 
499 	switch (m->m_pkthdr.csum_flags &
500 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
501 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
502 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
503 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
504 		UDP6_STATINC(UDP6_STAT_BADSUM);
505 		goto bad;
506 
507 #if 0 /* notyet */
508 	case M_CSUM_UDPv6|M_CSUM_DATA:
509 #endif
510 
511 	case M_CSUM_UDPv6:
512 		/* Checksum was okay. */
513 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
514 		break;
515 
516 	default:
517 		/*
518 		 * Need to compute it ourselves.  Maybe skip checksum
519 		 * on loopback interfaces.
520 		 */
521 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
522 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
523 			UDP6_STATINC(UDP6_STAT_BADSUM);
524 			goto bad;
525 		}
526 	}
527 
528 good:
529 	return 0;
530 bad:
531 	return -1;
532 }
533 
534 int
535 udp6_input(struct mbuf **mp, int *offp, int proto)
536 {
537 	struct mbuf *m = *mp;
538 	int off = *offp;
539 	struct sockaddr_in6 src, dst;
540 	struct ip6_hdr *ip6;
541 	struct udphdr *uh;
542 	u_int32_t plen, ulen;
543 
544 	ip6 = mtod(m, struct ip6_hdr *);
545 
546 #if defined(NFAITH) && 0 < NFAITH
547 	if (faithprefix(&ip6->ip6_dst)) {
548 		/* send icmp6 host unreach? */
549 		m_freem(m);
550 		return IPPROTO_DONE;
551 	}
552 #endif
553 
554 	UDP6_STATINC(UDP6_STAT_IPACKETS);
555 
556 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
557 	plen = m->m_pkthdr.len - off;
558 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
559 	if (uh == NULL) {
560 		IP6_STATINC(IP6_STAT_TOOSHORT);
561 		return IPPROTO_DONE;
562 	}
563 	KASSERT(UDP_HDR_ALIGNED_P(uh));
564 	ulen = ntohs((u_short)uh->uh_ulen);
565 	/*
566 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
567 	 * iff payload length > 0xffff.
568 	 */
569 	if (ulen == 0 && plen > 0xffff)
570 		ulen = plen;
571 
572 	if (plen != ulen) {
573 		UDP6_STATINC(UDP6_STAT_BADLEN);
574 		goto bad;
575 	}
576 
577 	/* destination port of 0 is illegal, based on RFC768. */
578 	if (uh->uh_dport == 0)
579 		goto bad;
580 
581 	/* Be proactive about malicious use of IPv4 mapped address */
582 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
583 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
584 		/* XXX stat */
585 		goto bad;
586 	}
587 
588 	/*
589 	 * Checksum extended UDP header and data.  Maybe skip checksum
590 	 * on loopback interfaces.
591 	 */
592 	if (udp6_input_checksum(m, uh, off, ulen))
593 		goto bad;
594 
595 	/*
596 	 * Construct source and dst sockaddrs.
597 	 */
598 	memset(&src, 0, sizeof(src));
599 	src.sin6_family = AF_INET6;
600 	src.sin6_len = sizeof(struct sockaddr_in6);
601 	src.sin6_addr = ip6->ip6_src;
602 	src.sin6_port = uh->uh_sport;
603 	memset(&dst, 0, sizeof(dst));
604 	dst.sin6_family = AF_INET6;
605 	dst.sin6_len = sizeof(struct sockaddr_in6);
606 	dst.sin6_addr = ip6->ip6_dst;
607 	dst.sin6_port = uh->uh_dport;
608 
609 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
610 		if (m->m_flags & M_MCAST) {
611 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
612 			goto bad;
613 		}
614 		UDP6_STATINC(UDP6_STAT_NOPORT);
615 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
616 		m = NULL;
617 	}
618 
619 bad:
620 	if (m)
621 		m_freem(m);
622 	return IPPROTO_DONE;
623 }
624 #endif
625 
626 #ifdef INET
627 static void
628 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
629 	struct sockaddr *src, struct socket *so)
630 {
631 	struct mbuf *opts = NULL;
632 	struct mbuf *n;
633 	struct inpcb *inp = NULL;
634 
635 	if (!so)
636 		return;
637 	switch (so->so_proto->pr_domain->dom_family) {
638 	case AF_INET:
639 		inp = sotoinpcb(so);
640 		break;
641 #ifdef INET6
642 	case AF_INET6:
643 		break;
644 #endif
645 	default:
646 		return;
647 	}
648 
649 #if defined(IPSEC)
650 	/* check AH/ESP integrity. */
651 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
652 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
653 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
654 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
655 			    0, 0);
656 		return;
657 	}
658 #endif /*IPSEC*/
659 
660 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
661 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
662 #ifdef SO_OTIMESTAMP
663 			 || so->so_options & SO_OTIMESTAMP
664 #endif
665 			 || so->so_options & SO_TIMESTAMP)) {
666 			struct ip *ip = mtod(n, struct ip *);
667 			ip_savecontrol(inp, &opts, ip, n);
668 		}
669 
670 		m_adj(n, off);
671 		if (sbappendaddr(&so->so_rcv, src, n,
672 				opts) == 0) {
673 			m_freem(n);
674 			if (opts)
675 				m_freem(opts);
676 			so->so_rcv.sb_overflowed++;
677 			UDP_STATINC(UDP_STAT_FULLSOCK);
678 		} else
679 			sorwakeup(so);
680 	}
681 }
682 #endif
683 
684 #ifdef INET6
685 static void
686 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
687 	struct sockaddr *src, struct socket *so)
688 {
689 	struct mbuf *opts = NULL;
690 	struct mbuf *n;
691 	struct in6pcb *in6p = NULL;
692 
693 	if (!so)
694 		return;
695 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
696 		return;
697 	in6p = sotoin6pcb(so);
698 
699 #if defined(IPSEC)
700 	/* check AH/ESP integrity. */
701 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
702 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
703 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
704 			icmp6_error(n, ICMP6_DST_UNREACH,
705 			    ICMP6_DST_UNREACH_ADMIN, 0);
706 		return;
707 	}
708 #endif /*IPSEC*/
709 
710 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
711 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
712 #ifdef SO_OTIMESTAMP
713 		    || in6p->in6p_socket->so_options & SO_OTIMESTAMP
714 #endif
715 		    || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
716 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
717 			ip6_savecontrol(in6p, &opts, ip6, n);
718 		}
719 
720 		m_adj(n, off);
721 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
722 			m_freem(n);
723 			if (opts)
724 				m_freem(opts);
725 			so->so_rcv.sb_overflowed++;
726 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
727 		} else
728 			sorwakeup(so);
729 	}
730 }
731 #endif
732 
733 #ifdef INET
734 static int
735 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
736 	struct mbuf **mp, int off /* offset of udphdr */)
737 {
738 	u_int16_t *sport, *dport;
739 	int rcvcnt;
740 	struct in_addr *src4, *dst4;
741 	struct inpcb_hdr *inph;
742 	struct inpcb *inp;
743 	struct mbuf *m = *mp;
744 
745 	rcvcnt = 0;
746 	off += sizeof(struct udphdr);	/* now, offset of payload */
747 
748 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
749 		goto bad;
750 
751 	src4 = &src->sin_addr;
752 	sport = &src->sin_port;
753 	dst4 = &dst->sin_addr;
754 	dport = &dst->sin_port;
755 
756 	if (IN_MULTICAST(dst4->s_addr) ||
757 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
758 		/*
759 		 * Deliver a multicast or broadcast datagram to *all* sockets
760 		 * for which the local and remote addresses and ports match
761 		 * those of the incoming datagram.  This allows more than
762 		 * one process to receive multi/broadcasts on the same port.
763 		 * (This really ought to be done for unicast datagrams as
764 		 * well, but that would cause problems with existing
765 		 * applications that open both address-specific sockets and
766 		 * a wildcard socket listening to the same port -- they would
767 		 * end up receiving duplicates of every unicast datagram.
768 		 * Those applications open the multiple sockets to overcome an
769 		 * inadequacy of the UDP socket interface, but for backwards
770 		 * compatibility we avoid the problem here rather than
771 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
772 		 */
773 
774 		/*
775 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
776 		 * we need udpiphdr for IPsec processing so we do that later.
777 		 */
778 		/*
779 		 * Locate pcb(s) for datagram.
780 		 */
781 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
782 			inp = (struct inpcb *)inph;
783 			if (inp->inp_af != AF_INET)
784 				continue;
785 
786 			if (inp->inp_lport != *dport)
787 				continue;
788 			if (!in_nullhost(inp->inp_laddr)) {
789 				if (!in_hosteq(inp->inp_laddr, *dst4))
790 					continue;
791 			}
792 			if (!in_nullhost(inp->inp_faddr)) {
793 				if (!in_hosteq(inp->inp_faddr, *src4) ||
794 				    inp->inp_fport != *sport)
795 					continue;
796 			}
797 
798 			udp4_sendup(m, off, (struct sockaddr *)src,
799 				inp->inp_socket);
800 			rcvcnt++;
801 
802 			/*
803 			 * Don't look for additional matches if this one does
804 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
805 			 * socket options set.  This heuristic avoids searching
806 			 * through all pcbs in the common case of a non-shared
807 			 * port.  It assumes that an application will never
808 			 * clear these options after setting them.
809 			 */
810 			if ((inp->inp_socket->so_options &
811 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
812 				break;
813 		}
814 	} else {
815 		/*
816 		 * Locate pcb for datagram.
817 		 */
818 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
819 		    *dport, 0);
820 		if (inp == 0) {
821 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
822 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
823 			if (inp == 0)
824 				return rcvcnt;
825 		}
826 
827 #ifdef IPSEC
828 		/* Handle ESP over UDP */
829 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
830 			struct sockaddr *sa = (struct sockaddr *)src;
831 
832 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
833 			case -1: 	/* Error, m was freeed */
834 				rcvcnt = -1;
835 				goto bad;
836 				break;
837 
838 			case 1:		/* ESP over UDP */
839 				rcvcnt++;
840 				goto bad;
841 				break;
842 
843 			case 0: 	/* plain UDP */
844 			default: 	/* Unexpected */
845 				/*
846 				 * Normal UDP processing will take place
847 				 * m may have changed.
848 				 */
849 				m = *mp;
850 				break;
851 			}
852 		}
853 #endif
854 
855 		/*
856 		 * Check the minimum TTL for socket.
857 		 */
858 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
859 			goto bad;
860 
861 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
862 		rcvcnt++;
863 	}
864 
865 bad:
866 	return rcvcnt;
867 }
868 #endif
869 
870 #ifdef INET6
871 static int
872 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
873 	struct mbuf *m, int off)
874 {
875 	u_int16_t sport, dport;
876 	int rcvcnt;
877 	struct in6_addr src6, *dst6;
878 	const struct in_addr *dst4;
879 	struct inpcb_hdr *inph;
880 	struct in6pcb *in6p;
881 
882 	rcvcnt = 0;
883 	off += sizeof(struct udphdr);	/* now, offset of payload */
884 
885 	if (af != AF_INET && af != AF_INET6)
886 		goto bad;
887 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
888 		goto bad;
889 
890 	src6 = src->sin6_addr;
891 	if (sa6_recoverscope(src) != 0) {
892 		/* XXX: should be impossible. */
893 		goto bad;
894 	}
895 	sport = src->sin6_port;
896 
897 	dport = dst->sin6_port;
898 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
899 	dst6 = &dst->sin6_addr;
900 
901 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
902 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
903 		/*
904 		 * Deliver a multicast or broadcast datagram to *all* sockets
905 		 * for which the local and remote addresses and ports match
906 		 * those of the incoming datagram.  This allows more than
907 		 * one process to receive multi/broadcasts on the same port.
908 		 * (This really ought to be done for unicast datagrams as
909 		 * well, but that would cause problems with existing
910 		 * applications that open both address-specific sockets and
911 		 * a wildcard socket listening to the same port -- they would
912 		 * end up receiving duplicates of every unicast datagram.
913 		 * Those applications open the multiple sockets to overcome an
914 		 * inadequacy of the UDP socket interface, but for backwards
915 		 * compatibility we avoid the problem here rather than
916 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
917 		 */
918 
919 		/*
920 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
921 		 * we need udpiphdr for IPsec processing so we do that later.
922 		 */
923 		/*
924 		 * Locate pcb(s) for datagram.
925 		 */
926 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
927 			in6p = (struct in6pcb *)inph;
928 			if (in6p->in6p_af != AF_INET6)
929 				continue;
930 
931 			if (in6p->in6p_lport != dport)
932 				continue;
933 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
934 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
935 				    dst6))
936 					continue;
937 			} else {
938 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
939 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
940 					continue;
941 			}
942 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
943 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
944 				    &src6) || in6p->in6p_fport != sport)
945 					continue;
946 			} else {
947 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
948 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
949 					continue;
950 			}
951 
952 			udp6_sendup(m, off, (struct sockaddr *)src,
953 				in6p->in6p_socket);
954 			rcvcnt++;
955 
956 			/*
957 			 * Don't look for additional matches if this one does
958 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
959 			 * socket options set.  This heuristic avoids searching
960 			 * through all pcbs in the common case of a non-shared
961 			 * port.  It assumes that an application will never
962 			 * clear these options after setting them.
963 			 */
964 			if ((in6p->in6p_socket->so_options &
965 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
966 				break;
967 		}
968 	} else {
969 		/*
970 		 * Locate pcb for datagram.
971 		 */
972 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
973 					     dport, 0, 0);
974 		if (in6p == 0) {
975 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
976 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
977 			if (in6p == 0)
978 				return rcvcnt;
979 		}
980 
981 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
982 		rcvcnt++;
983 	}
984 
985 bad:
986 	return rcvcnt;
987 }
988 #endif
989 
990 #ifdef INET
991 /*
992  * Notify a udp user of an asynchronous error;
993  * just wake up so that he can collect error status.
994  */
995 static void
996 udp_notify(struct inpcb *inp, int errno)
997 {
998 	inp->inp_socket->so_error = errno;
999 	sorwakeup(inp->inp_socket);
1000 	sowwakeup(inp->inp_socket);
1001 }
1002 
1003 void *
1004 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
1005 {
1006 	struct ip *ip = v;
1007 	struct udphdr *uh;
1008 	void (*notify)(struct inpcb *, int) = udp_notify;
1009 	int errno;
1010 
1011 	if (sa->sa_family != AF_INET
1012 	 || sa->sa_len != sizeof(struct sockaddr_in))
1013 		return NULL;
1014 	if ((unsigned)cmd >= PRC_NCMDS)
1015 		return NULL;
1016 	errno = inetctlerrmap[cmd];
1017 	if (PRC_IS_REDIRECT(cmd))
1018 		notify = in_rtchange, ip = 0;
1019 	else if (cmd == PRC_HOSTDEAD)
1020 		ip = 0;
1021 	else if (errno == 0)
1022 		return NULL;
1023 	if (ip) {
1024 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
1025 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
1026 		    ip->ip_src, uh->uh_sport, errno, notify);
1027 
1028 		/* XXX mapped address case */
1029 	} else
1030 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
1031 		    notify);
1032 	return NULL;
1033 }
1034 
1035 int
1036 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1037 {
1038 	int s;
1039 	int error = 0;
1040 	struct inpcb *inp;
1041 	int family;
1042 	int optval;
1043 
1044 	family = so->so_proto->pr_domain->dom_family;
1045 
1046 	s = splsoftnet();
1047 	switch (family) {
1048 #ifdef INET
1049 	case PF_INET:
1050 		if (sopt->sopt_level != IPPROTO_UDP) {
1051 			error = ip_ctloutput(op, so, sopt);
1052 			goto end;
1053 		}
1054 		break;
1055 #endif
1056 #ifdef INET6
1057 	case PF_INET6:
1058 		if (sopt->sopt_level != IPPROTO_UDP) {
1059 			error = ip6_ctloutput(op, so, sopt);
1060 			goto end;
1061 		}
1062 		break;
1063 #endif
1064 	default:
1065 		error = EAFNOSUPPORT;
1066 		goto end;
1067 	}
1068 
1069 
1070 	switch (op) {
1071 	case PRCO_SETOPT:
1072 		inp = sotoinpcb(so);
1073 
1074 		switch (sopt->sopt_name) {
1075 		case UDP_ENCAP:
1076 			error = sockopt_getint(sopt, &optval);
1077 			if (error)
1078 				break;
1079 
1080 			switch(optval) {
1081 			case 0:
1082 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1083 				break;
1084 
1085 			case UDP_ENCAP_ESPINUDP:
1086 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1087 				inp->inp_flags |= INP_ESPINUDP;
1088 				break;
1089 
1090 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1091 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1092 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1093 				break;
1094 			default:
1095 				error = EINVAL;
1096 				break;
1097 			}
1098 			break;
1099 
1100 		default:
1101 			error = ENOPROTOOPT;
1102 			break;
1103 		}
1104 		break;
1105 
1106 	default:
1107 		error = EINVAL;
1108 		break;
1109 	}
1110 
1111 end:
1112 	splx(s);
1113 	return error;
1114 }
1115 
1116 
1117 int
1118 udp_output(struct mbuf *m, ...)
1119 {
1120 	struct inpcb *inp;
1121 	struct udpiphdr *ui;
1122 	struct route *ro;
1123 	int len = m->m_pkthdr.len;
1124 	int error = 0;
1125 	va_list ap;
1126 
1127 	MCLAIM(m, &udp_tx_mowner);
1128 	va_start(ap, m);
1129 	inp = va_arg(ap, struct inpcb *);
1130 	va_end(ap);
1131 
1132 	/*
1133 	 * Calculate data length and get a mbuf
1134 	 * for UDP and IP headers.
1135 	 */
1136 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1137 	if (m == 0) {
1138 		error = ENOBUFS;
1139 		goto release;
1140 	}
1141 
1142 	/*
1143 	 * Compute the packet length of the IP header, and
1144 	 * punt if the length looks bogus.
1145 	 */
1146 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1147 		error = EMSGSIZE;
1148 		goto release;
1149 	}
1150 
1151 	/*
1152 	 * Fill in mbuf with extended UDP header
1153 	 * and addresses and length put into network format.
1154 	 */
1155 	ui = mtod(m, struct udpiphdr *);
1156 	ui->ui_pr = IPPROTO_UDP;
1157 	ui->ui_src = inp->inp_laddr;
1158 	ui->ui_dst = inp->inp_faddr;
1159 	ui->ui_sport = inp->inp_lport;
1160 	ui->ui_dport = inp->inp_fport;
1161 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1162 
1163 	ro = &inp->inp_route;
1164 
1165 	/*
1166 	 * Set up checksum and output datagram.
1167 	 */
1168 	if (udpcksum) {
1169 		/*
1170 		 * XXX Cache pseudo-header checksum part for
1171 		 * XXX "connected" UDP sockets.
1172 		 */
1173 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1174 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1175 		    sizeof(struct udphdr) + IPPROTO_UDP));
1176 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1177 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1178 	} else
1179 		ui->ui_sum = 0;
1180 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1181 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1182 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1183 	UDP_STATINC(UDP_STAT_OPACKETS);
1184 
1185 	return (ip_output(m, inp->inp_options, ro,
1186 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1187 	    inp->inp_moptions, inp->inp_socket));
1188 
1189 release:
1190 	m_freem(m);
1191 	return (error);
1192 }
1193 
1194 int	udp_sendspace = 9216;		/* really max datagram size */
1195 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1196 					/* 40 1K datagrams */
1197 
1198 /*ARGSUSED*/
1199 int
1200 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1201 	struct mbuf *control, struct lwp *l)
1202 {
1203 	struct inpcb *inp;
1204 	int s;
1205 	int error = 0;
1206 
1207 	if (req == PRU_CONTROL)
1208 		return (in_control(so, (long)m, (void *)nam,
1209 		    (struct ifnet *)control, l));
1210 
1211 	s = splsoftnet();
1212 
1213 	if (req == PRU_PURGEIF) {
1214 		mutex_enter(softnet_lock);
1215 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1216 		in_purgeif((struct ifnet *)control);
1217 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1218 		mutex_exit(softnet_lock);
1219 		splx(s);
1220 		return (0);
1221 	}
1222 
1223 	inp = sotoinpcb(so);
1224 #ifdef DIAGNOSTIC
1225 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1226 		panic("udp_usrreq: unexpected control mbuf");
1227 #endif
1228 	if (req == PRU_ATTACH) {
1229 		sosetlock(so);
1230 	} else if (inp == 0) {
1231 		error = EINVAL;
1232 		goto release;
1233 	}
1234 
1235 	/*
1236 	 * Note: need to block udp_input while changing
1237 	 * the udp pcb queue and/or pcb addresses.
1238 	 */
1239 	switch (req) {
1240 
1241 	case PRU_ATTACH:
1242 		if (inp != 0) {
1243 			error = EISCONN;
1244 			break;
1245 		}
1246 #ifdef MBUFTRACE
1247 		so->so_mowner = &udp_mowner;
1248 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1249 		so->so_snd.sb_mowner = &udp_tx_mowner;
1250 #endif
1251 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1252 			error = soreserve(so, udp_sendspace, udp_recvspace);
1253 			if (error)
1254 				break;
1255 		}
1256 		error = in_pcballoc(so, &udbtable);
1257 		if (error)
1258 			break;
1259 		inp = sotoinpcb(so);
1260 		inp->inp_ip.ip_ttl = ip_defttl;
1261 		break;
1262 
1263 	case PRU_DETACH:
1264 		in_pcbdetach(inp);
1265 		break;
1266 
1267 	case PRU_BIND:
1268 		error = in_pcbbind(inp, nam, l);
1269 		break;
1270 
1271 	case PRU_LISTEN:
1272 		error = EOPNOTSUPP;
1273 		break;
1274 
1275 	case PRU_CONNECT:
1276 		error = in_pcbconnect(inp, nam, l);
1277 		if (error)
1278 			break;
1279 		soisconnected(so);
1280 		break;
1281 
1282 	case PRU_CONNECT2:
1283 		error = EOPNOTSUPP;
1284 		break;
1285 
1286 	case PRU_DISCONNECT:
1287 		/*soisdisconnected(so);*/
1288 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1289 		in_pcbdisconnect(inp);
1290 		inp->inp_laddr = zeroin_addr;		/* XXX */
1291 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1292 		break;
1293 
1294 	case PRU_SHUTDOWN:
1295 		socantsendmore(so);
1296 		break;
1297 
1298 	case PRU_RCVD:
1299 		error = EOPNOTSUPP;
1300 		break;
1301 
1302 	case PRU_SEND:
1303 		if (control && control->m_len) {
1304 			m_freem(control);
1305 			m_freem(m);
1306 			error = EINVAL;
1307 			break;
1308 		}
1309 	{
1310 		struct in_addr laddr;			/* XXX */
1311 
1312 		memset(&laddr, 0, sizeof laddr);
1313 		if (nam) {
1314 			laddr = inp->inp_laddr;		/* XXX */
1315 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1316 				error = EISCONN;
1317 				goto die;
1318 			}
1319 			error = in_pcbconnect(inp, nam, l);
1320 			if (error)
1321 				goto die;
1322 		} else {
1323 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1324 				error = ENOTCONN;
1325 				goto die;
1326 			}
1327 		}
1328 		error = udp_output(m, inp);
1329 		m = NULL;
1330 		if (nam) {
1331 			in_pcbdisconnect(inp);
1332 			inp->inp_laddr = laddr;		/* XXX */
1333 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1334 		}
1335 	  die:
1336 		if (m)
1337 			m_freem(m);
1338 	}
1339 		break;
1340 
1341 	case PRU_SENSE:
1342 		/*
1343 		 * stat: don't bother with a blocksize.
1344 		 */
1345 		splx(s);
1346 		return (0);
1347 
1348 	case PRU_RCVOOB:
1349 		error =  EOPNOTSUPP;
1350 		break;
1351 
1352 	case PRU_SENDOOB:
1353 		m_freem(control);
1354 		m_freem(m);
1355 		error =  EOPNOTSUPP;
1356 		break;
1357 
1358 	case PRU_SOCKADDR:
1359 		in_setsockaddr(inp, nam);
1360 		break;
1361 
1362 	case PRU_PEERADDR:
1363 		in_setpeeraddr(inp, nam);
1364 		break;
1365 
1366 	default:
1367 		panic("udp_usrreq");
1368 	}
1369 
1370 release:
1371 	splx(s);
1372 	return (error);
1373 }
1374 
1375 static int
1376 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1377 {
1378 
1379 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1380 }
1381 
1382 /*
1383  * Sysctl for udp variables.
1384  */
1385 static void
1386 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1387 {
1388 	sysctl_createv(clog, 0, NULL, NULL,
1389 		       CTLFLAG_PERMANENT,
1390 		       CTLTYPE_NODE, "net", NULL,
1391 		       NULL, 0, NULL, 0,
1392 		       CTL_NET, CTL_EOL);
1393 	sysctl_createv(clog, 0, NULL, NULL,
1394 		       CTLFLAG_PERMANENT,
1395 		       CTLTYPE_NODE, "inet", NULL,
1396 		       NULL, 0, NULL, 0,
1397 		       CTL_NET, PF_INET, CTL_EOL);
1398 	sysctl_createv(clog, 0, NULL, NULL,
1399 		       CTLFLAG_PERMANENT,
1400 		       CTLTYPE_NODE, "udp",
1401 		       SYSCTL_DESCR("UDPv4 related settings"),
1402 		       NULL, 0, NULL, 0,
1403 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1404 
1405 	sysctl_createv(clog, 0, NULL, NULL,
1406 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1407 		       CTLTYPE_INT, "checksum",
1408 		       SYSCTL_DESCR("Compute UDP checksums"),
1409 		       NULL, 0, &udpcksum, 0,
1410 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1411 		       CTL_EOL);
1412 	sysctl_createv(clog, 0, NULL, NULL,
1413 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1414 		       CTLTYPE_INT, "sendspace",
1415 		       SYSCTL_DESCR("Default UDP send buffer size"),
1416 		       NULL, 0, &udp_sendspace, 0,
1417 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1418 		       CTL_EOL);
1419 	sysctl_createv(clog, 0, NULL, NULL,
1420 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1421 		       CTLTYPE_INT, "recvspace",
1422 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1423 		       NULL, 0, &udp_recvspace, 0,
1424 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1425 		       CTL_EOL);
1426 	sysctl_createv(clog, 0, NULL, NULL,
1427 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1428 		       CTLTYPE_INT, "do_loopback_cksum",
1429 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1430 		       NULL, 0, &udp_do_loopback_cksum, 0,
1431 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1432 		       CTL_EOL);
1433 	sysctl_createv(clog, 0, NULL, NULL,
1434 		       CTLFLAG_PERMANENT,
1435 		       CTLTYPE_STRUCT, "pcblist",
1436 		       SYSCTL_DESCR("UDP protocol control block list"),
1437 		       sysctl_inpcblist, 0, &udbtable, 0,
1438 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1439 		       CTL_EOL);
1440 	sysctl_createv(clog, 0, NULL, NULL,
1441 		       CTLFLAG_PERMANENT,
1442 		       CTLTYPE_STRUCT, "stats",
1443 		       SYSCTL_DESCR("UDP statistics"),
1444 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1445 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1446 		       CTL_EOL);
1447 }
1448 #endif
1449 
1450 void
1451 udp_statinc(u_int stat)
1452 {
1453 
1454 	KASSERT(stat < UDP_NSTATS);
1455 	UDP_STATINC(stat);
1456 }
1457 
1458 #if defined(INET) && defined(IPSEC)
1459 /*
1460  * Returns:
1461  * 1 if the packet was processed
1462  * 0 if normal UDP processing should take place
1463  * -1 if an error occurent and m was freed
1464  */
1465 static int
1466 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1467     struct socket *so)
1468 {
1469 	size_t len;
1470 	void *data;
1471 	struct inpcb *inp;
1472 	size_t skip = 0;
1473 	size_t minlen;
1474 	size_t iphdrlen;
1475 	struct ip *ip;
1476 	struct m_tag *tag;
1477 	struct udphdr *udphdr;
1478 	u_int16_t sport, dport;
1479 	struct mbuf *m = *mp;
1480 
1481 	/*
1482 	 * Collapse the mbuf chain if the first mbuf is too short
1483 	 * The longest case is: UDP + non ESP marker + ESP
1484 	 */
1485 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1486 	if (minlen > m->m_pkthdr.len)
1487 		minlen = m->m_pkthdr.len;
1488 
1489 	if (m->m_len < minlen) {
1490 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1491 			printf("udp4_espinudp: m_pullup failed\n");
1492 			return -1;
1493 		}
1494 		m = *mp;
1495 	}
1496 
1497 	len = m->m_len - off;
1498 	data = mtod(m, char *) + off;
1499 	inp = sotoinpcb(so);
1500 
1501 	/* Ignore keepalive packets */
1502 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1503 		m_free(m);
1504 		*mp = NULL; /* avoid any further processiong by caller ... */
1505 		return 1;
1506 	}
1507 
1508 	/*
1509 	 * Check that the payload is long enough to hold
1510 	 * an ESP header and compute the length of encapsulation
1511 	 * header to remove
1512 	 */
1513 	if (inp->inp_flags & INP_ESPINUDP) {
1514 		u_int32_t *st = (u_int32_t *)data;
1515 
1516 		if ((len <= sizeof(struct esp)) || (*st == 0))
1517 			return 0; /* Normal UDP processing */
1518 
1519 		skip = sizeof(struct udphdr);
1520 	}
1521 
1522 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1523 		u_int32_t *st = (u_int32_t *)data;
1524 
1525 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1526 		    || ((st[0] | st[1]) != 0))
1527 			return 0; /* Normal UDP processing */
1528 
1529 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1530 	}
1531 
1532 	/*
1533 	 * Get the UDP ports. They are handled in network
1534 	 * order everywhere in IPSEC_NAT_T code.
1535 	 */
1536 	udphdr = (struct udphdr *)((char *)data - skip);
1537 	sport = udphdr->uh_sport;
1538 	dport = udphdr->uh_dport;
1539 
1540 	/*
1541 	 * Remove the UDP header (and possibly the non ESP marker)
1542 	 * IP header lendth is iphdrlen
1543 	 * Before:
1544 	 *   <--- off --->
1545 	 *   +----+------+-----+
1546 	 *   | IP |  UDP | ESP |
1547 	 *   +----+------+-----+
1548 	 *        <-skip->
1549 	 * After:
1550 	 *          +----+-----+
1551 	 *          | IP | ESP |
1552 	 *          +----+-----+
1553 	 *   <-skip->
1554 	 */
1555 	iphdrlen = off - sizeof(struct udphdr);
1556 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1557 	m_adj(m, skip);
1558 
1559 	ip = mtod(m, struct ip *);
1560 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1561 	ip->ip_p = IPPROTO_ESP;
1562 
1563 	/*
1564 	 * We have modified the packet - it is now ESP, so we should not
1565 	 * return to UDP processing ...
1566 	 *
1567 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1568 	 * the source UDP port. This is required if we want
1569 	 * to select the right SPD for multiple hosts behind
1570 	 * same NAT
1571 	 */
1572 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1573 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1574 		printf("udp4_espinudp: m_tag_get failed\n");
1575 		m_freem(m);
1576 		return -1;
1577 	}
1578 	((u_int16_t *)(tag + 1))[0] = sport;
1579 	((u_int16_t *)(tag + 1))[1] = dport;
1580 	m_tag_prepend(m, tag);
1581 
1582 #ifdef IPSEC
1583 	ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1584 #else
1585 	esp4_input(m, iphdrlen);
1586 #endif
1587 
1588 	/* We handled it, it shouldn't be handled by UDP */
1589 	*mp = NULL; /* avoid free by caller ... */
1590 	return 1;
1591 }
1592 #endif
1593