1 /*	$NetBSD: udp_usrreq.c,v 1.226 2016/06/10 13:31:44 ozaki-r 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 /*
64  * UDP protocol implementation.
65  * Per RFC 768, August, 1980.
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.226 2016/06/10 13:31:44 ozaki-r Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_inet.h"
73 #include "opt_compat_netbsd.h"
74 #include "opt_ipsec.h"
75 #include "opt_inet_csum.h"
76 #include "opt_ipkdb.h"
77 #include "opt_mbuftrace.h"
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/mbuf.h>
82 #include <sys/once.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/systm.h>
87 #include <sys/proc.h>
88 #include <sys/domain.h>
89 #include <sys/sysctl.h>
90 
91 #include <net/if.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/udp.h>
101 #include <netinet/udp_var.h>
102 #include <netinet/udp_private.h>
103 
104 #ifdef INET6
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/ip6_private.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/udp6_var.h>
111 #include <netinet6/udp6_private.h>
112 #endif
113 
114 #ifndef INET6
115 /* always need ip6.h for IP6_EXTHDR_GET */
116 #include <netinet/ip6.h>
117 #endif
118 
119 #ifdef IPSEC
120 #include <netipsec/ipsec.h>
121 #include <netipsec/ipsec_var.h>
122 #include <netipsec/ipsec_private.h>
123 #include <netipsec/esp.h>
124 #ifdef INET6
125 #include <netipsec/ipsec6.h>
126 #endif
127 #endif	/* IPSEC */
128 
129 #ifdef COMPAT_50
130 #include <compat/sys/socket.h>
131 #endif
132 
133 #ifdef IPKDB
134 #include <ipkdb/ipkdb.h>
135 #endif
136 
137 int	udpcksum = 1;
138 int	udp_do_loopback_cksum = 0;
139 
140 struct	inpcbtable udbtable;
141 
142 percpu_t *udpstat_percpu;
143 
144 #ifdef INET
145 #ifdef IPSEC
146 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
147 	struct socket *);
148 #endif
149 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
150 	struct socket *);
151 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
152 	struct mbuf **, int);
153 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
154 #endif
155 #ifdef INET
156 static	void udp_notify (struct inpcb *, int);
157 #endif
158 
159 #ifndef UDBHASHSIZE
160 #define	UDBHASHSIZE	128
161 #endif
162 int	udbhashsize = UDBHASHSIZE;
163 
164 /*
165  * For send - really max datagram size; for receive - 40 1K datagrams.
166  */
167 static int	udp_sendspace = 9216;
168 static int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
169 
170 #ifdef MBUFTRACE
171 struct mowner udp_mowner = MOWNER_INIT("udp", "");
172 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
173 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
174 #endif
175 
176 #ifdef UDP_CSUM_COUNTERS
177 #include <sys/device.h>
178 
179 #if defined(INET)
180 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
181     NULL, "udp", "hwcsum bad");
182 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
183     NULL, "udp", "hwcsum ok");
184 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
185     NULL, "udp", "hwcsum data");
186 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
187     NULL, "udp", "swcsum");
188 
189 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
190 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
191 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
192 EVCNT_ATTACH_STATIC(udp_swcsum);
193 #endif /* defined(INET) */
194 
195 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
196 #else
197 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
198 #endif /* UDP_CSUM_COUNTERS */
199 
200 static void sysctl_net_inet_udp_setup(struct sysctllog **);
201 
202 static int
do_udpinit(void)203 do_udpinit(void)
204 {
205 
206 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
207 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
208 
209 	MOWNER_ATTACH(&udp_tx_mowner);
210 	MOWNER_ATTACH(&udp_rx_mowner);
211 	MOWNER_ATTACH(&udp_mowner);
212 
213 	return 0;
214 }
215 
216 void
udp_init_common(void)217 udp_init_common(void)
218 {
219 	static ONCE_DECL(doudpinit);
220 
221 	RUN_ONCE(&doudpinit, do_udpinit);
222 }
223 
224 void
udp_init(void)225 udp_init(void)
226 {
227 
228 	sysctl_net_inet_udp_setup(NULL);
229 
230 	udp_init_common();
231 }
232 
233 /*
234  * Checksum extended UDP header and data.
235  */
236 
237 int
udp_input_checksum(int af,struct mbuf * m,const struct udphdr * uh,int iphlen,int len)238 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
239     int iphlen, int len)
240 {
241 
242 	switch (af) {
243 #ifdef INET
244 	case AF_INET:
245 		return udp4_input_checksum(m, uh, iphlen, len);
246 #endif
247 #ifdef INET6
248 	case AF_INET6:
249 		return udp6_input_checksum(m, uh, iphlen, len);
250 #endif
251 	}
252 #ifdef DIAGNOSTIC
253 	panic("udp_input_checksum: unknown af %d", af);
254 #endif
255 	/* NOTREACHED */
256 	return -1;
257 }
258 
259 #ifdef INET
260 
261 /*
262  * Checksum extended UDP header and data.
263  */
264 
265 static int
udp4_input_checksum(struct mbuf * m,const struct udphdr * uh,int iphlen,int len)266 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
267     int iphlen, int len)
268 {
269 
270 	/*
271 	 * XXX it's better to record and check if this mbuf is
272 	 * already checked.
273 	 */
274 
275 	if (uh->uh_sum == 0)
276 		return 0;
277 
278 	switch (m->m_pkthdr.csum_flags &
279 	    ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
280 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
281 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
282 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
283 		goto badcsum;
284 
285 	case M_CSUM_UDPv4|M_CSUM_DATA: {
286 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
287 
288 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
289 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
290 			const struct ip *ip =
291 			    mtod(m, const struct ip *);
292 
293 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
294 			    ip->ip_dst.s_addr,
295 			    htons(hw_csum + len + IPPROTO_UDP));
296 		}
297 		if ((hw_csum ^ 0xffff) != 0)
298 			goto badcsum;
299 		break;
300 	}
301 
302 	case M_CSUM_UDPv4:
303 		/* Checksum was okay. */
304 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
305 		break;
306 
307 	default:
308 		/*
309 		 * Need to compute it ourselves.  Maybe skip checksum
310 		 * on loopback interfaces.
311 		 */
312 		if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
313 				     IFF_LOOPBACK) ||
314 				   udp_do_loopback_cksum)) {
315 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
316 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
317 				goto badcsum;
318 		}
319 		break;
320 	}
321 
322 	return 0;
323 
324 badcsum:
325 	UDP_STATINC(UDP_STAT_BADSUM);
326 	return -1;
327 }
328 
329 void
udp_input(struct mbuf * m,...)330 udp_input(struct mbuf *m, ...)
331 {
332 	va_list ap;
333 	struct sockaddr_in src, dst;
334 	struct ip *ip;
335 	struct udphdr *uh;
336 	int iphlen;
337 	int len;
338 	int n;
339 	u_int16_t ip_len;
340 
341 	va_start(ap, m);
342 	iphlen = va_arg(ap, int);
343 	(void)va_arg(ap, int);		/* ignore value, advance ap */
344 	va_end(ap);
345 
346 	MCLAIM(m, &udp_rx_mowner);
347 	UDP_STATINC(UDP_STAT_IPACKETS);
348 
349 	/*
350 	 * Get IP and UDP header together in first mbuf.
351 	 */
352 	ip = mtod(m, struct ip *);
353 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
354 	if (uh == NULL) {
355 		UDP_STATINC(UDP_STAT_HDROPS);
356 		return;
357 	}
358 	KASSERT(UDP_HDR_ALIGNED_P(uh));
359 
360 	/* destination port of 0 is illegal, based on RFC768. */
361 	if (uh->uh_dport == 0)
362 		goto bad;
363 
364 	/*
365 	 * Make mbuf data length reflect UDP length.
366 	 * If not enough data to reflect UDP length, drop.
367 	 */
368 	ip_len = ntohs(ip->ip_len);
369 	len = ntohs((u_int16_t)uh->uh_ulen);
370 	if (ip_len != iphlen + len) {
371 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
372 			UDP_STATINC(UDP_STAT_BADLEN);
373 			goto bad;
374 		}
375 		m_adj(m, iphlen + len - ip_len);
376 	}
377 
378 	/*
379 	 * Checksum extended UDP header and data.
380 	 */
381 	if (udp4_input_checksum(m, uh, iphlen, len))
382 		goto badcsum;
383 
384 	/* construct source and dst sockaddrs. */
385 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
386 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
387 
388 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
389 		UDP_STATINC(UDP_STAT_HDROPS);
390 		return;
391 	}
392 	if (m == NULL) {
393 		/*
394 		 * packet has been processed by ESP stuff -
395 		 * e.g. dropped NAT-T-keep-alive-packet ...
396 		 */
397 		return;
398 	}
399 	ip = mtod(m, struct ip *);
400 #ifdef INET6
401 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
402 		struct sockaddr_in6 src6, dst6;
403 
404 		memset(&src6, 0, sizeof(src6));
405 		src6.sin6_family = AF_INET6;
406 		src6.sin6_len = sizeof(struct sockaddr_in6);
407 		in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
408 		src6.sin6_port = uh->uh_sport;
409 		memset(&dst6, 0, sizeof(dst6));
410 		dst6.sin6_family = AF_INET6;
411 		dst6.sin6_len = sizeof(struct sockaddr_in6);
412 		in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
413 		dst6.sin6_port = uh->uh_dport;
414 
415 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
416 	}
417 #endif
418 
419 	if (n == 0) {
420 		if (m->m_flags & (M_BCAST | M_MCAST)) {
421 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
422 			goto bad;
423 		}
424 		UDP_STATINC(UDP_STAT_NOPORT);
425 #ifdef IPKDB
426 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
427 				m, iphlen + sizeof(struct udphdr),
428 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
429 			/*
430 			 * It was a debugger connect packet,
431 			 * just drop it now
432 			 */
433 			goto bad;
434 		}
435 #endif
436 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
437 		m = NULL;
438 	}
439 
440 bad:
441 	if (m)
442 		m_freem(m);
443 	return;
444 
445 badcsum:
446 	m_freem(m);
447 }
448 #endif
449 
450 #ifdef INET
451 static void
udp4_sendup(struct mbuf * m,int off,struct sockaddr * src,struct socket * so)452 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
453 	struct sockaddr *src, struct socket *so)
454 {
455 	struct mbuf *opts = NULL;
456 	struct mbuf *n;
457 	struct inpcb *inp = NULL;
458 
459 	if (!so)
460 		return;
461 	switch (so->so_proto->pr_domain->dom_family) {
462 	case AF_INET:
463 		inp = sotoinpcb(so);
464 		break;
465 #ifdef INET6
466 	case AF_INET6:
467 		break;
468 #endif
469 	default:
470 		return;
471 	}
472 
473 #if defined(IPSEC)
474 	/* check AH/ESP integrity. */
475 	if (ipsec_used && so != NULL && ipsec4_in_reject_so(m, so)) {
476 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
477 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
478 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
479 			    0, 0);
480 		return;
481 	}
482 #endif /*IPSEC*/
483 
484 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
485 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
486 #ifdef SO_OTIMESTAMP
487 			 || so->so_options & SO_OTIMESTAMP
488 #endif
489 			 || so->so_options & SO_TIMESTAMP)) {
490 			struct ip *ip = mtod(n, struct ip *);
491 			ip_savecontrol(inp, &opts, ip, n);
492 		}
493 
494 		m_adj(n, off);
495 		if (sbappendaddr(&so->so_rcv, src, n,
496 				opts) == 0) {
497 			m_freem(n);
498 			if (opts)
499 				m_freem(opts);
500 			so->so_rcv.sb_overflowed++;
501 			UDP_STATINC(UDP_STAT_FULLSOCK);
502 		} else
503 			sorwakeup(so);
504 	}
505 }
506 #endif
507 
508 #ifdef INET
509 static int
udp4_realinput(struct sockaddr_in * src,struct sockaddr_in * dst,struct mbuf ** mp,int off)510 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
511 	struct mbuf **mp, int off /* offset of udphdr */)
512 {
513 	u_int16_t *sport, *dport;
514 	int rcvcnt;
515 	struct in_addr *src4, *dst4;
516 	struct inpcb_hdr *inph;
517 	struct inpcb *inp;
518 	struct mbuf *m = *mp;
519 
520 	rcvcnt = 0;
521 	off += sizeof(struct udphdr);	/* now, offset of payload */
522 
523 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
524 		goto bad;
525 
526 	src4 = &src->sin_addr;
527 	sport = &src->sin_port;
528 	dst4 = &dst->sin_addr;
529 	dport = &dst->sin_port;
530 
531 	if (IN_MULTICAST(dst4->s_addr) ||
532 	    in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
533 		/*
534 		 * Deliver a multicast or broadcast datagram to *all* sockets
535 		 * for which the local and remote addresses and ports match
536 		 * those of the incoming datagram.  This allows more than
537 		 * one process to receive multi/broadcasts on the same port.
538 		 * (This really ought to be done for unicast datagrams as
539 		 * well, but that would cause problems with existing
540 		 * applications that open both address-specific sockets and
541 		 * a wildcard socket listening to the same port -- they would
542 		 * end up receiving duplicates of every unicast datagram.
543 		 * Those applications open the multiple sockets to overcome an
544 		 * inadequacy of the UDP socket interface, but for backwards
545 		 * compatibility we avoid the problem here rather than
546 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
547 		 */
548 
549 		/*
550 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
551 		 * we need udpiphdr for IPsec processing so we do that later.
552 		 */
553 		/*
554 		 * Locate pcb(s) for datagram.
555 		 */
556 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
557 			inp = (struct inpcb *)inph;
558 			if (inp->inp_af != AF_INET)
559 				continue;
560 
561 			if (inp->inp_lport != *dport)
562 				continue;
563 			if (!in_nullhost(inp->inp_laddr)) {
564 				if (!in_hosteq(inp->inp_laddr, *dst4))
565 					continue;
566 			}
567 			if (!in_nullhost(inp->inp_faddr)) {
568 				if (!in_hosteq(inp->inp_faddr, *src4) ||
569 				    inp->inp_fport != *sport)
570 					continue;
571 			}
572 
573 			udp4_sendup(m, off, (struct sockaddr *)src,
574 				inp->inp_socket);
575 			rcvcnt++;
576 
577 			/*
578 			 * Don't look for additional matches if this one does
579 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
580 			 * socket options set.  This heuristic avoids searching
581 			 * through all pcbs in the common case of a non-shared
582 			 * port.  It assumes that an application will never
583 			 * clear these options after setting them.
584 			 */
585 			if ((inp->inp_socket->so_options &
586 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
587 				break;
588 		}
589 	} else {
590 		/*
591 		 * Locate pcb for datagram.
592 		 */
593 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
594 		    *dport, 0);
595 		if (inp == 0) {
596 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
597 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
598 			if (inp == 0)
599 				return rcvcnt;
600 		}
601 
602 #ifdef IPSEC
603 		/* Handle ESP over UDP */
604 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
605 			struct sockaddr *sa = (struct sockaddr *)src;
606 
607 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
608 			case -1: 	/* Error, m was freeed */
609 				rcvcnt = -1;
610 				goto bad;
611 				break;
612 
613 			case 1:		/* ESP over UDP */
614 				rcvcnt++;
615 				goto bad;
616 				break;
617 
618 			case 0: 	/* plain UDP */
619 			default: 	/* Unexpected */
620 				/*
621 				 * Normal UDP processing will take place
622 				 * m may have changed.
623 				 */
624 				m = *mp;
625 				break;
626 			}
627 		}
628 #endif
629 
630 		/*
631 		 * Check the minimum TTL for socket.
632 		 */
633 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
634 			goto bad;
635 
636 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
637 		rcvcnt++;
638 	}
639 
640 bad:
641 	return rcvcnt;
642 }
643 #endif
644 
645 #ifdef INET
646 /*
647  * Notify a udp user of an asynchronous error;
648  * just wake up so that he can collect error status.
649  */
650 static void
udp_notify(struct inpcb * inp,int errno)651 udp_notify(struct inpcb *inp, int errno)
652 {
653 	inp->inp_socket->so_error = errno;
654 	sorwakeup(inp->inp_socket);
655 	sowwakeup(inp->inp_socket);
656 }
657 
658 void *
udp_ctlinput(int cmd,const struct sockaddr * sa,void * v)659 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
660 {
661 	struct ip *ip = v;
662 	struct udphdr *uh;
663 	void (*notify)(struct inpcb *, int) = udp_notify;
664 	int errno;
665 
666 	if (sa->sa_family != AF_INET
667 	 || sa->sa_len != sizeof(struct sockaddr_in))
668 		return NULL;
669 	if ((unsigned)cmd >= PRC_NCMDS)
670 		return NULL;
671 	errno = inetctlerrmap[cmd];
672 	if (PRC_IS_REDIRECT(cmd))
673 		notify = in_rtchange, ip = 0;
674 	else if (cmd == PRC_HOSTDEAD)
675 		ip = 0;
676 	else if (errno == 0)
677 		return NULL;
678 	if (ip) {
679 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
680 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
681 		    ip->ip_src, uh->uh_sport, errno, notify);
682 
683 		/* XXX mapped address case */
684 	} else
685 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
686 		    notify);
687 	return NULL;
688 }
689 
690 int
udp_ctloutput(int op,struct socket * so,struct sockopt * sopt)691 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
692 {
693 	int s;
694 	int error = 0;
695 	struct inpcb *inp;
696 	int family;
697 	int optval;
698 
699 	family = so->so_proto->pr_domain->dom_family;
700 
701 	s = splsoftnet();
702 	switch (family) {
703 #ifdef INET
704 	case PF_INET:
705 		if (sopt->sopt_level != IPPROTO_UDP) {
706 			error = ip_ctloutput(op, so, sopt);
707 			goto end;
708 		}
709 		break;
710 #endif
711 #ifdef INET6
712 	case PF_INET6:
713 		if (sopt->sopt_level != IPPROTO_UDP) {
714 			error = ip6_ctloutput(op, so, sopt);
715 			goto end;
716 		}
717 		break;
718 #endif
719 	default:
720 		error = EAFNOSUPPORT;
721 		goto end;
722 	}
723 
724 
725 	switch (op) {
726 	case PRCO_SETOPT:
727 		inp = sotoinpcb(so);
728 
729 		switch (sopt->sopt_name) {
730 		case UDP_ENCAP:
731 			error = sockopt_getint(sopt, &optval);
732 			if (error)
733 				break;
734 
735 			switch(optval) {
736 			case 0:
737 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
738 				break;
739 
740 			case UDP_ENCAP_ESPINUDP:
741 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
742 				inp->inp_flags |= INP_ESPINUDP;
743 				break;
744 
745 			case UDP_ENCAP_ESPINUDP_NON_IKE:
746 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
747 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
748 				break;
749 			default:
750 				error = EINVAL;
751 				break;
752 			}
753 			break;
754 
755 		default:
756 			error = ENOPROTOOPT;
757 			break;
758 		}
759 		break;
760 
761 	default:
762 		error = EINVAL;
763 		break;
764 	}
765 
766 end:
767 	splx(s);
768 	return error;
769 }
770 
771 
772 int
udp_output(struct mbuf * m,struct inpcb * inp)773 udp_output(struct mbuf *m, struct inpcb *inp)
774 {
775 	struct udpiphdr *ui;
776 	struct route *ro;
777 	int len = m->m_pkthdr.len;
778 	int error = 0;
779 
780 	MCLAIM(m, &udp_tx_mowner);
781 
782 	/*
783 	 * Calculate data length and get a mbuf
784 	 * for UDP and IP headers.
785 	 */
786 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
787 	if (m == 0) {
788 		error = ENOBUFS;
789 		goto release;
790 	}
791 
792 	/*
793 	 * Compute the packet length of the IP header, and
794 	 * punt if the length looks bogus.
795 	 */
796 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
797 		error = EMSGSIZE;
798 		goto release;
799 	}
800 
801 	/*
802 	 * Fill in mbuf with extended UDP header
803 	 * and addresses and length put into network format.
804 	 */
805 	ui = mtod(m, struct udpiphdr *);
806 	ui->ui_pr = IPPROTO_UDP;
807 	ui->ui_src = inp->inp_laddr;
808 	ui->ui_dst = inp->inp_faddr;
809 	ui->ui_sport = inp->inp_lport;
810 	ui->ui_dport = inp->inp_fport;
811 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
812 
813 	ro = &inp->inp_route;
814 
815 	/*
816 	 * Set up checksum and output datagram.
817 	 */
818 	if (udpcksum) {
819 		/*
820 		 * XXX Cache pseudo-header checksum part for
821 		 * XXX "connected" UDP sockets.
822 		 */
823 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
824 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
825 		    sizeof(struct udphdr) + IPPROTO_UDP));
826 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
827 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
828 	} else
829 		ui->ui_sum = 0;
830 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
831 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
832 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
833 	UDP_STATINC(UDP_STAT_OPACKETS);
834 
835 	return (ip_output(m, inp->inp_options, ro,
836 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
837 	    inp->inp_moptions, inp->inp_socket));
838 
839 release:
840 	m_freem(m);
841 	return (error);
842 }
843 
844 static int
udp_attach(struct socket * so,int proto)845 udp_attach(struct socket *so, int proto)
846 {
847 	struct inpcb *inp;
848 	int error;
849 
850 	KASSERT(sotoinpcb(so) == NULL);
851 
852 	/* Assign the lock (must happen even if we will error out). */
853 	sosetlock(so);
854 
855 #ifdef MBUFTRACE
856 	so->so_mowner = &udp_mowner;
857 	so->so_rcv.sb_mowner = &udp_rx_mowner;
858 	so->so_snd.sb_mowner = &udp_tx_mowner;
859 #endif
860 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
861 		error = soreserve(so, udp_sendspace, udp_recvspace);
862 		if (error) {
863 			return error;
864 		}
865 	}
866 
867 	error = in_pcballoc(so, &udbtable);
868 	if (error) {
869 		return error;
870 	}
871 	inp = sotoinpcb(so);
872 	inp->inp_ip.ip_ttl = ip_defttl;
873 	KASSERT(solocked(so));
874 
875 	return error;
876 }
877 
878 static void
udp_detach(struct socket * so)879 udp_detach(struct socket *so)
880 {
881 	struct inpcb *inp;
882 
883 	KASSERT(solocked(so));
884 	inp = sotoinpcb(so);
885 	KASSERT(inp != NULL);
886 	in_pcbdetach(inp);
887 }
888 
889 static int
udp_accept(struct socket * so,struct sockaddr * nam)890 udp_accept(struct socket *so, struct sockaddr *nam)
891 {
892 	KASSERT(solocked(so));
893 
894 	panic("udp_accept");
895 
896 	return EOPNOTSUPP;
897 }
898 
899 static int
udp_bind(struct socket * so,struct sockaddr * nam,struct lwp * l)900 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
901 {
902 	struct inpcb *inp = sotoinpcb(so);
903 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
904 	int error = 0;
905 	int s;
906 
907 	KASSERT(solocked(so));
908 	KASSERT(inp != NULL);
909 	KASSERT(nam != NULL);
910 
911 	s = splsoftnet();
912 	error = in_pcbbind(inp, sin, l);
913 	splx(s);
914 
915 	return error;
916 }
917 
918 static int
udp_listen(struct socket * so,struct lwp * l)919 udp_listen(struct socket *so, struct lwp *l)
920 {
921 	KASSERT(solocked(so));
922 
923 	return EOPNOTSUPP;
924 }
925 
926 static int
udp_connect(struct socket * so,struct sockaddr * nam,struct lwp * l)927 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
928 {
929 	struct inpcb *inp = sotoinpcb(so);
930 	int error = 0;
931 	int s;
932 
933 	KASSERT(solocked(so));
934 	KASSERT(inp != NULL);
935 	KASSERT(nam != NULL);
936 
937 	s = splsoftnet();
938 	error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
939 	if (! error)
940 		soisconnected(so);
941 	splx(s);
942 	return error;
943 }
944 
945 static int
udp_connect2(struct socket * so,struct socket * so2)946 udp_connect2(struct socket *so, struct socket *so2)
947 {
948 	KASSERT(solocked(so));
949 
950 	return EOPNOTSUPP;
951 }
952 
953 static int
udp_disconnect(struct socket * so)954 udp_disconnect(struct socket *so)
955 {
956 	struct inpcb *inp = sotoinpcb(so);
957 	int s;
958 
959 	KASSERT(solocked(so));
960 	KASSERT(inp != NULL);
961 
962 	s = splsoftnet();
963 	/*soisdisconnected(so);*/
964 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
965 	in_pcbdisconnect(inp);
966 	inp->inp_laddr = zeroin_addr;		/* XXX */
967 	in_pcbstate(inp, INP_BOUND);		/* XXX */
968 	splx(s);
969 
970 	return 0;
971 }
972 
973 static int
udp_shutdown(struct socket * so)974 udp_shutdown(struct socket *so)
975 {
976 	int s;
977 
978 	KASSERT(solocked(so));
979 
980 	s = splsoftnet();
981 	socantsendmore(so);
982 	splx(s);
983 
984 	return 0;
985 }
986 
987 static int
udp_abort(struct socket * so)988 udp_abort(struct socket *so)
989 {
990 	KASSERT(solocked(so));
991 
992 	panic("udp_abort");
993 
994 	return EOPNOTSUPP;
995 }
996 
997 static int
udp_ioctl(struct socket * so,u_long cmd,void * nam,struct ifnet * ifp)998 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
999 {
1000 	return in_control(so, cmd, nam, ifp);
1001 }
1002 
1003 static int
udp_stat(struct socket * so,struct stat * ub)1004 udp_stat(struct socket *so, struct stat *ub)
1005 {
1006 	KASSERT(solocked(so));
1007 
1008 	/* stat: don't bother with a blocksize. */
1009 	return 0;
1010 }
1011 
1012 static int
udp_peeraddr(struct socket * so,struct sockaddr * nam)1013 udp_peeraddr(struct socket *so, struct sockaddr *nam)
1014 {
1015 	int s;
1016 
1017 	KASSERT(solocked(so));
1018 	KASSERT(sotoinpcb(so) != NULL);
1019 	KASSERT(nam != NULL);
1020 
1021 	s = splsoftnet();
1022 	in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1023 	splx(s);
1024 
1025 	return 0;
1026 }
1027 
1028 static int
udp_sockaddr(struct socket * so,struct sockaddr * nam)1029 udp_sockaddr(struct socket *so, struct sockaddr *nam)
1030 {
1031 	int s;
1032 
1033 	KASSERT(solocked(so));
1034 	KASSERT(sotoinpcb(so) != NULL);
1035 	KASSERT(nam != NULL);
1036 
1037 	s = splsoftnet();
1038 	in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
1039 	splx(s);
1040 
1041 	return 0;
1042 }
1043 
1044 static int
udp_rcvd(struct socket * so,int flags,struct lwp * l)1045 udp_rcvd(struct socket *so, int flags, struct lwp *l)
1046 {
1047 	KASSERT(solocked(so));
1048 
1049 	return EOPNOTSUPP;
1050 }
1051 
1052 static int
udp_recvoob(struct socket * so,struct mbuf * m,int flags)1053 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
1054 {
1055 	KASSERT(solocked(so));
1056 
1057 	return EOPNOTSUPP;
1058 }
1059 
1060 static int
udp_send(struct socket * so,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct lwp * l)1061 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1062     struct mbuf *control, struct lwp *l)
1063 {
1064 	struct inpcb *inp = sotoinpcb(so);
1065 	int error = 0;
1066 	struct in_addr laddr;			/* XXX */
1067 	int s;
1068 
1069 	KASSERT(solocked(so));
1070 	KASSERT(inp != NULL);
1071 	KASSERT(m != NULL);
1072 
1073 	if (control && control->m_len) {
1074 		m_freem(control);
1075 		m_freem(m);
1076 		return EINVAL;
1077 	}
1078 
1079 	memset(&laddr, 0, sizeof laddr);
1080 
1081 	s = splsoftnet();
1082 	if (nam) {
1083 		laddr = inp->inp_laddr;		/* XXX */
1084 		if ((so->so_state & SS_ISCONNECTED) != 0) {
1085 			error = EISCONN;
1086 			goto die;
1087 		}
1088 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
1089 		if (error)
1090 			goto die;
1091 	} else {
1092 		if ((so->so_state & SS_ISCONNECTED) == 0) {
1093 			error = ENOTCONN;
1094 			goto die;
1095 		}
1096 	}
1097 	error = udp_output(m, inp);
1098 	m = NULL;
1099 	if (nam) {
1100 		in_pcbdisconnect(inp);
1101 		inp->inp_laddr = laddr;		/* XXX */
1102 		in_pcbstate(inp, INP_BOUND);	/* XXX */
1103 	}
1104   die:
1105 	if (m)
1106 		m_freem(m);
1107 
1108 	splx(s);
1109 	return error;
1110 }
1111 
1112 static int
udp_sendoob(struct socket * so,struct mbuf * m,struct mbuf * control)1113 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1114 {
1115 	KASSERT(solocked(so));
1116 
1117 	m_freem(m);
1118 	m_freem(control);
1119 
1120 	return EOPNOTSUPP;
1121 }
1122 
1123 static int
udp_purgeif(struct socket * so,struct ifnet * ifp)1124 udp_purgeif(struct socket *so, struct ifnet *ifp)
1125 {
1126 	int s;
1127 
1128 	s = splsoftnet();
1129 	mutex_enter(softnet_lock);
1130 	in_pcbpurgeif0(&udbtable, ifp);
1131 	in_purgeif(ifp);
1132 	in_pcbpurgeif(&udbtable, ifp);
1133 	mutex_exit(softnet_lock);
1134 	splx(s);
1135 
1136 	return 0;
1137 }
1138 
1139 static int
sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)1140 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1141 {
1142 
1143 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
1144 }
1145 
1146 /*
1147  * Sysctl for udp variables.
1148  */
1149 static void
sysctl_net_inet_udp_setup(struct sysctllog ** clog)1150 sysctl_net_inet_udp_setup(struct sysctllog **clog)
1151 {
1152 
1153 	sysctl_createv(clog, 0, NULL, NULL,
1154 		       CTLFLAG_PERMANENT,
1155 		       CTLTYPE_NODE, "inet", NULL,
1156 		       NULL, 0, NULL, 0,
1157 		       CTL_NET, PF_INET, CTL_EOL);
1158 	sysctl_createv(clog, 0, NULL, NULL,
1159 		       CTLFLAG_PERMANENT,
1160 		       CTLTYPE_NODE, "udp",
1161 		       SYSCTL_DESCR("UDPv4 related settings"),
1162 		       NULL, 0, NULL, 0,
1163 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1164 
1165 	sysctl_createv(clog, 0, NULL, NULL,
1166 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1167 		       CTLTYPE_INT, "checksum",
1168 		       SYSCTL_DESCR("Compute UDP checksums"),
1169 		       NULL, 0, &udpcksum, 0,
1170 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1171 		       CTL_EOL);
1172 	sysctl_createv(clog, 0, NULL, NULL,
1173 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1174 		       CTLTYPE_INT, "sendspace",
1175 		       SYSCTL_DESCR("Default UDP send buffer size"),
1176 		       NULL, 0, &udp_sendspace, 0,
1177 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1178 		       CTL_EOL);
1179 	sysctl_createv(clog, 0, NULL, NULL,
1180 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1181 		       CTLTYPE_INT, "recvspace",
1182 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1183 		       NULL, 0, &udp_recvspace, 0,
1184 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1185 		       CTL_EOL);
1186 	sysctl_createv(clog, 0, NULL, NULL,
1187 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1188 		       CTLTYPE_INT, "do_loopback_cksum",
1189 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1190 		       NULL, 0, &udp_do_loopback_cksum, 0,
1191 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1192 		       CTL_EOL);
1193 	sysctl_createv(clog, 0, NULL, NULL,
1194 		       CTLFLAG_PERMANENT,
1195 		       CTLTYPE_STRUCT, "pcblist",
1196 		       SYSCTL_DESCR("UDP protocol control block list"),
1197 		       sysctl_inpcblist, 0, &udbtable, 0,
1198 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1199 		       CTL_EOL);
1200 	sysctl_createv(clog, 0, NULL, NULL,
1201 		       CTLFLAG_PERMANENT,
1202 		       CTLTYPE_STRUCT, "stats",
1203 		       SYSCTL_DESCR("UDP statistics"),
1204 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1205 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1206 		       CTL_EOL);
1207 }
1208 #endif
1209 
1210 void
udp_statinc(u_int stat)1211 udp_statinc(u_int stat)
1212 {
1213 
1214 	KASSERT(stat < UDP_NSTATS);
1215 	UDP_STATINC(stat);
1216 }
1217 
1218 #if defined(INET) && defined(IPSEC)
1219 /*
1220  * Returns:
1221  * 1 if the packet was processed
1222  * 0 if normal UDP processing should take place
1223  * -1 if an error occurent and m was freed
1224  */
1225 static int
udp4_espinudp(struct mbuf ** mp,int off,struct sockaddr * src,struct socket * so)1226 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1227     struct socket *so)
1228 {
1229 	size_t len;
1230 	void *data;
1231 	struct inpcb *inp;
1232 	size_t skip = 0;
1233 	size_t minlen;
1234 	size_t iphdrlen;
1235 	struct ip *ip;
1236 	struct m_tag *tag;
1237 	struct udphdr *udphdr;
1238 	u_int16_t sport, dport;
1239 	struct mbuf *m = *mp;
1240 
1241 	/*
1242 	 * Collapse the mbuf chain if the first mbuf is too short
1243 	 * The longest case is: UDP + non ESP marker + ESP
1244 	 */
1245 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1246 	if (minlen > m->m_pkthdr.len)
1247 		minlen = m->m_pkthdr.len;
1248 
1249 	if (m->m_len < minlen) {
1250 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1251 			printf("udp4_espinudp: m_pullup failed\n");
1252 			return -1;
1253 		}
1254 		m = *mp;
1255 	}
1256 
1257 	len = m->m_len - off;
1258 	data = mtod(m, char *) + off;
1259 	inp = sotoinpcb(so);
1260 
1261 	/* Ignore keepalive packets */
1262 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1263 		m_free(m);
1264 		*mp = NULL; /* avoid any further processiong by caller ... */
1265 		return 1;
1266 	}
1267 
1268 	/*
1269 	 * Check that the payload is long enough to hold
1270 	 * an ESP header and compute the length of encapsulation
1271 	 * header to remove
1272 	 */
1273 	if (inp->inp_flags & INP_ESPINUDP) {
1274 		u_int32_t *st = (u_int32_t *)data;
1275 
1276 		if ((len <= sizeof(struct esp)) || (*st == 0))
1277 			return 0; /* Normal UDP processing */
1278 
1279 		skip = sizeof(struct udphdr);
1280 	}
1281 
1282 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1283 		u_int32_t *st = (u_int32_t *)data;
1284 
1285 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1286 		    || ((st[0] | st[1]) != 0))
1287 			return 0; /* Normal UDP processing */
1288 
1289 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1290 	}
1291 
1292 	/*
1293 	 * Get the UDP ports. They are handled in network
1294 	 * order everywhere in IPSEC_NAT_T code.
1295 	 */
1296 	udphdr = (struct udphdr *)((char *)data - skip);
1297 	sport = udphdr->uh_sport;
1298 	dport = udphdr->uh_dport;
1299 
1300 	/*
1301 	 * Remove the UDP header (and possibly the non ESP marker)
1302 	 * IP header lendth is iphdrlen
1303 	 * Before:
1304 	 *   <--- off --->
1305 	 *   +----+------+-----+
1306 	 *   | IP |  UDP | ESP |
1307 	 *   +----+------+-----+
1308 	 *        <-skip->
1309 	 * After:
1310 	 *          +----+-----+
1311 	 *          | IP | ESP |
1312 	 *          +----+-----+
1313 	 *   <-skip->
1314 	 */
1315 	iphdrlen = off - sizeof(struct udphdr);
1316 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1317 	m_adj(m, skip);
1318 
1319 	ip = mtod(m, struct ip *);
1320 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1321 	ip->ip_p = IPPROTO_ESP;
1322 
1323 	/*
1324 	 * We have modified the packet - it is now ESP, so we should not
1325 	 * return to UDP processing ...
1326 	 *
1327 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1328 	 * the source UDP port. This is required if we want
1329 	 * to select the right SPD for multiple hosts behind
1330 	 * same NAT
1331 	 */
1332 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1333 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1334 		printf("udp4_espinudp: m_tag_get failed\n");
1335 		m_freem(m);
1336 		return -1;
1337 	}
1338 	((u_int16_t *)(tag + 1))[0] = sport;
1339 	((u_int16_t *)(tag + 1))[1] = dport;
1340 	m_tag_prepend(m, tag);
1341 
1342 #ifdef IPSEC
1343 	if (ipsec_used)
1344 		ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
1345 	/* XXX: else */
1346 #else
1347 	esp4_input(m, iphdrlen);
1348 #endif
1349 
1350 	/* We handled it, it shouldn't be handled by UDP */
1351 	*mp = NULL; /* avoid free by caller ... */
1352 	return 1;
1353 }
1354 #endif
1355 
1356 PR_WRAP_USRREQS(udp)
1357 #define	udp_attach	udp_attach_wrapper
1358 #define	udp_detach	udp_detach_wrapper
1359 #define	udp_accept	udp_accept_wrapper
1360 #define	udp_bind	udp_bind_wrapper
1361 #define	udp_listen	udp_listen_wrapper
1362 #define	udp_connect	udp_connect_wrapper
1363 #define	udp_connect2	udp_connect2_wrapper
1364 #define	udp_disconnect	udp_disconnect_wrapper
1365 #define	udp_shutdown	udp_shutdown_wrapper
1366 #define	udp_abort	udp_abort_wrapper
1367 #define	udp_ioctl	udp_ioctl_wrapper
1368 #define	udp_stat	udp_stat_wrapper
1369 #define	udp_peeraddr	udp_peeraddr_wrapper
1370 #define	udp_sockaddr	udp_sockaddr_wrapper
1371 #define	udp_rcvd	udp_rcvd_wrapper
1372 #define	udp_recvoob	udp_recvoob_wrapper
1373 #define	udp_send	udp_send_wrapper
1374 #define	udp_sendoob	udp_sendoob_wrapper
1375 #define	udp_purgeif	udp_purgeif_wrapper
1376 
1377 const struct pr_usrreqs udp_usrreqs = {
1378 	.pr_attach	= udp_attach,
1379 	.pr_detach	= udp_detach,
1380 	.pr_accept	= udp_accept,
1381 	.pr_bind	= udp_bind,
1382 	.pr_listen	= udp_listen,
1383 	.pr_connect	= udp_connect,
1384 	.pr_connect2	= udp_connect2,
1385 	.pr_disconnect	= udp_disconnect,
1386 	.pr_shutdown	= udp_shutdown,
1387 	.pr_abort	= udp_abort,
1388 	.pr_ioctl	= udp_ioctl,
1389 	.pr_stat	= udp_stat,
1390 	.pr_peeraddr	= udp_peeraddr,
1391 	.pr_sockaddr	= udp_sockaddr,
1392 	.pr_rcvd	= udp_rcvd,
1393 	.pr_recvoob	= udp_recvoob,
1394 	.pr_send	= udp_send,
1395 	.pr_sendoob	= udp_sendoob,
1396 	.pr_purgeif	= udp_purgeif,
1397 };
1398