xref: /netbsd-src/sys/netinet/udp_usrreq.c (revision 7c3f385475147b6e1c4753f2bee961630e2dfc40)
1 /*	$NetBSD: udp_usrreq.c,v 1.168 2008/04/15 04:43:25 thorpej Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.168 2008/04/15 04:43:25 thorpej Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_ipsec.h"
68 #include "opt_inet_csum.h"
69 #include "opt_ipkdb.h"
70 #include "opt_mbuftrace.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/stat.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_icmp.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 #include <netinet/udp_private.h>
98 
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet/icmp6.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/ip6_private.h>
104 #include <netinet6/in6_pcb.h>
105 #include <netinet6/udp6_var.h>
106 #include <netinet6/udp6_private.h>
107 #include <netinet6/scope6_var.h>
108 #endif
109 
110 #ifndef INET6
111 /* always need ip6.h for IP6_EXTHDR_GET */
112 #include <netinet/ip6.h>
113 #endif
114 
115 #include "faith.h"
116 #if defined(NFAITH) && NFAITH > 0
117 #include <net/if_faith.h>
118 #endif
119 
120 #include <machine/stdarg.h>
121 
122 #ifdef FAST_IPSEC
123 #include <netipsec/ipsec.h>
124 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
125 #include <netipsec/esp.h>
126 #ifdef INET6
127 #include <netipsec/ipsec6.h>
128 #endif
129 #endif	/* FAST_IPSEC*/
130 
131 #ifdef IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netinet6/esp.h>
134 #include <netkey/key.h>
135 #endif /*IPSEC*/
136 
137 #ifdef IPKDB
138 #include <ipkdb/ipkdb.h>
139 #endif
140 
141 /*
142  * UDP protocol implementation.
143  * Per RFC 768, August, 1980.
144  */
145 int	udpcksum = 1;
146 int	udp_do_loopback_cksum = 0;
147 
148 struct	inpcbtable udbtable;
149 
150 percpu_t *udpstat_percpu;
151 
152 #ifdef INET
153 #ifdef IPSEC_NAT_T
154 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
155 	struct socket *);
156 #endif
157 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
158 	struct socket *);
159 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
160 	struct mbuf **, int);
161 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
162 #endif
163 #ifdef INET6
164 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
165 	struct socket *);
166 static int udp6_realinput (int, struct sockaddr_in6 *,
167 	struct sockaddr_in6 *, struct mbuf *, int);
168 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
169 #endif
170 #ifdef INET
171 static	void udp_notify (struct inpcb *, int);
172 #endif
173 
174 #ifndef UDBHASHSIZE
175 #define	UDBHASHSIZE	128
176 #endif
177 int	udbhashsize = UDBHASHSIZE;
178 
179 #ifdef MBUFTRACE
180 struct mowner udp_mowner = MOWNER_INIT("udp", "");
181 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
182 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
183 #endif
184 
185 #ifdef UDP_CSUM_COUNTERS
186 #include <sys/device.h>
187 
188 #if defined(INET)
189 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
190     NULL, "udp", "hwcsum bad");
191 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
192     NULL, "udp", "hwcsum ok");
193 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
194     NULL, "udp", "hwcsum data");
195 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
196     NULL, "udp", "swcsum");
197 
198 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
199 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
200 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
201 EVCNT_ATTACH_STATIC(udp_swcsum);
202 #endif /* defined(INET) */
203 
204 #if defined(INET6)
205 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
206     NULL, "udp6", "hwcsum bad");
207 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
208     NULL, "udp6", "hwcsum ok");
209 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
210     NULL, "udp6", "hwcsum data");
211 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
212     NULL, "udp6", "swcsum");
213 
214 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
215 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
216 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
217 EVCNT_ATTACH_STATIC(udp6_swcsum);
218 #endif /* defined(INET6) */
219 
220 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
221 
222 #else
223 
224 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
225 
226 #endif /* UDP_CSUM_COUNTERS */
227 
228 void
229 udp_init(void)
230 {
231 
232 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
233 
234 	MOWNER_ATTACH(&udp_tx_mowner);
235 	MOWNER_ATTACH(&udp_rx_mowner);
236 	MOWNER_ATTACH(&udp_mowner);
237 
238 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
239 }
240 
241 /*
242  * Checksum extended UDP header and data.
243  */
244 
245 int
246 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
247     int iphlen, int len)
248 {
249 
250 	switch (af) {
251 #ifdef INET
252 	case AF_INET:
253 		return udp4_input_checksum(m, uh, iphlen, len);
254 #endif
255 #ifdef INET6
256 	case AF_INET6:
257 		return udp6_input_checksum(m, uh, iphlen, len);
258 #endif
259 	}
260 #ifdef DIAGNOSTIC
261 	panic("udp_input_checksum: unknown af %d", af);
262 #endif
263 	/* NOTREACHED */
264 	return -1;
265 }
266 
267 #ifdef INET
268 
269 /*
270  * Checksum extended UDP header and data.
271  */
272 
273 static int
274 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
275     int iphlen, int len)
276 {
277 
278 	/*
279 	 * XXX it's better to record and check if this mbuf is
280 	 * already checked.
281 	 */
282 
283 	if (uh->uh_sum == 0)
284 		return 0;
285 
286 	switch (m->m_pkthdr.csum_flags &
287 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
288 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
289 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
290 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
291 		goto badcsum;
292 
293 	case M_CSUM_UDPv4|M_CSUM_DATA: {
294 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
295 
296 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
297 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
298 			const struct ip *ip =
299 			    mtod(m, const struct ip *);
300 
301 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
302 			    ip->ip_dst.s_addr,
303 			    htons(hw_csum + len + IPPROTO_UDP));
304 		}
305 		if ((hw_csum ^ 0xffff) != 0)
306 			goto badcsum;
307 		break;
308 	}
309 
310 	case M_CSUM_UDPv4:
311 		/* Checksum was okay. */
312 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
313 		break;
314 
315 	default:
316 		/*
317 		 * Need to compute it ourselves.  Maybe skip checksum
318 		 * on loopback interfaces.
319 		 */
320 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
321 				     IFF_LOOPBACK) ||
322 				   udp_do_loopback_cksum)) {
323 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
324 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
325 				goto badcsum;
326 		}
327 		break;
328 	}
329 
330 	return 0;
331 
332 badcsum:
333 	UDP_STATINC(UDP_STAT_BADSUM);
334 	return -1;
335 }
336 
337 void
338 udp_input(struct mbuf *m, ...)
339 {
340 	va_list ap;
341 	struct sockaddr_in src, dst;
342 	struct ip *ip;
343 	struct udphdr *uh;
344 	int iphlen;
345 	int len;
346 	int n;
347 	u_int16_t ip_len;
348 
349 	va_start(ap, m);
350 	iphlen = va_arg(ap, int);
351 	(void)va_arg(ap, int);		/* ignore value, advance ap */
352 	va_end(ap);
353 
354 	MCLAIM(m, &udp_rx_mowner);
355 	UDP_STATINC(UDP_STAT_IPACKETS);
356 
357 	/*
358 	 * Get IP and UDP header together in first mbuf.
359 	 */
360 	ip = mtod(m, struct ip *);
361 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
362 	if (uh == NULL) {
363 		UDP_STATINC(UDP_STAT_HDROPS);
364 		return;
365 	}
366 	KASSERT(UDP_HDR_ALIGNED_P(uh));
367 
368 	/* destination port of 0 is illegal, based on RFC768. */
369 	if (uh->uh_dport == 0)
370 		goto bad;
371 
372 	/*
373 	 * Make mbuf data length reflect UDP length.
374 	 * If not enough data to reflect UDP length, drop.
375 	 */
376 	ip_len = ntohs(ip->ip_len);
377 	len = ntohs((u_int16_t)uh->uh_ulen);
378 	if (ip_len != iphlen + len) {
379 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
380 			UDP_STATINC(UDP_STAT_BADLEN);
381 			goto bad;
382 		}
383 		m_adj(m, iphlen + len - ip_len);
384 	}
385 
386 	/*
387 	 * Checksum extended UDP header and data.
388 	 */
389 	if (udp4_input_checksum(m, uh, iphlen, len))
390 		goto badcsum;
391 
392 	/* construct source and dst sockaddrs. */
393 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
394 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
395 
396 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
397 		UDP_STATINC(UDP_STAT_HDROPS);
398 		return;
399 	}
400 #ifdef INET6
401 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
402 		struct sockaddr_in6 src6, dst6;
403 
404 		bzero(&src6, sizeof(src6));
405 		src6.sin6_family = AF_INET6;
406 		src6.sin6_len = sizeof(struct sockaddr_in6);
407 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
408 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
409 			sizeof(ip->ip_src));
410 		src6.sin6_port = uh->uh_sport;
411 		bzero(&dst6, sizeof(dst6));
412 		dst6.sin6_family = AF_INET6;
413 		dst6.sin6_len = sizeof(struct sockaddr_in6);
414 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
415 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
416 			sizeof(ip->ip_dst));
417 		dst6.sin6_port = uh->uh_dport;
418 
419 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
420 	}
421 #endif
422 
423 	if (n == 0) {
424 		if (m->m_flags & (M_BCAST | M_MCAST)) {
425 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
426 			goto bad;
427 		}
428 		UDP_STATINC(UDP_STAT_NOPORT);
429 #ifdef IPKDB
430 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
431 				m, iphlen + sizeof(struct udphdr),
432 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
433 			/*
434 			 * It was a debugger connect packet,
435 			 * just drop it now
436 			 */
437 			goto bad;
438 		}
439 #endif
440 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
441 		m = NULL;
442 	}
443 
444 bad:
445 	if (m)
446 		m_freem(m);
447 	return;
448 
449 badcsum:
450 	m_freem(m);
451 }
452 #endif
453 
454 #ifdef INET6
455 static int
456 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
457 {
458 
459 	/*
460 	 * XXX it's better to record and check if this mbuf is
461 	 * already checked.
462 	 */
463 
464 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
465 		goto good;
466 	}
467 	if (uh->uh_sum == 0) {
468 		UDP6_STATINC(UDP6_STAT_NOSUM);
469 		goto bad;
470 	}
471 
472 	switch (m->m_pkthdr.csum_flags &
473 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
474 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
475 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
476 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
477 		UDP6_STATINC(UDP6_STAT_BADSUM);
478 		goto bad;
479 
480 #if 0 /* notyet */
481 	case M_CSUM_UDPv6|M_CSUM_DATA:
482 #endif
483 
484 	case M_CSUM_UDPv6:
485 		/* Checksum was okay. */
486 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
487 		break;
488 
489 	default:
490 		/*
491 		 * Need to compute it ourselves.  Maybe skip checksum
492 		 * on loopback interfaces.
493 		 */
494 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
495 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
496 			UDP6_STATINC(UDP6_STAT_BADSUM);
497 			goto bad;
498 		}
499 	}
500 
501 good:
502 	return 0;
503 bad:
504 	return -1;
505 }
506 
507 int
508 udp6_input(struct mbuf **mp, int *offp, int proto)
509 {
510 	struct mbuf *m = *mp;
511 	int off = *offp;
512 	struct sockaddr_in6 src, dst;
513 	struct ip6_hdr *ip6;
514 	struct udphdr *uh;
515 	u_int32_t plen, ulen;
516 
517 	ip6 = mtod(m, struct ip6_hdr *);
518 
519 #if defined(NFAITH) && 0 < NFAITH
520 	if (faithprefix(&ip6->ip6_dst)) {
521 		/* send icmp6 host unreach? */
522 		m_freem(m);
523 		return IPPROTO_DONE;
524 	}
525 #endif
526 
527 	UDP6_STATINC(UDP6_STAT_IPACKETS);
528 
529 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
530 	plen = m->m_pkthdr.len - off;
531 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
532 	if (uh == NULL) {
533 		IP6_STATINC(IP6_STAT_TOOSHORT);
534 		return IPPROTO_DONE;
535 	}
536 	KASSERT(UDP_HDR_ALIGNED_P(uh));
537 	ulen = ntohs((u_short)uh->uh_ulen);
538 	/*
539 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
540 	 * iff payload length > 0xffff.
541 	 */
542 	if (ulen == 0 && plen > 0xffff)
543 		ulen = plen;
544 
545 	if (plen != ulen) {
546 		UDP6_STATINC(UDP6_STAT_BADLEN);
547 		goto bad;
548 	}
549 
550 	/* destination port of 0 is illegal, based on RFC768. */
551 	if (uh->uh_dport == 0)
552 		goto bad;
553 
554 	/* Be proactive about malicious use of IPv4 mapped address */
555 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
556 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
557 		/* XXX stat */
558 		goto bad;
559 	}
560 
561 	/*
562 	 * Checksum extended UDP header and data.  Maybe skip checksum
563 	 * on loopback interfaces.
564 	 */
565 	if (udp6_input_checksum(m, uh, off, ulen))
566 		goto bad;
567 
568 	/*
569 	 * Construct source and dst sockaddrs.
570 	 */
571 	bzero(&src, sizeof(src));
572 	src.sin6_family = AF_INET6;
573 	src.sin6_len = sizeof(struct sockaddr_in6);
574 	src.sin6_addr = ip6->ip6_src;
575 	src.sin6_port = uh->uh_sport;
576 	bzero(&dst, sizeof(dst));
577 	dst.sin6_family = AF_INET6;
578 	dst.sin6_len = sizeof(struct sockaddr_in6);
579 	dst.sin6_addr = ip6->ip6_dst;
580 	dst.sin6_port = uh->uh_dport;
581 
582 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
583 		if (m->m_flags & M_MCAST) {
584 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
585 			goto bad;
586 		}
587 		UDP6_STATINC(UDP6_STAT_NOPORT);
588 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
589 		m = NULL;
590 	}
591 
592 bad:
593 	if (m)
594 		m_freem(m);
595 	return IPPROTO_DONE;
596 }
597 #endif
598 
599 #ifdef INET
600 static void
601 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
602 	struct sockaddr *src, struct socket *so)
603 {
604 	struct mbuf *opts = NULL;
605 	struct mbuf *n;
606 	struct inpcb *inp = NULL;
607 
608 	if (!so)
609 		return;
610 	switch (so->so_proto->pr_domain->dom_family) {
611 	case AF_INET:
612 		inp = sotoinpcb(so);
613 		break;
614 #ifdef INET6
615 	case AF_INET6:
616 		break;
617 #endif
618 	default:
619 		return;
620 	}
621 
622 #if defined(IPSEC) || defined(FAST_IPSEC)
623 	/* check AH/ESP integrity. */
624 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
625 		ipsecstat.in_polvio++;
626 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
627 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
628 			    0, 0);
629 		return;
630 	}
631 #endif /*IPSEC*/
632 
633 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
634 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
635 			 || so->so_options & SO_TIMESTAMP)) {
636 			struct ip *ip = mtod(n, struct ip *);
637 			ip_savecontrol(inp, &opts, ip, n);
638 		}
639 
640 		m_adj(n, off);
641 		if (sbappendaddr(&so->so_rcv, src, n,
642 				opts) == 0) {
643 			m_freem(n);
644 			if (opts)
645 				m_freem(opts);
646 			so->so_rcv.sb_overflowed++;
647 			UDP_STATINC(UDP_STAT_FULLSOCK);
648 		} else
649 			sorwakeup(so);
650 	}
651 }
652 #endif
653 
654 #ifdef INET6
655 static void
656 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
657 	struct sockaddr *src, struct socket *so)
658 {
659 	struct mbuf *opts = NULL;
660 	struct mbuf *n;
661 	struct in6pcb *in6p = NULL;
662 
663 	if (!so)
664 		return;
665 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
666 		return;
667 	in6p = sotoin6pcb(so);
668 
669 #if defined(IPSEC) || defined(FAST_IPSEC)
670 	/* check AH/ESP integrity. */
671 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
672 		ipsec6stat.in_polvio++;
673 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
674 			icmp6_error(n, ICMP6_DST_UNREACH,
675 			    ICMP6_DST_UNREACH_ADMIN, 0);
676 		return;
677 	}
678 #endif /*IPSEC*/
679 
680 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
681 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
682 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
683 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
684 			ip6_savecontrol(in6p, &opts, ip6, n);
685 		}
686 
687 		m_adj(n, off);
688 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
689 			m_freem(n);
690 			if (opts)
691 				m_freem(opts);
692 			so->so_rcv.sb_overflowed++;
693 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
694 		} else
695 			sorwakeup(so);
696 	}
697 }
698 #endif
699 
700 #ifdef INET
701 static int
702 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
703 	struct mbuf **mp, int off /* offset of udphdr */)
704 {
705 	u_int16_t *sport, *dport;
706 	int rcvcnt;
707 	struct in_addr *src4, *dst4;
708 	struct inpcb_hdr *inph;
709 	struct inpcb *inp;
710 	struct mbuf *m = *mp;
711 
712 	rcvcnt = 0;
713 	off += sizeof(struct udphdr);	/* now, offset of payload */
714 
715 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
716 		goto bad;
717 
718 	src4 = &src->sin_addr;
719 	sport = &src->sin_port;
720 	dst4 = &dst->sin_addr;
721 	dport = &dst->sin_port;
722 
723 	if (IN_MULTICAST(dst4->s_addr) ||
724 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
725 		/*
726 		 * Deliver a multicast or broadcast datagram to *all* sockets
727 		 * for which the local and remote addresses and ports match
728 		 * those of the incoming datagram.  This allows more than
729 		 * one process to receive multi/broadcasts on the same port.
730 		 * (This really ought to be done for unicast datagrams as
731 		 * well, but that would cause problems with existing
732 		 * applications that open both address-specific sockets and
733 		 * a wildcard socket listening to the same port -- they would
734 		 * end up receiving duplicates of every unicast datagram.
735 		 * Those applications open the multiple sockets to overcome an
736 		 * inadequacy of the UDP socket interface, but for backwards
737 		 * compatibility we avoid the problem here rather than
738 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
739 		 */
740 
741 		/*
742 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
743 		 * we need udpiphdr for IPsec processing so we do that later.
744 		 */
745 		/*
746 		 * Locate pcb(s) for datagram.
747 		 */
748 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
749 			inp = (struct inpcb *)inph;
750 			if (inp->inp_af != AF_INET)
751 				continue;
752 
753 			if (inp->inp_lport != *dport)
754 				continue;
755 			if (!in_nullhost(inp->inp_laddr)) {
756 				if (!in_hosteq(inp->inp_laddr, *dst4))
757 					continue;
758 			}
759 			if (!in_nullhost(inp->inp_faddr)) {
760 				if (!in_hosteq(inp->inp_faddr, *src4) ||
761 				    inp->inp_fport != *sport)
762 					continue;
763 			}
764 
765 			udp4_sendup(m, off, (struct sockaddr *)src,
766 				inp->inp_socket);
767 			rcvcnt++;
768 
769 			/*
770 			 * Don't look for additional matches if this one does
771 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
772 			 * socket options set.  This heuristic avoids searching
773 			 * through all pcbs in the common case of a non-shared
774 			 * port.  It assumes that an application will never
775 			 * clear these options after setting them.
776 			 */
777 			if ((inp->inp_socket->so_options &
778 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
779 				break;
780 		}
781 	} else {
782 		/*
783 		 * Locate pcb for datagram.
784 		 */
785 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
786 		if (inp == 0) {
787 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
788 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
789 			if (inp == 0)
790 				return rcvcnt;
791 		}
792 
793 #ifdef IPSEC_NAT_T
794 		/* Handle ESP over UDP */
795 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
796 			struct sockaddr *sa = (struct sockaddr *)src;
797 
798 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
799 			case -1: 	/* Error, m was freeed */
800 				rcvcnt = -1;
801 				goto bad;
802 				break;
803 
804 			case 1:		/* ESP over UDP */
805 				rcvcnt++;
806 				goto bad;
807 				break;
808 
809 			case 0: 	/* plain UDP */
810 			default: 	/* Unexpected */
811 				/*
812 				 * Normal UDP processing will take place
813 				 * m may have changed.
814 				 */
815 				m = *mp;
816 				break;
817 			}
818 		}
819 #endif
820 
821 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
822 		rcvcnt++;
823 	}
824 
825 bad:
826 	return rcvcnt;
827 }
828 #endif
829 
830 #ifdef INET6
831 static int
832 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
833 	struct mbuf *m, int off)
834 {
835 	u_int16_t sport, dport;
836 	int rcvcnt;
837 	struct in6_addr src6, *dst6;
838 	const struct in_addr *dst4;
839 	struct inpcb_hdr *inph;
840 	struct in6pcb *in6p;
841 
842 	rcvcnt = 0;
843 	off += sizeof(struct udphdr);	/* now, offset of payload */
844 
845 	if (af != AF_INET && af != AF_INET6)
846 		goto bad;
847 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
848 		goto bad;
849 
850 	src6 = src->sin6_addr;
851 	if (sa6_recoverscope(src) != 0) {
852 		/* XXX: should be impossible. */
853 		goto bad;
854 	}
855 	sport = src->sin6_port;
856 
857 	dport = dst->sin6_port;
858 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
859 	dst6 = &dst->sin6_addr;
860 
861 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
862 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
863 		/*
864 		 * Deliver a multicast or broadcast datagram to *all* sockets
865 		 * for which the local and remote addresses and ports match
866 		 * those of the incoming datagram.  This allows more than
867 		 * one process to receive multi/broadcasts on the same port.
868 		 * (This really ought to be done for unicast datagrams as
869 		 * well, but that would cause problems with existing
870 		 * applications that open both address-specific sockets and
871 		 * a wildcard socket listening to the same port -- they would
872 		 * end up receiving duplicates of every unicast datagram.
873 		 * Those applications open the multiple sockets to overcome an
874 		 * inadequacy of the UDP socket interface, but for backwards
875 		 * compatibility we avoid the problem here rather than
876 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
877 		 */
878 
879 		/*
880 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
881 		 * we need udpiphdr for IPsec processing so we do that later.
882 		 */
883 		/*
884 		 * Locate pcb(s) for datagram.
885 		 */
886 		CIRCLEQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
887 			in6p = (struct in6pcb *)inph;
888 			if (in6p->in6p_af != AF_INET6)
889 				continue;
890 
891 			if (in6p->in6p_lport != dport)
892 				continue;
893 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
894 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
895 				    dst6))
896 					continue;
897 			} else {
898 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
899 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
900 					continue;
901 			}
902 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
903 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
904 				    &src6) || in6p->in6p_fport != sport)
905 					continue;
906 			} else {
907 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
908 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
909 					continue;
910 			}
911 
912 			udp6_sendup(m, off, (struct sockaddr *)src,
913 				in6p->in6p_socket);
914 			rcvcnt++;
915 
916 			/*
917 			 * Don't look for additional matches if this one does
918 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
919 			 * socket options set.  This heuristic avoids searching
920 			 * through all pcbs in the common case of a non-shared
921 			 * port.  It assumes that an application will never
922 			 * clear these options after setting them.
923 			 */
924 			if ((in6p->in6p_socket->so_options &
925 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
926 				break;
927 		}
928 	} else {
929 		/*
930 		 * Locate pcb for datagram.
931 		 */
932 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
933 		    dport, 0);
934 		if (in6p == 0) {
935 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
936 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
937 			if (in6p == 0)
938 				return rcvcnt;
939 		}
940 
941 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
942 		rcvcnt++;
943 	}
944 
945 bad:
946 	return rcvcnt;
947 }
948 #endif
949 
950 #ifdef INET
951 /*
952  * Notify a udp user of an asynchronous error;
953  * just wake up so that he can collect error status.
954  */
955 static void
956 udp_notify(struct inpcb *inp, int errno)
957 {
958 	inp->inp_socket->so_error = errno;
959 	sorwakeup(inp->inp_socket);
960 	sowwakeup(inp->inp_socket);
961 }
962 
963 void *
964 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
965 {
966 	struct ip *ip = v;
967 	struct udphdr *uh;
968 	void (*notify)(struct inpcb *, int) = udp_notify;
969 	int errno;
970 
971 	if (sa->sa_family != AF_INET
972 	 || sa->sa_len != sizeof(struct sockaddr_in))
973 		return NULL;
974 	if ((unsigned)cmd >= PRC_NCMDS)
975 		return NULL;
976 	errno = inetctlerrmap[cmd];
977 	if (PRC_IS_REDIRECT(cmd))
978 		notify = in_rtchange, ip = 0;
979 	else if (cmd == PRC_HOSTDEAD)
980 		ip = 0;
981 	else if (errno == 0)
982 		return NULL;
983 	if (ip) {
984 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
985 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
986 		    ip->ip_src, uh->uh_sport, errno, notify);
987 
988 		/* XXX mapped address case */
989 	} else
990 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
991 		    notify);
992 	return NULL;
993 }
994 
995 int
996 udp_ctloutput(int op, struct socket *so, int level, int optname,
997     struct mbuf **mp)
998 {
999 	int s;
1000 	int error = 0;
1001 	struct mbuf *m;
1002 	struct inpcb *inp;
1003 	int family;
1004 
1005 	family = so->so_proto->pr_domain->dom_family;
1006 
1007 	s = splsoftnet();
1008 	switch (family) {
1009 #ifdef INET
1010 	case PF_INET:
1011 		if (level != IPPROTO_UDP) {
1012 			error = ip_ctloutput(op, so, level, optname, mp);
1013 			goto end;
1014 		}
1015 		break;
1016 #endif
1017 #ifdef INET6
1018 	case PF_INET6:
1019 		if (level != IPPROTO_UDP) {
1020 			error = ip6_ctloutput(op, so, level, optname, mp);
1021 			goto end;
1022 		}
1023 		break;
1024 #endif
1025 	default:
1026 		error = EAFNOSUPPORT;
1027 		goto end;
1028 	}
1029 
1030 
1031 	switch (op) {
1032 	case PRCO_SETOPT:
1033 		m = *mp;
1034 		inp = sotoinpcb(so);
1035 
1036 		switch (optname) {
1037 		case UDP_ENCAP:
1038 			if (m == NULL || m->m_len != sizeof(int)) {
1039 				error = EINVAL;
1040 				break;
1041 			}
1042 
1043 			switch(*mtod(m, int *)) {
1044 #ifdef IPSEC_NAT_T
1045 			case 0:
1046 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1047 				break;
1048 
1049 			case UDP_ENCAP_ESPINUDP:
1050 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1051 				inp->inp_flags |= INP_ESPINUDP;
1052 				break;
1053 
1054 			case UDP_ENCAP_ESPINUDP_NON_IKE:
1055 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
1056 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
1057 				break;
1058 #endif
1059 			default:
1060 				error = EINVAL;
1061 				break;
1062 			}
1063 			break;
1064 
1065 		default:
1066 			error = ENOPROTOOPT;
1067 			break;
1068 		}
1069 		if (m != NULL) {
1070 			m_free(m);
1071 		}
1072 		break;
1073 
1074 	default:
1075 		error = EINVAL;
1076 		break;
1077 	}
1078 
1079 end:
1080 	splx(s);
1081 	return error;
1082 }
1083 
1084 
1085 int
1086 udp_output(struct mbuf *m, ...)
1087 {
1088 	struct inpcb *inp;
1089 	struct udpiphdr *ui;
1090 	struct route *ro;
1091 	int len = m->m_pkthdr.len;
1092 	int error = 0;
1093 	va_list ap;
1094 
1095 	MCLAIM(m, &udp_tx_mowner);
1096 	va_start(ap, m);
1097 	inp = va_arg(ap, struct inpcb *);
1098 	va_end(ap);
1099 
1100 	/*
1101 	 * Calculate data length and get a mbuf
1102 	 * for UDP and IP headers.
1103 	 */
1104 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
1105 	if (m == 0) {
1106 		error = ENOBUFS;
1107 		goto release;
1108 	}
1109 
1110 	/*
1111 	 * Compute the packet length of the IP header, and
1112 	 * punt if the length looks bogus.
1113 	 */
1114 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1115 		error = EMSGSIZE;
1116 		goto release;
1117 	}
1118 
1119 	/*
1120 	 * Fill in mbuf with extended UDP header
1121 	 * and addresses and length put into network format.
1122 	 */
1123 	ui = mtod(m, struct udpiphdr *);
1124 	ui->ui_pr = IPPROTO_UDP;
1125 	ui->ui_src = inp->inp_laddr;
1126 	ui->ui_dst = inp->inp_faddr;
1127 	ui->ui_sport = inp->inp_lport;
1128 	ui->ui_dport = inp->inp_fport;
1129 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
1130 
1131 	ro = &inp->inp_route;
1132 
1133 	/*
1134 	 * Set up checksum and output datagram.
1135 	 */
1136 	if (udpcksum) {
1137 		/*
1138 		 * XXX Cache pseudo-header checksum part for
1139 		 * XXX "connected" UDP sockets.
1140 		 */
1141 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
1142 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
1143 		    sizeof(struct udphdr) + IPPROTO_UDP));
1144 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
1145 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1146 	} else
1147 		ui->ui_sum = 0;
1148 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
1149 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
1150 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
1151 	UDP_STATINC(UDP_STAT_OPACKETS);
1152 
1153 	return (ip_output(m, inp->inp_options, ro,
1154 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
1155 	    inp->inp_moptions, inp->inp_socket));
1156 
1157 release:
1158 	m_freem(m);
1159 	return (error);
1160 }
1161 
1162 int	udp_sendspace = 9216;		/* really max datagram size */
1163 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1164 					/* 40 1K datagrams */
1165 
1166 /*ARGSUSED*/
1167 int
1168 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
1169 	struct mbuf *control, struct lwp *l)
1170 {
1171 	struct inpcb *inp;
1172 	int s;
1173 	int error = 0;
1174 
1175 	if (req == PRU_CONTROL)
1176 		return (in_control(so, (long)m, (void *)nam,
1177 		    (struct ifnet *)control, l));
1178 
1179 	s = splsoftnet();
1180 
1181 	if (req == PRU_PURGEIF) {
1182 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
1183 		in_purgeif((struct ifnet *)control);
1184 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
1185 		splx(s);
1186 		return (0);
1187 	}
1188 
1189 	inp = sotoinpcb(so);
1190 #ifdef DIAGNOSTIC
1191 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
1192 		panic("udp_usrreq: unexpected control mbuf");
1193 #endif
1194 	if (inp == 0 && req != PRU_ATTACH) {
1195 		error = EINVAL;
1196 		goto release;
1197 	}
1198 
1199 	/*
1200 	 * Note: need to block udp_input while changing
1201 	 * the udp pcb queue and/or pcb addresses.
1202 	 */
1203 	switch (req) {
1204 
1205 	case PRU_ATTACH:
1206 		if (inp != 0) {
1207 			error = EISCONN;
1208 			break;
1209 		}
1210 #ifdef MBUFTRACE
1211 		so->so_mowner = &udp_mowner;
1212 		so->so_rcv.sb_mowner = &udp_rx_mowner;
1213 		so->so_snd.sb_mowner = &udp_tx_mowner;
1214 #endif
1215 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1216 			error = soreserve(so, udp_sendspace, udp_recvspace);
1217 			if (error)
1218 				break;
1219 		}
1220 		error = in_pcballoc(so, &udbtable);
1221 		if (error)
1222 			break;
1223 		inp = sotoinpcb(so);
1224 		inp->inp_ip.ip_ttl = ip_defttl;
1225 		break;
1226 
1227 	case PRU_DETACH:
1228 		in_pcbdetach(inp);
1229 		break;
1230 
1231 	case PRU_BIND:
1232 		error = in_pcbbind(inp, nam, l);
1233 		break;
1234 
1235 	case PRU_LISTEN:
1236 		error = EOPNOTSUPP;
1237 		break;
1238 
1239 	case PRU_CONNECT:
1240 		error = in_pcbconnect(inp, nam, l);
1241 		if (error)
1242 			break;
1243 		soisconnected(so);
1244 		break;
1245 
1246 	case PRU_CONNECT2:
1247 		error = EOPNOTSUPP;
1248 		break;
1249 
1250 	case PRU_DISCONNECT:
1251 		/*soisdisconnected(so);*/
1252 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1253 		in_pcbdisconnect(inp);
1254 		inp->inp_laddr = zeroin_addr;		/* XXX */
1255 		in_pcbstate(inp, INP_BOUND);		/* XXX */
1256 		break;
1257 
1258 	case PRU_SHUTDOWN:
1259 		socantsendmore(so);
1260 		break;
1261 
1262 	case PRU_RCVD:
1263 		error = EOPNOTSUPP;
1264 		break;
1265 
1266 	case PRU_SEND:
1267 		if (control && control->m_len) {
1268 			m_freem(control);
1269 			m_freem(m);
1270 			error = EINVAL;
1271 			break;
1272 		}
1273 	{
1274 		struct in_addr laddr;			/* XXX */
1275 
1276 		if (nam) {
1277 			laddr = inp->inp_laddr;		/* XXX */
1278 			if ((so->so_state & SS_ISCONNECTED) != 0) {
1279 				error = EISCONN;
1280 				goto die;
1281 			}
1282 			error = in_pcbconnect(inp, nam, l);
1283 			if (error)
1284 				goto die;
1285 		} else {
1286 			if ((so->so_state & SS_ISCONNECTED) == 0) {
1287 				error = ENOTCONN;
1288 				goto die;
1289 			}
1290 		}
1291 		error = udp_output(m, inp);
1292 		m = NULL;
1293 		if (nam) {
1294 			in_pcbdisconnect(inp);
1295 			inp->inp_laddr = laddr;		/* XXX */
1296 			in_pcbstate(inp, INP_BOUND);	/* XXX */
1297 		}
1298 	  die:
1299 		if (m)
1300 			m_freem(m);
1301 	}
1302 		break;
1303 
1304 	case PRU_SENSE:
1305 		/*
1306 		 * stat: don't bother with a blocksize.
1307 		 */
1308 		splx(s);
1309 		return (0);
1310 
1311 	case PRU_RCVOOB:
1312 		error =  EOPNOTSUPP;
1313 		break;
1314 
1315 	case PRU_SENDOOB:
1316 		m_freem(control);
1317 		m_freem(m);
1318 		error =  EOPNOTSUPP;
1319 		break;
1320 
1321 	case PRU_SOCKADDR:
1322 		in_setsockaddr(inp, nam);
1323 		break;
1324 
1325 	case PRU_PEERADDR:
1326 		in_setpeeraddr(inp, nam);
1327 		break;
1328 
1329 	default:
1330 		panic("udp_usrreq");
1331 	}
1332 
1333 release:
1334 	splx(s);
1335 	return (error);
1336 }
1337 
1338 static void
1339 udpstat_convert_to_user_cb(void *v1, void *v2, struct cpu_info *ci)
1340 {
1341 	uint64_t *udpsc = v1;
1342 	uint64_t *udps = v2;
1343 	u_int i;
1344 
1345 	for (i = 0; i < UDP_NSTATS; i++)
1346 		udps[i] += udpsc[i];
1347 }
1348 
1349 static void
1350 udpstat_convert_to_user(uint64_t *udps)
1351 {
1352 
1353 	memset(udps, 0, sizeof(uint64_t) * UDP_NSTATS);
1354 	percpu_foreach(udpstat_percpu, udpstat_convert_to_user_cb, udps);
1355 }
1356 
1357 static int
1358 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
1359 {
1360 	struct sysctlnode node;
1361 	uint64_t udps[UDP_NSTATS];
1362 
1363 	udpstat_convert_to_user(udps);
1364 	node = *rnode;
1365 	node.sysctl_data = udps;
1366 	node.sysctl_size = sizeof(udps);
1367 	return (sysctl_lookup(SYSCTLFN_CALL(&node)));
1368 }
1369 
1370 /*
1371  * Sysctl for udp variables.
1372  */
1373 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
1374 {
1375 
1376 	sysctl_createv(clog, 0, NULL, NULL,
1377 		       CTLFLAG_PERMANENT,
1378 		       CTLTYPE_NODE, "net", NULL,
1379 		       NULL, 0, NULL, 0,
1380 		       CTL_NET, CTL_EOL);
1381 	sysctl_createv(clog, 0, NULL, NULL,
1382 		       CTLFLAG_PERMANENT,
1383 		       CTLTYPE_NODE, "inet", NULL,
1384 		       NULL, 0, NULL, 0,
1385 		       CTL_NET, PF_INET, CTL_EOL);
1386 	sysctl_createv(clog, 0, NULL, NULL,
1387 		       CTLFLAG_PERMANENT,
1388 		       CTLTYPE_NODE, "udp",
1389 		       SYSCTL_DESCR("UDPv4 related settings"),
1390 		       NULL, 0, NULL, 0,
1391 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
1392 
1393 	sysctl_createv(clog, 0, NULL, NULL,
1394 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1395 		       CTLTYPE_INT, "checksum",
1396 		       SYSCTL_DESCR("Compute UDP checksums"),
1397 		       NULL, 0, &udpcksum, 0,
1398 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
1399 		       CTL_EOL);
1400 	sysctl_createv(clog, 0, NULL, NULL,
1401 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1402 		       CTLTYPE_INT, "sendspace",
1403 		       SYSCTL_DESCR("Default UDP send buffer size"),
1404 		       NULL, 0, &udp_sendspace, 0,
1405 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
1406 		       CTL_EOL);
1407 	sysctl_createv(clog, 0, NULL, NULL,
1408 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1409 		       CTLTYPE_INT, "recvspace",
1410 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1411 		       NULL, 0, &udp_recvspace, 0,
1412 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
1413 		       CTL_EOL);
1414 	sysctl_createv(clog, 0, NULL, NULL,
1415 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1416 		       CTLTYPE_INT, "do_loopback_cksum",
1417 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1418 		       NULL, 0, &udp_do_loopback_cksum, 0,
1419 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
1420 		       CTL_EOL);
1421 	sysctl_createv(clog, 0, NULL, NULL,
1422 		       CTLFLAG_PERMANENT,
1423 		       CTLTYPE_STRUCT, "pcblist",
1424 		       SYSCTL_DESCR("UDP protocol control block list"),
1425 		       sysctl_inpcblist, 0, &udbtable, 0,
1426 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
1427 		       CTL_EOL);
1428 	sysctl_createv(clog, 0, NULL, NULL,
1429 		       CTLFLAG_PERMANENT,
1430 		       CTLTYPE_STRUCT, "stats",
1431 		       SYSCTL_DESCR("UDP statistics"),
1432 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
1433 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
1434 		       CTL_EOL);
1435 }
1436 #endif
1437 
1438 void
1439 udp_statinc(u_int stat)
1440 {
1441 
1442 	KASSERT(stat < UDP_NSTATS);
1443 	UDP_STATINC(stat);
1444 }
1445 
1446 #if (defined INET && defined IPSEC_NAT_T)
1447 /*
1448  * Returns:
1449  * 1 if the packet was processed
1450  * 0 if normal UDP processing should take place
1451  * -1 if an error occurent and m was freed
1452  */
1453 static int
1454 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
1455     struct socket *so)
1456 {
1457 	size_t len;
1458 	void *data;
1459 	struct inpcb *inp;
1460 	size_t skip = 0;
1461 	size_t minlen;
1462 	size_t iphdrlen;
1463 	struct ip *ip;
1464 	struct mbuf *n;
1465 	struct m_tag *tag;
1466 	struct udphdr *udphdr;
1467 	u_int16_t sport, dport;
1468 	struct mbuf *m = *mp;
1469 
1470 	/*
1471 	 * Collapse the mbuf chain if the first mbuf is too short
1472 	 * The longest case is: UDP + non ESP marker + ESP
1473 	 */
1474 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1475 	if (minlen > m->m_pkthdr.len)
1476 		minlen = m->m_pkthdr.len;
1477 
1478 	if (m->m_len < minlen) {
1479 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1480 			printf("udp4_espinudp: m_pullup failed\n");
1481 			return -1;
1482 		}
1483 		m = *mp;
1484 	}
1485 
1486 	len = m->m_len - off;
1487 	data = mtod(m, char *) + off;
1488 	inp = sotoinpcb(so);
1489 
1490 	/* Ignore keepalive packets */
1491 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1492 		return 1;
1493 	}
1494 
1495 	/*
1496 	 * Check that the payload is long enough to hold
1497 	 * an ESP header and compute the length of encapsulation
1498 	 * header to remove
1499 	 */
1500 	if (inp->inp_flags & INP_ESPINUDP) {
1501 		u_int32_t *st = (u_int32_t *)data;
1502 
1503 		if ((len <= sizeof(struct esp)) || (*st == 0))
1504 			return 0; /* Normal UDP processing */
1505 
1506 		skip = sizeof(struct udphdr);
1507 	}
1508 
1509 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
1510 		u_int32_t *st = (u_int32_t *)data;
1511 
1512 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
1513 		    || ((st[0] | st[1]) != 0))
1514 			return 0; /* Normal UDP processing */
1515 
1516 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
1517 	}
1518 
1519 	/*
1520 	 * Get the UDP ports. They are handled in network
1521 	 * order everywhere in IPSEC_NAT_T code.
1522 	 */
1523 	udphdr = (struct udphdr *)((char *)data - skip);
1524 	sport = udphdr->uh_sport;
1525 	dport = udphdr->uh_dport;
1526 
1527 	/*
1528 	 * Remove the UDP header (and possibly the non ESP marker)
1529 	 * IP header lendth is iphdrlen
1530 	 * Before:
1531 	 *   <--- off --->
1532 	 *   +----+------+-----+
1533 	 *   | IP |  UDP | ESP |
1534 	 *   +----+------+-----+
1535 	 *        <-skip->
1536 	 * After:
1537 	 *          +----+-----+
1538 	 *          | IP | ESP |
1539 	 *          +----+-----+
1540 	 *   <-skip->
1541 	 */
1542 	iphdrlen = off - sizeof(struct udphdr);
1543 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
1544 	m_adj(m, skip);
1545 
1546 	ip = mtod(m, struct ip *);
1547 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
1548 	ip->ip_p = IPPROTO_ESP;
1549 
1550 	/*
1551 	 * Copy the mbuf to avoid multiple free, as both
1552 	 * esp4_input (which we call) and udp_input (which
1553 	 * called us) free the mbuf.
1554 	 */
1555 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
1556 		printf("udp4_espinudp: m_dup failed\n");
1557 		return 0;
1558 	}
1559 
1560 	/*
1561 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1562 	 * the source UDP port. This is required if we want
1563 	 * to select the right SPD for multiple hosts behind
1564 	 * same NAT
1565 	 */
1566 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1567 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1568 		printf("udp4_espinudp: m_tag_get failed\n");
1569 		m_freem(n);
1570 		return 0;
1571 	}
1572 	((u_int16_t *)(tag + 1))[0] = sport;
1573 	((u_int16_t *)(tag + 1))[1] = dport;
1574 	m_tag_prepend(n, tag);
1575 
1576 #ifdef FAST_IPSEC
1577 	ipsec4_common_input(n, iphdrlen, IPPROTO_ESP);
1578 #else
1579 	esp4_input(n, iphdrlen);
1580 #endif
1581 
1582 	/* We handled it, it shoudln't be handled by UDP */
1583 	return 1;
1584 }
1585 #endif
1586