xref: /netbsd-src/sys/netinet6/udp6_usrreq.c (revision 5d781b6ffb671235621ca0d7453230b9969d0374)
1 /* $NetBSD: udp6_usrreq.c,v 1.156 2024/10/08 02:30:05 riastradh Exp $ */
2 /* $KAME: udp6_usrreq.c,v 1.86 2001/05/27 17:33:00 itojun Exp $ */
3 /* $KAME: udp6_output.c,v 1.43 2001/10/15 09:19:52 itojun Exp $ */
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1989, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)udp_var.h	8.1 (Berkeley) 6/10/93
63  */
64 
65 #include <sys/cdefs.h>
66 __KERNEL_RCSID(0, "$NetBSD: udp6_usrreq.c,v 1.156 2024/10/08 02:30:05 riastradh Exp $");
67 
68 #ifdef _KERNEL_OPT
69 #include "opt_inet.h"
70 #include "opt_inet_csum.h"
71 #include "opt_ipsec.h"
72 #include "opt_net_mpsafe.h"
73 #endif
74 
75 #include <sys/param.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 #include <sys/sysctl.h>
85 
86 #include <net/if.h>
87 #include <net/if_types.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/in_offload.h>
93 #include <netinet/ip.h>
94 #include <netinet/ip_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/udp.h>
97 #include <netinet/udp_var.h>
98 #include <netinet/udp_private.h>
99 
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/ip6protosw.h>
108 #include <netinet6/scope6_var.h>
109 
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/esp.h>
113 #ifdef INET6
114 #include <netipsec/ipsec6.h>
115 #endif
116 #endif
117 
118 #include "faith.h"
119 #if defined(NFAITH) && NFAITH > 0
120 #include <net/if_faith.h>
121 #endif
122 
123 /*
124  * UDP protocol implementation.
125  * Per RFC 768, August, 1980.
126  */
127 
128 extern struct inpcbtable udbtable;
129 
130 percpu_t *udp6stat_percpu;
131 
132 /* UDP on IP6 parameters */
133 static int udp6_sendspace = 9216;	/* really max datagram size */
134 static int udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
135 					/* 40 1K datagrams */
136 
137 static void udp6_notify(struct inpcb *, int);
138 static void sysctl_net_inet6_udp6_setup(struct sysctllog **);
139 #ifdef IPSEC
140 static int udp6_espinudp(struct mbuf **, int);
141 #endif
142 
143 #ifdef UDP_CSUM_COUNTERS
144 #include <sys/device.h>
145 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
146     NULL, "udp6", "hwcsum bad");
147 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
148     NULL, "udp6", "hwcsum ok");
149 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
150     NULL, "udp6", "hwcsum data");
151 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
152     NULL, "udp6", "swcsum");
153 
154 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
155 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
156 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
157 EVCNT_ATTACH_STATIC(udp6_swcsum);
158 
159 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
160 #else
161 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
162 #endif
163 
164 void
165 udp6_init(void)
166 {
167 	sysctl_net_inet6_udp6_setup(NULL);
168 	udp6stat_percpu = percpu_alloc(sizeof(uint64_t) * UDP6_NSTATS);
169 
170 	udp_init_common();
171 }
172 
173 /*
174  * Notify a udp user of an asynchronous error;
175  * just wake up so that he can collect error status.
176  */
177 static	void
178 udp6_notify(struct inpcb *inp, int errno)
179 {
180 	inp->inp_socket->so_error = errno;
181 	sorwakeup(inp->inp_socket);
182 	sowwakeup(inp->inp_socket);
183 }
184 
185 void *
186 udp6_ctlinput(int cmd, const struct sockaddr *sa, void *d)
187 {
188 	struct udphdr uh;
189 	struct ip6_hdr *ip6;
190 	const struct sockaddr_in6 *sa6 = (const struct sockaddr_in6 *)sa;
191 	struct mbuf *m;
192 	int off;
193 	void *cmdarg;
194 	struct ip6ctlparam *ip6cp = NULL;
195 	const struct sockaddr_in6 *sa6_src = NULL;
196 	void (*notify)(struct inpcb *, int) = udp6_notify;
197 	struct udp_portonly {
198 		u_int16_t uh_sport;
199 		u_int16_t uh_dport;
200 	} *uhp;
201 
202 	if (sa->sa_family != AF_INET6 ||
203 	    sa->sa_len != sizeof(struct sockaddr_in6))
204 		return NULL;
205 
206 	if ((unsigned)cmd >= PRC_NCMDS)
207 		return NULL;
208 	if (PRC_IS_REDIRECT(cmd))
209 		notify = in6pcb_rtchange, d = NULL;
210 	else if (cmd == PRC_HOSTDEAD)
211 		d = NULL;
212 	else if (cmd == PRC_MSGSIZE) {
213 		/* special code is present, see below */
214 		notify = in6pcb_rtchange;
215 	}
216 	else if (inet6ctlerrmap[cmd] == 0)
217 		return NULL;
218 
219 	/* if the parameter is from icmp6, decode it. */
220 	if (d != NULL) {
221 		ip6cp = (struct ip6ctlparam *)d;
222 		m = ip6cp->ip6c_m;
223 		ip6 = ip6cp->ip6c_ip6;
224 		off = ip6cp->ip6c_off;
225 		cmdarg = ip6cp->ip6c_cmdarg;
226 		sa6_src = ip6cp->ip6c_src;
227 	} else {
228 		m = NULL;
229 		ip6 = NULL;
230 		cmdarg = NULL;
231 		sa6_src = &sa6_any;
232 		off = 0;
233 	}
234 
235 	if (ip6) {
236 		/* check if we can safely examine src and dst ports */
237 		if (m->m_pkthdr.len < off + sizeof(*uhp)) {
238 			if (cmd == PRC_MSGSIZE)
239 				icmp6_mtudisc_update((struct ip6ctlparam *)d, 0);
240 			return NULL;
241 		}
242 
243 		memset(&uh, 0, sizeof(uh));
244 		m_copydata(m, off, sizeof(*uhp), (void *)&uh);
245 
246 		if (cmd == PRC_MSGSIZE) {
247 			int valid = 0;
248 
249 			/*
250 			 * Check to see if we have a valid UDP socket
251 			 * corresponding to the address in the ICMPv6 message
252 			 * payload.
253 			 */
254 			if (in6pcb_lookup(&udbtable, &sa6->sin6_addr,
255 			    uh.uh_dport, (const struct in6_addr *)&sa6_src->sin6_addr,
256 			    uh.uh_sport, 0, 0))
257 				valid++;
258 #if 0
259 			/*
260 			 * As the use of sendto(2) is fairly popular,
261 			 * we may want to allow non-connected pcb too.
262 			 * But it could be too weak against attacks...
263 			 * We should at least check if the local address (= s)
264 			 * is really ours.
265 			 */
266 			else if (in6pcb_lookup_bound(&udbtable, &sa6->sin6_addr,
267 			    uh.uh_dport, 0))
268 				valid++;
269 #endif
270 
271 			/*
272 			 * Depending on the value of "valid" and routing table
273 			 * size (mtudisc_{hi,lo}wat), we will:
274 			 * - recalculate the new MTU and create the
275 			 *   corresponding routing entry, or
276 			 * - ignore the MTU change notification.
277 			 */
278 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
279 
280 			/*
281 			 * regardless of if we called
282 			 * icmp6_mtudisc_update(), we need to call
283 			 * in6pcb_notify(), to notify path MTU change
284 			 * to the userland (RFC3542), because some
285 			 * unconnected sockets may share the same
286 			 * destination and want to know the path MTU.
287 			 */
288 		}
289 
290 		(void)in6pcb_notify(&udbtable, sa, uh.uh_dport,
291 		    sin6tocsa(sa6_src), uh.uh_sport, cmd, cmdarg,
292 		    notify);
293 	} else {
294 		(void)in6pcb_notify(&udbtable, sa, 0,
295 		    sin6tocsa(sa6_src), 0, cmd, cmdarg, notify);
296 	}
297 	return NULL;
298 }
299 
300 int
301 udp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
302 {
303 	int s;
304 	int error = 0;
305 	struct inpcb *inp;
306 	int family;
307 	int optval;
308 
309 	family = so->so_proto->pr_domain->dom_family;
310 
311 	s = splsoftnet();
312 	switch (family) {
313 #ifdef INET
314 	case PF_INET:
315 		if (sopt->sopt_level != IPPROTO_UDP) {
316 			error = ip_ctloutput(op, so, sopt);
317 			goto end;
318 		}
319 		break;
320 #endif
321 #ifdef INET6
322 	case PF_INET6:
323 		if (sopt->sopt_level != IPPROTO_UDP) {
324 			error = ip6_ctloutput(op, so, sopt);
325 			goto end;
326 		}
327 		break;
328 #endif
329 	default:
330 		error = EAFNOSUPPORT;
331 		goto end;
332 	}
333 
334 	switch (op) {
335 	case PRCO_SETOPT:
336 		inp = sotoinpcb(so);
337 
338 		switch (sopt->sopt_name) {
339 		case UDP_ENCAP:
340 			error = sockopt_getint(sopt, &optval);
341 			if (error)
342 				break;
343 
344 			switch(optval) {
345 			case 0:
346 				inp->inp_flags &= ~IN6P_ESPINUDP;
347 				break;
348 
349 			case UDP_ENCAP_ESPINUDP:
350 				inp->inp_flags |= IN6P_ESPINUDP;
351 				break;
352 
353 			default:
354 				error = EINVAL;
355 				break;
356 			}
357 			break;
358 
359 		default:
360 			error = ENOPROTOOPT;
361 			break;
362 		}
363 		break;
364 
365 	default:
366 		error = EINVAL;
367 		break;
368 	}
369 
370 end:
371 	splx(s);
372 	return error;
373 }
374 
375 static void
376 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
377     struct sockaddr *src, struct socket *so)
378 {
379 	struct mbuf *opts = NULL;
380 	struct mbuf *n;
381 	struct inpcb *inp;
382 
383 	KASSERT(so != NULL);
384 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET6);
385 	inp = sotoinpcb(so);
386 	KASSERT(inp != NULL);
387 
388 #if defined(IPSEC)
389 	if (ipsec_used && ipsec_in_reject(m, inp)) {
390 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
391 			icmp6_error(n, ICMP6_DST_UNREACH,
392 			    ICMP6_DST_UNREACH_ADMIN, 0);
393 		return;
394 	}
395 #endif
396 
397 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
398 		if (inp->inp_flags & IN6P_CONTROLOPTS ||
399 		    SOOPT_TIMESTAMP(inp->inp_socket->so_options)) {
400 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
401 			ip6_savecontrol(inp, &opts, ip6, n);
402 		}
403 
404 		m_adj(n, off);
405 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
406 			m_freem(n);
407 			m_freem(opts);
408 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
409 			soroverflow(so);
410 		} else
411 			sorwakeup(so);
412 	}
413 }
414 
415 int
416 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
417     struct mbuf **mp, int off)
418 {
419 	u_int16_t sport, dport;
420 	int rcvcnt;
421 	struct in6_addr src6, *dst6;
422 	const struct in_addr *dst4;
423 	struct inpcb *inp;
424 	struct mbuf *m = *mp;
425 
426 	rcvcnt = 0;
427 	off += sizeof(struct udphdr);	/* now, offset of payload */
428 
429 	if (af != AF_INET && af != AF_INET6)
430 		goto bad;
431 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
432 		goto bad;
433 
434 	src6 = src->sin6_addr;
435 	if (sa6_recoverscope(src) != 0) {
436 		/* XXX: should be impossible. */
437 		goto bad;
438 	}
439 	sport = src->sin6_port;
440 
441 	dport = dst->sin6_port;
442 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
443 	dst6 = &dst->sin6_addr;
444 
445 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
446 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
447 		/*
448 		 * Deliver a multicast or broadcast datagram to *all* sockets
449 		 * for which the local and remote addresses and ports match
450 		 * those of the incoming datagram.  This allows more than
451 		 * one process to receive multi/broadcasts on the same port.
452 		 * (This really ought to be done for unicast datagrams as
453 		 * well, but that would cause problems with existing
454 		 * applications that open both address-specific sockets and
455 		 * a wildcard socket listening to the same port -- they would
456 		 * end up receiving duplicates of every unicast datagram.
457 		 * Those applications open the multiple sockets to overcome an
458 		 * inadequacy of the UDP socket interface, but for backwards
459 		 * compatibility we avoid the problem here rather than
460 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
461 		 */
462 
463 		/*
464 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
465 		 * we need udpiphdr for IPsec processing so we do that later.
466 		 */
467 		/*
468 		 * Locate pcb(s) for datagram.
469 		 */
470 		TAILQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
471 			if (inp->inp_af != AF_INET6)
472 				continue;
473 
474 			if (inp->inp_lport != dport)
475 				continue;
476 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) {
477 				if (!IN6_ARE_ADDR_EQUAL(&in6p_laddr(inp),
478 				    dst6))
479 					continue;
480 			} else {
481 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
482 				    (inp->inp_flags & IN6P_IPV6_V6ONLY))
483 					continue;
484 			}
485 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
486 				if (!IN6_ARE_ADDR_EQUAL(&in6p_faddr(inp),
487 				    &src6) || inp->inp_fport != sport)
488 					continue;
489 			} else {
490 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
491 				    (inp->inp_flags & IN6P_IPV6_V6ONLY))
492 					continue;
493 			}
494 
495 			udp6_sendup(m, off, sin6tosa(src), inp->inp_socket);
496 			rcvcnt++;
497 
498 			/*
499 			 * Don't look for additional matches if this one does
500 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
501 			 * socket options set.  This heuristic avoids searching
502 			 * through all pcbs in the common case of a non-shared
503 			 * port.  It assumes that an application will never
504 			 * clear these options after setting them.
505 			 */
506 			if ((inp->inp_socket->so_options &
507 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
508 				break;
509 		}
510 	} else {
511 		/*
512 		 * Locate pcb for datagram.
513 		 */
514 		inp = in6pcb_lookup(&udbtable, &src6, sport, dst6,
515 					     dport, 0, 0);
516 		if (inp == NULL) {
517 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
518 			inp = in6pcb_lookup_bound(&udbtable, dst6, dport, 0);
519 			if (inp == NULL)
520 				return rcvcnt;
521 		}
522 
523 #ifdef IPSEC
524 		/* Handle ESP over UDP */
525 		if (inp->inp_flags & IN6P_ESPINUDP) {
526 			switch (udp6_espinudp(mp, off)) {
527 			case -1: /* Error, m was freed */
528 				KASSERT(*mp == NULL);
529 				rcvcnt = -1;
530 				goto bad;
531 
532 			case 1: /* ESP over UDP */
533 				KASSERT(*mp == NULL);
534 				rcvcnt++;
535 				goto bad;
536 
537 			case 0: /* plain UDP */
538 			default: /* Unexpected */
539 				/*
540 				 * Normal UDP processing will take place,
541 				 * m may have changed.
542 				 */
543 				m = *mp;
544 				break;
545 			}
546 		}
547 #endif
548 
549 		if (inp->inp_overudp_cb != NULL) {
550 			int ret;
551 			ret = inp->inp_overudp_cb(mp, off, inp->inp_socket,
552 			    sin6tosa(src), inp->inp_overudp_arg);
553 			switch (ret) {
554 			case -1: /* Error, m was freed */
555 				KASSERT(*mp == NULL);
556 				rcvcnt = -1;
557 				goto bad;
558 
559 			case 1: /* Foo over UDP */
560 				KASSERT(*mp == NULL);
561 				rcvcnt++;
562 				goto bad;
563 
564 			case 0: /* plain UDP */
565 			default: /* Unexpected */
566 				/*
567 				 * Normal UDP processing will take place,
568 				 * m may have changed.
569 				 */
570 				m = *mp;
571 				break;
572 			}
573 		}
574 
575 		udp6_sendup(m, off, sin6tosa(src), inp->inp_socket);
576 		rcvcnt++;
577 	}
578 
579 bad:
580 	return rcvcnt;
581 }
582 
583 int
584 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
585 {
586 
587 	/*
588 	 * XXX it's better to record and check if this mbuf is
589 	 * already checked.
590 	 */
591 
592 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
593 		goto good;
594 	}
595 	if (uh->uh_sum == 0) {
596 		UDP6_STATINC(UDP6_STAT_NOSUM);
597 		goto bad;
598 	}
599 
600 	switch (m->m_pkthdr.csum_flags &
601 	    ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv6) |
602 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
603 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
604 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
605 		UDP6_STATINC(UDP6_STAT_BADSUM);
606 		goto bad;
607 
608 #if 0 /* notyet */
609 	case M_CSUM_UDPv6|M_CSUM_DATA:
610 #endif
611 
612 	case M_CSUM_UDPv6:
613 		/* Checksum was okay. */
614 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
615 		break;
616 
617 	default:
618 		/*
619 		 * Need to compute it ourselves.  Maybe skip checksum
620 		 * on loopback interfaces.
621 		 */
622 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
623 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
624 			UDP6_STATINC(UDP6_STAT_BADSUM);
625 			goto bad;
626 		}
627 	}
628 
629 good:
630 	return 0;
631 bad:
632 	return -1;
633 }
634 
635 int
636 udp6_input(struct mbuf **mp, int *offp, int proto)
637 {
638 	struct mbuf *m = *mp;
639 	int off = *offp;
640 	struct sockaddr_in6 src, dst;
641 	struct ip6_hdr *ip6;
642 	struct udphdr *uh;
643 	u_int32_t plen, ulen;
644 
645 	ip6 = mtod(m, struct ip6_hdr *);
646 
647 #if defined(NFAITH) && 0 < NFAITH
648 	if (faithprefix(&ip6->ip6_dst)) {
649 		/* send icmp6 host unreach? */
650 		m_freem(m);
651 		return IPPROTO_DONE;
652 	}
653 #endif
654 
655 	UDP6_STATINC(UDP6_STAT_IPACKETS);
656 
657 	/* Check for jumbogram is done in ip6_input. We can trust pkthdr.len. */
658 	plen = m->m_pkthdr.len - off;
659 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
660 	if (uh == NULL) {
661 		IP6_STATINC(IP6_STAT_TOOSHORT);
662 		return IPPROTO_DONE;
663 	}
664 
665 	/*
666 	 * Enforce alignment requirements that are violated in
667 	 * some cases, see kern/50766 for details.
668 	 */
669 	if (ACCESSIBLE_POINTER(uh, struct udphdr) == 0) {
670 		m = m_copyup(m, off + sizeof(struct udphdr), 0);
671 		if (m == NULL) {
672 			IP6_STATINC(IP6_STAT_TOOSHORT);
673 			return IPPROTO_DONE;
674 		}
675 		ip6 = mtod(m, struct ip6_hdr *);
676 		uh = (struct udphdr *)(mtod(m, char *) + off);
677 	}
678 	KASSERT(ACCESSIBLE_POINTER(uh, struct udphdr));
679 	ulen = ntohs((u_short)uh->uh_ulen);
680 
681 	/*
682 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
683 	 * iff payload length > 0xffff.
684 	 */
685 	if (ulen == 0 && plen > 0xffff)
686 		ulen = plen;
687 
688 	if (plen != ulen) {
689 		UDP6_STATINC(UDP6_STAT_BADLEN);
690 		goto bad;
691 	}
692 
693 	/* destination port of 0 is illegal, based on RFC768. */
694 	if (uh->uh_dport == 0)
695 		goto bad;
696 
697 	/*
698 	 * Checksum extended UDP header and data.  Maybe skip checksum
699 	 * on loopback interfaces.
700 	 */
701 	if (udp6_input_checksum(m, uh, off, ulen))
702 		goto bad;
703 
704 	/*
705 	 * Construct source and dst sockaddrs.
706 	 */
707 	memset(&src, 0, sizeof(src));
708 	src.sin6_family = AF_INET6;
709 	src.sin6_len = sizeof(struct sockaddr_in6);
710 	src.sin6_addr = ip6->ip6_src;
711 	src.sin6_port = uh->uh_sport;
712 	memset(&dst, 0, sizeof(dst));
713 	dst.sin6_family = AF_INET6;
714 	dst.sin6_len = sizeof(struct sockaddr_in6);
715 	dst.sin6_addr = ip6->ip6_dst;
716 	dst.sin6_port = uh->uh_dport;
717 
718 	if (udp6_realinput(AF_INET6, &src, &dst, &m, off) == 0) {
719 		if (m->m_flags & M_MCAST) {
720 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
721 			goto bad;
722 		}
723 		UDP6_STATINC(UDP6_STAT_NOPORT);
724 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
725 		m = NULL;
726 	}
727 
728 bad:
729 	m_freem(m);
730 	return IPPROTO_DONE;
731 }
732 
733 int
734 udp6_output(struct inpcb * const inp, struct mbuf *m,
735     struct sockaddr_in6 * const addr6, struct mbuf * const control,
736     struct lwp * const l)
737 {
738 	u_int32_t ulen = m->m_pkthdr.len;
739 	u_int32_t plen = sizeof(struct udphdr) + ulen;
740 	struct ip6_hdr *ip6;
741 	struct udphdr *udp6;
742 	struct in6_addr _laddr, *laddr, *faddr;
743 	struct in6_addr laddr_mapped; /* XXX ugly */
744 	struct sockaddr_in6 *sin6 = NULL;
745 	struct ifnet *oifp = NULL;
746 	int scope_ambiguous = 0;
747 	u_int16_t fport;
748 	int error = 0;
749 	struct ip6_pktopts *optp = NULL;
750 	struct ip6_pktopts opt;
751 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
752 #ifdef INET
753 	struct ip *ip;
754 	struct udpiphdr *ui;
755 	int flags = 0;
756 #endif
757 	struct sockaddr_in6 tmp;
758 
759 	if (addr6) {
760 		sin6 = addr6;
761 		if (sin6->sin6_len != sizeof(*sin6)) {
762 			error = EINVAL;
763 			goto release;
764 		}
765 		if (sin6->sin6_family != AF_INET6) {
766 			error = EAFNOSUPPORT;
767 			goto release;
768 		}
769 
770 		/* protect *sin6 from overwrites */
771 		tmp = *sin6;
772 		sin6 = &tmp;
773 
774 		/*
775 		 * Application should provide a proper zone ID or the use of
776 		 * default zone IDs should be enabled.  Unfortunately, some
777 		 * applications do not behave as it should, so we need a
778 		 * workaround.  Even if an appropriate ID is not determined,
779 		 * we'll see if we can determine the outgoing interface.  If we
780 		 * can, determine the zone ID based on the interface below.
781 		 */
782 		if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
783 			scope_ambiguous = 1;
784 		if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
785 			goto release;
786 	}
787 
788 	if (control) {
789 		if (__predict_false(l == NULL)) {
790 			panic("%s: control but no lwp", __func__);
791 		}
792 		if ((error = ip6_setpktopts(control, &opt,
793 		    in6p_outputopts(inp), l->l_cred, IPPROTO_UDP)) != 0)
794 			goto release;
795 		optp = &opt;
796 	} else
797 		optp = in6p_outputopts(inp);
798 
799 
800 	if (sin6) {
801 		/*
802 		 * Slightly different than v4 version in that we call
803 		 * in6_selectsrc and in6pcb_set_port to fill in the local
804 		 * address and port rather than inpcb_connect. inpcb_connect
805 		 * sets inp_faddr which causes EISCONN below to be hit on
806 		 * subsequent sendto.
807 		 */
808 		if (sin6->sin6_port == 0) {
809 			error = EADDRNOTAVAIL;
810 			goto release;
811 		}
812 
813 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
814 			/* how about ::ffff:0.0.0.0 case? */
815 			error = EISCONN;
816 			goto release;
817 		}
818 
819 		faddr = &sin6->sin6_addr;
820 		fport = sin6->sin6_port; /* allow 0 port */
821 
822 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
823 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
824 				/*
825 				 * I believe we should explicitly discard the
826 				 * packet when mapped addresses are disabled,
827 				 * rather than send the packet as an IPv6 one.
828 				 * If we chose the latter approach, the packet
829 				 * might be sent out on the wire based on the
830 				 * default route, the situation which we'd
831 				 * probably want to avoid.
832 				 * (20010421 jinmei@kame.net)
833 				 */
834 				error = EINVAL;
835 				goto release;
836 			}
837 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp)) &&
838 			    !IN6_IS_ADDR_V4MAPPED(&in6p_laddr(inp))) {
839 				/*
840 				 * when remote addr is an IPv4-mapped address,
841 				 * local addr should not be an IPv6 address,
842 				 * since you cannot determine how to map IPv6
843 				 * source address to IPv4.
844 				 */
845 				error = EINVAL;
846 				goto release;
847 			}
848 
849 			af = AF_INET;
850 		}
851 
852 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
853 			struct psref psref;
854 			int bound = curlwp_bind();
855 
856 			error = in6_selectsrc(sin6, optp,
857 			    in6p_moptions(inp),
858 			    &inp->inp_route,
859 			    &in6p_laddr(inp), &oifp, &psref, &_laddr);
860 			if (error)
861 				laddr = NULL;
862 			else
863 				laddr = &_laddr;
864 			if (oifp && scope_ambiguous &&
865 			    (error = in6_setscope(&sin6->sin6_addr,
866 			    oifp, NULL))) {
867 				if_put(oifp, &psref);
868 				curlwp_bindx(bound);
869 				goto release;
870 			}
871 			if_put(oifp, &psref);
872 			curlwp_bindx(bound);
873 		} else {
874 			/*
875 			 * XXX: freebsd[34] does not have in_selectsrc, but
876 			 * we can omit the whole part because freebsd4 calls
877 			 * udp_output() directly in this case, and thus we'll
878 			 * never see this path.
879 			 */
880 			if (IN6_IS_ADDR_UNSPECIFIED(&in6p_laddr(inp))) {
881 				struct sockaddr_in sin_dst;
882 				struct in_addr ina;
883 				struct in_ifaddr *ia4;
884 				struct psref _psref;
885 				int bound;
886 
887 				memcpy(&ina, &faddr->s6_addr[12], sizeof(ina));
888 				sockaddr_in_init(&sin_dst, &ina, 0);
889 				bound = curlwp_bind();
890 				ia4 = in_selectsrc(&sin_dst, &inp->inp_route,
891 				    inp->inp_socket->so_options, NULL,
892 				    &error, &_psref);
893 				if (ia4 == NULL) {
894 					curlwp_bindx(bound);
895 					if (error == 0)
896 						error = EADDRNOTAVAIL;
897 					goto release;
898 				}
899 				memset(&laddr_mapped, 0, sizeof(laddr_mapped));
900 				laddr_mapped.s6_addr16[5] = 0xffff; /* ugly */
901 				memcpy(&laddr_mapped.s6_addr[12],
902 				      &IA_SIN(ia4)->sin_addr,
903 				      sizeof(IA_SIN(ia4)->sin_addr));
904 				ia4_release(ia4, &_psref);
905 				curlwp_bindx(bound);
906 				laddr = &laddr_mapped;
907 			} else
908 			{
909 				laddr = &in6p_laddr(inp);	/* XXX */
910 			}
911 		}
912 		if (laddr == NULL) {
913 			if (error == 0)
914 				error = EADDRNOTAVAIL;
915 			goto release;
916 		}
917 		if (inp->inp_lport == 0) {
918 			/*
919 			 * Craft a sockaddr_in6 for the local endpoint. Use the
920 			 * "any" as a base, set the address, and recover the
921 			 * scope.
922 			 */
923 			struct sockaddr_in6 lsin6 =
924 			    *((const struct sockaddr_in6 *)inp->inp_socket->so_proto->pr_domain->dom_sa_any);
925 			lsin6.sin6_addr = *laddr;
926 			error = sa6_recoverscope(&lsin6);
927 			if (error)
928 				goto release;
929 
930 			error = in6pcb_set_port(&lsin6, inp, l);
931 
932 			if (error) {
933 				in6p_laddr(inp) = in6addr_any;
934 				goto release;
935 			}
936 		}
937 	} else {
938 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp))) {
939 			error = ENOTCONN;
940 			goto release;
941 		}
942 		if (IN6_IS_ADDR_V4MAPPED(&in6p_faddr(inp))) {
943 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY))
944 			{
945 				/*
946 				 * XXX: this case would happen when the
947 				 * application sets the V6ONLY flag after
948 				 * connecting the foreign address.
949 				 * Such applications should be fixed,
950 				 * so we bark here.
951 				 */
952 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
953 				    "option was set for a connected socket\n");
954 				error = EINVAL;
955 				goto release;
956 			} else
957 				af = AF_INET;
958 		}
959 		laddr = &in6p_laddr(inp);
960 		faddr = &in6p_faddr(inp);
961 		fport = inp->inp_fport;
962 	}
963 
964 	if (af == AF_INET)
965 		hlen = sizeof(struct ip);
966 
967 	/*
968 	 * Calculate data length and get a mbuf
969 	 * for UDP and IP6 headers.
970 	 */
971 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
972 	if (m == NULL) {
973 		error = ENOBUFS;
974 		goto release;
975 	}
976 
977 	/*
978 	 * Stuff checksum and output datagram.
979 	 */
980 	udp6 = (struct udphdr *)(mtod(m, char *) + hlen);
981 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
982 	udp6->uh_dport = fport;
983 	if (plen <= 0xffff)
984 		udp6->uh_ulen = htons((u_int16_t)plen);
985 	else
986 		udp6->uh_ulen = 0;
987 	udp6->uh_sum = 0;
988 
989 	switch (af) {
990 	case AF_INET6:
991 		ip6 = mtod(m, struct ip6_hdr *);
992 		ip6->ip6_flow	= in6p_flowinfo(inp) & IPV6_FLOWINFO_MASK;
993 		ip6->ip6_vfc 	&= ~IPV6_VERSION_MASK;
994 		ip6->ip6_vfc 	|= IPV6_VERSION;
995 #if 0		/* ip6_plen will be filled in ip6_output. */
996 		ip6->ip6_plen	= htons((u_int16_t)plen);
997 #endif
998 		ip6->ip6_nxt	= IPPROTO_UDP;
999 		ip6->ip6_hlim	= in6pcb_selecthlim_rt(inp);
1000 		ip6->ip6_src	= *laddr;
1001 		ip6->ip6_dst	= *faddr;
1002 
1003 		udp6->uh_sum = in6_cksum_phdr(laddr, faddr,
1004 		    htonl(plen), htonl(IPPROTO_UDP));
1005 		m->m_pkthdr.csum_flags = M_CSUM_UDPv6;
1006 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1007 
1008 		UDP6_STATINC(UDP6_STAT_OPACKETS);
1009 		error = ip6_output(m, optp, &inp->inp_route, 0,
1010 		    in6p_moptions(inp), inp, NULL);
1011 		break;
1012 	case AF_INET:
1013 #ifdef INET
1014 		/* can't transmit jumbogram over IPv4 */
1015 		if (plen > 0xffff) {
1016 			error = EMSGSIZE;
1017 			goto release;
1018 		}
1019 
1020 		ip = mtod(m, struct ip *);
1021 		ui = (struct udpiphdr *)ip;
1022 		memset(ui->ui_x1, 0, sizeof(ui->ui_x1));
1023 		ui->ui_pr = IPPROTO_UDP;
1024 		ui->ui_len = htons(plen);
1025 		memcpy(&ui->ui_src, &laddr->s6_addr[12], sizeof(ui->ui_src));
1026 		ui->ui_ulen = ui->ui_len;
1027 
1028 		flags = (inp->inp_socket->so_options &
1029 			 (SO_DONTROUTE | SO_BROADCAST));
1030 		memcpy(&ui->ui_dst, &faddr->s6_addr[12], sizeof(ui->ui_dst));
1031 
1032 		udp6->uh_sum = in_cksum(m, hlen + plen);
1033 		if (udp6->uh_sum == 0)
1034 			udp6->uh_sum = 0xffff;
1035 
1036 		ip->ip_len = htons(hlen + plen);
1037 		ip->ip_ttl = in6pcb_selecthlim(inp, NULL); /* XXX */
1038 		ip->ip_tos = 0;	/* XXX */
1039 
1040 		UDP_STATINC(UDP_STAT_OPACKETS);
1041 		error = ip_output(m, NULL, &inp->inp_route, flags /* XXX */,
1042 		    inp->inp_moptions, NULL);
1043 		break;
1044 #else
1045 		error = EAFNOSUPPORT;
1046 		goto release;
1047 #endif
1048 	}
1049 	goto releaseopt;
1050 
1051 release:
1052 	m_freem(m);
1053 
1054 releaseopt:
1055 	if (control) {
1056 		if (optp == &opt)
1057 			ip6_clearpktopts(&opt, -1);
1058 		m_freem(control);
1059 	}
1060 	return (error);
1061 }
1062 
1063 static int
1064 udp6_attach(struct socket *so, int proto)
1065 {
1066 	struct inpcb *inp;
1067 	int s, error;
1068 
1069 	KASSERT(sotoinpcb(so) == NULL);
1070 	sosetlock(so);
1071 
1072 	error = soreserve(so, udp6_sendspace, udp6_recvspace);
1073 	if (error) {
1074 		return error;
1075 	}
1076 
1077 	/*
1078 	 * MAPPED_ADDR implementation spec:
1079 	 *  Always attach for IPv6, and only when necessary for IPv4.
1080 	 */
1081 	s = splsoftnet();
1082 	error = inpcb_create(so, &udbtable);
1083 	splx(s);
1084 	if (error) {
1085 		return error;
1086 	}
1087 
1088 	inp = sotoinpcb(so);
1089 	in6p_cksum(inp) = -1;	/* just to be sure */
1090 
1091 	KASSERT(solocked(so));
1092 	return 0;
1093 }
1094 
1095 static void
1096 udp6_detach(struct socket *so)
1097 {
1098 	struct inpcb *inp = sotoinpcb(so);
1099 	int s;
1100 
1101 	KASSERT(solocked(so));
1102 	KASSERT(inp != NULL);
1103 
1104 	s = splsoftnet();
1105 	inpcb_destroy(inp);
1106 	splx(s);
1107 }
1108 
1109 static int
1110 udp6_accept(struct socket *so, struct sockaddr *nam)
1111 {
1112 	KASSERT(solocked(so));
1113 
1114 	return EOPNOTSUPP;
1115 }
1116 
1117 static int
1118 udp6_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
1119 {
1120 	struct inpcb *inp = sotoinpcb(so);
1121 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1122 	int error = 0;
1123 	int s;
1124 
1125 	KASSERT(solocked(so));
1126 	KASSERT(inp != NULL);
1127 
1128 	s = splsoftnet();
1129 	error = in6pcb_bind(inp, sin6, l);
1130 	splx(s);
1131 	return error;
1132 }
1133 
1134 static int
1135 udp6_listen(struct socket *so, struct lwp *l)
1136 {
1137 	KASSERT(solocked(so));
1138 
1139 	return EOPNOTSUPP;
1140 }
1141 
1142 static int
1143 udp6_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
1144 {
1145 	struct inpcb *inp = sotoinpcb(so);
1146 	int error = 0;
1147 	int s;
1148 
1149 	KASSERT(solocked(so));
1150 	KASSERT(inp != NULL);
1151 
1152 	if (!IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp)))
1153 		return EISCONN;
1154 	s = splsoftnet();
1155 	error = in6pcb_connect(inp, (struct sockaddr_in6 *)nam, l);
1156 	splx(s);
1157 	if (error == 0)
1158 		soisconnected(so);
1159 
1160 	return error;
1161 }
1162 
1163 static int
1164 udp6_connect2(struct socket *so, struct socket *so2)
1165 {
1166 	KASSERT(solocked(so));
1167 
1168 	return EOPNOTSUPP;
1169 }
1170 
1171 static int
1172 udp6_disconnect(struct socket *so)
1173 {
1174 	struct inpcb *inp = sotoinpcb(so);
1175 	int s;
1176 
1177 	KASSERT(solocked(so));
1178 	KASSERT(inp != NULL);
1179 
1180 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p_faddr(inp)))
1181 		return ENOTCONN;
1182 
1183 	s = splsoftnet();
1184 	in6pcb_disconnect(inp);
1185 	memset((void *)&in6p_laddr(inp), 0, sizeof(in6p_laddr(inp)));
1186 	splx(s);
1187 
1188 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
1189 	in6pcb_set_state(inp, INP_BOUND);		/* XXX */
1190 	return 0;
1191 }
1192 
1193 static int
1194 udp6_shutdown(struct socket *so)
1195 {
1196 	int s;
1197 
1198 	s = splsoftnet();
1199 	socantsendmore(so);
1200 	splx(s);
1201 
1202 	return 0;
1203 }
1204 
1205 static int
1206 udp6_abort(struct socket *so)
1207 {
1208 	int s;
1209 
1210 	KASSERT(solocked(so));
1211 	KASSERT(sotoinpcb(so) != NULL);
1212 
1213 	s = splsoftnet();
1214 	soisdisconnected(so);
1215 	inpcb_destroy(sotoinpcb(so));
1216 	splx(s);
1217 
1218 	return 0;
1219 }
1220 
1221 static int
1222 udp6_ioctl(struct socket *so, u_long cmd, void *addr6, struct ifnet *ifp)
1223 {
1224 	/*
1225 	 * MAPPED_ADDR implementation info:
1226 	 *  Mapped addr support for PRU_CONTROL is not necessary.
1227 	 *  Because typical user of PRU_CONTROL is such as ifconfig,
1228 	 *  and they don't associate any addr to their socket.  Then
1229 	 *  socket family is only hint about the PRU_CONTROL'ed address
1230 	 *  family, especially when getting addrs from kernel.
1231 	 *  So AF_INET socket need to be used to control AF_INET addrs,
1232 	 *  and AF_INET6 socket for AF_INET6 addrs.
1233 	 */
1234 	return in6_control(so, cmd, addr6, ifp);
1235 }
1236 
1237 static int
1238 udp6_stat(struct socket *so, struct stat *ub)
1239 {
1240 	KASSERT(solocked(so));
1241 
1242 	/* stat: don't bother with a blocksize */
1243 	return 0;
1244 }
1245 
1246 static int
1247 udp6_peeraddr(struct socket *so, struct sockaddr *nam)
1248 {
1249 	KASSERT(solocked(so));
1250 	KASSERT(sotoinpcb(so) != NULL);
1251 	KASSERT(nam != NULL);
1252 
1253 	in6pcb_fetch_peeraddr(sotoinpcb(so), (struct sockaddr_in6 *)nam);
1254 	return 0;
1255 }
1256 
1257 static int
1258 udp6_sockaddr(struct socket *so, struct sockaddr *nam)
1259 {
1260 	KASSERT(solocked(so));
1261 	KASSERT(sotoinpcb(so) != NULL);
1262 	KASSERT(nam != NULL);
1263 
1264 	in6pcb_fetch_sockaddr(sotoinpcb(so), (struct sockaddr_in6 *)nam);
1265 	return 0;
1266 }
1267 
1268 static int
1269 udp6_rcvd(struct socket *so, int flags, struct lwp *l)
1270 {
1271 	KASSERT(solocked(so));
1272 
1273 	return EOPNOTSUPP;
1274 }
1275 
1276 static int
1277 udp6_recvoob(struct socket *so, struct mbuf *m, int flags)
1278 {
1279 	KASSERT(solocked(so));
1280 
1281 	return EOPNOTSUPP;
1282 }
1283 
1284 static int
1285 udp6_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
1286     struct mbuf *control, struct lwp *l)
1287 {
1288 	struct inpcb *inp = sotoinpcb(so);
1289 	int error = 0;
1290 	int s;
1291 
1292 	KASSERT(solocked(so));
1293 	KASSERT(inp != NULL);
1294 	KASSERT(m != NULL);
1295 
1296 	s = splsoftnet();
1297 	error = udp6_output(inp, m, (struct sockaddr_in6 *)nam, control, l);
1298 	splx(s);
1299 
1300 	return error;
1301 }
1302 
1303 static int
1304 udp6_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
1305 {
1306 	KASSERT(solocked(so));
1307 
1308 	m_freem(m);
1309 	m_freem(control);
1310 
1311 	return EOPNOTSUPP;
1312 }
1313 
1314 static int
1315 udp6_purgeif(struct socket *so, struct ifnet *ifp)
1316 {
1317 
1318 	mutex_enter(softnet_lock);
1319 	in6pcb_purgeif0(&udbtable, ifp);
1320 #ifdef NET_MPSAFE
1321 	mutex_exit(softnet_lock);
1322 #endif
1323 	in6_purgeif(ifp);
1324 #ifdef NET_MPSAFE
1325 	mutex_enter(softnet_lock);
1326 #endif
1327 	in6pcb_purgeif(&udbtable, ifp);
1328 	mutex_exit(softnet_lock);
1329 
1330 	return 0;
1331 }
1332 
1333 static int
1334 sysctl_net_inet6_udp6_stats(SYSCTLFN_ARGS)
1335 {
1336 
1337 	return (NETSTAT_SYSCTL(udp6stat_percpu, UDP6_NSTATS));
1338 }
1339 
1340 static void
1341 sysctl_net_inet6_udp6_setup(struct sysctllog **clog)
1342 {
1343 
1344 	sysctl_createv(clog, 0, NULL, NULL,
1345 		       CTLFLAG_PERMANENT,
1346 		       CTLTYPE_NODE, "inet6", NULL,
1347 		       NULL, 0, NULL, 0,
1348 		       CTL_NET, PF_INET6, CTL_EOL);
1349 	sysctl_createv(clog, 0, NULL, NULL,
1350 		       CTLFLAG_PERMANENT,
1351 		       CTLTYPE_NODE, "udp6",
1352 		       SYSCTL_DESCR("UDPv6 related settings"),
1353 		       NULL, 0, NULL, 0,
1354 		       CTL_NET, PF_INET6, IPPROTO_UDP, CTL_EOL);
1355 
1356 	sysctl_createv(clog, 0, NULL, NULL,
1357 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1358 		       CTLTYPE_INT, "sendspace",
1359 		       SYSCTL_DESCR("Default UDP send buffer size"),
1360 		       NULL, 0, &udp6_sendspace, 0,
1361 		       CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_SENDSPACE,
1362 		       CTL_EOL);
1363 	sysctl_createv(clog, 0, NULL, NULL,
1364 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1365 		       CTLTYPE_INT, "recvspace",
1366 		       SYSCTL_DESCR("Default UDP receive buffer size"),
1367 		       NULL, 0, &udp6_recvspace, 0,
1368 		       CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_RECVSPACE,
1369 		       CTL_EOL);
1370 	sysctl_createv(clog, 0, NULL, NULL,
1371 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1372 		       CTLTYPE_INT, "do_loopback_cksum",
1373 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
1374 		       NULL, 0, &udp_do_loopback_cksum, 0,
1375 		       CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_LOOPBACKCKSUM,
1376 		       CTL_EOL);
1377 	sysctl_createv(clog, 0, NULL, NULL,
1378 		       CTLFLAG_PERMANENT,
1379 		       CTLTYPE_STRUCT, "pcblist",
1380 		       SYSCTL_DESCR("UDP protocol control block list"),
1381 		       sysctl_inpcblist, 0, &udbtable, 0,
1382 		       CTL_NET, PF_INET6, IPPROTO_UDP, CTL_CREATE,
1383 		       CTL_EOL);
1384 	sysctl_createv(clog, 0, NULL, NULL,
1385 		       CTLFLAG_PERMANENT,
1386 		       CTLTYPE_STRUCT, "stats",
1387 		       SYSCTL_DESCR("UDPv6 statistics"),
1388 		       sysctl_net_inet6_udp6_stats, 0, NULL, 0,
1389 		       CTL_NET, PF_INET6, IPPROTO_UDP, UDP6CTL_STATS,
1390 		       CTL_EOL);
1391 }
1392 
1393 void
1394 udp6_statinc(u_int stat)
1395 {
1396 
1397 	KASSERT(stat < UDP6_NSTATS);
1398 	UDP6_STATINC(stat);
1399 }
1400 
1401 #ifdef IPSEC
1402 /*
1403  * Returns:
1404  *     1 if the packet was processed
1405  *     0 if normal UDP processing should take place
1406  *    -1 if an error occurred and m was freed
1407  */
1408 static int
1409 udp6_espinudp(struct mbuf **mp, int off)
1410 {
1411 	const size_t skip = sizeof(struct udphdr);
1412 	size_t len;
1413 	void *data;
1414 	size_t minlen;
1415 	int ip6hdrlen;
1416 	struct ip6_hdr *ip6;
1417 	struct m_tag *tag;
1418 	struct udphdr *udphdr;
1419 	u_int16_t sport, dport;
1420 	struct mbuf *m = *mp;
1421 	uint32_t *marker;
1422 
1423 	/*
1424 	 * Collapse the mbuf chain if the first mbuf is too short
1425 	 * The longest case is: UDP + non ESP marker + ESP
1426 	 */
1427 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
1428 	if (minlen > m->m_pkthdr.len)
1429 		minlen = m->m_pkthdr.len;
1430 
1431 	if (m->m_len < minlen) {
1432 		if ((*mp = m_pullup(m, minlen)) == NULL) {
1433 			return -1; /* dropped */
1434 		}
1435 		m = *mp;
1436 	}
1437 
1438 	len = m->m_len - off;
1439 	data = mtod(m, char *) + off;
1440 
1441 	/* Ignore keepalive packets */
1442 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
1443 		m_freem(m);
1444 		*mp = NULL; /* avoid any further processing by caller ... */
1445 		return 1;	/* consumed */
1446 	}
1447 
1448 	/* Handle Non-ESP marker (32bit). If zero, then IKE. */
1449 	marker = (uint32_t *)data;
1450 	if (len <= sizeof(uint32_t))
1451 		return 0;	/* passthrough */
1452 	if (marker[0] == 0)
1453 		return 0;	/* passthrough */
1454 
1455 	/*
1456 	 * Get the UDP ports. They are handled in network
1457 	 * order everywhere in IPSEC_NAT_T code.
1458 	 */
1459 	udphdr = (struct udphdr *)((char *)data - skip);
1460 	sport = udphdr->uh_sport;
1461 	dport = udphdr->uh_dport;
1462 
1463 	/*
1464 	 * Remove the UDP header (and possibly the non ESP marker)
1465 	 * IPv6 header length is ip6hdrlen
1466 	 * Before:
1467 	 *   <---- off --->
1468 	 *   +-----+------+-----+
1469 	 *   | IP6 |  UDP | ESP |
1470 	 *   +-----+------+-----+
1471 	 *         <-skip->
1472 	 * After:
1473 	 *          +-----+-----+
1474 	 *          | IP6 | ESP |
1475 	 *          +-----+-----+
1476 	 *   <-skip->
1477 	 */
1478 	ip6hdrlen = off - sizeof(struct udphdr);
1479 	memmove(mtod(m, char *) + skip, mtod(m, void *), ip6hdrlen);
1480 	m_adj(m, skip);
1481 
1482 	ip6 = mtod(m, struct ip6_hdr *);
1483 	ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) - skip);
1484 	ip6->ip6_nxt = IPPROTO_ESP;
1485 
1486 	/*
1487 	 * We have modified the packet - it is now ESP, so we should not
1488 	 * return to UDP processing ...
1489 	 *
1490 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
1491 	 * the source UDP port. This is required if we want
1492 	 * to select the right SPD for multiple hosts behind
1493 	 * same NAT
1494 	 */
1495 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
1496 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
1497 		m_freem(m);
1498 		*mp = NULL;
1499 		return -1;	/* dropped */
1500 	}
1501 	((u_int16_t *)(tag + 1))[0] = sport;
1502 	((u_int16_t *)(tag + 1))[1] = dport;
1503 	m_tag_prepend(m, tag);
1504 
1505 	if (ipsec_used)
1506 		ipsec6_common_input(&m, &ip6hdrlen, IPPROTO_ESP);
1507 	else
1508 		m_freem(m);
1509 
1510 	/* We handled it, it shouldn't be handled by UDP */
1511 	*mp = NULL; /* avoid free by caller ... */
1512 	return 1;		/* consumed */
1513 }
1514 #endif /* IPSEC */
1515 
1516 PR_WRAP_USRREQS(udp6)
1517 #define	udp6_attach	udp6_attach_wrapper
1518 #define	udp6_detach	udp6_detach_wrapper
1519 #define	udp6_accept	udp6_accept_wrapper
1520 #define	udp6_bind	udp6_bind_wrapper
1521 #define	udp6_listen	udp6_listen_wrapper
1522 #define	udp6_connect	udp6_connect_wrapper
1523 #define	udp6_connect2	udp6_connect2_wrapper
1524 #define	udp6_disconnect	udp6_disconnect_wrapper
1525 #define	udp6_shutdown	udp6_shutdown_wrapper
1526 #define	udp6_abort	udp6_abort_wrapper
1527 #define	udp6_ioctl	udp6_ioctl_wrapper
1528 #define	udp6_stat	udp6_stat_wrapper
1529 #define	udp6_peeraddr	udp6_peeraddr_wrapper
1530 #define	udp6_sockaddr	udp6_sockaddr_wrapper
1531 #define	udp6_rcvd	udp6_rcvd_wrapper
1532 #define	udp6_recvoob	udp6_recvoob_wrapper
1533 #define	udp6_send	udp6_send_wrapper
1534 #define	udp6_sendoob	udp6_sendoob_wrapper
1535 #define	udp6_purgeif	udp6_purgeif_wrapper
1536 
1537 const struct pr_usrreqs udp6_usrreqs = {
1538 	.pr_attach	= udp6_attach,
1539 	.pr_detach	= udp6_detach,
1540 	.pr_accept	= udp6_accept,
1541 	.pr_bind	= udp6_bind,
1542 	.pr_listen	= udp6_listen,
1543 	.pr_connect	= udp6_connect,
1544 	.pr_connect2	= udp6_connect2,
1545 	.pr_disconnect	= udp6_disconnect,
1546 	.pr_shutdown	= udp6_shutdown,
1547 	.pr_abort	= udp6_abort,
1548 	.pr_ioctl	= udp6_ioctl,
1549 	.pr_stat	= udp6_stat,
1550 	.pr_peeraddr	= udp6_peeraddr,
1551 	.pr_sockaddr	= udp6_sockaddr,
1552 	.pr_rcvd	= udp6_rcvd,
1553 	.pr_recvoob	= udp6_recvoob,
1554 	.pr_send	= udp6_send,
1555 	.pr_sendoob	= udp6_sendoob,
1556 	.pr_purgeif	= udp6_purgeif,
1557 };
1558