xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision f4f8f02054f3abb6ceb84aefcdecc78d5c8b462f)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
30  *	$KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
31  */
32 
33 /*-
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
35  *	The Regents of the University of California.
36  * 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  * 4. 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_usrreq.c	8.6 (Berkeley) 5/23/95
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_mac.h"
72 
73 #include <sys/param.h>
74 #include <sys/jail.h>
75 #include <sys/kernel.h>
76 #include <sys/lock.h>
77 #include <sys/mbuf.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/signalvar.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/sx.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 #include <sys/systm.h>
88 #include <sys/vimage.h>
89 
90 #include <net/if.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/ip6.h>
101 #include <netinet/icmp_var.h>
102 #include <netinet/icmp6.h>
103 #include <netinet/ip_var.h>
104 #include <netinet/udp.h>
105 #include <netinet/udp_var.h>
106 #include <netinet/vinet.h>
107 
108 #include <netinet6/ip6protosw.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/in6_pcb.h>
111 #include <netinet6/udp6_var.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/vinet6.h>
114 
115 #ifdef IPSEC
116 #include <netipsec/ipsec.h>
117 #include <netipsec/ipsec6.h>
118 #endif /* IPSEC */
119 
120 #include <security/mac/mac_framework.h>
121 
122 /*
123  * UDP protocol implementation.
124  * Per RFC 768, August, 1980.
125  */
126 
127 extern struct protosw	inetsw[];
128 static void		udp6_detach(struct socket *so);
129 
130 static void
131 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
132     struct sockaddr_in6 *fromsa)
133 {
134 	INIT_VNET_INET(inp->inp_vnet);
135 	struct socket *so;
136 	struct mbuf *opts;
137 
138 	INP_LOCK_ASSERT(inp);
139 
140 #ifdef IPSEC
141 	/* Check AH/ESP integrity. */
142 	if (ipsec6_in_reject(n, inp)) {
143 		INIT_VNET_IPSEC(inp->inp_vnet);
144 		m_freem(n);
145 		V_ipsec6stat.in_polvio++;
146 		return;
147 	}
148 #endif /* IPSEC */
149 #ifdef MAC
150 	if (mac_inpcb_check_deliver(inp, n) != 0) {
151 		m_freem(n);
152 		return;
153 	}
154 #endif
155 	opts = NULL;
156 	if (inp->inp_flags & INP_CONTROLOPTS ||
157 	    inp->inp_socket->so_options & SO_TIMESTAMP)
158 		ip6_savecontrol(inp, n, &opts);
159 	m_adj(n, off + sizeof(struct udphdr));
160 
161 	so = inp->inp_socket;
162 	SOCKBUF_LOCK(&so->so_rcv);
163 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
164 	    opts) == 0) {
165 		SOCKBUF_UNLOCK(&so->so_rcv);
166 		m_freem(n);
167 		if (opts)
168 			m_freem(opts);
169 		UDPSTAT_INC(udps_fullsock);
170 	} else
171 		sorwakeup_locked(so);
172 }
173 
174 int
175 udp6_input(struct mbuf **mp, int *offp, int proto)
176 {
177 	INIT_VNET_INET(curvnet);
178 	INIT_VNET_INET6(curvnet);
179 	struct mbuf *m = *mp;
180 	struct ifnet *ifp;
181 	struct ip6_hdr *ip6;
182 	struct udphdr *uh;
183 	struct inpcb *inp;
184 	int off = *offp;
185 	int plen, ulen;
186 	struct sockaddr_in6 fromsa;
187 
188 	ifp = m->m_pkthdr.rcvif;
189 	ip6 = mtod(m, struct ip6_hdr *);
190 
191 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
192 		/* XXX send icmp6 host/port unreach? */
193 		m_freem(m);
194 		return (IPPROTO_DONE);
195 	}
196 
197 #ifndef PULLDOWN_TEST
198 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
199 	ip6 = mtod(m, struct ip6_hdr *);
200 	uh = (struct udphdr *)((caddr_t)ip6 + off);
201 #else
202 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
203 	if (!uh)
204 		return (IPPROTO_DONE);
205 #endif
206 
207 	UDPSTAT_INC(udps_ipackets);
208 
209 	/*
210 	 * Destination port of 0 is illegal, based on RFC768.
211 	 */
212 	if (uh->uh_dport == 0)
213 		goto badunlocked;
214 
215 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
216 	ulen = ntohs((u_short)uh->uh_ulen);
217 
218 	if (plen != ulen) {
219 		UDPSTAT_INC(udps_badlen);
220 		goto badunlocked;
221 	}
222 
223 	/*
224 	 * Checksum extended UDP header and data.
225 	 */
226 	if (uh->uh_sum == 0) {
227 		UDPSTAT_INC(udps_nosum);
228 		goto badunlocked;
229 	}
230 	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
231 		UDPSTAT_INC(udps_badsum);
232 		goto badunlocked;
233 	}
234 
235 	/*
236 	 * Construct sockaddr format source address.
237 	 */
238 	init_sin6(&fromsa, m);
239 	fromsa.sin6_port = uh->uh_sport;
240 
241 	INP_INFO_RLOCK(&V_udbinfo);
242 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
243 		struct inpcb *last;
244 		struct ip6_moptions *imo;
245 
246 		/*
247 		 * In the event that laddr should be set to the link-local
248 		 * address (this happens in RIPng), the multicast address
249 		 * specified in the received packet will not match laddr.  To
250 		 * handle this situation, matching is relaxed if the
251 		 * receiving interface is the same as one specified in the
252 		 * socket and if the destination multicast address matches
253 		 * one of the multicast groups specified in the socket.
254 		 */
255 
256 		/*
257 		 * KAME note: traditionally we dropped udpiphdr from mbuf
258 		 * here.  We need udphdr for IPsec processing so we do that
259 		 * later.
260 		 */
261 		last = NULL;
262 		LIST_FOREACH(inp, &V_udb, inp_list) {
263 			if ((inp->inp_vflag & INP_IPV6) == 0)
264 				continue;
265 			if (inp->inp_lport != uh->uh_dport)
266 				continue;
267 			if (inp->inp_fport != 0 &&
268 			    inp->inp_fport != uh->uh_sport)
269 				continue;
270 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
271 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
272 							&ip6->ip6_dst))
273 					continue;
274 			}
275 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
276 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
277 							&ip6->ip6_src) ||
278 				    inp->inp_fport != uh->uh_sport)
279 					continue;
280 			}
281 
282 			/*
283 			 * Handle socket delivery policy for any-source
284 			 * and source-specific multicast. [RFC3678]
285 			 */
286 			imo = inp->in6p_moptions;
287 			if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
288 				struct sockaddr_in6	 mcaddr;
289 				int			 blocked;
290 
291 				INP_RLOCK(inp);
292 
293 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
294 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
295 				mcaddr.sin6_family = AF_INET6;
296 				mcaddr.sin6_addr = ip6->ip6_dst;
297 
298 				blocked = im6o_mc_filter(imo, ifp,
299 					(struct sockaddr *)&mcaddr,
300 					(struct sockaddr *)&fromsa);
301 				if (blocked != MCAST_PASS) {
302 					if (blocked == MCAST_NOTGMEMBER)
303 						IP6STAT_INC(ip6s_notmember);
304 					if (blocked == MCAST_NOTSMEMBER ||
305 					    blocked == MCAST_MUTED)
306 						UDPSTAT_INC(udps_filtermcast);
307 					INP_RUNLOCK(inp); /* XXX */
308 					continue;
309 				}
310 
311 				INP_RUNLOCK(inp);
312 			}
313 			if (last != NULL) {
314 				struct mbuf *n;
315 
316 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
317 					INP_RLOCK(last);
318 					if (last->inp_ppcb != NULL) {
319 						/*
320 						 * Engage the tunneling
321 						 * protocol we will have to
322 						 * leave the info_lock up,
323 						 * since we are hunting
324 						 * through multiple UDP's.
325 						 *
326 						 */
327 						udp_tun_func_t tunnel_func;
328 
329 						tunnel_func = (udp_tun_func_t)last->inp_ppcb;
330 						tunnel_func(n, off, last);
331 						INP_RUNLOCK(last);
332 					} else {
333 						udp6_append(last, n, off, &fromsa);
334 						INP_RUNLOCK(last);
335 					}
336 				}
337 			}
338 			last = inp;
339 			/*
340 			 * Don't look for additional matches if this one does
341 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
342 			 * socket options set.  This heuristic avoids
343 			 * searching through all pcbs in the common case of a
344 			 * non-shared port.  It assumes that an application
345 			 * will never clear these options after setting them.
346 			 */
347 			if ((last->inp_socket->so_options &
348 			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
349 				break;
350 		}
351 
352 		if (last == NULL) {
353 			/*
354 			 * No matching pcb found; discard datagram.  (No need
355 			 * to send an ICMP Port Unreachable for a broadcast
356 			 * or multicast datgram.)
357 			 */
358 			UDPSTAT_INC(udps_noport);
359 			UDPSTAT_INC(udps_noportmcast);
360 			goto badheadlocked;
361 		}
362 		INP_RLOCK(last);
363 		INP_INFO_RUNLOCK(&V_udbinfo);
364 		if (last->inp_ppcb != NULL) {
365 			/*
366 			 * Engage the tunneling protocol.
367 			 */
368 			udp_tun_func_t tunnel_func;
369 
370 			tunnel_func = (udp_tun_func_t)inp->inp_ppcb;
371 			tunnel_func(m, off, last);
372 			INP_RUNLOCK(last);
373 			return (IPPROTO_DONE);
374 		}
375 		udp6_append(last, m, off, &fromsa);
376 		INP_RUNLOCK(last);
377 		return (IPPROTO_DONE);
378 	}
379 	/*
380 	 * Locate pcb for datagram.
381 	 */
382 	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
383 	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
384 	if (inp == NULL) {
385 		if (udp_log_in_vain) {
386 			char ip6bufs[INET6_ADDRSTRLEN];
387 			char ip6bufd[INET6_ADDRSTRLEN];
388 
389 			log(LOG_INFO,
390 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
391 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
392 			    ntohs(uh->uh_dport),
393 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
394 			    ntohs(uh->uh_sport));
395 		}
396 		UDPSTAT_INC(udps_noport);
397 		if (m->m_flags & M_MCAST) {
398 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
399 			UDPSTAT_INC(udps_noportmcast);
400 			goto badheadlocked;
401 		}
402 		INP_INFO_RUNLOCK(&V_udbinfo);
403 		if (V_udp_blackhole)
404 			goto badunlocked;
405 		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
406 			goto badunlocked;
407 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
408 		return (IPPROTO_DONE);
409 	}
410 	INP_RLOCK(inp);
411 	INP_INFO_RUNLOCK(&V_udbinfo);
412 	if (inp->inp_ppcb != NULL) {
413 		/*
414 		 * Engage the tunneling protocol.
415 		 */
416 		udp_tun_func_t tunnel_func;
417 
418 		tunnel_func = (udp_tun_func_t)inp->inp_ppcb;
419 		tunnel_func(m, off, inp);
420 		INP_RUNLOCK(inp);
421 		return (IPPROTO_DONE);
422 	}
423 	udp6_append(inp, m, off, &fromsa);
424 	INP_RUNLOCK(inp);
425 	return (IPPROTO_DONE);
426 
427 badheadlocked:
428 	INP_INFO_RUNLOCK(&V_udbinfo);
429 badunlocked:
430 	if (m)
431 		m_freem(m);
432 	return (IPPROTO_DONE);
433 }
434 
435 void
436 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
437 {
438 	INIT_VNET_INET(curvnet);
439 	struct udphdr uh;
440 	struct ip6_hdr *ip6;
441 	struct mbuf *m;
442 	int off = 0;
443 	struct ip6ctlparam *ip6cp = NULL;
444 	const struct sockaddr_in6 *sa6_src = NULL;
445 	void *cmdarg;
446 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
447 	struct udp_portonly {
448 		u_int16_t uh_sport;
449 		u_int16_t uh_dport;
450 	} *uhp;
451 
452 	if (sa->sa_family != AF_INET6 ||
453 	    sa->sa_len != sizeof(struct sockaddr_in6))
454 		return;
455 
456 	if ((unsigned)cmd >= PRC_NCMDS)
457 		return;
458 	if (PRC_IS_REDIRECT(cmd))
459 		notify = in6_rtchange, d = NULL;
460 	else if (cmd == PRC_HOSTDEAD)
461 		d = NULL;
462 	else if (inet6ctlerrmap[cmd] == 0)
463 		return;
464 
465 	/* if the parameter is from icmp6, decode it. */
466 	if (d != NULL) {
467 		ip6cp = (struct ip6ctlparam *)d;
468 		m = ip6cp->ip6c_m;
469 		ip6 = ip6cp->ip6c_ip6;
470 		off = ip6cp->ip6c_off;
471 		cmdarg = ip6cp->ip6c_cmdarg;
472 		sa6_src = ip6cp->ip6c_src;
473 	} else {
474 		m = NULL;
475 		ip6 = NULL;
476 		cmdarg = NULL;
477 		sa6_src = &sa6_any;
478 	}
479 
480 	if (ip6) {
481 		/*
482 		 * XXX: We assume that when IPV6 is non NULL,
483 		 * M and OFF are valid.
484 		 */
485 
486 		/* Check if we can safely examine src and dst ports. */
487 		if (m->m_pkthdr.len < off + sizeof(*uhp))
488 			return;
489 
490 		bzero(&uh, sizeof(uh));
491 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
492 
493 		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
494 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
495 		    cmdarg, notify);
496 	} else
497 		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
498 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
499 }
500 
501 static int
502 udp6_getcred(SYSCTL_HANDLER_ARGS)
503 {
504 	INIT_VNET_INET(curvnet);
505 	INIT_VNET_INET6(curvnet);
506 	struct xucred xuc;
507 	struct sockaddr_in6 addrs[2];
508 	struct inpcb *inp;
509 	int error;
510 
511 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
512 	if (error)
513 		return (error);
514 
515 	if (req->newlen != sizeof(addrs))
516 		return (EINVAL);
517 	if (req->oldlen != sizeof(struct xucred))
518 		return (EINVAL);
519 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
520 	if (error)
521 		return (error);
522 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
523 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
524 		return (error);
525 	}
526 	INP_INFO_RLOCK(&V_udbinfo);
527 	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
528 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
529 	    NULL);
530 	if (inp != NULL) {
531 		INP_RLOCK(inp);
532 		INP_INFO_RUNLOCK(&V_udbinfo);
533 		if (inp->inp_socket == NULL)
534 			error = ENOENT;
535 		if (error == 0)
536 			error = cr_canseesocket(req->td->td_ucred,
537 			    inp->inp_socket);
538 		if (error == 0)
539 			cru2x(inp->inp_cred, &xuc);
540 		INP_RUNLOCK(inp);
541 	} else {
542 		INP_INFO_RUNLOCK(&V_udbinfo);
543 		error = ENOENT;
544 	}
545 	if (error == 0)
546 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
547 	return (error);
548 }
549 
550 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
551     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
552 
553 static int
554 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
555     struct mbuf *control, struct thread *td)
556 {
557 	INIT_VNET_INET(curvnet);
558 	INIT_VNET_INET6(curvnet);
559 	u_int32_t ulen = m->m_pkthdr.len;
560 	u_int32_t plen = sizeof(struct udphdr) + ulen;
561 	struct ip6_hdr *ip6;
562 	struct udphdr *udp6;
563 	struct in6_addr *laddr, *faddr;
564 	struct sockaddr_in6 *sin6 = NULL;
565 	struct ifnet *oifp = NULL;
566 	int scope_ambiguous = 0;
567 	u_short fport;
568 	int error = 0;
569 	struct ip6_pktopts *optp, opt;
570 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
571 	int flags;
572 	struct sockaddr_in6 tmp;
573 
574 	INP_WLOCK_ASSERT(inp);
575 
576 	if (addr6) {
577 		/* addr6 has been validated in udp6_send(). */
578 		sin6 = (struct sockaddr_in6 *)addr6;
579 
580 		/* protect *sin6 from overwrites */
581 		tmp = *sin6;
582 		sin6 = &tmp;
583 
584 		/*
585 		 * Application should provide a proper zone ID or the use of
586 		 * default zone IDs should be enabled.  Unfortunately, some
587 		 * applications do not behave as it should, so we need a
588 		 * workaround.  Even if an appropriate ID is not determined,
589 		 * we'll see if we can determine the outgoing interface.  If we
590 		 * can, determine the zone ID based on the interface below.
591 		 */
592 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
593 			scope_ambiguous = 1;
594 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
595 			return (error);
596 	}
597 
598 	if (control) {
599 		if ((error = ip6_setpktopts(control, &opt,
600 		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
601 			goto release;
602 		optp = &opt;
603 	} else
604 		optp = inp->in6p_outputopts;
605 
606 	if (sin6) {
607 		faddr = &sin6->sin6_addr;
608 
609 		/*
610 		 * IPv4 version of udp_output calls in_pcbconnect in this case,
611 		 * which needs splnet and affects performance.
612 		 * Since we saw no essential reason for calling in_pcbconnect,
613 		 * we get rid of such kind of logic, and call in6_selectsrc
614 		 * and in6_pcbsetport in order to fill in the local address
615 		 * and the local port.
616 		 */
617 		if (sin6->sin6_port == 0) {
618 			error = EADDRNOTAVAIL;
619 			goto release;
620 		}
621 
622 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
623 			/* how about ::ffff:0.0.0.0 case? */
624 			error = EISCONN;
625 			goto release;
626 		}
627 
628 		fport = sin6->sin6_port; /* allow 0 port */
629 
630 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
631 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
632 				/*
633 				 * I believe we should explicitly discard the
634 				 * packet when mapped addresses are disabled,
635 				 * rather than send the packet as an IPv6 one.
636 				 * If we chose the latter approach, the packet
637 				 * might be sent out on the wire based on the
638 				 * default route, the situation which we'd
639 				 * probably want to avoid.
640 				 * (20010421 jinmei@kame.net)
641 				 */
642 				error = EINVAL;
643 				goto release;
644 			}
645 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
646 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
647 				/*
648 				 * when remote addr is an IPv4-mapped address,
649 				 * local addr should not be an IPv6 address,
650 				 * since you cannot determine how to map IPv6
651 				 * source address to IPv4.
652 				 */
653 				error = EINVAL;
654 				goto release;
655 			}
656 
657 			af = AF_INET;
658 		}
659 
660 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
661 			laddr = in6_selectsrc(sin6, optp, inp, NULL,
662 			    td->td_ucred, &oifp, &error);
663 			if (oifp && scope_ambiguous &&
664 			    (error = in6_setscope(&sin6->sin6_addr,
665 			    oifp, NULL))) {
666 				goto release;
667 			}
668 		} else
669 			laddr = &inp->in6p_laddr;	/* XXX */
670 		if (laddr == NULL) {
671 			if (error == 0)
672 				error = EADDRNOTAVAIL;
673 			goto release;
674 		}
675 		if (inp->inp_lport == 0 &&
676 		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
677 			goto release;
678 	} else {
679 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
680 			error = ENOTCONN;
681 			goto release;
682 		}
683 		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
684 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
685 				/*
686 				 * XXX: this case would happen when the
687 				 * application sets the V6ONLY flag after
688 				 * connecting the foreign address.
689 				 * Such applications should be fixed,
690 				 * so we bark here.
691 				 */
692 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
693 				    "option was set for a connected socket\n");
694 				error = EINVAL;
695 				goto release;
696 			} else
697 				af = AF_INET;
698 		}
699 		laddr = &inp->in6p_laddr;
700 		faddr = &inp->in6p_faddr;
701 		fport = inp->inp_fport;
702 	}
703 
704 	if (af == AF_INET)
705 		hlen = sizeof(struct ip);
706 
707 	/*
708 	 * Calculate data length and get a mbuf
709 	 * for UDP and IP6 headers.
710 	 */
711 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
712 	if (m == 0) {
713 		error = ENOBUFS;
714 		goto release;
715 	}
716 
717 	/*
718 	 * Stuff checksum and output datagram.
719 	 */
720 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
721 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
722 	udp6->uh_dport = fport;
723 	if (plen <= 0xffff)
724 		udp6->uh_ulen = htons((u_short)plen);
725 	else
726 		udp6->uh_ulen = 0;
727 	udp6->uh_sum = 0;
728 
729 	switch (af) {
730 	case AF_INET6:
731 		ip6 = mtod(m, struct ip6_hdr *);
732 		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
733 		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
734 		ip6->ip6_vfc	|= IPV6_VERSION;
735 #if 0				/* ip6_plen will be filled in ip6_output. */
736 		ip6->ip6_plen	= htons((u_short)plen);
737 #endif
738 		ip6->ip6_nxt	= IPPROTO_UDP;
739 		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
740 		ip6->ip6_src	= *laddr;
741 		ip6->ip6_dst	= *faddr;
742 
743 		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
744 				sizeof(struct ip6_hdr), plen)) == 0) {
745 			udp6->uh_sum = 0xffff;
746 		}
747 
748 		flags = 0;
749 
750 		UDPSTAT_INC(udps_opackets);
751 		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
752 		    NULL, inp);
753 		break;
754 	case AF_INET:
755 		error = EAFNOSUPPORT;
756 		goto release;
757 	}
758 	goto releaseopt;
759 
760 release:
761 	m_freem(m);
762 
763 releaseopt:
764 	if (control) {
765 		ip6_clearpktopts(&opt, -1);
766 		m_freem(control);
767 	}
768 	return (error);
769 }
770 
771 static void
772 udp6_abort(struct socket *so)
773 {
774 	INIT_VNET_INET(so->so_vnet);
775 	struct inpcb *inp;
776 
777 	inp = sotoinpcb(so);
778 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
779 
780 #ifdef INET
781 	if (inp->inp_vflag & INP_IPV4) {
782 		struct pr_usrreqs *pru;
783 
784 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
785 		(*pru->pru_abort)(so);
786 		return;
787 	}
788 #endif
789 
790 	INP_INFO_WLOCK(&V_udbinfo);
791 	INP_WLOCK(inp);
792 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
793 		in6_pcbdisconnect(inp);
794 		inp->in6p_laddr = in6addr_any;
795 		soisdisconnected(so);
796 	}
797 	INP_WUNLOCK(inp);
798 	INP_INFO_WUNLOCK(&V_udbinfo);
799 }
800 
801 static int
802 udp6_attach(struct socket *so, int proto, struct thread *td)
803 {
804 	INIT_VNET_INET(so->so_vnet);
805 	struct inpcb *inp;
806 	int error;
807 
808 	inp = sotoinpcb(so);
809 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
810 
811 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
812 		error = soreserve(so, udp_sendspace, udp_recvspace);
813 		if (error)
814 			return (error);
815 	}
816 	INP_INFO_WLOCK(&V_udbinfo);
817 	error = in_pcballoc(so, &V_udbinfo);
818 	if (error) {
819 		INP_INFO_WUNLOCK(&V_udbinfo);
820 		return (error);
821 	}
822 	inp = (struct inpcb *)so->so_pcb;
823 	INP_INFO_WUNLOCK(&V_udbinfo);
824 	inp->inp_vflag |= INP_IPV6;
825 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
826 		inp->inp_vflag |= INP_IPV4;
827 	inp->in6p_hops = -1;	/* use kernel default */
828 	inp->in6p_cksum = -1;	/* just to be sure */
829 	/*
830 	 * XXX: ugly!!
831 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
832 	 * because the socket may be bound to an IPv6 wildcard address,
833 	 * which may match an IPv4-mapped IPv6 address.
834 	 */
835 	inp->inp_ip_ttl = V_ip_defttl;
836 	INP_WUNLOCK(inp);
837 	return (0);
838 }
839 
840 static int
841 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
842 {
843 	INIT_VNET_INET(so->so_vnet);
844 	struct inpcb *inp;
845 	int error;
846 
847 	inp = sotoinpcb(so);
848 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
849 
850 	INP_INFO_WLOCK(&V_udbinfo);
851 	INP_WLOCK(inp);
852 	inp->inp_vflag &= ~INP_IPV4;
853 	inp->inp_vflag |= INP_IPV6;
854 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
855 		struct sockaddr_in6 *sin6_p;
856 
857 		sin6_p = (struct sockaddr_in6 *)nam;
858 
859 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
860 			inp->inp_vflag |= INP_IPV4;
861 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
862 			struct sockaddr_in sin;
863 
864 			in6_sin6_2_sin(&sin, sin6_p);
865 			inp->inp_vflag |= INP_IPV4;
866 			inp->inp_vflag &= ~INP_IPV6;
867 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
868 			    td->td_ucred);
869 			goto out;
870 		}
871 	}
872 
873 	error = in6_pcbbind(inp, nam, td->td_ucred);
874 out:
875 	INP_WUNLOCK(inp);
876 	INP_INFO_WUNLOCK(&V_udbinfo);
877 	return (error);
878 }
879 
880 static void
881 udp6_close(struct socket *so)
882 {
883 	INIT_VNET_INET(so->so_vnet);
884 	struct inpcb *inp;
885 
886 	inp = sotoinpcb(so);
887 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
888 
889 #ifdef INET
890 	if (inp->inp_vflag & INP_IPV4) {
891 		struct pr_usrreqs *pru;
892 
893 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
894 		(*pru->pru_disconnect)(so);
895 		return;
896 	}
897 #endif
898 	INP_INFO_WLOCK(&V_udbinfo);
899 	INP_WLOCK(inp);
900 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
901 		in6_pcbdisconnect(inp);
902 		inp->in6p_laddr = in6addr_any;
903 		soisdisconnected(so);
904 	}
905 	INP_WUNLOCK(inp);
906 	INP_INFO_WUNLOCK(&V_udbinfo);
907 }
908 
909 static int
910 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
911 {
912 	INIT_VNET_INET(so->so_vnet);
913 	struct inpcb *inp;
914 	struct sockaddr_in6 *sin6;
915 	int error;
916 
917 	inp = sotoinpcb(so);
918 	sin6 = (struct sockaddr_in6 *)nam;
919 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
920 
921 	INP_INFO_WLOCK(&V_udbinfo);
922 	INP_WLOCK(inp);
923 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
924 	    IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
925 		struct sockaddr_in sin;
926 
927 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
928 			error = EISCONN;
929 			goto out;
930 		}
931 		in6_sin6_2_sin(&sin, sin6);
932 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
933 		if (error != 0)
934 			goto out;
935 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
936 		    td->td_ucred);
937 		if (error == 0) {
938 			inp->inp_vflag |= INP_IPV4;
939 			inp->inp_vflag &= ~INP_IPV6;
940 			soisconnected(so);
941 		}
942 		goto out;
943 	}
944 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
945 		error = EISCONN;
946 		goto out;
947 	}
948 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
949 	if (error != 0)
950 		goto out;
951 	error = in6_pcbconnect(inp, nam, td->td_ucred);
952 	if (error == 0) {
953 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
954 			/* should be non mapped addr */
955 			inp->inp_vflag &= ~INP_IPV4;
956 			inp->inp_vflag |= INP_IPV6;
957 		}
958 		soisconnected(so);
959 	}
960 out:
961 	INP_WUNLOCK(inp);
962 	INP_INFO_WUNLOCK(&V_udbinfo);
963 	return (error);
964 }
965 
966 static void
967 udp6_detach(struct socket *so)
968 {
969 	INIT_VNET_INET(so->so_vnet);
970 	struct inpcb *inp;
971 
972 	inp = sotoinpcb(so);
973 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
974 
975 	INP_INFO_WLOCK(&V_udbinfo);
976 	INP_WLOCK(inp);
977 	in_pcbdetach(inp);
978 	in_pcbfree(inp);
979 	INP_INFO_WUNLOCK(&V_udbinfo);
980 }
981 
982 static int
983 udp6_disconnect(struct socket *so)
984 {
985 	INIT_VNET_INET(so->so_vnet);
986 	struct inpcb *inp;
987 	int error;
988 
989 	inp = sotoinpcb(so);
990 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
991 
992 	INP_INFO_WLOCK(&V_udbinfo);
993 	INP_WLOCK(inp);
994 
995 #ifdef INET
996 	if (inp->inp_vflag & INP_IPV4) {
997 		struct pr_usrreqs *pru;
998 
999 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1000 		error = (*pru->pru_disconnect)(so);
1001 		goto out;
1002 	}
1003 #endif
1004 
1005 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1006 		error = ENOTCONN;
1007 		goto out;
1008 	}
1009 
1010 	in6_pcbdisconnect(inp);
1011 	inp->in6p_laddr = in6addr_any;
1012 	SOCK_LOCK(so);
1013 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1014 	SOCK_UNLOCK(so);
1015 out:
1016 	INP_WUNLOCK(inp);
1017 	INP_INFO_WUNLOCK(&V_udbinfo);
1018 	return (0);
1019 }
1020 
1021 static int
1022 udp6_send(struct socket *so, int flags, struct mbuf *m,
1023     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1024 {
1025 	INIT_VNET_INET(so->so_vnet);
1026 	struct inpcb *inp;
1027 	int error = 0;
1028 
1029 	inp = sotoinpcb(so);
1030 	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1031 
1032 	INP_INFO_WLOCK(&V_udbinfo);
1033 	INP_WLOCK(inp);
1034 	if (addr) {
1035 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1036 			error = EINVAL;
1037 			goto bad;
1038 		}
1039 		if (addr->sa_family != AF_INET6) {
1040 			error = EAFNOSUPPORT;
1041 			goto bad;
1042 		}
1043 	}
1044 
1045 #ifdef INET
1046 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1047 		int hasv4addr;
1048 		struct sockaddr_in6 *sin6 = 0;
1049 
1050 		if (addr == 0)
1051 			hasv4addr = (inp->inp_vflag & INP_IPV4);
1052 		else {
1053 			sin6 = (struct sockaddr_in6 *)addr;
1054 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1055 			    ? 1 : 0;
1056 		}
1057 		if (hasv4addr) {
1058 			struct pr_usrreqs *pru;
1059 
1060 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1061 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
1062 				/*
1063 				 * When remote addr is IPv4-mapped address,
1064 				 * local addr should not be an IPv6 address;
1065 				 * since you cannot determine how to map IPv6
1066 				 * source address to IPv4.
1067 				 */
1068 				error = EINVAL;
1069 				goto out;
1070 			}
1071 
1072 			/*
1073 			 * XXXRW: We release UDP-layer locks before calling
1074 			 * udp_send() in order to avoid recursion.  However,
1075 			 * this does mean there is a short window where inp's
1076 			 * fields are unstable.  Could this lead to a
1077 			 * potential race in which the factors causing us to
1078 			 * select the UDPv4 output routine are invalidated?
1079 			 */
1080 			INP_WUNLOCK(inp);
1081 			INP_INFO_WUNLOCK(&V_udbinfo);
1082 			if (sin6)
1083 				in6_sin6_2_sin_in_sock(addr);
1084 			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1085 			/* addr will just be freed in sendit(). */
1086 			return ((*pru->pru_send)(so, flags, m, addr, control,
1087 			    td));
1088 		}
1089 	}
1090 #endif
1091 #ifdef MAC
1092 	mac_inpcb_create_mbuf(inp, m);
1093 #endif
1094 	error = udp6_output(inp, m, addr, control, td);
1095 out:
1096 	INP_WUNLOCK(inp);
1097 	INP_INFO_WUNLOCK(&V_udbinfo);
1098 	return (error);
1099 
1100 bad:
1101 	INP_WUNLOCK(inp);
1102 	INP_INFO_WUNLOCK(&V_udbinfo);
1103 	m_freem(m);
1104 	return (error);
1105 }
1106 
1107 struct pr_usrreqs udp6_usrreqs = {
1108 	.pru_abort =		udp6_abort,
1109 	.pru_attach =		udp6_attach,
1110 	.pru_bind =		udp6_bind,
1111 	.pru_connect =		udp6_connect,
1112 	.pru_control =		in6_control,
1113 	.pru_detach =		udp6_detach,
1114 	.pru_disconnect =	udp6_disconnect,
1115 	.pru_peeraddr =		in6_mapped_peeraddr,
1116 	.pru_send =		udp6_send,
1117 	.pru_shutdown =		udp_shutdown,
1118 	.pru_sockaddr =		in6_mapped_sockaddr,
1119 	.pru_soreceive =	soreceive_dgram,
1120 	.pru_sosend =		sosend_dgram,
1121 	.pru_sosetlabel =	in_pcbsosetlabel,
1122 	.pru_close =		udp6_close
1123 };
1124