xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 39beb93c3f8bdbf72a61fda42300b5ebed7390c8)
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 		V_udpstat.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 ip6_hdr *ip6;
181 	struct udphdr *uh;
182 	struct inpcb *inp;
183 	int off = *offp;
184 	int plen, ulen;
185 	struct sockaddr_in6 fromsa;
186 
187 	ip6 = mtod(m, struct ip6_hdr *);
188 
189 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
190 		/* XXX send icmp6 host/port unreach? */
191 		m_freem(m);
192 		return (IPPROTO_DONE);
193 	}
194 
195 #ifndef PULLDOWN_TEST
196 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
197 	ip6 = mtod(m, struct ip6_hdr *);
198 	uh = (struct udphdr *)((caddr_t)ip6 + off);
199 #else
200 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
201 	if (!uh)
202 		return (IPPROTO_DONE);
203 #endif
204 
205 	V_udpstat.udps_ipackets++;
206 
207 	/*
208 	 * Destination port of 0 is illegal, based on RFC768.
209 	 */
210 	if (uh->uh_dport == 0)
211 		goto badunlocked;
212 
213 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
214 	ulen = ntohs((u_short)uh->uh_ulen);
215 
216 	if (plen != ulen) {
217 		V_udpstat.udps_badlen++;
218 		goto badunlocked;
219 	}
220 
221 	/*
222 	 * Checksum extended UDP header and data.
223 	 */
224 	if (uh->uh_sum == 0) {
225 		V_udpstat.udps_nosum++;
226 		goto badunlocked;
227 	}
228 	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
229 		V_udpstat.udps_badsum++;
230 		goto badunlocked;
231 	}
232 
233 	/*
234 	 * Construct sockaddr format source address.
235 	 */
236 	init_sin6(&fromsa, m);
237 	fromsa.sin6_port = uh->uh_sport;
238 
239 	INP_INFO_RLOCK(&V_udbinfo);
240 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
241 		struct inpcb *last;
242 
243 		/*
244 		 * In the event that laddr should be set to the link-local
245 		 * address (this happens in RIPng), the multicast address
246 		 * specified in the received packet will not match laddr.  To
247 		 * handle this situation, matching is relaxed if the
248 		 * receiving interface is the same as one specified in the
249 		 * socket and if the destination multicast address matches
250 		 * one of the multicast groups specified in the socket.
251 		 */
252 
253 		/*
254 		 * KAME note: traditionally we dropped udpiphdr from mbuf
255 		 * here.  We need udphdr for IPsec processing so we do that
256 		 * later.
257 		 */
258 		last = NULL;
259 		LIST_FOREACH(inp, &V_udb, inp_list) {
260 			if ((inp->inp_vflag & INP_IPV6) == 0)
261 				continue;
262 			if (inp->inp_lport != uh->uh_dport)
263 				continue;
264 			/*
265 			 * XXX: Do not check source port of incoming datagram
266 			 * unless inp_connect() has been called to bind the
267 			 * fport part of the 4-tuple; the source could be
268 			 * trying to talk to us with an ephemeral port.
269 			 */
270 			if (inp->inp_fport != 0 &&
271 			    inp->inp_fport != uh->uh_sport)
272 				continue;
273 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
274 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
275 							&ip6->ip6_dst))
276 					continue;
277 			}
278 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
279 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
280 							&ip6->ip6_src) ||
281 				    inp->inp_fport != uh->uh_sport)
282 					continue;
283 			}
284 
285 			if (last != NULL) {
286 				struct mbuf *n;
287 
288 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
289 					INP_RLOCK(last);
290 					if (last->inp_ppcb != NULL) {
291 						/*
292 						 * Engage the tunneling
293 						 * protocol we will have to
294 						 * leave the info_lock up,
295 						 * since we are hunting
296 						 * through multiple UDP's.
297 						 *
298 						 */
299 						udp_tun_func_t tunnel_func;
300 
301 						tunnel_func = (udp_tun_func_t)last->inp_ppcb;
302 						tunnel_func(n, off, last);
303 						INP_RUNLOCK(last);
304 					} else {
305 						udp6_append(last, n, off, &fromsa);
306 						INP_RUNLOCK(last);
307 					}
308 				}
309 			}
310 			last = inp;
311 			/*
312 			 * Don't look for additional matches if this one does
313 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
314 			 * socket options set.  This heuristic avoids
315 			 * searching through all pcbs in the common case of a
316 			 * non-shared port.  It assumes that an application
317 			 * will never clear these options after setting them.
318 			 */
319 			if ((last->inp_socket->so_options &
320 			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
321 				break;
322 		}
323 
324 		if (last == NULL) {
325 			/*
326 			 * No matching pcb found; discard datagram.  (No need
327 			 * to send an ICMP Port Unreachable for a broadcast
328 			 * or multicast datgram.)
329 			 */
330 			V_udpstat.udps_noport++;
331 			V_udpstat.udps_noportmcast++;
332 			goto badheadlocked;
333 		}
334 		INP_RLOCK(last);
335 		INP_INFO_RUNLOCK(&V_udbinfo);
336 		if (last->inp_ppcb != NULL) {
337 			/*
338 			 * Engage the tunneling protocol.
339 			 */
340 			udp_tun_func_t tunnel_func;
341 
342 			tunnel_func = (udp_tun_func_t)inp->inp_ppcb;
343 			tunnel_func(m, off, last);
344 			INP_RUNLOCK(last);
345 			return (IPPROTO_DONE);
346 		}
347 		udp6_append(last, m, off, &fromsa);
348 		INP_RUNLOCK(last);
349 		return (IPPROTO_DONE);
350 	}
351 	/*
352 	 * Locate pcb for datagram.
353 	 */
354 	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
355 	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
356 	if (inp == NULL) {
357 		if (udp_log_in_vain) {
358 			char ip6bufs[INET6_ADDRSTRLEN];
359 			char ip6bufd[INET6_ADDRSTRLEN];
360 
361 			log(LOG_INFO,
362 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
363 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
364 			    ntohs(uh->uh_dport),
365 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
366 			    ntohs(uh->uh_sport));
367 		}
368 		V_udpstat.udps_noport++;
369 		if (m->m_flags & M_MCAST) {
370 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
371 			V_udpstat.udps_noportmcast++;
372 			goto badheadlocked;
373 		}
374 		INP_INFO_RUNLOCK(&V_udbinfo);
375 		if (V_udp_blackhole)
376 			goto badunlocked;
377 		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
378 			goto badunlocked;
379 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
380 		return (IPPROTO_DONE);
381 	}
382 	INP_RLOCK(inp);
383 	INP_INFO_RUNLOCK(&V_udbinfo);
384 	if (inp->inp_ppcb != NULL) {
385 		/*
386 		 * Engage the tunneling protocol.
387 		 */
388 		udp_tun_func_t tunnel_func;
389 
390 		tunnel_func = (udp_tun_func_t)inp->inp_ppcb;
391 		tunnel_func(m, off, inp);
392 		INP_RUNLOCK(inp);
393 		return (IPPROTO_DONE);
394 	}
395 	udp6_append(inp, m, off, &fromsa);
396 	INP_RUNLOCK(inp);
397 	return (IPPROTO_DONE);
398 
399 badheadlocked:
400 	INP_INFO_RUNLOCK(&V_udbinfo);
401 badunlocked:
402 	if (m)
403 		m_freem(m);
404 	return (IPPROTO_DONE);
405 }
406 
407 void
408 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
409 {
410 	INIT_VNET_INET(curvnet);
411 	struct udphdr uh;
412 	struct ip6_hdr *ip6;
413 	struct mbuf *m;
414 	int off = 0;
415 	struct ip6ctlparam *ip6cp = NULL;
416 	const struct sockaddr_in6 *sa6_src = NULL;
417 	void *cmdarg;
418 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
419 	struct udp_portonly {
420 		u_int16_t uh_sport;
421 		u_int16_t uh_dport;
422 	} *uhp;
423 
424 	if (sa->sa_family != AF_INET6 ||
425 	    sa->sa_len != sizeof(struct sockaddr_in6))
426 		return;
427 
428 	if ((unsigned)cmd >= PRC_NCMDS)
429 		return;
430 	if (PRC_IS_REDIRECT(cmd))
431 		notify = in6_rtchange, d = NULL;
432 	else if (cmd == PRC_HOSTDEAD)
433 		d = NULL;
434 	else if (inet6ctlerrmap[cmd] == 0)
435 		return;
436 
437 	/* if the parameter is from icmp6, decode it. */
438 	if (d != NULL) {
439 		ip6cp = (struct ip6ctlparam *)d;
440 		m = ip6cp->ip6c_m;
441 		ip6 = ip6cp->ip6c_ip6;
442 		off = ip6cp->ip6c_off;
443 		cmdarg = ip6cp->ip6c_cmdarg;
444 		sa6_src = ip6cp->ip6c_src;
445 	} else {
446 		m = NULL;
447 		ip6 = NULL;
448 		cmdarg = NULL;
449 		sa6_src = &sa6_any;
450 	}
451 
452 	if (ip6) {
453 		/*
454 		 * XXX: We assume that when IPV6 is non NULL,
455 		 * M and OFF are valid.
456 		 */
457 
458 		/* Check if we can safely examine src and dst ports. */
459 		if (m->m_pkthdr.len < off + sizeof(*uhp))
460 			return;
461 
462 		bzero(&uh, sizeof(uh));
463 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
464 
465 		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
466 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
467 		    cmdarg, notify);
468 	} else
469 		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
470 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
471 }
472 
473 static int
474 udp6_getcred(SYSCTL_HANDLER_ARGS)
475 {
476 	INIT_VNET_INET(curvnet);
477 	INIT_VNET_INET6(curvnet);
478 	struct xucred xuc;
479 	struct sockaddr_in6 addrs[2];
480 	struct inpcb *inp;
481 	int error;
482 
483 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
484 	if (error)
485 		return (error);
486 
487 	if (req->newlen != sizeof(addrs))
488 		return (EINVAL);
489 	if (req->oldlen != sizeof(struct xucred))
490 		return (EINVAL);
491 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
492 	if (error)
493 		return (error);
494 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
495 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
496 		return (error);
497 	}
498 	INP_INFO_RLOCK(&V_udbinfo);
499 	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
500 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
501 	    NULL);
502 	if (inp != NULL) {
503 		INP_RLOCK(inp);
504 		INP_INFO_RUNLOCK(&V_udbinfo);
505 		if (inp->inp_socket == NULL)
506 			error = ENOENT;
507 		if (error == 0)
508 			error = cr_canseesocket(req->td->td_ucred,
509 			    inp->inp_socket);
510 		if (error == 0)
511 			cru2x(inp->inp_cred, &xuc);
512 		INP_RUNLOCK(inp);
513 	} else {
514 		INP_INFO_RUNLOCK(&V_udbinfo);
515 		error = ENOENT;
516 	}
517 	if (error == 0)
518 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
519 	return (error);
520 }
521 
522 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
523     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
524 
525 static int
526 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
527     struct mbuf *control, struct thread *td)
528 {
529 	INIT_VNET_INET(curvnet);
530 	INIT_VNET_INET6(curvnet);
531 	u_int32_t ulen = m->m_pkthdr.len;
532 	u_int32_t plen = sizeof(struct udphdr) + ulen;
533 	struct ip6_hdr *ip6;
534 	struct udphdr *udp6;
535 	struct in6_addr *laddr, *faddr;
536 	struct sockaddr_in6 *sin6 = NULL;
537 	struct ifnet *oifp = NULL;
538 	int scope_ambiguous = 0;
539 	u_short fport;
540 	int error = 0;
541 	struct ip6_pktopts *optp, opt;
542 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
543 	int flags;
544 	struct sockaddr_in6 tmp;
545 
546 	INP_WLOCK_ASSERT(inp);
547 
548 	if (addr6) {
549 		/* addr6 has been validated in udp6_send(). */
550 		sin6 = (struct sockaddr_in6 *)addr6;
551 
552 		/* protect *sin6 from overwrites */
553 		tmp = *sin6;
554 		sin6 = &tmp;
555 
556 		/*
557 		 * Application should provide a proper zone ID or the use of
558 		 * default zone IDs should be enabled.  Unfortunately, some
559 		 * applications do not behave as it should, so we need a
560 		 * workaround.  Even if an appropriate ID is not determined,
561 		 * we'll see if we can determine the outgoing interface.  If we
562 		 * can, determine the zone ID based on the interface below.
563 		 */
564 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
565 			scope_ambiguous = 1;
566 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
567 			return (error);
568 	}
569 
570 	if (control) {
571 		if ((error = ip6_setpktopts(control, &opt,
572 		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
573 			goto release;
574 		optp = &opt;
575 	} else
576 		optp = inp->in6p_outputopts;
577 
578 	if (sin6) {
579 		faddr = &sin6->sin6_addr;
580 
581 		/*
582 		 * IPv4 version of udp_output calls in_pcbconnect in this case,
583 		 * which needs splnet and affects performance.
584 		 * Since we saw no essential reason for calling in_pcbconnect,
585 		 * we get rid of such kind of logic, and call in6_selectsrc
586 		 * and in6_pcbsetport in order to fill in the local address
587 		 * and the local port.
588 		 */
589 		if (sin6->sin6_port == 0) {
590 			error = EADDRNOTAVAIL;
591 			goto release;
592 		}
593 
594 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
595 			/* how about ::ffff:0.0.0.0 case? */
596 			error = EISCONN;
597 			goto release;
598 		}
599 
600 		fport = sin6->sin6_port; /* allow 0 port */
601 
602 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
603 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
604 				/*
605 				 * I believe we should explicitly discard the
606 				 * packet when mapped addresses are disabled,
607 				 * rather than send the packet as an IPv6 one.
608 				 * If we chose the latter approach, the packet
609 				 * might be sent out on the wire based on the
610 				 * default route, the situation which we'd
611 				 * probably want to avoid.
612 				 * (20010421 jinmei@kame.net)
613 				 */
614 				error = EINVAL;
615 				goto release;
616 			}
617 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
618 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
619 				/*
620 				 * when remote addr is an IPv4-mapped address,
621 				 * local addr should not be an IPv6 address,
622 				 * since you cannot determine how to map IPv6
623 				 * source address to IPv4.
624 				 */
625 				error = EINVAL;
626 				goto release;
627 			}
628 
629 			af = AF_INET;
630 		}
631 
632 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
633 			laddr = in6_selectsrc(sin6, optp, inp, NULL,
634 			    td->td_ucred, &oifp, &error);
635 			if (oifp && scope_ambiguous &&
636 			    (error = in6_setscope(&sin6->sin6_addr,
637 			    oifp, NULL))) {
638 				goto release;
639 			}
640 		} else
641 			laddr = &inp->in6p_laddr;	/* XXX */
642 		if (laddr == NULL) {
643 			if (error == 0)
644 				error = EADDRNOTAVAIL;
645 			goto release;
646 		}
647 		if (inp->inp_lport == 0 &&
648 		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
649 			goto release;
650 	} else {
651 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
652 			error = ENOTCONN;
653 			goto release;
654 		}
655 		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
656 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
657 				/*
658 				 * XXX: this case would happen when the
659 				 * application sets the V6ONLY flag after
660 				 * connecting the foreign address.
661 				 * Such applications should be fixed,
662 				 * so we bark here.
663 				 */
664 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
665 				    "option was set for a connected socket\n");
666 				error = EINVAL;
667 				goto release;
668 			} else
669 				af = AF_INET;
670 		}
671 		laddr = &inp->in6p_laddr;
672 		faddr = &inp->in6p_faddr;
673 		fport = inp->inp_fport;
674 	}
675 
676 	if (af == AF_INET)
677 		hlen = sizeof(struct ip);
678 
679 	/*
680 	 * Calculate data length and get a mbuf
681 	 * for UDP and IP6 headers.
682 	 */
683 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
684 	if (m == 0) {
685 		error = ENOBUFS;
686 		goto release;
687 	}
688 
689 	/*
690 	 * Stuff checksum and output datagram.
691 	 */
692 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
693 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
694 	udp6->uh_dport = fport;
695 	if (plen <= 0xffff)
696 		udp6->uh_ulen = htons((u_short)plen);
697 	else
698 		udp6->uh_ulen = 0;
699 	udp6->uh_sum = 0;
700 
701 	switch (af) {
702 	case AF_INET6:
703 		ip6 = mtod(m, struct ip6_hdr *);
704 		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
705 		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
706 		ip6->ip6_vfc	|= IPV6_VERSION;
707 #if 0				/* ip6_plen will be filled in ip6_output. */
708 		ip6->ip6_plen	= htons((u_short)plen);
709 #endif
710 		ip6->ip6_nxt	= IPPROTO_UDP;
711 		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
712 		ip6->ip6_src	= *laddr;
713 		ip6->ip6_dst	= *faddr;
714 
715 		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
716 				sizeof(struct ip6_hdr), plen)) == 0) {
717 			udp6->uh_sum = 0xffff;
718 		}
719 
720 		flags = 0;
721 
722 		V_udpstat.udps_opackets++;
723 		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
724 		    NULL, inp);
725 		break;
726 	case AF_INET:
727 		error = EAFNOSUPPORT;
728 		goto release;
729 	}
730 	goto releaseopt;
731 
732 release:
733 	m_freem(m);
734 
735 releaseopt:
736 	if (control) {
737 		ip6_clearpktopts(&opt, -1);
738 		m_freem(control);
739 	}
740 	return (error);
741 }
742 
743 static void
744 udp6_abort(struct socket *so)
745 {
746 	INIT_VNET_INET(so->so_vnet);
747 	struct inpcb *inp;
748 
749 	inp = sotoinpcb(so);
750 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
751 
752 #ifdef INET
753 	if (inp->inp_vflag & INP_IPV4) {
754 		struct pr_usrreqs *pru;
755 
756 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
757 		(*pru->pru_abort)(so);
758 		return;
759 	}
760 #endif
761 
762 	INP_INFO_WLOCK(&V_udbinfo);
763 	INP_WLOCK(inp);
764 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
765 		in6_pcbdisconnect(inp);
766 		inp->in6p_laddr = in6addr_any;
767 		soisdisconnected(so);
768 	}
769 	INP_WUNLOCK(inp);
770 	INP_INFO_WUNLOCK(&V_udbinfo);
771 }
772 
773 static int
774 udp6_attach(struct socket *so, int proto, struct thread *td)
775 {
776 	INIT_VNET_INET(so->so_vnet);
777 	struct inpcb *inp;
778 	int error;
779 
780 	inp = sotoinpcb(so);
781 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
782 
783 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
784 		error = soreserve(so, udp_sendspace, udp_recvspace);
785 		if (error)
786 			return (error);
787 	}
788 	INP_INFO_WLOCK(&V_udbinfo);
789 	error = in_pcballoc(so, &V_udbinfo);
790 	if (error) {
791 		INP_INFO_WUNLOCK(&V_udbinfo);
792 		return (error);
793 	}
794 	inp = (struct inpcb *)so->so_pcb;
795 	INP_INFO_WUNLOCK(&V_udbinfo);
796 	inp->inp_vflag |= INP_IPV6;
797 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
798 		inp->inp_vflag |= INP_IPV4;
799 	inp->in6p_hops = -1;	/* use kernel default */
800 	inp->in6p_cksum = -1;	/* just to be sure */
801 	/*
802 	 * XXX: ugly!!
803 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
804 	 * because the socket may be bound to an IPv6 wildcard address,
805 	 * which may match an IPv4-mapped IPv6 address.
806 	 */
807 	inp->inp_ip_ttl = V_ip_defttl;
808 	INP_WUNLOCK(inp);
809 	return (0);
810 }
811 
812 static int
813 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
814 {
815 	INIT_VNET_INET(so->so_vnet);
816 	struct inpcb *inp;
817 	int error;
818 
819 	inp = sotoinpcb(so);
820 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
821 
822 	INP_INFO_WLOCK(&V_udbinfo);
823 	INP_WLOCK(inp);
824 	inp->inp_vflag &= ~INP_IPV4;
825 	inp->inp_vflag |= INP_IPV6;
826 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
827 		struct sockaddr_in6 *sin6_p;
828 
829 		sin6_p = (struct sockaddr_in6 *)nam;
830 
831 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
832 			inp->inp_vflag |= INP_IPV4;
833 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
834 			struct sockaddr_in sin;
835 
836 			in6_sin6_2_sin(&sin, sin6_p);
837 			inp->inp_vflag |= INP_IPV4;
838 			inp->inp_vflag &= ~INP_IPV6;
839 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
840 			    td->td_ucred);
841 			goto out;
842 		}
843 	}
844 
845 	error = in6_pcbbind(inp, nam, td->td_ucred);
846 out:
847 	INP_WUNLOCK(inp);
848 	INP_INFO_WUNLOCK(&V_udbinfo);
849 	return (error);
850 }
851 
852 static void
853 udp6_close(struct socket *so)
854 {
855 	INIT_VNET_INET(so->so_vnet);
856 	struct inpcb *inp;
857 
858 	inp = sotoinpcb(so);
859 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
860 
861 #ifdef INET
862 	if (inp->inp_vflag & INP_IPV4) {
863 		struct pr_usrreqs *pru;
864 
865 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
866 		(*pru->pru_disconnect)(so);
867 		return;
868 	}
869 #endif
870 	INP_INFO_WLOCK(&V_udbinfo);
871 	INP_WLOCK(inp);
872 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
873 		in6_pcbdisconnect(inp);
874 		inp->in6p_laddr = in6addr_any;
875 		soisdisconnected(so);
876 	}
877 	INP_WUNLOCK(inp);
878 	INP_INFO_WUNLOCK(&V_udbinfo);
879 }
880 
881 static int
882 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
883 {
884 	INIT_VNET_INET(so->so_vnet);
885 	struct inpcb *inp;
886 	struct sockaddr_in6 *sin6;
887 	int error;
888 
889 	inp = sotoinpcb(so);
890 	sin6 = (struct sockaddr_in6 *)nam;
891 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
892 
893 	INP_INFO_WLOCK(&V_udbinfo);
894 	INP_WLOCK(inp);
895 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
896 	    IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
897 		struct sockaddr_in sin;
898 
899 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
900 			error = EISCONN;
901 			goto out;
902 		}
903 		in6_sin6_2_sin(&sin, sin6);
904 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
905 		if (error != 0)
906 			goto out;
907 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
908 		    td->td_ucred);
909 		if (error == 0) {
910 			inp->inp_vflag |= INP_IPV4;
911 			inp->inp_vflag &= ~INP_IPV6;
912 			soisconnected(so);
913 		}
914 		goto out;
915 	}
916 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
917 		error = EISCONN;
918 		goto out;
919 	}
920 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
921 	if (error != 0)
922 		goto out;
923 	error = in6_pcbconnect(inp, nam, td->td_ucred);
924 	if (error == 0) {
925 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
926 			/* should be non mapped addr */
927 			inp->inp_vflag &= ~INP_IPV4;
928 			inp->inp_vflag |= INP_IPV6;
929 		}
930 		soisconnected(so);
931 	}
932 out:
933 	INP_WUNLOCK(inp);
934 	INP_INFO_WUNLOCK(&V_udbinfo);
935 	return (error);
936 }
937 
938 static void
939 udp6_detach(struct socket *so)
940 {
941 	INIT_VNET_INET(so->so_vnet);
942 	struct inpcb *inp;
943 
944 	inp = sotoinpcb(so);
945 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
946 
947 	INP_INFO_WLOCK(&V_udbinfo);
948 	INP_WLOCK(inp);
949 	in_pcbdetach(inp);
950 	in_pcbfree(inp);
951 	INP_INFO_WUNLOCK(&V_udbinfo);
952 }
953 
954 static int
955 udp6_disconnect(struct socket *so)
956 {
957 	INIT_VNET_INET(so->so_vnet);
958 	struct inpcb *inp;
959 	int error;
960 
961 	inp = sotoinpcb(so);
962 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
963 
964 	INP_INFO_WLOCK(&V_udbinfo);
965 	INP_WLOCK(inp);
966 
967 #ifdef INET
968 	if (inp->inp_vflag & INP_IPV4) {
969 		struct pr_usrreqs *pru;
970 
971 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
972 		error = (*pru->pru_disconnect)(so);
973 		goto out;
974 	}
975 #endif
976 
977 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
978 		error = ENOTCONN;
979 		goto out;
980 	}
981 
982 	in6_pcbdisconnect(inp);
983 	inp->in6p_laddr = in6addr_any;
984 	SOCK_LOCK(so);
985 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
986 	SOCK_UNLOCK(so);
987 out:
988 	INP_WUNLOCK(inp);
989 	INP_INFO_WUNLOCK(&V_udbinfo);
990 	return (0);
991 }
992 
993 static int
994 udp6_send(struct socket *so, int flags, struct mbuf *m,
995     struct sockaddr *addr, struct mbuf *control, struct thread *td)
996 {
997 	INIT_VNET_INET(so->so_vnet);
998 	struct inpcb *inp;
999 	int error = 0;
1000 
1001 	inp = sotoinpcb(so);
1002 	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1003 
1004 	INP_INFO_WLOCK(&V_udbinfo);
1005 	INP_WLOCK(inp);
1006 	if (addr) {
1007 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1008 			error = EINVAL;
1009 			goto bad;
1010 		}
1011 		if (addr->sa_family != AF_INET6) {
1012 			error = EAFNOSUPPORT;
1013 			goto bad;
1014 		}
1015 	}
1016 
1017 #ifdef INET
1018 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1019 		int hasv4addr;
1020 		struct sockaddr_in6 *sin6 = 0;
1021 
1022 		if (addr == 0)
1023 			hasv4addr = (inp->inp_vflag & INP_IPV4);
1024 		else {
1025 			sin6 = (struct sockaddr_in6 *)addr;
1026 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1027 			    ? 1 : 0;
1028 		}
1029 		if (hasv4addr) {
1030 			struct pr_usrreqs *pru;
1031 
1032 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1033 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
1034 				/*
1035 				 * When remote addr is IPv4-mapped address,
1036 				 * local addr should not be an IPv6 address;
1037 				 * since you cannot determine how to map IPv6
1038 				 * source address to IPv4.
1039 				 */
1040 				error = EINVAL;
1041 				goto out;
1042 			}
1043 
1044 			/*
1045 			 * XXXRW: We release UDP-layer locks before calling
1046 			 * udp_send() in order to avoid recursion.  However,
1047 			 * this does mean there is a short window where inp's
1048 			 * fields are unstable.  Could this lead to a
1049 			 * potential race in which the factors causing us to
1050 			 * select the UDPv4 output routine are invalidated?
1051 			 */
1052 			INP_WUNLOCK(inp);
1053 			INP_INFO_WUNLOCK(&V_udbinfo);
1054 			if (sin6)
1055 				in6_sin6_2_sin_in_sock(addr);
1056 			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1057 			/* addr will just be freed in sendit(). */
1058 			return ((*pru->pru_send)(so, flags, m, addr, control,
1059 			    td));
1060 		}
1061 	}
1062 #endif
1063 #ifdef MAC
1064 	mac_inpcb_create_mbuf(inp, m);
1065 #endif
1066 	error = udp6_output(inp, m, addr, control, td);
1067 out:
1068 	INP_WUNLOCK(inp);
1069 	INP_INFO_WUNLOCK(&V_udbinfo);
1070 	return (error);
1071 
1072 bad:
1073 	INP_WUNLOCK(inp);
1074 	INP_INFO_WUNLOCK(&V_udbinfo);
1075 	m_freem(m);
1076 	return (error);
1077 }
1078 
1079 struct pr_usrreqs udp6_usrreqs = {
1080 	.pru_abort =		udp6_abort,
1081 	.pru_attach =		udp6_attach,
1082 	.pru_bind =		udp6_bind,
1083 	.pru_connect =		udp6_connect,
1084 	.pru_control =		in6_control,
1085 	.pru_detach =		udp6_detach,
1086 	.pru_disconnect =	udp6_disconnect,
1087 	.pru_peeraddr =		in6_mapped_peeraddr,
1088 	.pru_send =		udp6_send,
1089 	.pru_shutdown =		udp_shutdown,
1090 	.pru_sockaddr =		in6_mapped_sockaddr,
1091 	.pru_soreceive =	soreceive_dgram,
1092 	.pru_sosend =		sosend_dgram,
1093 	.pru_sosetlabel =	in_pcbsosetlabel,
1094 	.pru_close =		udp6_close
1095 };
1096