xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 3157ba21)
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 
72 #include <sys/param.h>
73 #include <sys/jail.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/mbuf.h>
77 #include <sys/priv.h>
78 #include <sys/proc.h>
79 #include <sys/protosw.h>
80 #include <sys/signalvar.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/sx.h>
84 #include <sys/sysctl.h>
85 #include <sys/syslog.h>
86 #include <sys/systm.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip.h>
97 #include <netinet/ip_icmp.h>
98 #include <netinet/ip6.h>
99 #include <netinet/icmp_var.h>
100 #include <netinet/icmp6.h>
101 #include <netinet/ip_var.h>
102 #include <netinet/udp.h>
103 #include <netinet/udp_var.h>
104 
105 #include <netinet6/ip6protosw.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/udp6_var.h>
109 #include <netinet6/scope6_var.h>
110 
111 #ifdef IPSEC
112 #include <netipsec/ipsec.h>
113 #include <netipsec/ipsec6.h>
114 #endif /* IPSEC */
115 
116 #include <security/mac/mac_framework.h>
117 
118 /*
119  * UDP protocol implementation.
120  * Per RFC 768, August, 1980.
121  */
122 
123 extern struct protosw	inetsw[];
124 static void		udp6_detach(struct socket *so);
125 
126 static void
127 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
128     struct sockaddr_in6 *fromsa)
129 {
130 	struct socket *so;
131 	struct mbuf *opts;
132 
133 	INP_LOCK_ASSERT(inp);
134 
135 #ifdef IPSEC
136 	/* Check AH/ESP integrity. */
137 	if (ipsec6_in_reject(n, inp)) {
138 		m_freem(n);
139 		V_ipsec6stat.in_polvio++;
140 		return;
141 	}
142 #endif /* IPSEC */
143 #ifdef MAC
144 	if (mac_inpcb_check_deliver(inp, n) != 0) {
145 		m_freem(n);
146 		return;
147 	}
148 #endif
149 	opts = NULL;
150 	if (inp->inp_flags & INP_CONTROLOPTS ||
151 	    inp->inp_socket->so_options & SO_TIMESTAMP)
152 		ip6_savecontrol(inp, n, &opts);
153 	m_adj(n, off + sizeof(struct udphdr));
154 
155 	so = inp->inp_socket;
156 	SOCKBUF_LOCK(&so->so_rcv);
157 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
158 	    opts) == 0) {
159 		SOCKBUF_UNLOCK(&so->so_rcv);
160 		m_freem(n);
161 		if (opts)
162 			m_freem(opts);
163 		UDPSTAT_INC(udps_fullsock);
164 	} else
165 		sorwakeup_locked(so);
166 }
167 
168 int
169 udp6_input(struct mbuf **mp, int *offp, int proto)
170 {
171 	struct mbuf *m = *mp;
172 	struct ifnet *ifp;
173 	struct ip6_hdr *ip6;
174 	struct udphdr *uh;
175 	struct inpcb *inp;
176 	struct udpcb *up;
177 	int off = *offp;
178 	int plen, ulen;
179 	struct sockaddr_in6 fromsa;
180 
181 	ifp = m->m_pkthdr.rcvif;
182 	ip6 = mtod(m, struct ip6_hdr *);
183 
184 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
185 		/* XXX send icmp6 host/port unreach? */
186 		m_freem(m);
187 		return (IPPROTO_DONE);
188 	}
189 
190 #ifndef PULLDOWN_TEST
191 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
192 	ip6 = mtod(m, struct ip6_hdr *);
193 	uh = (struct udphdr *)((caddr_t)ip6 + off);
194 #else
195 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
196 	if (!uh)
197 		return (IPPROTO_DONE);
198 #endif
199 
200 	UDPSTAT_INC(udps_ipackets);
201 
202 	/*
203 	 * Destination port of 0 is illegal, based on RFC768.
204 	 */
205 	if (uh->uh_dport == 0)
206 		goto badunlocked;
207 
208 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
209 	ulen = ntohs((u_short)uh->uh_ulen);
210 
211 	if (plen != ulen) {
212 		UDPSTAT_INC(udps_badlen);
213 		goto badunlocked;
214 	}
215 
216 	/*
217 	 * Checksum extended UDP header and data.
218 	 */
219 	if (uh->uh_sum == 0) {
220 		UDPSTAT_INC(udps_nosum);
221 		goto badunlocked;
222 	}
223 	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
224 		UDPSTAT_INC(udps_badsum);
225 		goto badunlocked;
226 	}
227 
228 	/*
229 	 * Construct sockaddr format source address.
230 	 */
231 	init_sin6(&fromsa, m);
232 	fromsa.sin6_port = uh->uh_sport;
233 
234 	INP_INFO_RLOCK(&V_udbinfo);
235 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
236 		struct inpcb *last;
237 		struct ip6_moptions *imo;
238 
239 		/*
240 		 * In the event that laddr should be set to the link-local
241 		 * address (this happens in RIPng), the multicast address
242 		 * specified in the received packet will not match laddr.  To
243 		 * handle this situation, matching is relaxed if the
244 		 * receiving interface is the same as one specified in the
245 		 * socket and if the destination multicast address matches
246 		 * one of the multicast groups specified in the socket.
247 		 */
248 
249 		/*
250 		 * KAME note: traditionally we dropped udpiphdr from mbuf
251 		 * here.  We need udphdr for IPsec processing so we do that
252 		 * later.
253 		 */
254 		last = NULL;
255 		LIST_FOREACH(inp, &V_udb, inp_list) {
256 			if ((inp->inp_vflag & INP_IPV6) == 0)
257 				continue;
258 			if (inp->inp_lport != uh->uh_dport)
259 				continue;
260 			if (inp->inp_fport != 0 &&
261 			    inp->inp_fport != uh->uh_sport)
262 				continue;
263 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
264 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
265 							&ip6->ip6_dst))
266 					continue;
267 			}
268 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
269 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
270 							&ip6->ip6_src) ||
271 				    inp->inp_fport != uh->uh_sport)
272 					continue;
273 			}
274 
275 			/*
276 			 * Handle socket delivery policy for any-source
277 			 * and source-specific multicast. [RFC3678]
278 			 */
279 			imo = inp->in6p_moptions;
280 			if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
281 				struct sockaddr_in6	 mcaddr;
282 				int			 blocked;
283 
284 				INP_RLOCK(inp);
285 
286 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
287 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
288 				mcaddr.sin6_family = AF_INET6;
289 				mcaddr.sin6_addr = ip6->ip6_dst;
290 
291 				blocked = im6o_mc_filter(imo, ifp,
292 					(struct sockaddr *)&mcaddr,
293 					(struct sockaddr *)&fromsa);
294 				if (blocked != MCAST_PASS) {
295 					if (blocked == MCAST_NOTGMEMBER)
296 						IP6STAT_INC(ip6s_notmember);
297 					if (blocked == MCAST_NOTSMEMBER ||
298 					    blocked == MCAST_MUTED)
299 						UDPSTAT_INC(udps_filtermcast);
300 					INP_RUNLOCK(inp); /* XXX */
301 					continue;
302 				}
303 
304 				INP_RUNLOCK(inp);
305 			}
306 			if (last != NULL) {
307 				struct mbuf *n;
308 
309 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
310 					INP_RLOCK(last);
311 					up = intoudpcb(last);
312 					if (up->u_tun_func == NULL) {
313 						udp6_append(last, n, off, &fromsa);
314 					} else {
315 						/*
316 						 * Engage the tunneling
317 						 * protocol we will have to
318 						 * leave the info_lock up,
319 						 * since we are hunting
320 						 * through multiple UDP's.
321 						 *
322 						 */
323 						(*up->u_tun_func)(n, off, last);
324 					}
325 					INP_RUNLOCK(last);
326 				}
327 			}
328 			last = inp;
329 			/*
330 			 * Don't look for additional matches if this one does
331 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
332 			 * socket options set.  This heuristic avoids
333 			 * searching through all pcbs in the common case of a
334 			 * non-shared port.  It assumes that an application
335 			 * will never clear these options after setting them.
336 			 */
337 			if ((last->inp_socket->so_options &
338 			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
339 				break;
340 		}
341 
342 		if (last == NULL) {
343 			/*
344 			 * No matching pcb found; discard datagram.  (No need
345 			 * to send an ICMP Port Unreachable for a broadcast
346 			 * or multicast datgram.)
347 			 */
348 			UDPSTAT_INC(udps_noport);
349 			UDPSTAT_INC(udps_noportmcast);
350 			goto badheadlocked;
351 		}
352 		INP_RLOCK(last);
353 		INP_INFO_RUNLOCK(&V_udbinfo);
354 		up = intoudpcb(last);
355 		if (up->u_tun_func == NULL) {
356 			udp6_append(last, m, off, &fromsa);
357 		} else {
358 			/*
359 			 * Engage the tunneling protocol.
360 			 */
361 			(*up->u_tun_func)(m, off, last);
362 		}
363 		INP_RUNLOCK(last);
364 		return (IPPROTO_DONE);
365 	}
366 	/*
367 	 * Locate pcb for datagram.
368 	 */
369 	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
370 	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
371 	if (inp == NULL) {
372 		if (udp_log_in_vain) {
373 			char ip6bufs[INET6_ADDRSTRLEN];
374 			char ip6bufd[INET6_ADDRSTRLEN];
375 
376 			log(LOG_INFO,
377 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
378 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
379 			    ntohs(uh->uh_dport),
380 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
381 			    ntohs(uh->uh_sport));
382 		}
383 		UDPSTAT_INC(udps_noport);
384 		if (m->m_flags & M_MCAST) {
385 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
386 			UDPSTAT_INC(udps_noportmcast);
387 			goto badheadlocked;
388 		}
389 		INP_INFO_RUNLOCK(&V_udbinfo);
390 		if (V_udp_blackhole)
391 			goto badunlocked;
392 		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
393 			goto badunlocked;
394 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
395 		return (IPPROTO_DONE);
396 	}
397 	INP_RLOCK(inp);
398 	INP_INFO_RUNLOCK(&V_udbinfo);
399 	up = intoudpcb(inp);
400 	if (up->u_tun_func == NULL) {
401 		udp6_append(inp, m, off, &fromsa);
402 	} else {
403 		/*
404 		 * Engage the tunneling protocol.
405 		 */
406 
407 		(*up->u_tun_func)(m, off, inp);
408 	}
409 	INP_RUNLOCK(inp);
410 	return (IPPROTO_DONE);
411 
412 badheadlocked:
413 	INP_INFO_RUNLOCK(&V_udbinfo);
414 badunlocked:
415 	if (m)
416 		m_freem(m);
417 	return (IPPROTO_DONE);
418 }
419 
420 void
421 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
422 {
423 	struct udphdr uh;
424 	struct ip6_hdr *ip6;
425 	struct mbuf *m;
426 	int off = 0;
427 	struct ip6ctlparam *ip6cp = NULL;
428 	const struct sockaddr_in6 *sa6_src = NULL;
429 	void *cmdarg;
430 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
431 	struct udp_portonly {
432 		u_int16_t uh_sport;
433 		u_int16_t uh_dport;
434 	} *uhp;
435 
436 	if (sa->sa_family != AF_INET6 ||
437 	    sa->sa_len != sizeof(struct sockaddr_in6))
438 		return;
439 
440 	if ((unsigned)cmd >= PRC_NCMDS)
441 		return;
442 	if (PRC_IS_REDIRECT(cmd))
443 		notify = in6_rtchange, d = NULL;
444 	else if (cmd == PRC_HOSTDEAD)
445 		d = NULL;
446 	else if (inet6ctlerrmap[cmd] == 0)
447 		return;
448 
449 	/* if the parameter is from icmp6, decode it. */
450 	if (d != NULL) {
451 		ip6cp = (struct ip6ctlparam *)d;
452 		m = ip6cp->ip6c_m;
453 		ip6 = ip6cp->ip6c_ip6;
454 		off = ip6cp->ip6c_off;
455 		cmdarg = ip6cp->ip6c_cmdarg;
456 		sa6_src = ip6cp->ip6c_src;
457 	} else {
458 		m = NULL;
459 		ip6 = NULL;
460 		cmdarg = NULL;
461 		sa6_src = &sa6_any;
462 	}
463 
464 	if (ip6) {
465 		/*
466 		 * XXX: We assume that when IPV6 is non NULL,
467 		 * M and OFF are valid.
468 		 */
469 
470 		/* Check if we can safely examine src and dst ports. */
471 		if (m->m_pkthdr.len < off + sizeof(*uhp))
472 			return;
473 
474 		bzero(&uh, sizeof(uh));
475 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
476 
477 		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
478 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
479 		    cmdarg, notify);
480 	} else
481 		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
482 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
483 }
484 
485 static int
486 udp6_getcred(SYSCTL_HANDLER_ARGS)
487 {
488 	struct xucred xuc;
489 	struct sockaddr_in6 addrs[2];
490 	struct inpcb *inp;
491 	int error;
492 
493 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
494 	if (error)
495 		return (error);
496 
497 	if (req->newlen != sizeof(addrs))
498 		return (EINVAL);
499 	if (req->oldlen != sizeof(struct xucred))
500 		return (EINVAL);
501 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
502 	if (error)
503 		return (error);
504 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
505 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
506 		return (error);
507 	}
508 	INP_INFO_RLOCK(&V_udbinfo);
509 	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
510 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
511 	    NULL);
512 	if (inp != NULL) {
513 		INP_RLOCK(inp);
514 		INP_INFO_RUNLOCK(&V_udbinfo);
515 		if (inp->inp_socket == NULL)
516 			error = ENOENT;
517 		if (error == 0)
518 			error = cr_canseesocket(req->td->td_ucred,
519 			    inp->inp_socket);
520 		if (error == 0)
521 			cru2x(inp->inp_cred, &xuc);
522 		INP_RUNLOCK(inp);
523 	} else {
524 		INP_INFO_RUNLOCK(&V_udbinfo);
525 		error = ENOENT;
526 	}
527 	if (error == 0)
528 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
529 	return (error);
530 }
531 
532 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
533     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
534 
535 static int
536 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
537     struct mbuf *control, struct thread *td)
538 {
539 	u_int32_t ulen = m->m_pkthdr.len;
540 	u_int32_t plen = sizeof(struct udphdr) + ulen;
541 	struct ip6_hdr *ip6;
542 	struct udphdr *udp6;
543 	struct in6_addr *laddr, *faddr, in6a;
544 	struct sockaddr_in6 *sin6 = NULL;
545 	struct ifnet *oifp = NULL;
546 	int scope_ambiguous = 0;
547 	u_short fport;
548 	int error = 0;
549 	struct ip6_pktopts *optp, opt;
550 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
551 	int flags;
552 	struct sockaddr_in6 tmp;
553 
554 	INP_WLOCK_ASSERT(inp);
555 
556 	if (addr6) {
557 		/* addr6 has been validated in udp6_send(). */
558 		sin6 = (struct sockaddr_in6 *)addr6;
559 
560 		/* protect *sin6 from overwrites */
561 		tmp = *sin6;
562 		sin6 = &tmp;
563 
564 		/*
565 		 * Application should provide a proper zone ID or the use of
566 		 * default zone IDs should be enabled.  Unfortunately, some
567 		 * applications do not behave as it should, so we need a
568 		 * workaround.  Even if an appropriate ID is not determined,
569 		 * we'll see if we can determine the outgoing interface.  If we
570 		 * can, determine the zone ID based on the interface below.
571 		 */
572 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
573 			scope_ambiguous = 1;
574 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
575 			return (error);
576 	}
577 
578 	if (control) {
579 		if ((error = ip6_setpktopts(control, &opt,
580 		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
581 			goto release;
582 		optp = &opt;
583 	} else
584 		optp = inp->in6p_outputopts;
585 
586 	if (sin6) {
587 		faddr = &sin6->sin6_addr;
588 
589 		/*
590 		 * IPv4 version of udp_output calls in_pcbconnect in this case,
591 		 * which needs splnet and affects performance.
592 		 * Since we saw no essential reason for calling in_pcbconnect,
593 		 * we get rid of such kind of logic, and call in6_selectsrc
594 		 * and in6_pcbsetport in order to fill in the local address
595 		 * and the local port.
596 		 */
597 		if (sin6->sin6_port == 0) {
598 			error = EADDRNOTAVAIL;
599 			goto release;
600 		}
601 
602 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
603 			/* how about ::ffff:0.0.0.0 case? */
604 			error = EISCONN;
605 			goto release;
606 		}
607 
608 		fport = sin6->sin6_port; /* allow 0 port */
609 
610 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
611 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
612 				/*
613 				 * I believe we should explicitly discard the
614 				 * packet when mapped addresses are disabled,
615 				 * rather than send the packet as an IPv6 one.
616 				 * If we chose the latter approach, the packet
617 				 * might be sent out on the wire based on the
618 				 * default route, the situation which we'd
619 				 * probably want to avoid.
620 				 * (20010421 jinmei@kame.net)
621 				 */
622 				error = EINVAL;
623 				goto release;
624 			}
625 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
626 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
627 				/*
628 				 * when remote addr is an IPv4-mapped address,
629 				 * local addr should not be an IPv6 address,
630 				 * since you cannot determine how to map IPv6
631 				 * source address to IPv4.
632 				 */
633 				error = EINVAL;
634 				goto release;
635 			}
636 
637 			af = AF_INET;
638 		}
639 
640 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
641 			error = in6_selectsrc(sin6, optp, inp, NULL,
642 			    td->td_ucred, &oifp, &in6a);
643 			if (error)
644 				goto release;
645 			if (oifp && scope_ambiguous &&
646 			    (error = in6_setscope(&sin6->sin6_addr,
647 			    oifp, NULL))) {
648 				goto release;
649 			}
650 			laddr = &in6a;
651 		} else
652 			laddr = &inp->in6p_laddr;	/* XXX */
653 		if (laddr == NULL) {
654 			if (error == 0)
655 				error = EADDRNOTAVAIL;
656 			goto release;
657 		}
658 		if (inp->inp_lport == 0 &&
659 		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
660 			goto release;
661 	} else {
662 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
663 			error = ENOTCONN;
664 			goto release;
665 		}
666 		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
667 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
668 				/*
669 				 * XXX: this case would happen when the
670 				 * application sets the V6ONLY flag after
671 				 * connecting the foreign address.
672 				 * Such applications should be fixed,
673 				 * so we bark here.
674 				 */
675 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
676 				    "option was set for a connected socket\n");
677 				error = EINVAL;
678 				goto release;
679 			} else
680 				af = AF_INET;
681 		}
682 		laddr = &inp->in6p_laddr;
683 		faddr = &inp->in6p_faddr;
684 		fport = inp->inp_fport;
685 	}
686 
687 	if (af == AF_INET)
688 		hlen = sizeof(struct ip);
689 
690 	/*
691 	 * Calculate data length and get a mbuf
692 	 * for UDP and IP6 headers.
693 	 */
694 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
695 	if (m == 0) {
696 		error = ENOBUFS;
697 		goto release;
698 	}
699 
700 	/*
701 	 * Stuff checksum and output datagram.
702 	 */
703 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
704 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
705 	udp6->uh_dport = fport;
706 	if (plen <= 0xffff)
707 		udp6->uh_ulen = htons((u_short)plen);
708 	else
709 		udp6->uh_ulen = 0;
710 	udp6->uh_sum = 0;
711 
712 	switch (af) {
713 	case AF_INET6:
714 		ip6 = mtod(m, struct ip6_hdr *);
715 		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
716 		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
717 		ip6->ip6_vfc	|= IPV6_VERSION;
718 #if 0				/* ip6_plen will be filled in ip6_output. */
719 		ip6->ip6_plen	= htons((u_short)plen);
720 #endif
721 		ip6->ip6_nxt	= IPPROTO_UDP;
722 		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
723 		ip6->ip6_src	= *laddr;
724 		ip6->ip6_dst	= *faddr;
725 
726 		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
727 				sizeof(struct ip6_hdr), plen)) == 0) {
728 			udp6->uh_sum = 0xffff;
729 		}
730 
731 		flags = 0;
732 
733 		UDPSTAT_INC(udps_opackets);
734 		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
735 		    NULL, inp);
736 		break;
737 	case AF_INET:
738 		error = EAFNOSUPPORT;
739 		goto release;
740 	}
741 	goto releaseopt;
742 
743 release:
744 	m_freem(m);
745 
746 releaseopt:
747 	if (control) {
748 		ip6_clearpktopts(&opt, -1);
749 		m_freem(control);
750 	}
751 	return (error);
752 }
753 
754 static void
755 udp6_abort(struct socket *so)
756 {
757 	struct inpcb *inp;
758 
759 	inp = sotoinpcb(so);
760 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
761 
762 #ifdef INET
763 	if (inp->inp_vflag & INP_IPV4) {
764 		struct pr_usrreqs *pru;
765 
766 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
767 		(*pru->pru_abort)(so);
768 		return;
769 	}
770 #endif
771 
772 	INP_INFO_WLOCK(&V_udbinfo);
773 	INP_WLOCK(inp);
774 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
775 		in6_pcbdisconnect(inp);
776 		inp->in6p_laddr = in6addr_any;
777 		soisdisconnected(so);
778 	}
779 	INP_WUNLOCK(inp);
780 	INP_INFO_WUNLOCK(&V_udbinfo);
781 }
782 
783 static int
784 udp6_attach(struct socket *so, int proto, struct thread *td)
785 {
786 	struct inpcb *inp;
787 	int error;
788 
789 	inp = sotoinpcb(so);
790 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
791 
792 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
793 		error = soreserve(so, udp_sendspace, udp_recvspace);
794 		if (error)
795 			return (error);
796 	}
797 	INP_INFO_WLOCK(&V_udbinfo);
798 	error = in_pcballoc(so, &V_udbinfo);
799 	if (error) {
800 		INP_INFO_WUNLOCK(&V_udbinfo);
801 		return (error);
802 	}
803 	inp = (struct inpcb *)so->so_pcb;
804 	inp->inp_vflag |= INP_IPV6;
805 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
806 		inp->inp_vflag |= INP_IPV4;
807 	inp->in6p_hops = -1;	/* use kernel default */
808 	inp->in6p_cksum = -1;	/* just to be sure */
809 	/*
810 	 * XXX: ugly!!
811 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
812 	 * because the socket may be bound to an IPv6 wildcard address,
813 	 * which may match an IPv4-mapped IPv6 address.
814 	 */
815 	inp->inp_ip_ttl = V_ip_defttl;
816 
817 	error = udp_newudpcb(inp);
818 	if (error) {
819 		in_pcbdetach(inp);
820 		in_pcbfree(inp);
821 		INP_INFO_WUNLOCK(&V_udbinfo);
822 		return (error);
823 	}
824 	INP_WUNLOCK(inp);
825 	INP_INFO_WUNLOCK(&V_udbinfo);
826 	return (0);
827 }
828 
829 static int
830 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
831 {
832 	struct inpcb *inp;
833 	int error;
834 
835 	inp = sotoinpcb(so);
836 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
837 
838 	INP_INFO_WLOCK(&V_udbinfo);
839 	INP_WLOCK(inp);
840 	inp->inp_vflag &= ~INP_IPV4;
841 	inp->inp_vflag |= INP_IPV6;
842 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
843 		struct sockaddr_in6 *sin6_p;
844 
845 		sin6_p = (struct sockaddr_in6 *)nam;
846 
847 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
848 			inp->inp_vflag |= INP_IPV4;
849 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
850 			struct sockaddr_in sin;
851 
852 			in6_sin6_2_sin(&sin, sin6_p);
853 			inp->inp_vflag |= INP_IPV4;
854 			inp->inp_vflag &= ~INP_IPV6;
855 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
856 			    td->td_ucred);
857 			goto out;
858 		}
859 	}
860 
861 	error = in6_pcbbind(inp, nam, td->td_ucred);
862 out:
863 	INP_WUNLOCK(inp);
864 	INP_INFO_WUNLOCK(&V_udbinfo);
865 	return (error);
866 }
867 
868 static void
869 udp6_close(struct socket *so)
870 {
871 	struct inpcb *inp;
872 
873 	inp = sotoinpcb(so);
874 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
875 
876 #ifdef INET
877 	if (inp->inp_vflag & INP_IPV4) {
878 		struct pr_usrreqs *pru;
879 
880 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
881 		(*pru->pru_disconnect)(so);
882 		return;
883 	}
884 #endif
885 	INP_INFO_WLOCK(&V_udbinfo);
886 	INP_WLOCK(inp);
887 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
888 		in6_pcbdisconnect(inp);
889 		inp->in6p_laddr = in6addr_any;
890 		soisdisconnected(so);
891 	}
892 	INP_WUNLOCK(inp);
893 	INP_INFO_WUNLOCK(&V_udbinfo);
894 }
895 
896 static int
897 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
898 {
899 	struct inpcb *inp;
900 	struct sockaddr_in6 *sin6;
901 	int error;
902 
903 	inp = sotoinpcb(so);
904 	sin6 = (struct sockaddr_in6 *)nam;
905 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
906 
907 	INP_INFO_WLOCK(&V_udbinfo);
908 	INP_WLOCK(inp);
909 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
910 	    IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
911 		struct sockaddr_in sin;
912 
913 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
914 			error = EISCONN;
915 			goto out;
916 		}
917 		in6_sin6_2_sin(&sin, sin6);
918 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
919 		if (error != 0)
920 			goto out;
921 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
922 		    td->td_ucred);
923 		if (error == 0) {
924 			inp->inp_vflag |= INP_IPV4;
925 			inp->inp_vflag &= ~INP_IPV6;
926 			soisconnected(so);
927 		}
928 		goto out;
929 	}
930 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
931 		error = EISCONN;
932 		goto out;
933 	}
934 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
935 	if (error != 0)
936 		goto out;
937 	error = in6_pcbconnect(inp, nam, td->td_ucred);
938 	if (error == 0) {
939 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
940 			/* should be non mapped addr */
941 			inp->inp_vflag &= ~INP_IPV4;
942 			inp->inp_vflag |= INP_IPV6;
943 		}
944 		soisconnected(so);
945 	}
946 out:
947 	INP_WUNLOCK(inp);
948 	INP_INFO_WUNLOCK(&V_udbinfo);
949 	return (error);
950 }
951 
952 static void
953 udp6_detach(struct socket *so)
954 {
955 	struct inpcb *inp;
956 	struct udpcb *up;
957 
958 	inp = sotoinpcb(so);
959 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
960 
961 	INP_INFO_WLOCK(&V_udbinfo);
962 	INP_WLOCK(inp);
963 	up = intoudpcb(inp);
964 	KASSERT(up != NULL, ("%s: up == NULL", __func__));
965 	in_pcbdetach(inp);
966 	in_pcbfree(inp);
967 	INP_INFO_WUNLOCK(&V_udbinfo);
968 	udp_discardcb(up);
969 }
970 
971 static int
972 udp6_disconnect(struct socket *so)
973 {
974 	struct inpcb *inp;
975 	int error;
976 
977 	inp = sotoinpcb(so);
978 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
979 
980 	INP_INFO_WLOCK(&V_udbinfo);
981 	INP_WLOCK(inp);
982 
983 #ifdef INET
984 	if (inp->inp_vflag & INP_IPV4) {
985 		struct pr_usrreqs *pru;
986 
987 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
988 		error = (*pru->pru_disconnect)(so);
989 		goto out;
990 	}
991 #endif
992 
993 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
994 		error = ENOTCONN;
995 		goto out;
996 	}
997 
998 	in6_pcbdisconnect(inp);
999 	inp->in6p_laddr = in6addr_any;
1000 	SOCK_LOCK(so);
1001 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1002 	SOCK_UNLOCK(so);
1003 out:
1004 	INP_WUNLOCK(inp);
1005 	INP_INFO_WUNLOCK(&V_udbinfo);
1006 	return (0);
1007 }
1008 
1009 static int
1010 udp6_send(struct socket *so, int flags, struct mbuf *m,
1011     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1012 {
1013 	struct inpcb *inp;
1014 	int error = 0;
1015 
1016 	inp = sotoinpcb(so);
1017 	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1018 
1019 	INP_INFO_WLOCK(&V_udbinfo);
1020 	INP_WLOCK(inp);
1021 	if (addr) {
1022 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1023 			error = EINVAL;
1024 			goto bad;
1025 		}
1026 		if (addr->sa_family != AF_INET6) {
1027 			error = EAFNOSUPPORT;
1028 			goto bad;
1029 		}
1030 	}
1031 
1032 #ifdef INET
1033 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1034 		int hasv4addr;
1035 		struct sockaddr_in6 *sin6 = 0;
1036 
1037 		if (addr == 0)
1038 			hasv4addr = (inp->inp_vflag & INP_IPV4);
1039 		else {
1040 			sin6 = (struct sockaddr_in6 *)addr;
1041 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1042 			    ? 1 : 0;
1043 		}
1044 		if (hasv4addr) {
1045 			struct pr_usrreqs *pru;
1046 
1047 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1048 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
1049 				/*
1050 				 * When remote addr is IPv4-mapped address,
1051 				 * local addr should not be an IPv6 address;
1052 				 * since you cannot determine how to map IPv6
1053 				 * source address to IPv4.
1054 				 */
1055 				error = EINVAL;
1056 				goto out;
1057 			}
1058 
1059 			/*
1060 			 * XXXRW: We release UDP-layer locks before calling
1061 			 * udp_send() in order to avoid recursion.  However,
1062 			 * this does mean there is a short window where inp's
1063 			 * fields are unstable.  Could this lead to a
1064 			 * potential race in which the factors causing us to
1065 			 * select the UDPv4 output routine are invalidated?
1066 			 */
1067 			INP_WUNLOCK(inp);
1068 			INP_INFO_WUNLOCK(&V_udbinfo);
1069 			if (sin6)
1070 				in6_sin6_2_sin_in_sock(addr);
1071 			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1072 			/* addr will just be freed in sendit(). */
1073 			return ((*pru->pru_send)(so, flags, m, addr, control,
1074 			    td));
1075 		}
1076 	}
1077 #endif
1078 #ifdef MAC
1079 	mac_inpcb_create_mbuf(inp, m);
1080 #endif
1081 	error = udp6_output(inp, m, addr, control, td);
1082 #ifdef INET
1083 out:
1084 #endif
1085 	INP_WUNLOCK(inp);
1086 	INP_INFO_WUNLOCK(&V_udbinfo);
1087 	return (error);
1088 
1089 bad:
1090 	INP_WUNLOCK(inp);
1091 	INP_INFO_WUNLOCK(&V_udbinfo);
1092 	m_freem(m);
1093 	return (error);
1094 }
1095 
1096 struct pr_usrreqs udp6_usrreqs = {
1097 	.pru_abort =		udp6_abort,
1098 	.pru_attach =		udp6_attach,
1099 	.pru_bind =		udp6_bind,
1100 	.pru_connect =		udp6_connect,
1101 	.pru_control =		in6_control,
1102 	.pru_detach =		udp6_detach,
1103 	.pru_disconnect =	udp6_disconnect,
1104 	.pru_peeraddr =		in6_mapped_peeraddr,
1105 	.pru_send =		udp6_send,
1106 	.pru_shutdown =		udp_shutdown,
1107 	.pru_sockaddr =		in6_mapped_sockaddr,
1108 	.pru_soreceive =	soreceive_dgram,
1109 	.pru_sosend =		sosend_dgram,
1110 	.pru_sosetlabel =	in_pcbsosetlabel,
1111 	.pru_close =		udp6_close
1112 };
1113