udp6_usrreq.c revision 186223
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: head/sys/netinet6/udp6_usrreq.c 186223 2008-12-17 13:00:18Z bz $");
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
127extern struct protosw	inetsw[];
128static void		udp6_detach(struct socket *so);
129
130static void
131udp6_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
174int
175udp6_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					udp6_append(last, n, off, &fromsa);
291					INP_RUNLOCK(last);
292				}
293			}
294			last = inp;
295			/*
296			 * Don't look for additional matches if this one does
297			 * not have either the SO_REUSEPORT or SO_REUSEADDR
298			 * socket options set.  This heuristic avoids
299			 * searching through all pcbs in the common case of a
300			 * non-shared port.  It assumes that an application
301			 * will never clear these options after setting them.
302			 */
303			if ((last->inp_socket->so_options &
304			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
305				break;
306		}
307
308		if (last == NULL) {
309			/*
310			 * No matching pcb found; discard datagram.  (No need
311			 * to send an ICMP Port Unreachable for a broadcast
312			 * or multicast datgram.)
313			 */
314			V_udpstat.udps_noport++;
315			V_udpstat.udps_noportmcast++;
316			goto badheadlocked;
317		}
318		INP_RLOCK(last);
319		INP_INFO_RUNLOCK(&V_udbinfo);
320		udp6_append(last, m, off, &fromsa);
321		INP_RUNLOCK(last);
322		return (IPPROTO_DONE);
323	}
324	/*
325	 * Locate pcb for datagram.
326	 */
327	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
328	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
329	if (inp == NULL) {
330		if (udp_log_in_vain) {
331			char ip6bufs[INET6_ADDRSTRLEN];
332			char ip6bufd[INET6_ADDRSTRLEN];
333
334			log(LOG_INFO,
335			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
336			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
337			    ntohs(uh->uh_dport),
338			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
339			    ntohs(uh->uh_sport));
340		}
341		V_udpstat.udps_noport++;
342		if (m->m_flags & M_MCAST) {
343			printf("UDP6: M_MCAST is set in a unicast packet.\n");
344			V_udpstat.udps_noportmcast++;
345			goto badheadlocked;
346		}
347		INP_INFO_RUNLOCK(&V_udbinfo);
348		if (V_udp_blackhole)
349			goto badunlocked;
350		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
351			goto badunlocked;
352		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
353		return (IPPROTO_DONE);
354	}
355	INP_RLOCK(inp);
356	INP_INFO_RUNLOCK(&V_udbinfo);
357	udp6_append(inp, m, off, &fromsa);
358	INP_RUNLOCK(inp);
359	return (IPPROTO_DONE);
360
361badheadlocked:
362	INP_INFO_RUNLOCK(&V_udbinfo);
363badunlocked:
364	if (m)
365		m_freem(m);
366	return (IPPROTO_DONE);
367}
368
369void
370udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
371{
372	INIT_VNET_INET(curvnet);
373	struct udphdr uh;
374	struct ip6_hdr *ip6;
375	struct mbuf *m;
376	int off = 0;
377	struct ip6ctlparam *ip6cp = NULL;
378	const struct sockaddr_in6 *sa6_src = NULL;
379	void *cmdarg;
380	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
381	struct udp_portonly {
382		u_int16_t uh_sport;
383		u_int16_t uh_dport;
384	} *uhp;
385
386	if (sa->sa_family != AF_INET6 ||
387	    sa->sa_len != sizeof(struct sockaddr_in6))
388		return;
389
390	if ((unsigned)cmd >= PRC_NCMDS)
391		return;
392	if (PRC_IS_REDIRECT(cmd))
393		notify = in6_rtchange, d = NULL;
394	else if (cmd == PRC_HOSTDEAD)
395		d = NULL;
396	else if (inet6ctlerrmap[cmd] == 0)
397		return;
398
399	/* if the parameter is from icmp6, decode it. */
400	if (d != NULL) {
401		ip6cp = (struct ip6ctlparam *)d;
402		m = ip6cp->ip6c_m;
403		ip6 = ip6cp->ip6c_ip6;
404		off = ip6cp->ip6c_off;
405		cmdarg = ip6cp->ip6c_cmdarg;
406		sa6_src = ip6cp->ip6c_src;
407	} else {
408		m = NULL;
409		ip6 = NULL;
410		cmdarg = NULL;
411		sa6_src = &sa6_any;
412	}
413
414	if (ip6) {
415		/*
416		 * XXX: We assume that when IPV6 is non NULL,
417		 * M and OFF are valid.
418		 */
419
420		/* Check if we can safely examine src and dst ports. */
421		if (m->m_pkthdr.len < off + sizeof(*uhp))
422			return;
423
424		bzero(&uh, sizeof(uh));
425		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
426
427		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
428		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
429		    cmdarg, notify);
430	} else
431		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
432		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
433}
434
435static int
436udp6_getcred(SYSCTL_HANDLER_ARGS)
437{
438	INIT_VNET_INET(curvnet);
439	INIT_VNET_INET6(curvnet);
440	struct xucred xuc;
441	struct sockaddr_in6 addrs[2];
442	struct inpcb *inp;
443	int error;
444
445	error = priv_check(req->td, PRIV_NETINET_GETCRED);
446	if (error)
447		return (error);
448
449	if (req->newlen != sizeof(addrs))
450		return (EINVAL);
451	if (req->oldlen != sizeof(struct xucred))
452		return (EINVAL);
453	error = SYSCTL_IN(req, addrs, sizeof(addrs));
454	if (error)
455		return (error);
456	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
457	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
458		return (error);
459	}
460	INP_INFO_RLOCK(&V_udbinfo);
461	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
462	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
463	    NULL);
464	if (inp != NULL) {
465		INP_RLOCK(inp);
466		INP_INFO_RUNLOCK(&V_udbinfo);
467		if (inp->inp_socket == NULL)
468			error = ENOENT;
469		if (error == 0)
470			error = cr_canseesocket(req->td->td_ucred,
471			    inp->inp_socket);
472		if (error == 0)
473			cru2x(inp->inp_cred, &xuc);
474		INP_RUNLOCK(inp);
475	} else {
476		INP_INFO_RUNLOCK(&V_udbinfo);
477		error = ENOENT;
478	}
479	if (error == 0)
480		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
481	return (error);
482}
483
484SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
485    0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
486
487static int
488udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
489    struct mbuf *control, struct thread *td)
490{
491	INIT_VNET_INET(curvnet);
492	INIT_VNET_INET6(curvnet);
493	u_int32_t ulen = m->m_pkthdr.len;
494	u_int32_t plen = sizeof(struct udphdr) + ulen;
495	struct ip6_hdr *ip6;
496	struct udphdr *udp6;
497	struct in6_addr *laddr, *faddr;
498	struct sockaddr_in6 *sin6 = NULL;
499	struct ifnet *oifp = NULL;
500	int scope_ambiguous = 0;
501	u_short fport;
502	int error = 0;
503	struct ip6_pktopts *optp, opt;
504	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
505	int flags;
506	struct sockaddr_in6 tmp;
507
508	INP_WLOCK_ASSERT(inp);
509
510	if (addr6) {
511		/* addr6 has been validated in udp6_send(). */
512		sin6 = (struct sockaddr_in6 *)addr6;
513
514		/* protect *sin6 from overwrites */
515		tmp = *sin6;
516		sin6 = &tmp;
517
518		/*
519		 * Application should provide a proper zone ID or the use of
520		 * default zone IDs should be enabled.  Unfortunately, some
521		 * applications do not behave as it should, so we need a
522		 * workaround.  Even if an appropriate ID is not determined,
523		 * we'll see if we can determine the outgoing interface.  If we
524		 * can, determine the zone ID based on the interface below.
525		 */
526		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
527			scope_ambiguous = 1;
528		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
529			return (error);
530	}
531
532	if (control) {
533		if ((error = ip6_setpktopts(control, &opt,
534		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
535			goto release;
536		optp = &opt;
537	} else
538		optp = inp->in6p_outputopts;
539
540	if (sin6) {
541		faddr = &sin6->sin6_addr;
542
543		/*
544		 * IPv4 version of udp_output calls in_pcbconnect in this case,
545		 * which needs splnet and affects performance.
546		 * Since we saw no essential reason for calling in_pcbconnect,
547		 * we get rid of such kind of logic, and call in6_selectsrc
548		 * and in6_pcbsetport in order to fill in the local address
549		 * and the local port.
550		 */
551		if (sin6->sin6_port == 0) {
552			error = EADDRNOTAVAIL;
553			goto release;
554		}
555
556		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
557			/* how about ::ffff:0.0.0.0 case? */
558			error = EISCONN;
559			goto release;
560		}
561
562		fport = sin6->sin6_port; /* allow 0 port */
563
564		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
565			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
566				/*
567				 * I believe we should explicitly discard the
568				 * packet when mapped addresses are disabled,
569				 * rather than send the packet as an IPv6 one.
570				 * If we chose the latter approach, the packet
571				 * might be sent out on the wire based on the
572				 * default route, the situation which we'd
573				 * probably want to avoid.
574				 * (20010421 jinmei@kame.net)
575				 */
576				error = EINVAL;
577				goto release;
578			}
579			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
580			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
581				/*
582				 * when remote addr is an IPv4-mapped address,
583				 * local addr should not be an IPv6 address,
584				 * since you cannot determine how to map IPv6
585				 * source address to IPv4.
586				 */
587				error = EINVAL;
588				goto release;
589			}
590
591			af = AF_INET;
592		}
593
594		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
595			laddr = in6_selectsrc(sin6, optp, inp, NULL,
596			    td->td_ucred, &oifp, &error);
597			if (oifp && scope_ambiguous &&
598			    (error = in6_setscope(&sin6->sin6_addr,
599			    oifp, NULL))) {
600				goto release;
601			}
602		} else
603			laddr = &inp->in6p_laddr;	/* XXX */
604		if (laddr == NULL) {
605			if (error == 0)
606				error = EADDRNOTAVAIL;
607			goto release;
608		}
609		if (inp->inp_lport == 0 &&
610		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
611			goto release;
612	} else {
613		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
614			error = ENOTCONN;
615			goto release;
616		}
617		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
618			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
619				/*
620				 * XXX: this case would happen when the
621				 * application sets the V6ONLY flag after
622				 * connecting the foreign address.
623				 * Such applications should be fixed,
624				 * so we bark here.
625				 */
626				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
627				    "option was set for a connected socket\n");
628				error = EINVAL;
629				goto release;
630			} else
631				af = AF_INET;
632		}
633		laddr = &inp->in6p_laddr;
634		faddr = &inp->in6p_faddr;
635		fport = inp->inp_fport;
636	}
637
638	if (af == AF_INET)
639		hlen = sizeof(struct ip);
640
641	/*
642	 * Calculate data length and get a mbuf
643	 * for UDP and IP6 headers.
644	 */
645	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
646	if (m == 0) {
647		error = ENOBUFS;
648		goto release;
649	}
650
651	/*
652	 * Stuff checksum and output datagram.
653	 */
654	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
655	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
656	udp6->uh_dport = fport;
657	if (plen <= 0xffff)
658		udp6->uh_ulen = htons((u_short)plen);
659	else
660		udp6->uh_ulen = 0;
661	udp6->uh_sum = 0;
662
663	switch (af) {
664	case AF_INET6:
665		ip6 = mtod(m, struct ip6_hdr *);
666		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
667		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
668		ip6->ip6_vfc	|= IPV6_VERSION;
669#if 0				/* ip6_plen will be filled in ip6_output. */
670		ip6->ip6_plen	= htons((u_short)plen);
671#endif
672		ip6->ip6_nxt	= IPPROTO_UDP;
673		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
674		ip6->ip6_src	= *laddr;
675		ip6->ip6_dst	= *faddr;
676
677		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
678				sizeof(struct ip6_hdr), plen)) == 0) {
679			udp6->uh_sum = 0xffff;
680		}
681
682		flags = 0;
683
684		V_udpstat.udps_opackets++;
685		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
686		    NULL, inp);
687		break;
688	case AF_INET:
689		error = EAFNOSUPPORT;
690		goto release;
691	}
692	goto releaseopt;
693
694release:
695	m_freem(m);
696
697releaseopt:
698	if (control) {
699		ip6_clearpktopts(&opt, -1);
700		m_freem(control);
701	}
702	return (error);
703}
704
705static void
706udp6_abort(struct socket *so)
707{
708	INIT_VNET_INET(so->so_vnet);
709	struct inpcb *inp;
710
711	inp = sotoinpcb(so);
712	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
713
714#ifdef INET
715	if (inp->inp_vflag & INP_IPV4) {
716		struct pr_usrreqs *pru;
717
718		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
719		(*pru->pru_abort)(so);
720		return;
721	}
722#endif
723
724	INP_INFO_WLOCK(&V_udbinfo);
725	INP_WLOCK(inp);
726	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
727		in6_pcbdisconnect(inp);
728		inp->in6p_laddr = in6addr_any;
729		soisdisconnected(so);
730	}
731	INP_WUNLOCK(inp);
732	INP_INFO_WUNLOCK(&V_udbinfo);
733}
734
735static int
736udp6_attach(struct socket *so, int proto, struct thread *td)
737{
738	INIT_VNET_INET(so->so_vnet);
739	struct inpcb *inp;
740	int error;
741
742	inp = sotoinpcb(so);
743	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
744
745	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
746		error = soreserve(so, udp_sendspace, udp_recvspace);
747		if (error)
748			return (error);
749	}
750	INP_INFO_WLOCK(&V_udbinfo);
751	error = in_pcballoc(so, &V_udbinfo);
752	if (error) {
753		INP_INFO_WUNLOCK(&V_udbinfo);
754		return (error);
755	}
756	inp = (struct inpcb *)so->so_pcb;
757	INP_INFO_WUNLOCK(&V_udbinfo);
758	inp->inp_vflag |= INP_IPV6;
759	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
760		inp->inp_vflag |= INP_IPV4;
761	inp->in6p_hops = -1;	/* use kernel default */
762	inp->in6p_cksum = -1;	/* just to be sure */
763	/*
764	 * XXX: ugly!!
765	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
766	 * because the socket may be bound to an IPv6 wildcard address,
767	 * which may match an IPv4-mapped IPv6 address.
768	 */
769	inp->inp_ip_ttl = V_ip_defttl;
770	INP_WUNLOCK(inp);
771	return (0);
772}
773
774static int
775udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
776{
777	INIT_VNET_INET(so->so_vnet);
778	struct inpcb *inp;
779	int error;
780
781	inp = sotoinpcb(so);
782	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
783
784	INP_INFO_WLOCK(&V_udbinfo);
785	INP_WLOCK(inp);
786	inp->inp_vflag &= ~INP_IPV4;
787	inp->inp_vflag |= INP_IPV6;
788	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
789		struct sockaddr_in6 *sin6_p;
790
791		sin6_p = (struct sockaddr_in6 *)nam;
792
793		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
794			inp->inp_vflag |= INP_IPV4;
795		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
796			struct sockaddr_in sin;
797
798			in6_sin6_2_sin(&sin, sin6_p);
799			inp->inp_vflag |= INP_IPV4;
800			inp->inp_vflag &= ~INP_IPV6;
801			error = in_pcbbind(inp, (struct sockaddr *)&sin,
802			    td->td_ucred);
803			goto out;
804		}
805	}
806
807	error = in6_pcbbind(inp, nam, td->td_ucred);
808out:
809	INP_WUNLOCK(inp);
810	INP_INFO_WUNLOCK(&V_udbinfo);
811	return (error);
812}
813
814static void
815udp6_close(struct socket *so)
816{
817	INIT_VNET_INET(so->so_vnet);
818	struct inpcb *inp;
819
820	inp = sotoinpcb(so);
821	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
822
823#ifdef INET
824	if (inp->inp_vflag & INP_IPV4) {
825		struct pr_usrreqs *pru;
826
827		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
828		(*pru->pru_disconnect)(so);
829		return;
830	}
831#endif
832	INP_INFO_WLOCK(&V_udbinfo);
833	INP_WLOCK(inp);
834	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
835		in6_pcbdisconnect(inp);
836		inp->in6p_laddr = in6addr_any;
837		soisdisconnected(so);
838	}
839	INP_WUNLOCK(inp);
840	INP_INFO_WUNLOCK(&V_udbinfo);
841}
842
843static int
844udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
845{
846	INIT_VNET_INET(so->so_vnet);
847	struct inpcb *inp;
848	int error;
849
850	inp = sotoinpcb(so);
851	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
852
853	INP_INFO_WLOCK(&V_udbinfo);
854	INP_WLOCK(inp);
855	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
856		struct sockaddr_in6 *sin6_p;
857
858		sin6_p = (struct sockaddr_in6 *)nam;
859		if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
860			struct sockaddr_in sin;
861
862			if (inp->inp_faddr.s_addr != INADDR_ANY) {
863				error = EISCONN;
864				goto out;
865			}
866			in6_sin6_2_sin(&sin, sin6_p);
867			if (td && jailed(td->td_ucred))
868				if (prison_remote_ip4(td->td_ucred,
869				    &sin.sin_addr) != 0) {
870					error = EAFNOSUPPORT;
871					goto out;
872				}
873			error = in_pcbconnect(inp, (struct sockaddr *)&sin,
874			    td->td_ucred);
875			if (error == 0) {
876				inp->inp_vflag |= INP_IPV4;
877				inp->inp_vflag &= ~INP_IPV6;
878				soisconnected(so);
879			}
880			goto out;
881		}
882	}
883	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
884		error = EISCONN;
885		goto out;
886	}
887	if (td && jailed(td->td_ucred)) {
888		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
889		if (prison_remote_ip6(td->td_ucred, &sin6->sin6_addr) != 0) {
890			error = EAFNOSUPPORT;
891			goto out;
892		}
893	}
894	error = in6_pcbconnect(inp, nam, td->td_ucred);
895	if (error == 0) {
896		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
897			/* should be non mapped addr */
898			inp->inp_vflag &= ~INP_IPV4;
899			inp->inp_vflag |= INP_IPV6;
900		}
901		soisconnected(so);
902	}
903out:
904	INP_WUNLOCK(inp);
905	INP_INFO_WUNLOCK(&V_udbinfo);
906	return (error);
907}
908
909static void
910udp6_detach(struct socket *so)
911{
912	INIT_VNET_INET(so->so_vnet);
913	struct inpcb *inp;
914
915	inp = sotoinpcb(so);
916	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
917
918	INP_INFO_WLOCK(&V_udbinfo);
919	INP_WLOCK(inp);
920	in_pcbdetach(inp);
921	in_pcbfree(inp);
922	INP_INFO_WUNLOCK(&V_udbinfo);
923}
924
925static int
926udp6_disconnect(struct socket *so)
927{
928	INIT_VNET_INET(so->so_vnet);
929	struct inpcb *inp;
930	int error;
931
932	inp = sotoinpcb(so);
933	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
934
935	INP_INFO_WLOCK(&V_udbinfo);
936	INP_WLOCK(inp);
937
938#ifdef INET
939	if (inp->inp_vflag & INP_IPV4) {
940		struct pr_usrreqs *pru;
941
942		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
943		error = (*pru->pru_disconnect)(so);
944		goto out;
945	}
946#endif
947
948	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
949		error = ENOTCONN;
950		goto out;
951	}
952
953	in6_pcbdisconnect(inp);
954	inp->in6p_laddr = in6addr_any;
955	SOCK_LOCK(so);
956	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
957	SOCK_UNLOCK(so);
958out:
959	INP_WUNLOCK(inp);
960	INP_INFO_WUNLOCK(&V_udbinfo);
961	return (0);
962}
963
964static int
965udp6_send(struct socket *so, int flags, struct mbuf *m,
966    struct sockaddr *addr, struct mbuf *control, struct thread *td)
967{
968	INIT_VNET_INET(so->so_vnet);
969	struct inpcb *inp;
970	int error = 0;
971
972	inp = sotoinpcb(so);
973	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
974
975	INP_INFO_WLOCK(&V_udbinfo);
976	INP_WLOCK(inp);
977	if (addr) {
978		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
979			error = EINVAL;
980			goto bad;
981		}
982		if (addr->sa_family != AF_INET6) {
983			error = EAFNOSUPPORT;
984			goto bad;
985		}
986	}
987
988#ifdef INET
989	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
990		int hasv4addr;
991		struct sockaddr_in6 *sin6 = 0;
992
993		if (addr == 0)
994			hasv4addr = (inp->inp_vflag & INP_IPV4);
995		else {
996			sin6 = (struct sockaddr_in6 *)addr;
997			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
998			    ? 1 : 0;
999		}
1000		if (hasv4addr) {
1001			struct pr_usrreqs *pru;
1002
1003			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1004			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
1005				/*
1006				 * When remote addr is IPv4-mapped address,
1007				 * local addr should not be an IPv6 address;
1008				 * since you cannot determine how to map IPv6
1009				 * source address to IPv4.
1010				 */
1011				error = EINVAL;
1012				goto out;
1013			}
1014
1015			/*
1016			 * XXXRW: We release UDP-layer locks before calling
1017			 * udp_send() in order to avoid recursion.  However,
1018			 * this does mean there is a short window where inp's
1019			 * fields are unstable.  Could this lead to a
1020			 * potential race in which the factors causing us to
1021			 * select the UDPv4 output routine are invalidated?
1022			 */
1023			INP_WUNLOCK(inp);
1024			INP_INFO_WUNLOCK(&V_udbinfo);
1025			if (sin6)
1026				in6_sin6_2_sin_in_sock(addr);
1027			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1028			/* addr will just be freed in sendit(). */
1029			return ((*pru->pru_send)(so, flags, m, addr, control,
1030			    td));
1031		}
1032	}
1033#endif
1034#ifdef MAC
1035	mac_inpcb_create_mbuf(inp, m);
1036#endif
1037	error = udp6_output(inp, m, addr, control, td);
1038out:
1039	INP_WUNLOCK(inp);
1040	INP_INFO_WUNLOCK(&V_udbinfo);
1041	return (error);
1042
1043bad:
1044	INP_WUNLOCK(inp);
1045	INP_INFO_WUNLOCK(&V_udbinfo);
1046	m_freem(m);
1047	return (error);
1048}
1049
1050struct pr_usrreqs udp6_usrreqs = {
1051	.pru_abort =		udp6_abort,
1052	.pru_attach =		udp6_attach,
1053	.pru_bind =		udp6_bind,
1054	.pru_connect =		udp6_connect,
1055	.pru_control =		in6_control,
1056	.pru_detach =		udp6_detach,
1057	.pru_disconnect =	udp6_disconnect,
1058	.pru_peeraddr =		in6_mapped_peeraddr,
1059	.pru_send =		udp6_send,
1060	.pru_shutdown =		udp_shutdown,
1061	.pru_sockaddr =		in6_mapped_sockaddr,
1062	.pru_soreceive =	soreceive_dgram,
1063	.pru_sosend =		sosend_dgram,
1064	.pru_sosetlabel =	in_pcbsosetlabel,
1065	.pru_close =		udp6_close
1066};
1067