udp6_usrreq.c revision 183550
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 183550 2008-10-02 15:37:58Z zec $");
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/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#include <sys/vimage.h>
88
89#include <net/if.h>
90#include <net/if_types.h>
91#include <net/route.h>
92
93#include <netinet/in.h>
94#include <netinet/in_pcb.h>
95#include <netinet/in_systm.h>
96#include <netinet/in_var.h>
97#include <netinet/ip.h>
98#include <netinet/ip_icmp.h>
99#include <netinet/ip6.h>
100#include <netinet/icmp_var.h>
101#include <netinet/icmp6.h>
102#include <netinet/ip_var.h>
103#include <netinet/udp.h>
104#include <netinet/udp_var.h>
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
123extern struct protosw	inetsw[];
124static void		udp6_detach(struct socket *so);
125
126static void
127udp6_append(struct inpcb *inp, struct mbuf *n, int off,
128    struct sockaddr_in6 *fromsa)
129{
130	INIT_VNET_INET(inp->inp_vnet);
131	struct socket *so;
132	struct mbuf *opts;
133
134	INP_LOCK_ASSERT(inp);
135
136#ifdef IPSEC
137	/* Check AH/ESP integrity. */
138	if (ipsec6_in_reject(n, inp)) {
139		INIT_VNET_IPSEC(inp->inp_vnet);
140		m_freem(n);
141		V_ipsec6stat.in_polvio++;
142		return;
143	}
144#endif /* IPSEC */
145#ifdef MAC
146	if (mac_inpcb_check_deliver(inp, n) != 0) {
147		m_freem(n);
148		return;
149	}
150#endif
151	opts = NULL;
152	if (inp->in6p_flags & IN6P_CONTROLOPTS ||
153	    inp->inp_socket->so_options & SO_TIMESTAMP)
154		ip6_savecontrol(inp, n, &opts);
155	m_adj(n, off + sizeof(struct udphdr));
156
157	so = inp->inp_socket;
158	SOCKBUF_LOCK(&so->so_rcv);
159	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
160	    opts) == 0) {
161		SOCKBUF_UNLOCK(&so->so_rcv);
162		m_freem(n);
163		if (opts)
164			m_freem(opts);
165		V_udpstat.udps_fullsock++;
166	} else
167		sorwakeup_locked(so);
168}
169
170int
171udp6_input(struct mbuf **mp, int *offp, int proto)
172{
173	INIT_VNET_INET(curvnet);
174	INIT_VNET_INET6(curvnet);
175	struct mbuf *m = *mp;
176	struct ip6_hdr *ip6;
177	struct udphdr *uh;
178	struct inpcb *inp;
179	int off = *offp;
180	int plen, ulen;
181	struct sockaddr_in6 fromsa;
182
183	ip6 = mtod(m, struct ip6_hdr *);
184
185	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
186		/* XXX send icmp6 host/port unreach? */
187		m_freem(m);
188		return (IPPROTO_DONE);
189	}
190
191#ifndef PULLDOWN_TEST
192	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
193	ip6 = mtod(m, struct ip6_hdr *);
194	uh = (struct udphdr *)((caddr_t)ip6 + off);
195#else
196	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
197	if (!uh)
198		return (IPPROTO_DONE);
199#endif
200
201	V_udpstat.udps_ipackets++;
202
203	/*
204	 * Destination port of 0 is illegal, based on RFC768.
205	 */
206	if (uh->uh_dport == 0)
207		goto badunlocked;
208
209	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
210	ulen = ntohs((u_short)uh->uh_ulen);
211
212	if (plen != ulen) {
213		V_udpstat.udps_badlen++;
214		goto badunlocked;
215	}
216
217	/*
218	 * Checksum extended UDP header and data.
219	 */
220	if (uh->uh_sum == 0) {
221		V_udpstat.udps_nosum++;
222		goto badunlocked;
223	}
224	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
225		V_udpstat.udps_badsum++;
226		goto badunlocked;
227	}
228
229	/*
230	 * Construct sockaddr format source address.
231	 */
232	init_sin6(&fromsa, m);
233	fromsa.sin6_port = uh->uh_sport;
234
235	INP_INFO_RLOCK(&V_udbinfo);
236	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
237		struct inpcb *last;
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->in6p_lport != uh->uh_dport)
259				continue;
260			/*
261			 * XXX: Do not check source port of incoming datagram
262			 * unless inp_connect() has been called to bind the
263			 * fport part of the 4-tuple; the source could be
264			 * trying to talk to us with an ephemeral port.
265			 */
266			if (inp->inp_fport != 0 &&
267			    inp->inp_fport != uh->uh_sport)
268				continue;
269			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
270				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
271							&ip6->ip6_dst))
272					continue;
273			}
274			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
275				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
276							&ip6->ip6_src) ||
277				    inp->in6p_fport != uh->uh_sport)
278					continue;
279			}
280
281			if (last != NULL) {
282				struct mbuf *n;
283
284				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
285					INP_RLOCK(last);
286					udp6_append(last, n, off, &fromsa);
287					INP_RUNLOCK(last);
288				}
289			}
290			last = inp;
291			/*
292			 * Don't look for additional matches if this one does
293			 * not have either the SO_REUSEPORT or SO_REUSEADDR
294			 * socket options set.  This heuristic avoids
295			 * searching through all pcbs in the common case of a
296			 * non-shared port.  It assumes that an application
297			 * will never clear these options after setting them.
298			 */
299			if ((last->inp_socket->so_options &
300			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
301				break;
302		}
303
304		if (last == NULL) {
305			/*
306			 * No matching pcb found; discard datagram.  (No need
307			 * to send an ICMP Port Unreachable for a broadcast
308			 * or multicast datgram.)
309			 */
310			V_udpstat.udps_noport++;
311			V_udpstat.udps_noportmcast++;
312			goto badheadlocked;
313		}
314		INP_RLOCK(last);
315		INP_INFO_RUNLOCK(&V_udbinfo);
316		udp6_append(last, m, off, &fromsa);
317		INP_RUNLOCK(last);
318		return (IPPROTO_DONE);
319	}
320	/*
321	 * Locate pcb for datagram.
322	 */
323	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
324	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
325	if (inp == NULL) {
326		if (udp_log_in_vain) {
327			char ip6bufs[INET6_ADDRSTRLEN];
328			char ip6bufd[INET6_ADDRSTRLEN];
329
330			log(LOG_INFO,
331			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
332			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
333			    ntohs(uh->uh_dport),
334			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
335			    ntohs(uh->uh_sport));
336		}
337		V_udpstat.udps_noport++;
338		if (m->m_flags & M_MCAST) {
339			printf("UDP6: M_MCAST is set in a unicast packet.\n");
340			V_udpstat.udps_noportmcast++;
341			goto badheadlocked;
342		}
343		INP_INFO_RUNLOCK(&V_udbinfo);
344		if (V_udp_blackhole)
345			goto badunlocked;
346		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
347			goto badunlocked;
348		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
349		return (IPPROTO_DONE);
350	}
351	INP_RLOCK(inp);
352	INP_INFO_RUNLOCK(&V_udbinfo);
353	udp6_append(inp, m, off, &fromsa);
354	INP_RUNLOCK(inp);
355	return (IPPROTO_DONE);
356
357badheadlocked:
358	INP_INFO_RUNLOCK(&V_udbinfo);
359badunlocked:
360	if (m)
361		m_freem(m);
362	return (IPPROTO_DONE);
363}
364
365void
366udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
367{
368	INIT_VNET_INET(curvnet);
369	struct udphdr uh;
370	struct ip6_hdr *ip6;
371	struct mbuf *m;
372	int off = 0;
373	struct ip6ctlparam *ip6cp = NULL;
374	const struct sockaddr_in6 *sa6_src = NULL;
375	void *cmdarg;
376	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
377	struct udp_portonly {
378		u_int16_t uh_sport;
379		u_int16_t uh_dport;
380	} *uhp;
381
382	if (sa->sa_family != AF_INET6 ||
383	    sa->sa_len != sizeof(struct sockaddr_in6))
384		return;
385
386	if ((unsigned)cmd >= PRC_NCMDS)
387		return;
388	if (PRC_IS_REDIRECT(cmd))
389		notify = in6_rtchange, d = NULL;
390	else if (cmd == PRC_HOSTDEAD)
391		d = NULL;
392	else if (inet6ctlerrmap[cmd] == 0)
393		return;
394
395	/* if the parameter is from icmp6, decode it. */
396	if (d != NULL) {
397		ip6cp = (struct ip6ctlparam *)d;
398		m = ip6cp->ip6c_m;
399		ip6 = ip6cp->ip6c_ip6;
400		off = ip6cp->ip6c_off;
401		cmdarg = ip6cp->ip6c_cmdarg;
402		sa6_src = ip6cp->ip6c_src;
403	} else {
404		m = NULL;
405		ip6 = NULL;
406		cmdarg = NULL;
407		sa6_src = &sa6_any;
408	}
409
410	if (ip6) {
411		/*
412		 * XXX: We assume that when IPV6 is non NULL,
413		 * M and OFF are valid.
414		 */
415
416		/* Check if we can safely examine src and dst ports. */
417		if (m->m_pkthdr.len < off + sizeof(*uhp))
418			return;
419
420		bzero(&uh, sizeof(uh));
421		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
422
423		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
424		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
425		    cmdarg, notify);
426	} else
427		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
428		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
429}
430
431static int
432udp6_getcred(SYSCTL_HANDLER_ARGS)
433{
434	INIT_VNET_INET(curvnet);
435	INIT_VNET_INET6(curvnet);
436	struct xucred xuc;
437	struct sockaddr_in6 addrs[2];
438	struct inpcb *inp;
439	int error;
440
441	error = priv_check(req->td, PRIV_NETINET_GETCRED);
442	if (error)
443		return (error);
444
445	if (req->newlen != sizeof(addrs))
446		return (EINVAL);
447	if (req->oldlen != sizeof(struct xucred))
448		return (EINVAL);
449	error = SYSCTL_IN(req, addrs, sizeof(addrs));
450	if (error)
451		return (error);
452	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
453	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
454		return (error);
455	}
456	INP_INFO_RLOCK(&V_udbinfo);
457	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
458	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
459	    NULL);
460	if (inp != NULL) {
461		INP_RLOCK(inp);
462		INP_INFO_RUNLOCK(&V_udbinfo);
463		if (inp->inp_socket == NULL)
464			error = ENOENT;
465		if (error == 0)
466			error = cr_canseesocket(req->td->td_ucred,
467			    inp->inp_socket);
468		if (error == 0)
469			cru2x(inp->inp_socket->so_cred, &xuc);
470		INP_RUNLOCK(inp);
471	} else {
472		INP_INFO_RUNLOCK(&V_udbinfo);
473		error = ENOENT;
474	}
475	if (error == 0)
476		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
477	return (error);
478}
479
480SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
481    0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
482
483static int
484udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
485    struct mbuf *control, struct thread *td)
486{
487	INIT_VNET_INET(curvnet);
488	INIT_VNET_INET6(curvnet);
489	u_int32_t ulen = m->m_pkthdr.len;
490	u_int32_t plen = sizeof(struct udphdr) + ulen;
491	struct ip6_hdr *ip6;
492	struct udphdr *udp6;
493	struct in6_addr *laddr, *faddr;
494	struct sockaddr_in6 *sin6 = NULL;
495	struct ifnet *oifp = NULL;
496	int scope_ambiguous = 0;
497	u_short fport;
498	int error = 0;
499	struct ip6_pktopts *optp, opt;
500	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
501	int flags;
502	struct sockaddr_in6 tmp;
503
504	INP_WLOCK_ASSERT(inp);
505
506	if (addr6) {
507		/* addr6 has been validated in udp6_send(). */
508		sin6 = (struct sockaddr_in6 *)addr6;
509
510		/* protect *sin6 from overwrites */
511		tmp = *sin6;
512		sin6 = &tmp;
513
514		/*
515		 * Application should provide a proper zone ID or the use of
516		 * default zone IDs should be enabled.  Unfortunately, some
517		 * applications do not behave as it should, so we need a
518		 * workaround.  Even if an appropriate ID is not determined,
519		 * we'll see if we can determine the outgoing interface.  If we
520		 * can, determine the zone ID based on the interface below.
521		 */
522		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
523			scope_ambiguous = 1;
524		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
525			return (error);
526	}
527
528	if (control) {
529		if ((error = ip6_setpktopts(control, &opt,
530		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
531			goto release;
532		optp = &opt;
533	} else
534		optp = inp->in6p_outputopts;
535
536	if (sin6) {
537		faddr = &sin6->sin6_addr;
538
539		/*
540		 * IPv4 version of udp_output calls in_pcbconnect in this case,
541		 * which needs splnet and affects performance.
542		 * Since we saw no essential reason for calling in_pcbconnect,
543		 * we get rid of such kind of logic, and call in6_selectsrc
544		 * and in6_pcbsetport in order to fill in the local address
545		 * and the local port.
546		 */
547		if (sin6->sin6_port == 0) {
548			error = EADDRNOTAVAIL;
549			goto release;
550		}
551
552		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
553			/* how about ::ffff:0.0.0.0 case? */
554			error = EISCONN;
555			goto release;
556		}
557
558		fport = sin6->sin6_port; /* allow 0 port */
559
560		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
561			if ((inp->in6p_flags & IN6P_IPV6_V6ONLY)) {
562				/*
563				 * I believe we should explicitly discard the
564				 * packet when mapped addresses are disabled,
565				 * rather than send the packet as an IPv6 one.
566				 * If we chose the latter approach, the packet
567				 * might be sent out on the wire based on the
568				 * default route, the situation which we'd
569				 * probably want to avoid.
570				 * (20010421 jinmei@kame.net)
571				 */
572				error = EINVAL;
573				goto release;
574			}
575			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
576			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
577				/*
578				 * when remote addr is an IPv4-mapped address,
579				 * local addr should not be an IPv6 address,
580				 * since you cannot determine how to map IPv6
581				 * source address to IPv4.
582				 */
583				error = EINVAL;
584				goto release;
585			}
586
587			af = AF_INET;
588		}
589
590		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
591			laddr = in6_selectsrc(sin6, optp, inp, NULL,
592			    td->td_ucred, &oifp, &error);
593			if (oifp && scope_ambiguous &&
594			    (error = in6_setscope(&sin6->sin6_addr,
595			    oifp, NULL))) {
596				goto release;
597			}
598		} else
599			laddr = &inp->in6p_laddr;	/* XXX */
600		if (laddr == NULL) {
601			if (error == 0)
602				error = EADDRNOTAVAIL;
603			goto release;
604		}
605		if (inp->in6p_lport == 0 &&
606		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
607			goto release;
608	} else {
609		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
610			error = ENOTCONN;
611			goto release;
612		}
613		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
614			if ((inp->in6p_flags & IN6P_IPV6_V6ONLY)) {
615				/*
616				 * XXX: this case would happen when the
617				 * application sets the V6ONLY flag after
618				 * connecting the foreign address.
619				 * Such applications should be fixed,
620				 * so we bark here.
621				 */
622				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
623				    "option was set for a connected socket\n");
624				error = EINVAL;
625				goto release;
626			} else
627				af = AF_INET;
628		}
629		laddr = &inp->in6p_laddr;
630		faddr = &inp->in6p_faddr;
631		fport = inp->in6p_fport;
632	}
633
634	if (af == AF_INET)
635		hlen = sizeof(struct ip);
636
637	/*
638	 * Calculate data length and get a mbuf
639	 * for UDP and IP6 headers.
640	 */
641	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
642	if (m == 0) {
643		error = ENOBUFS;
644		goto release;
645	}
646
647	/*
648	 * Stuff checksum and output datagram.
649	 */
650	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
651	udp6->uh_sport = inp->in6p_lport; /* lport is always set in the PCB */
652	udp6->uh_dport = fport;
653	if (plen <= 0xffff)
654		udp6->uh_ulen = htons((u_short)plen);
655	else
656		udp6->uh_ulen = 0;
657	udp6->uh_sum = 0;
658
659	switch (af) {
660	case AF_INET6:
661		ip6 = mtod(m, struct ip6_hdr *);
662		ip6->ip6_flow	= inp->in6p_flowinfo & IPV6_FLOWINFO_MASK;
663		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
664		ip6->ip6_vfc	|= IPV6_VERSION;
665#if 0				/* ip6_plen will be filled in ip6_output. */
666		ip6->ip6_plen	= htons((u_short)plen);
667#endif
668		ip6->ip6_nxt	= IPPROTO_UDP;
669		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
670		ip6->ip6_src	= *laddr;
671		ip6->ip6_dst	= *faddr;
672
673		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
674				sizeof(struct ip6_hdr), plen)) == 0) {
675			udp6->uh_sum = 0xffff;
676		}
677
678		flags = 0;
679
680		V_udpstat.udps_opackets++;
681		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
682		    NULL, inp);
683		break;
684	case AF_INET:
685		error = EAFNOSUPPORT;
686		goto release;
687	}
688	goto releaseopt;
689
690release:
691	m_freem(m);
692
693releaseopt:
694	if (control) {
695		ip6_clearpktopts(&opt, -1);
696		m_freem(control);
697	}
698	return (error);
699}
700
701static void
702udp6_abort(struct socket *so)
703{
704	INIT_VNET_INET(so->so_vnet);
705	struct inpcb *inp;
706
707	inp = sotoinpcb(so);
708	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
709
710#ifdef INET
711	if (inp->inp_vflag & INP_IPV4) {
712		struct pr_usrreqs *pru;
713
714		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
715		(*pru->pru_abort)(so);
716		return;
717	}
718#endif
719
720	INP_INFO_WLOCK(&V_udbinfo);
721	INP_WLOCK(inp);
722	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
723		in6_pcbdisconnect(inp);
724		inp->in6p_laddr = in6addr_any;
725		soisdisconnected(so);
726	}
727	INP_WUNLOCK(inp);
728	INP_INFO_WUNLOCK(&V_udbinfo);
729}
730
731static int
732udp6_attach(struct socket *so, int proto, struct thread *td)
733{
734	INIT_VNET_INET(so->so_vnet);
735	struct inpcb *inp;
736	int error;
737
738	inp = sotoinpcb(so);
739	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
740
741	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
742		error = soreserve(so, udp_sendspace, udp_recvspace);
743		if (error)
744			return (error);
745	}
746	INP_INFO_WLOCK(&V_udbinfo);
747	error = in_pcballoc(so, &V_udbinfo);
748	if (error) {
749		INP_INFO_WUNLOCK(&V_udbinfo);
750		return (error);
751	}
752	inp = (struct inpcb *)so->so_pcb;
753	INP_INFO_WUNLOCK(&V_udbinfo);
754	inp->inp_vflag |= INP_IPV6;
755	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
756		inp->inp_vflag |= INP_IPV4;
757	inp->in6p_hops = -1;	/* use kernel default */
758	inp->in6p_cksum = -1;	/* just to be sure */
759	/*
760	 * XXX: ugly!!
761	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
762	 * because the socket may be bound to an IPv6 wildcard address,
763	 * which may match an IPv4-mapped IPv6 address.
764	 */
765	inp->inp_ip_ttl = V_ip_defttl;
766	INP_WUNLOCK(inp);
767	return (0);
768}
769
770static int
771udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
772{
773	INIT_VNET_INET(so->so_vnet);
774	struct inpcb *inp;
775	int error;
776
777	inp = sotoinpcb(so);
778	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
779
780	INP_INFO_WLOCK(&V_udbinfo);
781	INP_WLOCK(inp);
782	inp->inp_vflag &= ~INP_IPV4;
783	inp->inp_vflag |= INP_IPV6;
784	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
785		struct sockaddr_in6 *sin6_p;
786
787		sin6_p = (struct sockaddr_in6 *)nam;
788
789		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
790			inp->inp_vflag |= INP_IPV4;
791		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
792			struct sockaddr_in sin;
793
794			in6_sin6_2_sin(&sin, sin6_p);
795			inp->inp_vflag |= INP_IPV4;
796			inp->inp_vflag &= ~INP_IPV6;
797			error = in_pcbbind(inp, (struct sockaddr *)&sin,
798			    td->td_ucred);
799			goto out;
800		}
801	}
802
803	error = in6_pcbbind(inp, nam, td->td_ucred);
804out:
805	INP_WUNLOCK(inp);
806	INP_INFO_WUNLOCK(&V_udbinfo);
807	return (error);
808}
809
810static void
811udp6_close(struct socket *so)
812{
813	INIT_VNET_INET(so->so_vnet);
814	struct inpcb *inp;
815
816	inp = sotoinpcb(so);
817	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
818
819#ifdef INET
820	if (inp->inp_vflag & INP_IPV4) {
821		struct pr_usrreqs *pru;
822
823		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
824		(*pru->pru_disconnect)(so);
825		return;
826	}
827#endif
828	INP_INFO_WLOCK(&V_udbinfo);
829	INP_WLOCK(inp);
830	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
831		in6_pcbdisconnect(inp);
832		inp->in6p_laddr = in6addr_any;
833		soisdisconnected(so);
834	}
835	INP_WUNLOCK(inp);
836	INP_INFO_WUNLOCK(&V_udbinfo);
837}
838
839static int
840udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
841{
842	INIT_VNET_INET(so->so_vnet);
843	struct inpcb *inp;
844	int error;
845
846	inp = sotoinpcb(so);
847	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
848
849	INP_INFO_WLOCK(&V_udbinfo);
850	INP_WLOCK(inp);
851	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
852		struct sockaddr_in6 *sin6_p;
853
854		sin6_p = (struct sockaddr_in6 *)nam;
855		if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
856			struct sockaddr_in sin;
857
858			if (inp->inp_faddr.s_addr != INADDR_ANY) {
859				error = EISCONN;
860				goto out;
861			}
862			in6_sin6_2_sin(&sin, sin6_p);
863			error = in_pcbconnect(inp, (struct sockaddr *)&sin,
864			    td->td_ucred);
865			if (error == 0) {
866				inp->inp_vflag |= INP_IPV4;
867				inp->inp_vflag &= ~INP_IPV6;
868				soisconnected(so);
869			}
870			goto out;
871		}
872	}
873	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
874		error = EISCONN;
875		goto out;
876	}
877	error = in6_pcbconnect(inp, nam, td->td_ucred);
878	if (error == 0) {
879		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
880			/* should be non mapped addr */
881			inp->inp_vflag &= ~INP_IPV4;
882			inp->inp_vflag |= INP_IPV6;
883		}
884		soisconnected(so);
885	}
886out:
887	INP_WUNLOCK(inp);
888	INP_INFO_WUNLOCK(&V_udbinfo);
889	return (error);
890}
891
892static void
893udp6_detach(struct socket *so)
894{
895	INIT_VNET_INET(so->so_vnet);
896	struct inpcb *inp;
897
898	inp = sotoinpcb(so);
899	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
900
901	INP_INFO_WLOCK(&V_udbinfo);
902	INP_WLOCK(inp);
903	in6_pcbdetach(inp);
904	in6_pcbfree(inp);
905	INP_INFO_WUNLOCK(&V_udbinfo);
906}
907
908static int
909udp6_disconnect(struct socket *so)
910{
911	INIT_VNET_INET(so->so_vnet);
912	struct inpcb *inp;
913	int error;
914
915	inp = sotoinpcb(so);
916	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
917
918	INP_INFO_WLOCK(&V_udbinfo);
919	INP_WLOCK(inp);
920
921#ifdef INET
922	if (inp->inp_vflag & INP_IPV4) {
923		struct pr_usrreqs *pru;
924
925		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
926		error = (*pru->pru_disconnect)(so);
927		goto out;
928	}
929#endif
930
931	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
932		error = ENOTCONN;
933		goto out;
934	}
935
936	in6_pcbdisconnect(inp);
937	inp->in6p_laddr = in6addr_any;
938	/* XXXRW: so_state locking? */
939	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
940out:
941	INP_WUNLOCK(inp);
942	INP_INFO_WUNLOCK(&V_udbinfo);
943	return (0);
944}
945
946static int
947udp6_send(struct socket *so, int flags, struct mbuf *m,
948    struct sockaddr *addr, struct mbuf *control, struct thread *td)
949{
950	INIT_VNET_INET(so->so_vnet);
951	struct inpcb *inp;
952	int error = 0;
953
954	inp = sotoinpcb(so);
955	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
956
957	INP_INFO_WLOCK(&V_udbinfo);
958	INP_WLOCK(inp);
959	if (addr) {
960		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
961			error = EINVAL;
962			goto bad;
963		}
964		if (addr->sa_family != AF_INET6) {
965			error = EAFNOSUPPORT;
966			goto bad;
967		}
968	}
969
970#ifdef INET
971	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
972		int hasv4addr;
973		struct sockaddr_in6 *sin6 = 0;
974
975		if (addr == 0)
976			hasv4addr = (inp->inp_vflag & INP_IPV4);
977		else {
978			sin6 = (struct sockaddr_in6 *)addr;
979			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
980			    ? 1 : 0;
981		}
982		if (hasv4addr) {
983			struct pr_usrreqs *pru;
984
985			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
986			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
987				/*
988				 * When remote addr is IPv4-mapped address,
989				 * local addr should not be an IPv6 address;
990				 * since you cannot determine how to map IPv6
991				 * source address to IPv4.
992				 */
993				error = EINVAL;
994				goto out;
995			}
996
997			/*
998			 * XXXRW: We release UDP-layer locks before calling
999			 * udp_send() in order to avoid recursion.  However,
1000			 * this does mean there is a short window where inp's
1001			 * fields are unstable.  Could this lead to a
1002			 * potential race in which the factors causing us to
1003			 * select the UDPv4 output routine are invalidated?
1004			 */
1005			INP_WUNLOCK(inp);
1006			INP_INFO_WUNLOCK(&V_udbinfo);
1007			if (sin6)
1008				in6_sin6_2_sin_in_sock(addr);
1009			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1010			/* addr will just be freed in sendit(). */
1011			return ((*pru->pru_send)(so, flags, m, addr, control,
1012			    td));
1013		}
1014	}
1015#endif
1016#ifdef MAC
1017	mac_inpcb_create_mbuf(inp, m);
1018#endif
1019	error = udp6_output(inp, m, addr, control, td);
1020out:
1021	INP_WUNLOCK(inp);
1022	INP_INFO_WUNLOCK(&V_udbinfo);
1023	return (error);
1024
1025bad:
1026	INP_WUNLOCK(inp);
1027	INP_INFO_WUNLOCK(&V_udbinfo);
1028	m_freem(m);
1029	return (error);
1030}
1031
1032struct pr_usrreqs udp6_usrreqs = {
1033	.pru_abort =		udp6_abort,
1034	.pru_attach =		udp6_attach,
1035	.pru_bind =		udp6_bind,
1036	.pru_connect =		udp6_connect,
1037	.pru_control =		in6_control,
1038	.pru_detach =		udp6_detach,
1039	.pru_disconnect =	udp6_disconnect,
1040	.pru_peeraddr =		in6_mapped_peeraddr,
1041	.pru_send =		udp6_send,
1042	.pru_shutdown =		udp_shutdown,
1043	.pru_sockaddr =		in6_mapped_sockaddr,
1044	.pru_soreceive =	soreceive_dgram,
1045	.pru_sosend =		sosend_dgram,
1046	.pru_sosetlabel =	in_pcbsosetlabel,
1047	.pru_close =		udp6_close
1048};
1049