udp6_usrreq.c revision 181803
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 181803 2008-08-17 23:27:27Z 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/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	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->in6p_flags & IN6P_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		V_udpstat.udps_fullsock++;
164	} else
165		sorwakeup_locked(so);
166}
167
168int
169udp6_input(struct mbuf **mp, int *offp, int proto)
170{
171	struct mbuf *m = *mp;
172	struct ip6_hdr *ip6;
173	struct udphdr *uh;
174	struct inpcb *inp;
175	int off = *offp;
176	int plen, ulen;
177	struct sockaddr_in6 fromsa;
178
179	ip6 = mtod(m, struct ip6_hdr *);
180
181	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
182		/* XXX send icmp6 host/port unreach? */
183		m_freem(m);
184		return (IPPROTO_DONE);
185	}
186
187#ifndef PULLDOWN_TEST
188	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
189	ip6 = mtod(m, struct ip6_hdr *);
190	uh = (struct udphdr *)((caddr_t)ip6 + off);
191#else
192	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
193	if (!uh)
194		return (IPPROTO_DONE);
195#endif
196
197	V_udpstat.udps_ipackets++;
198
199	/*
200	 * Destination port of 0 is illegal, based on RFC768.
201	 */
202	if (uh->uh_dport == 0)
203		goto badunlocked;
204
205	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
206	ulen = ntohs((u_short)uh->uh_ulen);
207
208	if (plen != ulen) {
209		V_udpstat.udps_badlen++;
210		goto badunlocked;
211	}
212
213	/*
214	 * Checksum extended UDP header and data.
215	 */
216	if (uh->uh_sum == 0) {
217		V_udpstat.udps_nosum++;
218		goto badunlocked;
219	}
220	if (in6_cksum(m, IPPROTO_UDP, off, ulen) != 0) {
221		V_udpstat.udps_badsum++;
222		goto badunlocked;
223	}
224
225	/*
226	 * Construct sockaddr format source address.
227	 */
228	init_sin6(&fromsa, m);
229	fromsa.sin6_port = uh->uh_sport;
230
231	INP_INFO_RLOCK(&V_udbinfo);
232	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
233		struct inpcb *last;
234
235		/*
236		 * In the event that laddr should be set to the link-local
237		 * address (this happens in RIPng), the multicast address
238		 * specified in the received packet will not match laddr.  To
239		 * handle this situation, matching is relaxed if the
240		 * receiving interface is the same as one specified in the
241		 * socket and if the destination multicast address matches
242		 * one of the multicast groups specified in the socket.
243		 */
244
245		/*
246		 * KAME note: traditionally we dropped udpiphdr from mbuf
247		 * here.  We need udphdr for IPsec processing so we do that
248		 * later.
249		 */
250		last = NULL;
251		LIST_FOREACH(inp, &V_udb, inp_list) {
252			if ((inp->inp_vflag & INP_IPV6) == 0)
253				continue;
254			if (inp->in6p_lport != uh->uh_dport)
255				continue;
256			/*
257			 * XXX: Do not check source port of incoming datagram
258			 * unless inp_connect() has been called to bind the
259			 * fport part of the 4-tuple; the source could be
260			 * trying to talk to us with an ephemeral port.
261			 */
262			if (inp->inp_fport != 0 &&
263			    inp->inp_fport != uh->uh_sport)
264				continue;
265			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
266				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
267							&ip6->ip6_dst))
268					continue;
269			}
270			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
271				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
272							&ip6->ip6_src) ||
273				    inp->in6p_fport != uh->uh_sport)
274					continue;
275			}
276
277			if (last != NULL) {
278				struct mbuf *n;
279
280				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
281					INP_RLOCK(last);
282					udp6_append(last, n, off, &fromsa);
283					INP_RUNLOCK(last);
284				}
285			}
286			last = inp;
287			/*
288			 * Don't look for additional matches if this one does
289			 * not have either the SO_REUSEPORT or SO_REUSEADDR
290			 * socket options set.  This heuristic avoids
291			 * searching through all pcbs in the common case of a
292			 * non-shared port.  It assumes that an application
293			 * will never clear these options after setting them.
294			 */
295			if ((last->inp_socket->so_options &
296			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
297				break;
298		}
299
300		if (last == NULL) {
301			/*
302			 * No matching pcb found; discard datagram.  (No need
303			 * to send an ICMP Port Unreachable for a broadcast
304			 * or multicast datgram.)
305			 */
306			V_udpstat.udps_noport++;
307			V_udpstat.udps_noportmcast++;
308			goto badheadlocked;
309		}
310		INP_RLOCK(last);
311		udp6_append(last, m, off, &fromsa);
312		INP_RUNLOCK(last);
313		INP_INFO_RUNLOCK(&V_udbinfo);
314		return (IPPROTO_DONE);
315	}
316	/*
317	 * Locate pcb for datagram.
318	 */
319	inp = in6_pcblookup_hash(&V_udbinfo, &ip6->ip6_src, uh->uh_sport,
320	    &ip6->ip6_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
321	if (inp == NULL) {
322		if (udp_log_in_vain) {
323			char ip6bufs[INET6_ADDRSTRLEN];
324			char ip6bufd[INET6_ADDRSTRLEN];
325
326			log(LOG_INFO,
327			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
328			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
329			    ntohs(uh->uh_dport),
330			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
331			    ntohs(uh->uh_sport));
332		}
333		V_udpstat.udps_noport++;
334		if (m->m_flags & M_MCAST) {
335			printf("UDP6: M_MCAST is set in a unicast packet.\n");
336			V_udpstat.udps_noportmcast++;
337			goto badheadlocked;
338		}
339		INP_INFO_RUNLOCK(&V_udbinfo);
340		if (V_udp_blackhole)
341			goto badunlocked;
342		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
343			goto badunlocked;
344		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
345		return (IPPROTO_DONE);
346	}
347	INP_RLOCK(inp);
348	INP_INFO_RUNLOCK(&V_udbinfo);
349	udp6_append(inp, m, off, &fromsa);
350	INP_RUNLOCK(inp);
351	return (IPPROTO_DONE);
352
353badheadlocked:
354	INP_INFO_RUNLOCK(&V_udbinfo);
355badunlocked:
356	if (m)
357		m_freem(m);
358	return (IPPROTO_DONE);
359}
360
361void
362udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
363{
364	struct udphdr uh;
365	struct ip6_hdr *ip6;
366	struct mbuf *m;
367	int off = 0;
368	struct ip6ctlparam *ip6cp = NULL;
369	const struct sockaddr_in6 *sa6_src = NULL;
370	void *cmdarg;
371	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
372	struct udp_portonly {
373		u_int16_t uh_sport;
374		u_int16_t uh_dport;
375	} *uhp;
376
377	if (sa->sa_family != AF_INET6 ||
378	    sa->sa_len != sizeof(struct sockaddr_in6))
379		return;
380
381	if ((unsigned)cmd >= PRC_NCMDS)
382		return;
383	if (PRC_IS_REDIRECT(cmd))
384		notify = in6_rtchange, d = NULL;
385	else if (cmd == PRC_HOSTDEAD)
386		d = NULL;
387	else if (inet6ctlerrmap[cmd] == 0)
388		return;
389
390	/* if the parameter is from icmp6, decode it. */
391	if (d != NULL) {
392		ip6cp = (struct ip6ctlparam *)d;
393		m = ip6cp->ip6c_m;
394		ip6 = ip6cp->ip6c_ip6;
395		off = ip6cp->ip6c_off;
396		cmdarg = ip6cp->ip6c_cmdarg;
397		sa6_src = ip6cp->ip6c_src;
398	} else {
399		m = NULL;
400		ip6 = NULL;
401		cmdarg = NULL;
402		sa6_src = &sa6_any;
403	}
404
405	if (ip6) {
406		/*
407		 * XXX: We assume that when IPV6 is non NULL,
408		 * M and OFF are valid.
409		 */
410
411		/* Check if we can safely examine src and dst ports. */
412		if (m->m_pkthdr.len < off + sizeof(*uhp))
413			return;
414
415		bzero(&uh, sizeof(uh));
416		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
417
418		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
419		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
420		    cmdarg, notify);
421	} else
422		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
423		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
424}
425
426static int
427udp6_getcred(SYSCTL_HANDLER_ARGS)
428{
429	struct xucred xuc;
430	struct sockaddr_in6 addrs[2];
431	struct inpcb *inp;
432	int error;
433
434	error = priv_check(req->td, PRIV_NETINET_GETCRED);
435	if (error)
436		return (error);
437
438	if (req->newlen != sizeof(addrs))
439		return (EINVAL);
440	if (req->oldlen != sizeof(struct xucred))
441		return (EINVAL);
442	error = SYSCTL_IN(req, addrs, sizeof(addrs));
443	if (error)
444		return (error);
445	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
446	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
447		return (error);
448	}
449	INP_INFO_RLOCK(&V_udbinfo);
450	inp = in6_pcblookup_hash(&V_udbinfo, &addrs[1].sin6_addr,
451	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port, 1,
452	    NULL);
453	if (inp != NULL) {
454		INP_RLOCK(inp);
455		INP_INFO_RUNLOCK(&V_udbinfo);
456		if (inp->inp_socket == NULL)
457			error = ENOENT;
458		if (error == 0)
459			error = cr_canseesocket(req->td->td_ucred,
460			    inp->inp_socket);
461		if (error == 0)
462			cru2x(inp->inp_socket->so_cred, &xuc);
463		INP_RUNLOCK(inp);
464	} else {
465		INP_INFO_RUNLOCK(&V_udbinfo);
466		error = ENOENT;
467	}
468	INP_RUNLOCK(inp);
469	INP_INFO_RUNLOCK(&V_udbinfo);
470	if (error == 0)
471		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
472	return (error);
473}
474
475SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
476    0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
477
478static int
479udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
480    struct mbuf *control, struct thread *td)
481{
482	u_int32_t ulen = m->m_pkthdr.len;
483	u_int32_t plen = sizeof(struct udphdr) + ulen;
484	struct ip6_hdr *ip6;
485	struct udphdr *udp6;
486	struct in6_addr *laddr, *faddr;
487	struct sockaddr_in6 *sin6 = NULL;
488	struct ifnet *oifp = NULL;
489	int scope_ambiguous = 0;
490	u_short fport;
491	int error = 0;
492	struct ip6_pktopts *optp, opt;
493	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
494	int flags;
495	struct sockaddr_in6 tmp;
496
497	INP_WLOCK_ASSERT(inp);
498
499	if (addr6) {
500		/* addr6 has been validated in udp6_send(). */
501		sin6 = (struct sockaddr_in6 *)addr6;
502
503		/* protect *sin6 from overwrites */
504		tmp = *sin6;
505		sin6 = &tmp;
506
507		/*
508		 * Application should provide a proper zone ID or the use of
509		 * default zone IDs should be enabled.  Unfortunately, some
510		 * applications do not behave as it should, so we need a
511		 * workaround.  Even if an appropriate ID is not determined,
512		 * we'll see if we can determine the outgoing interface.  If we
513		 * can, determine the zone ID based on the interface below.
514		 */
515		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
516			scope_ambiguous = 1;
517		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
518			return (error);
519	}
520
521	if (control) {
522		if ((error = ip6_setpktopts(control, &opt,
523		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
524			goto release;
525		optp = &opt;
526	} else
527		optp = inp->in6p_outputopts;
528
529	if (sin6) {
530		faddr = &sin6->sin6_addr;
531
532		/*
533		 * IPv4 version of udp_output calls in_pcbconnect in this case,
534		 * which needs splnet and affects performance.
535		 * Since we saw no essential reason for calling in_pcbconnect,
536		 * we get rid of such kind of logic, and call in6_selectsrc
537		 * and in6_pcbsetport in order to fill in the local address
538		 * and the local port.
539		 */
540		if (sin6->sin6_port == 0) {
541			error = EADDRNOTAVAIL;
542			goto release;
543		}
544
545		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
546			/* how about ::ffff:0.0.0.0 case? */
547			error = EISCONN;
548			goto release;
549		}
550
551		fport = sin6->sin6_port; /* allow 0 port */
552
553		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
554			if ((inp->in6p_flags & IN6P_IPV6_V6ONLY)) {
555				/*
556				 * I believe we should explicitly discard the
557				 * packet when mapped addresses are disabled,
558				 * rather than send the packet as an IPv6 one.
559				 * If we chose the latter approach, the packet
560				 * might be sent out on the wire based on the
561				 * default route, the situation which we'd
562				 * probably want to avoid.
563				 * (20010421 jinmei@kame.net)
564				 */
565				error = EINVAL;
566				goto release;
567			}
568			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
569			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
570				/*
571				 * when remote addr is an IPv4-mapped address,
572				 * local addr should not be an IPv6 address,
573				 * since you cannot determine how to map IPv6
574				 * source address to IPv4.
575				 */
576				error = EINVAL;
577				goto release;
578			}
579
580			af = AF_INET;
581		}
582
583		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
584			laddr = in6_selectsrc(sin6, optp, inp, NULL,
585			    td->td_ucred, &oifp, &error);
586			if (oifp && scope_ambiguous &&
587			    (error = in6_setscope(&sin6->sin6_addr,
588			    oifp, NULL))) {
589				goto release;
590			}
591		} else
592			laddr = &inp->in6p_laddr;	/* XXX */
593		if (laddr == NULL) {
594			if (error == 0)
595				error = EADDRNOTAVAIL;
596			goto release;
597		}
598		if (inp->in6p_lport == 0 &&
599		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0)
600			goto release;
601	} else {
602		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
603			error = ENOTCONN;
604			goto release;
605		}
606		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
607			if ((inp->in6p_flags & IN6P_IPV6_V6ONLY)) {
608				/*
609				 * XXX: this case would happen when the
610				 * application sets the V6ONLY flag after
611				 * connecting the foreign address.
612				 * Such applications should be fixed,
613				 * so we bark here.
614				 */
615				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
616				    "option was set for a connected socket\n");
617				error = EINVAL;
618				goto release;
619			} else
620				af = AF_INET;
621		}
622		laddr = &inp->in6p_laddr;
623		faddr = &inp->in6p_faddr;
624		fport = inp->in6p_fport;
625	}
626
627	if (af == AF_INET)
628		hlen = sizeof(struct ip);
629
630	/*
631	 * Calculate data length and get a mbuf
632	 * for UDP and IP6 headers.
633	 */
634	M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
635	if (m == 0) {
636		error = ENOBUFS;
637		goto release;
638	}
639
640	/*
641	 * Stuff checksum and output datagram.
642	 */
643	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
644	udp6->uh_sport = inp->in6p_lport; /* lport is always set in the PCB */
645	udp6->uh_dport = fport;
646	if (plen <= 0xffff)
647		udp6->uh_ulen = htons((u_short)plen);
648	else
649		udp6->uh_ulen = 0;
650	udp6->uh_sum = 0;
651
652	switch (af) {
653	case AF_INET6:
654		ip6 = mtod(m, struct ip6_hdr *);
655		ip6->ip6_flow	= inp->in6p_flowinfo & IPV6_FLOWINFO_MASK;
656		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
657		ip6->ip6_vfc	|= IPV6_VERSION;
658#if 0				/* ip6_plen will be filled in ip6_output. */
659		ip6->ip6_plen	= htons((u_short)plen);
660#endif
661		ip6->ip6_nxt	= IPPROTO_UDP;
662		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
663		ip6->ip6_src	= *laddr;
664		ip6->ip6_dst	= *faddr;
665
666		if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
667				sizeof(struct ip6_hdr), plen)) == 0) {
668			udp6->uh_sum = 0xffff;
669		}
670
671		flags = 0;
672
673		V_udpstat.udps_opackets++;
674		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
675		    NULL, inp);
676		break;
677	case AF_INET:
678		error = EAFNOSUPPORT;
679		goto release;
680	}
681	goto releaseopt;
682
683release:
684	m_freem(m);
685
686releaseopt:
687	if (control) {
688		ip6_clearpktopts(&opt, -1);
689		m_freem(control);
690	}
691	return (error);
692}
693
694static void
695udp6_abort(struct socket *so)
696{
697	struct inpcb *inp;
698
699	inp = sotoinpcb(so);
700	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
701
702#ifdef INET
703	if (inp->inp_vflag & INP_IPV4) {
704		struct pr_usrreqs *pru;
705
706		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
707		(*pru->pru_abort)(so);
708		return;
709	}
710#endif
711
712	INP_INFO_WLOCK(&V_udbinfo);
713	INP_WLOCK(inp);
714	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
715		in6_pcbdisconnect(inp);
716		inp->in6p_laddr = in6addr_any;
717		soisdisconnected(so);
718	}
719	INP_WUNLOCK(inp);
720	INP_INFO_WUNLOCK(&V_udbinfo);
721}
722
723static int
724udp6_attach(struct socket *so, int proto, struct thread *td)
725{
726	struct inpcb *inp;
727	int error;
728
729	inp = sotoinpcb(so);
730	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
731
732	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
733		error = soreserve(so, udp_sendspace, udp_recvspace);
734		if (error)
735			return (error);
736	}
737	INP_INFO_WLOCK(&V_udbinfo);
738	error = in_pcballoc(so, &V_udbinfo);
739	if (error) {
740		INP_INFO_WUNLOCK(&V_udbinfo);
741		return (error);
742	}
743	inp = (struct inpcb *)so->so_pcb;
744	INP_INFO_WUNLOCK(&V_udbinfo);
745	inp->inp_vflag |= INP_IPV6;
746	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
747		inp->inp_vflag |= INP_IPV4;
748	inp->in6p_hops = -1;	/* use kernel default */
749	inp->in6p_cksum = -1;	/* just to be sure */
750	/*
751	 * XXX: ugly!!
752	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
753	 * because the socket may be bound to an IPv6 wildcard address,
754	 * which may match an IPv4-mapped IPv6 address.
755	 */
756	inp->inp_ip_ttl = V_ip_defttl;
757	INP_WUNLOCK(inp);
758	return (0);
759}
760
761static int
762udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
763{
764	struct inpcb *inp;
765	int error;
766
767	inp = sotoinpcb(so);
768	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
769
770	INP_INFO_WLOCK(&V_udbinfo);
771	INP_WLOCK(inp);
772	inp->inp_vflag &= ~INP_IPV4;
773	inp->inp_vflag |= INP_IPV6;
774	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
775		struct sockaddr_in6 *sin6_p;
776
777		sin6_p = (struct sockaddr_in6 *)nam;
778
779		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
780			inp->inp_vflag |= INP_IPV4;
781		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
782			struct sockaddr_in sin;
783
784			in6_sin6_2_sin(&sin, sin6_p);
785			inp->inp_vflag |= INP_IPV4;
786			inp->inp_vflag &= ~INP_IPV6;
787			error = in_pcbbind(inp, (struct sockaddr *)&sin,
788			    td->td_ucred);
789			goto out;
790		}
791	}
792
793	error = in6_pcbbind(inp, nam, td->td_ucred);
794out:
795	INP_WUNLOCK(inp);
796	INP_INFO_WUNLOCK(&V_udbinfo);
797	return (error);
798}
799
800static void
801udp6_close(struct socket *so)
802{
803	struct inpcb *inp;
804
805	inp = sotoinpcb(so);
806	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
807
808#ifdef INET
809	if (inp->inp_vflag & INP_IPV4) {
810		struct pr_usrreqs *pru;
811
812		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
813		(*pru->pru_disconnect)(so);
814		return;
815	}
816#endif
817	INP_INFO_WLOCK(&V_udbinfo);
818	INP_WLOCK(inp);
819	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
820		in6_pcbdisconnect(inp);
821		inp->in6p_laddr = in6addr_any;
822		soisdisconnected(so);
823	}
824	INP_WUNLOCK(inp);
825	INP_INFO_WUNLOCK(&V_udbinfo);
826}
827
828static int
829udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
830{
831	struct inpcb *inp;
832	int error;
833
834	inp = sotoinpcb(so);
835	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
836
837	INP_INFO_WLOCK(&V_udbinfo);
838	INP_WLOCK(inp);
839	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
840		struct sockaddr_in6 *sin6_p;
841
842		sin6_p = (struct sockaddr_in6 *)nam;
843		if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
844			struct sockaddr_in sin;
845
846			if (inp->inp_faddr.s_addr != INADDR_ANY) {
847				error = EISCONN;
848				goto out;
849			}
850			in6_sin6_2_sin(&sin, sin6_p);
851			error = in_pcbconnect(inp, (struct sockaddr *)&sin,
852			    td->td_ucred);
853			if (error == 0) {
854				inp->inp_vflag |= INP_IPV4;
855				inp->inp_vflag &= ~INP_IPV6;
856				soisconnected(so);
857			}
858			goto out;
859		}
860	}
861	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
862		error = EISCONN;
863		goto out;
864	}
865	error = in6_pcbconnect(inp, nam, td->td_ucred);
866	if (error == 0) {
867		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
868			/* should be non mapped addr */
869			inp->inp_vflag &= ~INP_IPV4;
870			inp->inp_vflag |= INP_IPV6;
871		}
872		soisconnected(so);
873	}
874out:
875	INP_WUNLOCK(inp);
876	INP_INFO_WUNLOCK(&V_udbinfo);
877	return (error);
878}
879
880static void
881udp6_detach(struct socket *so)
882{
883	struct inpcb *inp;
884
885	inp = sotoinpcb(so);
886	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
887
888	INP_INFO_WLOCK(&V_udbinfo);
889	INP_WLOCK(inp);
890	in6_pcbdetach(inp);
891	in6_pcbfree(inp);
892	INP_INFO_WUNLOCK(&V_udbinfo);
893}
894
895static int
896udp6_disconnect(struct socket *so)
897{
898	struct inpcb *inp;
899	int error;
900
901	inp = sotoinpcb(so);
902	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
903
904	INP_INFO_WLOCK(&V_udbinfo);
905	INP_WLOCK(inp);
906
907#ifdef INET
908	if (inp->inp_vflag & INP_IPV4) {
909		struct pr_usrreqs *pru;
910
911		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
912		error = (*pru->pru_disconnect)(so);
913		goto out;
914	}
915#endif
916
917	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
918		error = ENOTCONN;
919		goto out;
920	}
921
922	in6_pcbdisconnect(inp);
923	inp->in6p_laddr = in6addr_any;
924	/* XXXRW: so_state locking? */
925	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
926out:
927	INP_WUNLOCK(inp);
928	INP_INFO_WUNLOCK(&V_udbinfo);
929	return (0);
930}
931
932static int
933udp6_send(struct socket *so, int flags, struct mbuf *m,
934    struct sockaddr *addr, struct mbuf *control, struct thread *td)
935{
936	struct inpcb *inp;
937	int error = 0;
938
939	inp = sotoinpcb(so);
940	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
941
942	INP_INFO_WLOCK(&V_udbinfo);
943	INP_WLOCK(inp);
944	if (addr) {
945		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
946			error = EINVAL;
947			goto bad;
948		}
949		if (addr->sa_family != AF_INET6) {
950			error = EAFNOSUPPORT;
951			goto bad;
952		}
953	}
954
955#ifdef INET
956	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
957		int hasv4addr;
958		struct sockaddr_in6 *sin6 = 0;
959
960		if (addr == 0)
961			hasv4addr = (inp->inp_vflag & INP_IPV4);
962		else {
963			sin6 = (struct sockaddr_in6 *)addr;
964			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
965			    ? 1 : 0;
966		}
967		if (hasv4addr) {
968			struct pr_usrreqs *pru;
969
970			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
971			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
972				/*
973				 * When remote addr is IPv4-mapped address,
974				 * local addr should not be an IPv6 address;
975				 * since you cannot determine how to map IPv6
976				 * source address to IPv4.
977				 */
978				error = EINVAL;
979				goto out;
980			}
981			if (sin6)
982				in6_sin6_2_sin_in_sock(addr);
983			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
984			error = ((*pru->pru_send)(so, flags, m, addr, control,
985			    td));
986			/* addr will just be freed in sendit(). */
987			goto out;
988		}
989	}
990#endif
991#ifdef MAC
992	mac_inpcb_create_mbuf(inp, m);
993#endif
994	error = udp6_output(inp, m, addr, control, td);
995out:
996	INP_WUNLOCK(inp);
997	INP_INFO_WUNLOCK(&V_udbinfo);
998	return (error);
999
1000bad:
1001	INP_WUNLOCK(inp);
1002	INP_INFO_WUNLOCK(&V_udbinfo);
1003	m_freem(m);
1004	return (error);
1005}
1006
1007struct pr_usrreqs udp6_usrreqs = {
1008	.pru_abort =		udp6_abort,
1009	.pru_attach =		udp6_attach,
1010	.pru_bind =		udp6_bind,
1011	.pru_connect =		udp6_connect,
1012	.pru_control =		in6_control,
1013	.pru_detach =		udp6_detach,
1014	.pru_disconnect =	udp6_disconnect,
1015	.pru_peeraddr =		in6_mapped_peeraddr,
1016	.pru_send =		udp6_send,
1017	.pru_shutdown =		udp_shutdown,
1018	.pru_sockaddr =		in6_mapped_sockaddr,
1019	.pru_soreceive =	soreceive_dgram,
1020	.pru_sosend =		sosend_dgram,
1021	.pru_sosetlabel =	in_pcbsosetlabel,
1022	.pru_close =		udp6_close
1023};
1024