1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * Copyright (c) 2014 Kevin Lo
5 * All rights reserved.
6 *
7 * Portions of this software were developed by Robert N. M. Watson under
8 * contract to Juniper Networks, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the project nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	$KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
35 *	$KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
36 */
37
38/*-
39 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
40 *	The Regents of the University of California.
41 * All rights reserved.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 *    notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 *    notice, this list of conditions and the following disclaimer in the
50 *    documentation and/or other materials provided with the distribution.
51 * 4. Neither the name of the University nor the names of its contributors
52 *    may be used to endorse or promote products derived from this software
53 *    without specific prior written permission.
54 *
55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * SUCH DAMAGE.
66 *
67 *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
68 */
69
70#include <sys/cdefs.h>
71__FBSDID("$FreeBSD: stable/11/sys/netinet6/udp6_usrreq.c 347155 2019-05-05 12:32:47Z tuexen $");
72
73#include "opt_inet.h"
74#include "opt_inet6.h"
75#include "opt_ipsec.h"
76#include "opt_rss.h"
77
78#include <sys/param.h>
79#include <sys/jail.h>
80#include <sys/kernel.h>
81#include <sys/lock.h>
82#include <sys/mbuf.h>
83#include <sys/priv.h>
84#include <sys/proc.h>
85#include <sys/protosw.h>
86#include <sys/sdt.h>
87#include <sys/signalvar.h>
88#include <sys/socket.h>
89#include <sys/socketvar.h>
90#include <sys/sx.h>
91#include <sys/sysctl.h>
92#include <sys/syslog.h>
93#include <sys/systm.h>
94
95#include <net/if.h>
96#include <net/if_var.h>
97#include <net/if_types.h>
98#include <net/route.h>
99#include <net/rss_config.h>
100
101#include <netinet/in.h>
102#include <netinet/in_kdtrace.h>
103#include <netinet/in_pcb.h>
104#include <netinet/in_systm.h>
105#include <netinet/in_var.h>
106#include <netinet/ip.h>
107#include <netinet/ip6.h>
108#include <netinet/icmp6.h>
109#include <netinet/ip_var.h>
110#include <netinet/udp.h>
111#include <netinet/udp_var.h>
112#include <netinet/udplite.h>
113
114#include <netinet6/ip6protosw.h>
115#include <netinet6/ip6_var.h>
116#include <netinet6/in6_pcb.h>
117#include <netinet6/in6_rss.h>
118#include <netinet6/udp6_var.h>
119#include <netinet6/scope6_var.h>
120
121#include <netipsec/ipsec_support.h>
122
123#include <security/mac/mac_framework.h>
124
125/*
126 * UDP protocol implementation.
127 * Per RFC 768, August, 1980.
128 */
129
130extern struct protosw	inetsw[];
131static void		udp6_detach(struct socket *so);
132
133static int
134udp6_append(struct inpcb *inp, struct mbuf *n, int off,
135    struct sockaddr_in6 *fromsa)
136{
137	struct socket *so;
138	struct mbuf *opts;
139	struct udpcb *up;
140
141	INP_LOCK_ASSERT(inp);
142
143	/*
144	 * Engage the tunneling protocol.
145	 */
146	up = intoudpcb(inp);
147	if (up->u_tun_func != NULL) {
148		in_pcbref(inp);
149		INP_RUNLOCK(inp);
150		(*up->u_tun_func)(n, off, inp, (struct sockaddr *)fromsa,
151		    up->u_tun_ctx);
152		INP_RLOCK(inp);
153		return (in_pcbrele_rlocked(inp));
154	}
155#if defined(IPSEC) || defined(IPSEC_SUPPORT)
156	/* Check AH/ESP integrity. */
157	if (IPSEC_ENABLED(ipv6)) {
158		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
159			m_freem(n);
160			return (0);
161		}
162	}
163#endif /* IPSEC */
164#ifdef MAC
165	if (mac_inpcb_check_deliver(inp, n) != 0) {
166		m_freem(n);
167		return (0);
168	}
169#endif
170	opts = NULL;
171	if (inp->inp_flags & INP_CONTROLOPTS ||
172	    inp->inp_socket->so_options & SO_TIMESTAMP)
173		ip6_savecontrol(inp, n, &opts);
174	m_adj(n, off + sizeof(struct udphdr));
175
176	so = inp->inp_socket;
177	SOCKBUF_LOCK(&so->so_rcv);
178	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
179	    opts) == 0) {
180		SOCKBUF_UNLOCK(&so->so_rcv);
181		m_freem(n);
182		if (opts)
183			m_freem(opts);
184		UDPSTAT_INC(udps_fullsock);
185	} else
186		sorwakeup_locked(so);
187	return (0);
188}
189
190int
191udp6_input(struct mbuf **mp, int *offp, int proto)
192{
193	struct mbuf *m = *mp;
194	struct ifnet *ifp;
195	struct ip6_hdr *ip6;
196	struct udphdr *uh;
197	struct inpcb *inp;
198	struct inpcbinfo *pcbinfo;
199	struct udpcb *up;
200	int off = *offp;
201	int cscov_partial;
202	int plen, ulen;
203	struct sockaddr_in6 fromsa;
204	struct m_tag *fwd_tag;
205	uint16_t uh_sum;
206	uint8_t nxt;
207
208	ifp = m->m_pkthdr.rcvif;
209
210#ifndef PULLDOWN_TEST
211	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
212	ip6 = mtod(m, struct ip6_hdr *);
213	uh = (struct udphdr *)((caddr_t)ip6 + off);
214#else
215	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
216	if (!uh)
217		return (IPPROTO_DONE);
218	ip6 = mtod(m, struct ip6_hdr *);
219#endif
220
221	UDPSTAT_INC(udps_ipackets);
222
223	/*
224	 * Destination port of 0 is illegal, based on RFC768.
225	 */
226	if (uh->uh_dport == 0)
227		goto badunlocked;
228
229	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
230	ulen = ntohs((u_short)uh->uh_ulen);
231
232	nxt = proto;
233	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
234	if (nxt == IPPROTO_UDPLITE) {
235		/* Zero means checksum over the complete packet. */
236		if (ulen == 0)
237			ulen = plen;
238		if (ulen == plen)
239			cscov_partial = 0;
240		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
241			/* XXX: What is the right UDPLite MIB counter? */
242			goto badunlocked;
243		}
244		if (uh->uh_sum == 0) {
245			/* XXX: What is the right UDPLite MIB counter? */
246			goto badunlocked;
247		}
248	} else {
249		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
250			UDPSTAT_INC(udps_badlen);
251			goto badunlocked;
252		}
253		if (uh->uh_sum == 0) {
254			UDPSTAT_INC(udps_nosum);
255			goto badunlocked;
256		}
257	}
258
259	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
260	    !cscov_partial) {
261		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
262			uh_sum = m->m_pkthdr.csum_data;
263		else
264			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
265			    m->m_pkthdr.csum_data);
266		uh_sum ^= 0xffff;
267	} else
268		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
269
270	if (uh_sum != 0) {
271		UDPSTAT_INC(udps_badsum);
272		goto badunlocked;
273	}
274
275	/*
276	 * Construct sockaddr format source address.
277	 */
278	init_sin6(&fromsa, m);
279	fromsa.sin6_port = uh->uh_sport;
280
281	pcbinfo = udp_get_inpcbinfo(nxt);
282	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
283		struct inpcb *last;
284		struct inpcbhead *pcblist;
285		struct ip6_moptions *imo;
286
287		INP_INFO_RLOCK(pcbinfo);
288		/*
289		 * In the event that laddr should be set to the link-local
290		 * address (this happens in RIPng), the multicast address
291		 * specified in the received packet will not match laddr.  To
292		 * handle this situation, matching is relaxed if the
293		 * receiving interface is the same as one specified in the
294		 * socket and if the destination multicast address matches
295		 * one of the multicast groups specified in the socket.
296		 */
297
298		/*
299		 * KAME note: traditionally we dropped udpiphdr from mbuf
300		 * here.  We need udphdr for IPsec processing so we do that
301		 * later.
302		 */
303		pcblist = udp_get_pcblist(nxt);
304		last = NULL;
305		LIST_FOREACH(inp, pcblist, inp_list) {
306			if ((inp->inp_vflag & INP_IPV6) == 0)
307				continue;
308			if (inp->inp_lport != uh->uh_dport)
309				continue;
310			if (inp->inp_fport != 0 &&
311			    inp->inp_fport != uh->uh_sport)
312				continue;
313			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
314				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
315							&ip6->ip6_dst))
316					continue;
317			}
318			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
319				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
320							&ip6->ip6_src) ||
321				    inp->inp_fport != uh->uh_sport)
322					continue;
323			}
324
325			/*
326			 * XXXRW: Because we weren't holding either the inpcb
327			 * or the hash lock when we checked for a match
328			 * before, we should probably recheck now that the
329			 * inpcb lock is (supposed to be) held.
330			 */
331
332			/*
333			 * Handle socket delivery policy for any-source
334			 * and source-specific multicast. [RFC3678]
335			 */
336			imo = inp->in6p_moptions;
337			if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
338				struct sockaddr_in6	 mcaddr;
339				int			 blocked;
340
341				INP_RLOCK(inp);
342
343				bzero(&mcaddr, sizeof(struct sockaddr_in6));
344				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
345				mcaddr.sin6_family = AF_INET6;
346				mcaddr.sin6_addr = ip6->ip6_dst;
347
348				blocked = im6o_mc_filter(imo, ifp,
349					(struct sockaddr *)&mcaddr,
350					(struct sockaddr *)&fromsa);
351				if (blocked != MCAST_PASS) {
352					if (blocked == MCAST_NOTGMEMBER)
353						IP6STAT_INC(ip6s_notmember);
354					if (blocked == MCAST_NOTSMEMBER ||
355					    blocked == MCAST_MUTED)
356						UDPSTAT_INC(udps_filtermcast);
357					INP_RUNLOCK(inp); /* XXX */
358					continue;
359				}
360
361				INP_RUNLOCK(inp);
362			}
363			if (last != NULL) {
364				struct mbuf *n;
365
366				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
367					INP_RLOCK(last);
368					UDP_PROBE(receive, NULL, last, ip6,
369					    last, uh);
370					if (udp6_append(last, n, off, &fromsa))
371						goto inp_lost;
372					INP_RUNLOCK(last);
373				}
374			}
375			last = inp;
376			/*
377			 * Don't look for additional matches if this one does
378			 * not have either the SO_REUSEPORT or SO_REUSEADDR
379			 * socket options set.  This heuristic avoids
380			 * searching through all pcbs in the common case of a
381			 * non-shared port.  It assumes that an application
382			 * will never clear these options after setting them.
383			 */
384			if ((last->inp_socket->so_options &
385			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
386				break;
387		}
388
389		if (last == NULL) {
390			/*
391			 * No matching pcb found; discard datagram.  (No need
392			 * to send an ICMP Port Unreachable for a broadcast
393			 * or multicast datgram.)
394			 */
395			UDPSTAT_INC(udps_noport);
396			UDPSTAT_INC(udps_noportmcast);
397			goto badheadlocked;
398		}
399		INP_RLOCK(last);
400		INP_INFO_RUNLOCK(pcbinfo);
401		UDP_PROBE(receive, NULL, last, ip6, last, uh);
402		if (udp6_append(last, m, off, &fromsa) == 0)
403			INP_RUNLOCK(last);
404	inp_lost:
405		return (IPPROTO_DONE);
406	}
407	/*
408	 * Locate pcb for datagram.
409	 */
410
411	/*
412	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
413	 */
414	if ((m->m_flags & M_IP6_NEXTHOP) &&
415	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
416		struct sockaddr_in6 *next_hop6;
417
418		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
419
420		/*
421		 * Transparently forwarded. Pretend to be the destination.
422		 * Already got one like this?
423		 */
424		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
425		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
426		    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
427		if (!inp) {
428			/*
429			 * It's new.  Try to find the ambushing socket.
430			 * Because we've rewritten the destination address,
431			 * any hardware-generated hash is ignored.
432			 */
433			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
434			    uh->uh_sport, &next_hop6->sin6_addr,
435			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
436			    uh->uh_dport, INPLOOKUP_WILDCARD |
437			    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
438		}
439		/* Remove the tag from the packet. We don't need it anymore. */
440		m_tag_delete(m, fwd_tag);
441		m->m_flags &= ~M_IP6_NEXTHOP;
442	} else
443		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
444		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
445		    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
446		    m->m_pkthdr.rcvif, m);
447	if (inp == NULL) {
448		if (udp_log_in_vain) {
449			char ip6bufs[INET6_ADDRSTRLEN];
450			char ip6bufd[INET6_ADDRSTRLEN];
451
452			log(LOG_INFO,
453			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
454			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
455			    ntohs(uh->uh_dport),
456			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
457			    ntohs(uh->uh_sport));
458		}
459		UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
460		UDPSTAT_INC(udps_noport);
461		if (m->m_flags & M_MCAST) {
462			printf("UDP6: M_MCAST is set in a unicast packet.\n");
463			UDPSTAT_INC(udps_noportmcast);
464			goto badunlocked;
465		}
466		if (V_udp_blackhole)
467			goto badunlocked;
468		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
469		return (IPPROTO_DONE);
470	}
471	INP_RLOCK_ASSERT(inp);
472	up = intoudpcb(inp);
473	if (cscov_partial) {
474		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
475			INP_RUNLOCK(inp);
476			m_freem(m);
477			return (IPPROTO_DONE);
478		}
479	}
480	UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
481	if (udp6_append(inp, m, off, &fromsa) == 0)
482		INP_RUNLOCK(inp);
483	return (IPPROTO_DONE);
484
485badheadlocked:
486	INP_INFO_RUNLOCK(pcbinfo);
487badunlocked:
488	if (m)
489		m_freem(m);
490	return (IPPROTO_DONE);
491}
492
493static void
494udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
495    struct inpcbinfo *pcbinfo)
496{
497	struct udphdr uh;
498	struct ip6_hdr *ip6;
499	struct mbuf *m;
500	int off = 0;
501	struct ip6ctlparam *ip6cp = NULL;
502	const struct sockaddr_in6 *sa6_src = NULL;
503	void *cmdarg;
504	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
505	struct udp_portonly {
506		u_int16_t uh_sport;
507		u_int16_t uh_dport;
508	} *uhp;
509
510	if (sa->sa_family != AF_INET6 ||
511	    sa->sa_len != sizeof(struct sockaddr_in6))
512		return;
513
514	if ((unsigned)cmd >= PRC_NCMDS)
515		return;
516	if (PRC_IS_REDIRECT(cmd))
517		notify = in6_rtchange, d = NULL;
518	else if (cmd == PRC_HOSTDEAD)
519		d = NULL;
520	else if (inet6ctlerrmap[cmd] == 0)
521		return;
522
523	/* if the parameter is from icmp6, decode it. */
524	if (d != NULL) {
525		ip6cp = (struct ip6ctlparam *)d;
526		m = ip6cp->ip6c_m;
527		ip6 = ip6cp->ip6c_ip6;
528		off = ip6cp->ip6c_off;
529		cmdarg = ip6cp->ip6c_cmdarg;
530		sa6_src = ip6cp->ip6c_src;
531	} else {
532		m = NULL;
533		ip6 = NULL;
534		cmdarg = NULL;
535		sa6_src = &sa6_any;
536	}
537
538	if (ip6) {
539		/*
540		 * XXX: We assume that when IPV6 is non NULL,
541		 * M and OFF are valid.
542		 */
543
544		/* Check if we can safely examine src and dst ports. */
545		if (m->m_pkthdr.len < off + sizeof(*uhp))
546			return;
547
548		bzero(&uh, sizeof(uh));
549		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
550
551		if (!PRC_IS_REDIRECT(cmd)) {
552			/* Check to see if its tunneled */
553			struct inpcb *inp;
554			inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
555			    uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
556			    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
557			    m->m_pkthdr.rcvif, m);
558			if (inp != NULL) {
559				struct udpcb *up;
560
561				up = intoudpcb(inp);
562				if (up->u_icmp_func) {
563					/* Yes it is. */
564					INP_RUNLOCK(inp);
565					(*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
566					      d, up->u_tun_ctx);
567					return;
568				} else {
569					/* Can't find it. */
570					INP_RUNLOCK(inp);
571				}
572			}
573		}
574		(void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
575		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
576		    cmdarg, notify);
577	} else
578		(void)in6_pcbnotify(pcbinfo, sa, 0,
579		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
580}
581
582void
583udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
584{
585
586	return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
587}
588
589void
590udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
591{
592
593	return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
594}
595
596static int
597udp6_getcred(SYSCTL_HANDLER_ARGS)
598{
599	struct xucred xuc;
600	struct sockaddr_in6 addrs[2];
601	struct inpcb *inp;
602	int error;
603
604	error = priv_check(req->td, PRIV_NETINET_GETCRED);
605	if (error)
606		return (error);
607
608	if (req->newlen != sizeof(addrs))
609		return (EINVAL);
610	if (req->oldlen != sizeof(struct xucred))
611		return (EINVAL);
612	error = SYSCTL_IN(req, addrs, sizeof(addrs));
613	if (error)
614		return (error);
615	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
616	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
617		return (error);
618	}
619	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
620	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
621	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
622	if (inp != NULL) {
623		INP_RLOCK_ASSERT(inp);
624		if (inp->inp_socket == NULL)
625			error = ENOENT;
626		if (error == 0)
627			error = cr_canseesocket(req->td->td_ucred,
628			    inp->inp_socket);
629		if (error == 0)
630			cru2x(inp->inp_cred, &xuc);
631		INP_RUNLOCK(inp);
632	} else
633		error = ENOENT;
634	if (error == 0)
635		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
636	return (error);
637}
638
639SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
640    0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
641
642static int
643udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
644    struct mbuf *control, struct thread *td)
645{
646	u_int32_t ulen = m->m_pkthdr.len;
647	u_int32_t plen = sizeof(struct udphdr) + ulen;
648	struct ip6_hdr *ip6;
649	struct udphdr *udp6;
650	struct in6_addr *laddr, *faddr, in6a;
651	struct sockaddr_in6 *sin6 = NULL;
652	int cscov_partial = 0;
653	int scope_ambiguous = 0;
654	u_short fport;
655	int error = 0;
656	uint8_t nxt;
657	uint16_t cscov = 0;
658	struct ip6_pktopts *optp, opt;
659	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
660	int flags;
661	struct sockaddr_in6 tmp;
662
663	INP_WLOCK_ASSERT(inp);
664	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
665
666	if (addr6) {
667		/* addr6 has been validated in udp6_send(). */
668		sin6 = (struct sockaddr_in6 *)addr6;
669
670		/* protect *sin6 from overwrites */
671		tmp = *sin6;
672		sin6 = &tmp;
673
674		/*
675		 * Application should provide a proper zone ID or the use of
676		 * default zone IDs should be enabled.  Unfortunately, some
677		 * applications do not behave as it should, so we need a
678		 * workaround.  Even if an appropriate ID is not determined,
679		 * we'll see if we can determine the outgoing interface.  If we
680		 * can, determine the zone ID based on the interface below.
681		 */
682		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
683			scope_ambiguous = 1;
684		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
685			return (error);
686	}
687
688	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
689	    IPPROTO_UDP : IPPROTO_UDPLITE;
690	if (control) {
691		if ((error = ip6_setpktopts(control, &opt,
692		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0)
693			goto release;
694		optp = &opt;
695	} else
696		optp = inp->in6p_outputopts;
697
698	if (sin6) {
699		faddr = &sin6->sin6_addr;
700
701		/*
702		 * Since we saw no essential reason for calling in_pcbconnect,
703		 * we get rid of such kind of logic, and call in6_selectsrc
704		 * and in6_pcbsetport in order to fill in the local address
705		 * and the local port.
706		 */
707		if (sin6->sin6_port == 0) {
708			error = EADDRNOTAVAIL;
709			goto release;
710		}
711
712		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
713			/* how about ::ffff:0.0.0.0 case? */
714			error = EISCONN;
715			goto release;
716		}
717
718		fport = sin6->sin6_port; /* allow 0 port */
719
720		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
721			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
722				/*
723				 * I believe we should explicitly discard the
724				 * packet when mapped addresses are disabled,
725				 * rather than send the packet as an IPv6 one.
726				 * If we chose the latter approach, the packet
727				 * might be sent out on the wire based on the
728				 * default route, the situation which we'd
729				 * probably want to avoid.
730				 * (20010421 jinmei@kame.net)
731				 */
732				error = EINVAL;
733				goto release;
734			}
735			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
736			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
737				/*
738				 * when remote addr is an IPv4-mapped address,
739				 * local addr should not be an IPv6 address,
740				 * since you cannot determine how to map IPv6
741				 * source address to IPv4.
742				 */
743				error = EINVAL;
744				goto release;
745			}
746
747			af = AF_INET;
748		}
749
750		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
751			error = in6_selectsrc_socket(sin6, optp, inp,
752			    td->td_ucred, scope_ambiguous, &in6a, NULL);
753			if (error)
754				goto release;
755			laddr = &in6a;
756		} else
757			laddr = &inp->in6p_laddr;	/* XXX */
758		if (laddr == NULL) {
759			if (error == 0)
760				error = EADDRNOTAVAIL;
761			goto release;
762		}
763		if (inp->inp_lport == 0 &&
764		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
765			/* Undo an address bind that may have occurred. */
766			inp->in6p_laddr = in6addr_any;
767			goto release;
768		}
769	} else {
770		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
771			error = ENOTCONN;
772			goto release;
773		}
774		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
775			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
776				/*
777				 * XXX: this case would happen when the
778				 * application sets the V6ONLY flag after
779				 * connecting the foreign address.
780				 * Such applications should be fixed,
781				 * so we bark here.
782				 */
783				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
784				    "option was set for a connected socket\n");
785				error = EINVAL;
786				goto release;
787			} else
788				af = AF_INET;
789		}
790		laddr = &inp->in6p_laddr;
791		faddr = &inp->in6p_faddr;
792		fport = inp->inp_fport;
793	}
794
795	if (af == AF_INET)
796		hlen = sizeof(struct ip);
797
798	/*
799	 * Calculate data length and get a mbuf
800	 * for UDP and IP6 headers.
801	 */
802	M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
803	if (m == NULL) {
804		error = ENOBUFS;
805		goto release;
806	}
807
808	/*
809	 * Stuff checksum and output datagram.
810	 */
811	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
812	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
813	udp6->uh_dport = fport;
814	if (nxt == IPPROTO_UDPLITE) {
815		struct udpcb *up;
816
817		up = intoudpcb(inp);
818		cscov = up->u_txcslen;
819		if (cscov >= plen)
820			cscov = 0;
821		udp6->uh_ulen = htons(cscov);
822		/*
823		 * For UDP-Lite, checksum coverage length of zero means
824		 * the entire UDPLite packet is covered by the checksum.
825		 */
826		cscov_partial = (cscov == 0) ? 0 : 1;
827	} else if (plen <= 0xffff)
828		udp6->uh_ulen = htons((u_short)plen);
829	else
830		udp6->uh_ulen = 0;
831	udp6->uh_sum = 0;
832
833	switch (af) {
834	case AF_INET6:
835		ip6 = mtod(m, struct ip6_hdr *);
836		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
837		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
838		ip6->ip6_vfc	|= IPV6_VERSION;
839		ip6->ip6_plen	= htons((u_short)plen);
840		ip6->ip6_nxt	= nxt;
841		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
842		ip6->ip6_src	= *laddr;
843		ip6->ip6_dst	= *faddr;
844
845		if (cscov_partial) {
846			if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
847			    sizeof(struct ip6_hdr), plen, cscov)) == 0)
848				udp6->uh_sum = 0xffff;
849		} else {
850			udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
851			m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
852			m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
853		}
854
855#ifdef	RSS
856		{
857			uint32_t hash_val, hash_type;
858			uint8_t pr;
859
860			pr = inp->inp_socket->so_proto->pr_protocol;
861			/*
862			 * Calculate an appropriate RSS hash for UDP and
863			 * UDP Lite.
864			 *
865			 * The called function will take care of figuring out
866			 * whether a 2-tuple or 4-tuple hash is required based
867			 * on the currently configured scheme.
868			 *
869			 * Later later on connected socket values should be
870			 * cached in the inpcb and reused, rather than constantly
871			 * re-calculating it.
872			 *
873			 * UDP Lite is a different protocol number and will
874			 * likely end up being hashed as a 2-tuple until
875			 * RSS / NICs grow UDP Lite protocol awareness.
876			 */
877			if (rss_proto_software_hash_v6(faddr, laddr, fport,
878			    inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
879				m->m_pkthdr.flowid = hash_val;
880				M_HASHTYPE_SET(m, hash_type);
881			}
882		}
883#endif
884		flags = 0;
885#ifdef	RSS
886		/*
887		 * Don't override with the inp cached flowid.
888		 *
889		 * Until the whole UDP path is vetted, it may actually
890		 * be incorrect.
891		 */
892		flags |= IP_NODEFAULTFLOWID;
893#endif
894
895		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
896		UDPSTAT_INC(udps_opackets);
897		error = ip6_output(m, optp, &inp->inp_route6, flags,
898		    inp->in6p_moptions, NULL, inp);
899		break;
900	case AF_INET:
901		error = EAFNOSUPPORT;
902		goto release;
903	}
904	goto releaseopt;
905
906release:
907	m_freem(m);
908
909releaseopt:
910	if (control) {
911		ip6_clearpktopts(&opt, -1);
912		m_freem(control);
913	}
914	return (error);
915}
916
917static void
918udp6_abort(struct socket *so)
919{
920	struct inpcb *inp;
921	struct inpcbinfo *pcbinfo;
922
923	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
924	inp = sotoinpcb(so);
925	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
926
927	INP_WLOCK(inp);
928#ifdef INET
929	if (inp->inp_vflag & INP_IPV4) {
930		struct pr_usrreqs *pru;
931		uint8_t nxt;
932
933		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
934		    IPPROTO_UDP : IPPROTO_UDPLITE;
935		INP_WUNLOCK(inp);
936		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
937		(*pru->pru_abort)(so);
938		return;
939	}
940#endif
941
942	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
943		INP_HASH_WLOCK(pcbinfo);
944		in6_pcbdisconnect(inp);
945		inp->in6p_laddr = in6addr_any;
946		INP_HASH_WUNLOCK(pcbinfo);
947		soisdisconnected(so);
948	}
949	INP_WUNLOCK(inp);
950}
951
952static int
953udp6_attach(struct socket *so, int proto, struct thread *td)
954{
955	struct inpcb *inp;
956	struct inpcbinfo *pcbinfo;
957	int error;
958
959	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
960	inp = sotoinpcb(so);
961	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
962
963	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
964		error = soreserve(so, udp_sendspace, udp_recvspace);
965		if (error)
966			return (error);
967	}
968	INP_INFO_WLOCK(pcbinfo);
969	error = in_pcballoc(so, pcbinfo);
970	if (error) {
971		INP_INFO_WUNLOCK(pcbinfo);
972		return (error);
973	}
974	inp = (struct inpcb *)so->so_pcb;
975	inp->inp_vflag |= INP_IPV6;
976	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
977		inp->inp_vflag |= INP_IPV4;
978	inp->in6p_hops = -1;	/* use kernel default */
979	inp->in6p_cksum = -1;	/* just to be sure */
980	/*
981	 * XXX: ugly!!
982	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
983	 * because the socket may be bound to an IPv6 wildcard address,
984	 * which may match an IPv4-mapped IPv6 address.
985	 */
986	inp->inp_ip_ttl = V_ip_defttl;
987
988	error = udp_newudpcb(inp);
989	if (error) {
990		in_pcbdetach(inp);
991		in_pcbfree(inp);
992		INP_INFO_WUNLOCK(pcbinfo);
993		return (error);
994	}
995	INP_WUNLOCK(inp);
996	INP_INFO_WUNLOCK(pcbinfo);
997	return (0);
998}
999
1000static int
1001udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1002{
1003	struct inpcb *inp;
1004	struct inpcbinfo *pcbinfo;
1005	int error;
1006	u_char vflagsav;
1007
1008	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1009	inp = sotoinpcb(so);
1010	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1011
1012	INP_WLOCK(inp);
1013	INP_HASH_WLOCK(pcbinfo);
1014	vflagsav = inp->inp_vflag;
1015	inp->inp_vflag &= ~INP_IPV4;
1016	inp->inp_vflag |= INP_IPV6;
1017	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1018		struct sockaddr_in6 *sin6_p;
1019
1020		sin6_p = (struct sockaddr_in6 *)nam;
1021
1022		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1023			inp->inp_vflag |= INP_IPV4;
1024#ifdef INET
1025		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1026			struct sockaddr_in sin;
1027
1028			in6_sin6_2_sin(&sin, sin6_p);
1029			inp->inp_vflag |= INP_IPV4;
1030			inp->inp_vflag &= ~INP_IPV6;
1031			error = in_pcbbind(inp, (struct sockaddr *)&sin,
1032			    td->td_ucred);
1033			goto out;
1034		}
1035#endif
1036	}
1037
1038	error = in6_pcbbind(inp, nam, td->td_ucred);
1039#ifdef INET
1040out:
1041#endif
1042	if (error != 0)
1043		inp->inp_vflag = vflagsav;
1044	INP_HASH_WUNLOCK(pcbinfo);
1045	INP_WUNLOCK(inp);
1046	return (error);
1047}
1048
1049static void
1050udp6_close(struct socket *so)
1051{
1052	struct inpcb *inp;
1053	struct inpcbinfo *pcbinfo;
1054
1055	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1056	inp = sotoinpcb(so);
1057	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1058
1059	INP_WLOCK(inp);
1060#ifdef INET
1061	if (inp->inp_vflag & INP_IPV4) {
1062		struct pr_usrreqs *pru;
1063		uint8_t nxt;
1064
1065		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1066		    IPPROTO_UDP : IPPROTO_UDPLITE;
1067		INP_WUNLOCK(inp);
1068		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1069		(*pru->pru_disconnect)(so);
1070		return;
1071	}
1072#endif
1073	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1074		INP_HASH_WLOCK(pcbinfo);
1075		in6_pcbdisconnect(inp);
1076		inp->in6p_laddr = in6addr_any;
1077		INP_HASH_WUNLOCK(pcbinfo);
1078		soisdisconnected(so);
1079	}
1080	INP_WUNLOCK(inp);
1081}
1082
1083static int
1084udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1085{
1086	struct inpcb *inp;
1087	struct inpcbinfo *pcbinfo;
1088	struct sockaddr_in6 *sin6;
1089	int error;
1090	u_char vflagsav;
1091
1092	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1093	inp = sotoinpcb(so);
1094	sin6 = (struct sockaddr_in6 *)nam;
1095	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1096
1097	/*
1098	 * XXXRW: Need to clarify locking of v4/v6 flags.
1099	 */
1100	INP_WLOCK(inp);
1101#ifdef INET
1102	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1103		struct sockaddr_in sin;
1104
1105		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1106			error = EINVAL;
1107			goto out;
1108		}
1109		if ((inp->inp_vflag & INP_IPV4) == 0) {
1110			error = EAFNOSUPPORT;
1111			goto out;
1112		}
1113		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1114			error = EISCONN;
1115			goto out;
1116		}
1117		in6_sin6_2_sin(&sin, sin6);
1118		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1119		if (error != 0)
1120			goto out;
1121		vflagsav = inp->inp_vflag;
1122		inp->inp_vflag |= INP_IPV4;
1123		inp->inp_vflag &= ~INP_IPV6;
1124		INP_HASH_WLOCK(pcbinfo);
1125		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1126		    td->td_ucred);
1127		INP_HASH_WUNLOCK(pcbinfo);
1128		/*
1129		 * If connect succeeds, mark socket as connected. If
1130		 * connect fails and socket is unbound, reset inp_vflag
1131		 * field.
1132		 */
1133		if (error == 0)
1134			soisconnected(so);
1135		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1136		    inp->inp_lport == 0)
1137			inp->inp_vflag = vflagsav;
1138		goto out;
1139	} else {
1140		if ((inp->inp_vflag & INP_IPV6) == 0) {
1141			error = EAFNOSUPPORT;
1142			goto out;
1143		}
1144	}
1145#endif
1146	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1147		error = EISCONN;
1148		goto out;
1149	}
1150	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1151	if (error != 0)
1152		goto out;
1153	vflagsav = inp->inp_vflag;
1154	inp->inp_vflag &= ~INP_IPV4;
1155	inp->inp_vflag |= INP_IPV6;
1156	INP_HASH_WLOCK(pcbinfo);
1157	error = in6_pcbconnect(inp, nam, td->td_ucred);
1158	INP_HASH_WUNLOCK(pcbinfo);
1159	/*
1160	 * If connect succeeds, mark socket as connected. If
1161	 * connect fails and socket is unbound, reset inp_vflag
1162	 * field.
1163	 */
1164	if (error == 0)
1165		soisconnected(so);
1166	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1167	    inp->inp_lport == 0)
1168		inp->inp_vflag = vflagsav;
1169out:
1170	INP_WUNLOCK(inp);
1171	return (error);
1172}
1173
1174static void
1175udp6_detach(struct socket *so)
1176{
1177	struct inpcb *inp;
1178	struct inpcbinfo *pcbinfo;
1179	struct udpcb *up;
1180
1181	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1182	inp = sotoinpcb(so);
1183	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1184
1185	INP_INFO_WLOCK(pcbinfo);
1186	INP_WLOCK(inp);
1187	up = intoudpcb(inp);
1188	KASSERT(up != NULL, ("%s: up == NULL", __func__));
1189	in_pcbdetach(inp);
1190	in_pcbfree(inp);
1191	INP_INFO_WUNLOCK(pcbinfo);
1192	udp_discardcb(up);
1193}
1194
1195static int
1196udp6_disconnect(struct socket *so)
1197{
1198	struct inpcb *inp;
1199	struct inpcbinfo *pcbinfo;
1200	int error;
1201
1202	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1203	inp = sotoinpcb(so);
1204	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1205
1206	INP_WLOCK(inp);
1207#ifdef INET
1208	if (inp->inp_vflag & INP_IPV4) {
1209		struct pr_usrreqs *pru;
1210		uint8_t nxt;
1211
1212		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1213		    IPPROTO_UDP : IPPROTO_UDPLITE;
1214		INP_WUNLOCK(inp);
1215		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1216		(void)(*pru->pru_disconnect)(so);
1217		return (0);
1218	}
1219#endif
1220
1221	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1222		error = ENOTCONN;
1223		goto out;
1224	}
1225
1226	INP_HASH_WLOCK(pcbinfo);
1227	in6_pcbdisconnect(inp);
1228	inp->in6p_laddr = in6addr_any;
1229	INP_HASH_WUNLOCK(pcbinfo);
1230	SOCK_LOCK(so);
1231	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1232	SOCK_UNLOCK(so);
1233out:
1234	INP_WUNLOCK(inp);
1235	return (0);
1236}
1237
1238static int
1239udp6_send(struct socket *so, int flags, struct mbuf *m,
1240    struct sockaddr *addr, struct mbuf *control, struct thread *td)
1241{
1242	struct inpcb *inp;
1243	struct inpcbinfo *pcbinfo;
1244	int error = 0;
1245
1246	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1247	inp = sotoinpcb(so);
1248	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1249
1250	INP_WLOCK(inp);
1251	if (addr) {
1252		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1253			error = EINVAL;
1254			goto bad;
1255		}
1256		if (addr->sa_family != AF_INET6) {
1257			error = EAFNOSUPPORT;
1258			goto bad;
1259		}
1260	}
1261
1262#ifdef INET
1263	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1264		int hasv4addr;
1265		struct sockaddr_in6 *sin6 = NULL;
1266
1267		if (addr == NULL)
1268			hasv4addr = (inp->inp_vflag & INP_IPV4);
1269		else {
1270			sin6 = (struct sockaddr_in6 *)addr;
1271			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1272			    ? 1 : 0;
1273		}
1274		if (hasv4addr) {
1275			struct pr_usrreqs *pru;
1276			uint8_t nxt;
1277
1278			nxt = (inp->inp_socket->so_proto->pr_protocol ==
1279			    IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE;
1280			/*
1281			 * XXXRW: We release UDP-layer locks before calling
1282			 * udp_send() in order to avoid recursion.  However,
1283			 * this does mean there is a short window where inp's
1284			 * fields are unstable.  Could this lead to a
1285			 * potential race in which the factors causing us to
1286			 * select the UDPv4 output routine are invalidated?
1287			 */
1288			INP_WUNLOCK(inp);
1289			if (sin6)
1290				in6_sin6_2_sin_in_sock(addr);
1291			pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1292			/* addr will just be freed in sendit(). */
1293			return ((*pru->pru_send)(so, flags, m, addr, control,
1294			    td));
1295		}
1296	}
1297#endif
1298#ifdef MAC
1299	mac_inpcb_create_mbuf(inp, m);
1300#endif
1301	INP_HASH_WLOCK(pcbinfo);
1302	error = udp6_output(inp, m, addr, control, td);
1303	INP_HASH_WUNLOCK(pcbinfo);
1304	INP_WUNLOCK(inp);
1305	return (error);
1306
1307bad:
1308	INP_WUNLOCK(inp);
1309	m_freem(m);
1310	return (error);
1311}
1312
1313struct pr_usrreqs udp6_usrreqs = {
1314	.pru_abort =		udp6_abort,
1315	.pru_attach =		udp6_attach,
1316	.pru_bind =		udp6_bind,
1317	.pru_connect =		udp6_connect,
1318	.pru_control =		in6_control,
1319	.pru_detach =		udp6_detach,
1320	.pru_disconnect =	udp6_disconnect,
1321	.pru_peeraddr =		in6_mapped_peeraddr,
1322	.pru_send =		udp6_send,
1323	.pru_shutdown =		udp_shutdown,
1324	.pru_sockaddr =		in6_mapped_sockaddr,
1325	.pru_soreceive =	soreceive_dgram,
1326	.pru_sosend =		sosend_dgram,
1327	.pru_sosetlabel =	in_pcbsosetlabel,
1328	.pru_close =		udp6_close
1329};
1330