xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 328110da2661a8841f12000b99fea27ceacdd5b2)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * Copyright (c) 2010-2011 Juniper Networks, Inc.
6  * Copyright (c) 2014 Kevin Lo
7  * All rights reserved.
8  *
9  * Portions of this software were developed by Robert N. M. Watson under
10  * contract to Juniper Networks, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the project nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	$KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
37  *	$KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
38  */
39 
40 /*-
41  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
42  *	The Regents of the University of California.
43  * All rights reserved.
44  *
45  * Redistribution and use in source and binary forms, with or without
46  * modification, are permitted provided that the following conditions
47  * are met:
48  * 1. Redistributions of source code must retain the above copyright
49  *    notice, this list of conditions and the following disclaimer.
50  * 2. Redistributions in binary form must reproduce the above copyright
51  *    notice, this list of conditions and the following disclaimer in the
52  *    documentation and/or other materials provided with the distribution.
53  * 3. Neither the name of the University nor the names of its contributors
54  *    may be used to endorse or promote products derived from this software
55  *    without specific prior written permission.
56  *
57  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67  * SUCH DAMAGE.
68  */
69 
70 #include <sys/cdefs.h>
71 #include "opt_inet.h"
72 #include "opt_inet6.h"
73 #include "opt_ipsec.h"
74 #include "opt_route.h"
75 #include "opt_rss.h"
76 
77 #include <sys/param.h>
78 #include <sys/domain.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/ip6_var.h>
115 #include <netinet6/in6_fib.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 VNET_DEFINE(int, zero_checksum_port) = 0;
126 #define	V_zero_checksum_port	VNET(zero_checksum_port)
127 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
128     &VNET_NAME(zero_checksum_port), 0,
129     "Zero UDP checksum allowed for traffic to/from this port.");
130 
131 /*
132  * UDP protocol implementation.
133  * Per RFC 768, August, 1980.
134  */
135 
136 static void		udp6_detach(struct socket *so);
137 
138 static int
139 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
140     struct sockaddr_in6 *fromsa)
141 {
142 	struct socket *so;
143 	struct mbuf *opts = NULL, *tmp_opts;
144 	struct udpcb *up;
145 
146 	INP_LOCK_ASSERT(inp);
147 
148 	/*
149 	 * Engage the tunneling protocol.
150 	 */
151 	up = intoudpcb(inp);
152 	if (up->u_tun_func != NULL) {
153 		bool filtered;
154 
155 		in_pcbref(inp);
156 		INP_RUNLOCK(inp);
157 		filtered = (*up->u_tun_func)(n, off, inp,
158 		    (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
159 		INP_RLOCK(inp);
160 		if (in_pcbrele_rlocked(inp))
161 			return (1);
162 		if (filtered) {
163 			INP_RUNLOCK(inp);
164 			return (1);
165 		}
166 	}
167 
168 	off += sizeof(struct udphdr);
169 
170 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
171 	/* Check AH/ESP integrity. */
172 	if (IPSEC_ENABLED(ipv6)) {
173 		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
174 			m_freem(n);
175 			return (0);
176 		}
177 
178 		/* IPSec UDP encaps. */
179 		if ((up->u_flags & UF_ESPINUDP) != 0 &&
180 		    UDPENCAP_INPUT(ipv6, n, off, AF_INET6) != 0) {
181 			return (0); /* Consumed. */
182 		}
183 	}
184 #endif /* IPSEC */
185 #ifdef MAC
186 	if (mac_inpcb_check_deliver(inp, n) != 0) {
187 		m_freem(n);
188 		return (0);
189 	}
190 #endif
191 	opts = NULL;
192 	if (inp->inp_flags & INP_CONTROLOPTS ||
193 	    inp->inp_socket->so_options & SO_TIMESTAMP)
194 		ip6_savecontrol(inp, n, &opts);
195 	if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
196 		tmp_opts = sbcreatecontrol(&fromsa[1],
197 		    sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
198 		    IPPROTO_IPV6, M_NOWAIT);
199                 if (tmp_opts) {
200                         if (opts) {
201                                 tmp_opts->m_next = opts;
202                                 opts = tmp_opts;
203                         } else
204                                 opts = tmp_opts;
205                 }
206 	}
207 	m_adj(n, off);
208 
209 	so = inp->inp_socket;
210 	SOCKBUF_LOCK(&so->so_rcv);
211 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
212 	    opts) == 0) {
213 		soroverflow_locked(so);
214 		m_freem(n);
215 		if (opts)
216 			m_freem(opts);
217 		UDPSTAT_INC(udps_fullsock);
218 	} else
219 		sorwakeup_locked(so);
220 	return (0);
221 }
222 
223 struct udp6_multi_match_ctx {
224 	struct ip6_hdr *ip6;
225 	struct udphdr *uh;
226 };
227 
228 static bool
229 udp6_multi_match(const struct inpcb *inp, void *v)
230 {
231 	struct udp6_multi_match_ctx *ctx = v;
232 
233 	if ((inp->inp_vflag & INP_IPV6) == 0)
234 		return(false);
235 	if (inp->inp_lport != ctx->uh->uh_dport)
236 		return(false);
237 	if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
238 		return(false);
239 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
240 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
241 		return (false);
242 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
243 	    (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
244 	    inp->inp_fport != ctx->uh->uh_sport))
245 		return (false);
246 
247 	return (true);
248 }
249 
250 static int
251 udp6_multi_input(struct mbuf *m, int off, int proto,
252     struct sockaddr_in6 *fromsa)
253 {
254 	struct udp6_multi_match_ctx ctx;
255 	struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
256 	    INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
257 	struct inpcb *inp;
258 	struct ip6_moptions *imo;
259 	struct mbuf *n;
260 	int appends = 0;
261 
262 	/*
263 	 * In the event that laddr should be set to the link-local
264 	 * address (this happens in RIPng), the multicast address
265 	 * specified in the received packet will not match laddr.  To
266 	 * handle this situation, matching is relaxed if the
267 	 * receiving interface is the same as one specified in the
268 	 * socket and if the destination multicast address matches
269 	 * one of the multicast groups specified in the socket.
270 	 */
271 
272 	/*
273 	 * KAME note: traditionally we dropped udpiphdr from mbuf
274 	 * here.  We need udphdr for IPsec processing so we do that
275 	 * later.
276 	 */
277 	ctx.ip6 = mtod(m, struct ip6_hdr *);
278 	ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
279 	while ((inp = inp_next(&inpi)) != NULL) {
280 		INP_RLOCK_ASSERT(inp);
281 		/*
282 		 * XXXRW: Because we weren't holding either the inpcb
283 		 * or the hash lock when we checked for a match
284 		 * before, we should probably recheck now that the
285 		 * inpcb lock is (supposed to be) held.
286 		 */
287 		/*
288 		 * Handle socket delivery policy for any-source
289 		 * and source-specific multicast. [RFC3678]
290 		 */
291 		if ((imo = inp->in6p_moptions) != NULL) {
292 			struct sockaddr_in6	 mcaddr;
293 			int			 blocked;
294 
295 			bzero(&mcaddr, sizeof(struct sockaddr_in6));
296 			mcaddr.sin6_len = sizeof(struct sockaddr_in6);
297 			mcaddr.sin6_family = AF_INET6;
298 			mcaddr.sin6_addr = ctx.ip6->ip6_dst;
299 
300 			blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
301 				(struct sockaddr *)&mcaddr,
302 				(struct sockaddr *)&fromsa[0]);
303 			if (blocked != MCAST_PASS) {
304 				if (blocked == MCAST_NOTGMEMBER)
305 					IP6STAT_INC(ip6s_notmember);
306 				if (blocked == MCAST_NOTSMEMBER ||
307 				    blocked == MCAST_MUTED)
308 					UDPSTAT_INC(udps_filtermcast);
309 				continue;
310 			}
311 		}
312 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
313 			if (proto == IPPROTO_UDPLITE)
314 				UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
315 				    inp, ctx.uh);
316 			else
317 				UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
318 				    ctx.uh);
319 			if (udp6_append(inp, n, off, fromsa)) {
320 				break;
321 			} else
322 				appends++;
323 		}
324 		/*
325 		 * Don't look for additional matches if this one does
326 		 * not have either the SO_REUSEPORT or SO_REUSEADDR
327 		 * socket options set.  This heuristic avoids
328 		 * searching through all pcbs in the common case of a
329 		 * non-shared port.  It assumes that an application
330 		 * will never clear these options after setting them.
331 		 */
332 		if ((inp->inp_socket->so_options &
333 		     (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
334 			INP_RUNLOCK(inp);
335 			break;
336 		}
337 	}
338 	m_freem(m);
339 
340 	if (appends == 0) {
341 		/*
342 		 * No matching pcb found; discard datagram.  (No need
343 		 * to send an ICMP Port Unreachable for a broadcast
344 		 * or multicast datgram.)
345 		 */
346 		UDPSTAT_INC(udps_noport);
347 		UDPSTAT_INC(udps_noportmcast);
348 	}
349 
350 	return (IPPROTO_DONE);
351 }
352 
353 int
354 udp6_input(struct mbuf **mp, int *offp, int proto)
355 {
356 	struct mbuf *m = *mp;
357 	struct ip6_hdr *ip6;
358 	struct udphdr *uh;
359 	struct inpcb *inp;
360 	struct inpcbinfo *pcbinfo;
361 	struct udpcb *up;
362 	int off = *offp;
363 	int cscov_partial;
364 	int plen, ulen;
365 	int lookupflags;
366 	struct sockaddr_in6 fromsa[2];
367 	struct m_tag *fwd_tag;
368 	uint16_t uh_sum;
369 	uint8_t nxt;
370 
371 	NET_EPOCH_ASSERT();
372 
373 	if (m->m_len < off + sizeof(struct udphdr)) {
374 		m = m_pullup(m, off + sizeof(struct udphdr));
375 		if (m == NULL) {
376 			IP6STAT_INC(ip6s_exthdrtoolong);
377 			*mp = NULL;
378 			return (IPPROTO_DONE);
379 		}
380 	}
381 	ip6 = mtod(m, struct ip6_hdr *);
382 	uh = (struct udphdr *)((caddr_t)ip6 + off);
383 
384 	UDPSTAT_INC(udps_ipackets);
385 
386 	/*
387 	 * Destination port of 0 is illegal, based on RFC768.
388 	 */
389 	if (uh->uh_dport == 0)
390 		goto badunlocked;
391 
392 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
393 	ulen = ntohs((u_short)uh->uh_ulen);
394 
395 	nxt = proto;
396 	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
397 	if (nxt == IPPROTO_UDPLITE) {
398 		/* Zero means checksum over the complete packet. */
399 		if (ulen == 0)
400 			ulen = plen;
401 		if (ulen == plen)
402 			cscov_partial = 0;
403 		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
404 			/* XXX: What is the right UDPLite MIB counter? */
405 			goto badunlocked;
406 		}
407 		if (uh->uh_sum == 0) {
408 			/* XXX: What is the right UDPLite MIB counter? */
409 			goto badunlocked;
410 		}
411 	} else {
412 		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
413 			UDPSTAT_INC(udps_badlen);
414 			goto badunlocked;
415 		}
416 		if (uh->uh_sum == 0) {
417 			UDPSTAT_INC(udps_nosum);
418 			/*
419 			 * dport 0 was rejected earlier so this is OK even if
420 			 * zero_checksum_port is 0 (which is its default value).
421 			 */
422 			if (ntohs(uh->uh_dport) == V_zero_checksum_port)
423 				goto skip_checksum;
424 			else
425 				goto badunlocked;
426 		}
427 	}
428 
429 	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
430 	    !cscov_partial) {
431 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
432 			uh_sum = m->m_pkthdr.csum_data;
433 		else
434 			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
435 			    m->m_pkthdr.csum_data);
436 		uh_sum ^= 0xffff;
437 	} else
438 		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
439 
440 	if (uh_sum != 0) {
441 		UDPSTAT_INC(udps_badsum);
442 		goto badunlocked;
443 	}
444 
445 skip_checksum:
446 	/*
447 	 * Construct sockaddr format source address.
448 	 */
449 	init_sin6(&fromsa[0], m, 0);
450 	fromsa[0].sin6_port = uh->uh_sport;
451 	init_sin6(&fromsa[1], m, 1);
452 	fromsa[1].sin6_port = uh->uh_dport;
453 
454 	pcbinfo = udp_get_inpcbinfo(nxt);
455 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))  {
456 		*mp = NULL;
457 		return (udp6_multi_input(m, off, proto, fromsa));
458 	}
459 
460 	/*
461 	 * Locate pcb for datagram.
462 	 */
463 	lookupflags = INPLOOKUP_RLOCKPCB |
464 	    (V_udp_bind_all_fibs ? 0 : INPLOOKUP_FIB);
465 
466 	/*
467 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
468 	 */
469 	if ((m->m_flags & M_IP6_NEXTHOP) &&
470 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
471 		struct sockaddr_in6 *next_hop6;
472 
473 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
474 
475 		/*
476 		 * Transparently forwarded. Pretend to be the destination.
477 		 * Already got one like this?
478 		 */
479 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
480 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
481 		    lookupflags, m->m_pkthdr.rcvif, m);
482 		if (!inp) {
483 			/*
484 			 * It's new.  Try to find the ambushing socket.
485 			 * Because we've rewritten the destination address,
486 			 * any hardware-generated hash is ignored.
487 			 */
488 			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
489 			    uh->uh_sport, &next_hop6->sin6_addr,
490 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
491 			    uh->uh_dport, INPLOOKUP_WILDCARD | lookupflags,
492 			    m->m_pkthdr.rcvif);
493 		}
494 		/* Remove the tag from the packet. We don't need it anymore. */
495 		m_tag_delete(m, fwd_tag);
496 		m->m_flags &= ~M_IP6_NEXTHOP;
497 	} else
498 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
499 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
500 		    INPLOOKUP_WILDCARD | lookupflags,
501 		    m->m_pkthdr.rcvif, m);
502 	if (inp == NULL) {
503 		if (V_udp_log_in_vain) {
504 			char ip6bufs[INET6_ADDRSTRLEN];
505 			char ip6bufd[INET6_ADDRSTRLEN];
506 
507 			log(LOG_INFO,
508 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
509 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
510 			    ntohs(uh->uh_dport),
511 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
512 			    ntohs(uh->uh_sport));
513 		}
514 		if (nxt == IPPROTO_UDPLITE)
515 			UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
516 		else
517 			UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
518 		UDPSTAT_INC(udps_noport);
519 		if (m->m_flags & M_MCAST) {
520 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
521 			UDPSTAT_INC(udps_noportmcast);
522 			goto badunlocked;
523 		}
524 		if (V_udp_blackhole && (V_udp_blackhole_local ||
525 		    !in6_localip(&ip6->ip6_src)))
526 			goto badunlocked;
527 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
528 		*mp = NULL;
529 		return (IPPROTO_DONE);
530 	}
531 	INP_RLOCK_ASSERT(inp);
532 	up = intoudpcb(inp);
533 	if (cscov_partial) {
534 		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
535 			INP_RUNLOCK(inp);
536 			m_freem(m);
537 			*mp = NULL;
538 			return (IPPROTO_DONE);
539 		}
540 	}
541 	if (nxt == IPPROTO_UDPLITE)
542 		UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
543 	else
544 		UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
545 	if (udp6_append(inp, m, off, fromsa) == 0)
546 		INP_RUNLOCK(inp);
547 	*mp = NULL;
548 	return (IPPROTO_DONE);
549 
550 badunlocked:
551 	m_freem(m);
552 	*mp = NULL;
553 	return (IPPROTO_DONE);
554 }
555 
556 static void
557 udp6_common_ctlinput(struct ip6ctlparam *ip6cp, struct inpcbinfo *pcbinfo)
558 {
559 	struct udphdr uh;
560 	struct ip6_hdr *ip6;
561 	struct mbuf *m;
562 	struct inpcb *inp;
563 	int errno, off = 0;
564 	struct udp_portonly {
565 		u_int16_t uh_sport;
566 		u_int16_t uh_dport;
567 	} *uhp;
568 
569 	if ((errno = icmp6_errmap(ip6cp->ip6c_icmp6)) == 0)
570 		return;
571 
572 	m = ip6cp->ip6c_m;
573 	ip6 = ip6cp->ip6c_ip6;
574 	off = ip6cp->ip6c_off;
575 
576 	/* Check if we can safely examine src and dst ports. */
577 	if (m->m_pkthdr.len < off + sizeof(*uhp))
578 		return;
579 
580 	bzero(&uh, sizeof(uh));
581 	m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
582 
583 	/* Check to see if its tunneled */
584 	inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport,
585 	    &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
586 	    m->m_pkthdr.rcvif, m);
587 	if (inp != NULL) {
588 		struct udpcb *up;
589 		udp_tun_icmp_t *func;
590 
591 		up = intoudpcb(inp);
592 		func = up->u_icmp_func;
593 		INP_RUNLOCK(inp);
594 		if (func != NULL)
595 			func(ip6cp);
596 	}
597 	in6_pcbnotify(pcbinfo, ip6cp->ip6c_finaldst, uh.uh_dport,
598 	    ip6cp->ip6c_src, uh.uh_sport, errno, ip6cp->ip6c_cmdarg,
599 	    udp_notify);
600 }
601 
602 static void
603 udp6_ctlinput(struct ip6ctlparam *ctl)
604 {
605 
606 	return (udp6_common_ctlinput(ctl, &V_udbinfo));
607 }
608 
609 static void
610 udplite6_ctlinput(struct ip6ctlparam *ctl)
611 {
612 
613 	return (udp6_common_ctlinput(ctl, &V_ulitecbinfo));
614 }
615 
616 static int
617 udp6_getcred(SYSCTL_HANDLER_ARGS)
618 {
619 	struct xucred xuc;
620 	struct sockaddr_in6 addrs[2];
621 	struct epoch_tracker et;
622 	struct inpcb *inp;
623 	int error;
624 
625 	if (req->newptr == NULL)
626 		return (EINVAL);
627 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
628 	if (error)
629 		return (error);
630 
631 	if (req->newlen != sizeof(addrs))
632 		return (EINVAL);
633 	if (req->oldlen != sizeof(struct xucred))
634 		return (EINVAL);
635 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
636 	if (error)
637 		return (error);
638 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
639 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
640 		return (error);
641 	}
642 	NET_EPOCH_ENTER(et);
643 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
644 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
645 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
646 	NET_EPOCH_EXIT(et);
647 	if (inp != NULL) {
648 		INP_RLOCK_ASSERT(inp);
649 		if (inp->inp_socket == NULL)
650 			error = ENOENT;
651 		if (error == 0)
652 			error = cr_canseesocket(req->td->td_ucred,
653 			    inp->inp_socket);
654 		if (error == 0)
655 			cru2x(inp->inp_cred, &xuc);
656 		INP_RUNLOCK(inp);
657 	} else
658 		error = ENOENT;
659 	if (error == 0)
660 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
661 	return (error);
662 }
663 
664 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
665     CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
666     0, 0, udp6_getcred, "S,xucred",
667     "Get the xucred of a UDP6 connection");
668 
669 static int
670 udp6_send(struct socket *so, int flags_arg, struct mbuf *m,
671     struct sockaddr *addr6, struct mbuf *control, struct thread *td)
672 {
673 	struct inpcb *inp;
674 	struct ip6_hdr *ip6;
675 	struct udphdr *udp6;
676 	struct in6_addr *laddr, *faddr, in6a;
677 	struct ip6_pktopts *optp, opt;
678 	struct sockaddr_in6 *sin6, tmp;
679 	struct epoch_tracker et;
680 	int cscov_partial, error, flags, hlen, scope_ambiguous;
681 	u_int32_t ulen, plen;
682 	uint16_t cscov;
683 	u_short fport;
684 	uint8_t nxt;
685 
686 	if (addr6) {
687 		error = 0;
688 		if (addr6->sa_family != AF_INET6)
689 			error = EAFNOSUPPORT;
690 		else if (addr6->sa_len != sizeof(struct sockaddr_in6))
691 			error = EINVAL;
692 		if (__predict_false(error != 0)) {
693 			m_freem(control);
694 			m_freem(m);
695 			return (error);
696 		}
697 	}
698 
699 	sin6 = (struct sockaddr_in6 *)addr6;
700 
701 	/*
702 	 * In contrast to IPv4 we do not validate the max. packet length
703 	 * here due to IPv6 Jumbograms (RFC2675).
704 	 */
705 
706 	scope_ambiguous = 0;
707 	if (sin6) {
708 		/* Protect *addr6 from overwrites. */
709 		tmp = *sin6;
710 		sin6 = &tmp;
711 
712 		/*
713 		 * Application should provide a proper zone ID or the use of
714 		 * default zone IDs should be enabled.  Unfortunately, some
715 		 * applications do not behave as it should, so we need a
716 		 * workaround.  Even if an appropriate ID is not determined,
717 		 * we'll see if we can determine the outgoing interface.  If we
718 		 * can, determine the zone ID based on the interface below.
719 		 */
720 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
721 			scope_ambiguous = 1;
722 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
723 			if (control)
724 				m_freem(control);
725 			m_freem(m);
726 			return (error);
727 		}
728 	}
729 
730 	inp = sotoinpcb(so);
731 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
732 	/*
733 	 * In the following cases we want a write lock on the inp for either
734 	 * local operations or for possible route cache updates in the IPv6
735 	 * output path:
736 	 * - on connected sockets (sin6 is NULL) for route cache updates,
737 	 * - when we are not bound to an address and source port (it is
738 	 *   in6_pcbsetport() which will require the write lock).
739 	 *
740 	 * We check the inp fields before actually locking the inp, so
741 	 * here exists a race, and we may WLOCK the inp and end with already
742 	 * bound one by other thread. This is fine.
743 	 */
744 	if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
745 	    inp->inp_lport == 0))
746 		INP_WLOCK(inp);
747 	else
748 		INP_RLOCK(inp);
749 
750 	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
751 	    IPPROTO_UDP : IPPROTO_UDPLITE;
752 
753 #ifdef INET
754 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
755 		int hasv4addr;
756 
757 		if (sin6 == NULL)
758 			hasv4addr = (inp->inp_vflag & INP_IPV4);
759 		else
760 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
761 			    ? 1 : 0;
762 		if (hasv4addr) {
763 			/*
764 			 * XXXRW: We release UDP-layer locks before calling
765 			 * udp_send() in order to avoid recursion.  However,
766 			 * this does mean there is a short window where inp's
767 			 * fields are unstable.  Could this lead to a
768 			 * potential race in which the factors causing us to
769 			 * select the UDPv4 output routine are invalidated?
770 			 */
771 			INP_UNLOCK(inp);
772 			if (sin6)
773 				in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
774 			/* addr will just be freed in sendit(). */
775 			return (udp_send(so, flags_arg | PRUS_IPV6, m,
776 			    (struct sockaddr *)sin6, control, td));
777 		}
778 	} else
779 #endif
780 	if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
781 		/*
782 		 * Given this is either an IPv6-only socket or no INET is
783 		 * supported we will fail the send if the given destination
784 		 * address is a v4mapped address.
785 		 */
786 		INP_UNLOCK(inp);
787 		m_freem(m);
788 		m_freem(control);
789 		return (EINVAL);
790 	}
791 
792 	NET_EPOCH_ENTER(et);
793 	if (control) {
794 		if ((error = ip6_setpktopts(control, &opt,
795 		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
796 			goto release;
797 		}
798 		optp = &opt;
799 	} else
800 		optp = inp->in6p_outputopts;
801 
802 	if (sin6) {
803 		/*
804 		 * Since we saw no essential reason for calling in_pcbconnect,
805 		 * we get rid of such kind of logic, and call in6_selectsrc
806 		 * and in6_pcbsetport in order to fill in the local address
807 		 * and the local port.
808 		 */
809 		if (sin6->sin6_port == 0) {
810 			error = EADDRNOTAVAIL;
811 			goto release;
812 		}
813 
814 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
815 			/* how about ::ffff:0.0.0.0 case? */
816 			error = EISCONN;
817 			goto release;
818 		}
819 
820 		/*
821 		 * Given we handle the v4mapped case in the INET block above
822 		 * assert here that it must not happen anymore.
823 		 */
824 		KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
825 		    ("%s: sin6(%p)->sin6_addr is v4mapped which we "
826 		    "should have handled.", __func__, sin6));
827 
828 		/* This only requires read-locking. */
829 		error = in6_selectsrc_socket(sin6, optp, inp,
830 		    td->td_ucred, scope_ambiguous, &in6a, NULL);
831 		if (error)
832 			goto release;
833 		laddr = &in6a;
834 
835 		if (inp->inp_lport == 0) {
836 			struct inpcbinfo *pcbinfo;
837 
838 			INP_WLOCK_ASSERT(inp);
839 
840 			pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
841 			INP_HASH_WLOCK(pcbinfo);
842 			error = in6_pcbsetport(laddr, inp, td->td_ucred);
843 			INP_HASH_WUNLOCK(pcbinfo);
844 			if (error != 0) {
845 				/* Undo an address bind that may have occurred. */
846 				inp->in6p_laddr = in6addr_any;
847 				goto release;
848 			}
849 		}
850 		faddr = &sin6->sin6_addr;
851 		fport = sin6->sin6_port; /* allow 0 port */
852 
853 	} else {
854 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
855 			error = ENOTCONN;
856 			goto release;
857 		}
858 		laddr = &inp->in6p_laddr;
859 		faddr = &inp->in6p_faddr;
860 		fport = inp->inp_fport;
861 	}
862 
863 	ulen = m->m_pkthdr.len;
864 	plen = sizeof(struct udphdr) + ulen;
865 	hlen = sizeof(struct ip6_hdr);
866 
867 	/*
868 	 * Calculate data length and get a mbuf for UDP, IP6, and possible
869 	 * link-layer headers.  Immediate slide the data pointer back forward
870 	 * since we won't use that space at this layer.
871 	 */
872 	M_PREPEND(m, hlen + sizeof(struct udphdr) + max_linkhdr, M_NOWAIT);
873 	if (m == NULL) {
874 		error = ENOBUFS;
875 		goto release;
876 	}
877 	m->m_data += max_linkhdr;
878 	m->m_len -= max_linkhdr;
879 	m->m_pkthdr.len -= max_linkhdr;
880 
881 	/*
882 	 * Stuff checksum and output datagram.
883 	 */
884 	cscov = cscov_partial = 0;
885 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
886 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
887 	udp6->uh_dport = fport;
888 	if (nxt == IPPROTO_UDPLITE) {
889 		struct udpcb *up;
890 
891 		up = intoudpcb(inp);
892 		cscov = up->u_txcslen;
893 		if (cscov >= plen)
894 			cscov = 0;
895 		udp6->uh_ulen = htons(cscov);
896 		/*
897 		 * For UDP-Lite, checksum coverage length of zero means
898 		 * the entire UDPLite packet is covered by the checksum.
899 		 */
900 		cscov_partial = (cscov == 0) ? 0 : 1;
901 	} else if (plen <= 0xffff)
902 		udp6->uh_ulen = htons((u_short)plen);
903 	else
904 		udp6->uh_ulen = 0;
905 	udp6->uh_sum = 0;
906 
907 	ip6 = mtod(m, struct ip6_hdr *);
908 	ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
909 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
910 	ip6->ip6_vfc	|= IPV6_VERSION;
911 	ip6->ip6_plen	= htons((u_short)plen);
912 	ip6->ip6_nxt	= nxt;
913 	ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
914 	ip6->ip6_src	= *laddr;
915 	ip6->ip6_dst	= *faddr;
916 
917 #ifdef MAC
918 	mac_inpcb_create_mbuf(inp, m);
919 #endif
920 
921 	if (cscov_partial) {
922 		if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
923 		    sizeof(struct ip6_hdr), plen, cscov)) == 0)
924 			udp6->uh_sum = 0xffff;
925 	} else {
926 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
927 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
928 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
929 	}
930 
931 	flags = 0;
932 #if defined(ROUTE_MPATH) || defined(RSS)
933 	if (CALC_FLOWID_OUTBOUND_SENDTO) {
934 		uint32_t hash_type, hash_val;
935 		uint8_t pr;
936 
937 		pr = inp->inp_socket->so_proto->pr_protocol;
938 
939 		hash_val = fib6_calc_packet_hash(laddr, faddr,
940 		    inp->inp_lport, fport, pr, &hash_type);
941 		m->m_pkthdr.flowid = hash_val;
942 		M_HASHTYPE_SET(m, hash_type);
943 	}
944 	/* do not use inp flowid */
945 	flags |= IP_NODEFAULTFLOWID;
946 #endif
947 
948 	UDPSTAT_INC(udps_opackets);
949 	if (nxt == IPPROTO_UDPLITE)
950 		UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
951 	else
952 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
953 	error = ip6_output(m, optp,
954 	    INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
955 	    inp->in6p_moptions, NULL, inp);
956 	INP_UNLOCK(inp);
957 	NET_EPOCH_EXIT(et);
958 
959 	if (control) {
960 		ip6_clearpktopts(&opt, -1);
961 		m_freem(control);
962 	}
963 	return (error);
964 
965 release:
966 	INP_UNLOCK(inp);
967 	NET_EPOCH_EXIT(et);
968 	if (control) {
969 		ip6_clearpktopts(&opt, -1);
970 		m_freem(control);
971 	}
972 	m_freem(m);
973 
974 	return (error);
975 }
976 
977 static void
978 udp6_abort(struct socket *so)
979 {
980 	struct inpcb *inp;
981 	struct inpcbinfo *pcbinfo;
982 
983 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
984 	inp = sotoinpcb(so);
985 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
986 
987 	INP_WLOCK(inp);
988 #ifdef INET
989 	if (inp->inp_vflag & INP_IPV4) {
990 		INP_WUNLOCK(inp);
991 		udp_abort(so);
992 		return;
993 	}
994 #endif
995 
996 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
997 		INP_HASH_WLOCK(pcbinfo);
998 		in6_pcbdisconnect(inp);
999 		INP_HASH_WUNLOCK(pcbinfo);
1000 		soisdisconnected(so);
1001 	}
1002 	INP_WUNLOCK(inp);
1003 }
1004 
1005 static int
1006 udp6_attach(struct socket *so, int proto, struct thread *td)
1007 {
1008 	struct inpcbinfo *pcbinfo;
1009 	struct inpcb *inp;
1010 	struct udpcb *up;
1011 	int error;
1012 
1013 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1014 	inp = sotoinpcb(so);
1015 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1016 
1017 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1018 		error = soreserve(so, udp_sendspace, udp_recvspace);
1019 		if (error)
1020 			return (error);
1021 	}
1022 	error = in_pcballoc(so, pcbinfo);
1023 	if (error)
1024 		return (error);
1025 	inp = (struct inpcb *)so->so_pcb;
1026 	inp->in6p_cksum = -1;	/* just to be sure */
1027 	/*
1028 	 * XXX: ugly!!
1029 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1030 	 * because the socket may be bound to an IPv6 wildcard address,
1031 	 * which may match an IPv4-mapped IPv6 address.
1032 	 */
1033 	inp->inp_ip_ttl = V_ip_defttl;
1034 	up = intoudpcb(inp);
1035 	bzero(&up->u_start_zero, u_zero_size);
1036 	INP_WUNLOCK(inp);
1037 	return (0);
1038 }
1039 
1040 static int
1041 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1042 {
1043 	struct sockaddr_in6 *sin6_p;
1044 	struct inpcb *inp;
1045 	struct inpcbinfo *pcbinfo;
1046 	int error;
1047 	u_char vflagsav;
1048 
1049 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1050 	inp = sotoinpcb(so);
1051 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1052 
1053 	if (nam->sa_family != AF_INET6)
1054 		return (EAFNOSUPPORT);
1055 	if (nam->sa_len != sizeof(struct sockaddr_in6))
1056 		return (EINVAL);
1057 
1058 	sin6_p = (struct sockaddr_in6 *)nam;
1059 
1060 	INP_WLOCK(inp);
1061 	INP_HASH_WLOCK(pcbinfo);
1062 	vflagsav = inp->inp_vflag;
1063 	inp->inp_vflag &= ~INP_IPV4;
1064 	inp->inp_vflag |= INP_IPV6;
1065 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1066 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1067 			inp->inp_vflag |= INP_IPV4;
1068 #ifdef INET
1069 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1070 			struct sockaddr_in sin;
1071 
1072 			in6_sin6_2_sin(&sin, sin6_p);
1073 			inp->inp_vflag |= INP_IPV4;
1074 			inp->inp_vflag &= ~INP_IPV6;
1075 			error = in_pcbbind(inp, &sin,
1076 			    V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1077 			    td->td_ucred);
1078 			goto out;
1079 		}
1080 #endif
1081 	}
1082 
1083 	error = in6_pcbbind(inp, sin6_p, V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1084 	    td->td_ucred);
1085 #ifdef INET
1086 out:
1087 #endif
1088 	if (error != 0)
1089 		inp->inp_vflag = vflagsav;
1090 	INP_HASH_WUNLOCK(pcbinfo);
1091 	INP_WUNLOCK(inp);
1092 	return (error);
1093 }
1094 
1095 static void
1096 udp6_close(struct socket *so)
1097 {
1098 	struct inpcb *inp;
1099 	struct inpcbinfo *pcbinfo;
1100 
1101 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1102 	inp = sotoinpcb(so);
1103 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1104 
1105 	INP_WLOCK(inp);
1106 #ifdef INET
1107 	if (inp->inp_vflag & INP_IPV4) {
1108 		INP_WUNLOCK(inp);
1109 		(void)udp_disconnect(so);
1110 		return;
1111 	}
1112 #endif
1113 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1114 		INP_HASH_WLOCK(pcbinfo);
1115 		in6_pcbdisconnect(inp);
1116 		INP_HASH_WUNLOCK(pcbinfo);
1117 		soisdisconnected(so);
1118 	}
1119 	INP_WUNLOCK(inp);
1120 }
1121 
1122 static int
1123 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1124 {
1125 	struct epoch_tracker et;
1126 	struct inpcb *inp;
1127 	struct inpcbinfo *pcbinfo;
1128 	struct sockaddr_in6 *sin6;
1129 	int error;
1130 	u_char vflagsav;
1131 
1132 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1133 	inp = sotoinpcb(so);
1134 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1135 
1136 	sin6 = (struct sockaddr_in6 *)nam;
1137 	if (sin6->sin6_family != AF_INET6)
1138 		return (EAFNOSUPPORT);
1139 	if (sin6->sin6_len != sizeof(*sin6))
1140 		return (EINVAL);
1141 
1142 	/*
1143 	 * XXXRW: Need to clarify locking of v4/v6 flags.
1144 	 */
1145 	INP_WLOCK(inp);
1146 #ifdef INET
1147 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1148 		struct sockaddr_in sin;
1149 
1150 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1151 			error = EINVAL;
1152 			goto out;
1153 		}
1154 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1155 			error = EAFNOSUPPORT;
1156 			goto out;
1157 		}
1158 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1159 			error = EISCONN;
1160 			goto out;
1161 		}
1162 		in6_sin6_2_sin(&sin, sin6);
1163 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1164 		if (error != 0)
1165 			goto out;
1166 		vflagsav = inp->inp_vflag;
1167 		inp->inp_vflag |= INP_IPV4;
1168 		inp->inp_vflag &= ~INP_IPV6;
1169 		NET_EPOCH_ENTER(et);
1170 		INP_HASH_WLOCK(pcbinfo);
1171 		error = in_pcbconnect(inp, &sin, td->td_ucred);
1172 		INP_HASH_WUNLOCK(pcbinfo);
1173 		NET_EPOCH_EXIT(et);
1174 		/*
1175 		 * If connect succeeds, mark socket as connected. If
1176 		 * connect fails and socket is unbound, reset inp_vflag
1177 		 * field.
1178 		 */
1179 		if (error == 0)
1180 			soisconnected(so);
1181 		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1182 		    inp->inp_lport == 0)
1183 			inp->inp_vflag = vflagsav;
1184 		goto out;
1185 	} else {
1186 		if ((inp->inp_vflag & INP_IPV6) == 0) {
1187 			error = EAFNOSUPPORT;
1188 			goto out;
1189 		}
1190 	}
1191 #endif
1192 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1193 		error = EISCONN;
1194 		goto out;
1195 	}
1196 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1197 	if (error != 0)
1198 		goto out;
1199 	vflagsav = inp->inp_vflag;
1200 	inp->inp_vflag &= ~INP_IPV4;
1201 	inp->inp_vflag |= INP_IPV6;
1202 	NET_EPOCH_ENTER(et);
1203 	INP_HASH_WLOCK(pcbinfo);
1204 	error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
1205 	INP_HASH_WUNLOCK(pcbinfo);
1206 	NET_EPOCH_EXIT(et);
1207 	/*
1208 	 * If connect succeeds, mark socket as connected. If
1209 	 * connect fails and socket is unbound, reset inp_vflag
1210 	 * field.
1211 	 */
1212 	if (error == 0)
1213 		soisconnected(so);
1214 	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1215 	    inp->inp_lport == 0)
1216 		inp->inp_vflag = vflagsav;
1217 out:
1218 	INP_WUNLOCK(inp);
1219 	return (error);
1220 }
1221 
1222 static void
1223 udp6_detach(struct socket *so)
1224 {
1225 	struct inpcb *inp;
1226 
1227 	inp = sotoinpcb(so);
1228 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1229 
1230 	INP_WLOCK(inp);
1231 	in_pcbfree(inp);
1232 }
1233 
1234 static int
1235 udp6_disconnect(struct socket *so)
1236 {
1237 	struct inpcb *inp;
1238 	struct inpcbinfo *pcbinfo;
1239 
1240 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1241 	inp = sotoinpcb(so);
1242 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1243 
1244 	INP_WLOCK(inp);
1245 #ifdef INET
1246 	if (inp->inp_vflag & INP_IPV4) {
1247 		INP_WUNLOCK(inp);
1248 		(void)udp_disconnect(so);
1249 		return (0);
1250 	}
1251 #endif
1252 
1253 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1254 		INP_WUNLOCK(inp);
1255 		return (ENOTCONN);
1256 	}
1257 
1258 	INP_HASH_WLOCK(pcbinfo);
1259 	in6_pcbdisconnect(inp);
1260 	INP_HASH_WUNLOCK(pcbinfo);
1261 	SOCK_LOCK(so);
1262 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1263 	SOCK_UNLOCK(so);
1264 	INP_WUNLOCK(inp);
1265 	return (0);
1266 }
1267 
1268 #define	UDP6_PROTOSW							\
1269 	.pr_type =		SOCK_DGRAM,				\
1270 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_CAPATTACH,		\
1271 	.pr_ctloutput =		udp_ctloutput,				\
1272 	.pr_abort =		udp6_abort,				\
1273 	.pr_attach =		udp6_attach,				\
1274 	.pr_bind =		udp6_bind,				\
1275 	.pr_connect =		udp6_connect,				\
1276 	.pr_control =		in6_control,				\
1277 	.pr_detach =		udp6_detach,				\
1278 	.pr_disconnect =	udp6_disconnect,			\
1279 	.pr_peeraddr =		in6_mapped_peeraddr,			\
1280 	.pr_send =		udp6_send,				\
1281 	.pr_shutdown =		udp_shutdown,				\
1282 	.pr_sockaddr =		in6_mapped_sockaddr,			\
1283 	.pr_soreceive =		soreceive_dgram,			\
1284 	.pr_sosend =		sosend_dgram,				\
1285 	.pr_sosetlabel =	in_pcbsosetlabel,			\
1286 	.pr_close =		udp6_close
1287 
1288 struct protosw udp6_protosw = {
1289 	.pr_protocol =		IPPROTO_UDP,
1290 	UDP6_PROTOSW
1291 };
1292 
1293 struct protosw udplite6_protosw = {
1294 	.pr_protocol =		IPPROTO_UDPLITE,
1295 	UDP6_PROTOSW
1296 };
1297 
1298 static void
1299 udp6_init(void *arg __unused)
1300 {
1301 
1302 	IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1303 	IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1304 }
1305 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);
1306