xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 690b7ea081790eef2c890f63a4fe7e195cf51df0)
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  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
70  */
71 
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
74 
75 #include "opt_inet.h"
76 #include "opt_inet6.h"
77 #include "opt_ipsec.h"
78 #include "opt_route.h"
79 #include "opt_rss.h"
80 
81 #include <sys/param.h>
82 #include <sys/jail.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/mbuf.h>
86 #include <sys/priv.h>
87 #include <sys/proc.h>
88 #include <sys/protosw.h>
89 #include <sys/sdt.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sx.h>
94 #include <sys/sysctl.h>
95 #include <sys/syslog.h>
96 #include <sys/systm.h>
97 
98 #include <net/if.h>
99 #include <net/if_var.h>
100 #include <net/if_types.h>
101 #include <net/route.h>
102 #include <net/rss_config.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/in_kdtrace.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/in_var.h>
109 #include <netinet/ip.h>
110 #include <netinet/ip6.h>
111 #include <netinet/icmp6.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/udp.h>
114 #include <netinet/udp_var.h>
115 #include <netinet/udplite.h>
116 
117 #include <netinet6/ip6protosw.h>
118 #include <netinet6/ip6_var.h>
119 #include <netinet6/in6_fib.h>
120 #include <netinet6/in6_pcb.h>
121 #include <netinet6/in6_rss.h>
122 #include <netinet6/udp6_var.h>
123 #include <netinet6/scope6_var.h>
124 
125 #include <netipsec/ipsec_support.h>
126 
127 #include <security/mac/mac_framework.h>
128 
129 VNET_DEFINE(int, zero_checksum_port) = 0;
130 #define	V_zero_checksum_port	VNET(zero_checksum_port)
131 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
132     &VNET_NAME(zero_checksum_port), 0,
133     "Zero UDP checksum allowed for traffic to/from this port.");
134 /*
135  * UDP protocol implementation.
136  * Per RFC 768, August, 1980.
137  */
138 
139 extern struct protosw	inetsw[];
140 static void		udp6_detach(struct socket *so);
141 
142 static int
143 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
144     struct sockaddr_in6 *fromsa)
145 {
146 	struct socket *so;
147 	struct mbuf *opts = NULL, *tmp_opts;
148 	struct udpcb *up;
149 	bool filtered;
150 
151 	INP_LOCK_ASSERT(inp);
152 
153 	/*
154 	 * Engage the tunneling protocol.
155 	 */
156 	up = intoudpcb(inp);
157 	if (up->u_tun_func != NULL) {
158 		in_pcbref(inp);
159 		INP_RUNLOCK(inp);
160 		filtered = (*up->u_tun_func)(n, off, inp,
161 		    (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
162 		INP_RLOCK(inp);
163 		if (filtered)
164 			return (in_pcbrele_rlocked(inp));
165 	}
166 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
167 	/* Check AH/ESP integrity. */
168 	if (IPSEC_ENABLED(ipv6)) {
169 		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
170 			m_freem(n);
171 			return (0);
172 		}
173 	}
174 #endif /* IPSEC */
175 #ifdef MAC
176 	if (mac_inpcb_check_deliver(inp, n) != 0) {
177 		m_freem(n);
178 		return (0);
179 	}
180 #endif
181 	opts = NULL;
182 	if (inp->inp_flags & INP_CONTROLOPTS ||
183 	    inp->inp_socket->so_options & SO_TIMESTAMP)
184 		ip6_savecontrol(inp, n, &opts);
185 	if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
186 		tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
187                         sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
188                 if (tmp_opts) {
189                         if (opts) {
190                                 tmp_opts->m_next = opts;
191                                 opts = tmp_opts;
192                         } else
193                                 opts = tmp_opts;
194                 }
195 	}
196 	m_adj(n, off + sizeof(struct udphdr));
197 
198 	so = inp->inp_socket;
199 	SOCKBUF_LOCK(&so->so_rcv);
200 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
201 	    opts) == 0) {
202 		soroverflow_locked(so);
203 		m_freem(n);
204 		if (opts)
205 			m_freem(opts);
206 		UDPSTAT_INC(udps_fullsock);
207 	} else
208 		sorwakeup_locked(so);
209 	return (0);
210 }
211 
212 struct udp6_multi_match_ctx {
213 	struct ip6_hdr *ip6;
214 	struct udphdr *uh;
215 };
216 
217 static bool
218 udp6_multi_match(const struct inpcb *inp, void *v)
219 {
220 	struct udp6_multi_match_ctx *ctx = v;
221 
222 	if ((inp->inp_vflag & INP_IPV6) == 0)
223 		return(false);
224 	if (inp->inp_lport != ctx->uh->uh_dport)
225 		return(false);
226 	if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
227 		return(false);
228 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
229 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
230 		return (false);
231 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
232 	    (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
233 	    inp->inp_fport != ctx->uh->uh_sport))
234 		return (false);
235 
236 	return (true);
237 }
238 
239 static int
240 udp6_multi_input(struct mbuf *m, int off, int proto,
241     struct sockaddr_in6 *fromsa)
242 {
243 	struct udp6_multi_match_ctx ctx;
244 	struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
245 	    INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
246 	struct inpcb *inp;
247 	struct ip6_moptions *imo;
248 	struct mbuf *n;
249 	int appends = 0;
250 
251 	/*
252 	 * In the event that laddr should be set to the link-local
253 	 * address (this happens in RIPng), the multicast address
254 	 * specified in the received packet will not match laddr.  To
255 	 * handle this situation, matching is relaxed if the
256 	 * receiving interface is the same as one specified in the
257 	 * socket and if the destination multicast address matches
258 	 * one of the multicast groups specified in the socket.
259 	 */
260 
261 	/*
262 	 * KAME note: traditionally we dropped udpiphdr from mbuf
263 	 * here.  We need udphdr for IPsec processing so we do that
264 	 * later.
265 	 */
266 	ctx.ip6 = mtod(m, struct ip6_hdr *);
267 	ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
268 	while ((inp = inp_next(&inpi)) != NULL) {
269 		INP_RLOCK_ASSERT(inp);
270 		/*
271 		 * XXXRW: Because we weren't holding either the inpcb
272 		 * or the hash lock when we checked for a match
273 		 * before, we should probably recheck now that the
274 		 * inpcb lock is (supposed to be) held.
275 		 */
276 		/*
277 		 * Handle socket delivery policy for any-source
278 		 * and source-specific multicast. [RFC3678]
279 		 */
280 		if ((imo = inp->in6p_moptions) != NULL) {
281 			struct sockaddr_in6	 mcaddr;
282 			int			 blocked;
283 
284 			bzero(&mcaddr, sizeof(struct sockaddr_in6));
285 			mcaddr.sin6_len = sizeof(struct sockaddr_in6);
286 			mcaddr.sin6_family = AF_INET6;
287 			mcaddr.sin6_addr = ctx.ip6->ip6_dst;
288 
289 			blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
290 				(struct sockaddr *)&mcaddr,
291 				(struct sockaddr *)&fromsa[0]);
292 			if (blocked != MCAST_PASS) {
293 				if (blocked == MCAST_NOTGMEMBER)
294 					IP6STAT_INC(ip6s_notmember);
295 				if (blocked == MCAST_NOTSMEMBER ||
296 				    blocked == MCAST_MUTED)
297 					UDPSTAT_INC(udps_filtermcast);
298 				continue;
299 			}
300 		}
301 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
302 			if (proto == IPPROTO_UDPLITE)
303 				UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
304 				    inp, ctx.uh);
305 			else
306 				UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
307 				    ctx.uh);
308 			if (udp6_append(inp, n, off, fromsa)) {
309 				INP_RUNLOCK(inp);
310 				break;
311 			} else
312 				appends++;
313 		}
314 		/*
315 		 * Don't look for additional matches if this one does
316 		 * not have either the SO_REUSEPORT or SO_REUSEADDR
317 		 * socket options set.  This heuristic avoids
318 		 * searching through all pcbs in the common case of a
319 		 * non-shared port.  It assumes that an application
320 		 * will never clear these options after setting them.
321 		 */
322 		if ((inp->inp_socket->so_options &
323 		     (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
324 			INP_RUNLOCK(inp);
325 			break;
326 		}
327 	}
328 	m_freem(m);
329 
330 	if (appends == 0) {
331 		/*
332 		 * No matching pcb found; discard datagram.  (No need
333 		 * to send an ICMP Port Unreachable for a broadcast
334 		 * or multicast datgram.)
335 		 */
336 		UDPSTAT_INC(udps_noport);
337 		UDPSTAT_INC(udps_noportmcast);
338 	}
339 
340 	return (IPPROTO_DONE);
341 }
342 
343 int
344 udp6_input(struct mbuf **mp, int *offp, int proto)
345 {
346 	struct mbuf *m = *mp;
347 	struct ip6_hdr *ip6;
348 	struct udphdr *uh;
349 	struct inpcb *inp;
350 	struct inpcbinfo *pcbinfo;
351 	struct udpcb *up;
352 	int off = *offp;
353 	int cscov_partial;
354 	int plen, ulen;
355 	struct sockaddr_in6 fromsa[2];
356 	struct m_tag *fwd_tag;
357 	uint16_t uh_sum;
358 	uint8_t nxt;
359 
360 	NET_EPOCH_ASSERT();
361 
362 	if (m->m_len < off + sizeof(struct udphdr)) {
363 		m = m_pullup(m, off + sizeof(struct udphdr));
364 		if (m == NULL) {
365 			IP6STAT_INC(ip6s_exthdrtoolong);
366 			*mp = NULL;
367 			return (IPPROTO_DONE);
368 		}
369 	}
370 	ip6 = mtod(m, struct ip6_hdr *);
371 	uh = (struct udphdr *)((caddr_t)ip6 + off);
372 
373 	UDPSTAT_INC(udps_ipackets);
374 
375 	/*
376 	 * Destination port of 0 is illegal, based on RFC768.
377 	 */
378 	if (uh->uh_dport == 0)
379 		goto badunlocked;
380 
381 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
382 	ulen = ntohs((u_short)uh->uh_ulen);
383 
384 	nxt = proto;
385 	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
386 	if (nxt == IPPROTO_UDPLITE) {
387 		/* Zero means checksum over the complete packet. */
388 		if (ulen == 0)
389 			ulen = plen;
390 		if (ulen == plen)
391 			cscov_partial = 0;
392 		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
393 			/* XXX: What is the right UDPLite MIB counter? */
394 			goto badunlocked;
395 		}
396 		if (uh->uh_sum == 0) {
397 			/* XXX: What is the right UDPLite MIB counter? */
398 			goto badunlocked;
399 		}
400 	} else {
401 		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
402 			UDPSTAT_INC(udps_badlen);
403 			goto badunlocked;
404 		}
405 		if (uh->uh_sum == 0) {
406 			UDPSTAT_INC(udps_nosum);
407 			/*
408 			 * dport 0 was rejected earlier so this is OK even if
409 			 * zero_checksum_port is 0 (which is its default value).
410 			 */
411 			if (ntohs(uh->uh_dport) == V_zero_checksum_port)
412 				goto skip_checksum;
413 			else
414 				goto badunlocked;
415 		}
416 	}
417 
418 	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
419 	    !cscov_partial) {
420 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
421 			uh_sum = m->m_pkthdr.csum_data;
422 		else
423 			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
424 			    m->m_pkthdr.csum_data);
425 		uh_sum ^= 0xffff;
426 	} else
427 		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
428 
429 	if (uh_sum != 0) {
430 		UDPSTAT_INC(udps_badsum);
431 		goto badunlocked;
432 	}
433 
434 skip_checksum:
435 	/*
436 	 * Construct sockaddr format source address.
437 	 */
438 	init_sin6(&fromsa[0], m, 0);
439 	fromsa[0].sin6_port = uh->uh_sport;
440 	init_sin6(&fromsa[1], m, 1);
441 	fromsa[1].sin6_port = uh->uh_dport;
442 
443 	pcbinfo = udp_get_inpcbinfo(nxt);
444 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))  {
445 		*mp = NULL;
446 		return (udp6_multi_input(m, off, proto, fromsa));
447 	}
448 
449 	/*
450 	 * Locate pcb for datagram.
451 	 */
452 
453 	/*
454 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
455 	 */
456 	if ((m->m_flags & M_IP6_NEXTHOP) &&
457 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
458 		struct sockaddr_in6 *next_hop6;
459 
460 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
461 
462 		/*
463 		 * Transparently forwarded. Pretend to be the destination.
464 		 * Already got one like this?
465 		 */
466 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
467 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
468 		    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
469 		if (!inp) {
470 			/*
471 			 * It's new.  Try to find the ambushing socket.
472 			 * Because we've rewritten the destination address,
473 			 * any hardware-generated hash is ignored.
474 			 */
475 			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
476 			    uh->uh_sport, &next_hop6->sin6_addr,
477 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
478 			    uh->uh_dport, INPLOOKUP_WILDCARD |
479 			    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
480 		}
481 		/* Remove the tag from the packet. We don't need it anymore. */
482 		m_tag_delete(m, fwd_tag);
483 		m->m_flags &= ~M_IP6_NEXTHOP;
484 	} else
485 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
486 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
487 		    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
488 		    m->m_pkthdr.rcvif, m);
489 	if (inp == NULL) {
490 		if (V_udp_log_in_vain) {
491 			char ip6bufs[INET6_ADDRSTRLEN];
492 			char ip6bufd[INET6_ADDRSTRLEN];
493 
494 			log(LOG_INFO,
495 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
496 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
497 			    ntohs(uh->uh_dport),
498 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
499 			    ntohs(uh->uh_sport));
500 		}
501 		if (nxt == IPPROTO_UDPLITE)
502 			UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
503 		else
504 			UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
505 		UDPSTAT_INC(udps_noport);
506 		if (m->m_flags & M_MCAST) {
507 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
508 			UDPSTAT_INC(udps_noportmcast);
509 			goto badunlocked;
510 		}
511 		if (V_udp_blackhole && (V_udp_blackhole_local ||
512 		    !in6_localaddr(&ip6->ip6_src)))
513 			goto badunlocked;
514 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
515 		*mp = NULL;
516 		return (IPPROTO_DONE);
517 	}
518 	INP_RLOCK_ASSERT(inp);
519 	up = intoudpcb(inp);
520 	if (cscov_partial) {
521 		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
522 			INP_RUNLOCK(inp);
523 			m_freem(m);
524 			*mp = NULL;
525 			return (IPPROTO_DONE);
526 		}
527 	}
528 	if (nxt == IPPROTO_UDPLITE)
529 		UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
530 	else
531 		UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
532 	if (udp6_append(inp, m, off, fromsa) == 0)
533 		INP_RUNLOCK(inp);
534 	*mp = NULL;
535 	return (IPPROTO_DONE);
536 
537 badunlocked:
538 	m_freem(m);
539 	*mp = NULL;
540 	return (IPPROTO_DONE);
541 }
542 
543 static void
544 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
545     struct inpcbinfo *pcbinfo)
546 {
547 	struct udphdr uh;
548 	struct ip6_hdr *ip6;
549 	struct mbuf *m;
550 	int off = 0;
551 	struct ip6ctlparam *ip6cp = NULL;
552 	const struct sockaddr_in6 *sa6_src = NULL;
553 	void *cmdarg;
554 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
555 	struct udp_portonly {
556 		u_int16_t uh_sport;
557 		u_int16_t uh_dport;
558 	} *uhp;
559 
560 	if (sa->sa_family != AF_INET6 ||
561 	    sa->sa_len != sizeof(struct sockaddr_in6))
562 		return;
563 
564 	if ((unsigned)cmd >= PRC_NCMDS)
565 		return;
566 	if (PRC_IS_REDIRECT(cmd))
567 		notify = in6_rtchange, d = NULL;
568 	else if (cmd == PRC_HOSTDEAD)
569 		d = NULL;
570 	else if (inet6ctlerrmap[cmd] == 0)
571 		return;
572 
573 	/* if the parameter is from icmp6, decode it. */
574 	if (d != NULL) {
575 		ip6cp = (struct ip6ctlparam *)d;
576 		m = ip6cp->ip6c_m;
577 		ip6 = ip6cp->ip6c_ip6;
578 		off = ip6cp->ip6c_off;
579 		cmdarg = ip6cp->ip6c_cmdarg;
580 		sa6_src = ip6cp->ip6c_src;
581 	} else {
582 		m = NULL;
583 		ip6 = NULL;
584 		cmdarg = NULL;
585 		sa6_src = &sa6_any;
586 	}
587 
588 	if (ip6) {
589 		/*
590 		 * XXX: We assume that when IPV6 is non NULL,
591 		 * M and OFF are valid.
592 		 */
593 
594 		/* Check if we can safely examine src and dst ports. */
595 		if (m->m_pkthdr.len < off + sizeof(*uhp))
596 			return;
597 
598 		bzero(&uh, sizeof(uh));
599 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
600 
601 		if (!PRC_IS_REDIRECT(cmd)) {
602 			/* Check to see if its tunneled */
603 			struct inpcb *inp;
604 			inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
605 			    uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
606 			    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
607 			    m->m_pkthdr.rcvif, m);
608 			if (inp != NULL) {
609 				struct udpcb *up;
610 
611 				up = intoudpcb(inp);
612 				if (up->u_icmp_func) {
613 					/* Yes it is. */
614 					INP_RUNLOCK(inp);
615 					(*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
616 					      d, up->u_tun_ctx);
617 					return;
618 				} else {
619 					/* Can't find it. */
620 					INP_RUNLOCK(inp);
621 				}
622 			}
623 		}
624 		(void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
625 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
626 		    cmdarg, notify);
627 	} else
628 		(void)in6_pcbnotify(pcbinfo, sa, 0,
629 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
630 }
631 
632 void
633 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
634 {
635 
636 	return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
637 }
638 
639 void
640 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
641 {
642 
643 	return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
644 }
645 
646 static int
647 udp6_getcred(SYSCTL_HANDLER_ARGS)
648 {
649 	struct xucred xuc;
650 	struct sockaddr_in6 addrs[2];
651 	struct epoch_tracker et;
652 	struct inpcb *inp;
653 	int error;
654 
655 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
656 	if (error)
657 		return (error);
658 
659 	if (req->newlen != sizeof(addrs))
660 		return (EINVAL);
661 	if (req->oldlen != sizeof(struct xucred))
662 		return (EINVAL);
663 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
664 	if (error)
665 		return (error);
666 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
667 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
668 		return (error);
669 	}
670 	NET_EPOCH_ENTER(et);
671 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
672 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
673 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
674 	NET_EPOCH_EXIT(et);
675 	if (inp != NULL) {
676 		INP_RLOCK_ASSERT(inp);
677 		if (inp->inp_socket == NULL)
678 			error = ENOENT;
679 		if (error == 0)
680 			error = cr_canseesocket(req->td->td_ucred,
681 			    inp->inp_socket);
682 		if (error == 0)
683 			cru2x(inp->inp_cred, &xuc);
684 		INP_RUNLOCK(inp);
685 	} else
686 		error = ENOENT;
687 	if (error == 0)
688 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
689 	return (error);
690 }
691 
692 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
693     CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
694     0, 0, udp6_getcred, "S,xucred",
695     "Get the xucred of a UDP6 connection");
696 
697 static int
698 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
699     struct sockaddr *addr6, struct mbuf *control, struct thread *td)
700 {
701 	struct inpcb *inp;
702 	struct ip6_hdr *ip6;
703 	struct udphdr *udp6;
704 	struct in6_addr *laddr, *faddr, in6a;
705 	struct ip6_pktopts *optp, opt;
706 	struct sockaddr_in6 *sin6, tmp;
707 	struct epoch_tracker et;
708 	int cscov_partial, error, flags, hlen, scope_ambiguous;
709 	u_int32_t ulen, plen;
710 	uint16_t cscov;
711 	u_short fport;
712 	uint8_t nxt;
713 
714 	/* addr6 has been validated in udp6_send(). */
715 	sin6 = (struct sockaddr_in6 *)addr6;
716 
717 	/*
718 	 * In contrast to to IPv4 we do not validate the max. packet length
719 	 * here due to IPv6 Jumbograms (RFC2675).
720 	 */
721 
722 	scope_ambiguous = 0;
723 	if (sin6) {
724 		/* Protect *addr6 from overwrites. */
725 		tmp = *sin6;
726 		sin6 = &tmp;
727 
728 		/*
729 		 * Application should provide a proper zone ID or the use of
730 		 * default zone IDs should be enabled.  Unfortunately, some
731 		 * applications do not behave as it should, so we need a
732 		 * workaround.  Even if an appropriate ID is not determined,
733 		 * we'll see if we can determine the outgoing interface.  If we
734 		 * can, determine the zone ID based on the interface below.
735 		 */
736 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
737 			scope_ambiguous = 1;
738 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
739 			if (control)
740 				m_freem(control);
741 			m_freem(m);
742 			return (error);
743 		}
744 	}
745 
746 	inp = sotoinpcb(so);
747 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
748 	/*
749 	 * In the following cases we want a write lock on the inp for either
750 	 * local operations or for possible route cache updates in the IPv6
751 	 * output path:
752 	 * - on connected sockets (sin6 is NULL) for route cache updates,
753 	 * - when we are not bound to an address and source port (it is
754 	 *   in6_pcbsetport() which will require the write lock).
755 	 *
756 	 * We check the inp fields before actually locking the inp, so
757 	 * here exists a race, and we may WLOCK the inp and end with already
758 	 * bound one by other thread. This is fine.
759 	 */
760 	if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
761 	    inp->inp_lport == 0))
762 		INP_WLOCK(inp);
763 	else
764 		INP_RLOCK(inp);
765 
766 	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
767 	    IPPROTO_UDP : IPPROTO_UDPLITE;
768 
769 #ifdef INET
770 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
771 		int hasv4addr;
772 
773 		if (sin6 == NULL)
774 			hasv4addr = (inp->inp_vflag & INP_IPV4);
775 		else
776 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
777 			    ? 1 : 0;
778 		if (hasv4addr) {
779 			struct pr_usrreqs *pru;
780 
781 			/*
782 			 * XXXRW: We release UDP-layer locks before calling
783 			 * udp_send() in order to avoid recursion.  However,
784 			 * this does mean there is a short window where inp's
785 			 * fields are unstable.  Could this lead to a
786 			 * potential race in which the factors causing us to
787 			 * select the UDPv4 output routine are invalidated?
788 			 */
789 			INP_UNLOCK(inp);
790 			if (sin6)
791 				in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
792 			pru = inetsw[ip_protox[nxt]].pr_usrreqs;
793 			/* addr will just be freed in sendit(). */
794 			return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
795 			    (struct sockaddr *)sin6, control, td));
796 		}
797 	} else
798 #endif
799 	if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
800 		/*
801 		 * Given this is either an IPv6-only socket or no INET is
802 		 * supported we will fail the send if the given destination
803 		 * address is a v4mapped address.
804 		 *
805 		 * XXXGL: do we leak m and control?
806 		 */
807 		INP_UNLOCK(inp);
808 		return (EINVAL);
809 	}
810 
811 	NET_EPOCH_ENTER(et);
812 	if (control) {
813 		if ((error = ip6_setpktopts(control, &opt,
814 		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
815 			goto release;
816 		}
817 		optp = &opt;
818 	} else
819 		optp = inp->in6p_outputopts;
820 
821 	if (sin6) {
822 		/*
823 		 * Since we saw no essential reason for calling in_pcbconnect,
824 		 * we get rid of such kind of logic, and call in6_selectsrc
825 		 * and in6_pcbsetport in order to fill in the local address
826 		 * and the local port.
827 		 */
828 		if (sin6->sin6_port == 0) {
829 			error = EADDRNOTAVAIL;
830 			goto release;
831 		}
832 
833 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
834 			/* how about ::ffff:0.0.0.0 case? */
835 			error = EISCONN;
836 			goto release;
837 		}
838 
839 		/*
840 		 * Given we handle the v4mapped case in the INET block above
841 		 * assert here that it must not happen anymore.
842 		 */
843 		KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
844 		    ("%s: sin6(%p)->sin6_addr is v4mapped which we "
845 		    "should have handled.", __func__, sin6));
846 
847 		/* This only requires read-locking. */
848 		error = in6_selectsrc_socket(sin6, optp, inp,
849 		    td->td_ucred, scope_ambiguous, &in6a, NULL);
850 		if (error)
851 			goto release;
852 		laddr = &in6a;
853 
854 		if (inp->inp_lport == 0) {
855 			struct inpcbinfo *pcbinfo;
856 
857 			INP_WLOCK_ASSERT(inp);
858 
859 			pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
860 			INP_HASH_WLOCK(pcbinfo);
861 			error = in6_pcbsetport(laddr, inp, td->td_ucred);
862 			INP_HASH_WUNLOCK(pcbinfo);
863 			if (error != 0) {
864 				/* Undo an address bind that may have occurred. */
865 				inp->in6p_laddr = in6addr_any;
866 				goto release;
867 			}
868 		}
869 		faddr = &sin6->sin6_addr;
870 		fport = sin6->sin6_port; /* allow 0 port */
871 
872 	} else {
873 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
874 			error = ENOTCONN;
875 			goto release;
876 		}
877 		laddr = &inp->in6p_laddr;
878 		faddr = &inp->in6p_faddr;
879 		fport = inp->inp_fport;
880 	}
881 
882 	ulen = m->m_pkthdr.len;
883 	plen = sizeof(struct udphdr) + ulen;
884 	hlen = sizeof(struct ip6_hdr);
885 
886 	/*
887 	 * Calculate data length and get a mbuf
888 	 * for UDP and IP6 headers.
889 	 */
890 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
891 	if (m == NULL) {
892 		error = ENOBUFS;
893 		goto release;
894 	}
895 
896 	/*
897 	 * Stuff checksum and output datagram.
898 	 */
899 	cscov = cscov_partial = 0;
900 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
901 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
902 	udp6->uh_dport = fport;
903 	if (nxt == IPPROTO_UDPLITE) {
904 		struct udpcb *up;
905 
906 		up = intoudpcb(inp);
907 		cscov = up->u_txcslen;
908 		if (cscov >= plen)
909 			cscov = 0;
910 		udp6->uh_ulen = htons(cscov);
911 		/*
912 		 * For UDP-Lite, checksum coverage length of zero means
913 		 * the entire UDPLite packet is covered by the checksum.
914 		 */
915 		cscov_partial = (cscov == 0) ? 0 : 1;
916 	} else if (plen <= 0xffff)
917 		udp6->uh_ulen = htons((u_short)plen);
918 	else
919 		udp6->uh_ulen = 0;
920 	udp6->uh_sum = 0;
921 
922 	ip6 = mtod(m, struct ip6_hdr *);
923 	ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
924 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
925 	ip6->ip6_vfc	|= IPV6_VERSION;
926 	ip6->ip6_plen	= htons((u_short)plen);
927 	ip6->ip6_nxt	= nxt;
928 	ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
929 	ip6->ip6_src	= *laddr;
930 	ip6->ip6_dst	= *faddr;
931 
932 #ifdef MAC
933 	mac_inpcb_create_mbuf(inp, m);
934 #endif
935 
936 	if (cscov_partial) {
937 		if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
938 		    sizeof(struct ip6_hdr), plen, cscov)) == 0)
939 			udp6->uh_sum = 0xffff;
940 	} else {
941 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
942 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
943 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
944 	}
945 
946 	flags = 0;
947 #if defined(ROUTE_MPATH) || defined(RSS)
948 	if (CALC_FLOWID_OUTBOUND_SENDTO) {
949 		uint32_t hash_type, hash_val;
950 		uint8_t pr;
951 
952 		pr = inp->inp_socket->so_proto->pr_protocol;
953 
954 		hash_val = fib6_calc_packet_hash(laddr, faddr,
955 		    inp->inp_lport, fport, pr, &hash_type);
956 		m->m_pkthdr.flowid = hash_val;
957 		M_HASHTYPE_SET(m, hash_type);
958 	}
959 	/* do not use inp flowid */
960 	flags |= IP_NODEFAULTFLOWID;
961 #endif
962 
963 	UDPSTAT_INC(udps_opackets);
964 	if (nxt == IPPROTO_UDPLITE)
965 		UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
966 	else
967 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
968 	error = ip6_output(m, optp,
969 	    INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
970 	    inp->in6p_moptions, NULL, inp);
971 	INP_UNLOCK(inp);
972 	NET_EPOCH_EXIT(et);
973 
974 	if (control) {
975 		ip6_clearpktopts(&opt, -1);
976 		m_freem(control);
977 	}
978 	return (error);
979 
980 release:
981 	INP_UNLOCK(inp);
982 	NET_EPOCH_EXIT(et);
983 	if (control) {
984 		ip6_clearpktopts(&opt, -1);
985 		m_freem(control);
986 	}
987 	m_freem(m);
988 
989 	return (error);
990 }
991 
992 static void
993 udp6_abort(struct socket *so)
994 {
995 	struct inpcb *inp;
996 	struct inpcbinfo *pcbinfo;
997 
998 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
999 	inp = sotoinpcb(so);
1000 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1001 
1002 	INP_WLOCK(inp);
1003 #ifdef INET
1004 	if (inp->inp_vflag & INP_IPV4) {
1005 		struct pr_usrreqs *pru;
1006 		uint8_t nxt;
1007 
1008 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1009 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1010 		INP_WUNLOCK(inp);
1011 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1012 		(*pru->pru_abort)(so);
1013 		return;
1014 	}
1015 #endif
1016 
1017 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1018 		INP_HASH_WLOCK(pcbinfo);
1019 		in6_pcbdisconnect(inp);
1020 		inp->in6p_laddr = in6addr_any;
1021 		INP_HASH_WUNLOCK(pcbinfo);
1022 		soisdisconnected(so);
1023 	}
1024 	INP_WUNLOCK(inp);
1025 }
1026 
1027 static int
1028 udp6_attach(struct socket *so, int proto, struct thread *td)
1029 {
1030 	struct inpcb *inp;
1031 	struct inpcbinfo *pcbinfo;
1032 	int error;
1033 
1034 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1035 	inp = sotoinpcb(so);
1036 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1037 
1038 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1039 		error = soreserve(so, udp_sendspace, udp_recvspace);
1040 		if (error)
1041 			return (error);
1042 	}
1043 	error = in_pcballoc(so, pcbinfo);
1044 	if (error)
1045 		return (error);
1046 	inp = (struct inpcb *)so->so_pcb;
1047 	inp->inp_vflag |= INP_IPV6;
1048 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1049 		inp->inp_vflag |= INP_IPV4;
1050 	inp->in6p_hops = -1;	/* use kernel default */
1051 	inp->in6p_cksum = -1;	/* just to be sure */
1052 	/*
1053 	 * XXX: ugly!!
1054 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1055 	 * because the socket may be bound to an IPv6 wildcard address,
1056 	 * which may match an IPv4-mapped IPv6 address.
1057 	 */
1058 	inp->inp_ip_ttl = V_ip_defttl;
1059 
1060 	error = udp_newudpcb(inp);
1061 	if (error) {
1062 		in_pcbdetach(inp);
1063 		in_pcbfree(inp);
1064 		return (error);
1065 	}
1066 	INP_WUNLOCK(inp);
1067 	return (0);
1068 }
1069 
1070 static int
1071 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1072 {
1073 	struct inpcb *inp;
1074 	struct inpcbinfo *pcbinfo;
1075 	int error;
1076 	u_char vflagsav;
1077 
1078 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1079 	inp = sotoinpcb(so);
1080 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1081 
1082 	if (nam->sa_family != AF_INET6)
1083 		return (EAFNOSUPPORT);
1084 	if (nam->sa_len != sizeof(struct sockaddr_in6))
1085 		return (EINVAL);
1086 
1087 	INP_WLOCK(inp);
1088 	INP_HASH_WLOCK(pcbinfo);
1089 	vflagsav = inp->inp_vflag;
1090 	inp->inp_vflag &= ~INP_IPV4;
1091 	inp->inp_vflag |= INP_IPV6;
1092 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1093 		struct sockaddr_in6 *sin6_p;
1094 
1095 		sin6_p = (struct sockaddr_in6 *)nam;
1096 
1097 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1098 			inp->inp_vflag |= INP_IPV4;
1099 #ifdef INET
1100 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1101 			struct sockaddr_in sin;
1102 
1103 			in6_sin6_2_sin(&sin, sin6_p);
1104 			inp->inp_vflag |= INP_IPV4;
1105 			inp->inp_vflag &= ~INP_IPV6;
1106 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
1107 			    td->td_ucred);
1108 			goto out;
1109 		}
1110 #endif
1111 	}
1112 
1113 	error = in6_pcbbind(inp, nam, td->td_ucred);
1114 #ifdef INET
1115 out:
1116 #endif
1117 	if (error != 0)
1118 		inp->inp_vflag = vflagsav;
1119 	INP_HASH_WUNLOCK(pcbinfo);
1120 	INP_WUNLOCK(inp);
1121 	return (error);
1122 }
1123 
1124 static void
1125 udp6_close(struct socket *so)
1126 {
1127 	struct inpcb *inp;
1128 	struct inpcbinfo *pcbinfo;
1129 
1130 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1131 	inp = sotoinpcb(so);
1132 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1133 
1134 	INP_WLOCK(inp);
1135 #ifdef INET
1136 	if (inp->inp_vflag & INP_IPV4) {
1137 		struct pr_usrreqs *pru;
1138 		uint8_t nxt;
1139 
1140 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1141 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1142 		INP_WUNLOCK(inp);
1143 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1144 		(*pru->pru_disconnect)(so);
1145 		return;
1146 	}
1147 #endif
1148 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1149 		INP_HASH_WLOCK(pcbinfo);
1150 		in6_pcbdisconnect(inp);
1151 		inp->in6p_laddr = in6addr_any;
1152 		INP_HASH_WUNLOCK(pcbinfo);
1153 		soisdisconnected(so);
1154 	}
1155 	INP_WUNLOCK(inp);
1156 }
1157 
1158 static int
1159 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1160 {
1161 #ifdef INET
1162 	struct epoch_tracker et;
1163 #endif
1164 	struct inpcb *inp;
1165 	struct inpcbinfo *pcbinfo;
1166 	struct sockaddr_in6 *sin6;
1167 	int error;
1168 	u_char vflagsav;
1169 
1170 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1171 	inp = sotoinpcb(so);
1172 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1173 
1174 	sin6 = (struct sockaddr_in6 *)nam;
1175 	if (sin6->sin6_family != AF_INET6)
1176 		return (EAFNOSUPPORT);
1177 	if (sin6->sin6_len != sizeof(*sin6))
1178 		return (EINVAL);
1179 
1180 	/*
1181 	 * XXXRW: Need to clarify locking of v4/v6 flags.
1182 	 */
1183 	INP_WLOCK(inp);
1184 #ifdef INET
1185 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1186 		struct sockaddr_in sin;
1187 
1188 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1189 			error = EINVAL;
1190 			goto out;
1191 		}
1192 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1193 			error = EAFNOSUPPORT;
1194 			goto out;
1195 		}
1196 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1197 			error = EISCONN;
1198 			goto out;
1199 		}
1200 		in6_sin6_2_sin(&sin, sin6);
1201 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1202 		if (error != 0)
1203 			goto out;
1204 		vflagsav = inp->inp_vflag;
1205 		inp->inp_vflag |= INP_IPV4;
1206 		inp->inp_vflag &= ~INP_IPV6;
1207 		NET_EPOCH_ENTER(et);
1208 		INP_HASH_WLOCK(pcbinfo);
1209 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1210 		    td->td_ucred, true);
1211 		INP_HASH_WUNLOCK(pcbinfo);
1212 		NET_EPOCH_EXIT(et);
1213 		/*
1214 		 * If connect succeeds, mark socket as connected. If
1215 		 * connect fails and socket is unbound, reset inp_vflag
1216 		 * field.
1217 		 */
1218 		if (error == 0)
1219 			soisconnected(so);
1220 		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1221 		    inp->inp_lport == 0)
1222 			inp->inp_vflag = vflagsav;
1223 		goto out;
1224 	} else {
1225 		if ((inp->inp_vflag & INP_IPV6) == 0) {
1226 			error = EAFNOSUPPORT;
1227 			goto out;
1228 		}
1229 	}
1230 #endif
1231 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1232 		error = EISCONN;
1233 		goto out;
1234 	}
1235 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1236 	if (error != 0)
1237 		goto out;
1238 	vflagsav = inp->inp_vflag;
1239 	inp->inp_vflag &= ~INP_IPV4;
1240 	inp->inp_vflag |= INP_IPV6;
1241 	INP_HASH_WLOCK(pcbinfo);
1242 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1243 	INP_HASH_WUNLOCK(pcbinfo);
1244 	/*
1245 	 * If connect succeeds, mark socket as connected. If
1246 	 * connect fails and socket is unbound, reset inp_vflag
1247 	 * field.
1248 	 */
1249 	if (error == 0)
1250 		soisconnected(so);
1251 	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1252 	    inp->inp_lport == 0)
1253 		inp->inp_vflag = vflagsav;
1254 out:
1255 	INP_WUNLOCK(inp);
1256 	return (error);
1257 }
1258 
1259 static void
1260 udp6_detach(struct socket *so)
1261 {
1262 	struct inpcb *inp;
1263 	struct udpcb *up;
1264 
1265 	inp = sotoinpcb(so);
1266 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1267 
1268 	INP_WLOCK(inp);
1269 	up = intoudpcb(inp);
1270 	KASSERT(up != NULL, ("%s: up == NULL", __func__));
1271 	in_pcbdetach(inp);
1272 	in_pcbfree(inp);
1273 	udp_discardcb(up);
1274 }
1275 
1276 static int
1277 udp6_disconnect(struct socket *so)
1278 {
1279 	struct inpcb *inp;
1280 	struct inpcbinfo *pcbinfo;
1281 
1282 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1283 	inp = sotoinpcb(so);
1284 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1285 
1286 	INP_WLOCK(inp);
1287 #ifdef INET
1288 	if (inp->inp_vflag & INP_IPV4) {
1289 		struct pr_usrreqs *pru;
1290 		uint8_t nxt;
1291 
1292 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1293 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1294 		INP_WUNLOCK(inp);
1295 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1296 		(void)(*pru->pru_disconnect)(so);
1297 		return (0);
1298 	}
1299 #endif
1300 
1301 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1302 		INP_WUNLOCK(inp);
1303 		return (ENOTCONN);
1304 	}
1305 
1306 	INP_HASH_WLOCK(pcbinfo);
1307 	in6_pcbdisconnect(inp);
1308 	inp->in6p_laddr = in6addr_any;
1309 	INP_HASH_WUNLOCK(pcbinfo);
1310 	SOCK_LOCK(so);
1311 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1312 	SOCK_UNLOCK(so);
1313 	INP_WUNLOCK(inp);
1314 	return (0);
1315 }
1316 
1317 static int
1318 udp6_send(struct socket *so, int flags, struct mbuf *m,
1319     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1320 {
1321 	int error;
1322 
1323 	if (addr) {
1324 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1325 			error = EINVAL;
1326 			goto bad;
1327 		}
1328 		if (addr->sa_family != AF_INET6) {
1329 			error = EAFNOSUPPORT;
1330 			goto bad;
1331 		}
1332 	}
1333 
1334 	return (udp6_output(so, flags, m, addr, control, td));
1335 
1336 bad:
1337 	if (control)
1338 		m_freem(control);
1339 	m_freem(m);
1340 	return (error);
1341 }
1342 
1343 struct pr_usrreqs udp6_usrreqs = {
1344 	.pru_abort =		udp6_abort,
1345 	.pru_attach =		udp6_attach,
1346 	.pru_bind =		udp6_bind,
1347 	.pru_connect =		udp6_connect,
1348 	.pru_control =		in6_control,
1349 	.pru_detach =		udp6_detach,
1350 	.pru_disconnect =	udp6_disconnect,
1351 	.pru_peeraddr =		in6_mapped_peeraddr,
1352 	.pru_send =		udp6_send,
1353 	.pru_shutdown =		udp_shutdown,
1354 	.pru_sockaddr =		in6_mapped_sockaddr,
1355 	.pru_soreceive =	soreceive_dgram,
1356 	.pru_sosend =		sosend_dgram,
1357 	.pru_sosetlabel =	in_pcbsosetlabel,
1358 	.pru_close =		udp6_close
1359 };
1360