xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 3332f1b444d4a73238e9f59cca27bfc95fe936bd)
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 
150 	INP_LOCK_ASSERT(inp);
151 
152 	/*
153 	 * Engage the tunneling protocol.
154 	 */
155 	up = intoudpcb(inp);
156 	if (up->u_tun_func != NULL) {
157 		in_pcbref(inp);
158 		INP_RUNLOCK(inp);
159 		(*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
160 		    up->u_tun_ctx);
161 		INP_RLOCK(inp);
162 		return (in_pcbrele_rlocked(inp));
163 	}
164 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
165 	/* Check AH/ESP integrity. */
166 	if (IPSEC_ENABLED(ipv6)) {
167 		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
168 			m_freem(n);
169 			return (0);
170 		}
171 	}
172 #endif /* IPSEC */
173 #ifdef MAC
174 	if (mac_inpcb_check_deliver(inp, n) != 0) {
175 		m_freem(n);
176 		return (0);
177 	}
178 #endif
179 	opts = NULL;
180 	if (inp->inp_flags & INP_CONTROLOPTS ||
181 	    inp->inp_socket->so_options & SO_TIMESTAMP)
182 		ip6_savecontrol(inp, n, &opts);
183 	if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
184 		tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
185                         sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
186                 if (tmp_opts) {
187                         if (opts) {
188                                 tmp_opts->m_next = opts;
189                                 opts = tmp_opts;
190                         } else
191                                 opts = tmp_opts;
192                 }
193 	}
194 	m_adj(n, off + sizeof(struct udphdr));
195 
196 	so = inp->inp_socket;
197 	SOCKBUF_LOCK(&so->so_rcv);
198 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
199 	    opts) == 0) {
200 		soroverflow_locked(so);
201 		m_freem(n);
202 		if (opts)
203 			m_freem(opts);
204 		UDPSTAT_INC(udps_fullsock);
205 	} else
206 		sorwakeup_locked(so);
207 	return (0);
208 }
209 
210 int
211 udp6_input(struct mbuf **mp, int *offp, int proto)
212 {
213 	struct mbuf *m = *mp;
214 	struct ifnet *ifp;
215 	struct ip6_hdr *ip6;
216 	struct udphdr *uh;
217 	struct inpcb *inp;
218 	struct inpcbinfo *pcbinfo;
219 	struct udpcb *up;
220 	int off = *offp;
221 	int cscov_partial;
222 	int plen, ulen;
223 	struct sockaddr_in6 fromsa[2];
224 	struct m_tag *fwd_tag;
225 	uint16_t uh_sum;
226 	uint8_t nxt;
227 
228 	NET_EPOCH_ASSERT();
229 
230 	ifp = m->m_pkthdr.rcvif;
231 
232 	if (m->m_len < off + sizeof(struct udphdr)) {
233 		m = m_pullup(m, off + sizeof(struct udphdr));
234 		if (m == NULL) {
235 			IP6STAT_INC(ip6s_exthdrtoolong);
236 			*mp = NULL;
237 			return (IPPROTO_DONE);
238 		}
239 	}
240 	ip6 = mtod(m, struct ip6_hdr *);
241 	uh = (struct udphdr *)((caddr_t)ip6 + off);
242 
243 	UDPSTAT_INC(udps_ipackets);
244 
245 	/*
246 	 * Destination port of 0 is illegal, based on RFC768.
247 	 */
248 	if (uh->uh_dport == 0)
249 		goto badunlocked;
250 
251 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
252 	ulen = ntohs((u_short)uh->uh_ulen);
253 
254 	nxt = proto;
255 	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
256 	if (nxt == IPPROTO_UDPLITE) {
257 		/* Zero means checksum over the complete packet. */
258 		if (ulen == 0)
259 			ulen = plen;
260 		if (ulen == plen)
261 			cscov_partial = 0;
262 		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
263 			/* XXX: What is the right UDPLite MIB counter? */
264 			goto badunlocked;
265 		}
266 		if (uh->uh_sum == 0) {
267 			/* XXX: What is the right UDPLite MIB counter? */
268 			goto badunlocked;
269 		}
270 	} else {
271 		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
272 			UDPSTAT_INC(udps_badlen);
273 			goto badunlocked;
274 		}
275 		if (uh->uh_sum == 0) {
276 			UDPSTAT_INC(udps_nosum);
277 			/*
278 			 * dport 0 was rejected earlier so this is OK even if
279 			 * zero_checksum_port is 0 (which is its default value).
280 			 */
281 			if (ntohs(uh->uh_dport) == V_zero_checksum_port)
282 				goto skip_checksum;
283 			else
284 				goto badunlocked;
285 		}
286 	}
287 
288 	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
289 	    !cscov_partial) {
290 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
291 			uh_sum = m->m_pkthdr.csum_data;
292 		else
293 			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
294 			    m->m_pkthdr.csum_data);
295 		uh_sum ^= 0xffff;
296 	} else
297 		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
298 
299 	if (uh_sum != 0) {
300 		UDPSTAT_INC(udps_badsum);
301 		goto badunlocked;
302 	}
303 
304 skip_checksum:
305 	/*
306 	 * Construct sockaddr format source address.
307 	 */
308 	init_sin6(&fromsa[0], m, 0);
309 	fromsa[0].sin6_port = uh->uh_sport;
310 	init_sin6(&fromsa[1], m, 1);
311 	fromsa[1].sin6_port = uh->uh_dport;
312 
313 	pcbinfo = udp_get_inpcbinfo(nxt);
314 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
315 		struct inpcb *last;
316 		struct inpcbhead *pcblist;
317 		struct ip6_moptions *imo;
318 
319 		/*
320 		 * In the event that laddr should be set to the link-local
321 		 * address (this happens in RIPng), the multicast address
322 		 * specified in the received packet will not match laddr.  To
323 		 * handle this situation, matching is relaxed if the
324 		 * receiving interface is the same as one specified in the
325 		 * socket and if the destination multicast address matches
326 		 * one of the multicast groups specified in the socket.
327 		 */
328 
329 		/*
330 		 * KAME note: traditionally we dropped udpiphdr from mbuf
331 		 * here.  We need udphdr for IPsec processing so we do that
332 		 * later.
333 		 */
334 		pcblist = udp_get_pcblist(nxt);
335 		last = NULL;
336 		CK_LIST_FOREACH(inp, pcblist, inp_list) {
337 			if ((inp->inp_vflag & INP_IPV6) == 0)
338 				continue;
339 			if (inp->inp_lport != uh->uh_dport)
340 				continue;
341 			if (inp->inp_fport != 0 &&
342 			    inp->inp_fport != uh->uh_sport)
343 				continue;
344 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
345 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
346 							&ip6->ip6_dst))
347 					continue;
348 			}
349 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
350 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
351 							&ip6->ip6_src) ||
352 				    inp->inp_fport != uh->uh_sport)
353 					continue;
354 			}
355 
356 			INP_RLOCK(inp);
357 
358 			if (__predict_false(inp->inp_flags2 & INP_FREED)) {
359 				INP_RUNLOCK(inp);
360 				continue;
361 			}
362 
363 			/*
364 			 * XXXRW: Because we weren't holding either the inpcb
365 			 * or the hash lock when we checked for a match
366 			 * before, we should probably recheck now that the
367 			 * inpcb lock is (supposed to be) held.
368 			 */
369 
370 			/*
371 			 * Handle socket delivery policy for any-source
372 			 * and source-specific multicast. [RFC3678]
373 			 */
374 			imo = inp->in6p_moptions;
375 			if (imo != NULL) {
376 				struct sockaddr_in6	 mcaddr;
377 				int			 blocked;
378 
379 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
380 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
381 				mcaddr.sin6_family = AF_INET6;
382 				mcaddr.sin6_addr = ip6->ip6_dst;
383 
384 				blocked = im6o_mc_filter(imo, ifp,
385 					(struct sockaddr *)&mcaddr,
386 					(struct sockaddr *)&fromsa[0]);
387 				if (blocked != MCAST_PASS) {
388 					if (blocked == MCAST_NOTGMEMBER)
389 						IP6STAT_INC(ip6s_notmember);
390 					if (blocked == MCAST_NOTSMEMBER ||
391 					    blocked == MCAST_MUTED)
392 						UDPSTAT_INC(udps_filtermcast);
393 					INP_RUNLOCK(inp);
394 					continue;
395 				}
396 			}
397 
398 			if (last != NULL) {
399 				struct mbuf *n;
400 
401 				if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
402 				    NULL) {
403 					if (nxt == IPPROTO_UDPLITE)
404 						UDPLITE_PROBE(receive, NULL,
405 						    last, ip6, last, uh);
406 					else
407 						UDP_PROBE(receive, NULL, last,
408 						    ip6, last, uh);
409 					if (udp6_append(last, n, off,
410 					    fromsa)) {
411 						INP_RUNLOCK(inp);
412 						goto badunlocked;
413 					}
414 				}
415 				/* Release PCB lock taken on previous pass. */
416 				INP_RUNLOCK(last);
417 			}
418 			last = inp;
419 			/*
420 			 * Don't look for additional matches if this one does
421 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
422 			 * socket options set.  This heuristic avoids
423 			 * searching through all pcbs in the common case of a
424 			 * non-shared port.  It assumes that an application
425 			 * will never clear these options after setting them.
426 			 */
427 			if ((last->inp_socket->so_options &
428 			     (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
429 				break;
430 		}
431 
432 		if (last == NULL) {
433 			/*
434 			 * No matching pcb found; discard datagram.  (No need
435 			 * to send an ICMP Port Unreachable for a broadcast
436 			 * or multicast datgram.)
437 			 */
438 			UDPSTAT_INC(udps_noport);
439 			UDPSTAT_INC(udps_noportmcast);
440 			goto badunlocked;
441 		}
442 
443 		if (nxt == IPPROTO_UDPLITE)
444 			UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
445 		else
446 			UDP_PROBE(receive, NULL, last, ip6, last, uh);
447 		if (udp6_append(last, m, off, fromsa) == 0)
448 			INP_RUNLOCK(last);
449 		*mp = NULL;
450 		return (IPPROTO_DONE);
451 	}
452 	/*
453 	 * Locate pcb for datagram.
454 	 */
455 
456 	/*
457 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
458 	 */
459 	if ((m->m_flags & M_IP6_NEXTHOP) &&
460 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
461 		struct sockaddr_in6 *next_hop6;
462 
463 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
464 
465 		/*
466 		 * Transparently forwarded. Pretend to be the destination.
467 		 * Already got one like this?
468 		 */
469 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
470 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
471 		    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
472 		if (!inp) {
473 			/*
474 			 * It's new.  Try to find the ambushing socket.
475 			 * Because we've rewritten the destination address,
476 			 * any hardware-generated hash is ignored.
477 			 */
478 			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
479 			    uh->uh_sport, &next_hop6->sin6_addr,
480 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
481 			    uh->uh_dport, INPLOOKUP_WILDCARD |
482 			    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
483 		}
484 		/* Remove the tag from the packet. We don't need it anymore. */
485 		m_tag_delete(m, fwd_tag);
486 		m->m_flags &= ~M_IP6_NEXTHOP;
487 	} else
488 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
489 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
490 		    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
491 		    m->m_pkthdr.rcvif, m);
492 	if (inp == NULL) {
493 		if (V_udp_log_in_vain) {
494 			char ip6bufs[INET6_ADDRSTRLEN];
495 			char ip6bufd[INET6_ADDRSTRLEN];
496 
497 			log(LOG_INFO,
498 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
499 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
500 			    ntohs(uh->uh_dport),
501 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
502 			    ntohs(uh->uh_sport));
503 		}
504 		if (nxt == IPPROTO_UDPLITE)
505 			UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
506 		else
507 			UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
508 		UDPSTAT_INC(udps_noport);
509 		if (m->m_flags & M_MCAST) {
510 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
511 			UDPSTAT_INC(udps_noportmcast);
512 			goto badunlocked;
513 		}
514 		if (V_udp_blackhole && (V_udp_blackhole_local ||
515 		    !in6_localaddr(&ip6->ip6_src)))
516 			goto badunlocked;
517 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
518 		*mp = NULL;
519 		return (IPPROTO_DONE);
520 	}
521 	INP_RLOCK_ASSERT(inp);
522 	up = intoudpcb(inp);
523 	if (cscov_partial) {
524 		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
525 			INP_RUNLOCK(inp);
526 			m_freem(m);
527 			*mp = NULL;
528 			return (IPPROTO_DONE);
529 		}
530 	}
531 	if (nxt == IPPROTO_UDPLITE)
532 		UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
533 	else
534 		UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
535 	if (udp6_append(inp, m, off, fromsa) == 0)
536 		INP_RUNLOCK(inp);
537 	*mp = NULL;
538 	return (IPPROTO_DONE);
539 
540 badunlocked:
541 	m_freem(m);
542 	*mp = NULL;
543 	return (IPPROTO_DONE);
544 }
545 
546 static void
547 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
548     struct inpcbinfo *pcbinfo)
549 {
550 	struct udphdr uh;
551 	struct ip6_hdr *ip6;
552 	struct mbuf *m;
553 	int off = 0;
554 	struct ip6ctlparam *ip6cp = NULL;
555 	const struct sockaddr_in6 *sa6_src = NULL;
556 	void *cmdarg;
557 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
558 	struct udp_portonly {
559 		u_int16_t uh_sport;
560 		u_int16_t uh_dport;
561 	} *uhp;
562 
563 	if (sa->sa_family != AF_INET6 ||
564 	    sa->sa_len != sizeof(struct sockaddr_in6))
565 		return;
566 
567 	if ((unsigned)cmd >= PRC_NCMDS)
568 		return;
569 	if (PRC_IS_REDIRECT(cmd))
570 		notify = in6_rtchange, d = NULL;
571 	else if (cmd == PRC_HOSTDEAD)
572 		d = NULL;
573 	else if (inet6ctlerrmap[cmd] == 0)
574 		return;
575 
576 	/* if the parameter is from icmp6, decode it. */
577 	if (d != NULL) {
578 		ip6cp = (struct ip6ctlparam *)d;
579 		m = ip6cp->ip6c_m;
580 		ip6 = ip6cp->ip6c_ip6;
581 		off = ip6cp->ip6c_off;
582 		cmdarg = ip6cp->ip6c_cmdarg;
583 		sa6_src = ip6cp->ip6c_src;
584 	} else {
585 		m = NULL;
586 		ip6 = NULL;
587 		cmdarg = NULL;
588 		sa6_src = &sa6_any;
589 	}
590 
591 	if (ip6) {
592 		/*
593 		 * XXX: We assume that when IPV6 is non NULL,
594 		 * M and OFF are valid.
595 		 */
596 
597 		/* Check if we can safely examine src and dst ports. */
598 		if (m->m_pkthdr.len < off + sizeof(*uhp))
599 			return;
600 
601 		bzero(&uh, sizeof(uh));
602 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
603 
604 		if (!PRC_IS_REDIRECT(cmd)) {
605 			/* Check to see if its tunneled */
606 			struct inpcb *inp;
607 			inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
608 			    uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
609 			    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
610 			    m->m_pkthdr.rcvif, m);
611 			if (inp != NULL) {
612 				struct udpcb *up;
613 
614 				up = intoudpcb(inp);
615 				if (up->u_icmp_func) {
616 					/* Yes it is. */
617 					INP_RUNLOCK(inp);
618 					(*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
619 					      d, up->u_tun_ctx);
620 					return;
621 				} else {
622 					/* Can't find it. */
623 					INP_RUNLOCK(inp);
624 				}
625 			}
626 		}
627 		(void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
628 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
629 		    cmdarg, notify);
630 	} else
631 		(void)in6_pcbnotify(pcbinfo, sa, 0,
632 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
633 }
634 
635 void
636 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
637 {
638 
639 	return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
640 }
641 
642 void
643 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
644 {
645 
646 	return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
647 }
648 
649 static int
650 udp6_getcred(SYSCTL_HANDLER_ARGS)
651 {
652 	struct xucred xuc;
653 	struct sockaddr_in6 addrs[2];
654 	struct epoch_tracker et;
655 	struct inpcb *inp;
656 	int error;
657 
658 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
659 	if (error)
660 		return (error);
661 
662 	if (req->newlen != sizeof(addrs))
663 		return (EINVAL);
664 	if (req->oldlen != sizeof(struct xucred))
665 		return (EINVAL);
666 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
667 	if (error)
668 		return (error);
669 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
670 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
671 		return (error);
672 	}
673 	NET_EPOCH_ENTER(et);
674 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
675 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
676 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
677 	NET_EPOCH_EXIT(et);
678 	if (inp != NULL) {
679 		INP_RLOCK_ASSERT(inp);
680 		if (inp->inp_socket == NULL)
681 			error = ENOENT;
682 		if (error == 0)
683 			error = cr_canseesocket(req->td->td_ucred,
684 			    inp->inp_socket);
685 		if (error == 0)
686 			cru2x(inp->inp_cred, &xuc);
687 		INP_RUNLOCK(inp);
688 	} else
689 		error = ENOENT;
690 	if (error == 0)
691 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
692 	return (error);
693 }
694 
695 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
696     CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
697     0, 0, udp6_getcred, "S,xucred",
698     "Get the xucred of a UDP6 connection");
699 
700 static int
701 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
702     struct sockaddr *addr6, struct mbuf *control, struct thread *td)
703 {
704 	struct inpcb *inp;
705 	struct ip6_hdr *ip6;
706 	struct udphdr *udp6;
707 	struct in6_addr *laddr, *faddr, in6a;
708 	struct ip6_pktopts *optp, opt;
709 	struct sockaddr_in6 *sin6, tmp;
710 	struct epoch_tracker et;
711 	int cscov_partial, error, flags, hlen, scope_ambiguous;
712 	u_int32_t ulen, plen;
713 	uint16_t cscov;
714 	u_short fport;
715 	uint8_t nxt;
716 
717 	/* addr6 has been validated in udp6_send(). */
718 	sin6 = (struct sockaddr_in6 *)addr6;
719 
720 	/*
721 	 * In contrast to to IPv4 we do not validate the max. packet length
722 	 * here due to IPv6 Jumbograms (RFC2675).
723 	 */
724 
725 	scope_ambiguous = 0;
726 	if (sin6) {
727 		/* Protect *addr6 from overwrites. */
728 		tmp = *sin6;
729 		sin6 = &tmp;
730 
731 		/*
732 		 * Application should provide a proper zone ID or the use of
733 		 * default zone IDs should be enabled.  Unfortunately, some
734 		 * applications do not behave as it should, so we need a
735 		 * workaround.  Even if an appropriate ID is not determined,
736 		 * we'll see if we can determine the outgoing interface.  If we
737 		 * can, determine the zone ID based on the interface below.
738 		 */
739 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
740 			scope_ambiguous = 1;
741 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
742 			if (control)
743 				m_freem(control);
744 			m_freem(m);
745 			return (error);
746 		}
747 	}
748 
749 	inp = sotoinpcb(so);
750 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
751 	/*
752 	 * In the following cases we want a write lock on the inp for either
753 	 * local operations or for possible route cache updates in the IPv6
754 	 * output path:
755 	 * - on connected sockets (sin6 is NULL) for route cache updates,
756 	 * - when we are not bound to an address and source port (it is
757 	 *   in6_pcbsetport() which will require the write lock).
758 	 *
759 	 * We check the inp fields before actually locking the inp, so
760 	 * here exists a race, and we may WLOCK the inp and end with already
761 	 * bound one by other thread. This is fine.
762 	 */
763 	if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
764 	    inp->inp_lport == 0))
765 		INP_WLOCK(inp);
766 	else
767 		INP_RLOCK(inp);
768 
769 	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
770 	    IPPROTO_UDP : IPPROTO_UDPLITE;
771 
772 #ifdef INET
773 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
774 		int hasv4addr;
775 
776 		if (sin6 == NULL)
777 			hasv4addr = (inp->inp_vflag & INP_IPV4);
778 		else
779 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
780 			    ? 1 : 0;
781 		if (hasv4addr) {
782 			struct pr_usrreqs *pru;
783 
784 			/*
785 			 * XXXRW: We release UDP-layer locks before calling
786 			 * udp_send() in order to avoid recursion.  However,
787 			 * this does mean there is a short window where inp's
788 			 * fields are unstable.  Could this lead to a
789 			 * potential race in which the factors causing us to
790 			 * select the UDPv4 output routine are invalidated?
791 			 */
792 			INP_UNLOCK(inp);
793 			if (sin6)
794 				in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
795 			pru = inetsw[ip_protox[nxt]].pr_usrreqs;
796 			/* addr will just be freed in sendit(). */
797 			return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
798 			    (struct sockaddr *)sin6, control, td));
799 		}
800 	} else
801 #endif
802 	if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
803 		/*
804 		 * Given this is either an IPv6-only socket or no INET is
805 		 * supported we will fail the send if the given destination
806 		 * address is a v4mapped address.
807 		 *
808 		 * XXXGL: do we leak m and control?
809 		 */
810 		INP_UNLOCK(inp);
811 		return (EINVAL);
812 	}
813 
814 	NET_EPOCH_ENTER(et);
815 	if (control) {
816 		if ((error = ip6_setpktopts(control, &opt,
817 		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
818 			goto release;
819 		}
820 		optp = &opt;
821 	} else
822 		optp = inp->in6p_outputopts;
823 
824 	if (sin6) {
825 		/*
826 		 * Since we saw no essential reason for calling in_pcbconnect,
827 		 * we get rid of such kind of logic, and call in6_selectsrc
828 		 * and in6_pcbsetport in order to fill in the local address
829 		 * and the local port.
830 		 */
831 		if (sin6->sin6_port == 0) {
832 			error = EADDRNOTAVAIL;
833 			goto release;
834 		}
835 
836 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
837 			/* how about ::ffff:0.0.0.0 case? */
838 			error = EISCONN;
839 			goto release;
840 		}
841 
842 		/*
843 		 * Given we handle the v4mapped case in the INET block above
844 		 * assert here that it must not happen anymore.
845 		 */
846 		KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
847 		    ("%s: sin6(%p)->sin6_addr is v4mapped which we "
848 		    "should have handled.", __func__, sin6));
849 
850 		/* This only requires read-locking. */
851 		error = in6_selectsrc_socket(sin6, optp, inp,
852 		    td->td_ucred, scope_ambiguous, &in6a, NULL);
853 		if (error)
854 			goto release;
855 		laddr = &in6a;
856 
857 		if (inp->inp_lport == 0) {
858 			struct inpcbinfo *pcbinfo;
859 
860 			INP_WLOCK_ASSERT(inp);
861 
862 			pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
863 			INP_HASH_WLOCK(pcbinfo);
864 			error = in6_pcbsetport(laddr, inp, td->td_ucred);
865 			INP_HASH_WUNLOCK(pcbinfo);
866 			if (error != 0) {
867 				/* Undo an address bind that may have occurred. */
868 				inp->in6p_laddr = in6addr_any;
869 				goto release;
870 			}
871 		}
872 		faddr = &sin6->sin6_addr;
873 		fport = sin6->sin6_port; /* allow 0 port */
874 
875 	} else {
876 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
877 			error = ENOTCONN;
878 			goto release;
879 		}
880 		laddr = &inp->in6p_laddr;
881 		faddr = &inp->in6p_faddr;
882 		fport = inp->inp_fport;
883 	}
884 
885 	ulen = m->m_pkthdr.len;
886 	plen = sizeof(struct udphdr) + ulen;
887 	hlen = sizeof(struct ip6_hdr);
888 
889 	/*
890 	 * Calculate data length and get a mbuf
891 	 * for UDP and IP6 headers.
892 	 */
893 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
894 	if (m == NULL) {
895 		error = ENOBUFS;
896 		goto release;
897 	}
898 
899 	/*
900 	 * Stuff checksum and output datagram.
901 	 */
902 	cscov = cscov_partial = 0;
903 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
904 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
905 	udp6->uh_dport = fport;
906 	if (nxt == IPPROTO_UDPLITE) {
907 		struct udpcb *up;
908 
909 		up = intoudpcb(inp);
910 		cscov = up->u_txcslen;
911 		if (cscov >= plen)
912 			cscov = 0;
913 		udp6->uh_ulen = htons(cscov);
914 		/*
915 		 * For UDP-Lite, checksum coverage length of zero means
916 		 * the entire UDPLite packet is covered by the checksum.
917 		 */
918 		cscov_partial = (cscov == 0) ? 0 : 1;
919 	} else if (plen <= 0xffff)
920 		udp6->uh_ulen = htons((u_short)plen);
921 	else
922 		udp6->uh_ulen = 0;
923 	udp6->uh_sum = 0;
924 
925 	ip6 = mtod(m, struct ip6_hdr *);
926 	ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
927 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
928 	ip6->ip6_vfc	|= IPV6_VERSION;
929 	ip6->ip6_plen	= htons((u_short)plen);
930 	ip6->ip6_nxt	= nxt;
931 	ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
932 	ip6->ip6_src	= *laddr;
933 	ip6->ip6_dst	= *faddr;
934 
935 #ifdef MAC
936 	mac_inpcb_create_mbuf(inp, m);
937 #endif
938 
939 	if (cscov_partial) {
940 		if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
941 		    sizeof(struct ip6_hdr), plen, cscov)) == 0)
942 			udp6->uh_sum = 0xffff;
943 	} else {
944 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
945 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
946 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
947 	}
948 
949 	flags = 0;
950 #if defined(ROUTE_MPATH) || defined(RSS)
951 	if (CALC_FLOWID_OUTBOUND_SENDTO) {
952 		uint32_t hash_type, hash_val;
953 		uint8_t pr;
954 
955 		pr = inp->inp_socket->so_proto->pr_protocol;
956 
957 		hash_val = fib6_calc_packet_hash(laddr, faddr,
958 		    inp->inp_lport, fport, pr, &hash_type);
959 		m->m_pkthdr.flowid = hash_val;
960 		M_HASHTYPE_SET(m, hash_type);
961 	}
962 	/* do not use inp flowid */
963 	flags |= IP_NODEFAULTFLOWID;
964 #endif
965 
966 	UDPSTAT_INC(udps_opackets);
967 	if (nxt == IPPROTO_UDPLITE)
968 		UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
969 	else
970 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
971 	error = ip6_output(m, optp,
972 	    INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
973 	    inp->in6p_moptions, NULL, inp);
974 	INP_UNLOCK(inp);
975 	NET_EPOCH_EXIT(et);
976 
977 	if (control) {
978 		ip6_clearpktopts(&opt, -1);
979 		m_freem(control);
980 	}
981 	return (error);
982 
983 release:
984 	INP_UNLOCK(inp);
985 	NET_EPOCH_EXIT(et);
986 	if (control) {
987 		ip6_clearpktopts(&opt, -1);
988 		m_freem(control);
989 	}
990 	m_freem(m);
991 
992 	return (error);
993 }
994 
995 static void
996 udp6_abort(struct socket *so)
997 {
998 	struct inpcb *inp;
999 	struct inpcbinfo *pcbinfo;
1000 
1001 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1002 	inp = sotoinpcb(so);
1003 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1004 
1005 	INP_WLOCK(inp);
1006 #ifdef INET
1007 	if (inp->inp_vflag & INP_IPV4) {
1008 		struct pr_usrreqs *pru;
1009 		uint8_t nxt;
1010 
1011 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1012 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1013 		INP_WUNLOCK(inp);
1014 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1015 		(*pru->pru_abort)(so);
1016 		return;
1017 	}
1018 #endif
1019 
1020 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1021 		INP_HASH_WLOCK(pcbinfo);
1022 		in6_pcbdisconnect(inp);
1023 		inp->in6p_laddr = in6addr_any;
1024 		INP_HASH_WUNLOCK(pcbinfo);
1025 		soisdisconnected(so);
1026 	}
1027 	INP_WUNLOCK(inp);
1028 }
1029 
1030 static int
1031 udp6_attach(struct socket *so, int proto, struct thread *td)
1032 {
1033 	struct inpcb *inp;
1034 	struct inpcbinfo *pcbinfo;
1035 	int error;
1036 
1037 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1038 	inp = sotoinpcb(so);
1039 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1040 
1041 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1042 		error = soreserve(so, udp_sendspace, udp_recvspace);
1043 		if (error)
1044 			return (error);
1045 	}
1046 	INP_INFO_WLOCK(pcbinfo);
1047 	error = in_pcballoc(so, pcbinfo);
1048 	if (error) {
1049 		INP_INFO_WUNLOCK(pcbinfo);
1050 		return (error);
1051 	}
1052 	inp = (struct inpcb *)so->so_pcb;
1053 	inp->inp_vflag |= INP_IPV6;
1054 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1055 		inp->inp_vflag |= INP_IPV4;
1056 	inp->in6p_hops = -1;	/* use kernel default */
1057 	inp->in6p_cksum = -1;	/* just to be sure */
1058 	/*
1059 	 * XXX: ugly!!
1060 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1061 	 * because the socket may be bound to an IPv6 wildcard address,
1062 	 * which may match an IPv4-mapped IPv6 address.
1063 	 */
1064 	inp->inp_ip_ttl = V_ip_defttl;
1065 
1066 	error = udp_newudpcb(inp);
1067 	if (error) {
1068 		in_pcbdetach(inp);
1069 		in_pcbfree(inp);
1070 		INP_INFO_WUNLOCK(pcbinfo);
1071 		return (error);
1072 	}
1073 	INP_WUNLOCK(inp);
1074 	INP_INFO_WUNLOCK(pcbinfo);
1075 	return (0);
1076 }
1077 
1078 static int
1079 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1080 {
1081 	struct inpcb *inp;
1082 	struct inpcbinfo *pcbinfo;
1083 	int error;
1084 	u_char vflagsav;
1085 
1086 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1087 	inp = sotoinpcb(so);
1088 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1089 
1090 	if (nam->sa_family != AF_INET6)
1091 		return (EAFNOSUPPORT);
1092 	if (nam->sa_len != sizeof(struct sockaddr_in6))
1093 		return (EINVAL);
1094 
1095 	INP_WLOCK(inp);
1096 	INP_HASH_WLOCK(pcbinfo);
1097 	vflagsav = inp->inp_vflag;
1098 	inp->inp_vflag &= ~INP_IPV4;
1099 	inp->inp_vflag |= INP_IPV6;
1100 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1101 		struct sockaddr_in6 *sin6_p;
1102 
1103 		sin6_p = (struct sockaddr_in6 *)nam;
1104 
1105 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1106 			inp->inp_vflag |= INP_IPV4;
1107 #ifdef INET
1108 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1109 			struct sockaddr_in sin;
1110 
1111 			in6_sin6_2_sin(&sin, sin6_p);
1112 			inp->inp_vflag |= INP_IPV4;
1113 			inp->inp_vflag &= ~INP_IPV6;
1114 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
1115 			    td->td_ucred);
1116 			goto out;
1117 		}
1118 #endif
1119 	}
1120 
1121 	error = in6_pcbbind(inp, nam, td->td_ucred);
1122 #ifdef INET
1123 out:
1124 #endif
1125 	if (error != 0)
1126 		inp->inp_vflag = vflagsav;
1127 	INP_HASH_WUNLOCK(pcbinfo);
1128 	INP_WUNLOCK(inp);
1129 	return (error);
1130 }
1131 
1132 static void
1133 udp6_close(struct socket *so)
1134 {
1135 	struct inpcb *inp;
1136 	struct inpcbinfo *pcbinfo;
1137 
1138 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1139 	inp = sotoinpcb(so);
1140 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1141 
1142 	INP_WLOCK(inp);
1143 #ifdef INET
1144 	if (inp->inp_vflag & INP_IPV4) {
1145 		struct pr_usrreqs *pru;
1146 		uint8_t nxt;
1147 
1148 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1149 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1150 		INP_WUNLOCK(inp);
1151 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1152 		(*pru->pru_disconnect)(so);
1153 		return;
1154 	}
1155 #endif
1156 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1157 		INP_HASH_WLOCK(pcbinfo);
1158 		in6_pcbdisconnect(inp);
1159 		inp->in6p_laddr = in6addr_any;
1160 		INP_HASH_WUNLOCK(pcbinfo);
1161 		soisdisconnected(so);
1162 	}
1163 	INP_WUNLOCK(inp);
1164 }
1165 
1166 static int
1167 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1168 {
1169 #ifdef INET
1170 	struct epoch_tracker et;
1171 #endif
1172 	struct inpcb *inp;
1173 	struct inpcbinfo *pcbinfo;
1174 	struct sockaddr_in6 *sin6;
1175 	int error;
1176 	u_char vflagsav;
1177 
1178 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1179 	inp = sotoinpcb(so);
1180 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1181 
1182 	sin6 = (struct sockaddr_in6 *)nam;
1183 	if (sin6->sin6_family != AF_INET6)
1184 		return (EAFNOSUPPORT);
1185 	if (sin6->sin6_len != sizeof(*sin6))
1186 		return (EINVAL);
1187 
1188 	/*
1189 	 * XXXRW: Need to clarify locking of v4/v6 flags.
1190 	 */
1191 	INP_WLOCK(inp);
1192 #ifdef INET
1193 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1194 		struct sockaddr_in sin;
1195 
1196 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1197 			error = EINVAL;
1198 			goto out;
1199 		}
1200 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1201 			error = EAFNOSUPPORT;
1202 			goto out;
1203 		}
1204 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1205 			error = EISCONN;
1206 			goto out;
1207 		}
1208 		in6_sin6_2_sin(&sin, sin6);
1209 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1210 		if (error != 0)
1211 			goto out;
1212 		vflagsav = inp->inp_vflag;
1213 		inp->inp_vflag |= INP_IPV4;
1214 		inp->inp_vflag &= ~INP_IPV6;
1215 		NET_EPOCH_ENTER(et);
1216 		INP_HASH_WLOCK(pcbinfo);
1217 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1218 		    td->td_ucred);
1219 		INP_HASH_WUNLOCK(pcbinfo);
1220 		NET_EPOCH_EXIT(et);
1221 		/*
1222 		 * If connect succeeds, mark socket as connected. If
1223 		 * connect fails and socket is unbound, reset inp_vflag
1224 		 * field.
1225 		 */
1226 		if (error == 0)
1227 			soisconnected(so);
1228 		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1229 		    inp->inp_lport == 0)
1230 			inp->inp_vflag = vflagsav;
1231 		goto out;
1232 	} else {
1233 		if ((inp->inp_vflag & INP_IPV6) == 0) {
1234 			error = EAFNOSUPPORT;
1235 			goto out;
1236 		}
1237 	}
1238 #endif
1239 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1240 		error = EISCONN;
1241 		goto out;
1242 	}
1243 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1244 	if (error != 0)
1245 		goto out;
1246 	vflagsav = inp->inp_vflag;
1247 	inp->inp_vflag &= ~INP_IPV4;
1248 	inp->inp_vflag |= INP_IPV6;
1249 	INP_HASH_WLOCK(pcbinfo);
1250 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1251 	INP_HASH_WUNLOCK(pcbinfo);
1252 	/*
1253 	 * If connect succeeds, mark socket as connected. If
1254 	 * connect fails and socket is unbound, reset inp_vflag
1255 	 * field.
1256 	 */
1257 	if (error == 0)
1258 		soisconnected(so);
1259 	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1260 	    inp->inp_lport == 0)
1261 		inp->inp_vflag = vflagsav;
1262 out:
1263 	INP_WUNLOCK(inp);
1264 	return (error);
1265 }
1266 
1267 static void
1268 udp6_detach(struct socket *so)
1269 {
1270 	struct inpcb *inp;
1271 	struct inpcbinfo *pcbinfo;
1272 	struct udpcb *up;
1273 
1274 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1275 	inp = sotoinpcb(so);
1276 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1277 
1278 	INP_INFO_WLOCK(pcbinfo);
1279 	INP_WLOCK(inp);
1280 	up = intoudpcb(inp);
1281 	KASSERT(up != NULL, ("%s: up == NULL", __func__));
1282 	in_pcbdetach(inp);
1283 	in_pcbfree(inp);
1284 	INP_INFO_WUNLOCK(pcbinfo);
1285 	udp_discardcb(up);
1286 }
1287 
1288 static int
1289 udp6_disconnect(struct socket *so)
1290 {
1291 	struct inpcb *inp;
1292 	struct inpcbinfo *pcbinfo;
1293 
1294 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1295 	inp = sotoinpcb(so);
1296 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1297 
1298 	INP_WLOCK(inp);
1299 #ifdef INET
1300 	if (inp->inp_vflag & INP_IPV4) {
1301 		struct pr_usrreqs *pru;
1302 		uint8_t nxt;
1303 
1304 		nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1305 		    IPPROTO_UDP : IPPROTO_UDPLITE;
1306 		INP_WUNLOCK(inp);
1307 		pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1308 		(void)(*pru->pru_disconnect)(so);
1309 		return (0);
1310 	}
1311 #endif
1312 
1313 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1314 		INP_WUNLOCK(inp);
1315 		return (ENOTCONN);
1316 	}
1317 
1318 	INP_HASH_WLOCK(pcbinfo);
1319 	in6_pcbdisconnect(inp);
1320 	inp->in6p_laddr = in6addr_any;
1321 	INP_HASH_WUNLOCK(pcbinfo);
1322 	SOCK_LOCK(so);
1323 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1324 	SOCK_UNLOCK(so);
1325 	INP_WUNLOCK(inp);
1326 	return (0);
1327 }
1328 
1329 static int
1330 udp6_send(struct socket *so, int flags, struct mbuf *m,
1331     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1332 {
1333 	int error;
1334 
1335 	if (addr) {
1336 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1337 			error = EINVAL;
1338 			goto bad;
1339 		}
1340 		if (addr->sa_family != AF_INET6) {
1341 			error = EAFNOSUPPORT;
1342 			goto bad;
1343 		}
1344 	}
1345 
1346 	return (udp6_output(so, flags, m, addr, control, td));
1347 
1348 bad:
1349 	if (control)
1350 		m_freem(control);
1351 	m_freem(m);
1352 	return (error);
1353 }
1354 
1355 struct pr_usrreqs udp6_usrreqs = {
1356 	.pru_abort =		udp6_abort,
1357 	.pru_attach =		udp6_attach,
1358 	.pru_bind =		udp6_bind,
1359 	.pru_connect =		udp6_connect,
1360 	.pru_control =		in6_control,
1361 	.pru_detach =		udp6_detach,
1362 	.pru_disconnect =	udp6_disconnect,
1363 	.pru_peeraddr =		in6_mapped_peeraddr,
1364 	.pru_send =		udp6_send,
1365 	.pru_shutdown =		udp_shutdown,
1366 	.pru_sockaddr =		in6_mapped_sockaddr,
1367 	.pru_soreceive =	soreceive_dgram,
1368 	.pru_sosend =		sosend_dgram,
1369 	.pru_sosetlabel =	in_pcbsosetlabel,
1370 	.pru_close =		udp6_close
1371 };
1372