xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 7b71f57f4e514a2ab7308ce4147e14d90e099ad0)
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 "opt_inet.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
73 #include "opt_route.h"
74 #include "opt_rss.h"
75 
76 #include <sys/param.h>
77 #include <sys/domain.h>
78 #include <sys/jail.h>
79 #include <sys/kernel.h>
80 #include <sys/lock.h>
81 #include <sys/mbuf.h>
82 #include <sys/priv.h>
83 #include <sys/proc.h>
84 #include <sys/protosw.h>
85 #include <sys/sdt.h>
86 #include <sys/signalvar.h>
87 #include <sys/socket.h>
88 #include <sys/socketvar.h>
89 #include <sys/sx.h>
90 #include <sys/sysctl.h>
91 #include <sys/syslog.h>
92 #include <sys/systm.h>
93 
94 #include <net/if.h>
95 #include <net/if_var.h>
96 #include <net/if_types.h>
97 #include <net/route.h>
98 #include <net/rss_config.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_kdtrace.h>
102 #include <netinet/in_pcb.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/in_var.h>
105 #include <netinet/ip.h>
106 #include <netinet/ip6.h>
107 #include <netinet/icmp6.h>
108 #include <netinet/ip_var.h>
109 #include <netinet/udp.h>
110 #include <netinet/udp_var.h>
111 #include <netinet/udplite.h>
112 
113 #include <netinet6/ip6_var.h>
114 #include <netinet6/in6_fib.h>
115 #include <netinet6/in6_pcb.h>
116 #include <netinet6/in6_rss.h>
117 #include <netinet6/udp6_var.h>
118 #include <netinet6/scope6_var.h>
119 
120 #include <netipsec/ipsec_support.h>
121 
122 #include <security/mac/mac_framework.h>
123 
124 VNET_DEFINE(int, zero_checksum_port) = 0;
125 #define	V_zero_checksum_port	VNET(zero_checksum_port)
126 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
127     &VNET_NAME(zero_checksum_port), 0,
128     "Zero UDP checksum allowed for traffic to/from this port.");
129 
130 /*
131  * UDP protocol implementation.
132  * Per RFC 768, August, 1980.
133  */
134 
135 static void		udp6_detach(struct socket *so);
136 
137 static int
udp6_append(struct inpcb * inp,struct mbuf * n,int off,struct sockaddr_in6 * fromsa)138 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
139     struct sockaddr_in6 *fromsa)
140 {
141 	struct socket *so;
142 	struct mbuf *opts = NULL, *tmp_opts;
143 	struct udpcb *up;
144 
145 	INP_LOCK_ASSERT(inp);
146 
147 	/*
148 	 * Engage the tunneling protocol.
149 	 */
150 	up = intoudpcb(inp);
151 	if (up->u_tun_func != NULL) {
152 		bool filtered;
153 
154 		in_pcbref(inp);
155 		INP_RUNLOCK(inp);
156 		filtered = (*up->u_tun_func)(n, off, inp,
157 		    (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
158 		INP_RLOCK(inp);
159 		if (in_pcbrele_rlocked(inp))
160 			return (1);
161 		if (filtered) {
162 			INP_RUNLOCK(inp);
163 			return (1);
164 		}
165 	}
166 
167 	off += sizeof(struct udphdr);
168 
169 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
170 	/* Check AH/ESP integrity. */
171 	if (IPSEC_ENABLED(ipv6)) {
172 		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
173 			m_freem(n);
174 			return (0);
175 		}
176 
177 		/* IPSec UDP encaps. */
178 		if ((up->u_flags & UF_ESPINUDP) != 0 &&
179 		    UDPENCAP_INPUT(ipv6, n, off, AF_INET6) != 0) {
180 			return (0); /* Consumed. */
181 		}
182 	}
183 #endif /* IPSEC */
184 #ifdef MAC
185 	if (mac_inpcb_check_deliver(inp, n) != 0) {
186 		m_freem(n);
187 		return (0);
188 	}
189 #endif
190 	opts = NULL;
191 	if (inp->inp_flags & INP_CONTROLOPTS ||
192 	    inp->inp_socket->so_options & SO_TIMESTAMP)
193 		ip6_savecontrol(inp, n, &opts);
194 	if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
195 		tmp_opts = sbcreatecontrol(&fromsa[1],
196 		    sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
197 		    IPPROTO_IPV6, M_NOWAIT);
198                 if (tmp_opts) {
199                         if (opts) {
200                                 tmp_opts->m_next = opts;
201                                 opts = tmp_opts;
202                         } else
203                                 opts = tmp_opts;
204                 }
205 	}
206 	m_adj(n, off);
207 
208 	so = inp->inp_socket;
209 	SOCKBUF_LOCK(&so->so_rcv);
210 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
211 	    opts) == 0) {
212 		soroverflow_locked(so);
213 		m_freem(n);
214 		if (opts)
215 			m_freem(opts);
216 		UDPSTAT_INC(udps_fullsock);
217 	} else
218 		sorwakeup_locked(so);
219 	return (0);
220 }
221 
222 struct udp6_multi_match_ctx {
223 	struct ip6_hdr *ip6;
224 	struct udphdr *uh;
225 };
226 
227 static bool
udp6_multi_match(const struct inpcb * inp,void * v)228 udp6_multi_match(const struct inpcb *inp, void *v)
229 {
230 	struct udp6_multi_match_ctx *ctx = v;
231 
232 	if ((inp->inp_vflag & INP_IPV6) == 0)
233 		return(false);
234 	if (inp->inp_lport != ctx->uh->uh_dport)
235 		return(false);
236 	if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
237 		return(false);
238 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
239 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
240 		return (false);
241 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
242 	    (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
243 	    inp->inp_fport != ctx->uh->uh_sport))
244 		return (false);
245 
246 	return (true);
247 }
248 
249 static int
udp6_multi_input(struct mbuf * m,int off,int proto,struct sockaddr_in6 * fromsa)250 udp6_multi_input(struct mbuf *m, int off, int proto,
251     struct sockaddr_in6 *fromsa)
252 {
253 	struct udp6_multi_match_ctx ctx;
254 	struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
255 	    INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
256 	struct inpcb *inp;
257 	struct ip6_moptions *imo;
258 	struct mbuf *n;
259 	int appends = 0;
260 
261 	/*
262 	 * In the event that laddr should be set to the link-local
263 	 * address (this happens in RIPng), the multicast address
264 	 * specified in the received packet will not match laddr.  To
265 	 * handle this situation, matching is relaxed if the
266 	 * receiving interface is the same as one specified in the
267 	 * socket and if the destination multicast address matches
268 	 * one of the multicast groups specified in the socket.
269 	 */
270 
271 	/*
272 	 * KAME note: traditionally we dropped udpiphdr from mbuf
273 	 * here.  We need udphdr for IPsec processing so we do that
274 	 * later.
275 	 */
276 	ctx.ip6 = mtod(m, struct ip6_hdr *);
277 	ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
278 	while ((inp = inp_next(&inpi)) != NULL) {
279 		INP_RLOCK_ASSERT(inp);
280 		/*
281 		 * XXXRW: Because we weren't holding either the inpcb
282 		 * or the hash lock when we checked for a match
283 		 * before, we should probably recheck now that the
284 		 * inpcb lock is (supposed to be) held.
285 		 */
286 		/*
287 		 * Handle socket delivery policy for any-source
288 		 * and source-specific multicast. [RFC3678]
289 		 */
290 		if ((imo = inp->in6p_moptions) != NULL) {
291 			struct sockaddr_in6	 mcaddr;
292 			int			 blocked;
293 
294 			bzero(&mcaddr, sizeof(struct sockaddr_in6));
295 			mcaddr.sin6_len = sizeof(struct sockaddr_in6);
296 			mcaddr.sin6_family = AF_INET6;
297 			mcaddr.sin6_addr = ctx.ip6->ip6_dst;
298 
299 			blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
300 				(struct sockaddr *)&mcaddr,
301 				(struct sockaddr *)&fromsa[0]);
302 			if (blocked != MCAST_PASS) {
303 				if (blocked == MCAST_NOTGMEMBER)
304 					IP6STAT_INC(ip6s_notmember);
305 				if (blocked == MCAST_NOTSMEMBER ||
306 				    blocked == MCAST_MUTED)
307 					UDPSTAT_INC(udps_filtermcast);
308 				continue;
309 			}
310 		}
311 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
312 			if (proto == IPPROTO_UDPLITE)
313 				UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
314 				    inp, ctx.uh);
315 			else
316 				UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
317 				    ctx.uh);
318 			if (udp6_append(inp, n, off, fromsa)) {
319 				break;
320 			} else
321 				appends++;
322 		}
323 		/*
324 		 * Don't look for additional matches if this one does
325 		 * not have either the SO_REUSEPORT or SO_REUSEADDR
326 		 * socket options set.  This heuristic avoids
327 		 * searching through all pcbs in the common case of a
328 		 * non-shared port.  It assumes that an application
329 		 * will never clear these options after setting them.
330 		 */
331 		if ((inp->inp_socket->so_options &
332 		     (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
333 			INP_RUNLOCK(inp);
334 			break;
335 		}
336 	}
337 	m_freem(m);
338 
339 	if (appends == 0) {
340 		/*
341 		 * No matching pcb found; discard datagram.  (No need
342 		 * to send an ICMP Port Unreachable for a broadcast
343 		 * or multicast datagram.)
344 		 */
345 		UDPSTAT_INC(udps_noport);
346 		UDPSTAT_INC(udps_noportmcast);
347 	}
348 
349 	return (IPPROTO_DONE);
350 }
351 
352 int
udp6_input(struct mbuf ** mp,int * offp,int proto)353 udp6_input(struct mbuf **mp, int *offp, int proto)
354 {
355 	struct mbuf *m = *mp;
356 	struct ip6_hdr *ip6;
357 	struct udphdr *uh;
358 	struct inpcb *inp;
359 	struct inpcbinfo *pcbinfo;
360 	struct udpcb *up;
361 	int off = *offp;
362 	int cscov_partial;
363 	int plen, ulen;
364 	int lookupflags;
365 	struct sockaddr_in6 fromsa[2];
366 	struct m_tag *fwd_tag;
367 	uint16_t uh_sum;
368 	uint8_t nxt;
369 
370 	NET_EPOCH_ASSERT();
371 
372 	if (m->m_len < off + sizeof(struct udphdr)) {
373 		m = m_pullup(m, off + sizeof(struct udphdr));
374 		if (m == NULL) {
375 			IP6STAT_INC(ip6s_exthdrtoolong);
376 			*mp = NULL;
377 			return (IPPROTO_DONE);
378 		}
379 	}
380 	ip6 = mtod(m, struct ip6_hdr *);
381 	uh = (struct udphdr *)((caddr_t)ip6 + off);
382 
383 	UDPSTAT_INC(udps_ipackets);
384 
385 	/*
386 	 * Destination port of 0 is illegal, based on RFC768.
387 	 */
388 	if (uh->uh_dport == 0)
389 		goto badunlocked;
390 
391 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
392 	ulen = ntohs((u_short)uh->uh_ulen);
393 
394 	nxt = proto;
395 	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
396 	if (nxt == IPPROTO_UDPLITE) {
397 		/* Zero means checksum over the complete packet. */
398 		if (ulen == 0)
399 			ulen = plen;
400 		if (ulen == plen)
401 			cscov_partial = 0;
402 		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
403 			/* XXX: What is the right UDPLite MIB counter? */
404 			goto badunlocked;
405 		}
406 		if (uh->uh_sum == 0) {
407 			/* XXX: What is the right UDPLite MIB counter? */
408 			goto badunlocked;
409 		}
410 	} else {
411 		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
412 			UDPSTAT_INC(udps_badlen);
413 			goto badunlocked;
414 		}
415 		if (uh->uh_sum == 0) {
416 			UDPSTAT_INC(udps_nosum);
417 			/*
418 			 * dport 0 was rejected earlier so this is OK even if
419 			 * zero_checksum_port is 0 (which is its default value).
420 			 */
421 			if (ntohs(uh->uh_dport) == V_zero_checksum_port)
422 				goto skip_checksum;
423 			else
424 				goto badunlocked;
425 		}
426 	}
427 
428 	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
429 	    !cscov_partial) {
430 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
431 			uh_sum = m->m_pkthdr.csum_data;
432 		else
433 			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
434 			    m->m_pkthdr.csum_data);
435 		uh_sum ^= 0xffff;
436 	} else if (m->m_pkthdr.csum_flags & CSUM_IP6_UDP) {
437 		/*
438 		 * Packet from local host (maybe from a VM).
439 		 * Checksum not required.
440 		 */
441 		uh_sum = 0;
442 	} else
443 		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
444 
445 	if (uh_sum != 0) {
446 		UDPSTAT_INC(udps_badsum);
447 		goto badunlocked;
448 	}
449 
450 skip_checksum:
451 	/*
452 	 * Construct sockaddr format source address.
453 	 */
454 	init_sin6(&fromsa[0], m, 0);
455 	fromsa[0].sin6_port = uh->uh_sport;
456 	init_sin6(&fromsa[1], m, 1);
457 	fromsa[1].sin6_port = uh->uh_dport;
458 
459 	pcbinfo = udp_get_inpcbinfo(nxt);
460 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))  {
461 		*mp = NULL;
462 		return (udp6_multi_input(m, off, proto, fromsa));
463 	}
464 
465 	/*
466 	 * Locate pcb for datagram.
467 	 */
468 	lookupflags = INPLOOKUP_RLOCKPCB |
469 	    (V_udp_bind_all_fibs ? 0 : INPLOOKUP_FIB);
470 
471 	/*
472 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
473 	 */
474 	if ((m->m_flags & M_IP6_NEXTHOP) &&
475 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
476 		struct sockaddr_in6 *next_hop6;
477 
478 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
479 
480 		/*
481 		 * Transparently forwarded. Pretend to be the destination.
482 		 * Already got one like this?
483 		 */
484 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
485 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
486 		    lookupflags, m->m_pkthdr.rcvif, m);
487 		if (!inp) {
488 			/*
489 			 * It's new.  Try to find the ambushing socket.
490 			 * Because we've rewritten the destination address,
491 			 * any hardware-generated hash is ignored.
492 			 */
493 			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
494 			    uh->uh_sport, &next_hop6->sin6_addr,
495 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
496 			    uh->uh_dport, INPLOOKUP_WILDCARD | lookupflags,
497 			    m->m_pkthdr.rcvif);
498 		}
499 		/* Remove the tag from the packet. We don't need it anymore. */
500 		m_tag_delete(m, fwd_tag);
501 		m->m_flags &= ~M_IP6_NEXTHOP;
502 	} else
503 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
504 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
505 		    INPLOOKUP_WILDCARD | lookupflags,
506 		    m->m_pkthdr.rcvif, m);
507 	if (inp == NULL) {
508 		if (V_udp_log_in_vain) {
509 			char ip6bufs[INET6_ADDRSTRLEN];
510 			char ip6bufd[INET6_ADDRSTRLEN];
511 
512 			log(LOG_INFO,
513 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
514 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
515 			    ntohs(uh->uh_dport),
516 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
517 			    ntohs(uh->uh_sport));
518 		}
519 		if (nxt == IPPROTO_UDPLITE)
520 			UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
521 		else
522 			UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
523 		UDPSTAT_INC(udps_noport);
524 		if (m->m_flags & M_MCAST) {
525 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
526 			UDPSTAT_INC(udps_noportmcast);
527 			goto badunlocked;
528 		}
529 		if (V_udp_blackhole && (V_udp_blackhole_local ||
530 		    !in6_localip(&ip6->ip6_src)))
531 			goto badunlocked;
532 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
533 		*mp = NULL;
534 		return (IPPROTO_DONE);
535 	}
536 	INP_RLOCK_ASSERT(inp);
537 	up = intoudpcb(inp);
538 	if (cscov_partial) {
539 		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
540 			INP_RUNLOCK(inp);
541 			m_freem(m);
542 			*mp = NULL;
543 			return (IPPROTO_DONE);
544 		}
545 	}
546 	if (nxt == IPPROTO_UDPLITE)
547 		UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
548 	else
549 		UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
550 	if (udp6_append(inp, m, off, fromsa) == 0)
551 		INP_RUNLOCK(inp);
552 	*mp = NULL;
553 	return (IPPROTO_DONE);
554 
555 badunlocked:
556 	m_freem(m);
557 	*mp = NULL;
558 	return (IPPROTO_DONE);
559 }
560 
561 static void
udp6_common_ctlinput(struct ip6ctlparam * ip6cp,struct inpcbinfo * pcbinfo)562 udp6_common_ctlinput(struct ip6ctlparam *ip6cp, struct inpcbinfo *pcbinfo)
563 {
564 	struct udphdr uh;
565 	struct ip6_hdr *ip6;
566 	struct mbuf *m;
567 	struct inpcb *inp;
568 	int errno, off = 0;
569 	struct udp_portonly {
570 		u_int16_t uh_sport;
571 		u_int16_t uh_dport;
572 	} *uhp;
573 
574 	if ((errno = icmp6_errmap(ip6cp->ip6c_icmp6)) == 0)
575 		return;
576 
577 	m = ip6cp->ip6c_m;
578 	ip6 = ip6cp->ip6c_ip6;
579 	off = ip6cp->ip6c_off;
580 
581 	/* Check if we can safely examine src and dst ports. */
582 	if (m->m_pkthdr.len < off + sizeof(*uhp))
583 		return;
584 
585 	bzero(&uh, sizeof(uh));
586 	m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
587 
588 	/* Check to see if its tunneled */
589 	inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport,
590 	    &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
591 	    m->m_pkthdr.rcvif, m);
592 	if (inp != NULL) {
593 		struct udpcb *up;
594 		udp_tun_icmp_t *func;
595 
596 		up = intoudpcb(inp);
597 		func = up->u_icmp_func;
598 		INP_RUNLOCK(inp);
599 		if (func != NULL)
600 			func(ip6cp);
601 	}
602 	in6_pcbnotify(pcbinfo, ip6cp->ip6c_finaldst, uh.uh_dport,
603 	    ip6cp->ip6c_src, uh.uh_sport, errno, ip6cp->ip6c_cmdarg,
604 	    udp_notify);
605 }
606 
607 static void
udp6_ctlinput(struct ip6ctlparam * ctl)608 udp6_ctlinput(struct ip6ctlparam *ctl)
609 {
610 
611 	return (udp6_common_ctlinput(ctl, &V_udbinfo));
612 }
613 
614 static void
udplite6_ctlinput(struct ip6ctlparam * ctl)615 udplite6_ctlinput(struct ip6ctlparam *ctl)
616 {
617 
618 	return (udp6_common_ctlinput(ctl, &V_ulitecbinfo));
619 }
620 
621 static int
udp6_getcred(SYSCTL_HANDLER_ARGS)622 udp6_getcred(SYSCTL_HANDLER_ARGS)
623 {
624 	struct xucred xuc;
625 	struct sockaddr_in6 addrs[2];
626 	struct epoch_tracker et;
627 	struct inpcb *inp;
628 	int error;
629 
630 	if (req->newptr == NULL)
631 		return (EINVAL);
632 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
633 	if (error)
634 		return (error);
635 
636 	if (req->newlen != sizeof(addrs))
637 		return (EINVAL);
638 	if (req->oldlen != sizeof(struct xucred))
639 		return (EINVAL);
640 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
641 	if (error)
642 		return (error);
643 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
644 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
645 		return (error);
646 	}
647 	NET_EPOCH_ENTER(et);
648 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
649 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
650 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
651 	NET_EPOCH_EXIT(et);
652 	if (inp != NULL) {
653 		INP_RLOCK_ASSERT(inp);
654 		if (inp->inp_socket == NULL)
655 			error = ENOENT;
656 		if (error == 0)
657 			error = cr_canseesocket(req->td->td_ucred,
658 			    inp->inp_socket);
659 		if (error == 0)
660 			cru2x(inp->inp_cred, &xuc);
661 		INP_RUNLOCK(inp);
662 	} else
663 		error = ENOENT;
664 	if (error == 0)
665 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
666 	return (error);
667 }
668 
669 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
670     CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
671     0, 0, udp6_getcred, "S,xucred",
672     "Get the xucred of a UDP6 connection");
673 
674 static int
udp6_send(struct socket * so,int flags_arg,struct mbuf * m,struct sockaddr * addr6,struct mbuf * control,struct thread * td)675 udp6_send(struct socket *so, int flags_arg, struct mbuf *m,
676     struct sockaddr *addr6, struct mbuf *control, struct thread *td)
677 {
678 	struct inpcb *inp;
679 	struct ip6_hdr *ip6;
680 	struct udphdr *udp6;
681 	struct in6_addr *laddr, *faddr, in6a;
682 	struct ip6_pktopts *optp, opt;
683 	struct sockaddr_in6 *sin6, tmp;
684 	struct epoch_tracker et;
685 	int cscov_partial, error, flags, hlen, scope_ambiguous;
686 	u_int32_t ulen, plen;
687 	uint16_t cscov;
688 	u_short fport;
689 	uint8_t nxt;
690 
691 	if (addr6) {
692 		error = 0;
693 		if (addr6->sa_family != AF_INET6)
694 			error = EAFNOSUPPORT;
695 		else if (addr6->sa_len != sizeof(struct sockaddr_in6))
696 			error = EINVAL;
697 		if (__predict_false(error != 0)) {
698 			m_freem(control);
699 			m_freem(m);
700 			return (error);
701 		}
702 	}
703 
704 	sin6 = (struct sockaddr_in6 *)addr6;
705 
706 	/*
707 	 * In contrast to IPv4 we do not validate the max. packet length
708 	 * here due to IPv6 Jumbograms (RFC2675).
709 	 */
710 
711 	scope_ambiguous = 0;
712 	if (sin6) {
713 		/* Protect *addr6 from overwrites. */
714 		tmp = *sin6;
715 		sin6 = &tmp;
716 
717 		/*
718 		 * Application should provide a proper zone ID or the use of
719 		 * default zone IDs should be enabled.  Unfortunately, some
720 		 * applications do not behave as it should, so we need a
721 		 * workaround.  Even if an appropriate ID is not determined,
722 		 * we'll see if we can determine the outgoing interface.  If we
723 		 * can, determine the zone ID based on the interface below.
724 		 */
725 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
726 			scope_ambiguous = 1;
727 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
728 			if (control)
729 				m_freem(control);
730 			m_freem(m);
731 			return (error);
732 		}
733 	}
734 
735 	inp = sotoinpcb(so);
736 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
737 	/*
738 	 * In the following cases we want a write lock on the inp for either
739 	 * local operations or for possible route cache updates in the IPv6
740 	 * output path:
741 	 * - on connected sockets (sin6 is NULL) for route cache updates,
742 	 * - when we are not bound to an address and source port (it is
743 	 *   in6_pcbsetport() which will require the write lock).
744 	 *
745 	 * We check the inp fields before actually locking the inp, so
746 	 * here exists a race, and we may WLOCK the inp and end with already
747 	 * bound one by other thread. This is fine.
748 	 */
749 	if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
750 	    inp->inp_lport == 0))
751 		INP_WLOCK(inp);
752 	else
753 		INP_RLOCK(inp);
754 
755 	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
756 	    IPPROTO_UDP : IPPROTO_UDPLITE;
757 
758 #ifdef INET
759 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
760 		int hasv4addr;
761 
762 		if (sin6 == NULL)
763 			hasv4addr = (inp->inp_vflag & INP_IPV4);
764 		else
765 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
766 			    ? 1 : 0;
767 		if (hasv4addr) {
768 			/*
769 			 * XXXRW: We release UDP-layer locks before calling
770 			 * udp_send() in order to avoid recursion.  However,
771 			 * this does mean there is a short window where inp's
772 			 * fields are unstable.  Could this lead to a
773 			 * potential race in which the factors causing us to
774 			 * select the UDPv4 output routine are invalidated?
775 			 */
776 			INP_UNLOCK(inp);
777 			if (sin6)
778 				in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
779 			/* addr will just be freed in sendit(). */
780 			return (udp_send(so, flags_arg | PRUS_IPV6, m,
781 			    (struct sockaddr *)sin6, control, td));
782 		}
783 	} else
784 #endif
785 	if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
786 		/*
787 		 * Given this is either an IPv6-only socket or no INET is
788 		 * supported we will fail the send if the given destination
789 		 * address is a v4mapped address.
790 		 */
791 		INP_UNLOCK(inp);
792 		m_freem(m);
793 		m_freem(control);
794 		return (EINVAL);
795 	}
796 
797 	NET_EPOCH_ENTER(et);
798 	if (control) {
799 		if ((error = ip6_setpktopts(control, &opt,
800 		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
801 			goto release;
802 		}
803 		optp = &opt;
804 	} else
805 		optp = inp->in6p_outputopts;
806 
807 	if (sin6) {
808 		/*
809 		 * Since we saw no essential reason for calling in_pcbconnect,
810 		 * we get rid of such kind of logic, and call in6_selectsrc
811 		 * and in6_pcbsetport in order to fill in the local address
812 		 * and the local port.
813 		 */
814 		if (sin6->sin6_port == 0) {
815 			error = EADDRNOTAVAIL;
816 			goto release;
817 		}
818 
819 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
820 			/* how about ::ffff:0.0.0.0 case? */
821 			error = EISCONN;
822 			goto release;
823 		}
824 
825 		/*
826 		 * Given we handle the v4mapped case in the INET block above
827 		 * assert here that it must not happen anymore.
828 		 */
829 		KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
830 		    ("%s: sin6(%p)->sin6_addr is v4mapped which we "
831 		    "should have handled.", __func__, sin6));
832 
833 		/* This only requires read-locking. */
834 		error = in6_selectsrc_socket(sin6, optp, inp,
835 		    td->td_ucred, scope_ambiguous, &in6a, NULL);
836 		if (error)
837 			goto release;
838 		laddr = &in6a;
839 
840 		if (inp->inp_lport == 0) {
841 			struct inpcbinfo *pcbinfo;
842 
843 			INP_WLOCK_ASSERT(inp);
844 
845 			pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
846 			INP_HASH_WLOCK(pcbinfo);
847 			error = in6_pcbsetport(laddr, inp, td->td_ucred);
848 			INP_HASH_WUNLOCK(pcbinfo);
849 			if (error != 0) {
850 				/* Undo an address bind that may have occurred. */
851 				inp->in6p_laddr = in6addr_any;
852 				goto release;
853 			}
854 		}
855 		faddr = &sin6->sin6_addr;
856 		fport = sin6->sin6_port; /* allow 0 port */
857 
858 	} else {
859 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
860 			error = ENOTCONN;
861 			goto release;
862 		}
863 		laddr = &inp->in6p_laddr;
864 		faddr = &inp->in6p_faddr;
865 		fport = inp->inp_fport;
866 	}
867 
868 	ulen = m->m_pkthdr.len;
869 	plen = sizeof(struct udphdr) + ulen;
870 	hlen = sizeof(struct ip6_hdr);
871 
872 	/*
873 	 * Calculate data length and get a mbuf for UDP, IP6, and possible
874 	 * link-layer headers.  Immediate slide the data pointer back forward
875 	 * since we won't use that space at this layer.
876 	 */
877 	M_PREPEND(m, hlen + sizeof(struct udphdr) + max_linkhdr, M_NOWAIT);
878 	if (m == NULL) {
879 		error = ENOBUFS;
880 		goto release;
881 	}
882 	m->m_data += max_linkhdr;
883 	m->m_len -= max_linkhdr;
884 	m->m_pkthdr.len -= max_linkhdr;
885 
886 	/*
887 	 * Stuff checksum and output datagram.
888 	 */
889 	cscov = cscov_partial = 0;
890 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
891 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
892 	udp6->uh_dport = fport;
893 	if (nxt == IPPROTO_UDPLITE) {
894 		struct udpcb *up;
895 
896 		up = intoudpcb(inp);
897 		cscov = up->u_txcslen;
898 		if (cscov >= plen)
899 			cscov = 0;
900 		udp6->uh_ulen = htons(cscov);
901 		/*
902 		 * For UDP-Lite, checksum coverage length of zero means
903 		 * the entire UDPLite packet is covered by the checksum.
904 		 */
905 		cscov_partial = (cscov == 0) ? 0 : 1;
906 	} else if (plen <= 0xffff)
907 		udp6->uh_ulen = htons((u_short)plen);
908 	else
909 		udp6->uh_ulen = 0;
910 	udp6->uh_sum = 0;
911 
912 	ip6 = mtod(m, struct ip6_hdr *);
913 	ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
914 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
915 	ip6->ip6_vfc	|= IPV6_VERSION;
916 	ip6->ip6_plen	= htons((u_short)plen);
917 	ip6->ip6_nxt	= nxt;
918 	ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
919 	ip6->ip6_src	= *laddr;
920 	ip6->ip6_dst	= *faddr;
921 
922 #ifdef MAC
923 	mac_inpcb_create_mbuf(inp, m);
924 #endif
925 
926 	if (cscov_partial) {
927 		if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
928 		    sizeof(struct ip6_hdr), plen, cscov)) == 0)
929 			udp6->uh_sum = 0xffff;
930 	} else {
931 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
932 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
933 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
934 	}
935 
936 	flags = 0;
937 #if defined(ROUTE_MPATH) || defined(RSS)
938 	if (CALC_FLOWID_OUTBOUND_SENDTO) {
939 		uint32_t hash_type, hash_val;
940 		uint8_t pr;
941 
942 		pr = inp->inp_socket->so_proto->pr_protocol;
943 
944 		hash_val = fib6_calc_packet_hash(laddr, faddr,
945 		    inp->inp_lport, fport, pr, &hash_type);
946 		m->m_pkthdr.flowid = hash_val;
947 		M_HASHTYPE_SET(m, hash_type);
948 	}
949 	/* do not use inp flowid */
950 	flags |= IP_NODEFAULTFLOWID;
951 #endif
952 
953 	UDPSTAT_INC(udps_opackets);
954 	if (nxt == IPPROTO_UDPLITE)
955 		UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
956 	else
957 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
958 	error = ip6_output(m, optp,
959 	    INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
960 	    inp->in6p_moptions, NULL, inp);
961 	INP_UNLOCK(inp);
962 	NET_EPOCH_EXIT(et);
963 
964 	if (control) {
965 		ip6_clearpktopts(&opt, -1);
966 		m_freem(control);
967 	}
968 	return (error);
969 
970 release:
971 	INP_UNLOCK(inp);
972 	NET_EPOCH_EXIT(et);
973 	if (control) {
974 		ip6_clearpktopts(&opt, -1);
975 		m_freem(control);
976 	}
977 	m_freem(m);
978 
979 	return (error);
980 }
981 
982 static void
udp6_abort(struct socket * so)983 udp6_abort(struct socket *so)
984 {
985 	struct inpcb *inp;
986 	struct inpcbinfo *pcbinfo;
987 
988 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
989 	inp = sotoinpcb(so);
990 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
991 
992 	INP_WLOCK(inp);
993 #ifdef INET
994 	if (inp->inp_vflag & INP_IPV4) {
995 		INP_WUNLOCK(inp);
996 		udp_abort(so);
997 		return;
998 	}
999 #endif
1000 
1001 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1002 		INP_HASH_WLOCK(pcbinfo);
1003 		in6_pcbdisconnect(inp);
1004 		INP_HASH_WUNLOCK(pcbinfo);
1005 		soisdisconnected(so);
1006 	}
1007 	INP_WUNLOCK(inp);
1008 }
1009 
1010 static int
udp6_attach(struct socket * so,int proto,struct thread * td)1011 udp6_attach(struct socket *so, int proto, struct thread *td)
1012 {
1013 	struct inpcbinfo *pcbinfo;
1014 	struct inpcb *inp;
1015 	struct udpcb *up;
1016 	int error;
1017 
1018 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1019 	inp = sotoinpcb(so);
1020 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1021 
1022 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1023 		error = soreserve(so, udp_sendspace, udp_recvspace);
1024 		if (error)
1025 			return (error);
1026 	}
1027 	error = in_pcballoc(so, pcbinfo);
1028 	if (error)
1029 		return (error);
1030 	inp = (struct inpcb *)so->so_pcb;
1031 	inp->in6p_cksum = -1;	/* just to be sure */
1032 	/*
1033 	 * XXX: ugly!!
1034 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1035 	 * because the socket may be bound to an IPv6 wildcard address,
1036 	 * which may match an IPv4-mapped IPv6 address.
1037 	 */
1038 	inp->inp_ip_ttl = V_ip_defttl;
1039 	up = intoudpcb(inp);
1040 	bzero(&up->u_start_zero, u_zero_size);
1041 	INP_WUNLOCK(inp);
1042 	return (0);
1043 }
1044 
1045 static int
udp6_bind(struct socket * so,struct sockaddr * nam,struct thread * td)1046 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1047 {
1048 	struct sockaddr_in6 *sin6_p;
1049 	struct inpcb *inp;
1050 	struct inpcbinfo *pcbinfo;
1051 	int error;
1052 	u_char vflagsav;
1053 
1054 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1055 	inp = sotoinpcb(so);
1056 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1057 
1058 	if (nam->sa_family != AF_INET6)
1059 		return (EAFNOSUPPORT);
1060 	if (nam->sa_len != sizeof(struct sockaddr_in6))
1061 		return (EINVAL);
1062 
1063 	sin6_p = (struct sockaddr_in6 *)nam;
1064 
1065 	INP_WLOCK(inp);
1066 	INP_HASH_WLOCK(pcbinfo);
1067 	vflagsav = inp->inp_vflag;
1068 	inp->inp_vflag &= ~INP_IPV4;
1069 	inp->inp_vflag |= INP_IPV6;
1070 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1071 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1072 			inp->inp_vflag |= INP_IPV4;
1073 #ifdef INET
1074 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1075 			struct sockaddr_in sin;
1076 
1077 			in6_sin6_2_sin(&sin, sin6_p);
1078 			inp->inp_vflag |= INP_IPV4;
1079 			inp->inp_vflag &= ~INP_IPV6;
1080 			error = in_pcbbind(inp, &sin,
1081 			    V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1082 			    td->td_ucred);
1083 			goto out;
1084 		}
1085 #endif
1086 	}
1087 
1088 	error = in6_pcbbind(inp, sin6_p, V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1089 	    td->td_ucred);
1090 #ifdef INET
1091 out:
1092 #endif
1093 	if (error != 0)
1094 		inp->inp_vflag = vflagsav;
1095 	INP_HASH_WUNLOCK(pcbinfo);
1096 	INP_WUNLOCK(inp);
1097 	return (error);
1098 }
1099 
1100 static void
udp6_close(struct socket * so)1101 udp6_close(struct socket *so)
1102 {
1103 	struct inpcb *inp;
1104 	struct inpcbinfo *pcbinfo;
1105 
1106 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1107 	inp = sotoinpcb(so);
1108 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1109 
1110 	INP_WLOCK(inp);
1111 #ifdef INET
1112 	if (inp->inp_vflag & INP_IPV4) {
1113 		INP_WUNLOCK(inp);
1114 		(void)udp_disconnect(so);
1115 		return;
1116 	}
1117 #endif
1118 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1119 		INP_HASH_WLOCK(pcbinfo);
1120 		in6_pcbdisconnect(inp);
1121 		INP_HASH_WUNLOCK(pcbinfo);
1122 		soisdisconnected(so);
1123 	}
1124 	INP_WUNLOCK(inp);
1125 }
1126 
1127 static int
udp6_connect(struct socket * so,struct sockaddr * nam,struct thread * td)1128 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1129 {
1130 	struct epoch_tracker et;
1131 	struct inpcb *inp;
1132 	struct inpcbinfo *pcbinfo;
1133 	struct sockaddr_in6 *sin6;
1134 	int error;
1135 	u_char vflagsav;
1136 
1137 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1138 	inp = sotoinpcb(so);
1139 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1140 
1141 	sin6 = (struct sockaddr_in6 *)nam;
1142 	if (sin6->sin6_family != AF_INET6)
1143 		return (EAFNOSUPPORT);
1144 	if (sin6->sin6_len != sizeof(*sin6))
1145 		return (EINVAL);
1146 
1147 	/*
1148 	 * XXXRW: Need to clarify locking of v4/v6 flags.
1149 	 */
1150 	INP_WLOCK(inp);
1151 #ifdef INET
1152 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1153 		struct sockaddr_in sin;
1154 
1155 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1156 			error = EINVAL;
1157 			goto out;
1158 		}
1159 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1160 			error = EAFNOSUPPORT;
1161 			goto out;
1162 		}
1163 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1164 			error = EISCONN;
1165 			goto out;
1166 		}
1167 		in6_sin6_2_sin(&sin, sin6);
1168 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1169 		if (error != 0)
1170 			goto out;
1171 		vflagsav = inp->inp_vflag;
1172 		inp->inp_vflag |= INP_IPV4;
1173 		inp->inp_vflag &= ~INP_IPV6;
1174 		NET_EPOCH_ENTER(et);
1175 		INP_HASH_WLOCK(pcbinfo);
1176 		error = in_pcbconnect(inp, &sin, td->td_ucred);
1177 		INP_HASH_WUNLOCK(pcbinfo);
1178 		NET_EPOCH_EXIT(et);
1179 		/*
1180 		 * If connect succeeds, mark socket as connected. If
1181 		 * connect fails and socket is unbound, reset inp_vflag
1182 		 * field.
1183 		 */
1184 		if (error == 0)
1185 			soisconnected(so);
1186 		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1187 		    inp->inp_lport == 0)
1188 			inp->inp_vflag = vflagsav;
1189 		goto out;
1190 	} else {
1191 		if ((inp->inp_vflag & INP_IPV6) == 0) {
1192 			error = EAFNOSUPPORT;
1193 			goto out;
1194 		}
1195 	}
1196 #endif
1197 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1198 		error = EISCONN;
1199 		goto out;
1200 	}
1201 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_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 = in6_pcbconnect(inp, sin6, td->td_ucred, true);
1210 	INP_HASH_WUNLOCK(pcbinfo);
1211 	NET_EPOCH_EXIT(et);
1212 	/*
1213 	 * If connect succeeds, mark socket as connected. If
1214 	 * connect fails and socket is unbound, reset inp_vflag
1215 	 * field.
1216 	 */
1217 	if (error == 0)
1218 		soisconnected(so);
1219 	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1220 	    inp->inp_lport == 0)
1221 		inp->inp_vflag = vflagsav;
1222 out:
1223 	INP_WUNLOCK(inp);
1224 	return (error);
1225 }
1226 
1227 static void
udp6_detach(struct socket * so)1228 udp6_detach(struct socket *so)
1229 {
1230 	struct inpcb *inp;
1231 
1232 	inp = sotoinpcb(so);
1233 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1234 
1235 	INP_WLOCK(inp);
1236 	in_pcbfree(inp);
1237 }
1238 
1239 static int
udp6_disconnect(struct socket * so)1240 udp6_disconnect(struct socket *so)
1241 {
1242 	struct inpcb *inp;
1243 	struct inpcbinfo *pcbinfo;
1244 
1245 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1246 	inp = sotoinpcb(so);
1247 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1248 
1249 	INP_WLOCK(inp);
1250 #ifdef INET
1251 	if (inp->inp_vflag & INP_IPV4) {
1252 		INP_WUNLOCK(inp);
1253 		(void)udp_disconnect(so);
1254 		return (0);
1255 	}
1256 #endif
1257 
1258 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1259 		INP_WUNLOCK(inp);
1260 		return (ENOTCONN);
1261 	}
1262 
1263 	INP_HASH_WLOCK(pcbinfo);
1264 	in6_pcbdisconnect(inp);
1265 	INP_HASH_WUNLOCK(pcbinfo);
1266 	SOCK_LOCK(so);
1267 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1268 	SOCK_UNLOCK(so);
1269 	INP_WUNLOCK(inp);
1270 	return (0);
1271 }
1272 
1273 #define	UDP6_PROTOSW							\
1274 	.pr_type =		SOCK_DGRAM,				\
1275 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_CAPATTACH,		\
1276 	.pr_ctloutput =		udp_ctloutput,				\
1277 	.pr_abort =		udp6_abort,				\
1278 	.pr_attach =		udp6_attach,				\
1279 	.pr_bind =		udp6_bind,				\
1280 	.pr_connect =		udp6_connect,				\
1281 	.pr_control =		in6_control,				\
1282 	.pr_detach =		udp6_detach,				\
1283 	.pr_disconnect =	udp6_disconnect,			\
1284 	.pr_peeraddr =		in6_mapped_peeraddr,			\
1285 	.pr_send =		udp6_send,				\
1286 	.pr_shutdown =		udp_shutdown,				\
1287 	.pr_sockaddr =		in6_mapped_sockaddr,			\
1288 	.pr_soreceive =		soreceive_dgram,			\
1289 	.pr_sosend =		sosend_dgram,				\
1290 	.pr_sosetlabel =	in_pcbsosetlabel,			\
1291 	.pr_close =		udp6_close
1292 
1293 struct protosw udp6_protosw = {
1294 	.pr_protocol =		IPPROTO_UDP,
1295 	UDP6_PROTOSW
1296 };
1297 
1298 struct protosw udplite6_protosw = {
1299 	.pr_protocol =		IPPROTO_UDPLITE,
1300 	UDP6_PROTOSW
1301 };
1302 
1303 static void
udp6_init(void * arg __unused)1304 udp6_init(void *arg __unused)
1305 {
1306 
1307 	IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1308 	IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1309 }
1310 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);
1311