xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 254b23eb1f540844cf2a90f2781ae4231c5701ce)
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_rss.h"
74 
75 #include <sys/param.h>
76 #include <sys/domain.h>
77 #include <sys/jail.h>
78 #include <sys/kernel.h>
79 #include <sys/lock.h>
80 #include <sys/mbuf.h>
81 #include <sys/priv.h>
82 #include <sys/proc.h>
83 #include <sys/protosw.h>
84 #include <sys/sdt.h>
85 #include <sys/signalvar.h>
86 #include <sys/socket.h>
87 #include <sys/socketvar.h>
88 #include <sys/sx.h>
89 #include <sys/sysctl.h>
90 #include <sys/syslog.h>
91 #include <sys/systm.h>
92 
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_types.h>
96 #include <net/route.h>
97 #include <net/rss_config.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_kdtrace.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip.h>
105 #include <netinet/ip6.h>
106 #include <netinet/icmp6.h>
107 #include <netinet/ip_var.h>
108 #include <netinet/udp.h>
109 #include <netinet/udp_var.h>
110 #include <netinet/udplite.h>
111 
112 #include <netinet6/ip6_var.h>
113 #include <netinet6/in6_fib.h>
114 #include <netinet6/in6_pcb.h>
115 #include <netinet6/in6_rss.h>
116 #include <netinet6/udp6_var.h>
117 #include <netinet6/scope6_var.h>
118 
119 #include <netipsec/ipsec_support.h>
120 
121 #include <security/mac/mac_framework.h>
122 
123 VNET_DEFINE(int, zero_checksum_port) = 0;
124 #define	V_zero_checksum_port	VNET(zero_checksum_port)
125 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
126     &VNET_NAME(zero_checksum_port), 0,
127     "Zero UDP checksum allowed for traffic to/from this port.");
128 
129 /*
130  * UDP protocol implementation.
131  * Per RFC 768, August, 1980.
132  */
133 
134 static void		udp6_detach(struct socket *so);
135 
136 static int
udp6_append(struct inpcb * inp,struct mbuf * n,int off,struct sockaddr_in6 * fromsa)137 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
138     struct sockaddr_in6 *fromsa)
139 {
140 	struct socket *so;
141 	struct mbuf *opts = NULL, *tmp_opts;
142 	struct udpcb *up;
143 
144 	INP_LOCK_ASSERT(inp);
145 
146 	/*
147 	 * Engage the tunneling protocol.
148 	 */
149 	up = intoudpcb(inp);
150 	if (up->u_tun_func != NULL) {
151 		bool filtered;
152 
153 		in_pcbref(inp);
154 		INP_RUNLOCK(inp);
155 		filtered = (*up->u_tun_func)(n, off, inp,
156 		    (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
157 		INP_RLOCK(inp);
158 		if (in_pcbrele_rlocked(inp))
159 			return (1);
160 		if (filtered) {
161 			INP_RUNLOCK(inp);
162 			return (1);
163 		}
164 	}
165 
166 	off += sizeof(struct udphdr);
167 
168 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
169 	/* Check AH/ESP integrity. */
170 	if (IPSEC_ENABLED(ipv6)) {
171 		if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
172 			m_freem(n);
173 			return (0);
174 		}
175 
176 		/* IPSec UDP encaps. */
177 		if ((up->u_flags & UF_ESPINUDP) != 0 &&
178 		    UDPENCAP_INPUT(ipv6, n, off, AF_INET6) != 0) {
179 			return (0); /* Consumed. */
180 		}
181 	}
182 #endif /* IPSEC */
183 #ifdef MAC
184 	if (mac_inpcb_check_deliver(inp, n) != 0) {
185 		m_freem(n);
186 		return (0);
187 	}
188 #endif
189 	opts = NULL;
190 	if (inp->inp_flags & INP_CONTROLOPTS ||
191 	    inp->inp_socket->so_options & SO_TIMESTAMP)
192 		ip6_savecontrol(inp, n, &opts);
193 	if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
194 		tmp_opts = sbcreatecontrol(&fromsa[1],
195 		    sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
196 		    IPPROTO_IPV6, M_NOWAIT);
197                 if (tmp_opts) {
198                         if (opts) {
199                                 tmp_opts->m_next = opts;
200                                 opts = tmp_opts;
201                         } else
202                                 opts = tmp_opts;
203                 }
204 	}
205 	m_adj(n, off);
206 
207 	so = inp->inp_socket;
208 	SOCKBUF_LOCK(&so->so_rcv);
209 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
210 	    opts) == 0) {
211 		soroverflow_locked(so);
212 		m_freem(n);
213 		if (opts)
214 			m_freem(opts);
215 		UDPSTAT_INC(udps_fullsock);
216 	} else
217 		sorwakeup_locked(so);
218 	return (0);
219 }
220 
221 struct udp6_multi_match_ctx {
222 	struct ip6_hdr *ip6;
223 	struct udphdr *uh;
224 };
225 
226 static bool
udp6_multi_match(const struct inpcb * inp,void * v)227 udp6_multi_match(const struct inpcb *inp, void *v)
228 {
229 	struct udp6_multi_match_ctx *ctx = v;
230 
231 	if ((inp->inp_vflag & INP_IPV6) == 0)
232 		return(false);
233 	if (inp->inp_lport != ctx->uh->uh_dport)
234 		return(false);
235 	if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
236 		return(false);
237 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
238 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
239 		return (false);
240 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
241 	    (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
242 	    inp->inp_fport != ctx->uh->uh_sport))
243 		return (false);
244 
245 	return (true);
246 }
247 
248 static int
udp6_multi_input(struct mbuf * m,int off,int proto,struct sockaddr_in6 * fromsa)249 udp6_multi_input(struct mbuf *m, int off, int proto,
250     struct sockaddr_in6 *fromsa)
251 {
252 	struct udp6_multi_match_ctx ctx;
253 	struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
254 	    INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
255 	struct inpcb *inp;
256 	struct ip6_moptions *imo;
257 	struct mbuf *n;
258 	int appends = 0;
259 
260 	/*
261 	 * In the event that laddr should be set to the link-local
262 	 * address (this happens in RIPng), the multicast address
263 	 * specified in the received packet will not match laddr.  To
264 	 * handle this situation, matching is relaxed if the
265 	 * receiving interface is the same as one specified in the
266 	 * socket and if the destination multicast address matches
267 	 * one of the multicast groups specified in the socket.
268 	 */
269 
270 	/*
271 	 * KAME note: traditionally we dropped udpiphdr from mbuf
272 	 * here.  We need udphdr for IPsec processing so we do that
273 	 * later.
274 	 */
275 	ctx.ip6 = mtod(m, struct ip6_hdr *);
276 	ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
277 	while ((inp = inp_next(&inpi)) != NULL) {
278 		INP_RLOCK_ASSERT(inp);
279 		/*
280 		 * XXXRW: Because we weren't holding either the inpcb
281 		 * or the hash lock when we checked for a match
282 		 * before, we should probably recheck now that the
283 		 * inpcb lock is (supposed to be) held.
284 		 */
285 		/*
286 		 * Handle socket delivery policy for any-source
287 		 * and source-specific multicast. [RFC3678]
288 		 */
289 		if ((imo = inp->in6p_moptions) != NULL) {
290 			struct sockaddr_in6	 mcaddr;
291 			int			 blocked;
292 
293 			bzero(&mcaddr, sizeof(struct sockaddr_in6));
294 			mcaddr.sin6_len = sizeof(struct sockaddr_in6);
295 			mcaddr.sin6_family = AF_INET6;
296 			mcaddr.sin6_addr = ctx.ip6->ip6_dst;
297 
298 			blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
299 				(struct sockaddr *)&mcaddr,
300 				(struct sockaddr *)&fromsa[0]);
301 			if (blocked != MCAST_PASS) {
302 				if (blocked == MCAST_NOTGMEMBER)
303 					IP6STAT_INC(ip6s_notmember);
304 				if (blocked == MCAST_NOTSMEMBER ||
305 				    blocked == MCAST_MUTED)
306 					UDPSTAT_INC(udps_filtermcast);
307 				continue;
308 			}
309 		}
310 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
311 			if (proto == IPPROTO_UDPLITE)
312 				UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
313 				    inp, ctx.uh);
314 			else
315 				UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
316 				    ctx.uh);
317 			if (udp6_append(inp, n, off, fromsa)) {
318 				break;
319 			} else
320 				appends++;
321 		}
322 		/*
323 		 * Don't look for additional matches if this one does
324 		 * not have either the SO_REUSEPORT or SO_REUSEADDR
325 		 * socket options set.  This heuristic avoids
326 		 * searching through all pcbs in the common case of a
327 		 * non-shared port.  It assumes that an application
328 		 * will never clear these options after setting them.
329 		 */
330 		if ((inp->inp_socket->so_options &
331 		     (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
332 			INP_RUNLOCK(inp);
333 			break;
334 		}
335 	}
336 	m_freem(m);
337 
338 	if (appends == 0) {
339 		/*
340 		 * No matching pcb found; discard datagram.  (No need
341 		 * to send an ICMP Port Unreachable for a broadcast
342 		 * or multicast datagram.)
343 		 */
344 		UDPSTAT_INC(udps_noport);
345 		UDPSTAT_INC(udps_noportmcast);
346 	}
347 
348 	return (IPPROTO_DONE);
349 }
350 
351 int
udp6_input(struct mbuf ** mp,int * offp,int proto)352 udp6_input(struct mbuf **mp, int *offp, int proto)
353 {
354 	struct mbuf *m = *mp;
355 	struct ip6_hdr *ip6;
356 	struct udphdr *uh;
357 	struct inpcb *inp;
358 	struct inpcbinfo *pcbinfo;
359 	struct udpcb *up;
360 	int off = *offp;
361 	int cscov_partial;
362 	int plen, ulen;
363 	int lookupflags;
364 	struct sockaddr_in6 fromsa[2];
365 	struct m_tag *fwd_tag;
366 	uint16_t uh_sum;
367 	uint8_t nxt;
368 
369 	NET_EPOCH_ASSERT();
370 
371 	if (m->m_len < off + sizeof(struct udphdr)) {
372 		m = m_pullup(m, off + sizeof(struct udphdr));
373 		if (m == NULL) {
374 			IP6STAT_INC(ip6s_exthdrtoolong);
375 			*mp = NULL;
376 			return (IPPROTO_DONE);
377 		}
378 	}
379 	ip6 = mtod(m, struct ip6_hdr *);
380 	uh = (struct udphdr *)((caddr_t)ip6 + off);
381 
382 	UDPSTAT_INC(udps_ipackets);
383 
384 	/*
385 	 * Destination port of 0 is illegal, based on RFC768.
386 	 */
387 	if (uh->uh_dport == 0)
388 		goto badunlocked;
389 
390 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
391 	ulen = ntohs((u_short)uh->uh_ulen);
392 
393 	nxt = proto;
394 	cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
395 	if (nxt == IPPROTO_UDPLITE) {
396 		/* Zero means checksum over the complete packet. */
397 		if (ulen == 0)
398 			ulen = plen;
399 		if (ulen == plen)
400 			cscov_partial = 0;
401 		if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
402 			/* XXX: What is the right UDPLite MIB counter? */
403 			goto badunlocked;
404 		}
405 		if (uh->uh_sum == 0) {
406 			/* XXX: What is the right UDPLite MIB counter? */
407 			goto badunlocked;
408 		}
409 	} else {
410 		if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
411 			UDPSTAT_INC(udps_badlen);
412 			goto badunlocked;
413 		}
414 		if (uh->uh_sum == 0) {
415 			UDPSTAT_INC(udps_nosum);
416 			/*
417 			 * dport 0 was rejected earlier so this is OK even if
418 			 * zero_checksum_port is 0 (which is its default value).
419 			 */
420 			if (ntohs(uh->uh_dport) == V_zero_checksum_port)
421 				goto skip_checksum;
422 			else
423 				goto badunlocked;
424 		}
425 	}
426 
427 	if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
428 	    !cscov_partial) {
429 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
430 			uh_sum = m->m_pkthdr.csum_data;
431 		else
432 			uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
433 			    m->m_pkthdr.csum_data);
434 		uh_sum ^= 0xffff;
435 	} else if (m->m_pkthdr.csum_flags & CSUM_IP6_UDP) {
436 		/*
437 		 * Packet from local host (maybe from a VM).
438 		 * Checksum not required.
439 		 */
440 		uh_sum = 0;
441 	} else
442 		uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
443 
444 	if (uh_sum != 0) {
445 		UDPSTAT_INC(udps_badsum);
446 		goto badunlocked;
447 	}
448 
449 skip_checksum:
450 	/*
451 	 * Construct sockaddr format source address.
452 	 */
453 	init_sin6(&fromsa[0], m, 0);
454 	fromsa[0].sin6_port = uh->uh_sport;
455 	init_sin6(&fromsa[1], m, 1);
456 	fromsa[1].sin6_port = uh->uh_dport;
457 
458 	pcbinfo = udp_get_inpcbinfo(nxt);
459 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))  {
460 		*mp = NULL;
461 		return (udp6_multi_input(m, off, proto, fromsa));
462 	}
463 
464 	/*
465 	 * Locate pcb for datagram.
466 	 */
467 	lookupflags = INPLOOKUP_RLOCKPCB |
468 	    (V_udp_bind_all_fibs ? 0 : INPLOOKUP_FIB);
469 
470 	/*
471 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
472 	 */
473 	if ((m->m_flags & M_IP6_NEXTHOP) &&
474 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
475 		struct sockaddr_in6 *next_hop6;
476 
477 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
478 
479 		/*
480 		 * Transparently forwarded. Pretend to be the destination.
481 		 * Already got one like this?
482 		 */
483 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
484 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
485 		    lookupflags, m->m_pkthdr.rcvif, m);
486 		if (!inp) {
487 			/*
488 			 * It's new.  Try to find the ambushing socket.
489 			 * Because we've rewritten the destination address,
490 			 * any hardware-generated hash is ignored.
491 			 */
492 			inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
493 			    uh->uh_sport, &next_hop6->sin6_addr,
494 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
495 			    uh->uh_dport, INPLOOKUP_WILDCARD | lookupflags,
496 			    m->m_pkthdr.rcvif);
497 		}
498 		/* Remove the tag from the packet. We don't need it anymore. */
499 		m_tag_delete(m, fwd_tag);
500 		m->m_flags &= ~M_IP6_NEXTHOP;
501 	} else
502 		inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
503 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
504 		    INPLOOKUP_WILDCARD | lookupflags,
505 		    m->m_pkthdr.rcvif, m);
506 	if (inp == NULL) {
507 		if (V_udp_log_in_vain) {
508 			char ip6bufs[INET6_ADDRSTRLEN];
509 			char ip6bufd[INET6_ADDRSTRLEN];
510 
511 			log(LOG_INFO,
512 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
513 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
514 			    ntohs(uh->uh_dport),
515 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
516 			    ntohs(uh->uh_sport));
517 		}
518 		if (nxt == IPPROTO_UDPLITE)
519 			UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
520 		else
521 			UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
522 		UDPSTAT_INC(udps_noport);
523 		if (m->m_flags & M_MCAST) {
524 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
525 			UDPSTAT_INC(udps_noportmcast);
526 			goto badunlocked;
527 		}
528 		if (V_udp_blackhole && (V_udp_blackhole_local ||
529 		    !in6_localip(&ip6->ip6_src)))
530 			goto badunlocked;
531 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
532 		*mp = NULL;
533 		return (IPPROTO_DONE);
534 	}
535 	INP_RLOCK_ASSERT(inp);
536 	up = intoudpcb(inp);
537 	if (cscov_partial) {
538 		if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
539 			INP_RUNLOCK(inp);
540 			m_freem(m);
541 			*mp = NULL;
542 			return (IPPROTO_DONE);
543 		}
544 	}
545 	if (nxt == IPPROTO_UDPLITE)
546 		UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
547 	else
548 		UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
549 	if (udp6_append(inp, m, off, fromsa) == 0)
550 		INP_RUNLOCK(inp);
551 	*mp = NULL;
552 	return (IPPROTO_DONE);
553 
554 badunlocked:
555 	m_freem(m);
556 	*mp = NULL;
557 	return (IPPROTO_DONE);
558 }
559 
560 static void
udp6_common_ctlinput(struct ip6ctlparam * ip6cp,struct inpcbinfo * pcbinfo)561 udp6_common_ctlinput(struct ip6ctlparam *ip6cp, struct inpcbinfo *pcbinfo)
562 {
563 	struct udphdr uh;
564 	struct ip6_hdr *ip6;
565 	struct mbuf *m;
566 	struct inpcb *inp;
567 	int errno, off = 0;
568 	struct udp_portonly {
569 		u_int16_t uh_sport;
570 		u_int16_t uh_dport;
571 	} *uhp;
572 
573 	if ((errno = icmp6_errmap(ip6cp->ip6c_icmp6)) == 0)
574 		return;
575 
576 	m = ip6cp->ip6c_m;
577 	ip6 = ip6cp->ip6c_ip6;
578 	off = ip6cp->ip6c_off;
579 
580 	/* Check if we can safely examine src and dst ports. */
581 	if (m->m_pkthdr.len < off + sizeof(*uhp))
582 		return;
583 
584 	bzero(&uh, sizeof(uh));
585 	m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
586 
587 	/* Check to see if its tunneled */
588 	inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport,
589 	    &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
590 	    m->m_pkthdr.rcvif, m);
591 	if (inp != NULL) {
592 		struct udpcb *up;
593 		udp_tun_icmp_t *func;
594 
595 		up = intoudpcb(inp);
596 		func = up->u_icmp_func;
597 		INP_RUNLOCK(inp);
598 		if (func != NULL)
599 			func(ip6cp);
600 	}
601 	in6_pcbnotify(pcbinfo, ip6cp->ip6c_finaldst, uh.uh_dport,
602 	    ip6cp->ip6c_src, uh.uh_sport, errno, ip6cp->ip6c_cmdarg,
603 	    udp_notify);
604 }
605 
606 static void
udp6_ctlinput(struct ip6ctlparam * ctl)607 udp6_ctlinput(struct ip6ctlparam *ctl)
608 {
609 
610 	return (udp6_common_ctlinput(ctl, &V_udbinfo));
611 }
612 
613 static void
udplite6_ctlinput(struct ip6ctlparam * ctl)614 udplite6_ctlinput(struct ip6ctlparam *ctl)
615 {
616 
617 	return (udp6_common_ctlinput(ctl, &V_ulitecbinfo));
618 }
619 
620 static int
udp6_getcred(SYSCTL_HANDLER_ARGS)621 udp6_getcred(SYSCTL_HANDLER_ARGS)
622 {
623 	struct xucred xuc;
624 	struct sockaddr_in6 addrs[2];
625 	struct epoch_tracker et;
626 	struct inpcb *inp;
627 	int error;
628 
629 	if (req->newptr == NULL)
630 		return (EINVAL);
631 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
632 	if (error)
633 		return (error);
634 
635 	if (req->newlen != sizeof(addrs))
636 		return (EINVAL);
637 	if (req->oldlen != sizeof(struct xucred))
638 		return (EINVAL);
639 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
640 	if (error)
641 		return (error);
642 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
643 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
644 		return (error);
645 	}
646 	NET_EPOCH_ENTER(et);
647 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
648 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
649 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
650 	NET_EPOCH_EXIT(et);
651 	if (inp != NULL) {
652 		INP_RLOCK_ASSERT(inp);
653 		if (inp->inp_socket == NULL)
654 			error = ENOENT;
655 		if (error == 0)
656 			error = cr_canseesocket(req->td->td_ucred,
657 			    inp->inp_socket);
658 		if (error == 0)
659 			cru2x(inp->inp_cred, &xuc);
660 		INP_RUNLOCK(inp);
661 	} else
662 		error = ENOENT;
663 	if (error == 0)
664 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
665 	return (error);
666 }
667 
668 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
669     CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
670     0, 0, udp6_getcred, "S,xucred",
671     "Get the xucred of a UDP6 connection");
672 
673 static int
udp6_send(struct socket * so,int flags_arg,struct mbuf * m,struct sockaddr * addr6,struct mbuf * control,struct thread * td)674 udp6_send(struct socket *so, int flags_arg, struct mbuf *m,
675     struct sockaddr *addr6, struct mbuf *control, struct thread *td)
676 {
677 	struct inpcb *inp;
678 	struct ip6_hdr *ip6;
679 	struct udphdr *udp6;
680 	struct in6_addr *laddr, *faddr, in6a;
681 	struct ip6_pktopts *optp, opt;
682 	struct sockaddr_in6 *sin6, tmp;
683 	struct epoch_tracker et;
684 	int cscov_partial, error, flags, hlen, scope_ambiguous;
685 	u_int32_t ulen, plen;
686 	uint16_t cscov;
687 	u_short fport;
688 	uint8_t nxt;
689 
690 	if (addr6) {
691 		error = 0;
692 		if (addr6->sa_family != AF_INET6)
693 			error = EAFNOSUPPORT;
694 		else if (addr6->sa_len != sizeof(struct sockaddr_in6))
695 			error = EINVAL;
696 		if (__predict_false(error != 0)) {
697 			m_freem(control);
698 			m_freem(m);
699 			return (error);
700 		}
701 	}
702 
703 	sin6 = (struct sockaddr_in6 *)addr6;
704 
705 	scope_ambiguous = 0;
706 	if (sin6) {
707 		/* Protect *addr6 from overwrites. */
708 		tmp = *sin6;
709 		sin6 = &tmp;
710 
711 		/*
712 		 * Application should provide a proper zone ID or the use of
713 		 * default zone IDs should be enabled.  Unfortunately, some
714 		 * applications do not behave as it should, so we need a
715 		 * workaround.  Even if an appropriate ID is not determined,
716 		 * we'll see if we can determine the outgoing interface.  If we
717 		 * can, determine the zone ID based on the interface below.
718 		 */
719 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
720 			scope_ambiguous = 1;
721 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
722 			if (control)
723 				m_freem(control);
724 			m_freem(m);
725 			return (error);
726 		}
727 	}
728 
729 	inp = sotoinpcb(so);
730 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
731 	/*
732 	 * In the following cases we want a write lock on the inp for either
733 	 * local operations or for possible route cache updates in the IPv6
734 	 * output path:
735 	 * - on connected sockets (sin6 is NULL) for route cache updates,
736 	 * - when we are not bound to an address and source port (it is
737 	 *   in6_pcbsetport() which will require the write lock).
738 	 *
739 	 * We check the inp fields before actually locking the inp, so
740 	 * here exists a race, and we may WLOCK the inp and end with already
741 	 * bound one by other thread. This is fine.
742 	 */
743 	if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
744 	    inp->inp_lport == 0))
745 		INP_WLOCK(inp);
746 	else
747 		INP_RLOCK(inp);
748 
749 	nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
750 	    IPPROTO_UDP : IPPROTO_UDPLITE;
751 
752 #ifdef INET
753 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
754 		int hasv4addr;
755 
756 		if (sin6 == NULL)
757 			hasv4addr = (inp->inp_vflag & INP_IPV4);
758 		else
759 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
760 			    ? 1 : 0;
761 		if (hasv4addr) {
762 			/*
763 			 * XXXRW: We release UDP-layer locks before calling
764 			 * udp_send() in order to avoid recursion.  However,
765 			 * this does mean there is a short window where inp's
766 			 * fields are unstable.  Could this lead to a
767 			 * potential race in which the factors causing us to
768 			 * select the UDPv4 output routine are invalidated?
769 			 */
770 			INP_UNLOCK(inp);
771 			if (sin6)
772 				in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
773 			/* addr will just be freed in sendit(). */
774 			return (udp_send(so, flags_arg | PRUS_IPV6, m,
775 			    (struct sockaddr *)sin6, control, td));
776 		}
777 	} else
778 #endif
779 	if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
780 		/*
781 		 * Given this is either an IPv6-only socket or no INET is
782 		 * supported we will fail the send if the given destination
783 		 * address is a v4mapped address.
784 		 */
785 		INP_UNLOCK(inp);
786 		m_freem(m);
787 		m_freem(control);
788 		return (EINVAL);
789 	}
790 
791 	NET_EPOCH_ENTER(et);
792 	if (control) {
793 		if ((error = ip6_setpktopts(control, &opt,
794 		    inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
795 			goto release;
796 		}
797 		optp = &opt;
798 	} else
799 		optp = inp->in6p_outputopts;
800 
801 	if (sin6) {
802 		/*
803 		 * Since we saw no essential reason for calling in_pcbconnect,
804 		 * we get rid of such kind of logic, and call in6_selectsrc
805 		 * and in6_pcbsetport in order to fill in the local address
806 		 * and the local port.
807 		 */
808 		if (sin6->sin6_port == 0) {
809 			error = EADDRNOTAVAIL;
810 			goto release;
811 		}
812 
813 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
814 			/* how about ::ffff:0.0.0.0 case? */
815 			error = EISCONN;
816 			goto release;
817 		}
818 
819 		/*
820 		 * Given we handle the v4mapped case in the INET block above
821 		 * assert here that it must not happen anymore.
822 		 */
823 		KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
824 		    ("%s: sin6(%p)->sin6_addr is v4mapped which we "
825 		    "should have handled.", __func__, sin6));
826 
827 		/* This only requires read-locking. */
828 		error = in6_selectsrc_socket(sin6, optp, inp,
829 		    td->td_ucred, scope_ambiguous, &in6a, NULL);
830 		if (error)
831 			goto release;
832 		laddr = &in6a;
833 
834 		if (inp->inp_lport == 0) {
835 			struct inpcbinfo *pcbinfo;
836 
837 			INP_WLOCK_ASSERT(inp);
838 
839 			pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
840 			INP_HASH_WLOCK(pcbinfo);
841 			error = in6_pcbsetport(laddr, inp, td->td_ucred);
842 			INP_HASH_WUNLOCK(pcbinfo);
843 			if (error != 0) {
844 				/* Undo an address bind that may have occurred. */
845 				inp->in6p_laddr = in6addr_any;
846 				goto release;
847 			}
848 		}
849 		faddr = &sin6->sin6_addr;
850 		fport = sin6->sin6_port; /* allow 0 port */
851 
852 	} else {
853 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
854 			error = ENOTCONN;
855 			goto release;
856 		}
857 		laddr = &inp->in6p_laddr;
858 		faddr = &inp->in6p_faddr;
859 		fport = inp->inp_fport;
860 	}
861 
862 
863 	/*
864 	 * We do not support IPv6 Jumbograms (RFC2675), so validate the payload
865 	 * length fits in a normal gram.
866 	 */
867 	ulen = m->m_pkthdr.len;
868 	plen = sizeof(struct udphdr) + ulen;
869 	hlen = sizeof(struct ip6_hdr);
870 
871 	if (plen > IPV6_MAXPAYLOAD) {
872 		m_freem(control);
873 		m_freem(m);
874 		return (EMSGSIZE);
875 	}
876 
877 	/*
878 	 * Calculate data length and get a mbuf for UDP, IP6, and possible
879 	 * link-layer headers.  Immediate slide the data pointer back forward
880 	 * since we won't use that space at this layer.
881 	 */
882 	M_PREPEND(m, hlen + sizeof(struct udphdr) + max_linkhdr, M_NOWAIT);
883 	if (m == NULL) {
884 		error = ENOBUFS;
885 		goto release;
886 	}
887 	m->m_data += max_linkhdr;
888 	m->m_len -= max_linkhdr;
889 	m->m_pkthdr.len -= max_linkhdr;
890 
891 	/*
892 	 * Stuff checksum and output datagram.
893 	 */
894 	cscov = cscov_partial = 0;
895 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
896 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
897 	udp6->uh_dport = fport;
898 	if (nxt == IPPROTO_UDPLITE) {
899 		struct udpcb *up;
900 
901 		up = intoudpcb(inp);
902 		cscov = up->u_txcslen;
903 		if (cscov >= plen)
904 			cscov = 0;
905 		udp6->uh_ulen = htons(cscov);
906 		/*
907 		 * For UDP-Lite, checksum coverage length of zero means
908 		 * the entire UDPLite packet is covered by the checksum.
909 		 */
910 		cscov_partial = (cscov == 0) ? 0 : 1;
911 	} else {
912 		MPASS(plen <= IPV6_MAXPAYLOAD);
913 		udp6->uh_ulen = htons((u_short)plen);
914 	}
915 	udp6->uh_sum = 0;
916 
917 	ip6 = mtod(m, struct ip6_hdr *);
918 	ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
919 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
920 	ip6->ip6_vfc	|= IPV6_VERSION;
921 	ip6->ip6_plen	= htons((u_short)plen);
922 	ip6->ip6_nxt	= nxt;
923 	ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
924 	ip6->ip6_src	= *laddr;
925 	ip6->ip6_dst	= *faddr;
926 
927 #ifdef MAC
928 	mac_inpcb_create_mbuf(inp, m);
929 #endif
930 
931 	if (cscov_partial) {
932 		if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
933 		    sizeof(struct ip6_hdr), plen, cscov)) == 0)
934 			udp6->uh_sum = 0xffff;
935 	} else {
936 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
937 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
938 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
939 	}
940 
941 	flags = 0;
942 	if (CALC_FLOWID_OUTBOUND_SENDTO) {
943 		uint32_t hash_type, hash_val;
944 		uint8_t pr;
945 
946 		pr = inp->inp_socket->so_proto->pr_protocol;
947 
948 		hash_val = fib6_calc_packet_hash(laddr, faddr,
949 		    inp->inp_lport, fport, pr, &hash_type);
950 		m->m_pkthdr.flowid = hash_val;
951 		M_HASHTYPE_SET(m, hash_type);
952 	}
953 	/* do not use inp flowid */
954 	flags |= IP_NODEFAULTFLOWID;
955 
956 	UDPSTAT_INC(udps_opackets);
957 	if (nxt == IPPROTO_UDPLITE)
958 		UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
959 	else
960 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
961 	error = ip6_output(m, optp,
962 	    INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
963 	    inp->in6p_moptions, NULL, inp);
964 	INP_UNLOCK(inp);
965 	NET_EPOCH_EXIT(et);
966 
967 	if (control) {
968 		ip6_clearpktopts(&opt, -1);
969 		m_freem(control);
970 	}
971 	return (error);
972 
973 release:
974 	INP_UNLOCK(inp);
975 	NET_EPOCH_EXIT(et);
976 	if (control) {
977 		ip6_clearpktopts(&opt, -1);
978 		m_freem(control);
979 	}
980 	m_freem(m);
981 
982 	return (error);
983 }
984 
985 static void
udp6_abort(struct socket * so)986 udp6_abort(struct socket *so)
987 {
988 	struct inpcb *inp;
989 	struct inpcbinfo *pcbinfo;
990 
991 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
992 	inp = sotoinpcb(so);
993 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
994 
995 	INP_WLOCK(inp);
996 #ifdef INET
997 	if (inp->inp_vflag & INP_IPV4) {
998 		INP_WUNLOCK(inp);
999 		udp_abort(so);
1000 		return;
1001 	}
1002 #endif
1003 
1004 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1005 		INP_HASH_WLOCK(pcbinfo);
1006 		in6_pcbdisconnect(inp);
1007 		INP_HASH_WUNLOCK(pcbinfo);
1008 		soisdisconnected(so);
1009 	}
1010 	INP_WUNLOCK(inp);
1011 }
1012 
1013 static int
udp6_attach(struct socket * so,int proto,struct thread * td)1014 udp6_attach(struct socket *so, int proto, struct thread *td)
1015 {
1016 	struct inpcbinfo *pcbinfo;
1017 	struct inpcb *inp;
1018 	struct udpcb *up;
1019 	int error;
1020 
1021 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1022 	inp = sotoinpcb(so);
1023 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1024 
1025 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1026 		error = soreserve(so, udp_sendspace, udp_recvspace);
1027 		if (error)
1028 			return (error);
1029 	}
1030 	error = in_pcballoc(so, pcbinfo);
1031 	if (error)
1032 		return (error);
1033 	inp = (struct inpcb *)so->so_pcb;
1034 	inp->in6p_cksum = -1;	/* just to be sure */
1035 	/*
1036 	 * XXX: ugly!!
1037 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1038 	 * because the socket may be bound to an IPv6 wildcard address,
1039 	 * which may match an IPv4-mapped IPv6 address.
1040 	 */
1041 	inp->inp_ip_ttl = V_ip_defttl;
1042 	up = intoudpcb(inp);
1043 	bzero(&up->u_start_zero, u_zero_size);
1044 	INP_WUNLOCK(inp);
1045 	return (0);
1046 }
1047 
1048 static int
udp6_bind(struct socket * so,struct sockaddr * nam,struct thread * td)1049 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1050 {
1051 	struct sockaddr_in6 *sin6_p;
1052 	struct inpcb *inp;
1053 	struct inpcbinfo *pcbinfo;
1054 	int error;
1055 	u_char vflagsav;
1056 
1057 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1058 	inp = sotoinpcb(so);
1059 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1060 
1061 	if (nam->sa_family != AF_INET6)
1062 		return (EAFNOSUPPORT);
1063 	if (nam->sa_len != sizeof(struct sockaddr_in6))
1064 		return (EINVAL);
1065 
1066 	sin6_p = (struct sockaddr_in6 *)nam;
1067 
1068 	INP_WLOCK(inp);
1069 	INP_HASH_WLOCK(pcbinfo);
1070 	vflagsav = inp->inp_vflag;
1071 	inp->inp_vflag &= ~INP_IPV4;
1072 	inp->inp_vflag |= INP_IPV6;
1073 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1074 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1075 			inp->inp_vflag |= INP_IPV4;
1076 #ifdef INET
1077 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1078 			struct sockaddr_in sin;
1079 
1080 			in6_sin6_2_sin(&sin, sin6_p);
1081 			inp->inp_vflag |= INP_IPV4;
1082 			inp->inp_vflag &= ~INP_IPV6;
1083 			error = in_pcbbind(inp, &sin,
1084 			    V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1085 			    td->td_ucred);
1086 			goto out;
1087 		}
1088 #endif
1089 	}
1090 
1091 	error = in6_pcbbind(inp, sin6_p, V_udp_bind_all_fibs ? 0 : INPBIND_FIB,
1092 	    td->td_ucred);
1093 #ifdef INET
1094 out:
1095 #endif
1096 	if (error != 0)
1097 		inp->inp_vflag = vflagsav;
1098 	INP_HASH_WUNLOCK(pcbinfo);
1099 	INP_WUNLOCK(inp);
1100 	return (error);
1101 }
1102 
1103 static void
udp6_close(struct socket * so)1104 udp6_close(struct socket *so)
1105 {
1106 	struct inpcb *inp;
1107 	struct inpcbinfo *pcbinfo;
1108 
1109 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1110 	inp = sotoinpcb(so);
1111 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1112 
1113 	INP_WLOCK(inp);
1114 #ifdef INET
1115 	if (inp->inp_vflag & INP_IPV4) {
1116 		INP_WUNLOCK(inp);
1117 		(void)udp_disconnect(so);
1118 		return;
1119 	}
1120 #endif
1121 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1122 		INP_HASH_WLOCK(pcbinfo);
1123 		in6_pcbdisconnect(inp);
1124 		INP_HASH_WUNLOCK(pcbinfo);
1125 		soisdisconnected(so);
1126 	}
1127 	INP_WUNLOCK(inp);
1128 }
1129 
1130 static int
udp6_connect(struct socket * so,struct sockaddr * nam,struct thread * td)1131 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1132 {
1133 	struct epoch_tracker et;
1134 	struct inpcb *inp;
1135 	struct inpcbinfo *pcbinfo;
1136 	struct sockaddr_in6 *sin6;
1137 	int error;
1138 	u_char vflagsav;
1139 
1140 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1141 	inp = sotoinpcb(so);
1142 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1143 
1144 	sin6 = (struct sockaddr_in6 *)nam;
1145 	if (sin6->sin6_family != AF_INET6)
1146 		return (EAFNOSUPPORT);
1147 	if (sin6->sin6_len != sizeof(*sin6))
1148 		return (EINVAL);
1149 
1150 	/*
1151 	 * XXXRW: Need to clarify locking of v4/v6 flags.
1152 	 */
1153 	INP_WLOCK(inp);
1154 #ifdef INET
1155 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1156 		struct sockaddr_in sin;
1157 
1158 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1159 			error = EINVAL;
1160 			goto out;
1161 		}
1162 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1163 			error = EAFNOSUPPORT;
1164 			goto out;
1165 		}
1166 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
1167 			error = EISCONN;
1168 			goto out;
1169 		}
1170 		in6_sin6_2_sin(&sin, sin6);
1171 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1172 		if (error != 0)
1173 			goto out;
1174 		vflagsav = inp->inp_vflag;
1175 		inp->inp_vflag |= INP_IPV4;
1176 		inp->inp_vflag &= ~INP_IPV6;
1177 		NET_EPOCH_ENTER(et);
1178 		INP_HASH_WLOCK(pcbinfo);
1179 		error = in_pcbconnect(inp, &sin, td->td_ucred);
1180 		INP_HASH_WUNLOCK(pcbinfo);
1181 		NET_EPOCH_EXIT(et);
1182 		/*
1183 		 * If connect succeeds, mark socket as connected. If
1184 		 * connect fails and socket is unbound, reset inp_vflag
1185 		 * field.
1186 		 */
1187 		if (error == 0)
1188 			soisconnected(so);
1189 		else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1190 		    inp->inp_lport == 0)
1191 			inp->inp_vflag = vflagsav;
1192 		goto out;
1193 	} else {
1194 		if ((inp->inp_vflag & INP_IPV6) == 0) {
1195 			error = EAFNOSUPPORT;
1196 			goto out;
1197 		}
1198 	}
1199 #endif
1200 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1201 		error = EISCONN;
1202 		goto out;
1203 	}
1204 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1205 	if (error != 0)
1206 		goto out;
1207 	vflagsav = inp->inp_vflag;
1208 	inp->inp_vflag &= ~INP_IPV4;
1209 	inp->inp_vflag |= INP_IPV6;
1210 	NET_EPOCH_ENTER(et);
1211 	INP_HASH_WLOCK(pcbinfo);
1212 	error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
1213 	INP_HASH_WUNLOCK(pcbinfo);
1214 	NET_EPOCH_EXIT(et);
1215 	/*
1216 	 * If connect succeeds, mark socket as connected. If
1217 	 * connect fails and socket is unbound, reset inp_vflag
1218 	 * field.
1219 	 */
1220 	if (error == 0)
1221 		soisconnected(so);
1222 	else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1223 	    inp->inp_lport == 0)
1224 		inp->inp_vflag = vflagsav;
1225 out:
1226 	INP_WUNLOCK(inp);
1227 	return (error);
1228 }
1229 
1230 static void
udp6_detach(struct socket * so)1231 udp6_detach(struct socket *so)
1232 {
1233 	struct inpcb *inp;
1234 
1235 	inp = sotoinpcb(so);
1236 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1237 
1238 	INP_WLOCK(inp);
1239 	in_pcbfree(inp);
1240 }
1241 
1242 static int
udp6_disconnect(struct socket * so)1243 udp6_disconnect(struct socket *so)
1244 {
1245 	struct inpcb *inp;
1246 	struct inpcbinfo *pcbinfo;
1247 
1248 	pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1249 	inp = sotoinpcb(so);
1250 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1251 
1252 	INP_WLOCK(inp);
1253 #ifdef INET
1254 	if (inp->inp_vflag & INP_IPV4) {
1255 		INP_WUNLOCK(inp);
1256 		(void)udp_disconnect(so);
1257 		return (0);
1258 	}
1259 #endif
1260 
1261 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1262 		INP_WUNLOCK(inp);
1263 		return (ENOTCONN);
1264 	}
1265 
1266 	INP_HASH_WLOCK(pcbinfo);
1267 	in6_pcbdisconnect(inp);
1268 	INP_HASH_WUNLOCK(pcbinfo);
1269 	SOCK_LOCK(so);
1270 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1271 	SOCK_UNLOCK(so);
1272 	INP_WUNLOCK(inp);
1273 	return (0);
1274 }
1275 
1276 #define	UDP6_PROTOSW							\
1277 	.pr_type =		SOCK_DGRAM,				\
1278 	.pr_flags =		PR_ATOMIC|PR_ADDR|PR_CAPATTACH,		\
1279 	.pr_ctloutput =		udp_ctloutput,				\
1280 	.pr_abort =		udp6_abort,				\
1281 	.pr_attach =		udp6_attach,				\
1282 	.pr_bind =		udp6_bind,				\
1283 	.pr_connect =		udp6_connect,				\
1284 	.pr_control =		in6_control,				\
1285 	.pr_detach =		udp6_detach,				\
1286 	.pr_disconnect =	udp6_disconnect,			\
1287 	.pr_peeraddr =		in6_mapped_peeraddr,			\
1288 	.pr_send =		udp6_send,				\
1289 	.pr_shutdown =		udp_shutdown,				\
1290 	.pr_sockaddr =		in6_mapped_sockaddr,			\
1291 	.pr_soreceive =		soreceive_dgram,			\
1292 	.pr_sosend =		sosend_dgram,				\
1293 	.pr_sosetlabel =	in_pcbsosetlabel,			\
1294 	.pr_close =		udp6_close
1295 
1296 struct protosw udp6_protosw = {
1297 	.pr_protocol =		IPPROTO_UDP,
1298 	UDP6_PROTOSW
1299 };
1300 
1301 struct protosw udplite6_protosw = {
1302 	.pr_protocol =		IPPROTO_UDPLITE,
1303 	UDP6_PROTOSW
1304 };
1305 
1306 static void
udp6_init(void * arg __unused)1307 udp6_init(void *arg __unused)
1308 {
1309 
1310 	IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1311 	IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1312 }
1313 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);
1314