xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 6ef6ba9950260f42b47499d17874d00ca9290955)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * Copyright (c) 2010-2011 Juniper Networks, Inc.
4  * All rights reserved.
5  *
6  * Portions of this software were developed by Robert N. M. Watson under
7  * contract to Juniper Networks, Inc.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	$KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
34  *	$KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
35  */
36 
37 /*-
38  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
39  *	The Regents of the University of California.
40  * All rights reserved.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
67  */
68 
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71 
72 #include "opt_inet.h"
73 #include "opt_inet6.h"
74 #include "opt_ipfw.h"
75 #include "opt_ipsec.h"
76 #include "opt_kdtrace.h"
77 
78 #include <sys/param.h>
79 #include <sys/jail.h>
80 #include <sys/kernel.h>
81 #include <sys/lock.h>
82 #include <sys/mbuf.h>
83 #include <sys/priv.h>
84 #include <sys/proc.h>
85 #include <sys/protosw.h>
86 #include <sys/sdt.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sx.h>
91 #include <sys/sysctl.h>
92 #include <sys/syslog.h>
93 #include <sys/systm.h>
94 
95 #include <net/if.h>
96 #include <net/if_var.h>
97 #include <net/if_types.h>
98 #include <net/route.h>
99 
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/ip_icmp.h>
107 #include <netinet/ip6.h>
108 #include <netinet/icmp_var.h>
109 #include <netinet/icmp6.h>
110 #include <netinet/ip_var.h>
111 #include <netinet/udp.h>
112 #include <netinet/udp_var.h>
113 
114 #include <netinet6/ip6protosw.h>
115 #include <netinet6/ip6_var.h>
116 #include <netinet6/in6_pcb.h>
117 #include <netinet6/udp6_var.h>
118 #include <netinet6/scope6_var.h>
119 
120 #ifdef IPSEC
121 #include <netipsec/ipsec.h>
122 #include <netipsec/ipsec6.h>
123 #endif /* IPSEC */
124 
125 #include <security/mac/mac_framework.h>
126 
127 /*
128  * UDP protocol implementation.
129  * Per RFC 768, August, 1980.
130  */
131 
132 extern struct protosw	inetsw[];
133 static void		udp6_detach(struct socket *so);
134 
135 static void
136 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
137     struct sockaddr_in6 *fromsa)
138 {
139 	struct socket *so;
140 	struct mbuf *opts;
141 
142 	INP_LOCK_ASSERT(inp);
143 
144 #ifdef IPSEC
145 	/* Check AH/ESP integrity. */
146 	if (ipsec6_in_reject(n, inp)) {
147 		m_freem(n);
148 		IPSEC6STAT_INC(ips_in_polvio);
149 		return;
150 	}
151 #endif /* IPSEC */
152 #ifdef MAC
153 	if (mac_inpcb_check_deliver(inp, n) != 0) {
154 		m_freem(n);
155 		return;
156 	}
157 #endif
158 	opts = NULL;
159 	if (inp->inp_flags & INP_CONTROLOPTS ||
160 	    inp->inp_socket->so_options & SO_TIMESTAMP)
161 		ip6_savecontrol(inp, n, &opts);
162 	m_adj(n, off + sizeof(struct udphdr));
163 
164 	so = inp->inp_socket;
165 	SOCKBUF_LOCK(&so->so_rcv);
166 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
167 	    opts) == 0) {
168 		SOCKBUF_UNLOCK(&so->so_rcv);
169 		m_freem(n);
170 		if (opts)
171 			m_freem(opts);
172 		UDPSTAT_INC(udps_fullsock);
173 	} else
174 		sorwakeup_locked(so);
175 }
176 
177 int
178 udp6_input(struct mbuf **mp, int *offp, int proto)
179 {
180 	struct mbuf *m = *mp;
181 	struct ifnet *ifp;
182 	struct ip6_hdr *ip6;
183 	struct udphdr *uh;
184 	struct inpcb *inp;
185 	struct udpcb *up;
186 	int off = *offp;
187 	int plen, ulen;
188 	struct sockaddr_in6 fromsa;
189 	struct m_tag *fwd_tag;
190 	uint16_t uh_sum;
191 
192 	ifp = m->m_pkthdr.rcvif;
193 	ip6 = mtod(m, struct ip6_hdr *);
194 
195 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
196 		/* XXX send icmp6 host/port unreach? */
197 		m_freem(m);
198 		return (IPPROTO_DONE);
199 	}
200 
201 #ifndef PULLDOWN_TEST
202 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
203 	ip6 = mtod(m, struct ip6_hdr *);
204 	uh = (struct udphdr *)((caddr_t)ip6 + off);
205 #else
206 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
207 	if (!uh)
208 		return (IPPROTO_DONE);
209 #endif
210 
211 	UDPSTAT_INC(udps_ipackets);
212 
213 	/*
214 	 * Destination port of 0 is illegal, based on RFC768.
215 	 */
216 	if (uh->uh_dport == 0)
217 		goto badunlocked;
218 
219 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
220 	ulen = ntohs((u_short)uh->uh_ulen);
221 
222 	if (plen != ulen) {
223 		UDPSTAT_INC(udps_badlen);
224 		goto badunlocked;
225 	}
226 
227 	/*
228 	 * Checksum extended UDP header and data.
229 	 */
230 	if (uh->uh_sum == 0) {
231 		UDPSTAT_INC(udps_nosum);
232 		goto badunlocked;
233 	}
234 
235 	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) {
236 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
237 			uh_sum = m->m_pkthdr.csum_data;
238 		else
239 			uh_sum = in6_cksum_pseudo(ip6, ulen,
240 			    IPPROTO_UDP, m->m_pkthdr.csum_data);
241 		uh_sum ^= 0xffff;
242 	} else
243 		uh_sum = in6_cksum(m, IPPROTO_UDP, off, ulen);
244 
245 	if (uh_sum != 0) {
246 		UDPSTAT_INC(udps_badsum);
247 		goto badunlocked;
248 	}
249 
250 	/*
251 	 * Construct sockaddr format source address.
252 	 */
253 	init_sin6(&fromsa, m);
254 	fromsa.sin6_port = uh->uh_sport;
255 
256 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
257 		struct inpcb *last;
258 		struct ip6_moptions *imo;
259 
260 		INP_INFO_RLOCK(&V_udbinfo);
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 		last = NULL;
277 		LIST_FOREACH(inp, &V_udb, inp_list) {
278 			if ((inp->inp_vflag & INP_IPV6) == 0)
279 				continue;
280 			if (inp->inp_lport != uh->uh_dport)
281 				continue;
282 			if (inp->inp_fport != 0 &&
283 			    inp->inp_fport != uh->uh_sport)
284 				continue;
285 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
286 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
287 							&ip6->ip6_dst))
288 					continue;
289 			}
290 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
291 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
292 							&ip6->ip6_src) ||
293 				    inp->inp_fport != uh->uh_sport)
294 					continue;
295 			}
296 
297 			/*
298 			 * XXXRW: Because we weren't holding either the inpcb
299 			 * or the hash lock when we checked for a match
300 			 * before, we should probably recheck now that the
301 			 * inpcb lock is (supposed to be) held.
302 			 */
303 
304 			/*
305 			 * Handle socket delivery policy for any-source
306 			 * and source-specific multicast. [RFC3678]
307 			 */
308 			imo = inp->in6p_moptions;
309 			if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
310 				struct sockaddr_in6	 mcaddr;
311 				int			 blocked;
312 
313 				INP_RLOCK(inp);
314 
315 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
316 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
317 				mcaddr.sin6_family = AF_INET6;
318 				mcaddr.sin6_addr = ip6->ip6_dst;
319 
320 				blocked = im6o_mc_filter(imo, ifp,
321 					(struct sockaddr *)&mcaddr,
322 					(struct sockaddr *)&fromsa);
323 				if (blocked != MCAST_PASS) {
324 					if (blocked == MCAST_NOTGMEMBER)
325 						IP6STAT_INC(ip6s_notmember);
326 					if (blocked == MCAST_NOTSMEMBER ||
327 					    blocked == MCAST_MUTED)
328 						UDPSTAT_INC(udps_filtermcast);
329 					INP_RUNLOCK(inp); /* XXX */
330 					continue;
331 				}
332 
333 				INP_RUNLOCK(inp);
334 			}
335 			if (last != NULL) {
336 				struct mbuf *n;
337 
338 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
339 					INP_RLOCK(last);
340 					up = intoudpcb(last);
341 					if (up->u_tun_func == NULL) {
342 						udp6_append(last, n, off, &fromsa);
343 					} else {
344 						/*
345 						 * Engage the tunneling
346 						 * protocol we will have to
347 						 * leave the info_lock up,
348 						 * since we are hunting
349 						 * through multiple UDP's.
350 						 *
351 						 */
352 						(*up->u_tun_func)(n, off, last);
353 					}
354 					INP_RUNLOCK(last);
355 				}
356 			}
357 			last = inp;
358 			/*
359 			 * Don't look for additional matches if this one does
360 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
361 			 * socket options set.  This heuristic avoids
362 			 * searching through all pcbs in the common case of a
363 			 * non-shared port.  It assumes that an application
364 			 * will never clear these options after setting them.
365 			 */
366 			if ((last->inp_socket->so_options &
367 			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
368 				break;
369 		}
370 
371 		if (last == NULL) {
372 			/*
373 			 * No matching pcb found; discard datagram.  (No need
374 			 * to send an ICMP Port Unreachable for a broadcast
375 			 * or multicast datgram.)
376 			 */
377 			UDPSTAT_INC(udps_noport);
378 			UDPSTAT_INC(udps_noportmcast);
379 			goto badheadlocked;
380 		}
381 		INP_RLOCK(last);
382 		INP_INFO_RUNLOCK(&V_udbinfo);
383 		up = intoudpcb(last);
384 		UDP_PROBE(receive, NULL, last, ip6, last, uh);
385 		if (up->u_tun_func == NULL) {
386 			udp6_append(last, m, off, &fromsa);
387 		} else {
388 			/*
389 			 * Engage the tunneling protocol.
390 			 */
391 			(*up->u_tun_func)(m, off, last);
392 		}
393 		INP_RUNLOCK(last);
394 		return (IPPROTO_DONE);
395 	}
396 	/*
397 	 * Locate pcb for datagram.
398 	 */
399 
400 	/*
401 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
402 	 */
403 	if ((m->m_flags & M_IP6_NEXTHOP) &&
404 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
405 		struct sockaddr_in6 *next_hop6;
406 
407 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
408 
409 		/*
410 		 * Transparently forwarded. Pretend to be the destination.
411 		 * Already got one like this?
412 		 */
413 		inp = in6_pcblookup_mbuf(&V_udbinfo,
414 		    &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
415 		    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
416 		if (!inp) {
417 			/*
418 			 * It's new.  Try to find the ambushing socket.
419 			 * Because we've rewritten the destination address,
420 			 * any hardware-generated hash is ignored.
421 			 */
422 			inp = in6_pcblookup(&V_udbinfo, &ip6->ip6_src,
423 			    uh->uh_sport, &next_hop6->sin6_addr,
424 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
425 			    uh->uh_dport, INPLOOKUP_WILDCARD |
426 			    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
427 		}
428 		/* Remove the tag from the packet. We don't need it anymore. */
429 		m_tag_delete(m, fwd_tag);
430 		m->m_flags &= ~M_IP6_NEXTHOP;
431 	} else
432 		inp = in6_pcblookup_mbuf(&V_udbinfo, &ip6->ip6_src,
433 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
434 		    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
435 		    m->m_pkthdr.rcvif, m);
436 	if (inp == NULL) {
437 		if (udp_log_in_vain) {
438 			char ip6bufs[INET6_ADDRSTRLEN];
439 			char ip6bufd[INET6_ADDRSTRLEN];
440 
441 			log(LOG_INFO,
442 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
443 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
444 			    ntohs(uh->uh_dport),
445 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
446 			    ntohs(uh->uh_sport));
447 		}
448 		UDPSTAT_INC(udps_noport);
449 		if (m->m_flags & M_MCAST) {
450 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
451 			UDPSTAT_INC(udps_noportmcast);
452 			goto badunlocked;
453 		}
454 		if (V_udp_blackhole)
455 			goto badunlocked;
456 		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
457 			goto badunlocked;
458 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
459 		return (IPPROTO_DONE);
460 	}
461 	INP_RLOCK_ASSERT(inp);
462 	up = intoudpcb(inp);
463 	UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
464 	if (up->u_tun_func == NULL) {
465 		udp6_append(inp, m, off, &fromsa);
466 	} else {
467 		/*
468 		 * Engage the tunneling protocol.
469 		 */
470 
471 		(*up->u_tun_func)(m, off, inp);
472 	}
473 	INP_RUNLOCK(inp);
474 	return (IPPROTO_DONE);
475 
476 badheadlocked:
477 	INP_INFO_RUNLOCK(&V_udbinfo);
478 badunlocked:
479 	if (m)
480 		m_freem(m);
481 	return (IPPROTO_DONE);
482 }
483 
484 void
485 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
486 {
487 	struct udphdr uh;
488 	struct ip6_hdr *ip6;
489 	struct mbuf *m;
490 	int off = 0;
491 	struct ip6ctlparam *ip6cp = NULL;
492 	const struct sockaddr_in6 *sa6_src = NULL;
493 	void *cmdarg;
494 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
495 	struct udp_portonly {
496 		u_int16_t uh_sport;
497 		u_int16_t uh_dport;
498 	} *uhp;
499 
500 	if (sa->sa_family != AF_INET6 ||
501 	    sa->sa_len != sizeof(struct sockaddr_in6))
502 		return;
503 
504 	if ((unsigned)cmd >= PRC_NCMDS)
505 		return;
506 	if (PRC_IS_REDIRECT(cmd))
507 		notify = in6_rtchange, d = NULL;
508 	else if (cmd == PRC_HOSTDEAD)
509 		d = NULL;
510 	else if (inet6ctlerrmap[cmd] == 0)
511 		return;
512 
513 	/* if the parameter is from icmp6, decode it. */
514 	if (d != NULL) {
515 		ip6cp = (struct ip6ctlparam *)d;
516 		m = ip6cp->ip6c_m;
517 		ip6 = ip6cp->ip6c_ip6;
518 		off = ip6cp->ip6c_off;
519 		cmdarg = ip6cp->ip6c_cmdarg;
520 		sa6_src = ip6cp->ip6c_src;
521 	} else {
522 		m = NULL;
523 		ip6 = NULL;
524 		cmdarg = NULL;
525 		sa6_src = &sa6_any;
526 	}
527 
528 	if (ip6) {
529 		/*
530 		 * XXX: We assume that when IPV6 is non NULL,
531 		 * M and OFF are valid.
532 		 */
533 
534 		/* Check if we can safely examine src and dst ports. */
535 		if (m->m_pkthdr.len < off + sizeof(*uhp))
536 			return;
537 
538 		bzero(&uh, sizeof(uh));
539 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
540 
541 		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
542 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
543 		    cmdarg, notify);
544 	} else
545 		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
546 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
547 }
548 
549 static int
550 udp6_getcred(SYSCTL_HANDLER_ARGS)
551 {
552 	struct xucred xuc;
553 	struct sockaddr_in6 addrs[2];
554 	struct inpcb *inp;
555 	int error;
556 
557 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
558 	if (error)
559 		return (error);
560 
561 	if (req->newlen != sizeof(addrs))
562 		return (EINVAL);
563 	if (req->oldlen != sizeof(struct xucred))
564 		return (EINVAL);
565 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
566 	if (error)
567 		return (error);
568 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
569 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
570 		return (error);
571 	}
572 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
573 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
574 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
575 	if (inp != NULL) {
576 		INP_RLOCK_ASSERT(inp);
577 		if (inp->inp_socket == NULL)
578 			error = ENOENT;
579 		if (error == 0)
580 			error = cr_canseesocket(req->td->td_ucred,
581 			    inp->inp_socket);
582 		if (error == 0)
583 			cru2x(inp->inp_cred, &xuc);
584 		INP_RUNLOCK(inp);
585 	} else
586 		error = ENOENT;
587 	if (error == 0)
588 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
589 	return (error);
590 }
591 
592 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
593     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
594 
595 static int
596 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
597     struct mbuf *control, struct thread *td)
598 {
599 	u_int32_t ulen = m->m_pkthdr.len;
600 	u_int32_t plen = sizeof(struct udphdr) + ulen;
601 	struct ip6_hdr *ip6;
602 	struct udphdr *udp6;
603 	struct in6_addr *laddr, *faddr, in6a;
604 	struct sockaddr_in6 *sin6 = NULL;
605 	struct ifnet *oifp = NULL;
606 	int scope_ambiguous = 0;
607 	u_short fport;
608 	int error = 0;
609 	struct ip6_pktopts *optp, opt;
610 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
611 	int flags;
612 	struct sockaddr_in6 tmp;
613 
614 	INP_WLOCK_ASSERT(inp);
615 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
616 
617 	if (addr6) {
618 		/* addr6 has been validated in udp6_send(). */
619 		sin6 = (struct sockaddr_in6 *)addr6;
620 
621 		/* protect *sin6 from overwrites */
622 		tmp = *sin6;
623 		sin6 = &tmp;
624 
625 		/*
626 		 * Application should provide a proper zone ID or the use of
627 		 * default zone IDs should be enabled.  Unfortunately, some
628 		 * applications do not behave as it should, so we need a
629 		 * workaround.  Even if an appropriate ID is not determined,
630 		 * we'll see if we can determine the outgoing interface.  If we
631 		 * can, determine the zone ID based on the interface below.
632 		 */
633 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
634 			scope_ambiguous = 1;
635 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
636 			return (error);
637 	}
638 
639 	if (control) {
640 		if ((error = ip6_setpktopts(control, &opt,
641 		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
642 			goto release;
643 		optp = &opt;
644 	} else
645 		optp = inp->in6p_outputopts;
646 
647 	if (sin6) {
648 		faddr = &sin6->sin6_addr;
649 
650 		/*
651 		 * Since we saw no essential reason for calling in_pcbconnect,
652 		 * we get rid of such kind of logic, and call in6_selectsrc
653 		 * and in6_pcbsetport in order to fill in the local address
654 		 * and the local port.
655 		 */
656 		if (sin6->sin6_port == 0) {
657 			error = EADDRNOTAVAIL;
658 			goto release;
659 		}
660 
661 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
662 			/* how about ::ffff:0.0.0.0 case? */
663 			error = EISCONN;
664 			goto release;
665 		}
666 
667 		fport = sin6->sin6_port; /* allow 0 port */
668 
669 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
670 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
671 				/*
672 				 * I believe we should explicitly discard the
673 				 * packet when mapped addresses are disabled,
674 				 * rather than send the packet as an IPv6 one.
675 				 * If we chose the latter approach, the packet
676 				 * might be sent out on the wire based on the
677 				 * default route, the situation which we'd
678 				 * probably want to avoid.
679 				 * (20010421 jinmei@kame.net)
680 				 */
681 				error = EINVAL;
682 				goto release;
683 			}
684 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
685 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
686 				/*
687 				 * when remote addr is an IPv4-mapped address,
688 				 * local addr should not be an IPv6 address,
689 				 * since you cannot determine how to map IPv6
690 				 * source address to IPv4.
691 				 */
692 				error = EINVAL;
693 				goto release;
694 			}
695 
696 			af = AF_INET;
697 		}
698 
699 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
700 			error = in6_selectsrc(sin6, optp, inp, NULL,
701 			    td->td_ucred, &oifp, &in6a);
702 			if (error)
703 				goto release;
704 			if (oifp && scope_ambiguous &&
705 			    (error = in6_setscope(&sin6->sin6_addr,
706 			    oifp, NULL))) {
707 				goto release;
708 			}
709 			laddr = &in6a;
710 		} else
711 			laddr = &inp->in6p_laddr;	/* XXX */
712 		if (laddr == NULL) {
713 			if (error == 0)
714 				error = EADDRNOTAVAIL;
715 			goto release;
716 		}
717 		if (inp->inp_lport == 0 &&
718 		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
719 			/* Undo an address bind that may have occurred. */
720 			inp->in6p_laddr = in6addr_any;
721 			goto release;
722 		}
723 	} else {
724 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
725 			error = ENOTCONN;
726 			goto release;
727 		}
728 		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
729 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
730 				/*
731 				 * XXX: this case would happen when the
732 				 * application sets the V6ONLY flag after
733 				 * connecting the foreign address.
734 				 * Such applications should be fixed,
735 				 * so we bark here.
736 				 */
737 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
738 				    "option was set for a connected socket\n");
739 				error = EINVAL;
740 				goto release;
741 			} else
742 				af = AF_INET;
743 		}
744 		laddr = &inp->in6p_laddr;
745 		faddr = &inp->in6p_faddr;
746 		fport = inp->inp_fport;
747 	}
748 
749 	if (af == AF_INET)
750 		hlen = sizeof(struct ip);
751 
752 	/*
753 	 * Calculate data length and get a mbuf
754 	 * for UDP and IP6 headers.
755 	 */
756 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
757 	if (m == 0) {
758 		error = ENOBUFS;
759 		goto release;
760 	}
761 
762 	/*
763 	 * Stuff checksum and output datagram.
764 	 */
765 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
766 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
767 	udp6->uh_dport = fport;
768 	if (plen <= 0xffff)
769 		udp6->uh_ulen = htons((u_short)plen);
770 	else
771 		udp6->uh_ulen = 0;
772 	udp6->uh_sum = 0;
773 
774 	switch (af) {
775 	case AF_INET6:
776 		ip6 = mtod(m, struct ip6_hdr *);
777 		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
778 		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
779 		ip6->ip6_vfc	|= IPV6_VERSION;
780 		ip6->ip6_plen	= htons((u_short)plen);
781 		ip6->ip6_nxt	= IPPROTO_UDP;
782 		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
783 		ip6->ip6_src	= *laddr;
784 		ip6->ip6_dst	= *faddr;
785 
786 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, IPPROTO_UDP, 0);
787 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
788 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
789 
790 		flags = 0;
791 
792 		UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
793 		UDPSTAT_INC(udps_opackets);
794 		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
795 		    NULL, inp);
796 		break;
797 	case AF_INET:
798 		error = EAFNOSUPPORT;
799 		goto release;
800 	}
801 	goto releaseopt;
802 
803 release:
804 	m_freem(m);
805 
806 releaseopt:
807 	if (control) {
808 		ip6_clearpktopts(&opt, -1);
809 		m_freem(control);
810 	}
811 	return (error);
812 }
813 
814 static void
815 udp6_abort(struct socket *so)
816 {
817 	struct inpcb *inp;
818 
819 	inp = sotoinpcb(so);
820 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
821 
822 #ifdef INET
823 	if (inp->inp_vflag & INP_IPV4) {
824 		struct pr_usrreqs *pru;
825 
826 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
827 		(*pru->pru_abort)(so);
828 		return;
829 	}
830 #endif
831 
832 	INP_WLOCK(inp);
833 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
834 		INP_HASH_WLOCK(&V_udbinfo);
835 		in6_pcbdisconnect(inp);
836 		inp->in6p_laddr = in6addr_any;
837 		INP_HASH_WUNLOCK(&V_udbinfo);
838 		soisdisconnected(so);
839 	}
840 	INP_WUNLOCK(inp);
841 }
842 
843 static int
844 udp6_attach(struct socket *so, int proto, struct thread *td)
845 {
846 	struct inpcb *inp;
847 	int error;
848 
849 	inp = sotoinpcb(so);
850 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
851 
852 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
853 		error = soreserve(so, udp_sendspace, udp_recvspace);
854 		if (error)
855 			return (error);
856 	}
857 	INP_INFO_WLOCK(&V_udbinfo);
858 	error = in_pcballoc(so, &V_udbinfo);
859 	if (error) {
860 		INP_INFO_WUNLOCK(&V_udbinfo);
861 		return (error);
862 	}
863 	inp = (struct inpcb *)so->so_pcb;
864 	inp->inp_vflag |= INP_IPV6;
865 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
866 		inp->inp_vflag |= INP_IPV4;
867 	inp->in6p_hops = -1;	/* use kernel default */
868 	inp->in6p_cksum = -1;	/* just to be sure */
869 	/*
870 	 * XXX: ugly!!
871 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
872 	 * because the socket may be bound to an IPv6 wildcard address,
873 	 * which may match an IPv4-mapped IPv6 address.
874 	 */
875 	inp->inp_ip_ttl = V_ip_defttl;
876 
877 	error = udp_newudpcb(inp);
878 	if (error) {
879 		in_pcbdetach(inp);
880 		in_pcbfree(inp);
881 		INP_INFO_WUNLOCK(&V_udbinfo);
882 		return (error);
883 	}
884 	INP_WUNLOCK(inp);
885 	INP_INFO_WUNLOCK(&V_udbinfo);
886 	return (0);
887 }
888 
889 static int
890 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
891 {
892 	struct inpcb *inp;
893 	int error;
894 
895 	inp = sotoinpcb(so);
896 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
897 
898 	INP_WLOCK(inp);
899 	INP_HASH_WLOCK(&V_udbinfo);
900 	inp->inp_vflag &= ~INP_IPV4;
901 	inp->inp_vflag |= INP_IPV6;
902 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
903 		struct sockaddr_in6 *sin6_p;
904 
905 		sin6_p = (struct sockaddr_in6 *)nam;
906 
907 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
908 			inp->inp_vflag |= INP_IPV4;
909 #ifdef INET
910 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
911 			struct sockaddr_in sin;
912 
913 			in6_sin6_2_sin(&sin, sin6_p);
914 			inp->inp_vflag |= INP_IPV4;
915 			inp->inp_vflag &= ~INP_IPV6;
916 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
917 			    td->td_ucred);
918 			goto out;
919 		}
920 #endif
921 	}
922 
923 	error = in6_pcbbind(inp, nam, td->td_ucred);
924 #ifdef INET
925 out:
926 #endif
927 	INP_HASH_WUNLOCK(&V_udbinfo);
928 	INP_WUNLOCK(inp);
929 	return (error);
930 }
931 
932 static void
933 udp6_close(struct socket *so)
934 {
935 	struct inpcb *inp;
936 
937 	inp = sotoinpcb(so);
938 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
939 
940 #ifdef INET
941 	if (inp->inp_vflag & INP_IPV4) {
942 		struct pr_usrreqs *pru;
943 
944 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
945 		(*pru->pru_disconnect)(so);
946 		return;
947 	}
948 #endif
949 	INP_WLOCK(inp);
950 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
951 		INP_HASH_WLOCK(&V_udbinfo);
952 		in6_pcbdisconnect(inp);
953 		inp->in6p_laddr = in6addr_any;
954 		INP_HASH_WUNLOCK(&V_udbinfo);
955 		soisdisconnected(so);
956 	}
957 	INP_WUNLOCK(inp);
958 }
959 
960 static int
961 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
962 {
963 	struct inpcb *inp;
964 	struct sockaddr_in6 *sin6;
965 	int error;
966 
967 	inp = sotoinpcb(so);
968 	sin6 = (struct sockaddr_in6 *)nam;
969 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
970 
971 	/*
972 	 * XXXRW: Need to clarify locking of v4/v6 flags.
973 	 */
974 	INP_WLOCK(inp);
975 #ifdef INET
976 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
977 		struct sockaddr_in sin;
978 
979 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
980 			error = EINVAL;
981 			goto out;
982 		}
983 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
984 			error = EISCONN;
985 			goto out;
986 		}
987 		in6_sin6_2_sin(&sin, sin6);
988 		inp->inp_vflag |= INP_IPV4;
989 		inp->inp_vflag &= ~INP_IPV6;
990 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
991 		if (error != 0)
992 			goto out;
993 		INP_HASH_WLOCK(&V_udbinfo);
994 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
995 		    td->td_ucred);
996 		INP_HASH_WUNLOCK(&V_udbinfo);
997 		if (error == 0)
998 			soisconnected(so);
999 		goto out;
1000 	}
1001 #endif
1002 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1003 		error = EISCONN;
1004 		goto out;
1005 	}
1006 	inp->inp_vflag &= ~INP_IPV4;
1007 	inp->inp_vflag |= INP_IPV6;
1008 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1009 	if (error != 0)
1010 		goto out;
1011 	INP_HASH_WLOCK(&V_udbinfo);
1012 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1013 	INP_HASH_WUNLOCK(&V_udbinfo);
1014 	if (error == 0)
1015 		soisconnected(so);
1016 out:
1017 	INP_WUNLOCK(inp);
1018 	return (error);
1019 }
1020 
1021 static void
1022 udp6_detach(struct socket *so)
1023 {
1024 	struct inpcb *inp;
1025 	struct udpcb *up;
1026 
1027 	inp = sotoinpcb(so);
1028 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1029 
1030 	INP_INFO_WLOCK(&V_udbinfo);
1031 	INP_WLOCK(inp);
1032 	up = intoudpcb(inp);
1033 	KASSERT(up != NULL, ("%s: up == NULL", __func__));
1034 	in_pcbdetach(inp);
1035 	in_pcbfree(inp);
1036 	INP_INFO_WUNLOCK(&V_udbinfo);
1037 	udp_discardcb(up);
1038 }
1039 
1040 static int
1041 udp6_disconnect(struct socket *so)
1042 {
1043 	struct inpcb *inp;
1044 	int error;
1045 
1046 	inp = sotoinpcb(so);
1047 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1048 
1049 #ifdef INET
1050 	if (inp->inp_vflag & INP_IPV4) {
1051 		struct pr_usrreqs *pru;
1052 
1053 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1054 		(void)(*pru->pru_disconnect)(so);
1055 		return (0);
1056 	}
1057 #endif
1058 
1059 	INP_WLOCK(inp);
1060 
1061 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1062 		error = ENOTCONN;
1063 		goto out;
1064 	}
1065 
1066 	INP_HASH_WLOCK(&V_udbinfo);
1067 	in6_pcbdisconnect(inp);
1068 	inp->in6p_laddr = in6addr_any;
1069 	INP_HASH_WUNLOCK(&V_udbinfo);
1070 	SOCK_LOCK(so);
1071 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1072 	SOCK_UNLOCK(so);
1073 out:
1074 	INP_WUNLOCK(inp);
1075 	return (0);
1076 }
1077 
1078 static int
1079 udp6_send(struct socket *so, int flags, struct mbuf *m,
1080     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1081 {
1082 	struct inpcb *inp;
1083 	int error = 0;
1084 
1085 	inp = sotoinpcb(so);
1086 	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1087 
1088 	INP_WLOCK(inp);
1089 	if (addr) {
1090 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1091 			error = EINVAL;
1092 			goto bad;
1093 		}
1094 		if (addr->sa_family != AF_INET6) {
1095 			error = EAFNOSUPPORT;
1096 			goto bad;
1097 		}
1098 	}
1099 
1100 #ifdef INET
1101 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1102 		int hasv4addr;
1103 		struct sockaddr_in6 *sin6 = 0;
1104 
1105 		if (addr == 0)
1106 			hasv4addr = (inp->inp_vflag & INP_IPV4);
1107 		else {
1108 			sin6 = (struct sockaddr_in6 *)addr;
1109 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1110 			    ? 1 : 0;
1111 		}
1112 		if (hasv4addr) {
1113 			struct pr_usrreqs *pru;
1114 
1115 			/*
1116 			 * XXXRW: We release UDP-layer locks before calling
1117 			 * udp_send() in order to avoid recursion.  However,
1118 			 * this does mean there is a short window where inp's
1119 			 * fields are unstable.  Could this lead to a
1120 			 * potential race in which the factors causing us to
1121 			 * select the UDPv4 output routine are invalidated?
1122 			 */
1123 			INP_WUNLOCK(inp);
1124 			if (sin6)
1125 				in6_sin6_2_sin_in_sock(addr);
1126 			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1127 			/* addr will just be freed in sendit(). */
1128 			return ((*pru->pru_send)(so, flags, m, addr, control,
1129 			    td));
1130 		}
1131 	}
1132 #endif
1133 #ifdef MAC
1134 	mac_inpcb_create_mbuf(inp, m);
1135 #endif
1136 	INP_HASH_WLOCK(&V_udbinfo);
1137 	error = udp6_output(inp, m, addr, control, td);
1138 	INP_HASH_WUNLOCK(&V_udbinfo);
1139 #ifdef INET
1140 #endif
1141 	INP_WUNLOCK(inp);
1142 	return (error);
1143 
1144 bad:
1145 	INP_WUNLOCK(inp);
1146 	m_freem(m);
1147 	return (error);
1148 }
1149 
1150 struct pr_usrreqs udp6_usrreqs = {
1151 	.pru_abort =		udp6_abort,
1152 	.pru_attach =		udp6_attach,
1153 	.pru_bind =		udp6_bind,
1154 	.pru_connect =		udp6_connect,
1155 	.pru_control =		in6_control,
1156 	.pru_detach =		udp6_detach,
1157 	.pru_disconnect =	udp6_disconnect,
1158 	.pru_peeraddr =		in6_mapped_peeraddr,
1159 	.pru_send =		udp6_send,
1160 	.pru_shutdown =		udp_shutdown,
1161 	.pru_sockaddr =		in6_mapped_sockaddr,
1162 	.pru_soreceive =	soreceive_dgram,
1163 	.pru_sosend =		sosend_dgram,
1164 	.pru_sosetlabel =	in_pcbsosetlabel,
1165 	.pru_close =		udp6_close
1166 };
1167