xref: /freebsd/sys/netinet6/udp6_usrreq.c (revision 298cf604ccf133b101c6fad42d1a078a1fac58ca)
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 
77 #include <sys/param.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/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_types.h>
95 #include <net/route.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/in_var.h>
101 #include <netinet/ip.h>
102 #include <netinet/ip_icmp.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp_var.h>
105 #include <netinet/icmp6.h>
106 #include <netinet/ip_var.h>
107 #include <netinet/udp.h>
108 #include <netinet/udp_var.h>
109 
110 #include <netinet6/ip6protosw.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/in6_pcb.h>
113 #include <netinet6/udp6_var.h>
114 #include <netinet6/scope6_var.h>
115 
116 #ifdef IPSEC
117 #include <netipsec/ipsec.h>
118 #include <netipsec/ipsec6.h>
119 #endif /* IPSEC */
120 
121 #include <security/mac/mac_framework.h>
122 
123 /*
124  * UDP protocol implementation.
125  * Per RFC 768, August, 1980.
126  */
127 
128 extern struct protosw	inetsw[];
129 static void		udp6_detach(struct socket *so);
130 
131 static void
132 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
133     struct sockaddr_in6 *fromsa)
134 {
135 	struct socket *so;
136 	struct mbuf *opts;
137 
138 	INP_LOCK_ASSERT(inp);
139 
140 #ifdef IPSEC
141 	/* Check AH/ESP integrity. */
142 	if (ipsec6_in_reject(n, inp)) {
143 		m_freem(n);
144 		V_ipsec6stat.in_polvio++;
145 		return;
146 	}
147 #endif /* IPSEC */
148 #ifdef MAC
149 	if (mac_inpcb_check_deliver(inp, n) != 0) {
150 		m_freem(n);
151 		return;
152 	}
153 #endif
154 	opts = NULL;
155 	if (inp->inp_flags & INP_CONTROLOPTS ||
156 	    inp->inp_socket->so_options & SO_TIMESTAMP)
157 		ip6_savecontrol(inp, n, &opts);
158 	m_adj(n, off + sizeof(struct udphdr));
159 
160 	so = inp->inp_socket;
161 	SOCKBUF_LOCK(&so->so_rcv);
162 	if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
163 	    opts) == 0) {
164 		SOCKBUF_UNLOCK(&so->so_rcv);
165 		m_freem(n);
166 		if (opts)
167 			m_freem(opts);
168 		UDPSTAT_INC(udps_fullsock);
169 	} else
170 		sorwakeup_locked(so);
171 }
172 
173 int
174 udp6_input(struct mbuf **mp, int *offp, int proto)
175 {
176 	struct mbuf *m = *mp;
177 	struct ifnet *ifp;
178 	struct ip6_hdr *ip6;
179 	struct udphdr *uh;
180 	struct inpcb *inp;
181 	struct udpcb *up;
182 	int off = *offp;
183 	int plen, ulen;
184 	struct sockaddr_in6 fromsa;
185 	struct m_tag *fwd_tag;
186 	uint16_t uh_sum;
187 
188 	ifp = m->m_pkthdr.rcvif;
189 	ip6 = mtod(m, struct ip6_hdr *);
190 
191 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
192 		/* XXX send icmp6 host/port unreach? */
193 		m_freem(m);
194 		return (IPPROTO_DONE);
195 	}
196 
197 #ifndef PULLDOWN_TEST
198 	IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
199 	ip6 = mtod(m, struct ip6_hdr *);
200 	uh = (struct udphdr *)((caddr_t)ip6 + off);
201 #else
202 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
203 	if (!uh)
204 		return (IPPROTO_DONE);
205 #endif
206 
207 	UDPSTAT_INC(udps_ipackets);
208 
209 	/*
210 	 * Destination port of 0 is illegal, based on RFC768.
211 	 */
212 	if (uh->uh_dport == 0)
213 		goto badunlocked;
214 
215 	plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
216 	ulen = ntohs((u_short)uh->uh_ulen);
217 
218 	if (plen != ulen) {
219 		UDPSTAT_INC(udps_badlen);
220 		goto badunlocked;
221 	}
222 
223 	/*
224 	 * Checksum extended UDP header and data.
225 	 */
226 	if (uh->uh_sum == 0) {
227 		UDPSTAT_INC(udps_nosum);
228 		goto badunlocked;
229 	}
230 
231 	if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) {
232 		if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
233 			uh_sum = m->m_pkthdr.csum_data;
234 		else
235 			uh_sum = in6_cksum_pseudo(ip6, ulen,
236 			    IPPROTO_UDP, m->m_pkthdr.csum_data);
237 		uh_sum ^= 0xffff;
238 	} else
239 		uh_sum = in6_cksum(m, IPPROTO_UDP, off, ulen);
240 
241 	if (uh_sum != 0) {
242 		UDPSTAT_INC(udps_badsum);
243 		goto badunlocked;
244 	}
245 
246 	/*
247 	 * Construct sockaddr format source address.
248 	 */
249 	init_sin6(&fromsa, m);
250 	fromsa.sin6_port = uh->uh_sport;
251 
252 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
253 		struct inpcb *last;
254 		struct ip6_moptions *imo;
255 
256 		INP_INFO_RLOCK(&V_udbinfo);
257 		/*
258 		 * In the event that laddr should be set to the link-local
259 		 * address (this happens in RIPng), the multicast address
260 		 * specified in the received packet will not match laddr.  To
261 		 * handle this situation, matching is relaxed if the
262 		 * receiving interface is the same as one specified in the
263 		 * socket and if the destination multicast address matches
264 		 * one of the multicast groups specified in the socket.
265 		 */
266 
267 		/*
268 		 * KAME note: traditionally we dropped udpiphdr from mbuf
269 		 * here.  We need udphdr for IPsec processing so we do that
270 		 * later.
271 		 */
272 		last = NULL;
273 		LIST_FOREACH(inp, &V_udb, inp_list) {
274 			if ((inp->inp_vflag & INP_IPV6) == 0)
275 				continue;
276 			if (inp->inp_lport != uh->uh_dport)
277 				continue;
278 			if (inp->inp_fport != 0 &&
279 			    inp->inp_fport != uh->uh_sport)
280 				continue;
281 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
282 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
283 							&ip6->ip6_dst))
284 					continue;
285 			}
286 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
287 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
288 							&ip6->ip6_src) ||
289 				    inp->inp_fport != uh->uh_sport)
290 					continue;
291 			}
292 
293 			/*
294 			 * XXXRW: Because we weren't holding either the inpcb
295 			 * or the hash lock when we checked for a match
296 			 * before, we should probably recheck now that the
297 			 * inpcb lock is (supposed to be) held.
298 			 */
299 
300 			/*
301 			 * Handle socket delivery policy for any-source
302 			 * and source-specific multicast. [RFC3678]
303 			 */
304 			imo = inp->in6p_moptions;
305 			if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
306 				struct sockaddr_in6	 mcaddr;
307 				int			 blocked;
308 
309 				INP_RLOCK(inp);
310 
311 				bzero(&mcaddr, sizeof(struct sockaddr_in6));
312 				mcaddr.sin6_len = sizeof(struct sockaddr_in6);
313 				mcaddr.sin6_family = AF_INET6;
314 				mcaddr.sin6_addr = ip6->ip6_dst;
315 
316 				blocked = im6o_mc_filter(imo, ifp,
317 					(struct sockaddr *)&mcaddr,
318 					(struct sockaddr *)&fromsa);
319 				if (blocked != MCAST_PASS) {
320 					if (blocked == MCAST_NOTGMEMBER)
321 						IP6STAT_INC(ip6s_notmember);
322 					if (blocked == MCAST_NOTSMEMBER ||
323 					    blocked == MCAST_MUTED)
324 						UDPSTAT_INC(udps_filtermcast);
325 					INP_RUNLOCK(inp); /* XXX */
326 					continue;
327 				}
328 
329 				INP_RUNLOCK(inp);
330 			}
331 			if (last != NULL) {
332 				struct mbuf *n;
333 
334 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
335 					INP_RLOCK(last);
336 					up = intoudpcb(last);
337 					if (up->u_tun_func == NULL) {
338 						udp6_append(last, n, off, &fromsa);
339 					} else {
340 						/*
341 						 * Engage the tunneling
342 						 * protocol we will have to
343 						 * leave the info_lock up,
344 						 * since we are hunting
345 						 * through multiple UDP's.
346 						 *
347 						 */
348 						(*up->u_tun_func)(n, off, last);
349 					}
350 					INP_RUNLOCK(last);
351 				}
352 			}
353 			last = inp;
354 			/*
355 			 * Don't look for additional matches if this one does
356 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
357 			 * socket options set.  This heuristic avoids
358 			 * searching through all pcbs in the common case of a
359 			 * non-shared port.  It assumes that an application
360 			 * will never clear these options after setting them.
361 			 */
362 			if ((last->inp_socket->so_options &
363 			     (SO_REUSEPORT|SO_REUSEADDR)) == 0)
364 				break;
365 		}
366 
367 		if (last == NULL) {
368 			/*
369 			 * No matching pcb found; discard datagram.  (No need
370 			 * to send an ICMP Port Unreachable for a broadcast
371 			 * or multicast datgram.)
372 			 */
373 			UDPSTAT_INC(udps_noport);
374 			UDPSTAT_INC(udps_noportmcast);
375 			goto badheadlocked;
376 		}
377 		INP_RLOCK(last);
378 		INP_INFO_RUNLOCK(&V_udbinfo);
379 		up = intoudpcb(last);
380 		if (up->u_tun_func == NULL) {
381 			udp6_append(last, m, off, &fromsa);
382 		} else {
383 			/*
384 			 * Engage the tunneling protocol.
385 			 */
386 			(*up->u_tun_func)(m, off, last);
387 		}
388 		INP_RUNLOCK(last);
389 		return (IPPROTO_DONE);
390 	}
391 	/*
392 	 * Locate pcb for datagram.
393 	 */
394 
395 	/*
396 	 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
397 	 */
398 	if ((m->m_flags & M_IP6_NEXTHOP) &&
399 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
400 		struct sockaddr_in6 *next_hop6;
401 
402 		next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
403 
404 		/*
405 		 * Transparently forwarded. Pretend to be the destination.
406 		 * Already got one like this?
407 		 */
408 		inp = in6_pcblookup_mbuf(&V_udbinfo,
409 		    &ip6->ip6_src, uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
410 		    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
411 		if (!inp) {
412 			/*
413 			 * It's new.  Try to find the ambushing socket.
414 			 * Because we've rewritten the destination address,
415 			 * any hardware-generated hash is ignored.
416 			 */
417 			inp = in6_pcblookup(&V_udbinfo, &ip6->ip6_src,
418 			    uh->uh_sport, &next_hop6->sin6_addr,
419 			    next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
420 			    uh->uh_dport, INPLOOKUP_WILDCARD |
421 			    INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
422 		}
423 		/* Remove the tag from the packet. We don't need it anymore. */
424 		m_tag_delete(m, fwd_tag);
425 		m->m_flags &= ~M_IP6_NEXTHOP;
426 	} else
427 		inp = in6_pcblookup_mbuf(&V_udbinfo, &ip6->ip6_src,
428 		    uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
429 		    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
430 		    m->m_pkthdr.rcvif, m);
431 	if (inp == NULL) {
432 		if (udp_log_in_vain) {
433 			char ip6bufs[INET6_ADDRSTRLEN];
434 			char ip6bufd[INET6_ADDRSTRLEN];
435 
436 			log(LOG_INFO,
437 			    "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
438 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
439 			    ntohs(uh->uh_dport),
440 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
441 			    ntohs(uh->uh_sport));
442 		}
443 		UDPSTAT_INC(udps_noport);
444 		if (m->m_flags & M_MCAST) {
445 			printf("UDP6: M_MCAST is set in a unicast packet.\n");
446 			UDPSTAT_INC(udps_noportmcast);
447 			goto badunlocked;
448 		}
449 		if (V_udp_blackhole)
450 			goto badunlocked;
451 		if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
452 			goto badunlocked;
453 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
454 		return (IPPROTO_DONE);
455 	}
456 	INP_RLOCK_ASSERT(inp);
457 	up = intoudpcb(inp);
458 	if (up->u_tun_func == NULL) {
459 		udp6_append(inp, m, off, &fromsa);
460 	} else {
461 		/*
462 		 * Engage the tunneling protocol.
463 		 */
464 
465 		(*up->u_tun_func)(m, off, inp);
466 	}
467 	INP_RUNLOCK(inp);
468 	return (IPPROTO_DONE);
469 
470 badheadlocked:
471 	INP_INFO_RUNLOCK(&V_udbinfo);
472 badunlocked:
473 	if (m)
474 		m_freem(m);
475 	return (IPPROTO_DONE);
476 }
477 
478 void
479 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
480 {
481 	struct udphdr uh;
482 	struct ip6_hdr *ip6;
483 	struct mbuf *m;
484 	int off = 0;
485 	struct ip6ctlparam *ip6cp = NULL;
486 	const struct sockaddr_in6 *sa6_src = NULL;
487 	void *cmdarg;
488 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
489 	struct udp_portonly {
490 		u_int16_t uh_sport;
491 		u_int16_t uh_dport;
492 	} *uhp;
493 
494 	if (sa->sa_family != AF_INET6 ||
495 	    sa->sa_len != sizeof(struct sockaddr_in6))
496 		return;
497 
498 	if ((unsigned)cmd >= PRC_NCMDS)
499 		return;
500 	if (PRC_IS_REDIRECT(cmd))
501 		notify = in6_rtchange, d = NULL;
502 	else if (cmd == PRC_HOSTDEAD)
503 		d = NULL;
504 	else if (inet6ctlerrmap[cmd] == 0)
505 		return;
506 
507 	/* if the parameter is from icmp6, decode it. */
508 	if (d != NULL) {
509 		ip6cp = (struct ip6ctlparam *)d;
510 		m = ip6cp->ip6c_m;
511 		ip6 = ip6cp->ip6c_ip6;
512 		off = ip6cp->ip6c_off;
513 		cmdarg = ip6cp->ip6c_cmdarg;
514 		sa6_src = ip6cp->ip6c_src;
515 	} else {
516 		m = NULL;
517 		ip6 = NULL;
518 		cmdarg = NULL;
519 		sa6_src = &sa6_any;
520 	}
521 
522 	if (ip6) {
523 		/*
524 		 * XXX: We assume that when IPV6 is non NULL,
525 		 * M and OFF are valid.
526 		 */
527 
528 		/* Check if we can safely examine src and dst ports. */
529 		if (m->m_pkthdr.len < off + sizeof(*uhp))
530 			return;
531 
532 		bzero(&uh, sizeof(uh));
533 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
534 
535 		(void) in6_pcbnotify(&V_udbinfo, sa, uh.uh_dport,
536 		    (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
537 		    cmdarg, notify);
538 	} else
539 		(void) in6_pcbnotify(&V_udbinfo, sa, 0,
540 		    (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
541 }
542 
543 static int
544 udp6_getcred(SYSCTL_HANDLER_ARGS)
545 {
546 	struct xucred xuc;
547 	struct sockaddr_in6 addrs[2];
548 	struct inpcb *inp;
549 	int error;
550 
551 	error = priv_check(req->td, PRIV_NETINET_GETCRED);
552 	if (error)
553 		return (error);
554 
555 	if (req->newlen != sizeof(addrs))
556 		return (EINVAL);
557 	if (req->oldlen != sizeof(struct xucred))
558 		return (EINVAL);
559 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
560 	if (error)
561 		return (error);
562 	if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
563 	    (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
564 		return (error);
565 	}
566 	inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
567 	    addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
568 	    INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
569 	if (inp != NULL) {
570 		INP_RLOCK_ASSERT(inp);
571 		if (inp->inp_socket == NULL)
572 			error = ENOENT;
573 		if (error == 0)
574 			error = cr_canseesocket(req->td->td_ucred,
575 			    inp->inp_socket);
576 		if (error == 0)
577 			cru2x(inp->inp_cred, &xuc);
578 		INP_RUNLOCK(inp);
579 	} else
580 		error = ENOENT;
581 	if (error == 0)
582 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
583 	return (error);
584 }
585 
586 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
587     0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
588 
589 static int
590 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
591     struct mbuf *control, struct thread *td)
592 {
593 	u_int32_t ulen = m->m_pkthdr.len;
594 	u_int32_t plen = sizeof(struct udphdr) + ulen;
595 	struct ip6_hdr *ip6;
596 	struct udphdr *udp6;
597 	struct in6_addr *laddr, *faddr, in6a;
598 	struct sockaddr_in6 *sin6 = NULL;
599 	struct ifnet *oifp = NULL;
600 	int scope_ambiguous = 0;
601 	u_short fport;
602 	int error = 0;
603 	struct ip6_pktopts *optp, opt;
604 	int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
605 	int flags;
606 	struct sockaddr_in6 tmp;
607 
608 	INP_WLOCK_ASSERT(inp);
609 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
610 
611 	if (addr6) {
612 		/* addr6 has been validated in udp6_send(). */
613 		sin6 = (struct sockaddr_in6 *)addr6;
614 
615 		/* protect *sin6 from overwrites */
616 		tmp = *sin6;
617 		sin6 = &tmp;
618 
619 		/*
620 		 * Application should provide a proper zone ID or the use of
621 		 * default zone IDs should be enabled.  Unfortunately, some
622 		 * applications do not behave as it should, so we need a
623 		 * workaround.  Even if an appropriate ID is not determined,
624 		 * we'll see if we can determine the outgoing interface.  If we
625 		 * can, determine the zone ID based on the interface below.
626 		 */
627 		if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
628 			scope_ambiguous = 1;
629 		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
630 			return (error);
631 	}
632 
633 	if (control) {
634 		if ((error = ip6_setpktopts(control, &opt,
635 		    inp->in6p_outputopts, td->td_ucred, IPPROTO_UDP)) != 0)
636 			goto release;
637 		optp = &opt;
638 	} else
639 		optp = inp->in6p_outputopts;
640 
641 	if (sin6) {
642 		faddr = &sin6->sin6_addr;
643 
644 		/*
645 		 * Since we saw no essential reason for calling in_pcbconnect,
646 		 * we get rid of such kind of logic, and call in6_selectsrc
647 		 * and in6_pcbsetport in order to fill in the local address
648 		 * and the local port.
649 		 */
650 		if (sin6->sin6_port == 0) {
651 			error = EADDRNOTAVAIL;
652 			goto release;
653 		}
654 
655 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
656 			/* how about ::ffff:0.0.0.0 case? */
657 			error = EISCONN;
658 			goto release;
659 		}
660 
661 		fport = sin6->sin6_port; /* allow 0 port */
662 
663 		if (IN6_IS_ADDR_V4MAPPED(faddr)) {
664 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
665 				/*
666 				 * I believe we should explicitly discard the
667 				 * packet when mapped addresses are disabled,
668 				 * rather than send the packet as an IPv6 one.
669 				 * If we chose the latter approach, the packet
670 				 * might be sent out on the wire based on the
671 				 * default route, the situation which we'd
672 				 * probably want to avoid.
673 				 * (20010421 jinmei@kame.net)
674 				 */
675 				error = EINVAL;
676 				goto release;
677 			}
678 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
679 			    !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
680 				/*
681 				 * when remote addr is an IPv4-mapped address,
682 				 * local addr should not be an IPv6 address,
683 				 * since you cannot determine how to map IPv6
684 				 * source address to IPv4.
685 				 */
686 				error = EINVAL;
687 				goto release;
688 			}
689 
690 			af = AF_INET;
691 		}
692 
693 		if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
694 			error = in6_selectsrc(sin6, optp, inp, NULL,
695 			    td->td_ucred, &oifp, &in6a);
696 			if (error)
697 				goto release;
698 			if (oifp && scope_ambiguous &&
699 			    (error = in6_setscope(&sin6->sin6_addr,
700 			    oifp, NULL))) {
701 				goto release;
702 			}
703 			laddr = &in6a;
704 		} else
705 			laddr = &inp->in6p_laddr;	/* XXX */
706 		if (laddr == NULL) {
707 			if (error == 0)
708 				error = EADDRNOTAVAIL;
709 			goto release;
710 		}
711 		if (inp->inp_lport == 0 &&
712 		    (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
713 			/* Undo an address bind that may have occurred. */
714 			inp->in6p_laddr = in6addr_any;
715 			goto release;
716 		}
717 	} else {
718 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
719 			error = ENOTCONN;
720 			goto release;
721 		}
722 		if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
723 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
724 				/*
725 				 * XXX: this case would happen when the
726 				 * application sets the V6ONLY flag after
727 				 * connecting the foreign address.
728 				 * Such applications should be fixed,
729 				 * so we bark here.
730 				 */
731 				log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
732 				    "option was set for a connected socket\n");
733 				error = EINVAL;
734 				goto release;
735 			} else
736 				af = AF_INET;
737 		}
738 		laddr = &inp->in6p_laddr;
739 		faddr = &inp->in6p_faddr;
740 		fport = inp->inp_fport;
741 	}
742 
743 	if (af == AF_INET)
744 		hlen = sizeof(struct ip);
745 
746 	/*
747 	 * Calculate data length and get a mbuf
748 	 * for UDP and IP6 headers.
749 	 */
750 	M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
751 	if (m == 0) {
752 		error = ENOBUFS;
753 		goto release;
754 	}
755 
756 	/*
757 	 * Stuff checksum and output datagram.
758 	 */
759 	udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
760 	udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
761 	udp6->uh_dport = fport;
762 	if (plen <= 0xffff)
763 		udp6->uh_ulen = htons((u_short)plen);
764 	else
765 		udp6->uh_ulen = 0;
766 	udp6->uh_sum = 0;
767 
768 	switch (af) {
769 	case AF_INET6:
770 		ip6 = mtod(m, struct ip6_hdr *);
771 		ip6->ip6_flow	= inp->inp_flow & IPV6_FLOWINFO_MASK;
772 		ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
773 		ip6->ip6_vfc	|= IPV6_VERSION;
774 #if 0				/* ip6_plen will be filled in ip6_output. */
775 		ip6->ip6_plen	= htons((u_short)plen);
776 #endif
777 		ip6->ip6_nxt	= IPPROTO_UDP;
778 		ip6->ip6_hlim	= in6_selecthlim(inp, NULL);
779 		ip6->ip6_src	= *laddr;
780 		ip6->ip6_dst	= *faddr;
781 
782 		udp6->uh_sum = in6_cksum_pseudo(ip6, plen, IPPROTO_UDP, 0);
783 		m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
784 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
785 
786 		flags = 0;
787 
788 		UDPSTAT_INC(udps_opackets);
789 		error = ip6_output(m, optp, NULL, flags, inp->in6p_moptions,
790 		    NULL, inp);
791 		break;
792 	case AF_INET:
793 		error = EAFNOSUPPORT;
794 		goto release;
795 	}
796 	goto releaseopt;
797 
798 release:
799 	m_freem(m);
800 
801 releaseopt:
802 	if (control) {
803 		ip6_clearpktopts(&opt, -1);
804 		m_freem(control);
805 	}
806 	return (error);
807 }
808 
809 static void
810 udp6_abort(struct socket *so)
811 {
812 	struct inpcb *inp;
813 
814 	inp = sotoinpcb(so);
815 	KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
816 
817 #ifdef INET
818 	if (inp->inp_vflag & INP_IPV4) {
819 		struct pr_usrreqs *pru;
820 
821 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
822 		(*pru->pru_abort)(so);
823 		return;
824 	}
825 #endif
826 
827 	INP_WLOCK(inp);
828 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
829 		INP_HASH_WLOCK(&V_udbinfo);
830 		in6_pcbdisconnect(inp);
831 		inp->in6p_laddr = in6addr_any;
832 		INP_HASH_WUNLOCK(&V_udbinfo);
833 		soisdisconnected(so);
834 	}
835 	INP_WUNLOCK(inp);
836 }
837 
838 static int
839 udp6_attach(struct socket *so, int proto, struct thread *td)
840 {
841 	struct inpcb *inp;
842 	int error;
843 
844 	inp = sotoinpcb(so);
845 	KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
846 
847 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
848 		error = soreserve(so, udp_sendspace, udp_recvspace);
849 		if (error)
850 			return (error);
851 	}
852 	INP_INFO_WLOCK(&V_udbinfo);
853 	error = in_pcballoc(so, &V_udbinfo);
854 	if (error) {
855 		INP_INFO_WUNLOCK(&V_udbinfo);
856 		return (error);
857 	}
858 	inp = (struct inpcb *)so->so_pcb;
859 	inp->inp_vflag |= INP_IPV6;
860 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
861 		inp->inp_vflag |= INP_IPV4;
862 	inp->in6p_hops = -1;	/* use kernel default */
863 	inp->in6p_cksum = -1;	/* just to be sure */
864 	/*
865 	 * XXX: ugly!!
866 	 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
867 	 * because the socket may be bound to an IPv6 wildcard address,
868 	 * which may match an IPv4-mapped IPv6 address.
869 	 */
870 	inp->inp_ip_ttl = V_ip_defttl;
871 
872 	error = udp_newudpcb(inp);
873 	if (error) {
874 		in_pcbdetach(inp);
875 		in_pcbfree(inp);
876 		INP_INFO_WUNLOCK(&V_udbinfo);
877 		return (error);
878 	}
879 	INP_WUNLOCK(inp);
880 	INP_INFO_WUNLOCK(&V_udbinfo);
881 	return (0);
882 }
883 
884 static int
885 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
886 {
887 	struct inpcb *inp;
888 	int error;
889 
890 	inp = sotoinpcb(so);
891 	KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
892 
893 	INP_WLOCK(inp);
894 	INP_HASH_WLOCK(&V_udbinfo);
895 	inp->inp_vflag &= ~INP_IPV4;
896 	inp->inp_vflag |= INP_IPV6;
897 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
898 		struct sockaddr_in6 *sin6_p;
899 
900 		sin6_p = (struct sockaddr_in6 *)nam;
901 
902 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
903 			inp->inp_vflag |= INP_IPV4;
904 #ifdef INET
905 		else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
906 			struct sockaddr_in sin;
907 
908 			in6_sin6_2_sin(&sin, sin6_p);
909 			inp->inp_vflag |= INP_IPV4;
910 			inp->inp_vflag &= ~INP_IPV6;
911 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
912 			    td->td_ucred);
913 			goto out;
914 		}
915 #endif
916 	}
917 
918 	error = in6_pcbbind(inp, nam, td->td_ucred);
919 #ifdef INET
920 out:
921 #endif
922 	INP_HASH_WUNLOCK(&V_udbinfo);
923 	INP_WUNLOCK(inp);
924 	return (error);
925 }
926 
927 static void
928 udp6_close(struct socket *so)
929 {
930 	struct inpcb *inp;
931 
932 	inp = sotoinpcb(so);
933 	KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
934 
935 #ifdef INET
936 	if (inp->inp_vflag & INP_IPV4) {
937 		struct pr_usrreqs *pru;
938 
939 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
940 		(*pru->pru_disconnect)(so);
941 		return;
942 	}
943 #endif
944 	INP_WLOCK(inp);
945 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
946 		INP_HASH_WLOCK(&V_udbinfo);
947 		in6_pcbdisconnect(inp);
948 		inp->in6p_laddr = in6addr_any;
949 		INP_HASH_WUNLOCK(&V_udbinfo);
950 		soisdisconnected(so);
951 	}
952 	INP_WUNLOCK(inp);
953 }
954 
955 static int
956 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
957 {
958 	struct inpcb *inp;
959 	struct sockaddr_in6 *sin6;
960 	int error;
961 
962 	inp = sotoinpcb(so);
963 	sin6 = (struct sockaddr_in6 *)nam;
964 	KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
965 
966 	/*
967 	 * XXXRW: Need to clarify locking of v4/v6 flags.
968 	 */
969 	INP_WLOCK(inp);
970 #ifdef INET
971 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
972 		struct sockaddr_in sin;
973 
974 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
975 			error = EINVAL;
976 			goto out;
977 		}
978 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
979 			error = EISCONN;
980 			goto out;
981 		}
982 		in6_sin6_2_sin(&sin, sin6);
983 		inp->inp_vflag |= INP_IPV4;
984 		inp->inp_vflag &= ~INP_IPV6;
985 		error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
986 		if (error != 0)
987 			goto out;
988 		INP_HASH_WLOCK(&V_udbinfo);
989 		error = in_pcbconnect(inp, (struct sockaddr *)&sin,
990 		    td->td_ucred);
991 		INP_HASH_WUNLOCK(&V_udbinfo);
992 		if (error == 0)
993 			soisconnected(so);
994 		goto out;
995 	}
996 #endif
997 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
998 		error = EISCONN;
999 		goto out;
1000 	}
1001 	inp->inp_vflag &= ~INP_IPV4;
1002 	inp->inp_vflag |= INP_IPV6;
1003 	error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1004 	if (error != 0)
1005 		goto out;
1006 	INP_HASH_WLOCK(&V_udbinfo);
1007 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1008 	INP_HASH_WUNLOCK(&V_udbinfo);
1009 	if (error == 0)
1010 		soisconnected(so);
1011 out:
1012 	INP_WUNLOCK(inp);
1013 	return (error);
1014 }
1015 
1016 static void
1017 udp6_detach(struct socket *so)
1018 {
1019 	struct inpcb *inp;
1020 	struct udpcb *up;
1021 
1022 	inp = sotoinpcb(so);
1023 	KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1024 
1025 	INP_INFO_WLOCK(&V_udbinfo);
1026 	INP_WLOCK(inp);
1027 	up = intoudpcb(inp);
1028 	KASSERT(up != NULL, ("%s: up == NULL", __func__));
1029 	in_pcbdetach(inp);
1030 	in_pcbfree(inp);
1031 	INP_INFO_WUNLOCK(&V_udbinfo);
1032 	udp_discardcb(up);
1033 }
1034 
1035 static int
1036 udp6_disconnect(struct socket *so)
1037 {
1038 	struct inpcb *inp;
1039 	int error;
1040 
1041 	inp = sotoinpcb(so);
1042 	KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1043 
1044 #ifdef INET
1045 	if (inp->inp_vflag & INP_IPV4) {
1046 		struct pr_usrreqs *pru;
1047 
1048 		pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1049 		(void)(*pru->pru_disconnect)(so);
1050 		return (0);
1051 	}
1052 #endif
1053 
1054 	INP_WLOCK(inp);
1055 
1056 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1057 		error = ENOTCONN;
1058 		goto out;
1059 	}
1060 
1061 	INP_HASH_WLOCK(&V_udbinfo);
1062 	in6_pcbdisconnect(inp);
1063 	inp->in6p_laddr = in6addr_any;
1064 	INP_HASH_WUNLOCK(&V_udbinfo);
1065 	SOCK_LOCK(so);
1066 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1067 	SOCK_UNLOCK(so);
1068 out:
1069 	INP_WUNLOCK(inp);
1070 	return (0);
1071 }
1072 
1073 static int
1074 udp6_send(struct socket *so, int flags, struct mbuf *m,
1075     struct sockaddr *addr, struct mbuf *control, struct thread *td)
1076 {
1077 	struct inpcb *inp;
1078 	int error = 0;
1079 
1080 	inp = sotoinpcb(so);
1081 	KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1082 
1083 	INP_WLOCK(inp);
1084 	if (addr) {
1085 		if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1086 			error = EINVAL;
1087 			goto bad;
1088 		}
1089 		if (addr->sa_family != AF_INET6) {
1090 			error = EAFNOSUPPORT;
1091 			goto bad;
1092 		}
1093 	}
1094 
1095 #ifdef INET
1096 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1097 		int hasv4addr;
1098 		struct sockaddr_in6 *sin6 = 0;
1099 
1100 		if (addr == 0)
1101 			hasv4addr = (inp->inp_vflag & INP_IPV4);
1102 		else {
1103 			sin6 = (struct sockaddr_in6 *)addr;
1104 			hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1105 			    ? 1 : 0;
1106 		}
1107 		if (hasv4addr) {
1108 			struct pr_usrreqs *pru;
1109 
1110 			/*
1111 			 * XXXRW: We release UDP-layer locks before calling
1112 			 * udp_send() in order to avoid recursion.  However,
1113 			 * this does mean there is a short window where inp's
1114 			 * fields are unstable.  Could this lead to a
1115 			 * potential race in which the factors causing us to
1116 			 * select the UDPv4 output routine are invalidated?
1117 			 */
1118 			INP_WUNLOCK(inp);
1119 			if (sin6)
1120 				in6_sin6_2_sin_in_sock(addr);
1121 			pru = inetsw[ip_protox[IPPROTO_UDP]].pr_usrreqs;
1122 			/* addr will just be freed in sendit(). */
1123 			return ((*pru->pru_send)(so, flags, m, addr, control,
1124 			    td));
1125 		}
1126 	}
1127 #endif
1128 #ifdef MAC
1129 	mac_inpcb_create_mbuf(inp, m);
1130 #endif
1131 	INP_HASH_WLOCK(&V_udbinfo);
1132 	error = udp6_output(inp, m, addr, control, td);
1133 	INP_HASH_WUNLOCK(&V_udbinfo);
1134 #ifdef INET
1135 #endif
1136 	INP_WUNLOCK(inp);
1137 	return (error);
1138 
1139 bad:
1140 	INP_WUNLOCK(inp);
1141 	m_freem(m);
1142 	return (error);
1143 }
1144 
1145 struct pr_usrreqs udp6_usrreqs = {
1146 	.pru_abort =		udp6_abort,
1147 	.pru_attach =		udp6_attach,
1148 	.pru_bind =		udp6_bind,
1149 	.pru_connect =		udp6_connect,
1150 	.pru_control =		in6_control,
1151 	.pru_detach =		udp6_detach,
1152 	.pru_disconnect =	udp6_disconnect,
1153 	.pru_peeraddr =		in6_mapped_peeraddr,
1154 	.pru_send =		udp6_send,
1155 	.pru_shutdown =		udp_shutdown,
1156 	.pru_sockaddr =		in6_mapped_sockaddr,
1157 	.pru_soreceive =	soreceive_dgram,
1158 	.pru_sosend =		sosend_dgram,
1159 	.pru_sosetlabel =	in_pcbsosetlabel,
1160 	.pru_close =		udp6_close
1161 };
1162