xref: /openbsd-src/sys/netinet/udp_usrreq.c (revision 47911bd667ac77dc523b8a13ef40b012dbffa741)
1 /*	$OpenBSD: udp_usrreq.c,v 1.86 2002/08/28 15:43:03 pefo Exp $	*/
2 /*	$NetBSD: udp_usrreq.c,v 1.28 1996/03/16 23:54:03 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University 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 REGENTS 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 REGENTS 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  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
37  *
38  * NRL grants permission for redistribution and use in source and binary
39  * forms, with or without modification, of the software and documentation
40  * created at NRL provided that the following conditions are met:
41  *
42  * 1. Redistributions of source code must retain the above copyright
43  *    notice, this list of conditions and the following disclaimer.
44  * 2. Redistributions in binary form must reproduce the above copyright
45  *    notice, this list of conditions and the following disclaimer in the
46  *    documentation and/or other materials provided with the distribution.
47  * 3. All advertising materials mentioning features or use of this software
48  *    must display the following acknowledgements:
49  * 	This product includes software developed by the University of
50  * 	California, Berkeley and its contributors.
51  * 	This product includes software developed at the Information
52  * 	Technology Division, US Naval Research Laboratory.
53  * 4. Neither the name of the NRL 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  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
58  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
59  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
60  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
61  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
62  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
63  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
64  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
65  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
66  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
67  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
68  *
69  * The views and conclusions contained in the software and documentation
70  * are those of the authors and should not be interpreted as representing
71  * official policies, either expressed or implied, of the US Naval
72  * Research Laboratory (NRL).
73  */
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sysctl.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/ip_var.h>
92 #include <netinet/ip_icmp.h>
93 #include <netinet/udp.h>
94 #include <netinet/udp_var.h>
95 
96 #ifdef INET6
97 #ifndef INET
98 #include <netinet/in.h>
99 #endif
100 #include <netinet6/ip6protosw.h>
101 
102 extern int ip6_defhlim;
103 #endif /* INET6 */
104 
105 /*
106  * UDP protocol implementation.
107  * Per RFC 768, August, 1980.
108  */
109 int	udpcksum = 1;
110 
111 
112 static	void udp_detach(struct inpcb *);
113 static	void udp_notify(struct inpcb *, int);
114 static	struct mbuf *udp_saveopt(caddr_t, int, int);
115 
116 #ifndef UDBHASHSIZE
117 #define	UDBHASHSIZE	128
118 #endif
119 int	udbhashsize = UDBHASHSIZE;
120 
121 /* from in_pcb.c */
122 extern	struct baddynamicports baddynamicports;
123 
124 void
125 udp_init()
126 {
127 	in_pcbinit(&udbtable, udbhashsize);
128 }
129 
130 #ifdef INET6
131 int
132 udp6_input(mp, offp, proto)
133 	struct mbuf **mp;
134 	int *offp, proto;
135 {
136 	struct mbuf *m = *mp;
137 
138 #if defined(NFAITH) && 0 < NFAITH
139 	if (m->m_pkthdr.rcvif) {
140 		if (m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
141 			/* XXX send icmp6 host/port unreach? */
142 			m_freem(m);
143 			return IPPROTO_DONE;
144 		}
145 	}
146 #endif
147 
148 	udp_input(m, *offp, proto);
149 	return IPPROTO_DONE;
150 }
151 #endif
152 
153 void
154 udp_input(struct mbuf *m, ...)
155 {
156 	register struct ip *ip;
157 	register struct udphdr *uh;
158 	register struct inpcb *inp;
159 	struct mbuf *opts = 0;
160 	struct ip save_ip;
161 	int iphlen, len;
162 	va_list ap;
163 	u_int16_t savesum;
164 	union {
165 		struct sockaddr sa;
166 		struct sockaddr_in sin;
167 #ifdef INET6
168 		struct sockaddr_in6 sin6;
169 #endif /* INET6 */
170 	} srcsa, dstsa;
171 #ifdef INET6
172 	struct ip6_hdr *ip6;
173 #endif /* INET6 */
174 #ifdef IPSEC
175 	struct m_tag *mtag;
176 	struct tdb_ident *tdbi;
177 	struct tdb *tdb;
178 	int error, s;
179 #endif /* IPSEC */
180 
181 	va_start(ap, m);
182 	iphlen = va_arg(ap, int);
183 	va_end(ap);
184 
185 	udpstat.udps_ipackets++;
186 
187 	switch (mtod(m, struct ip *)->ip_v) {
188 	case 4:
189 		ip = mtod(m, struct ip *);
190 #ifdef INET6
191 		ip6 = NULL;
192 #endif /* INET6 */
193 		srcsa.sa.sa_family = AF_INET;
194 		break;
195 #ifdef INET6
196 	case 6:
197 		ip = NULL;
198 		ip6 = mtod(m, struct ip6_hdr *);
199 		srcsa.sa.sa_family = AF_INET6;
200 		break;
201 #endif /* INET6 */
202 	default:
203 		goto bad;
204 	}
205 
206 	/*
207 	 * Strip IP options, if any; should skip this,
208 	 * make available to user, and use on returned packets,
209 	 * but we don't yet have a way to check the checksum
210 	 * with options still present.
211 	 */
212 	/*
213 	 * (contd. from above...)  Furthermore, we may want to strip options
214 	 * for such things as ICMP errors, where options just get in the way.
215 	 */
216 	if (ip && iphlen > sizeof (struct ip)) {
217 		ip_stripoptions(m, (struct mbuf *)0);
218 		iphlen = sizeof(struct ip);
219 	}
220 
221 	/*
222 	 * Get IP and UDP header together in first mbuf.
223 	 */
224 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
225 		if ((m = m_pullup2(m, iphlen + sizeof(struct udphdr))) ==
226 		    NULL) {
227 			udpstat.udps_hdrops++;
228 			return;
229 		}
230 #ifdef INET6
231 		if (ip6)
232 			ip6 = mtod(m, struct ip6_hdr *);
233 		else
234 #endif /* INET6 */
235 			ip = mtod(m, struct ip *);
236 	}
237 	uh = (struct udphdr *)(mtod(m, caddr_t) + iphlen);
238 
239 	/* Check for illegal destination port 0 */
240 	if (uh->uh_dport == 0) {
241 		udpstat.udps_noport++;
242 		goto bad;
243 	}
244 
245 	/*
246 	 * Make mbuf data length reflect UDP length.
247 	 * If not enough data to reflect UDP length, drop.
248 	 */
249 	len = ntohs((u_int16_t)uh->uh_ulen);
250 	if (m->m_pkthdr.len - iphlen != len) {
251 		if (len > (m->m_pkthdr.len - iphlen) ||
252 		    len < sizeof(struct udphdr)) {
253 			udpstat.udps_badlen++;
254 			goto bad;
255 		}
256 		m_adj(m, len - (m->m_pkthdr.len - iphlen));
257 	}
258 	/*
259 	 * Save a copy of the IP header in case we want restore it
260 	 * for sending an ICMP error message in response.
261 	 */
262 	if (ip)
263 		save_ip = *ip;
264 
265 	/*
266 	 * Checksum extended UDP header and data.
267 	 * from W.R.Stevens: check incoming udp cksums even if
268 	 *	udpcksum is not set.
269 	 */
270 	savesum = uh->uh_sum;
271 #ifdef INET6
272 	if (ip6) {
273 		/* Be proactive about malicious use of IPv4 mapped address */
274 		if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
275 		    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
276 			/* XXX stat */
277 			goto bad;
278 		}
279 
280 		/*
281 		 * In IPv6, the UDP checksum is ALWAYS used.
282 		 */
283 		if ((uh->uh_sum = in6_cksum(m, IPPROTO_UDP, iphlen, len))) {
284 			udpstat.udps_badsum++;
285 			goto bad;
286 		}
287 	} else
288 #endif /* INET6 */
289 	if (uh->uh_sum) {
290 		if ((m->m_pkthdr.csum & M_UDP_CSUM_IN_OK) == 0) {
291 			if (m->m_pkthdr.csum & M_UDP_CSUM_IN_BAD) {
292 				udpstat.udps_badsum++;
293 				udpstat.udps_inhwcsum++;
294 				m_freem(m);
295 				return;
296 			}
297 
298 			bzero(((struct ipovly *)ip)->ih_x1,
299 			    sizeof ((struct ipovly *)ip)->ih_x1);
300 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
301 
302 			if ((uh->uh_sum = in_cksum(m, len +
303 			    sizeof (struct ip))) != 0) {
304 				udpstat.udps_badsum++;
305 				m_freem(m);
306 				return;
307 			}
308 		} else {
309 			m->m_pkthdr.csum &= ~M_UDP_CSUM_IN_OK;
310 			udpstat.udps_inhwcsum++;
311 		}
312 	} else
313 		udpstat.udps_nosum++;
314 
315 	switch (srcsa.sa.sa_family) {
316 	case AF_INET:
317 		bzero(&srcsa, sizeof(struct sockaddr_in));
318 		srcsa.sin.sin_len = sizeof(struct sockaddr_in);
319 		srcsa.sin.sin_family = AF_INET;
320 		srcsa.sin.sin_port = uh->uh_sport;
321 		srcsa.sin.sin_addr = ip->ip_src;
322 
323 		bzero(&dstsa, sizeof(struct sockaddr_in));
324 		dstsa.sin.sin_len = sizeof(struct sockaddr_in);
325 		dstsa.sin.sin_family = AF_INET;
326 		dstsa.sin.sin_port = uh->uh_dport;
327 		dstsa.sin.sin_addr = ip->ip_dst;
328 		break;
329 #ifdef INET6
330 	case AF_INET6:
331 		bzero(&srcsa, sizeof(struct sockaddr_in6));
332 		srcsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
333 		srcsa.sin6.sin6_family = AF_INET6;
334 		srcsa.sin6.sin6_port = uh->uh_sport;
335 #if 0 /*XXX inbound flowinfo */
336 		srcsa.sin6.sin6_flowinfo = htonl(0x0fffffff) & ip6->ip6_flow;
337 #endif
338 		/* KAME hack: recover scopeid */
339 		(void)in6_recoverscope(&srcsa.sin6, &ip6->ip6_src,
340 		    m->m_pkthdr.rcvif);
341 
342 		bzero(&dstsa, sizeof(struct sockaddr_in6));
343 		dstsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
344 		dstsa.sin6.sin6_family = AF_INET6;
345 		dstsa.sin6.sin6_port = uh->uh_dport;
346 		/* KAME hack: recover scopeid */
347 		(void)in6_recoverscope(&dstsa.sin6, &ip6->ip6_dst,
348 		    m->m_pkthdr.rcvif);
349 		break;
350 #endif /* INET6 */
351 	}
352 
353 #ifdef INET6
354 	if ((ip6 && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) ||
355 	    (ip && IN_MULTICAST(ip->ip_dst.s_addr)) ||
356 	    (ip && in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))) {
357 #else /* INET6 */
358 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
359 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
360 #endif /* INET6 */
361 		struct socket *last;
362 		/*
363 		 * Deliver a multicast or broadcast datagram to *all* sockets
364 		 * for which the local and remote addresses and ports match
365 		 * those of the incoming datagram.  This allows more than
366 		 * one process to receive multi/broadcasts on the same port.
367 		 * (This really ought to be done for unicast datagrams as
368 		 * well, but that would cause problems with existing
369 		 * applications that open both address-specific sockets and
370 		 * a wildcard socket listening to the same port -- they would
371 		 * end up receiving duplicates of every unicast datagram.
372 		 * Those applications open the multiple sockets to overcome an
373 		 * inadequacy of the UDP socket interface, but for backwards
374 		 * compatibility we avoid the problem here rather than
375 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
376 		 */
377 
378 		iphlen += sizeof(struct udphdr);
379 
380 		/*
381 		 * Locate pcb(s) for datagram.
382 		 * (Algorithm copied from raw_intr().)
383 		 */
384 		last = NULL;
385 		for (inp = udbtable.inpt_queue.cqh_first;
386 		    inp != (struct inpcb *)&udbtable.inpt_queue;
387 		    inp = inp->inp_queue.cqe_next) {
388 #ifdef INET6
389 			/* don't accept it if AF does not match */
390 			if (ip6 && !(inp->inp_flags & INP_IPV6))
391 				continue;
392 			if (!ip6 && (inp->inp_flags & INP_IPV6))
393 				continue;
394 #endif
395 			if (inp->inp_lport != uh->uh_dport)
396 				continue;
397 #ifdef INET6
398 			if (ip6) {
399 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
400 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
401 					    &ip6->ip6_dst))
402 						continue;
403 			} else
404 #endif /* INET6 */
405 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
406 				if (inp->inp_laddr.s_addr !=
407 				    ip->ip_dst.s_addr)
408 					continue;
409 			}
410 #ifdef INET6
411 			if (ip6) {
412 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
413 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
414 					    &ip6->ip6_src) ||
415 					    inp->inp_fport != uh->uh_sport)
416 						continue;
417 			} else
418 #endif /* INET6 */
419 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
420 				if (inp->inp_faddr.s_addr !=
421 				    ip->ip_src.s_addr ||
422 				    inp->inp_fport != uh->uh_sport)
423 					continue;
424 			}
425 
426 			if (last != NULL) {
427 				struct mbuf *n;
428 
429 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
430 					opts = NULL;
431 #ifdef INET6
432 					if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
433 						ip6_savecontrol(inp, &opts, ip6, n);
434 #endif /* INET6 */
435 					m_adj(n, iphlen);
436 					if (sbappendaddr(&last->so_rcv,
437 					    &srcsa.sa, n, opts) == 0) {
438 						m_freem(n);
439 						if (opts)
440 							m_freem(opts);
441 						udpstat.udps_fullsock++;
442 					} else
443 						sorwakeup(last);
444 					opts = NULL;
445 				}
446 			}
447 			last = inp->inp_socket;
448 			/*
449 			 * Don't look for additional matches if this one does
450 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
451 			 * socket options set.  This heuristic avoids searching
452 			 * through all pcbs in the common case of a non-shared
453 			 * port.  It * assumes that an application will never
454 			 * clear these options after setting them.
455 			 */
456 			if ((last->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
457 				break;
458 		}
459 
460 		if (last == NULL) {
461 			/*
462 			 * No matching pcb found; discard datagram.
463 			 * (No need to send an ICMP Port Unreachable
464 			 * for a broadcast or multicast datgram.)
465 			 */
466 			udpstat.udps_noportbcast++;
467 			goto bad;
468 		}
469 
470 		opts = NULL;
471 #ifdef INET6
472 		if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
473 			ip6_savecontrol(inp, &opts, ip6, m);
474 #endif /* INET6 */
475 		m_adj(m, iphlen);
476 		if (sbappendaddr(&last->so_rcv,
477 		    &srcsa.sa, m, opts) == 0) {
478 			udpstat.udps_fullsock++;
479 			goto bad;
480 		}
481 		sorwakeup(last);
482 		return;
483 	}
484 	/*
485 	 * Locate pcb for datagram.
486 	 */
487 #ifdef INET6
488 	if (ip6)
489 		inp = in6_pcbhashlookup(&udbtable, &ip6->ip6_src, uh->uh_sport,
490 		    &ip6->ip6_dst, uh->uh_dport);
491 	else
492 #endif /* INET6 */
493 	inp = in_pcbhashlookup(&udbtable, ip->ip_src, uh->uh_sport,
494 	    ip->ip_dst, uh->uh_dport);
495 	if (inp == 0) {
496 		++udpstat.udps_pcbhashmiss;
497 #ifdef INET6
498 		if (ip6) {
499 			inp = in_pcblookup(&udbtable,
500 			    (struct in_addr *)&(ip6->ip6_src),
501 			    uh->uh_sport, (struct in_addr *)&(ip6->ip6_dst),
502 			    uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_IPV6);
503 		} else
504 #endif /* INET6 */
505 		inp = in_pcblookup(&udbtable, &ip->ip_src, uh->uh_sport,
506 		    &ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD);
507 		if (inp == 0) {
508 			udpstat.udps_noport++;
509 			if (m->m_flags & (M_BCAST | M_MCAST)) {
510 				udpstat.udps_noportbcast++;
511 				goto bad;
512 			}
513 #ifdef INET6
514 			if (ip6) {
515 				icmp6_error(m, ICMP6_DST_UNREACH,
516 				    ICMP6_DST_UNREACH_NOPORT,0);
517 			} else
518 #endif /* INET6 */
519 			{
520 				*ip = save_ip;
521 				uh->uh_sum = savesum;
522 				icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT,
523 				    0, 0);
524 			}
525 			return;
526 		}
527 	}
528 
529 #ifdef IPSEC
530 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
531 	s = splnet();
532 	if (mtag != NULL) {
533 		tdbi = (struct tdb_ident *)(mtag + 1);
534 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
535 	} else
536 		tdb = NULL;
537 	ipsp_spd_lookup(m, srcsa.sa.sa_family, iphlen, &error,
538 	    IPSP_DIRECTION_IN, tdb, inp);
539 	if (error) {
540 		splx(s);
541 		goto bad;
542 	}
543 
544 	/* Latch SA only if the socket is connected */
545 	if (inp->inp_tdb_in != tdb &&
546 	    (inp->inp_socket->so_state & SS_ISCONNECTED)) {
547 		if (tdb) {
548 			tdb_add_inp(tdb, inp, 1);
549 			if (inp->inp_ipo == NULL) {
550 				inp->inp_ipo = ipsec_add_policy(inp,
551 				    srcsa.sa.sa_family, IPSP_DIRECTION_OUT);
552 				if (inp->inp_ipo == NULL) {
553 					splx(s);
554 					goto bad;
555 				}
556 			}
557 			if (inp->inp_ipo->ipo_dstid == NULL &&
558 			    tdb->tdb_srcid != NULL) {
559 				inp->inp_ipo->ipo_dstid = tdb->tdb_srcid;
560 				tdb->tdb_srcid->ref_count++;
561 			}
562 			if (inp->inp_ipsec_remotecred == NULL &&
563 			    tdb->tdb_remote_cred != NULL) {
564 				inp->inp_ipsec_remotecred =
565 				    tdb->tdb_remote_cred;
566 				tdb->tdb_remote_cred->ref_count++;
567 			}
568 			if (inp->inp_ipsec_remoteauth == NULL &&
569 			    tdb->tdb_remote_auth != NULL) {
570 				inp->inp_ipsec_remoteauth =
571 				    tdb->tdb_remote_auth;
572 				tdb->tdb_remote_auth->ref_count++;
573 			}
574 		} else { /* Just reset */
575 			TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in, inp,
576 			    inp_tdb_in_next);
577 			inp->inp_tdb_in = NULL;
578 		}
579 	}
580 	splx(s);
581 #endif /*IPSEC */
582 
583 	opts = NULL;
584 #ifdef INET6
585 	if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
586 		ip6_savecontrol(inp, &opts, ip6, m);
587 #endif /* INET6 */
588 	if (ip && (inp->inp_flags & INP_CONTROLOPTS)) {
589 		struct mbuf **mp = &opts;
590 
591 		if (inp->inp_flags & INP_RECVDSTADDR) {
592 			*mp = udp_saveopt((caddr_t) &ip->ip_dst,
593 			    sizeof(struct in_addr), IP_RECVDSTADDR);
594 			if (*mp)
595 				mp = &(*mp)->m_next;
596 		}
597 #ifdef notyet
598 		/* options were tossed above */
599 		if (inp->inp_flags & INP_RECVOPTS) {
600 			*mp = udp_saveopt((caddr_t) opts_deleted_above,
601 			    sizeof(struct in_addr), IP_RECVOPTS);
602 			if (*mp)
603 				mp = &(*mp)->m_next;
604 		}
605 		/* ip_srcroute doesn't do what we want here, need to fix */
606 		if (inp->inp_flags & INP_RECVRETOPTS) {
607 			*mp = udp_saveopt((caddr_t) ip_srcroute(),
608 			    sizeof(struct in_addr), IP_RECVRETOPTS);
609 			if (*mp)
610 				mp = &(*mp)->m_next;
611 		}
612 #endif
613 	}
614 	iphlen += sizeof(struct udphdr);
615 	m_adj(m, iphlen);
616 	if (sbappendaddr(&inp->inp_socket->so_rcv,
617 		&srcsa.sa, m, opts) == 0) {
618 		udpstat.udps_fullsock++;
619 		goto bad;
620 	}
621 	sorwakeup(inp->inp_socket);
622 	return;
623 bad:
624 	m_freem(m);
625 	if (opts)
626 		m_freem(opts);
627 }
628 
629 /*
630  * Create a "control" mbuf containing the specified data
631  * with the specified type for presentation with a datagram.
632  */
633 struct mbuf *
634 udp_saveopt(p, size, type)
635 	caddr_t p;
636 	register int size;
637 	int type;
638 {
639 	register struct cmsghdr *cp;
640 	struct mbuf *m;
641 
642 	if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
643 		return ((struct mbuf *) NULL);
644 	cp = (struct cmsghdr *) mtod(m, struct cmsghdr *);
645 	bcopy(p, CMSG_DATA(cp), size);
646 	size = CMSG_LEN(size);
647 	m->m_len = size;
648 	cp->cmsg_len = size;
649 	cp->cmsg_level = IPPROTO_IP;
650 	cp->cmsg_type = type;
651 	return (m);
652 }
653 
654 /*
655  * Notify a udp user of an asynchronous error;
656  * just wake up so that he can collect error status.
657  */
658 static void
659 udp_notify(inp, errno)
660 	register struct inpcb *inp;
661 	int errno;
662 {
663 	inp->inp_socket->so_error = errno;
664 	sorwakeup(inp->inp_socket);
665 	sowwakeup(inp->inp_socket);
666 }
667 
668 #ifdef INET6
669 void
670 udp6_ctlinput(cmd, sa, d)
671 	int cmd;
672 	struct sockaddr *sa;
673 	void *d;
674 {
675 	struct udphdr uh;
676 	struct sockaddr_in6 sa6;
677 	register struct ip6_hdr *ip6;
678 	struct mbuf *m;
679 	int off;
680 	void *cmdarg;
681 	struct ip6ctlparam *ip6cp = NULL;
682 	struct in6_addr finaldst;
683 	struct udp_portonly {
684 		u_int16_t uh_sport;
685 		u_int16_t uh_dport;
686 	} *uhp;
687 	void (*notify)(struct inpcb *, int) = udp_notify;
688 
689 	if (sa == NULL)
690 		return;
691 	if (sa->sa_family != AF_INET6 ||
692 	    sa->sa_len != sizeof(struct sockaddr_in6))
693 		return;
694 
695 	if ((unsigned)cmd >= PRC_NCMDS)
696 		return;
697 	if (PRC_IS_REDIRECT(cmd))
698 		notify = in_rtchange, d = NULL;
699 	else if (cmd == PRC_HOSTDEAD)
700 		d = NULL;
701 	else if (cmd == PRC_MSGSIZE)
702 		; /* special code is present, see below */
703 	else if (inet6ctlerrmap[cmd] == 0)
704 		return;
705 
706 	/* if the parameter is from icmp6, decode it. */
707 	if (d != NULL) {
708 		ip6cp = (struct ip6ctlparam *)d;
709 		m = ip6cp->ip6c_m;
710 		ip6 = ip6cp->ip6c_ip6;
711 		off = ip6cp->ip6c_off;
712 		cmdarg = ip6cp->ip6c_cmdarg;
713 	} else {
714 		m = NULL;
715 		ip6 = NULL;
716 		cmdarg = NULL;
717 		/* XXX: translate addresses into internal form */
718 		sa6 = *(struct sockaddr_in6 *)sa;
719 #ifndef SCOPEDROUTING
720 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
721 			/* should be impossible */
722 			return;
723 		}
724 #endif
725 	}
726 
727 	if (ip6cp && ip6cp->ip6c_finaldst) {
728 		bzero(&sa6, sizeof(sa6));
729 		sa6.sin6_family = AF_INET6;
730 		sa6.sin6_len = sizeof(sa6);
731 		sa6.sin6_addr = *ip6cp->ip6c_finaldst;
732 		/* XXX: assuming M is valid in this case */
733 		sa6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
734 		    ip6cp->ip6c_finaldst);
735 #ifndef SCOPEDROUTING
736 		if (in6_embedscope(ip6cp->ip6c_finaldst, &sa6, NULL, NULL)) {
737 			/* should be impossible */
738 			return;
739 		}
740 #endif
741 	} else {
742 		/* XXX: translate addresses into internal form */
743 		sa6 = *(struct sockaddr_in6 *)sa;
744 #ifndef SCOPEDROUTING
745 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
746 			/* should be impossible */
747 			return;
748 		}
749 #endif
750 	}
751 
752 	if (ip6) {
753 		/*
754 		 * XXX: We assume that when IPV6 is non NULL,
755 		 * M and OFF are valid.
756 		 */
757 		struct sockaddr_in6 sa6_src;
758 
759 		/* check if we can safely examine src and dst ports */
760 		if (m->m_pkthdr.len < off + sizeof(*uhp))
761 			return;
762 
763 		bzero(&uh, sizeof(uh));
764 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
765 
766 		bzero(&sa6_src, sizeof(sa6_src));
767 		sa6_src.sin6_family = AF_INET6;
768 		sa6_src.sin6_len = sizeof(sa6_src);
769 		sa6_src.sin6_addr = ip6->ip6_src;
770 		sa6_src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
771 		    &ip6->ip6_src);
772 #ifndef SCOPEDROUTING
773 		if (in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL)) {
774 			/* should be impossible */
775 			return;
776 		}
777 #endif
778 
779 		if (cmd == PRC_MSGSIZE) {
780 			int valid = 0;
781 
782 			/*
783 			 * Check to see if we have a valid UDP socket
784 			 * corresponding to the address in the ICMPv6 message
785 			 * payload.
786 			 */
787 			if (in6_pcbhashlookup(&udbtable, &finaldst,
788 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport))
789 				valid = 1;
790 			else if (in_pcblookup(&udbtable, &sa6.sin6_addr,
791 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport,
792 			    INPLOOKUP_IPV6))
793 				valid = 1;
794 #if 0
795 			/*
796 			 * As the use of sendto(2) is fairly popular,
797 			 * we may want to allow non-connected pcb too.
798 			 * But it could be too weak against attacks...
799 			 * We should at least check if the local address (= s)
800 			 * is really ours.
801 			 */
802 			else if (in_pcblookup(&udbtable, &sa6.sin6_addr,
803 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport,
804 			    INPLOOKUP_WILDCARD | INPLOOKUP_IPV6))
805 				valid = 1;
806 #endif
807 
808 			/*
809 			 * Depending on the value of "valid" and routing table
810 			 * size (mtudisc_{hi,lo}wat), we will:
811 			 * - recalcurate the new MTU and create the
812 			 *   corresponding routing entry, or
813 			 * - ignore the MTU change notification.
814 			 */
815 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
816 
817 			/*
818 			 * regardless of if we called icmp6_mtudisc_update(),
819 			 * we need to call in6_pcbnotify(), to notify path
820 			 * MTU change to the userland (2292bis-02), because
821 			 * some unconnected sockets may share the same
822 			 * destination and want to know the path MTU.
823 			 */
824 		}
825 
826 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6,
827 		    uh.uh_dport, (struct sockaddr *)&sa6_src,
828 		    uh.uh_sport, cmd, cmdarg, notify);
829 	} else {
830 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6, 0,
831 		    (struct sockaddr *)&sa6_any, 0, cmd, cmdarg, notify);
832 	}
833 }
834 #endif
835 
836 void *
837 udp_ctlinput(cmd, sa, v)
838 	int cmd;
839 	struct sockaddr *sa;
840 	void *v;
841 {
842 	register struct ip *ip = v;
843 	register struct udphdr *uhp;
844 	extern int inetctlerrmap[];
845 	void (*notify)(struct inpcb *, int) = udp_notify;
846 	int errno;
847 
848 	if (sa == NULL)
849 		return NULL;
850 	if (sa->sa_family != AF_INET ||
851 	    sa->sa_len != sizeof(struct sockaddr_in))
852 		return NULL;
853 
854 	if ((unsigned)cmd >= PRC_NCMDS)
855 		return NULL;
856 	errno = inetctlerrmap[cmd];
857 	if (PRC_IS_REDIRECT(cmd))
858 		notify = in_rtchange, ip = 0;
859 	else if (cmd == PRC_HOSTDEAD)
860 		ip = 0;
861 	else if (errno == 0)
862 		return NULL;
863 	if (ip) {
864 		uhp = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
865 		in_pcbnotify(&udbtable, sa, uhp->uh_dport, ip->ip_src,
866 		    uhp->uh_sport, errno, notify);
867 	} else
868 		in_pcbnotifyall(&udbtable, sa, errno, notify);
869 	return NULL;
870 }
871 
872 int
873 udp_output(struct mbuf *m, ...)
874 {
875 	register struct inpcb *inp;
876 	struct mbuf *addr, *control;
877 	register struct udpiphdr *ui;
878 	register int len = m->m_pkthdr.len;
879 	struct in_addr laddr;
880 	int s = 0, error = 0;
881 	va_list ap;
882 	int pcbflags = 0;
883 
884 	va_start(ap, m);
885 	inp = va_arg(ap, struct inpcb *);
886 	addr = va_arg(ap, struct mbuf *);
887 	control = va_arg(ap, struct mbuf *);
888 	va_end(ap);
889 
890 #ifdef DIAGNOSTIC
891 	if ((inp->inp_flags & INP_IPV6) != 0)
892 		panic("IPv6 inpcb to udp_output");
893 #endif
894 
895 	/*
896 	 * Compute the packet length of the IP header, and
897 	 * punt if the length looks bogus.
898 	 */
899 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
900 		error = EMSGSIZE;
901 		goto release;
902 	}
903 
904 	if (addr) {
905 		/*
906 		 * Save current PCB flags because they may change during
907 		 * temporary connection.
908 		 */
909 		pcbflags = inp->inp_flags;
910 
911 		laddr = inp->inp_laddr;
912 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
913 			error = EISCONN;
914 			goto release;
915 		}
916 		/*
917 		 * Must block input while temporarily connected.
918 		 */
919 		s = splsoftnet();
920 		error = in_pcbconnect(inp, addr);
921 		if (error) {
922 			splx(s);
923 			goto release;
924 		}
925 	} else {
926 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
927 			error = ENOTCONN;
928 			goto release;
929 		}
930 	}
931 	/*
932 	 * Calculate data length and get a mbuf
933 	 * for UDP and IP headers.
934 	 */
935 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
936 	if (m == 0) {
937 		error = ENOBUFS;
938 		goto bail;
939 	}
940 
941 	/*
942 	 * Fill in mbuf with extended UDP header
943 	 * and addresses and length put into network format.
944 	 */
945 	ui = mtod(m, struct udpiphdr *);
946 	bzero(ui->ui_x1, sizeof ui->ui_x1);
947 	ui->ui_pr = IPPROTO_UDP;
948 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
949 	ui->ui_src = inp->inp_laddr;
950 	ui->ui_dst = inp->inp_faddr;
951 	ui->ui_sport = inp->inp_lport;
952 	ui->ui_dport = inp->inp_fport;
953 	ui->ui_ulen = ui->ui_len;
954 
955 	/*
956 	 * Compute the pseudo-header checksum; defer further checksumming
957 	 * until ip_output() or hardware (if it exists).
958 	 */
959 	if (udpcksum) {
960 		m->m_pkthdr.csum |= M_UDPV4_CSUM_OUT;
961 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
962 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
963 		    sizeof (struct udphdr) + IPPROTO_UDP));
964 	} else
965 		ui->ui_sum = 0;
966 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
967 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;
968 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;
969 
970 	udpstat.udps_opackets++;
971 	error = ip_output(m, inp->inp_options, &inp->inp_route,
972 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
973 	    inp->inp_moptions, inp, (void *)NULL);
974 
975 bail:
976 	if (addr) {
977 		in_pcbdisconnect(inp);
978 		inp->inp_flags = pcbflags;
979 		inp->inp_laddr = laddr;
980 		splx(s);
981 	}
982 	if (control)
983 		m_freem(control);
984 	return (error);
985 
986 release:
987 	m_freem(m);
988 	if (control)
989 		m_freem(control);
990 	return (error);
991 }
992 
993 u_int	udp_sendspace = 9216;		/* really max datagram size */
994 u_int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
995 					/* 40 1K datagrams */
996 
997 #ifdef INET6
998 /*ARGSUSED*/
999 int
1000 udp6_usrreq(so, req, m, addr, control, p)
1001 	struct socket *so;
1002 	int req;
1003 	struct mbuf *m, *addr, *control;
1004 	struct proc *p;
1005 {
1006 
1007 	return udp_usrreq(so, req, m, addr, control);
1008 }
1009 #endif
1010 
1011 /*ARGSUSED*/
1012 int
1013 udp_usrreq(so, req, m, addr, control)
1014 	struct socket *so;
1015 	int req;
1016 	struct mbuf *m, *addr, *control;
1017 {
1018 	struct inpcb *inp = sotoinpcb(so);
1019 	int error = 0;
1020 	int s;
1021 
1022 	if (req == PRU_CONTROL) {
1023 #ifdef INET6
1024 		if (inp->inp_flags & INP_IPV6)
1025 			return (in6_control(so, (u_long)m, (caddr_t)addr,
1026 			    (struct ifnet *)control, 0));
1027 		else
1028 #endif /* INET6 */
1029 			return (in_control(so, (u_long)m, (caddr_t)addr,
1030 			    (struct ifnet *)control));
1031 	}
1032 	if (inp == NULL && req != PRU_ATTACH) {
1033 		error = EINVAL;
1034 		goto release;
1035 	}
1036 	/*
1037 	 * Note: need to block udp_input while changing
1038 	 * the udp pcb queue and/or pcb addresses.
1039 	 */
1040 	switch (req) {
1041 
1042 	case PRU_ATTACH:
1043 		if (inp != NULL) {
1044 			error = EINVAL;
1045 			break;
1046 		}
1047 		s = splsoftnet();
1048 		error = in_pcballoc(so, &udbtable);
1049 		splx(s);
1050 		if (error)
1051 			break;
1052 		error = soreserve(so, udp_sendspace, udp_recvspace);
1053 		if (error)
1054 			break;
1055 #ifdef INET6
1056 		if (((struct inpcb *)so->so_pcb)->inp_flags & INP_IPV6)
1057 			((struct inpcb *) so->so_pcb)->inp_ipv6.ip6_hlim =
1058 			    ip6_defhlim;
1059 		else
1060 #endif /* INET6 */
1061 			((struct inpcb *) so->so_pcb)->inp_ip.ip_ttl = ip_defttl;
1062 		break;
1063 
1064 	case PRU_DETACH:
1065 		udp_detach(inp);
1066 		break;
1067 
1068 	case PRU_BIND:
1069 		s = splsoftnet();
1070 #ifdef INET6
1071 		if (inp->inp_flags & INP_IPV6)
1072 			error = in6_pcbbind(inp, addr);
1073 		else
1074 #endif
1075 			error = in_pcbbind(inp, addr);
1076 		splx(s);
1077 		break;
1078 
1079 	case PRU_LISTEN:
1080 		error = EOPNOTSUPP;
1081 		break;
1082 
1083 	case PRU_CONNECT:
1084 #ifdef INET6
1085 		if (inp->inp_flags & INP_IPV6) {
1086 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1087 				error = EISCONN;
1088 				break;
1089 			}
1090 			s = splsoftnet();
1091 			error = in6_pcbconnect(inp, addr);
1092 			splx(s);
1093 		} else
1094 #endif /* INET6 */
1095 		{
1096 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
1097 				error = EISCONN;
1098 				break;
1099 			}
1100 			s = splsoftnet();
1101 			error = in_pcbconnect(inp, addr);
1102 			splx(s);
1103 		}
1104 
1105 		if (error == 0)
1106 			soisconnected(so);
1107 		break;
1108 
1109 	case PRU_CONNECT2:
1110 		error = EOPNOTSUPP;
1111 		break;
1112 
1113 	case PRU_ACCEPT:
1114 		error = EOPNOTSUPP;
1115 		break;
1116 
1117 	case PRU_DISCONNECT:
1118 #ifdef INET6
1119 		if (inp->inp_flags & INP_IPV6) {
1120 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1121 				error = ENOTCONN;
1122 				break;
1123 			}
1124 		} else
1125 #endif /* INET6 */
1126 		{
1127 			if (inp->inp_faddr.s_addr == INADDR_ANY) {
1128 				error = ENOTCONN;
1129 				break;
1130 			}
1131 		}
1132 
1133 		s = splsoftnet();
1134 		in_pcbdisconnect(inp);
1135 #ifdef INET6
1136 		if (inp->inp_flags & INP_IPV6)
1137 			inp->inp_laddr6 = in6addr_any;
1138 		else
1139 #endif /* INET6 */
1140 			inp->inp_laddr.s_addr = INADDR_ANY;
1141 
1142 		splx(s);
1143 		so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1144 		break;
1145 
1146 	case PRU_SHUTDOWN:
1147 		socantsendmore(so);
1148 		break;
1149 
1150 	case PRU_SEND:
1151 #ifdef INET6
1152 		if (inp->inp_flags & INP_IPV6)
1153 			return (udp6_output(inp, m, addr, control));
1154 		else
1155 			return (udp_output(m, inp, addr, control));
1156 #else
1157 		return (udp_output(m, inp, addr, control));
1158 #endif
1159 
1160 	case PRU_ABORT:
1161 		soisdisconnected(so);
1162 		udp_detach(inp);
1163 		break;
1164 
1165 	case PRU_SOCKADDR:
1166 #ifdef INET6
1167 		if (inp->inp_flags & INP_IPV6)
1168 			in6_setsockaddr(inp, addr);
1169 		else
1170 #endif /* INET6 */
1171 			in_setsockaddr(inp, addr);
1172 		break;
1173 
1174 	case PRU_PEERADDR:
1175 #ifdef INET6
1176 		if (inp->inp_flags & INP_IPV6)
1177 			in6_setpeeraddr(inp, addr);
1178 		else
1179 #endif /* INET6 */
1180 			in_setpeeraddr(inp, addr);
1181 		break;
1182 
1183 	case PRU_SENSE:
1184 		/*
1185 		 * stat: don't bother with a blocksize.
1186 		 */
1187 		/*
1188 		 * Perhaps Path MTU might be returned for a connected
1189 		 * UDP socket in this case.
1190 		 */
1191 		return (0);
1192 
1193 	case PRU_SENDOOB:
1194 	case PRU_FASTTIMO:
1195 	case PRU_SLOWTIMO:
1196 	case PRU_PROTORCV:
1197 	case PRU_PROTOSEND:
1198 		error =  EOPNOTSUPP;
1199 		break;
1200 
1201 	case PRU_RCVD:
1202 	case PRU_RCVOOB:
1203 		return (EOPNOTSUPP);	/* do not free mbuf's */
1204 
1205 	default:
1206 		panic("udp_usrreq");
1207 	}
1208 
1209 release:
1210 	if (control) {
1211 		m_freem(control);
1212 	}
1213 	if (m)
1214 		m_freem(m);
1215 	return (error);
1216 }
1217 
1218 static void
1219 udp_detach(inp)
1220 	struct inpcb *inp;
1221 {
1222 	int s = splsoftnet();
1223 
1224 	in_pcbdetach(inp);
1225 	splx(s);
1226 }
1227 
1228 /*
1229  * Sysctl for udp variables.
1230  */
1231 int
1232 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1233 	int *name;
1234 	u_int namelen;
1235 	void *oldp;
1236 	size_t *oldlenp;
1237 	void *newp;
1238 	size_t newlen;
1239 {
1240 	/* All sysctl names at this level are terminal. */
1241 	if (namelen != 1)
1242 		return (ENOTDIR);
1243 
1244 	switch (name[0]) {
1245 	case UDPCTL_CHECKSUM:
1246 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1247 	case UDPCTL_BADDYNAMIC:
1248 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1249 		    baddynamicports.udp, sizeof(baddynamicports.udp)));
1250 	case UDPCTL_RECVSPACE:
1251 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_recvspace));
1252 	case UDPCTL_SENDSPACE:
1253 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_sendspace));
1254 	default:
1255 		return (ENOPROTOOPT);
1256 	}
1257 	/* NOTREACHED */
1258 }
1259