xref: /openbsd-src/sys/netinet/udp_usrreq.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*	$OpenBSD: udp_usrreq.c,v 1.74 2001/06/25 02:06:40 angelos 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 __P((struct inpcb *));
113 static	void udp_notify __P((struct inpcb *, int));
114 static	struct mbuf *udp_saveopt __P((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 #if __STDC__
155 udp_input(struct mbuf *m, ...)
156 #else
157 udp_input(m, va_alist)
158 	struct mbuf *m;
159 	va_dcl
160 #endif
161 {
162 	register struct ip *ip;
163 	register struct udphdr *uh;
164 	register struct inpcb *inp;
165 	struct mbuf *opts = 0;
166 	int len;
167 	struct ip save_ip;
168 	int iphlen;
169 	va_list ap;
170 	u_int16_t savesum;
171 	union {
172 		struct sockaddr sa;
173 		struct sockaddr_in sin;
174 #ifdef INET6
175 		struct sockaddr_in6 sin6;
176 #endif /* INET6 */
177 	} srcsa, dstsa;
178 #ifdef INET6
179 	struct ip6_hdr *ipv6;
180 #endif /* INET6 */
181 #ifdef IPSEC
182 	struct m_tag *mtag;
183 	struct tdb_ident *tdbi;
184 	struct tdb *tdb;
185 	int error, s;
186 #endif /* IPSEC */
187 
188 	va_start(ap, m);
189 	iphlen = va_arg(ap, int);
190 	va_end(ap);
191 
192 	udpstat.udps_ipackets++;
193 
194 	switch (mtod(m, struct ip *)->ip_v) {
195 	case 4:
196 		ip = mtod(m, struct ip *);
197 #ifdef INET6
198 		ipv6 = NULL;
199 #endif /* INET6 */
200 		srcsa.sa.sa_family = AF_INET;
201 		break;
202 #ifdef INET6
203 	case 6:
204 		ip = NULL;
205 		ipv6 = mtod(m, struct ip6_hdr *);
206 		srcsa.sa.sa_family = AF_INET6;
207 		break;
208 #endif /* INET6 */
209 	default:
210 		goto bad;
211 	}
212 
213 	/*
214 	 * Strip IP options, if any; should skip this,
215 	 * make available to user, and use on returned packets,
216 	 * but we don't yet have a way to check the checksum
217 	 * with options still present.
218 	 */
219 	/*
220 	 * (contd. from above...)  Furthermore, we may want to strip options
221 	 * for such things as ICMP errors, where options just get in the way.
222 	 */
223 	if (ip && iphlen > sizeof (struct ip)) {
224 		ip_stripoptions(m, (struct mbuf *)0);
225 		iphlen = sizeof(struct ip);
226 	}
227 
228 	/*
229 	 * Get IP and UDP header together in first mbuf.
230 	 */
231 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
232 		if ((m = m_pullup2(m, iphlen + sizeof(struct udphdr))) ==
233 		    NULL) {
234 			udpstat.udps_hdrops++;
235 			return;
236 		}
237 #ifdef INET6
238 		if (ipv6)
239 			ipv6 = mtod(m, struct ip6_hdr *);
240 		else
241 #endif /* INET6 */
242 			ip = mtod(m, struct ip *);
243 	}
244 	uh = (struct udphdr *)(mtod(m, caddr_t) + iphlen);
245 
246 	/* Check for illegal destination port 0 */
247 	if (uh->uh_dport == 0) {
248 		udpstat.udps_noport++;
249 		goto bad;
250 	}
251 
252 	/*
253 	 * Make mbuf data length reflect UDP length.
254 	 * If not enough data to reflect UDP length, drop.
255 	 */
256 	len = ntohs((u_int16_t)uh->uh_ulen);
257 	if (m->m_pkthdr.len - iphlen != len) {
258 		if (len > (m->m_pkthdr.len - iphlen) ||
259 			len < sizeof(struct udphdr)) {
260 			udpstat.udps_badlen++;
261 			goto bad;
262 		}
263 		m_adj(m, len - (m->m_pkthdr.len - iphlen));
264 	}
265 	/*
266 	 * Save a copy of the IP header in case we want restore it
267 	 * for sending an ICMP error message in response.
268 	 */
269 	if (ip)
270 		save_ip = *ip;
271 
272 	/*
273 	 * Checksum extended UDP header and data.
274 	 * from W.R.Stevens: check incoming udp cksums even if
275 	 *	udpcksum is not set.
276 	 */
277 	savesum = uh->uh_sum;
278 #ifdef INET6
279 	if (ipv6) {
280 		/* Be proactive about malicious use of IPv4 mapped address */
281 		if (IN6_IS_ADDR_V4MAPPED(&ipv6->ip6_src) ||
282 		    IN6_IS_ADDR_V4MAPPED(&ipv6->ip6_dst)) {
283 			/* XXX stat */
284 			goto bad;
285 		}
286 
287 		/*
288 		 * In IPv6, the UDP checksum is ALWAYS used.
289 		 */
290 		if ((uh->uh_sum = in6_cksum(m, IPPROTO_UDP, iphlen, len))) {
291 			udpstat.udps_badsum++;
292 			goto bad;
293 		}
294 	} else
295 #endif /* INET6 */
296 	if (uh->uh_sum) {
297 		if ((m->m_pkthdr.csum & M_UDP_CSUM_IN_OK) == 0) {
298 			if (m->m_pkthdr.csum & M_UDP_CSUM_IN_BAD) {
299 				udpstat.udps_badsum++;
300 				udpstat.udps_inhwcsum++;
301 				m_freem(m);
302 				return;
303 			}
304 
305 			bzero(((struct ipovly *)ip)->ih_x1,
306 			      sizeof ((struct ipovly *)ip)->ih_x1);
307 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
308 
309 			if ((uh->uh_sum = in_cksum(m, len +
310 			    sizeof (struct ip))) != 0) {
311 				udpstat.udps_badsum++;
312 				m_freem(m);
313 				return;
314 			}
315 		} else {
316 			m->m_pkthdr.csum &= ~M_UDP_CSUM_IN_OK;
317 			udpstat.udps_inhwcsum++;
318 		}
319 	} else
320 		udpstat.udps_nosum++;
321 
322 	switch (srcsa.sa.sa_family) {
323 	case AF_INET:
324 		bzero(&srcsa, sizeof(struct sockaddr_in));
325 		srcsa.sin.sin_len = sizeof(struct sockaddr_in);
326 		srcsa.sin.sin_family = AF_INET;
327 		srcsa.sin.sin_port = uh->uh_sport;
328 		srcsa.sin.sin_addr = ip->ip_src;
329 
330 		bzero(&dstsa, sizeof(struct sockaddr_in));
331 		dstsa.sin.sin_len = sizeof(struct sockaddr_in);
332 		dstsa.sin.sin_family = AF_INET;
333 		dstsa.sin.sin_port = uh->uh_dport;
334 		dstsa.sin.sin_addr = ip->ip_dst;
335 		break;
336 #ifdef INET6
337 	case AF_INET6:
338 		bzero(&srcsa, sizeof(struct sockaddr_in6));
339 		srcsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
340 		srcsa.sin6.sin6_family = AF_INET6;
341 		srcsa.sin6.sin6_port = uh->uh_sport;
342 #if 0 /*XXX inbound flowinfo */
343 		srcsa.sin6.sin6_flowinfo = htonl(0x0fffffff) & ipv6->ip6_flow;
344 #endif
345 		/* KAME hack: recover scopeid */
346 		(void)in6_recoverscope(&srcsa.sin6, &ipv6->ip6_src,
347 		    m->m_pkthdr.rcvif);
348 
349 		bzero(&dstsa, sizeof(struct sockaddr_in6));
350 		dstsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
351 		dstsa.sin6.sin6_family = AF_INET6;
352 		dstsa.sin6.sin6_port = uh->uh_dport;
353 		/* KAME hack: recover scopeid */
354 		(void)in6_recoverscope(&dstsa.sin6, &ipv6->ip6_dst,
355 		    m->m_pkthdr.rcvif);
356 		break;
357 #endif /* INET6 */
358 	}
359 
360 #ifdef INET6
361 	if ((ipv6 && IN6_IS_ADDR_MULTICAST(&ipv6->ip6_dst)) ||
362 	    (ip && IN_MULTICAST(ip->ip_dst.s_addr)) ||
363 	    (ip && in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)))
364 #else /* INET6 */
365 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
366 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))
367 #endif /* INET6 */
368 	{
369 		struct socket *last;
370 		/*
371 		 * Deliver a multicast or broadcast datagram to *all* sockets
372 		 * for which the local and remote addresses and ports match
373 		 * those of the incoming datagram.  This allows more than
374 		 * one process to receive multi/broadcasts on the same port.
375 		 * (This really ought to be done for unicast datagrams as
376 		 * well, but that would cause problems with existing
377 		 * applications that open both address-specific sockets and
378 		 * a wildcard socket listening to the same port -- they would
379 		 * end up receiving duplicates of every unicast datagram.
380 		 * Those applications open the multiple sockets to overcome an
381 		 * inadequacy of the UDP socket interface, but for backwards
382 		 * compatibility we avoid the problem here rather than
383 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
384 		 */
385 
386 		iphlen += sizeof(struct udphdr);
387 
388 		/*
389 		 * Locate pcb(s) for datagram.
390 		 * (Algorithm copied from raw_intr().)
391 		 */
392 		last = NULL;
393 		for (inp = udbtable.inpt_queue.cqh_first;
394 		    inp != (struct inpcb *)&udbtable.inpt_queue;
395 		    inp = inp->inp_queue.cqe_next) {
396 #ifdef INET6
397 			/* don't accept it if AF does not match */
398 			if (ipv6 && !(inp->inp_flags & INP_IPV6))
399 				continue;
400 			if (!ipv6 && (inp->inp_flags & INP_IPV6))
401 				continue;
402 #endif
403 			if (inp->inp_lport != uh->uh_dport)
404 				continue;
405 #ifdef INET6
406 			if (ipv6) {
407 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
408 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
409 					    &ipv6->ip6_dst))
410 						continue;
411 			} else
412 #endif /* INET6 */
413 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
414 				if (inp->inp_laddr.s_addr !=
415 				    ip->ip_dst.s_addr)
416 					continue;
417 			}
418 #ifdef INET6
419 			if (ipv6) {
420 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
421 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
422 					    &ipv6->ip6_src) ||
423 					    inp->inp_fport != uh->uh_sport)
424 			        continue;
425 			} else
426 #endif /* INET6 */
427 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
428 				if (inp->inp_faddr.s_addr !=
429 				    ip->ip_src.s_addr ||
430 				    inp->inp_fport != uh->uh_sport)
431 					continue;
432 			}
433 
434 			if (last != NULL) {
435 				struct mbuf *n;
436 
437 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
438 					opts = NULL;
439 #ifdef INET6
440 					if (ipv6 && (inp->inp_flags & IN6P_CONTROLOPTS))
441 						ip6_savecontrol(inp, &opts, ipv6, n);
442 #endif /* INET6 */
443 					m_adj(n, iphlen);
444 					if (sbappendaddr(&last->so_rcv,
445 					    &srcsa.sa, n, opts) == 0) {
446 						m_freem(n);
447 						if (opts)
448 							m_freem(opts);
449 						udpstat.udps_fullsock++;
450 					} else
451 						sorwakeup(last);
452 					opts = NULL;
453 				}
454 			}
455 			last = inp->inp_socket;
456 			/*
457 			 * Don't look for additional matches if this one does
458 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
459 			 * socket options set.  This heuristic avoids searching
460 			 * through all pcbs in the common case of a non-shared
461 			 * port.  It * assumes that an application will never
462 			 * clear these options after setting them.
463 			 */
464 			if ((last->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
465 				break;
466 		}
467 
468 		if (last == NULL) {
469 			/*
470 			 * No matching pcb found; discard datagram.
471 			 * (No need to send an ICMP Port Unreachable
472 			 * for a broadcast or multicast datgram.)
473 			 */
474 			udpstat.udps_noportbcast++;
475 			goto bad;
476 		}
477 
478 		opts = NULL;
479 #ifdef INET6
480 		if (ipv6 && (inp->inp_flags & IN6P_CONTROLOPTS))
481 			ip6_savecontrol(inp, &opts, ipv6, m);
482 #endif /* INET6 */
483 		m_adj(m, iphlen);
484 		if (sbappendaddr(&last->so_rcv,
485 		    &srcsa.sa, m, opts) == 0) {
486 			udpstat.udps_fullsock++;
487 			goto bad;
488 		}
489 		sorwakeup(last);
490 		return;
491 	}
492 	/*
493 	 * Locate pcb for datagram.
494 	 */
495 #ifdef INET6
496 	if (ipv6)
497 		inp = in6_pcbhashlookup(&udbtable, &ipv6->ip6_src, uh->uh_sport,
498 		    &ipv6->ip6_dst, uh->uh_dport);
499 	else
500 #endif /* INET6 */
501 	inp = in_pcbhashlookup(&udbtable, ip->ip_src, uh->uh_sport,
502 	    ip->ip_dst, uh->uh_dport);
503 	if (inp == 0) {
504 		++udpstat.udps_pcbhashmiss;
505 #ifdef INET6
506 		if (ipv6) {
507 			inp = in_pcblookup(&udbtable,
508 			    (struct in_addr *)&(ipv6->ip6_src),
509 			    uh->uh_sport, (struct in_addr *)&(ipv6->ip6_dst),
510 			    uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_IPV6);
511 		} else
512 #endif /* INET6 */
513 		inp = in_pcblookup(&udbtable, &ip->ip_src, uh->uh_sport,
514 		    &ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD);
515 		if (inp == 0) {
516 			udpstat.udps_noport++;
517 			if (m->m_flags & (M_BCAST | M_MCAST)) {
518 				udpstat.udps_noportbcast++;
519 				goto bad;
520 			}
521 #ifdef INET6
522 			if (ipv6) {
523 				icmp6_error(m, ICMP6_DST_UNREACH,
524 				    ICMP6_DST_UNREACH_NOPORT,0);
525 			} else
526 #endif /* INET6 */
527 			{
528 				*ip = save_ip;
529 				uh->uh_sum = savesum;
530 				icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT,
531 					0, 0);
532 			}
533 			return;
534 		}
535 	}
536 
537 #ifdef IPSEC
538 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
539         s = splnet();
540 	if (mtag != NULL) {
541 		tdbi = (struct tdb_ident *)(mtag + 1);
542 	        tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
543 	} else
544 		tdb = NULL;
545 	ipsp_spd_lookup(m, srcsa.sa.sa_family, iphlen, &error,
546 	    IPSP_DIRECTION_IN, tdb, inp);
547 
548 	/* Latch SA only if the socket is connected */
549 	if (inp->inp_tdb_in != tdb &&
550 	    (inp->inp_socket->so_state & SS_ISCONNECTED)) {
551 		if (tdb) {
552 		        tdb_add_inp(tdb, inp, 1);
553 			if (inp->inp_ipsec_remoteid == NULL &&
554 			    tdb->tdb_srcid != NULL) {
555 				inp->inp_ipsec_remoteid = tdb->tdb_srcid;
556 				tdb->tdb_srcid->ref_count++;
557 			}
558 			if (inp->inp_ipsec_remotecred == NULL &&
559 			    tdb->tdb_remote_cred != NULL) {
560 				inp->inp_ipsec_remotecred =
561 				    tdb->tdb_remote_cred;
562 				tdb->tdb_remote_cred->ref_count++;
563 			}
564 			if (inp->inp_ipsec_remoteauth == NULL &&
565 			    tdb->tdb_remote_auth != NULL) {
566 				inp->inp_ipsec_remoteauth =
567 				    tdb->tdb_remote_auth;
568 				tdb->tdb_remote_auth->ref_count++;
569 			}
570 		} else { /* Just reset */
571 		        TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in, inp,
572 				     inp_tdb_in_next);
573 			inp->inp_tdb_in = NULL;
574 		}
575 	}
576         splx(s);
577 
578 	/* Error or otherwise drop-packet indication. */
579 	if (error)
580 		goto bad;
581 #endif /*IPSEC */
582 
583 	opts = NULL;
584 #ifdef INET6
585 	if (ipv6 && (inp->inp_flags & IN6P_CONTROLOPTS))
586 		ip6_savecontrol(inp, &opts, ipv6, 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) __P((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) __P((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 #if __STDC__
874 udp_output(struct mbuf *m, ...)
875 #else
876 udp_output(m, va_alist)
877 	struct mbuf *m;
878 	va_dcl
879 #endif
880 {
881 	register struct inpcb *inp;
882 	struct mbuf *addr, *control;
883 	register struct udpiphdr *ui;
884 	register int len = m->m_pkthdr.len;
885 	struct in_addr laddr;
886 	int s = 0, error = 0;
887 	va_list ap;
888 	int pcbflags = 0;
889 
890 	va_start(ap, m);
891 	inp = va_arg(ap, struct inpcb *);
892 	addr = va_arg(ap, struct mbuf *);
893 	control = va_arg(ap, struct mbuf *);
894 	va_end(ap);
895 
896 #ifdef DIAGNOSTIC
897 	if ((inp->inp_flags & INP_IPV6) != 0)
898 		panic("IPv6 inpcb to udp_output");
899 #endif
900 
901 	/*
902 	 * Compute the packet length of the IP header, and
903 	 * punt if the length looks bogus.
904 	 */
905 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
906 		error = EMSGSIZE;
907 		goto release;
908 	}
909 
910 	if (addr) {
911 	        /*
912 		 * Save current PCB flags because they may change during
913 		 * temporary connection.
914 		 */
915                 pcbflags = inp->inp_flags;
916 
917 			laddr = inp->inp_laddr;
918 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
919 			error = EISCONN;
920 			goto release;
921 		}
922 		/*
923 		 * Must block input while temporarily connected.
924 		 */
925 		s = splsoftnet();
926 		error = in_pcbconnect(inp, addr);
927 		if (error) {
928 			splx(s);
929 			goto release;
930 		}
931 	} else {
932 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
933 			error = ENOTCONN;
934 			goto release;
935 		}
936 	}
937 	/*
938 	 * Calculate data length and get a mbuf
939 	 * for UDP and IP headers.
940 	 */
941 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
942 	if (m == 0) {
943 		error = ENOBUFS;
944 		goto bail;
945 	}
946 
947 	/*
948 	 * Fill in mbuf with extended UDP header
949 	 * and addresses and length put into network format.
950 	 */
951 	ui = mtod(m, struct udpiphdr *);
952 	bzero(ui->ui_x1, sizeof ui->ui_x1);
953 	ui->ui_pr = IPPROTO_UDP;
954 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
955 	ui->ui_src = inp->inp_laddr;
956 	ui->ui_dst = inp->inp_faddr;
957 	ui->ui_sport = inp->inp_lport;
958 	ui->ui_dport = inp->inp_fport;
959 	ui->ui_ulen = ui->ui_len;
960 
961 	/*
962 	 * Compute the pseudo-header checksum; defer further checksumming
963 	 * until ip_output() or hardware (if it exists).
964 	 */
965 	if (udpcksum) {
966 		m->m_pkthdr.csum |= M_UDPV4_CSUM_OUT;
967 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
968 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
969 			sizeof (struct udphdr) + IPPROTO_UDP));
970 	} else
971 		ui->ui_sum = 0;
972 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
973 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;
974 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;
975 
976 	udpstat.udps_opackets++;
977 	error = ip_output(m, inp->inp_options, &inp->inp_route,
978 		inp->inp_socket->so_options &
979 		(SO_DONTROUTE | SO_BROADCAST),
980 		inp->inp_moptions, inp, NULL);
981 
982 bail:
983 	if (addr) {
984 		in_pcbdisconnect(inp);
985                 inp->inp_flags = pcbflags;
986 			inp->inp_laddr = laddr;
987 		splx(s);
988 	}
989 	if (control)
990 		m_freem(control);
991 	return (error);
992 
993 release:
994 	m_freem(m);
995 	if (control)
996 		m_freem(control);
997 	return (error);
998 }
999 
1000 u_int	udp_sendspace = 9216;		/* really max datagram size */
1001 u_int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
1002 					/* 40 1K datagrams */
1003 
1004 #ifdef INET6
1005 /*ARGSUSED*/
1006 int
1007 udp6_usrreq(so, req, m, addr, control, p)
1008 	struct socket *so;
1009 	int req;
1010 	struct mbuf *m, *addr, *control;
1011 	struct proc *p;
1012 {
1013 
1014 	return udp_usrreq(so, req, m, addr, control);
1015 }
1016 #endif
1017 
1018 /*ARGSUSED*/
1019 int
1020 udp_usrreq(so, req, m, addr, control)
1021 	struct socket *so;
1022 	int req;
1023 	struct mbuf *m, *addr, *control;
1024 {
1025 	struct inpcb *inp = sotoinpcb(so);
1026 	int error = 0;
1027 	int s;
1028 
1029 	if (req == PRU_CONTROL) {
1030 #ifdef INET6
1031 		if (inp->inp_flags & INP_IPV6)
1032 			return (in6_control(so, (u_long)m, (caddr_t)addr,
1033 			    (struct ifnet *)control, 0));
1034 		else
1035 #endif /* INET6 */
1036 			return (in_control(so, (u_long)m, (caddr_t)addr,
1037 			    (struct ifnet *)control));
1038 	}
1039 	if (inp == NULL && req != PRU_ATTACH) {
1040 		error = EINVAL;
1041 		goto release;
1042 	}
1043 	/*
1044 	 * Note: need to block udp_input while changing
1045 	 * the udp pcb queue and/or pcb addresses.
1046 	 */
1047 	switch (req) {
1048 
1049 	case PRU_ATTACH:
1050 		if (inp != NULL) {
1051 			error = EINVAL;
1052 			break;
1053 		}
1054 		s = splsoftnet();
1055 		error = in_pcballoc(so, &udbtable);
1056 		splx(s);
1057 		if (error)
1058 			break;
1059 		error = soreserve(so, udp_sendspace, udp_recvspace);
1060 		if (error)
1061 			break;
1062 #ifdef INET6
1063 		if (((struct inpcb *)so->so_pcb)->inp_flags & INP_IPV6)
1064 			((struct inpcb *) so->so_pcb)->inp_ipv6.ip6_hlim =
1065 			    ip6_defhlim;
1066 		else
1067 #endif /* INET6 */
1068 			((struct inpcb *) so->so_pcb)->inp_ip.ip_ttl = ip_defttl;
1069 		break;
1070 
1071 	case PRU_DETACH:
1072 		udp_detach(inp);
1073 		break;
1074 
1075 	case PRU_BIND:
1076 		s = splsoftnet();
1077 #ifdef INET6
1078 		if (inp->inp_flags & INP_IPV6)
1079 			error = in6_pcbbind(inp, addr);
1080 		else
1081 #endif
1082 			error = in_pcbbind(inp, addr);
1083 		splx(s);
1084 		break;
1085 
1086 	case PRU_LISTEN:
1087 		error = EOPNOTSUPP;
1088 		break;
1089 
1090 	case PRU_CONNECT:
1091 #ifdef INET6
1092 		if (inp->inp_flags & INP_IPV6) {
1093 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1094 				error = EISCONN;
1095 				break;
1096 			}
1097 			s = splsoftnet();
1098 			error = in6_pcbconnect(inp, addr);
1099 			splx(s);
1100 		} else
1101 #endif /* INET6 */
1102 		{
1103 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
1104 				error = EISCONN;
1105 				break;
1106 			}
1107 			s = splsoftnet();
1108 			error = in_pcbconnect(inp, addr);
1109 			splx(s);
1110 		}
1111 
1112 		if (error == 0)
1113 			soisconnected(so);
1114 		break;
1115 
1116 	case PRU_CONNECT2:
1117 		error = EOPNOTSUPP;
1118 		break;
1119 
1120 	case PRU_ACCEPT:
1121 		error = EOPNOTSUPP;
1122 		break;
1123 
1124 	case PRU_DISCONNECT:
1125 #ifdef INET6
1126 		if (inp->inp_flags & INP_IPV6) {
1127 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1128 				error = ENOTCONN;
1129 				break;
1130 			}
1131 		} else
1132 #endif /* INET6 */
1133 		{
1134 			if (inp->inp_faddr.s_addr == INADDR_ANY) {
1135 				error = ENOTCONN;
1136 				break;
1137 			}
1138 		}
1139 
1140 		s = splsoftnet();
1141 		in_pcbdisconnect(inp);
1142 #ifdef INET6
1143 		if (inp->inp_flags & INP_IPV6)
1144 			inp->inp_laddr6 = in6addr_any;
1145 		else
1146 #endif /* INET6 */
1147 			inp->inp_laddr.s_addr = INADDR_ANY;
1148 
1149 		splx(s);
1150 		so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1151 		break;
1152 
1153 	case PRU_SHUTDOWN:
1154 		socantsendmore(so);
1155 		break;
1156 
1157 	case PRU_SEND:
1158 #ifdef INET6
1159 		if (inp->inp_flags & INP_IPV6)
1160 			return (udp6_output(inp, m, addr, control));
1161 		else
1162 			return (udp_output(m, inp, addr, control));
1163 #else
1164 		return (udp_output(m, inp, addr, control));
1165 #endif
1166 
1167 	case PRU_ABORT:
1168 		soisdisconnected(so);
1169 		udp_detach(inp);
1170 		break;
1171 
1172 	case PRU_SOCKADDR:
1173 #ifdef INET6
1174 		if (inp->inp_flags & INP_IPV6)
1175 			in6_setsockaddr(inp, addr);
1176 		else
1177 #endif /* INET6 */
1178 			in_setsockaddr(inp, addr);
1179 		break;
1180 
1181 	case PRU_PEERADDR:
1182 #ifdef INET6
1183 		if (inp->inp_flags & INP_IPV6)
1184 			in6_setpeeraddr(inp, addr);
1185 		else
1186 #endif /* INET6 */
1187 			in_setpeeraddr(inp, addr);
1188 		break;
1189 
1190 	case PRU_SENSE:
1191 		/*
1192 		 * stat: don't bother with a blocksize.
1193 		 */
1194 		/*
1195 		 * Perhaps Path MTU might be returned for a connected
1196 		 * UDP socket in this case.
1197 		 */
1198 		return (0);
1199 
1200 	case PRU_SENDOOB:
1201 	case PRU_FASTTIMO:
1202 	case PRU_SLOWTIMO:
1203 	case PRU_PROTORCV:
1204 	case PRU_PROTOSEND:
1205 		error =  EOPNOTSUPP;
1206 		break;
1207 
1208 	case PRU_RCVD:
1209 	case PRU_RCVOOB:
1210 		return (EOPNOTSUPP);	/* do not free mbuf's */
1211 
1212 	default:
1213 		panic("udp_usrreq");
1214 	}
1215 
1216 release:
1217 	if (control) {
1218 		m_freem(control);
1219 	}
1220 	if (m)
1221 		m_freem(m);
1222 	return (error);
1223 }
1224 
1225 static void
1226 udp_detach(inp)
1227 	struct inpcb *inp;
1228 {
1229 	int s = splsoftnet();
1230 
1231 	in_pcbdetach(inp);
1232 	splx(s);
1233 }
1234 
1235 /*
1236  * Sysctl for udp variables.
1237  */
1238 int
1239 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1240 	int *name;
1241 	u_int namelen;
1242 	void *oldp;
1243 	size_t *oldlenp;
1244 	void *newp;
1245 	size_t newlen;
1246 {
1247 	/* All sysctl names at this level are terminal. */
1248 	if (namelen != 1)
1249 		return (ENOTDIR);
1250 
1251 	switch (name[0]) {
1252 	case UDPCTL_CHECKSUM:
1253 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1254 	case UDPCTL_BADDYNAMIC:
1255 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1256 		    baddynamicports.udp, sizeof(baddynamicports.udp)));
1257 	case UDPCTL_RECVSPACE:
1258 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_recvspace));
1259 	case UDPCTL_SENDSPACE:
1260 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_sendspace));
1261 	default:
1262 		return (ENOPROTOOPT);
1263 	}
1264 	/* NOTREACHED */
1265 }
1266