xref: /dragonfly/sys/netinet/udp_usrreq.c (revision d600454b)
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
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. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
36  *
37  * License terms: all terms for the DragonFly license above plus the following:
38  *
39  * 4. All advertising materials mentioning features or use of this software
40  *    must display the following acknowledgement:
41  *
42  *	This product includes software developed by Jeffrey M. Hsu
43  *	for the DragonFly Project.
44  *
45  *    This requirement may be waived with permission from Jeffrey Hsu.
46  *    This requirement will sunset and may be removed on July 8 2005,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
84  * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
85  * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.36 2006/01/14 11:33:50 swildner Exp $
86  */
87 
88 #include "opt_ipsec.h"
89 #include "opt_inet6.h"
90 
91 #include <sys/param.h>
92 #include <sys/systm.h>
93 #include <sys/kernel.h>
94 #include <sys/malloc.h>
95 #include <sys/mbuf.h>
96 #include <sys/domain.h>
97 #include <sys/proc.h>
98 #include <sys/protosw.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/sysctl.h>
102 #include <sys/syslog.h>
103 #include <sys/thread2.h>
104 #include <sys/in_cksum.h>
105 
106 #include <machine/stdarg.h>
107 
108 #include <vm/vm_zone.h>
109 
110 #include <net/if.h>
111 #include <net/route.h>
112 
113 #include <netinet/in.h>
114 #include <netinet/in_systm.h>
115 #include <netinet/ip.h>
116 #ifdef INET6
117 #include <netinet/ip6.h>
118 #endif
119 #include <netinet/in_pcb.h>
120 #include <netinet/in_var.h>
121 #include <netinet/ip_var.h>
122 #ifdef INET6
123 #include <netinet6/ip6_var.h>
124 #endif
125 #include <netinet/ip_icmp.h>
126 #include <netinet/icmp_var.h>
127 #include <netinet/udp.h>
128 #include <netinet/udp_var.h>
129 
130 #ifdef FAST_IPSEC
131 #include <netproto/ipsec/ipsec.h>
132 #endif
133 
134 #ifdef IPSEC
135 #include <netinet6/ipsec.h>
136 #endif
137 
138 /*
139  * UDP protocol implementation.
140  * Per RFC 768, August, 1980.
141  */
142 #ifndef	COMPAT_42
143 static int	udpcksum = 1;
144 #else
145 static int	udpcksum = 0;		/* XXX */
146 #endif
147 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
148 		&udpcksum, 0, "");
149 
150 int	log_in_vain = 0;
151 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
152     &log_in_vain, 0, "Log all incoming UDP packets");
153 
154 static int	blackhole = 0;
155 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
156 	&blackhole, 0, "Do not send port unreachables for refused connects");
157 
158 static int	strict_mcast_mship = 1;
159 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
160 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
161 
162 struct	inpcbinfo udbinfo;
163 
164 #ifndef UDBHASHSIZE
165 #define UDBHASHSIZE 16
166 #endif
167 
168 struct	udpstat udpstat;	/* from udp_var.h */
169 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
170     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
171 
172 static struct	sockaddr_in udp_in = { sizeof udp_in, AF_INET };
173 #ifdef INET6
174 struct udp_in6 {
175 	struct sockaddr_in6	uin6_sin;
176 	u_char			uin6_init_done : 1;
177 } udp_in6 = {
178 	{ sizeof udp_in6.uin6_sin, AF_INET6 },
179 	0
180 };
181 struct udp_ip6 {
182 	struct ip6_hdr		uip6_ip6;
183 	u_char			uip6_init_done : 1;
184 } udp_ip6;
185 #endif /* INET6 */
186 
187 static void udp_append (struct inpcb *last, struct ip *ip,
188 			    struct mbuf *n, int off);
189 #ifdef INET6
190 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
191 #endif
192 
193 static int udp_detach (struct socket *so);
194 static	int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
195 			    struct mbuf *, struct thread *);
196 
197 void
198 udp_init(void)
199 {
200 	in_pcbinfo_init(&udbinfo);
201 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
202 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
203 					&udbinfo.porthashmask);
204 	udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
205 					    &udbinfo.wildcardhashmask);
206 	udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
207 				 ZONE_INTERRUPT, 0);
208 	udp_thread_init();
209 }
210 
211 /*
212  * Check multicast packets to make sure they are only sent to sockets with
213  * multicast memberships for the packet's destination address and arrival
214  * interface.  Multicast packets to multicast-unaware sockets are also
215  * disallowed.
216  *
217  * Returns 0 if the packet is acceptable, -1 if it is not.
218  */
219 static __inline int
220 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
221 {
222 	int mshipno;
223 	struct ip_moptions *mopt;
224 
225 	if (strict_mcast_mship == 0 ||
226 	    !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
227 		return (0);
228 	}
229 	mopt = inp->inp_moptions;
230 	if (mopt == NULL)
231 		return (-1);
232 	for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
233 		struct in_multi *maddr = mopt->imo_membership[mshipno];
234 
235 		if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
236 		    m->m_pkthdr.rcvif == maddr->inm_ifp) {
237 			return (0);
238 		}
239 	}
240 	return (-1);
241 }
242 
243 void
244 udp_input(struct mbuf *m, ...)
245 {
246 	int iphlen;
247 	struct ip *ip;
248 	struct udphdr *uh;
249 	struct inpcb *inp;
250 	struct mbuf *opts = NULL;
251 	int len, off, proto;
252 	struct ip save_ip;
253 	struct sockaddr *append_sa;
254 	__va_list ap;
255 
256 	__va_start(ap, m);
257 	off = __va_arg(ap, int);
258 	proto = __va_arg(ap, int);
259 	__va_end(ap);
260 
261 	iphlen = off;
262 	udpstat.udps_ipackets++;
263 
264 	/*
265 	 * Strip IP options, if any; should skip this,
266 	 * make available to user, and use on returned packets,
267 	 * but we don't yet have a way to check the checksum
268 	 * with options still present.
269 	 */
270 	if (iphlen > sizeof(struct ip)) {
271 		ip_stripoptions(m);
272 		iphlen = sizeof(struct ip);
273 	}
274 
275 	/*
276 	 * IP and UDP headers are together in first mbuf.
277 	 * Already checked and pulled up in ip_demux().
278 	 */
279 	KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
280 	    ("UDP header not in one mbuf"));
281 
282 	ip = mtod(m, struct ip *);
283 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
284 
285 	/* destination port of 0 is illegal, based on RFC768. */
286 	if (uh->uh_dport == 0)
287 		goto bad;
288 
289 	/*
290 	 * Make mbuf data length reflect UDP length.
291 	 * If not enough data to reflect UDP length, drop.
292 	 */
293 	len = ntohs((u_short)uh->uh_ulen);
294 	if (ip->ip_len != len) {
295 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
296 			udpstat.udps_badlen++;
297 			goto bad;
298 		}
299 		m_adj(m, len - ip->ip_len);
300 		/* ip->ip_len = len; */
301 	}
302 	/*
303 	 * Save a copy of the IP header in case we want restore it
304 	 * for sending an ICMP error message in response.
305 	 */
306 	save_ip = *ip;
307 
308 	/*
309 	 * Checksum extended UDP header and data.
310 	 */
311 	if (uh->uh_sum) {
312 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
313 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
314 				uh->uh_sum = m->m_pkthdr.csum_data;
315 			else
316 				uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
317 				    ip->ip_dst.s_addr, htonl((u_short)len +
318 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
319 			uh->uh_sum ^= 0xffff;
320 		} else {
321 			char b[9];
322 
323 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
324 			bzero(((struct ipovly *)ip)->ih_x1, 9);
325 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
326 			uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
327 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
328 		}
329 		if (uh->uh_sum) {
330 			udpstat.udps_badsum++;
331 			m_freem(m);
332 			return;
333 		}
334 	} else
335 		udpstat.udps_nosum++;
336 
337 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
338 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
339 		struct inpcb *last;
340 
341 		/*
342 		 * Deliver a multicast or broadcast datagram to *all* sockets
343 		 * for which the local and remote addresses and ports match
344 		 * those of the incoming datagram.  This allows more than
345 		 * one process to receive multi/broadcasts on the same port.
346 		 * (This really ought to be done for unicast datagrams as
347 		 * well, but that would cause problems with existing
348 		 * applications that open both address-specific sockets and
349 		 * a wildcard socket listening to the same port -- they would
350 		 * end up receiving duplicates of every unicast datagram.
351 		 * Those applications open the multiple sockets to overcome an
352 		 * inadequacy of the UDP socket interface, but for backwards
353 		 * compatibility we avoid the problem here rather than
354 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
355 		 */
356 
357 		/*
358 		 * Construct sockaddr format source address.
359 		 */
360 		udp_in.sin_port = uh->uh_sport;
361 		udp_in.sin_addr = ip->ip_src;
362 		/*
363 		 * Locate pcb(s) for datagram.
364 		 * (Algorithm copied from raw_intr().)
365 		 */
366 		last = NULL;
367 #ifdef INET6
368 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
369 #endif
370 		LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
371 			if (inp->inp_flags & INP_PLACEMARKER)
372 				continue;
373 #ifdef INET6
374 			if (!(inp->inp_vflag & INP_IPV4))
375 				continue;
376 #endif
377 			if (inp->inp_lport != uh->uh_dport)
378 				continue;
379 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
380 				if (inp->inp_laddr.s_addr !=
381 				    ip->ip_dst.s_addr)
382 					continue;
383 			}
384 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
385 				if (inp->inp_faddr.s_addr !=
386 				    ip->ip_src.s_addr ||
387 				    inp->inp_fport != uh->uh_sport)
388 					continue;
389 			}
390 
391 			if (check_multicast_membership(ip, inp, m) < 0)
392 				continue;
393 
394 			if (last != NULL) {
395 				struct mbuf *n;
396 
397 #ifdef IPSEC
398 				/* check AH/ESP integrity. */
399 				if (ipsec4_in_reject_so(m, last->inp_socket))
400 					ipsecstat.in_polvio++;
401 					/* do not inject data to pcb */
402 				else
403 #endif /*IPSEC*/
404 #ifdef FAST_IPSEC
405 				/* check AH/ESP integrity. */
406 				if (ipsec4_in_reject(m, last))
407 					;
408 				else
409 #endif /*FAST_IPSEC*/
410 				if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
411 					udp_append(last, ip, n,
412 						   iphlen +
413 						   sizeof(struct udphdr));
414 			}
415 			last = inp;
416 			/*
417 			 * Don't look for additional matches if this one does
418 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
419 			 * socket options set.  This heuristic avoids searching
420 			 * through all pcbs in the common case of a non-shared
421 			 * port.  It * assumes that an application will never
422 			 * clear these options after setting them.
423 			 */
424 			if (!(last->inp_socket->so_options &
425 			    (SO_REUSEPORT | SO_REUSEADDR)))
426 				break;
427 		}
428 
429 		if (last == NULL) {
430 			/*
431 			 * No matching pcb found; discard datagram.
432 			 * (No need to send an ICMP Port Unreachable
433 			 * for a broadcast or multicast datgram.)
434 			 */
435 			udpstat.udps_noportbcast++;
436 			goto bad;
437 		}
438 #ifdef IPSEC
439 		/* check AH/ESP integrity. */
440 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
441 			ipsecstat.in_polvio++;
442 			goto bad;
443 		}
444 #endif /*IPSEC*/
445 #ifdef FAST_IPSEC
446 		/* check AH/ESP integrity. */
447 		if (ipsec4_in_reject(m, last))
448 			goto bad;
449 #endif /*FAST_IPSEC*/
450 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
451 		return;
452 	}
453 	/*
454 	 * Locate pcb for datagram.
455 	 */
456 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
457 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
458 	if (inp == NULL) {
459 		if (log_in_vain) {
460 			char buf[sizeof "aaa.bbb.ccc.ddd"];
461 
462 			strcpy(buf, inet_ntoa(ip->ip_dst));
463 			log(LOG_INFO,
464 			    "Connection attempt to UDP %s:%d from %s:%d\n",
465 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
466 			    ntohs(uh->uh_sport));
467 		}
468 		udpstat.udps_noport++;
469 		if (m->m_flags & (M_BCAST | M_MCAST)) {
470 			udpstat.udps_noportbcast++;
471 			goto bad;
472 		}
473 		if (blackhole)
474 			goto bad;
475 #ifdef ICMP_BANDLIM
476 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
477 			goto bad;
478 #endif
479 		*ip = save_ip;
480 		ip->ip_len += iphlen;
481 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
482 		return;
483 	}
484 #ifdef IPSEC
485 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
486 		ipsecstat.in_polvio++;
487 		goto bad;
488 	}
489 #endif /*IPSEC*/
490 #ifdef FAST_IPSEC
491 	if (ipsec4_in_reject(m, inp))
492 		goto bad;
493 #endif /*FAST_IPSEC*/
494 
495 	/*
496 	 * Construct sockaddr format source address.
497 	 * Stuff source address and datagram in user buffer.
498 	 */
499 	udp_in.sin_port = uh->uh_sport;
500 	udp_in.sin_addr = ip->ip_src;
501 	if ((inp->inp_flags & INP_CONTROLOPTS) ||
502 	    (inp->inp_socket->so_options & SO_TIMESTAMP)) {
503 #ifdef INET6
504 		if (inp->inp_vflag & INP_IPV6) {
505 			int savedflags;
506 
507 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
508 			savedflags = inp->inp_flags;
509 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
510 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
511 			inp->inp_flags = savedflags;
512 		} else
513 #endif
514 		ip_savecontrol(inp, &opts, ip, m);
515 	}
516 	m_adj(m, iphlen + sizeof(struct udphdr));
517 #ifdef INET6
518 	if (inp->inp_vflag & INP_IPV6) {
519 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
520 		append_sa = (struct sockaddr *)&udp_in6;
521 	} else
522 #endif
523 		append_sa = (struct sockaddr *)&udp_in;
524 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
525 		udpstat.udps_fullsock++;
526 		goto bad;
527 	}
528 	sorwakeup(inp->inp_socket);
529 	return;
530 bad:
531 	m_freem(m);
532 	if (opts)
533 		m_freem(opts);
534 	return;
535 }
536 
537 #ifdef INET6
538 static void
539 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
540 {
541 	bzero(ip6, sizeof *ip6);
542 
543 	ip6->ip6_vfc = IPV6_VERSION;
544 	ip6->ip6_plen = ip->ip_len;
545 	ip6->ip6_nxt = ip->ip_p;
546 	ip6->ip6_hlim = ip->ip_ttl;
547 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
548 		IPV6_ADDR_INT32_SMP;
549 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
550 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
551 }
552 #endif
553 
554 /*
555  * subroutine of udp_input(), mainly for source code readability.
556  * caller must properly init udp_ip6 and udp_in6 beforehand.
557  */
558 static void
559 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
560 {
561 	struct sockaddr *append_sa;
562 	struct mbuf *opts = NULL;
563 
564 	if (last->inp_flags & INP_CONTROLOPTS ||
565 	    last->inp_socket->so_options & SO_TIMESTAMP) {
566 #ifdef INET6
567 		if (last->inp_vflag & INP_IPV6) {
568 			int savedflags;
569 
570 			if (udp_ip6.uip6_init_done == 0) {
571 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
572 				udp_ip6.uip6_init_done = 1;
573 			}
574 			savedflags = last->inp_flags;
575 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
576 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
577 			last->inp_flags = savedflags;
578 		} else
579 #endif
580 		ip_savecontrol(last, &opts, ip, n);
581 	}
582 #ifdef INET6
583 	if (last->inp_vflag & INP_IPV6) {
584 		if (udp_in6.uin6_init_done == 0) {
585 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
586 			udp_in6.uin6_init_done = 1;
587 		}
588 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
589 	} else
590 #endif
591 		append_sa = (struct sockaddr *)&udp_in;
592 	m_adj(n, off);
593 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
594 		m_freem(n);
595 		if (opts)
596 			m_freem(opts);
597 		udpstat.udps_fullsock++;
598 	} else
599 		sorwakeup(last->inp_socket);
600 }
601 
602 /*
603  * Notify a udp user of an asynchronous error;
604  * just wake up so that he can collect error status.
605  */
606 void
607 udp_notify(struct inpcb *inp, int errno)
608 {
609 	inp->inp_socket->so_error = errno;
610 	sorwakeup(inp->inp_socket);
611 	sowwakeup(inp->inp_socket);
612 }
613 
614 void
615 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
616 {
617 	struct ip *ip = vip;
618 	struct udphdr *uh;
619 	void (*notify) (struct inpcb *, int) = udp_notify;
620 	struct in_addr faddr;
621 	struct inpcb *inp;
622 
623 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
624 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
625 		return;
626 
627 	if (PRC_IS_REDIRECT(cmd)) {
628 		ip = NULL;
629 		notify = in_rtchange;
630 	} else if (cmd == PRC_HOSTDEAD)
631 		ip = NULL;
632 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
633 		return;
634 	if (ip) {
635 		crit_enter();
636 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
637 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
638 					ip->ip_src, uh->uh_sport, 0, NULL);
639 		if (inp != NULL && inp->inp_socket != NULL)
640 			(*notify)(inp, inetctlerrmap[cmd]);
641 		crit_exit();
642 	} else
643 		in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
644 				notify);
645 }
646 
647 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
648 	    in_pcblist_global, "S,xinpcb", "List of active UDP sockets");
649 
650 static int
651 udp_getcred(SYSCTL_HANDLER_ARGS)
652 {
653 	struct sockaddr_in addrs[2];
654 	struct inpcb *inp;
655 	int error;
656 
657 	error = suser(req->td);
658 	if (error)
659 		return (error);
660 	error = SYSCTL_IN(req, addrs, sizeof addrs);
661 	if (error)
662 		return (error);
663 	crit_enter();
664 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
665 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
666 	if (inp == NULL || inp->inp_socket == NULL) {
667 		error = ENOENT;
668 		goto out;
669 	}
670 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
671 out:
672 	crit_exit();
673 	return (error);
674 }
675 
676 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
677     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
678 
679 static int
680 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
681 	   struct mbuf *control, struct thread *td)
682 {
683 	struct udpiphdr *ui;
684 	int len = m->m_pkthdr.len;
685 	struct sockaddr_in *sin;	/* really is initialized before use */
686 	int error = 0;
687 
688 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
689 		error = EMSGSIZE;
690 		goto release;
691 	}
692 
693 	if (inp->inp_lport == 0) {	/* unbound socket */
694 		error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
695 		if (error)
696 			goto release;
697 		in_pcbinswildcardhash(inp);
698 	}
699 
700 	if (dstaddr != NULL) {		/* destination address specified */
701 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
702 			/* already connected */
703 			error = EISCONN;
704 			goto release;
705 		}
706 		sin = (struct sockaddr_in *)dstaddr;
707 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
708 	} else {
709 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
710 			/* no destination specified and not already connected */
711 			error = ENOTCONN;
712 			goto release;
713 		}
714 		sin = NULL;
715 	}
716 
717 	/*
718 	 * Calculate data length and get a mbuf
719 	 * for UDP and IP headers.
720 	 */
721 	M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
722 	if (m == NULL) {
723 		error = ENOBUFS;
724 		goto release;
725 	}
726 
727 	/*
728 	 * Fill in mbuf with extended UDP header
729 	 * and addresses and length put into network format.
730 	 */
731 	ui = mtod(m, struct udpiphdr *);
732 	bzero(ui->ui_x1, sizeof ui->ui_x1);	/* XXX still needed? */
733 	ui->ui_pr = IPPROTO_UDP;
734 
735 	/*
736 	 * Set destination address.
737 	 */
738 	if (dstaddr != NULL) {			/* use specified destination */
739 		ui->ui_dst = sin->sin_addr;
740 		ui->ui_dport = sin->sin_port;
741 	} else {				/* use connected destination */
742 		ui->ui_dst = inp->inp_faddr;
743 		ui->ui_dport = inp->inp_fport;
744 	}
745 
746 	/*
747 	 * Set source address.
748 	 */
749 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
750 		struct sockaddr_in *if_sin;
751 
752 		if (dstaddr == NULL) {
753 			/*
754 			 * connect() had (or should have) failed because
755 			 * the interface had no IP address, but the
756 			 * application proceeded to call send() anyways.
757 			 */
758 			error = ENOTCONN;
759 			goto release;
760 		}
761 
762 		/* Look up outgoing interface. */
763 		if ((error = in_pcbladdr(inp, dstaddr, &if_sin)))
764 			goto release;
765 		ui->ui_src = if_sin->sin_addr;	/* use address of interface */
766 	} else {
767 		ui->ui_src = inp->inp_laddr;	/* use non-null bound address */
768 	}
769 	ui->ui_sport = inp->inp_lport;
770 	KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
771 
772 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
773 
774 	/*
775 	 * Set up checksum and output datagram.
776 	 */
777 	if (udpcksum) {
778 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
779 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
780 		m->m_pkthdr.csum_flags = CSUM_UDP;
781 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
782 	} else {
783 		ui->ui_sum = 0;
784 	}
785 	((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
786 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
787 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
788 	udpstat.udps_opackets++;
789 
790 	error = ip_output(m, inp->inp_options, &inp->inp_route,
791 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
792 	    inp->inp_moptions, inp);
793 
794 	return (error);
795 
796 release:
797 	m_freem(m);
798 	return (error);
799 }
800 
801 u_long	udp_sendspace = 9216;		/* really max datagram size */
802 					/* 40 1K datagrams */
803 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
804     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
805 
806 u_long	udp_recvspace = 40 * (1024 +
807 #ifdef INET6
808 				      sizeof(struct sockaddr_in6)
809 #else
810 				      sizeof(struct sockaddr_in)
811 #endif
812 				      );
813 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
814     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
815 
816 static int
817 udp_abort(struct socket *so)
818 {
819 	struct inpcb *inp;
820 
821 	inp = so->so_pcb;
822 	if (inp == NULL)
823 		return EINVAL;	/* ??? possible? panic instead? */
824 	soisdisconnected(so);
825 	crit_enter();
826 	in_pcbdetach(inp);
827 	crit_exit();
828 	return 0;
829 }
830 
831 static int
832 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai)
833 {
834 	struct inpcb *inp;
835 	int error;
836 
837 	inp = so->so_pcb;
838 	if (inp != NULL)
839 		return EINVAL;
840 
841 	error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
842 	if (error)
843 		return error;
844 	crit_enter();
845 	error = in_pcballoc(so, &udbinfo);
846 	crit_exit();
847 	if (error)
848 		return error;
849 
850 	inp = (struct inpcb *)so->so_pcb;
851 	inp->inp_vflag |= INP_IPV4;
852 	inp->inp_ip_ttl = ip_defttl;
853 	return 0;
854 }
855 
856 static int
857 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
858 {
859 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
860 	struct inpcb *inp;
861 	int error;
862 
863 	inp = so->so_pcb;
864 	if (inp == NULL)
865 		return EINVAL;
866 	crit_enter();
867 	error = in_pcbbind(inp, nam, td);
868 	crit_exit();
869 	if (error == 0) {
870 		if (sin->sin_addr.s_addr != INADDR_ANY)
871 			inp->inp_flags |= INP_WASBOUND_NOTANY;
872 		in_pcbinswildcardhash(inp);
873 	}
874 	return error;
875 }
876 
877 static int
878 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
879 {
880 	struct inpcb *inp;
881 	int error;
882 	struct sockaddr_in *sin;
883 
884 	inp = so->so_pcb;
885 	if (inp == NULL)
886 		return EINVAL;
887 	if (inp->inp_faddr.s_addr != INADDR_ANY)
888 		return EISCONN;
889 	error = 0;
890 	crit_enter();
891 	if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
892 	    inp->inp_laddr.s_addr == INADDR_ANY) {
893 		error = in_pcbbind(inp, NULL, td);
894 	}
895 	if (error == 0) {
896 		sin = (struct sockaddr_in *)nam;
897 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
898 		if (inp->inp_flags & INP_WILDCARD)
899 			in_pcbremwildcardhash(inp);
900 		error = in_pcbconnect(inp, nam, td);
901 	}
902 	crit_exit();
903 	if (error == 0)
904 		soisconnected(so);
905 	else if (error == EAFNOSUPPORT) {	/* connection dissolved */
906 		/*
907 		 * Follow traditional BSD behavior and retain
908 		 * the local port binding.  But, fix the old misbehavior
909 		 * of overwriting any previously bound local address.
910 		 */
911 		if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
912 			inp->inp_laddr.s_addr = INADDR_ANY;
913 		in_pcbinswildcardhash(inp);
914 	}
915 	return error;
916 }
917 
918 static int
919 udp_detach(struct socket *so)
920 {
921 	struct inpcb *inp;
922 
923 	inp = so->so_pcb;
924 	if (inp == NULL)
925 		return EINVAL;
926 	crit_enter();
927 	in_pcbdetach(inp);
928 	crit_exit();
929 	return 0;
930 }
931 
932 static int
933 udp_disconnect(struct socket *so)
934 {
935 	struct inpcb *inp;
936 
937 	inp = so->so_pcb;
938 	if (inp == NULL)
939 		return EINVAL;
940 	if (inp->inp_faddr.s_addr == INADDR_ANY)
941 		return ENOTCONN;
942 
943 	crit_enter();
944 	in_pcbdisconnect(inp);
945 	crit_exit();
946 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
947 	return 0;
948 }
949 
950 static int
951 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
952 	    struct mbuf *control, struct thread *td)
953 {
954 	struct inpcb *inp;
955 
956 	inp = so->so_pcb;
957 	if (inp == NULL) {
958 		m_freem(m);
959 		return EINVAL;
960 	}
961 	return udp_output(inp, m, addr, control, td);
962 }
963 
964 int
965 udp_shutdown(struct socket *so)
966 {
967 	struct inpcb *inp;
968 
969 	inp = so->so_pcb;
970 	if (inp == NULL)
971 		return EINVAL;
972 	socantsendmore(so);
973 	return 0;
974 }
975 
976 struct pr_usrreqs udp_usrreqs = {
977 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
978 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
979 	pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
980 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
981 	in_setsockaddr, sosendudp, soreceive, sopoll
982 };
983 
984