xref: /dflybsd-src/sys/netinet/udp_usrreq.c (revision c3c96e4421a1087a390825eac6c01c9ed9182387)
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) 1982, 1986, 1988, 1990, 1993, 1995
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
67  * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
68  * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.47 2008/11/11 10:46:58 sephe Exp $
69  */
70 
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/proc.h>
81 #include <sys/priv.h>
82 #include <sys/protosw.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 #include <sys/thread2.h>
88 #include <sys/in_cksum.h>
89 
90 #include <machine/stdarg.h>
91 
92 #include <net/if.h>
93 #include <net/route.h>
94 #include <net/netmsg2.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/ip.h>
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #endif
102 #include <netinet/in_pcb.h>
103 #include <netinet/in_var.h>
104 #include <netinet/ip_var.h>
105 #ifdef INET6
106 #include <netinet6/ip6_var.h>
107 #endif
108 #include <netinet/ip_icmp.h>
109 #include <netinet/icmp_var.h>
110 #include <netinet/udp.h>
111 #include <netinet/udp_var.h>
112 
113 #ifdef FAST_IPSEC
114 #include <netproto/ipsec/ipsec.h>
115 #endif
116 
117 #ifdef IPSEC
118 #include <netinet6/ipsec.h>
119 #endif
120 
121 /*
122  * UDP protocol implementation.
123  * Per RFC 768, August, 1980.
124  */
125 #ifndef	COMPAT_42
126 static int	udpcksum = 1;
127 #else
128 static int	udpcksum = 0;		/* XXX */
129 #endif
130 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
131 		&udpcksum, 0, "");
132 
133 int	log_in_vain = 0;
134 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
135     &log_in_vain, 0, "Log all incoming UDP packets");
136 
137 static int	blackhole = 0;
138 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
139 	&blackhole, 0, "Do not send port unreachables for refused connects");
140 
141 static int	strict_mcast_mship = 1;
142 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
143 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
144 
145 struct	inpcbinfo udbinfo;
146 
147 #ifndef UDBHASHSIZE
148 #define UDBHASHSIZE 16
149 #endif
150 
151 struct	udpstat udpstat;	/* from udp_var.h */
152 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
153     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
154 
155 static struct	sockaddr_in udp_in = { sizeof udp_in, AF_INET };
156 #ifdef INET6
157 struct udp_in6 {
158 	struct sockaddr_in6	uin6_sin;
159 	u_char			uin6_init_done : 1;
160 } udp_in6 = {
161 	{ sizeof udp_in6.uin6_sin, AF_INET6 },
162 	0
163 };
164 struct udp_ip6 {
165 	struct ip6_hdr		uip6_ip6;
166 	u_char			uip6_init_done : 1;
167 } udp_ip6;
168 #endif /* INET6 */
169 
170 static void udp_append (struct inpcb *last, struct ip *ip,
171 			    struct mbuf *n, int off);
172 #ifdef INET6
173 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
174 #endif
175 
176 static int udp_connect_oncpu(struct socket *so, struct thread *td,
177 			struct sockaddr_in *sin, struct sockaddr_in *if_sin);
178 static int udp_detach (struct socket *so);
179 static	int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
180 			    struct mbuf *, struct thread *);
181 
182 void
183 udp_init(void)
184 {
185 	in_pcbinfo_init(&udbinfo);
186 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
187 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
188 					&udbinfo.porthashmask);
189 	udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
190 					    &udbinfo.wildcardhashmask);
191 	udbinfo.ipi_size = sizeof(struct inpcb);
192 }
193 
194 /*
195  * Check multicast packets to make sure they are only sent to sockets with
196  * multicast memberships for the packet's destination address and arrival
197  * interface.  Multicast packets to multicast-unaware sockets are also
198  * disallowed.
199  *
200  * Returns 0 if the packet is acceptable, -1 if it is not.
201  */
202 static __inline int
203 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
204 {
205 	int mshipno;
206 	struct ip_moptions *mopt;
207 
208 	if (strict_mcast_mship == 0 ||
209 	    !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
210 		return (0);
211 	}
212 	mopt = inp->inp_moptions;
213 	if (mopt == NULL)
214 		return (-1);
215 	for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
216 		struct in_multi *maddr = mopt->imo_membership[mshipno];
217 
218 		if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
219 		    m->m_pkthdr.rcvif == maddr->inm_ifp) {
220 			return (0);
221 		}
222 	}
223 	return (-1);
224 }
225 
226 void
227 udp_input(struct mbuf *m, ...)
228 {
229 	int iphlen;
230 	struct ip *ip;
231 	struct udphdr *uh;
232 	struct inpcb *inp;
233 	struct mbuf *opts = NULL;
234 	int len, off, proto;
235 	struct ip save_ip;
236 	struct sockaddr *append_sa;
237 	__va_list ap;
238 
239 	__va_start(ap, m);
240 	off = __va_arg(ap, int);
241 	proto = __va_arg(ap, int);
242 	__va_end(ap);
243 
244 	iphlen = off;
245 	udpstat.udps_ipackets++;
246 
247 	/*
248 	 * Strip IP options, if any; should skip this,
249 	 * make available to user, and use on returned packets,
250 	 * but we don't yet have a way to check the checksum
251 	 * with options still present.
252 	 */
253 	if (iphlen > sizeof(struct ip)) {
254 		ip_stripoptions(m);
255 		iphlen = sizeof(struct ip);
256 	}
257 
258 	/*
259 	 * IP and UDP headers are together in first mbuf.
260 	 * Already checked and pulled up in ip_demux().
261 	 */
262 	KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
263 	    ("UDP header not in one mbuf"));
264 
265 	ip = mtod(m, struct ip *);
266 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
267 
268 	/* destination port of 0 is illegal, based on RFC768. */
269 	if (uh->uh_dport == 0)
270 		goto bad;
271 
272 	/*
273 	 * Make mbuf data length reflect UDP length.
274 	 * If not enough data to reflect UDP length, drop.
275 	 */
276 	len = ntohs((u_short)uh->uh_ulen);
277 	if (ip->ip_len != len) {
278 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
279 			udpstat.udps_badlen++;
280 			goto bad;
281 		}
282 		m_adj(m, len - ip->ip_len);
283 		/* ip->ip_len = len; */
284 	}
285 	/*
286 	 * Save a copy of the IP header in case we want restore it
287 	 * for sending an ICMP error message in response.
288 	 */
289 	save_ip = *ip;
290 
291 	/*
292 	 * Checksum extended UDP header and data.
293 	 */
294 	if (uh->uh_sum) {
295 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
296 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
297 				uh->uh_sum = m->m_pkthdr.csum_data;
298 			else
299 				uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
300 				    ip->ip_dst.s_addr, htonl((u_short)len +
301 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
302 			uh->uh_sum ^= 0xffff;
303 		} else {
304 			char b[9];
305 
306 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
307 			bzero(((struct ipovly *)ip)->ih_x1, 9);
308 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
309 			uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
310 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
311 		}
312 		if (uh->uh_sum) {
313 			udpstat.udps_badsum++;
314 			m_freem(m);
315 			return;
316 		}
317 	} else
318 		udpstat.udps_nosum++;
319 
320 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
321 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
322 		struct inpcb *last;
323 
324 		/*
325 		 * Deliver a multicast or broadcast datagram to *all* sockets
326 		 * for which the local and remote addresses and ports match
327 		 * those of the incoming datagram.  This allows more than
328 		 * one process to receive multi/broadcasts on the same port.
329 		 * (This really ought to be done for unicast datagrams as
330 		 * well, but that would cause problems with existing
331 		 * applications that open both address-specific sockets and
332 		 * a wildcard socket listening to the same port -- they would
333 		 * end up receiving duplicates of every unicast datagram.
334 		 * Those applications open the multiple sockets to overcome an
335 		 * inadequacy of the UDP socket interface, but for backwards
336 		 * compatibility we avoid the problem here rather than
337 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
338 		 */
339 
340 		/*
341 		 * Construct sockaddr format source address.
342 		 */
343 		udp_in.sin_port = uh->uh_sport;
344 		udp_in.sin_addr = ip->ip_src;
345 		/*
346 		 * Locate pcb(s) for datagram.
347 		 * (Algorithm copied from raw_intr().)
348 		 */
349 		last = NULL;
350 #ifdef INET6
351 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
352 #endif
353 		LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
354 			if (inp->inp_flags & INP_PLACEMARKER)
355 				continue;
356 #ifdef INET6
357 			if (!(inp->inp_vflag & INP_IPV4))
358 				continue;
359 #endif
360 			if (inp->inp_lport != uh->uh_dport)
361 				continue;
362 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
363 				if (inp->inp_laddr.s_addr !=
364 				    ip->ip_dst.s_addr)
365 					continue;
366 			}
367 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
368 				if (inp->inp_faddr.s_addr !=
369 				    ip->ip_src.s_addr ||
370 				    inp->inp_fport != uh->uh_sport)
371 					continue;
372 			}
373 
374 			if (check_multicast_membership(ip, inp, m) < 0)
375 				continue;
376 
377 			if (last != NULL) {
378 				struct mbuf *n;
379 
380 #ifdef IPSEC
381 				/* check AH/ESP integrity. */
382 				if (ipsec4_in_reject_so(m, last->inp_socket))
383 					ipsecstat.in_polvio++;
384 					/* do not inject data to pcb */
385 				else
386 #endif /*IPSEC*/
387 #ifdef FAST_IPSEC
388 				/* check AH/ESP integrity. */
389 				if (ipsec4_in_reject(m, last))
390 					;
391 				else
392 #endif /*FAST_IPSEC*/
393 				if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
394 					udp_append(last, ip, n,
395 						   iphlen +
396 						   sizeof(struct udphdr));
397 			}
398 			last = inp;
399 			/*
400 			 * Don't look for additional matches if this one does
401 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
402 			 * socket options set.  This heuristic avoids searching
403 			 * through all pcbs in the common case of a non-shared
404 			 * port.  It * assumes that an application will never
405 			 * clear these options after setting them.
406 			 */
407 			if (!(last->inp_socket->so_options &
408 			    (SO_REUSEPORT | SO_REUSEADDR)))
409 				break;
410 		}
411 
412 		if (last == NULL) {
413 			/*
414 			 * No matching pcb found; discard datagram.
415 			 * (No need to send an ICMP Port Unreachable
416 			 * for a broadcast or multicast datgram.)
417 			 */
418 			udpstat.udps_noportbcast++;
419 			goto bad;
420 		}
421 #ifdef IPSEC
422 		/* check AH/ESP integrity. */
423 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
424 			ipsecstat.in_polvio++;
425 			goto bad;
426 		}
427 #endif /*IPSEC*/
428 #ifdef FAST_IPSEC
429 		/* check AH/ESP integrity. */
430 		if (ipsec4_in_reject(m, last))
431 			goto bad;
432 #endif /*FAST_IPSEC*/
433 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
434 		return;
435 	}
436 	/*
437 	 * Locate pcb for datagram.
438 	 */
439 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
440 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
441 	if (inp == NULL) {
442 		if (log_in_vain) {
443 			char buf[sizeof "aaa.bbb.ccc.ddd"];
444 
445 			strcpy(buf, inet_ntoa(ip->ip_dst));
446 			log(LOG_INFO,
447 			    "Connection attempt to UDP %s:%d from %s:%d\n",
448 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
449 			    ntohs(uh->uh_sport));
450 		}
451 		udpstat.udps_noport++;
452 		if (m->m_flags & (M_BCAST | M_MCAST)) {
453 			udpstat.udps_noportbcast++;
454 			goto bad;
455 		}
456 		if (blackhole)
457 			goto bad;
458 #ifdef ICMP_BANDLIM
459 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
460 			goto bad;
461 #endif
462 		*ip = save_ip;
463 		ip->ip_len += iphlen;
464 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
465 		return;
466 	}
467 #ifdef IPSEC
468 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
469 		ipsecstat.in_polvio++;
470 		goto bad;
471 	}
472 #endif /*IPSEC*/
473 #ifdef FAST_IPSEC
474 	if (ipsec4_in_reject(m, inp))
475 		goto bad;
476 #endif /*FAST_IPSEC*/
477 	/*
478 	 * Check the minimum TTL for socket.
479 	 */
480 	if (ip->ip_ttl < inp->inp_ip_minttl)
481 		goto bad;
482 
483 	/*
484 	 * Construct sockaddr format source address.
485 	 * Stuff source address and datagram in user buffer.
486 	 */
487 	udp_in.sin_port = uh->uh_sport;
488 	udp_in.sin_addr = ip->ip_src;
489 	if ((inp->inp_flags & INP_CONTROLOPTS) ||
490 	    (inp->inp_socket->so_options & SO_TIMESTAMP)) {
491 #ifdef INET6
492 		if (inp->inp_vflag & INP_IPV6) {
493 			int savedflags;
494 
495 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
496 			savedflags = inp->inp_flags;
497 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
498 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
499 			inp->inp_flags = savedflags;
500 		} else
501 #endif
502 		ip_savecontrol(inp, &opts, ip, m);
503 	}
504 	m_adj(m, iphlen + sizeof(struct udphdr));
505 #ifdef INET6
506 	if (inp->inp_vflag & INP_IPV6) {
507 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
508 		append_sa = (struct sockaddr *)&udp_in6;
509 	} else
510 #endif
511 		append_sa = (struct sockaddr *)&udp_in;
512 	if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
513 		udpstat.udps_fullsock++;
514 		goto bad;
515 	}
516 	sorwakeup(inp->inp_socket);
517 	return;
518 bad:
519 	m_freem(m);
520 	if (opts)
521 		m_freem(opts);
522 	return;
523 }
524 
525 #ifdef INET6
526 static void
527 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
528 {
529 	bzero(ip6, sizeof *ip6);
530 
531 	ip6->ip6_vfc = IPV6_VERSION;
532 	ip6->ip6_plen = ip->ip_len;
533 	ip6->ip6_nxt = ip->ip_p;
534 	ip6->ip6_hlim = ip->ip_ttl;
535 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
536 		IPV6_ADDR_INT32_SMP;
537 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
538 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
539 }
540 #endif
541 
542 /*
543  * subroutine of udp_input(), mainly for source code readability.
544  * caller must properly init udp_ip6 and udp_in6 beforehand.
545  */
546 static void
547 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
548 {
549 	struct sockaddr *append_sa;
550 	struct mbuf *opts = NULL;
551 
552 	if (last->inp_flags & INP_CONTROLOPTS ||
553 	    last->inp_socket->so_options & SO_TIMESTAMP) {
554 #ifdef INET6
555 		if (last->inp_vflag & INP_IPV6) {
556 			int savedflags;
557 
558 			if (udp_ip6.uip6_init_done == 0) {
559 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
560 				udp_ip6.uip6_init_done = 1;
561 			}
562 			savedflags = last->inp_flags;
563 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
564 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
565 			last->inp_flags = savedflags;
566 		} else
567 #endif
568 		ip_savecontrol(last, &opts, ip, n);
569 	}
570 #ifdef INET6
571 	if (last->inp_vflag & INP_IPV6) {
572 		if (udp_in6.uin6_init_done == 0) {
573 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
574 			udp_in6.uin6_init_done = 1;
575 		}
576 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
577 	} else
578 #endif
579 		append_sa = (struct sockaddr *)&udp_in;
580 	m_adj(n, off);
581 	if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
582 		m_freem(n);
583 		if (opts)
584 			m_freem(opts);
585 		udpstat.udps_fullsock++;
586 	} else
587 		sorwakeup(last->inp_socket);
588 }
589 
590 /*
591  * Notify a udp user of an asynchronous error;
592  * just wake up so that he can collect error status.
593  */
594 void
595 udp_notify(struct inpcb *inp, int error)
596 {
597 	inp->inp_socket->so_error = error;
598 	sorwakeup(inp->inp_socket);
599 	sowwakeup(inp->inp_socket);
600 }
601 
602 struct netmsg_udp_notify {
603 	struct netmsg	nm_nmsg;
604 	void		(*nm_notify)(struct inpcb *, int);
605 	struct in_addr	nm_faddr;
606 	int		nm_arg;
607 };
608 
609 static void
610 udp_notifyall_oncpu(struct netmsg *netmsg)
611 {
612 	struct netmsg_udp_notify *nmsg = (struct netmsg_udp_notify *)netmsg;
613 	int nextcpu;
614 
615 	in_pcbnotifyall(&udbinfo.pcblisthead, nmsg->nm_faddr, nmsg->nm_arg,
616 			nmsg->nm_notify);
617 
618 	nextcpu = mycpuid + 1;
619 	if (nextcpu < ncpus2)
620 		lwkt_forwardmsg(cpu_portfn(nextcpu), &netmsg->nm_lmsg);
621 	else
622 		lwkt_replymsg(&netmsg->nm_lmsg, 0);
623 }
624 
625 static void
626 udp_rtchange(struct inpcb *inp, int err)
627 {
628 #ifdef SMP
629 	/* XXX Nuke this, once UDP inpcbs are CPU localized */
630 	if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
631 		rtfree(inp->inp_route.ro_rt);
632 		inp->inp_route.ro_rt = NULL;
633 		/*
634 		 * A new route can be allocated the next time
635 		 * output is attempted.
636 		 */
637 	}
638 #else
639 	in_rtchange(inp, err);
640 #endif
641 }
642 
643 void
644 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
645 {
646 	struct ip *ip = vip;
647 	struct udphdr *uh;
648 	void (*notify) (struct inpcb *, int) = udp_notify;
649 	struct in_addr faddr;
650 	struct inpcb *inp;
651 
652 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
653 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
654 		return;
655 
656 	if (PRC_IS_REDIRECT(cmd)) {
657 		ip = NULL;
658 		notify = udp_rtchange;
659 	} else if (cmd == PRC_HOSTDEAD)
660 		ip = NULL;
661 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
662 		return;
663 	if (ip) {
664 		crit_enter();
665 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
666 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
667 					ip->ip_src, uh->uh_sport, 0, NULL);
668 		if (inp != NULL && inp->inp_socket != NULL)
669 			(*notify)(inp, inetctlerrmap[cmd]);
670 		crit_exit();
671 	} else if (PRC_IS_REDIRECT(cmd)) {
672 		struct netmsg_udp_notify nmsg;
673 
674 		KKASSERT(&curthread->td_msgport == cpu_portfn(0));
675 		netmsg_init(&nmsg.nm_nmsg, NULL, &curthread->td_msgport,
676 			    0, udp_notifyall_oncpu);
677 		nmsg.nm_faddr = faddr;
678 		nmsg.nm_arg = inetctlerrmap[cmd];
679 		nmsg.nm_notify = notify;
680 
681 		lwkt_domsg(cpu_portfn(0), &nmsg.nm_nmsg.nm_lmsg, 0);
682 	} else {
683 		/*
684 		 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
685 		 * once UDP inpcbs are CPU localized
686 		 */
687 		KKASSERT(&curthread->td_msgport == cpu_portfn(0));
688 		in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
689 				notify);
690 	}
691 }
692 
693 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
694 	    in_pcblist_global, "S,xinpcb", "List of active UDP sockets");
695 
696 static int
697 udp_getcred(SYSCTL_HANDLER_ARGS)
698 {
699 	struct sockaddr_in addrs[2];
700 	struct inpcb *inp;
701 	int error;
702 
703 	error = priv_check(req->td, PRIV_ROOT);
704 	if (error)
705 		return (error);
706 	error = SYSCTL_IN(req, addrs, sizeof addrs);
707 	if (error)
708 		return (error);
709 	crit_enter();
710 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
711 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
712 	if (inp == NULL || inp->inp_socket == NULL) {
713 		error = ENOENT;
714 		goto out;
715 	}
716 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
717 out:
718 	crit_exit();
719 	return (error);
720 }
721 
722 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
723     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
724 
725 static int
726 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
727 	   struct mbuf *control, struct thread *td)
728 {
729 	struct udpiphdr *ui;
730 	int len = m->m_pkthdr.len;
731 	struct sockaddr_in *sin;	/* really is initialized before use */
732 	int error = 0, lport_any = 0;
733 
734 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
735 		error = EMSGSIZE;
736 		goto release;
737 	}
738 
739 	if (inp->inp_lport == 0) {	/* unbound socket */
740 		error = in_pcbbind(inp, NULL, td);
741 		if (error)
742 			goto release;
743 		in_pcbinswildcardhash(inp);
744 		lport_any = 1;
745 	}
746 
747 	if (dstaddr != NULL) {		/* destination address specified */
748 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
749 			/* already connected */
750 			error = EISCONN;
751 			goto release;
752 		}
753 		sin = (struct sockaddr_in *)dstaddr;
754 		if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
755 			error = EAFNOSUPPORT; /* IPv6 only jail */
756 			goto release;
757 		}
758 	} else {
759 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
760 			/* no destination specified and not already connected */
761 			error = ENOTCONN;
762 			goto release;
763 		}
764 		sin = NULL;
765 	}
766 
767 	/*
768 	 * Calculate data length and get a mbuf
769 	 * for UDP and IP headers.
770 	 */
771 	M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
772 	if (m == NULL) {
773 		error = ENOBUFS;
774 		goto release;
775 	}
776 
777 	/*
778 	 * Fill in mbuf with extended UDP header
779 	 * and addresses and length put into network format.
780 	 */
781 	ui = mtod(m, struct udpiphdr *);
782 	bzero(ui->ui_x1, sizeof ui->ui_x1);	/* XXX still needed? */
783 	ui->ui_pr = IPPROTO_UDP;
784 
785 	/*
786 	 * Set destination address.
787 	 */
788 	if (dstaddr != NULL) {			/* use specified destination */
789 		ui->ui_dst = sin->sin_addr;
790 		ui->ui_dport = sin->sin_port;
791 	} else {				/* use connected destination */
792 		ui->ui_dst = inp->inp_faddr;
793 		ui->ui_dport = inp->inp_fport;
794 	}
795 
796 	/*
797 	 * Set source address.
798 	 */
799 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
800 		struct sockaddr_in *if_sin;
801 
802 		if (dstaddr == NULL) {
803 			/*
804 			 * connect() had (or should have) failed because
805 			 * the interface had no IP address, but the
806 			 * application proceeded to call send() anyways.
807 			 */
808 			error = ENOTCONN;
809 			goto release;
810 		}
811 
812 		/* Look up outgoing interface. */
813 		if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
814 			goto release;
815 		ui->ui_src = if_sin->sin_addr;	/* use address of interface */
816 	} else {
817 		ui->ui_src = inp->inp_laddr;	/* use non-null bound address */
818 	}
819 	ui->ui_sport = inp->inp_lport;
820 	KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
821 
822 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
823 
824 	/*
825 	 * Set up checksum and output datagram.
826 	 */
827 	if (udpcksum) {
828 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
829 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
830 		m->m_pkthdr.csum_flags = CSUM_UDP;
831 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
832 	} else {
833 		ui->ui_sum = 0;
834 	}
835 	((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
836 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
837 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
838 	udpstat.udps_opackets++;
839 
840 	error = ip_output(m, inp->inp_options, &inp->inp_route,
841 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
842 	    IP_DEBUGROUTE,
843 	    inp->inp_moptions, inp);
844 
845 	/*
846 	 * If this is the first data gram sent on an unbound and unconnected
847 	 * UDP socket, lport will be changed in this function.  If target
848 	 * CPU after this lport changing is no longer the current CPU, then
849 	 * free the route entry allocated on the current CPU.
850 	 */
851 	if (lport_any) {
852 		if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
853 		    inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
854 			struct route *ro = &inp->inp_route;
855 
856 			if (ro->ro_rt != NULL)
857 				RTFREE(ro->ro_rt);
858 			bzero(ro, sizeof(*ro));
859 		}
860 	}
861 	return (error);
862 
863 release:
864 	m_freem(m);
865 	return (error);
866 }
867 
868 u_long	udp_sendspace = 9216;		/* really max datagram size */
869 					/* 40 1K datagrams */
870 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
871     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
872 
873 u_long	udp_recvspace = 40 * (1024 +
874 #ifdef INET6
875 				      sizeof(struct sockaddr_in6)
876 #else
877 				      sizeof(struct sockaddr_in)
878 #endif
879 				      );
880 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
881     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
882 
883 static int
884 udp_abort(struct socket *so)
885 {
886 	struct inpcb *inp;
887 
888 	inp = so->so_pcb;
889 	if (inp == NULL)
890 		return EINVAL;	/* ??? possible? panic instead? */
891 	soisdisconnected(so);
892 	crit_enter();
893 	in_pcbdetach(inp);
894 	crit_exit();
895 	return 0;
896 }
897 
898 static int
899 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai)
900 {
901 	struct inpcb *inp;
902 	int error;
903 
904 	inp = so->so_pcb;
905 	if (inp != NULL)
906 		return EINVAL;
907 
908 	error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
909 	if (error)
910 		return error;
911 	crit_enter();
912 	error = in_pcballoc(so, &udbinfo);
913 	crit_exit();
914 	if (error)
915 		return error;
916 	so->so_port = udp_soport_attach(so);
917 
918 	inp = (struct inpcb *)so->so_pcb;
919 	inp->inp_vflag |= INP_IPV4;
920 	inp->inp_ip_ttl = ip_defttl;
921 	return 0;
922 }
923 
924 static int
925 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
926 {
927 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
928 	struct inpcb *inp;
929 	int error;
930 
931 	inp = so->so_pcb;
932 	if (inp == NULL)
933 		return EINVAL;
934 	crit_enter();
935 	error = in_pcbbind(inp, nam, td);
936 	crit_exit();
937 	if (error == 0) {
938 		if (sin->sin_addr.s_addr != INADDR_ANY)
939 			inp->inp_flags |= INP_WASBOUND_NOTANY;
940 		in_pcbinswildcardhash(inp);
941 	}
942 	return error;
943 }
944 
945 #ifdef SMP
946 
947 struct netmsg_udp_connect {
948 	struct netmsg		nm_netmsg;
949 	struct socket		*nm_so;
950 	struct sockaddr_in	*nm_sin;
951 	struct sockaddr_in	*nm_ifsin;
952 	struct thread		*nm_td;
953 };
954 
955 static void
956 udp_connect_handler(netmsg_t netmsg)
957 {
958 	struct netmsg_udp_connect *msg = (void *)netmsg;
959 	int error;
960 
961 	error = udp_connect_oncpu(msg->nm_so, msg->nm_td,
962 				  msg->nm_sin, msg->nm_ifsin);
963 	lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error);
964 }
965 
966 #endif
967 
968 static int
969 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
970 {
971 	struct inpcb *inp;
972 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
973 	struct sockaddr_in *if_sin;
974 	lwkt_port_t port;
975 	int error;
976 
977 	inp = so->so_pcb;
978 	if (inp == NULL)
979 		return EINVAL;
980 	if (inp->inp_faddr.s_addr != INADDR_ANY)
981 		return EISCONN;
982 	error = 0;
983 
984 	/*
985 	 * Bind if we have to
986 	 */
987 	if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
988 	    inp->inp_laddr.s_addr == INADDR_ANY) {
989 		error = in_pcbbind(inp, NULL, td);
990 		if (error)
991 			return (error);
992 	}
993 
994 	/*
995 	 * Calculate the correct protocol processing thread.  The connect
996 	 * operation must run there.
997 	 */
998 	error = in_pcbladdr(inp, nam, &if_sin, td);
999 	if (error)
1000 		return(error);
1001 	if (!prison_remote_ip(td, nam))
1002 		return(EAFNOSUPPORT); /* IPv6 only jail */
1003 
1004 	port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1005 			    inp->inp_laddr.s_addr, inp->inp_lport);
1006 #ifdef SMP
1007 	if (port != &curthread->td_msgport) {
1008 		struct netmsg_udp_connect msg;
1009 		struct route *ro = &inp->inp_route;
1010 
1011 		/*
1012 		 * in_pcbladdr() may have allocated a route entry for us
1013 		 * on the current CPU, but we need a route entry on the
1014 		 * inpcb's owner CPU, so free it here.
1015 		 */
1016 		if (ro->ro_rt != NULL)
1017 			RTFREE(ro->ro_rt);
1018 		bzero(ro, sizeof(*ro));
1019 
1020 		/*
1021 		 * NOTE: We haven't set so->so_port yet do not pass so
1022 		 *       to netmsg_init() or it will be improperly forwarded.
1023 		 */
1024 		netmsg_init(&msg.nm_netmsg, NULL, &curthread->td_msgport,
1025 			    0, udp_connect_handler);
1026 		msg.nm_so = so;
1027 		msg.nm_sin = sin;
1028 		msg.nm_ifsin = if_sin;
1029 		msg.nm_td = td;
1030 		error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0);
1031 	} else {
1032 		error = udp_connect_oncpu(so, td, sin, if_sin);
1033 	}
1034 #else
1035 	error = udp_connect_oncpu(so, td, sin, if_sin);
1036 #endif
1037 	return (error);
1038 }
1039 
1040 static int
1041 udp_connect_oncpu(struct socket *so, struct thread *td,
1042 		  struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1043 {
1044 	struct inpcb *inp;
1045 	int error;
1046 
1047 	inp = so->so_pcb;
1048 	if (inp->inp_flags & INP_WILDCARD)
1049 		in_pcbremwildcardhash(inp);
1050 	error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1051 
1052 	if (error == 0) {
1053 		/*
1054 		 * No more errors can occur, finish adjusting the socket
1055 		 * and change the processing port to reflect the connected
1056 		 * socket.  Once set we can no longer safely mess with the
1057 		 * socket.
1058 		 */
1059 		soisconnected(so);
1060 		sosetport(so, &curthread->td_msgport);
1061 	} else if (error == EAFNOSUPPORT) {	/* connection dissolved */
1062 		/*
1063 		 * Follow traditional BSD behavior and retain
1064 		 * the local port binding.  But, fix the old misbehavior
1065 		 * of overwriting any previously bound local address.
1066 		 */
1067 		if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1068 			inp->inp_laddr.s_addr = INADDR_ANY;
1069 		in_pcbinswildcardhash(inp);
1070 	}
1071 	return error;
1072 }
1073 
1074 static int
1075 udp_detach(struct socket *so)
1076 {
1077 	struct inpcb *inp;
1078 
1079 	inp = so->so_pcb;
1080 	if (inp == NULL)
1081 		return EINVAL;
1082 	crit_enter();
1083 	in_pcbdetach(inp);
1084 	crit_exit();
1085 	return 0;
1086 }
1087 
1088 static int
1089 udp_disconnect(struct socket *so)
1090 {
1091 	struct route *ro;
1092 	struct inpcb *inp;
1093 
1094 	inp = so->so_pcb;
1095 	if (inp == NULL)
1096 		return EINVAL;
1097 	if (inp->inp_faddr.s_addr == INADDR_ANY)
1098 		return ENOTCONN;
1099 
1100 	crit_enter();
1101 	in_pcbdisconnect(inp);
1102 	crit_exit();
1103 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1104 
1105 	ro = &inp->inp_route;
1106 	if (ro->ro_rt != NULL)
1107 		RTFREE(ro->ro_rt);
1108 	bzero(ro, sizeof(*ro));
1109 
1110 	return 0;
1111 }
1112 
1113 static int
1114 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1115 	    struct mbuf *control, struct thread *td)
1116 {
1117 	struct inpcb *inp;
1118 
1119 	inp = so->so_pcb;
1120 	if (inp == NULL) {
1121 		m_freem(m);
1122 		return EINVAL;
1123 	}
1124 	return udp_output(inp, m, addr, control, td);
1125 }
1126 
1127 int
1128 udp_shutdown(struct socket *so)
1129 {
1130 	struct inpcb *inp;
1131 
1132 	inp = so->so_pcb;
1133 	if (inp == NULL)
1134 		return EINVAL;
1135 	socantsendmore(so);
1136 	return 0;
1137 }
1138 
1139 struct pr_usrreqs udp_usrreqs = {
1140 	.pru_abort = udp_abort,
1141 	.pru_accept = pru_accept_notsupp,
1142 	.pru_attach = udp_attach,
1143 	.pru_bind = udp_bind,
1144 	.pru_connect = udp_connect,
1145 	.pru_connect2 = pru_connect2_notsupp,
1146 	.pru_control = in_control,
1147 	.pru_detach = udp_detach,
1148 	.pru_disconnect = udp_disconnect,
1149 	.pru_listen = pru_listen_notsupp,
1150 	.pru_peeraddr = in_setpeeraddr,
1151 	.pru_rcvd = pru_rcvd_notsupp,
1152 	.pru_rcvoob = pru_rcvoob_notsupp,
1153 	.pru_send = udp_send,
1154 	.pru_sense = pru_sense_null,
1155 	.pru_shutdown = udp_shutdown,
1156 	.pru_sockaddr = in_setsockaddr,
1157 	.pru_sosend = sosendudp,
1158 	.pru_soreceive = soreceive
1159 };
1160 
1161