xref: /freebsd/sys/netinet/udp_usrreq.c (revision 6ccbb635d7b228a34d0eb8bb16b767a233c21166)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_ipsec.h"
34 #include "opt_inet6.h"
35 #include "opt_mac.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/domain.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/mac.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sx.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 
55 #include <vm/uma.h>
56 
57 #include <net/if.h>
58 #include <net/route.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_pcb.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #ifdef INET6
66 #include <netinet/ip6.h>
67 #endif
68 #include <netinet/ip_icmp.h>
69 #include <netinet/icmp_var.h>
70 #include <netinet/ip_var.h>
71 #ifdef INET6
72 #include <netinet6/ip6_var.h>
73 #endif
74 #include <netinet/udp.h>
75 #include <netinet/udp_var.h>
76 
77 #ifdef FAST_IPSEC
78 #include <netipsec/ipsec.h>
79 #endif /*FAST_IPSEC*/
80 
81 #ifdef IPSEC
82 #include <netinet6/ipsec.h>
83 #endif /*IPSEC*/
84 
85 #include <machine/in_cksum.h>
86 
87 /*
88  * UDP protocol implementation.
89  * Per RFC 768, August, 1980.
90  */
91 #ifndef	COMPAT_42
92 static int	udpcksum = 1;
93 #else
94 static int	udpcksum = 0;		/* XXX */
95 #endif
96 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
97 		&udpcksum, 0, "");
98 
99 int	log_in_vain = 0;
100 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
101     &log_in_vain, 0, "Log all incoming UDP packets");
102 
103 static int	blackhole = 0;
104 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
105 	&blackhole, 0, "Do not send port unreachables for refused connects");
106 
107 static int	strict_mcast_mship = 0;
108 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
109 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
110 
111 struct	inpcbhead udb;		/* from udp_var.h */
112 #define	udb6	udb  /* for KAME src sync over BSD*'s */
113 struct	inpcbinfo udbinfo;
114 
115 #ifndef UDBHASHSIZE
116 #define UDBHASHSIZE 16
117 #endif
118 
119 struct	udpstat udpstat;	/* from udp_var.h */
120 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
121     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
122 
123 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
124 #ifdef INET6
125 struct udp_in6 {
126 	struct sockaddr_in6	uin6_sin;
127 	u_char			uin6_init_done : 1;
128 } udp_in6 = {
129 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
130 	0
131 };
132 struct udp_ip6 {
133 	struct ip6_hdr		uip6_ip6;
134 	u_char			uip6_init_done : 1;
135 } udp_ip6;
136 #endif /* INET6 */
137 
138 static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
139 		int off);
140 #ifdef INET6
141 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
142 #endif
143 
144 static int udp_detach(struct socket *so);
145 static	int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
146 		struct mbuf *, struct thread *);
147 
148 void
149 udp_init()
150 {
151 	INP_INFO_LOCK_INIT(&udbinfo, "udp");
152 	LIST_INIT(&udb);
153 	udbinfo.listhead = &udb;
154 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
155 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
156 					&udbinfo.porthashmask);
157 	udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL,
158 	    NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
159 	uma_zone_set_max(udbinfo.ipi_zone, maxsockets);
160 }
161 
162 void
163 udp_input(m, off)
164 	register struct mbuf *m;
165 	int off;
166 {
167 	int iphlen = off;
168 	register struct ip *ip;
169 	register struct udphdr *uh;
170 	register struct inpcb *inp;
171 	struct mbuf *opts = 0;
172 	int len;
173 	struct ip save_ip;
174 
175 	udpstat.udps_ipackets++;
176 
177 	/*
178 	 * Strip IP options, if any; should skip this,
179 	 * make available to user, and use on returned packets,
180 	 * but we don't yet have a way to check the checksum
181 	 * with options still present.
182 	 */
183 	if (iphlen > sizeof (struct ip)) {
184 		ip_stripoptions(m, (struct mbuf *)0);
185 		iphlen = sizeof(struct ip);
186 	}
187 
188 	/*
189 	 * Get IP and UDP header together in first mbuf.
190 	 */
191 	ip = mtod(m, struct ip *);
192 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
193 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
194 			udpstat.udps_hdrops++;
195 			return;
196 		}
197 		ip = mtod(m, struct ip *);
198 	}
199 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
200 
201 	/* destination port of 0 is illegal, based on RFC768. */
202 	if (uh->uh_dport == 0)
203 		goto badunlocked;
204 
205 	/*
206 	 * Construct sockaddr format source address.
207 	 * Stuff source address and datagram in user buffer.
208 	 */
209 	udp_in.sin_port = uh->uh_sport;
210 	udp_in.sin_addr = ip->ip_src;
211 #ifdef INET6
212 	udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
213 #endif
214 
215 	/*
216 	 * Make mbuf data length reflect UDP length.
217 	 * If not enough data to reflect UDP length, drop.
218 	 */
219 	len = ntohs((u_short)uh->uh_ulen);
220 	if (ip->ip_len != len) {
221 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
222 			udpstat.udps_badlen++;
223 			goto badunlocked;
224 		}
225 		m_adj(m, len - ip->ip_len);
226 		/* ip->ip_len = len; */
227 	}
228 	/*
229 	 * Save a copy of the IP header in case we want restore it
230 	 * for sending an ICMP error message in response.
231 	 */
232 	if (!blackhole)
233 		save_ip = *ip;
234 
235 	/*
236 	 * Checksum extended UDP header and data.
237 	 */
238 	if (uh->uh_sum) {
239 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
240 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
241 				uh->uh_sum = m->m_pkthdr.csum_data;
242 			else
243 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
244 				    ip->ip_dst.s_addr, htonl((u_short)len +
245 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
246 			uh->uh_sum ^= 0xffff;
247 		} else {
248 			char b[9];
249 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
250 			bzero(((struct ipovly *)ip)->ih_x1, 9);
251 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
252 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
253 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
254 		}
255 		if (uh->uh_sum) {
256 			udpstat.udps_badsum++;
257 			m_freem(m);
258 			return;
259 		}
260 	} else
261 		udpstat.udps_nosum++;
262 
263 	INP_INFO_RLOCK(&udbinfo);
264 
265 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
266 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
267 		struct inpcb *last;
268 		/*
269 		 * Deliver a multicast or broadcast datagram to *all* sockets
270 		 * for which the local and remote addresses and ports match
271 		 * those of the incoming datagram.  This allows more than
272 		 * one process to receive multi/broadcasts on the same port.
273 		 * (This really ought to be done for unicast datagrams as
274 		 * well, but that would cause problems with existing
275 		 * applications that open both address-specific sockets and
276 		 * a wildcard socket listening to the same port -- they would
277 		 * end up receiving duplicates of every unicast datagram.
278 		 * Those applications open the multiple sockets to overcome an
279 		 * inadequacy of the UDP socket interface, but for backwards
280 		 * compatibility we avoid the problem here rather than
281 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
282 		 */
283 
284 		/*
285 		 * Locate pcb(s) for datagram.
286 		 * (Algorithm copied from raw_intr().)
287 		 */
288 		last = NULL;
289 		LIST_FOREACH(inp, &udb, inp_list) {
290 			INP_LOCK(inp);
291 			if (inp->inp_lport != uh->uh_dport) {
292 		docontinue:
293 				INP_UNLOCK(inp);
294 				continue;
295 			}
296 #ifdef INET6
297 			if ((inp->inp_vflag & INP_IPV4) == 0)
298 				goto docontinue;
299 #endif
300 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
301 				if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
302 					goto docontinue;
303 			}
304 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
305 				if (inp->inp_faddr.s_addr !=
306 				    ip->ip_src.s_addr ||
307 				    inp->inp_fport != uh->uh_sport)
308 					goto docontinue;
309 			}
310 
311 			/*
312 			 * Check multicast packets to make sure they are only
313 			 * sent to sockets with multicast memberships for the
314 			 * packet's destination address and arrival interface
315 			 */
316 #define MSHIP(_inp, n) ((_inp)->inp_moptions->imo_membership[(n)])
317 #define NMSHIPS(_inp) ((_inp)->inp_moptions->imo_num_memberships)
318 			if (strict_mcast_mship && inp->inp_moptions != NULL) {
319 				int mship, foundmship = 0;
320 
321 				for (mship = 0; mship < NMSHIPS(inp); mship++) {
322 					if (MSHIP(inp, mship)->inm_addr.s_addr
323 					    == ip->ip_dst.s_addr &&
324 					    MSHIP(inp, mship)->inm_ifp
325 					    == m->m_pkthdr.rcvif) {
326 						foundmship = 1;
327 						break;
328 					}
329 				}
330 				if (foundmship == 0)
331 					goto docontinue;
332 			}
333 #undef NMSHIPS
334 #undef MSHIP
335 			if (last != NULL) {
336 				struct mbuf *n;
337 
338 				n = m_copy(m, 0, M_COPYALL);
339 				if (n != NULL)
340 					udp_append(last, ip, n,
341 						   iphlen +
342 						   sizeof(struct udphdr));
343 				INP_UNLOCK(last);
344 			}
345 			last = inp;
346 			/*
347 			 * Don't look for additional matches if this one does
348 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
349 			 * socket options set.  This heuristic avoids searching
350 			 * through all pcbs in the common case of a non-shared
351 			 * port.  It * assumes that an application will never
352 			 * clear these options after setting them.
353 			 */
354 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
355 				break;
356 		}
357 
358 		if (last == NULL) {
359 			/*
360 			 * No matching pcb found; discard datagram.
361 			 * (No need to send an ICMP Port Unreachable
362 			 * for a broadcast or multicast datgram.)
363 			 */
364 			udpstat.udps_noportbcast++;
365 			goto badheadlocked;
366 		}
367 		INP_INFO_RUNLOCK(&udbinfo);
368 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
369 		INP_UNLOCK(last);
370 		return;
371 	}
372 	/*
373 	 * Locate pcb for datagram.
374 	 */
375 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
376 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
377 	if (inp == NULL) {
378 		if (log_in_vain) {
379 			char buf[4*sizeof "123"];
380 
381 			strcpy(buf, inet_ntoa(ip->ip_dst));
382 			log(LOG_INFO,
383 			    "Connection attempt to UDP %s:%d from %s:%d\n",
384 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
385 			    ntohs(uh->uh_sport));
386 		}
387 		udpstat.udps_noport++;
388 		if (m->m_flags & (M_BCAST | M_MCAST)) {
389 			udpstat.udps_noportbcast++;
390 			goto badheadlocked;
391 		}
392 		if (blackhole)
393 			goto badheadlocked;
394 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
395 			goto badheadlocked;
396 		*ip = save_ip;
397 		ip->ip_len += iphlen;
398 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
399 		INP_INFO_RUNLOCK(&udbinfo);
400 		return;
401 	}
402 	INP_LOCK(inp);
403 	INP_INFO_RUNLOCK(&udbinfo);
404 	udp_append(inp, ip, m, iphlen + sizeof(struct udphdr));
405 	INP_UNLOCK(inp);
406 	return;
407 
408 badheadlocked:
409 	INP_INFO_RUNLOCK(&udbinfo);
410 	if (inp)
411 		INP_UNLOCK(inp);
412 badunlocked:
413 	m_freem(m);
414 	if (opts)
415 		m_freem(opts);
416 	return;
417 }
418 
419 #ifdef INET6
420 static void
421 ip_2_ip6_hdr(ip6, ip)
422 	struct ip6_hdr *ip6;
423 	struct ip *ip;
424 {
425 	bzero(ip6, sizeof(*ip6));
426 
427 	ip6->ip6_vfc = IPV6_VERSION;
428 	ip6->ip6_plen = ip->ip_len;
429 	ip6->ip6_nxt = ip->ip_p;
430 	ip6->ip6_hlim = ip->ip_ttl;
431 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
432 		IPV6_ADDR_INT32_SMP;
433 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
434 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
435 }
436 #endif
437 
438 /*
439  * subroutine of udp_input(), mainly for source code readability.
440  * caller must properly init udp_ip6 and udp_in6 beforehand.
441  */
442 static void
443 udp_append(last, ip, n, off)
444 	struct inpcb *last;
445 	struct ip *ip;
446 	struct mbuf *n;
447 	int off;
448 {
449 	struct sockaddr *append_sa;
450 	struct mbuf *opts = 0;
451 
452 #if defined(IPSEC) || defined(FAST_IPSEC)
453 	/* check AH/ESP integrity. */
454 	if (ipsec4_in_reject(n, last)) {
455 #ifdef IPSEC
456 		ipsecstat.in_polvio++;
457 #endif /*IPSEC*/
458 		m_freem(n);
459 		return;
460 	}
461 #endif /*IPSEC || FAST_IPSEC*/
462 #ifdef MAC
463 	if (mac_check_inpcb_deliver(last, n) != 0) {
464 		m_freem(n);
465 		return;
466 	}
467 #endif
468 	if (last->inp_flags & INP_CONTROLOPTS ||
469 	    last->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) {
470 #ifdef INET6
471 		if (last->inp_vflag & INP_IPV6) {
472 			int savedflags;
473 
474 			if (udp_ip6.uip6_init_done == 0) {
475 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
476 				udp_ip6.uip6_init_done = 1;
477 			}
478 			savedflags = last->inp_flags;
479 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
480 			ip6_savecontrol(last, n, &opts);
481 			last->inp_flags = savedflags;
482 		} else
483 #endif
484 		ip_savecontrol(last, &opts, ip, n);
485 	}
486 #ifdef INET6
487 	if (last->inp_vflag & INP_IPV6) {
488 		if (udp_in6.uin6_init_done == 0) {
489 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
490 			udp_in6.uin6_init_done = 1;
491 		}
492 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
493 	} else
494 #endif
495 	append_sa = (struct sockaddr *)&udp_in;
496 	m_adj(n, off);
497 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
498 		m_freem(n);
499 		if (opts)
500 			m_freem(opts);
501 		udpstat.udps_fullsock++;
502 	} else
503 		sorwakeup(last->inp_socket);
504 }
505 
506 /*
507  * Notify a udp user of an asynchronous error;
508  * just wake up so that he can collect error status.
509  */
510 struct inpcb *
511 udp_notify(inp, errno)
512 	register struct inpcb *inp;
513 	int errno;
514 {
515 	inp->inp_socket->so_error = errno;
516 	sorwakeup(inp->inp_socket);
517 	sowwakeup(inp->inp_socket);
518 	return inp;
519 }
520 
521 void
522 udp_ctlinput(cmd, sa, vip)
523 	int cmd;
524 	struct sockaddr *sa;
525 	void *vip;
526 {
527 	struct ip *ip = vip;
528 	struct udphdr *uh;
529 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
530         struct in_addr faddr;
531 	struct inpcb *inp;
532 	int s;
533 
534 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
535 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
536         	return;
537 
538 	/*
539 	 * Redirects don't need to be handled up here.
540 	 */
541 	if (PRC_IS_REDIRECT(cmd))
542 		return;
543 	/*
544 	 * Hostdead is ugly because it goes linearly through all PCBs.
545 	 * XXX: We never get this from ICMP, otherwise it makes an
546 	 * excellent DoS attack on machines with many connections.
547 	 */
548 	if (cmd == PRC_HOSTDEAD)
549 		ip = 0;
550 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
551 		return;
552 	if (ip) {
553 		s = splnet();
554 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
555 		INP_INFO_RLOCK(&udbinfo);
556 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
557                     ip->ip_src, uh->uh_sport, 0, NULL);
558 		if (inp != NULL) {
559 			INP_LOCK(inp);
560 			if (inp->inp_socket != NULL) {
561 				(*notify)(inp, inetctlerrmap[cmd]);
562 			}
563 			INP_UNLOCK(inp);
564 		}
565 		INP_INFO_RUNLOCK(&udbinfo);
566 		splx(s);
567 	} else
568 		in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
569 }
570 
571 static int
572 udp_pcblist(SYSCTL_HANDLER_ARGS)
573 {
574 	int error, i, n, s;
575 	struct inpcb *inp, **inp_list;
576 	inp_gen_t gencnt;
577 	struct xinpgen xig;
578 
579 	/*
580 	 * The process of preparing the TCB list is too time-consuming and
581 	 * resource-intensive to repeat twice on every request.
582 	 */
583 	if (req->oldptr == 0) {
584 		n = udbinfo.ipi_count;
585 		req->oldidx = 2 * (sizeof xig)
586 			+ (n + n/8) * sizeof(struct xinpcb);
587 		return 0;
588 	}
589 
590 	if (req->newptr != 0)
591 		return EPERM;
592 
593 	/*
594 	 * OK, now we're committed to doing something.
595 	 */
596 	s = splnet();
597 	INP_INFO_RLOCK(&udbinfo);
598 	gencnt = udbinfo.ipi_gencnt;
599 	n = udbinfo.ipi_count;
600 	INP_INFO_RUNLOCK(&udbinfo);
601 	splx(s);
602 
603 	error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
604 		+ n * sizeof(struct xinpcb));
605 	if (error != 0)
606 		return (error);
607 
608 	xig.xig_len = sizeof xig;
609 	xig.xig_count = n;
610 	xig.xig_gen = gencnt;
611 	xig.xig_sogen = so_gencnt;
612 	error = SYSCTL_OUT(req, &xig, sizeof xig);
613 	if (error)
614 		return error;
615 
616 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
617 	if (inp_list == 0)
618 		return ENOMEM;
619 
620 	s = splnet();
621 	INP_INFO_RLOCK(&udbinfo);
622 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
623 	     inp = LIST_NEXT(inp, inp_list)) {
624 		INP_LOCK(inp);
625 		if (inp->inp_gencnt <= gencnt &&
626 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
627 			inp_list[i++] = inp;
628 		INP_UNLOCK(inp);
629 	}
630 	INP_INFO_RUNLOCK(&udbinfo);
631 	splx(s);
632 	n = i;
633 
634 	error = 0;
635 	for (i = 0; i < n; i++) {
636 		inp = inp_list[i];
637 		if (inp->inp_gencnt <= gencnt) {
638 			struct xinpcb xi;
639 			xi.xi_len = sizeof xi;
640 			/* XXX should avoid extra copy */
641 			bcopy(inp, &xi.xi_inp, sizeof *inp);
642 			if (inp->inp_socket)
643 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
644 			xi.xi_inp.inp_gencnt = inp->inp_gencnt;
645 			error = SYSCTL_OUT(req, &xi, sizeof xi);
646 		}
647 	}
648 	if (!error) {
649 		/*
650 		 * Give the user an updated idea of our state.
651 		 * If the generation differs from what we told
652 		 * her before, she knows that something happened
653 		 * while we were processing this request, and it
654 		 * might be necessary to retry.
655 		 */
656 		s = splnet();
657 		INP_INFO_RLOCK(&udbinfo);
658 		xig.xig_gen = udbinfo.ipi_gencnt;
659 		xig.xig_sogen = so_gencnt;
660 		xig.xig_count = udbinfo.ipi_count;
661 		INP_INFO_RUNLOCK(&udbinfo);
662 		splx(s);
663 		error = SYSCTL_OUT(req, &xig, sizeof xig);
664 	}
665 	free(inp_list, M_TEMP);
666 	return error;
667 }
668 
669 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
670 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
671 
672 static int
673 udp_getcred(SYSCTL_HANDLER_ARGS)
674 {
675 	struct xucred xuc;
676 	struct sockaddr_in addrs[2];
677 	struct inpcb *inp;
678 	int error, s;
679 
680 	error = suser_cred(req->td->td_ucred, PRISON_ROOT);
681 	if (error)
682 		return (error);
683 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
684 	if (error)
685 		return (error);
686 	s = splnet();
687 	INP_INFO_RLOCK(&udbinfo);
688 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
689 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
690 	if (inp == NULL || inp->inp_socket == NULL) {
691 		error = ENOENT;
692 		goto out;
693 	}
694 	error = cr_canseesocket(req->td->td_ucred, inp->inp_socket);
695 	if (error)
696 		goto out;
697 	cru2x(inp->inp_socket->so_cred, &xuc);
698 out:
699 	INP_INFO_RUNLOCK(&udbinfo);
700 	splx(s);
701 	if (error == 0)
702 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
703 	return (error);
704 }
705 
706 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
707     CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
708     udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
709 
710 static int
711 udp_output(inp, m, addr, control, td)
712 	register struct inpcb *inp;
713 	struct mbuf *m;
714 	struct sockaddr *addr;
715 	struct mbuf *control;
716 	struct thread *td;
717 {
718 	register struct udpiphdr *ui;
719 	register int len = m->m_pkthdr.len;
720 	struct in_addr faddr, laddr;
721 	struct cmsghdr *cm;
722 	struct sockaddr_in *sin, src;
723 	int error = 0;
724 	int ipflags;
725 	u_short fport, lport;
726 
727 	INP_LOCK_ASSERT(inp);
728 #ifdef MAC
729 	mac_create_mbuf_from_socket(inp->inp_socket, m);
730 #endif
731 
732 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
733 		error = EMSGSIZE;
734 		if (control)
735 			m_freem(control);
736 		goto release;
737 	}
738 
739 	src.sin_addr.s_addr = INADDR_ANY;
740 	if (control != NULL) {
741 		/*
742 		 * XXX: Currently, we assume all the optional information
743 		 * is stored in a single mbuf.
744 		 */
745 		if (control->m_next) {
746 			error = EINVAL;
747 			m_freem(control);
748 			goto release;
749 		}
750 		for (; control->m_len > 0;
751 		    control->m_data += CMSG_ALIGN(cm->cmsg_len),
752 		    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
753 			cm = mtod(control, struct cmsghdr *);
754 			if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 ||
755 			    cm->cmsg_len > control->m_len) {
756 				error = EINVAL;
757 				break;
758 			}
759 			if (cm->cmsg_level != IPPROTO_IP)
760 				continue;
761 
762 			switch (cm->cmsg_type) {
763 			case IP_SENDSRCADDR:
764 				if (cm->cmsg_len !=
765 				    CMSG_LEN(sizeof(struct in_addr))) {
766 					error = EINVAL;
767 					break;
768 				}
769 				bzero(&src, sizeof(src));
770 				src.sin_family = AF_INET;
771 				src.sin_len = sizeof(src);
772 				src.sin_port = inp->inp_lport;
773 				src.sin_addr = *(struct in_addr *)CMSG_DATA(cm);
774 				break;
775 			default:
776 				error = ENOPROTOOPT;
777 				break;
778 			}
779 			if (error)
780 				break;
781 		}
782 		m_freem(control);
783 	}
784 	if (error)
785 		goto release;
786 	laddr = inp->inp_laddr;
787 	lport = inp->inp_lport;
788 	if (src.sin_addr.s_addr != INADDR_ANY) {
789 		if (lport == 0) {
790 			error = EINVAL;
791 			goto release;
792 		}
793 		error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
794 		    &laddr.s_addr, &lport, td->td_ucred);
795 		if (error)
796 			goto release;
797 	}
798 
799 	if (addr) {
800 		sin = (struct sockaddr_in *)addr;
801 		if (td && jailed(td->td_ucred))
802 			prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
803 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
804 			error = EISCONN;
805 			goto release;
806 		}
807 		error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport,
808 		    &faddr.s_addr, &fport, NULL, td->td_ucred);
809 		if (error)
810 			goto release;
811 
812 		/* Commit the local port if newly assigned. */
813 		if (inp->inp_laddr.s_addr == INADDR_ANY &&
814 		    inp->inp_lport == 0) {
815 			inp->inp_lport = lport;
816 			if (in_pcbinshash(inp) != 0) {
817 				inp->inp_lport = 0;
818 				error = EAGAIN;
819 				goto release;
820 			}
821 			inp->inp_flags |= INP_ANONPORT;
822 		}
823 	} else {
824 		faddr = inp->inp_faddr;
825 		fport = inp->inp_fport;
826 		if (faddr.s_addr == INADDR_ANY) {
827 			error = ENOTCONN;
828 			goto release;
829 		}
830 	}
831 	/*
832 	 * Calculate data length and get a mbuf
833 	 * for UDP and IP headers.
834 	 */
835 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
836 	if (m == 0) {
837 		error = ENOBUFS;
838 		goto release;
839 	}
840 
841 	/*
842 	 * Fill in mbuf with extended UDP header
843 	 * and addresses and length put into network format.
844 	 */
845 	ui = mtod(m, struct udpiphdr *);
846 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
847 	ui->ui_pr = IPPROTO_UDP;
848 	ui->ui_src = laddr;
849 	ui->ui_dst = faddr;
850 	ui->ui_sport = lport;
851 	ui->ui_dport = fport;
852 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
853 
854 	ipflags = inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST);
855 	if (inp->inp_flags & INP_ONESBCAST)
856 		ipflags |= IP_SENDONES;
857 
858 	/*
859 	 * Set up checksum and output datagram.
860 	 */
861 	if (udpcksum) {
862 		if (inp->inp_flags & INP_ONESBCAST)
863 			faddr.s_addr = INADDR_BROADCAST;
864 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
865 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
866 		m->m_pkthdr.csum_flags = CSUM_UDP;
867 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
868 	} else {
869 		ui->ui_sum = 0;
870 	}
871 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
872 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
873 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
874 	udpstat.udps_opackets++;
875 
876 	error = ip_output(m, inp->inp_options, NULL, ipflags,
877 	    inp->inp_moptions, inp);
878 	return (error);
879 
880 release:
881 	m_freem(m);
882 	return (error);
883 }
884 
885 u_long	udp_sendspace = 9216;		/* really max datagram size */
886 					/* 40 1K datagrams */
887 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
888     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
889 
890 u_long	udp_recvspace = 40 * (1024 +
891 #ifdef INET6
892 				      sizeof(struct sockaddr_in6)
893 #else
894 				      sizeof(struct sockaddr_in)
895 #endif
896 				      );
897 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
898     &udp_recvspace, 0, "Maximum space for incoming UDP datagrams");
899 
900 static int
901 udp_abort(struct socket *so)
902 {
903 	struct inpcb *inp;
904 	int s;
905 
906 	INP_INFO_WLOCK(&udbinfo);
907 	inp = sotoinpcb(so);
908 	if (inp == 0) {
909 		INP_INFO_WUNLOCK(&udbinfo);
910 		return EINVAL;	/* ??? possible? panic instead? */
911 	}
912 	INP_LOCK(inp);
913 	soisdisconnected(so);
914 	s = splnet();
915 	in_pcbdetach(inp);
916 	INP_INFO_WUNLOCK(&udbinfo);
917 	splx(s);
918 	return 0;
919 }
920 
921 static int
922 udp_attach(struct socket *so, int proto, struct thread *td)
923 {
924 	struct inpcb *inp;
925 	int s, error;
926 
927 	INP_INFO_WLOCK(&udbinfo);
928 	inp = sotoinpcb(so);
929 	if (inp != 0) {
930 		INP_INFO_WUNLOCK(&udbinfo);
931 		return EINVAL;
932 	}
933 	error = soreserve(so, udp_sendspace, udp_recvspace);
934 	if (error) {
935 		INP_INFO_WUNLOCK(&udbinfo);
936 		return error;
937 	}
938 	s = splnet();
939 	error = in_pcballoc(so, &udbinfo, "udpinp");
940 	splx(s);
941 	if (error) {
942 		INP_INFO_WUNLOCK(&udbinfo);
943 		return error;
944 	}
945 
946 	inp = (struct inpcb *)so->so_pcb;
947 	INP_LOCK(inp);
948 	INP_INFO_WUNLOCK(&udbinfo);
949 	inp->inp_vflag |= INP_IPV4;
950 	inp->inp_ip_ttl = ip_defttl;
951 	INP_UNLOCK(inp);
952 	return 0;
953 }
954 
955 static int
956 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
957 {
958 	struct inpcb *inp;
959 	int s, error;
960 
961 	INP_INFO_WLOCK(&udbinfo);
962 	inp = sotoinpcb(so);
963 	if (inp == 0) {
964 		INP_INFO_WUNLOCK(&udbinfo);
965 		return EINVAL;
966 	}
967 	INP_LOCK(inp);
968 	s = splnet();
969 	error = in_pcbbind(inp, nam, td->td_ucred);
970 	splx(s);
971 	INP_UNLOCK(inp);
972 	INP_INFO_WUNLOCK(&udbinfo);
973 	return error;
974 }
975 
976 static int
977 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
978 {
979 	struct inpcb *inp;
980 	int s, error;
981 	struct sockaddr_in *sin;
982 
983 	INP_INFO_WLOCK(&udbinfo);
984 	inp = sotoinpcb(so);
985 	if (inp == 0) {
986 		INP_INFO_WUNLOCK(&udbinfo);
987 		return EINVAL;
988 	}
989 	INP_LOCK(inp);
990 	if (inp->inp_faddr.s_addr != INADDR_ANY) {
991 		INP_UNLOCK(inp);
992 		INP_INFO_WUNLOCK(&udbinfo);
993 		return EISCONN;
994 	}
995 	s = splnet();
996 	sin = (struct sockaddr_in *)nam;
997 	if (td && jailed(td->td_ucred))
998 		prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
999 	error = in_pcbconnect(inp, nam, td->td_ucred);
1000 	splx(s);
1001 	if (error == 0)
1002 		soisconnected(so);
1003 	INP_UNLOCK(inp);
1004 	INP_INFO_WUNLOCK(&udbinfo);
1005 	return error;
1006 }
1007 
1008 static int
1009 udp_detach(struct socket *so)
1010 {
1011 	struct inpcb *inp;
1012 	int s;
1013 
1014 	INP_INFO_WLOCK(&udbinfo);
1015 	inp = sotoinpcb(so);
1016 	if (inp == 0) {
1017 		INP_INFO_WUNLOCK(&udbinfo);
1018 		return EINVAL;
1019 	}
1020 	INP_LOCK(inp);
1021 	s = splnet();
1022 	in_pcbdetach(inp);
1023 	INP_INFO_WUNLOCK(&udbinfo);
1024 	splx(s);
1025 	return 0;
1026 }
1027 
1028 static int
1029 udp_disconnect(struct socket *so)
1030 {
1031 	struct inpcb *inp;
1032 	int s;
1033 
1034 	INP_INFO_WLOCK(&udbinfo);
1035 	inp = sotoinpcb(so);
1036 	if (inp == 0) {
1037 		INP_INFO_WUNLOCK(&udbinfo);
1038 		return EINVAL;
1039 	}
1040 	INP_LOCK(inp);
1041 	if (inp->inp_faddr.s_addr == INADDR_ANY) {
1042 		INP_INFO_WUNLOCK(&udbinfo);
1043 		INP_UNLOCK(inp);
1044 		return ENOTCONN;
1045 	}
1046 
1047 	s = splnet();
1048 	in_pcbdisconnect(inp);
1049 	inp->inp_laddr.s_addr = INADDR_ANY;
1050 	INP_UNLOCK(inp);
1051 	INP_INFO_WUNLOCK(&udbinfo);
1052 	splx(s);
1053 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1054 	return 0;
1055 }
1056 
1057 static int
1058 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1059 	    struct mbuf *control, struct thread *td)
1060 {
1061 	struct inpcb *inp;
1062 	int ret;
1063 
1064 	INP_INFO_WLOCK(&udbinfo);
1065 	inp = sotoinpcb(so);
1066 	if (inp == 0) {
1067 		INP_INFO_WUNLOCK(&udbinfo);
1068 		m_freem(m);
1069 		return EINVAL;
1070 	}
1071 	INP_LOCK(inp);
1072 	ret = udp_output(inp, m, addr, control, td);
1073 	INP_UNLOCK(inp);
1074 	INP_INFO_WUNLOCK(&udbinfo);
1075 	return ret;
1076 }
1077 
1078 int
1079 udp_shutdown(struct socket *so)
1080 {
1081 	struct inpcb *inp;
1082 
1083 	INP_INFO_RLOCK(&udbinfo);
1084 	inp = sotoinpcb(so);
1085 	if (inp == 0) {
1086 		INP_INFO_RUNLOCK(&udbinfo);
1087 		return EINVAL;
1088 	}
1089 	INP_LOCK(inp);
1090 	INP_INFO_RUNLOCK(&udbinfo);
1091 	socantsendmore(so);
1092 	INP_UNLOCK(inp);
1093 	return 0;
1094 }
1095 
1096 /*
1097  * This is the wrapper function for in_setsockaddr.  We just pass down
1098  * the pcbinfo for in_setsockaddr to lock.  We don't want to do the locking
1099  * here because in_setsockaddr will call malloc and might block.
1100  */
1101 static int
1102 udp_sockaddr(struct socket *so, struct sockaddr **nam)
1103 {
1104 	return (in_setsockaddr(so, nam, &udbinfo));
1105 }
1106 
1107 /*
1108  * This is the wrapper function for in_setpeeraddr.  We just pass down
1109  * the pcbinfo for in_setpeeraddr to lock.
1110  */
1111 static int
1112 udp_peeraddr(struct socket *so, struct sockaddr **nam)
1113 {
1114 	return (in_setpeeraddr(so, nam, &udbinfo));
1115 }
1116 
1117 struct pr_usrreqs udp_usrreqs = {
1118 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
1119 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
1120 	pru_listen_notsupp, udp_peeraddr, pru_rcvd_notsupp,
1121 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
1122 	udp_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
1123 };
1124