xref: /freebsd/sys/netinet/udp_usrreq.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
34  * $FreeBSD$
35  */
36 
37 #include "opt_ipsec.h"
38 #include "opt_inet6.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/sysctl.h>
51 #include <sys/syslog.h>
52 #include <sys/jail.h>
53 
54 #include <vm/uma.h>
55 
56 #include <net/if.h>
57 #include <net/route.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/ip.h>
62 #ifdef INET6
63 #include <netinet/ip6.h>
64 #endif
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
68 #ifdef INET6
69 #include <netinet6/ip6_var.h>
70 #endif
71 #include <netinet/ip_icmp.h>
72 #include <netinet/icmp_var.h>
73 #include <netinet/udp.h>
74 #include <netinet/udp_var.h>
75 
76 #ifdef IPSEC
77 #include <netinet6/ipsec.h>
78 #endif /*IPSEC*/
79 
80 #include <machine/in_cksum.h>
81 
82 /*
83  * UDP protocol implementation.
84  * Per RFC 768, August, 1980.
85  */
86 #ifndef	COMPAT_42
87 static int	udpcksum = 1;
88 #else
89 static int	udpcksum = 0;		/* XXX */
90 #endif
91 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
92 		&udpcksum, 0, "");
93 
94 int	log_in_vain = 0;
95 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
96     &log_in_vain, 0, "Log all incoming UDP packets");
97 
98 static int	blackhole = 0;
99 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
100 	&blackhole, 0, "Do not send port unreachables for refused connects");
101 
102 struct	inpcbhead udb;		/* from udp_var.h */
103 #define	udb6	udb  /* for KAME src sync over BSD*'s */
104 struct	inpcbinfo udbinfo;
105 
106 #ifndef UDBHASHSIZE
107 #define UDBHASHSIZE 16
108 #endif
109 
110 struct	udpstat udpstat;	/* from udp_var.h */
111 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
112     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
113 
114 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
115 #ifdef INET6
116 struct udp_in6 {
117 	struct sockaddr_in6	uin6_sin;
118 	u_char			uin6_init_done : 1;
119 } udp_in6 = {
120 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
121 	0
122 };
123 struct udp_ip6 {
124 	struct ip6_hdr		uip6_ip6;
125 	u_char			uip6_init_done : 1;
126 } udp_ip6;
127 #endif /* INET6 */
128 
129 static void udp_append(struct inpcb *last, struct ip *ip,
130 			    struct mbuf *n, int off);
131 #ifdef INET6
132 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
133 #endif
134 
135 static int udp_detach(struct socket *so);
136 static	int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
137 			    struct mbuf *, struct thread *);
138 
139 void
140 udp_init()
141 {
142 	LIST_INIT(&udb);
143 	udbinfo.listhead = &udb;
144 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
145 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
146 					&udbinfo.porthashmask);
147 	udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL,
148 	    NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
149 	uma_zone_set_max(udbinfo.ipi_zone, maxsockets);
150 }
151 
152 void
153 udp_input(m, off)
154 	register struct mbuf *m;
155 	int off;
156 {
157 	int iphlen = off;
158 	register struct ip *ip;
159 	register struct udphdr *uh;
160 	register struct inpcb *inp;
161 	struct mbuf *opts = 0;
162 	int len;
163 	struct ip save_ip;
164 	struct sockaddr *append_sa;
165 
166 	udpstat.udps_ipackets++;
167 
168 	/*
169 	 * Strip IP options, if any; should skip this,
170 	 * make available to user, and use on returned packets,
171 	 * but we don't yet have a way to check the checksum
172 	 * with options still present.
173 	 */
174 	if (iphlen > sizeof (struct ip)) {
175 		ip_stripoptions(m, (struct mbuf *)0);
176 		iphlen = sizeof(struct ip);
177 	}
178 
179 	/*
180 	 * Get IP and UDP header together in first mbuf.
181 	 */
182 	ip = mtod(m, struct ip *);
183 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
184 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
185 			udpstat.udps_hdrops++;
186 			return;
187 		}
188 		ip = mtod(m, struct ip *);
189 	}
190 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
191 
192 	/* destination port of 0 is illegal, based on RFC768. */
193 	if (uh->uh_dport == 0)
194 		goto bad;
195 
196 	/*
197 	 * Make mbuf data length reflect UDP length.
198 	 * If not enough data to reflect UDP length, drop.
199 	 */
200 	len = ntohs((u_short)uh->uh_ulen);
201 	if (ip->ip_len != len) {
202 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
203 			udpstat.udps_badlen++;
204 			goto bad;
205 		}
206 		m_adj(m, len - ip->ip_len);
207 		/* ip->ip_len = len; */
208 	}
209 	/*
210 	 * Save a copy of the IP header in case we want restore it
211 	 * for sending an ICMP error message in response.
212 	 */
213 	if (!blackhole)
214 		save_ip = *ip;
215 
216 	/*
217 	 * Checksum extended UDP header and data.
218 	 */
219 	if (uh->uh_sum) {
220 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
221 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
222 				uh->uh_sum = m->m_pkthdr.csum_data;
223 			else
224 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
225 				    ip->ip_dst.s_addr, htonl((u_short)len +
226 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
227 			uh->uh_sum ^= 0xffff;
228 		} else {
229 			char b[9];
230 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
231 			bzero(((struct ipovly *)ip)->ih_x1, 9);
232 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
233 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
234 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
235 		}
236 		if (uh->uh_sum) {
237 			udpstat.udps_badsum++;
238 			m_freem(m);
239 			return;
240 		}
241 	} else
242 		udpstat.udps_nosum++;
243 
244 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
245 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
246 		struct inpcb *last;
247 		/*
248 		 * Deliver a multicast or broadcast datagram to *all* sockets
249 		 * for which the local and remote addresses and ports match
250 		 * those of the incoming datagram.  This allows more than
251 		 * one process to receive multi/broadcasts on the same port.
252 		 * (This really ought to be done for unicast datagrams as
253 		 * well, but that would cause problems with existing
254 		 * applications that open both address-specific sockets and
255 		 * a wildcard socket listening to the same port -- they would
256 		 * end up receiving duplicates of every unicast datagram.
257 		 * Those applications open the multiple sockets to overcome an
258 		 * inadequacy of the UDP socket interface, but for backwards
259 		 * compatibility we avoid the problem here rather than
260 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
261 		 */
262 
263 		/*
264 		 * Construct sockaddr format source address.
265 		 */
266 		udp_in.sin_port = uh->uh_sport;
267 		udp_in.sin_addr = ip->ip_src;
268 		/*
269 		 * Locate pcb(s) for datagram.
270 		 * (Algorithm copied from raw_intr().)
271 		 */
272 		last = NULL;
273 #ifdef INET6
274 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
275 #endif
276 		LIST_FOREACH(inp, &udb, inp_list) {
277 #ifdef INET6
278 			if ((inp->inp_vflag & INP_IPV4) == 0)
279 				continue;
280 #endif
281 			if (inp->inp_lport != uh->uh_dport)
282 				continue;
283 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
284 				if (inp->inp_laddr.s_addr !=
285 				    ip->ip_dst.s_addr)
286 					continue;
287 			}
288 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
289 				if (inp->inp_faddr.s_addr !=
290 				    ip->ip_src.s_addr ||
291 				    inp->inp_fport != uh->uh_sport)
292 					continue;
293 			}
294 
295 			if (last != NULL) {
296 				struct mbuf *n;
297 
298 #ifdef IPSEC
299 				/* check AH/ESP integrity. */
300 				if (ipsec4_in_reject_so(m, last->inp_socket))
301 					ipsecstat.in_polvio++;
302 					/* do not inject data to pcb */
303 				else
304 #endif /*IPSEC*/
305 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
306 					udp_append(last, ip, n,
307 						   iphlen +
308 						   sizeof(struct udphdr));
309 			}
310 			last = inp;
311 			/*
312 			 * Don't look for additional matches if this one does
313 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
314 			 * socket options set.  This heuristic avoids searching
315 			 * through all pcbs in the common case of a non-shared
316 			 * port.  It * assumes that an application will never
317 			 * clear these options after setting them.
318 			 */
319 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
320 				break;
321 		}
322 
323 		if (last == NULL) {
324 			/*
325 			 * No matching pcb found; discard datagram.
326 			 * (No need to send an ICMP Port Unreachable
327 			 * for a broadcast or multicast datgram.)
328 			 */
329 			udpstat.udps_noportbcast++;
330 			goto bad;
331 		}
332 #ifdef IPSEC
333 		/* check AH/ESP integrity. */
334 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
335 			ipsecstat.in_polvio++;
336 			goto bad;
337 		}
338 #endif /*IPSEC*/
339 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
340 		return;
341 	}
342 	/*
343 	 * Locate pcb for datagram.
344 	 */
345 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
346 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
347 	if (inp == NULL) {
348 		if (log_in_vain) {
349 			char buf[4*sizeof "123"];
350 
351 			strcpy(buf, inet_ntoa(ip->ip_dst));
352 			log(LOG_INFO,
353 			    "Connection attempt to UDP %s:%d from %s:%d\n",
354 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
355 			    ntohs(uh->uh_sport));
356 		}
357 		udpstat.udps_noport++;
358 		if (m->m_flags & (M_BCAST | M_MCAST)) {
359 			udpstat.udps_noportbcast++;
360 			goto bad;
361 		}
362 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
363 			goto bad;
364 		if (blackhole)
365 			goto bad;
366 		*ip = save_ip;
367 		ip->ip_len += iphlen;
368 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
369 		return;
370 	}
371 #ifdef IPSEC
372 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
373 		ipsecstat.in_polvio++;
374 		goto bad;
375 	}
376 #endif /*IPSEC*/
377 
378 	/*
379 	 * Construct sockaddr format source address.
380 	 * Stuff source address and datagram in user buffer.
381 	 */
382 	udp_in.sin_port = uh->uh_sport;
383 	udp_in.sin_addr = ip->ip_src;
384 	if (inp->inp_flags & INP_CONTROLOPTS
385 	    || inp->inp_socket->so_options & SO_TIMESTAMP) {
386 #ifdef INET6
387 		if (inp->inp_vflag & INP_IPV6) {
388 			int savedflags;
389 
390 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
391 			savedflags = inp->inp_flags;
392 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
393 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
394 			inp->inp_flags = savedflags;
395 		} else
396 #endif
397 		ip_savecontrol(inp, &opts, ip, m);
398 	}
399  	m_adj(m, iphlen + sizeof(struct udphdr));
400 #ifdef INET6
401 	if (inp->inp_vflag & INP_IPV6) {
402 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
403 		append_sa = (struct sockaddr *)&udp_in6;
404 	} else
405 #endif
406 	append_sa = (struct sockaddr *)&udp_in;
407 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
408 		udpstat.udps_fullsock++;
409 		goto bad;
410 	}
411 	sorwakeup(inp->inp_socket);
412 	return;
413 bad:
414 	m_freem(m);
415 	if (opts)
416 		m_freem(opts);
417 	return;
418 }
419 
420 #ifdef INET6
421 static void
422 ip_2_ip6_hdr(ip6, ip)
423 	struct ip6_hdr *ip6;
424 	struct ip *ip;
425 {
426 	bzero(ip6, sizeof(*ip6));
427 
428 	ip6->ip6_vfc = IPV6_VERSION;
429 	ip6->ip6_plen = ip->ip_len;
430 	ip6->ip6_nxt = ip->ip_p;
431 	ip6->ip6_hlim = ip->ip_ttl;
432 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
433 		IPV6_ADDR_INT32_SMP;
434 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
435 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
436 }
437 #endif
438 
439 /*
440  * subroutine of udp_input(), mainly for source code readability.
441  * caller must properly init udp_ip6 and udp_in6 beforehand.
442  */
443 static void
444 udp_append(last, ip, n, off)
445 	struct inpcb *last;
446 	struct ip *ip;
447 	struct mbuf *n;
448 	int off;
449 {
450 	struct sockaddr *append_sa;
451 	struct mbuf *opts = 0;
452 
453 	if (last->inp_flags & INP_CONTROLOPTS ||
454 	    last->inp_socket->so_options & SO_TIMESTAMP) {
455 #ifdef INET6
456 		if (last->inp_vflag & INP_IPV6) {
457 			int savedflags;
458 
459 			if (udp_ip6.uip6_init_done == 0) {
460 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
461 				udp_ip6.uip6_init_done = 1;
462 			}
463 			savedflags = last->inp_flags;
464 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
465 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
466 			last->inp_flags = savedflags;
467 		} else
468 #endif
469 		ip_savecontrol(last, &opts, ip, n);
470 	}
471 #ifdef INET6
472 	if (last->inp_vflag & INP_IPV6) {
473 		if (udp_in6.uin6_init_done == 0) {
474 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
475 			udp_in6.uin6_init_done = 1;
476 		}
477 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
478 	} else
479 #endif
480 	append_sa = (struct sockaddr *)&udp_in;
481 	m_adj(n, off);
482 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
483 		m_freem(n);
484 		if (opts)
485 			m_freem(opts);
486 		udpstat.udps_fullsock++;
487 	} else
488 		sorwakeup(last->inp_socket);
489 }
490 
491 /*
492  * Notify a udp user of an asynchronous error;
493  * just wake up so that he can collect error status.
494  */
495 void
496 udp_notify(inp, errno)
497 	register struct inpcb *inp;
498 	int errno;
499 {
500 	inp->inp_socket->so_error = errno;
501 	sorwakeup(inp->inp_socket);
502 	sowwakeup(inp->inp_socket);
503 }
504 
505 void
506 udp_ctlinput(cmd, sa, vip)
507 	int cmd;
508 	struct sockaddr *sa;
509 	void *vip;
510 {
511 	struct ip *ip = vip;
512 	struct udphdr *uh;
513 	void (*notify)(struct inpcb *, int) = udp_notify;
514         struct in_addr faddr;
515 	struct inpcb *inp;
516 	int s;
517 
518 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
519 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
520         	return;
521 
522 	if (PRC_IS_REDIRECT(cmd)) {
523 		ip = 0;
524 		notify = in_rtchange;
525 	} else if (cmd == PRC_HOSTDEAD)
526 		ip = 0;
527 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
528 		return;
529 	if (ip) {
530 		s = splnet();
531 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
532 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
533                     ip->ip_src, uh->uh_sport, 0, NULL);
534 		if (inp != NULL && inp->inp_socket != NULL)
535 			(*notify)(inp, inetctlerrmap[cmd]);
536 		splx(s);
537 	} else
538 		in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
539 }
540 
541 static int
542 udp_pcblist(SYSCTL_HANDLER_ARGS)
543 {
544 	int error, i, n, s;
545 	struct inpcb *inp, **inp_list;
546 	inp_gen_t gencnt;
547 	struct xinpgen xig;
548 
549 	/*
550 	 * The process of preparing the TCB list is too time-consuming and
551 	 * resource-intensive to repeat twice on every request.
552 	 */
553 	if (req->oldptr == 0) {
554 		n = udbinfo.ipi_count;
555 		req->oldidx = 2 * (sizeof xig)
556 			+ (n + n/8) * sizeof(struct xinpcb);
557 		return 0;
558 	}
559 
560 	if (req->newptr != 0)
561 		return EPERM;
562 
563 	/*
564 	 * OK, now we're committed to doing something.
565 	 */
566 	s = splnet();
567 	gencnt = udbinfo.ipi_gencnt;
568 	n = udbinfo.ipi_count;
569 	splx(s);
570 
571 	xig.xig_len = sizeof xig;
572 	xig.xig_count = n;
573 	xig.xig_gen = gencnt;
574 	xig.xig_sogen = so_gencnt;
575 	error = SYSCTL_OUT(req, &xig, sizeof xig);
576 	if (error)
577 		return error;
578 
579 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
580 	if (inp_list == 0)
581 		return ENOMEM;
582 
583 	s = splnet();
584 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
585 	     inp = LIST_NEXT(inp, inp_list)) {
586 		if (inp->inp_gencnt <= gencnt) {
587 			if (cr_canseesocket(req->td->td_ucred,
588 			    inp->inp_socket))
589 				continue;
590 			inp_list[i++] = inp;
591 		}
592 	}
593 	splx(s);
594 	n = i;
595 
596 	error = 0;
597 	for (i = 0; i < n; i++) {
598 		inp = inp_list[i];
599 		if (inp->inp_gencnt <= gencnt) {
600 			struct xinpcb xi;
601 			xi.xi_len = sizeof xi;
602 			/* XXX should avoid extra copy */
603 			bcopy(inp, &xi.xi_inp, sizeof *inp);
604 			if (inp->inp_socket)
605 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
606 			error = SYSCTL_OUT(req, &xi, sizeof xi);
607 		}
608 	}
609 	if (!error) {
610 		/*
611 		 * Give the user an updated idea of our state.
612 		 * If the generation differs from what we told
613 		 * her before, she knows that something happened
614 		 * while we were processing this request, and it
615 		 * might be necessary to retry.
616 		 */
617 		s = splnet();
618 		xig.xig_gen = udbinfo.ipi_gencnt;
619 		xig.xig_sogen = so_gencnt;
620 		xig.xig_count = udbinfo.ipi_count;
621 		splx(s);
622 		error = SYSCTL_OUT(req, &xig, sizeof xig);
623 	}
624 	free(inp_list, M_TEMP);
625 	return error;
626 }
627 
628 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
629 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
630 
631 static int
632 udp_getcred(SYSCTL_HANDLER_ARGS)
633 {
634 	struct xucred xuc;
635 	struct sockaddr_in addrs[2];
636 	struct inpcb *inp;
637 	int error, s;
638 
639 	error = suser_xxx(0, req->td->td_proc, PRISON_ROOT);
640 	if (error)
641 		return (error);
642 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
643 	if (error)
644 		return (error);
645 	s = splnet();
646 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
647 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
648 	if (inp == NULL || inp->inp_socket == NULL) {
649 		error = ENOENT;
650 		goto out;
651 	}
652 	error = cr_canseesocket(req->td->td_ucred, inp->inp_socket);
653 	if (error)
654 		goto out;
655 	cru2x(inp->inp_socket->so_cred, &xuc);
656 	error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
657 out:
658 	splx(s);
659 	return (error);
660 }
661 
662 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
663     CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
664     udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
665 
666 static int
667 udp_output(inp, m, addr, control, td)
668 	register struct inpcb *inp;
669 	struct mbuf *m;
670 	struct sockaddr *addr;
671 	struct mbuf *control;
672 	struct thread *td;
673 {
674 	register struct udpiphdr *ui;
675 	register int len = m->m_pkthdr.len;
676 	struct in_addr laddr;
677 	struct sockaddr_in *sin;
678 	int s = 0, error = 0;
679 
680 	if (control)
681 		m_freem(control);		/* XXX */
682 
683 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
684 		error = EMSGSIZE;
685 		goto release;
686 	}
687 
688 	if (addr) {
689 		sin = (struct sockaddr_in *)addr;
690 		if (td && jailed(td->td_ucred))
691 			prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
692 		laddr = inp->inp_laddr;
693 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
694 			error = EISCONN;
695 			goto release;
696 		}
697 		/*
698 		 * Must block input while temporarily connected.
699 		 */
700 		s = splnet();
701 		error = in_pcbconnect(inp, addr, td);
702 		if (error) {
703 			splx(s);
704 			goto release;
705 		}
706 	} else {
707 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
708 			error = ENOTCONN;
709 			goto release;
710 		}
711 	}
712 	/*
713 	 * Calculate data length and get a mbuf
714 	 * for UDP and IP headers.
715 	 */
716 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
717 	if (m == 0) {
718 		error = ENOBUFS;
719 		if (addr)
720 			splx(s);
721 		goto release;
722 	}
723 
724 	/*
725 	 * Fill in mbuf with extended UDP header
726 	 * and addresses and length put into network format.
727 	 */
728 	ui = mtod(m, struct udpiphdr *);
729 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
730 	ui->ui_pr = IPPROTO_UDP;
731 	ui->ui_src = inp->inp_laddr;
732 	ui->ui_dst = inp->inp_faddr;
733 	ui->ui_sport = inp->inp_lport;
734 	ui->ui_dport = inp->inp_fport;
735 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
736 
737 	/*
738 	 * Set up checksum and output datagram.
739 	 */
740 	if (udpcksum) {
741         	ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
742 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
743 		m->m_pkthdr.csum_flags = CSUM_UDP;
744 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
745 	} else {
746 		ui->ui_sum = 0;
747 	}
748 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
749 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
750 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
751 	udpstat.udps_opackets++;
752 
753 #ifdef IPSEC
754 	if (ipsec_setsocket(m, inp->inp_socket) != 0) {
755 		error = ENOBUFS;
756 		goto release;
757 	}
758 #endif /*IPSEC*/
759 	error = ip_output(m, inp->inp_options, &inp->inp_route,
760 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
761 	    inp->inp_moptions);
762 
763 	if (addr) {
764 		in_pcbdisconnect(inp);
765 		inp->inp_laddr = laddr;	/* XXX rehash? */
766 		splx(s);
767 	}
768 	return (error);
769 
770 release:
771 	m_freem(m);
772 	return (error);
773 }
774 
775 u_long	udp_sendspace = 9216;		/* really max datagram size */
776 					/* 40 1K datagrams */
777 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
778     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
779 
780 u_long	udp_recvspace = 40 * (1024 +
781 #ifdef INET6
782 				      sizeof(struct sockaddr_in6)
783 #else
784 				      sizeof(struct sockaddr_in)
785 #endif
786 				      );
787 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
788     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
789 
790 static int
791 udp_abort(struct socket *so)
792 {
793 	struct inpcb *inp;
794 	int s;
795 
796 	inp = sotoinpcb(so);
797 	if (inp == 0)
798 		return EINVAL;	/* ??? possible? panic instead? */
799 	soisdisconnected(so);
800 	s = splnet();
801 	in_pcbdetach(inp);
802 	splx(s);
803 	return 0;
804 }
805 
806 static int
807 udp_attach(struct socket *so, int proto, struct thread *td)
808 {
809 	struct inpcb *inp;
810 	int s, error;
811 
812 	inp = sotoinpcb(so);
813 	if (inp != 0)
814 		return EINVAL;
815 
816 	error = soreserve(so, udp_sendspace, udp_recvspace);
817 	if (error)
818 		return error;
819 	s = splnet();
820 	error = in_pcballoc(so, &udbinfo, td);
821 	splx(s);
822 	if (error)
823 		return error;
824 
825 	inp = (struct inpcb *)so->so_pcb;
826 	inp->inp_vflag |= INP_IPV4;
827 	inp->inp_ip_ttl = ip_defttl;
828 	return 0;
829 }
830 
831 static int
832 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
833 {
834 	struct inpcb *inp;
835 	int s, error;
836 
837 	inp = sotoinpcb(so);
838 	if (inp == 0)
839 		return EINVAL;
840 	s = splnet();
841 	error = in_pcbbind(inp, nam, td);
842 	splx(s);
843 	return error;
844 }
845 
846 static int
847 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
848 {
849 	struct inpcb *inp;
850 	int s, error;
851 	struct sockaddr_in *sin;
852 
853 	inp = sotoinpcb(so);
854 	if (inp == 0)
855 		return EINVAL;
856 	if (inp->inp_faddr.s_addr != INADDR_ANY)
857 		return EISCONN;
858 	s = splnet();
859 	sin = (struct sockaddr_in *)nam;
860 	if (td && jailed(td->td_ucred))
861 		prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
862 	error = in_pcbconnect(inp, nam, td);
863 	splx(s);
864 	if (error == 0)
865 		soisconnected(so);
866 	return error;
867 }
868 
869 static int
870 udp_detach(struct socket *so)
871 {
872 	struct inpcb *inp;
873 	int s;
874 
875 	inp = sotoinpcb(so);
876 	if (inp == 0)
877 		return EINVAL;
878 	s = splnet();
879 	in_pcbdetach(inp);
880 	splx(s);
881 	return 0;
882 }
883 
884 static int
885 udp_disconnect(struct socket *so)
886 {
887 	struct inpcb *inp;
888 	int s;
889 
890 	inp = sotoinpcb(so);
891 	if (inp == 0)
892 		return EINVAL;
893 	if (inp->inp_faddr.s_addr == INADDR_ANY)
894 		return ENOTCONN;
895 
896 	s = splnet();
897 	in_pcbdisconnect(inp);
898 	inp->inp_laddr.s_addr = INADDR_ANY;
899 	splx(s);
900 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
901 	return 0;
902 }
903 
904 static int
905 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
906 	    struct mbuf *control, struct thread *td)
907 {
908 	struct inpcb *inp;
909 
910 	inp = sotoinpcb(so);
911 	if (inp == 0) {
912 		m_freem(m);
913 		return EINVAL;
914 	}
915 	return udp_output(inp, m, addr, control, td);
916 }
917 
918 int
919 udp_shutdown(struct socket *so)
920 {
921 	struct inpcb *inp;
922 
923 	inp = sotoinpcb(so);
924 	if (inp == 0)
925 		return EINVAL;
926 	socantsendmore(so);
927 	return 0;
928 }
929 
930 struct pr_usrreqs udp_usrreqs = {
931 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
932 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
933 	pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
934 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
935 	in_setsockaddr, sosend, soreceive, sopoll
936 };
937