xref: /dragonfly/sys/netinet/udp_usrreq.c (revision 1d1731fa)
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: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
35  * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.6 2003/08/23 11:18:00 rob Exp $
36  */
37 
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/syslog.h>
53 
54 #include <vm/vm_zone.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 FAST_IPSEC
77 #include <netipsec/ipsec.h>
78 #endif /*FAST_IPSEC*/
79 
80 #ifdef IPSEC
81 #include <netinet6/ipsec.h>
82 #endif /*IPSEC*/
83 
84 #include <machine/in_cksum.h>
85 
86 /*
87  * UDP protocol implementation.
88  * Per RFC 768, August, 1980.
89  */
90 #ifndef	COMPAT_42
91 static int	udpcksum = 1;
92 #else
93 static int	udpcksum = 0;		/* XXX */
94 #endif
95 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
96 		&udpcksum, 0, "");
97 
98 int	log_in_vain = 0;
99 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
100     &log_in_vain, 0, "Log all incoming UDP packets");
101 
102 static int	blackhole = 0;
103 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
104 	&blackhole, 0, "Do not send port unreachables for refused connects");
105 
106 struct	inpcbhead udb;		/* from udp_var.h */
107 #define	udb6	udb  /* for KAME src sync over BSD*'s */
108 struct	inpcbinfo udbinfo;
109 
110 #ifndef UDBHASHSIZE
111 #define UDBHASHSIZE 16
112 #endif
113 
114 struct	udpstat udpstat;	/* from udp_var.h */
115 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
116     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
117 
118 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
119 #ifdef INET6
120 struct udp_in6 {
121 	struct sockaddr_in6	uin6_sin;
122 	u_char			uin6_init_done : 1;
123 } udp_in6 = {
124 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
125 	0
126 };
127 struct udp_ip6 {
128 	struct ip6_hdr		uip6_ip6;
129 	u_char			uip6_init_done : 1;
130 } udp_ip6;
131 #endif /* INET6 */
132 
133 static void udp_append (struct inpcb *last, struct ip *ip,
134 			    struct mbuf *n, int off);
135 #ifdef INET6
136 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
137 #endif
138 
139 static int udp_detach (struct socket *so);
140 static	int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
141 			    struct mbuf *, struct thread *);
142 
143 void
144 udp_init()
145 {
146 	LIST_INIT(&udb);
147 	udbinfo.listhead = &udb;
148 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
149 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
150 					&udbinfo.porthashmask);
151 	udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
152 				 ZONE_INTERRUPT, 0);
153 }
154 
155 void
156 udp_input(m, off, proto)
157 	struct mbuf *m;
158 	int off, proto;
159 {
160 	int iphlen = off;
161 	struct ip *ip;
162 	struct udphdr *uh;
163 	struct inpcb *inp;
164 	struct mbuf *opts = 0;
165 	int len;
166 	struct ip save_ip;
167 	struct sockaddr *append_sa;
168 
169 	udpstat.udps_ipackets++;
170 
171 	/*
172 	 * Strip IP options, if any; should skip this,
173 	 * make available to user, and use on returned packets,
174 	 * but we don't yet have a way to check the checksum
175 	 * with options still present.
176 	 */
177 	if (iphlen > sizeof (struct ip)) {
178 		ip_stripoptions(m, (struct mbuf *)0);
179 		iphlen = sizeof(struct ip);
180 	}
181 
182 	/*
183 	 * Get IP and UDP header together in first mbuf.
184 	 */
185 	ip = mtod(m, struct ip *);
186 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
187 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
188 			udpstat.udps_hdrops++;
189 			return;
190 		}
191 		ip = mtod(m, struct ip *);
192 	}
193 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
194 
195 	/* destination port of 0 is illegal, based on RFC768. */
196 	if (uh->uh_dport == 0)
197 		goto bad;
198 
199 	/*
200 	 * Make mbuf data length reflect UDP length.
201 	 * If not enough data to reflect UDP length, drop.
202 	 */
203 	len = ntohs((u_short)uh->uh_ulen);
204 	if (ip->ip_len != len) {
205 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
206 			udpstat.udps_badlen++;
207 			goto bad;
208 		}
209 		m_adj(m, len - ip->ip_len);
210 		/* ip->ip_len = len; */
211 	}
212 	/*
213 	 * Save a copy of the IP header in case we want restore it
214 	 * for sending an ICMP error message in response.
215 	 */
216 	save_ip = *ip;
217 
218 	/*
219 	 * Checksum extended UDP header and data.
220 	 */
221 	if (uh->uh_sum) {
222 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
223 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
224 				uh->uh_sum = m->m_pkthdr.csum_data;
225 			else
226 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
227 				    ip->ip_dst.s_addr, htonl((u_short)len +
228 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
229 			uh->uh_sum ^= 0xffff;
230 		} else {
231 			char b[9];
232 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
233 			bzero(((struct ipovly *)ip)->ih_x1, 9);
234 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
235 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
236 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
237 		}
238 		if (uh->uh_sum) {
239 			udpstat.udps_badsum++;
240 			m_freem(m);
241 			return;
242 		}
243 	} else
244 		udpstat.udps_nosum++;
245 
246 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
247 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
248 		struct inpcb *last;
249 		/*
250 		 * Deliver a multicast or broadcast datagram to *all* sockets
251 		 * for which the local and remote addresses and ports match
252 		 * those of the incoming datagram.  This allows more than
253 		 * one process to receive multi/broadcasts on the same port.
254 		 * (This really ought to be done for unicast datagrams as
255 		 * well, but that would cause problems with existing
256 		 * applications that open both address-specific sockets and
257 		 * a wildcard socket listening to the same port -- they would
258 		 * end up receiving duplicates of every unicast datagram.
259 		 * Those applications open the multiple sockets to overcome an
260 		 * inadequacy of the UDP socket interface, but for backwards
261 		 * compatibility we avoid the problem here rather than
262 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
263 		 */
264 
265 		/*
266 		 * Construct sockaddr format source address.
267 		 */
268 		udp_in.sin_port = uh->uh_sport;
269 		udp_in.sin_addr = ip->ip_src;
270 		/*
271 		 * Locate pcb(s) for datagram.
272 		 * (Algorithm copied from raw_intr().)
273 		 */
274 		last = NULL;
275 #ifdef INET6
276 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
277 #endif
278 		LIST_FOREACH(inp, &udb, inp_list) {
279 #ifdef INET6
280 			if ((inp->inp_vflag & INP_IPV4) == 0)
281 				continue;
282 #endif
283 			if (inp->inp_lport != uh->uh_dport)
284 				continue;
285 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
286 				if (inp->inp_laddr.s_addr !=
287 				    ip->ip_dst.s_addr)
288 					continue;
289 			}
290 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
291 				if (inp->inp_faddr.s_addr !=
292 				    ip->ip_src.s_addr ||
293 				    inp->inp_fport != uh->uh_sport)
294 					continue;
295 			}
296 
297 			if (last != NULL) {
298 				struct mbuf *n;
299 
300 #ifdef IPSEC
301 				/* check AH/ESP integrity. */
302 				if (ipsec4_in_reject_so(m, last->inp_socket))
303 					ipsecstat.in_polvio++;
304 					/* do not inject data to pcb */
305 				else
306 #endif /*IPSEC*/
307 #ifdef FAST_IPSEC
308 				/* check AH/ESP integrity. */
309 				if (ipsec4_in_reject(m, last))
310 					;
311 				else
312 #endif /*FAST_IPSEC*/
313 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
314 					udp_append(last, ip, n,
315 						   iphlen +
316 						   sizeof(struct udphdr));
317 			}
318 			last = inp;
319 			/*
320 			 * Don't look for additional matches if this one does
321 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
322 			 * socket options set.  This heuristic avoids searching
323 			 * through all pcbs in the common case of a non-shared
324 			 * port.  It * assumes that an application will never
325 			 * clear these options after setting them.
326 			 */
327 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
328 				break;
329 		}
330 
331 		if (last == NULL) {
332 			/*
333 			 * No matching pcb found; discard datagram.
334 			 * (No need to send an ICMP Port Unreachable
335 			 * for a broadcast or multicast datgram.)
336 			 */
337 			udpstat.udps_noportbcast++;
338 			goto bad;
339 		}
340 #ifdef IPSEC
341 		/* check AH/ESP integrity. */
342 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
343 			ipsecstat.in_polvio++;
344 			goto bad;
345 		}
346 #endif /*IPSEC*/
347 #ifdef FAST_IPSEC
348 		/* check AH/ESP integrity. */
349 		if (ipsec4_in_reject(m, last))
350 			goto bad;
351 #endif /*FAST_IPSEC*/
352 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
353 		return;
354 	}
355 	/*
356 	 * Locate pcb for datagram.
357 	 */
358 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
359 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
360 	if (inp == NULL) {
361 		if (log_in_vain) {
362 			char buf[4*sizeof "123"];
363 
364 			strcpy(buf, inet_ntoa(ip->ip_dst));
365 			log(LOG_INFO,
366 			    "Connection attempt to UDP %s:%d from %s:%d\n",
367 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
368 			    ntohs(uh->uh_sport));
369 		}
370 		udpstat.udps_noport++;
371 		if (m->m_flags & (M_BCAST | M_MCAST)) {
372 			udpstat.udps_noportbcast++;
373 			goto bad;
374 		}
375 		if (blackhole)
376 			goto bad;
377 #ifdef ICMP_BANDLIM
378 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
379 			goto bad;
380 #endif
381 		*ip = save_ip;
382 		ip->ip_len += iphlen;
383 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
384 		return;
385 	}
386 #ifdef IPSEC
387 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
388 		ipsecstat.in_polvio++;
389 		goto bad;
390 	}
391 #endif /*IPSEC*/
392 #ifdef FAST_IPSEC
393 	if (ipsec4_in_reject(m, inp))
394 		goto bad;
395 #endif /*FAST_IPSEC*/
396 
397 	/*
398 	 * Construct sockaddr format source address.
399 	 * Stuff source address and datagram in user buffer.
400 	 */
401 	udp_in.sin_port = uh->uh_sport;
402 	udp_in.sin_addr = ip->ip_src;
403 	if (inp->inp_flags & INP_CONTROLOPTS
404 	    || inp->inp_socket->so_options & SO_TIMESTAMP) {
405 #ifdef INET6
406 		if (inp->inp_vflag & INP_IPV6) {
407 			int savedflags;
408 
409 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
410 			savedflags = inp->inp_flags;
411 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
412 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
413 			inp->inp_flags = savedflags;
414 		} else
415 #endif
416 		ip_savecontrol(inp, &opts, ip, m);
417 	}
418  	m_adj(m, iphlen + sizeof(struct udphdr));
419 #ifdef INET6
420 	if (inp->inp_vflag & INP_IPV6) {
421 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
422 		append_sa = (struct sockaddr *)&udp_in6;
423 	} else
424 #endif
425 	append_sa = (struct sockaddr *)&udp_in;
426 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
427 		udpstat.udps_fullsock++;
428 		goto bad;
429 	}
430 	sorwakeup(inp->inp_socket);
431 	return;
432 bad:
433 	m_freem(m);
434 	if (opts)
435 		m_freem(opts);
436 	return;
437 }
438 
439 #ifdef INET6
440 static void
441 ip_2_ip6_hdr(ip6, ip)
442 	struct ip6_hdr *ip6;
443 	struct ip *ip;
444 {
445 	bzero(ip6, sizeof(*ip6));
446 
447 	ip6->ip6_vfc = IPV6_VERSION;
448 	ip6->ip6_plen = ip->ip_len;
449 	ip6->ip6_nxt = ip->ip_p;
450 	ip6->ip6_hlim = ip->ip_ttl;
451 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
452 		IPV6_ADDR_INT32_SMP;
453 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
454 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
455 }
456 #endif
457 
458 /*
459  * subroutine of udp_input(), mainly for source code readability.
460  * caller must properly init udp_ip6 and udp_in6 beforehand.
461  */
462 static void
463 udp_append(last, ip, n, off)
464 	struct inpcb *last;
465 	struct ip *ip;
466 	struct mbuf *n;
467 	int off;
468 {
469 	struct sockaddr *append_sa;
470 	struct mbuf *opts = 0;
471 
472 	if (last->inp_flags & INP_CONTROLOPTS ||
473 	    last->inp_socket->so_options & SO_TIMESTAMP) {
474 #ifdef INET6
475 		if (last->inp_vflag & INP_IPV6) {
476 			int savedflags;
477 
478 			if (udp_ip6.uip6_init_done == 0) {
479 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
480 				udp_ip6.uip6_init_done = 1;
481 			}
482 			savedflags = last->inp_flags;
483 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
484 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
485 			last->inp_flags = savedflags;
486 		} else
487 #endif
488 		ip_savecontrol(last, &opts, ip, n);
489 	}
490 #ifdef INET6
491 	if (last->inp_vflag & INP_IPV6) {
492 		if (udp_in6.uin6_init_done == 0) {
493 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
494 			udp_in6.uin6_init_done = 1;
495 		}
496 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
497 	} else
498 #endif
499 	append_sa = (struct sockaddr *)&udp_in;
500 	m_adj(n, off);
501 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
502 		m_freem(n);
503 		if (opts)
504 			m_freem(opts);
505 		udpstat.udps_fullsock++;
506 	} else
507 		sorwakeup(last->inp_socket);
508 }
509 
510 /*
511  * Notify a udp user of an asynchronous error;
512  * just wake up so that he can collect error status.
513  */
514 void
515 udp_notify(inp, errno)
516 	struct inpcb *inp;
517 	int errno;
518 {
519 	inp->inp_socket->so_error = errno;
520 	sorwakeup(inp->inp_socket);
521 	sowwakeup(inp->inp_socket);
522 }
523 
524 void
525 udp_ctlinput(cmd, sa, vip)
526 	int cmd;
527 	struct sockaddr *sa;
528 	void *vip;
529 {
530 	struct ip *ip = vip;
531 	struct udphdr *uh;
532 	void (*notify) (struct inpcb *, int) = udp_notify;
533         struct in_addr faddr;
534 	struct inpcb *inp;
535 	int s;
536 
537 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
538 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
539         	return;
540 
541 	if (PRC_IS_REDIRECT(cmd)) {
542 		ip = 0;
543 		notify = in_rtchange;
544 	} else if (cmd == PRC_HOSTDEAD)
545 		ip = 0;
546 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
547 		return;
548 	if (ip) {
549 		s = splnet();
550 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
551 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
552                     ip->ip_src, uh->uh_sport, 0, NULL);
553 		if (inp != NULL && inp->inp_socket != NULL)
554 			(*notify)(inp, inetctlerrmap[cmd]);
555 		splx(s);
556 	} else
557 		in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
558 }
559 
560 static int
561 udp_pcblist(SYSCTL_HANDLER_ARGS)
562 {
563 	int error, i, n, s;
564 	struct inpcb *inp, **inp_list;
565 	inp_gen_t gencnt;
566 	struct xinpgen xig;
567 
568 	/*
569 	 * The process of preparing the TCB list is too time-consuming and
570 	 * resource-intensive to repeat twice on every request.
571 	 */
572 	if (req->oldptr == 0) {
573 		n = udbinfo.ipi_count;
574 		req->oldidx = 2 * (sizeof xig)
575 			+ (n + n/8) * sizeof(struct xinpcb);
576 		return 0;
577 	}
578 
579 	if (req->newptr != 0)
580 		return EPERM;
581 
582 	/*
583 	 * OK, now we're committed to doing something.
584 	 */
585 	s = splnet();
586 	gencnt = udbinfo.ipi_gencnt;
587 	n = udbinfo.ipi_count;
588 	splx(s);
589 
590 	xig.xig_len = sizeof xig;
591 	xig.xig_count = n;
592 	xig.xig_gen = gencnt;
593 	xig.xig_sogen = so_gencnt;
594 	error = SYSCTL_OUT(req, &xig, sizeof xig);
595 	if (error)
596 		return error;
597 
598 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
599 	if (inp_list == 0)
600 		return ENOMEM;
601 
602 	s = splnet();
603 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
604 	     inp = LIST_NEXT(inp, inp_list)) {
605 		if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
606 			inp_list[i++] = inp;
607 	}
608 	splx(s);
609 	n = i;
610 
611 	error = 0;
612 	for (i = 0; i < n; i++) {
613 		inp = inp_list[i];
614 		if (inp->inp_gencnt <= gencnt) {
615 			struct xinpcb xi;
616 			xi.xi_len = sizeof xi;
617 			/* XXX should avoid extra copy */
618 			bcopy(inp, &xi.xi_inp, sizeof *inp);
619 			if (inp->inp_socket)
620 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
621 			error = SYSCTL_OUT(req, &xi, sizeof xi);
622 		}
623 	}
624 	if (!error) {
625 		/*
626 		 * Give the user an updated idea of our state.
627 		 * If the generation differs from what we told
628 		 * her before, she knows that something happened
629 		 * while we were processing this request, and it
630 		 * might be necessary to retry.
631 		 */
632 		s = splnet();
633 		xig.xig_gen = udbinfo.ipi_gencnt;
634 		xig.xig_sogen = so_gencnt;
635 		xig.xig_count = udbinfo.ipi_count;
636 		splx(s);
637 		error = SYSCTL_OUT(req, &xig, sizeof xig);
638 	}
639 	free(inp_list, M_TEMP);
640 	return error;
641 }
642 
643 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
644 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
645 
646 static int
647 udp_getcred(SYSCTL_HANDLER_ARGS)
648 {
649 	struct sockaddr_in addrs[2];
650 	struct inpcb *inp;
651 	int error, s;
652 
653 	error = suser(req->td);
654 	if (error)
655 		return (error);
656 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
657 	if (error)
658 		return (error);
659 	s = splnet();
660 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
661 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
662 	if (inp == NULL || inp->inp_socket == NULL) {
663 		error = ENOENT;
664 		goto out;
665 	}
666 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
667 out:
668 	splx(s);
669 	return (error);
670 }
671 
672 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
673     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
674 
675 static int
676 udp_output(inp, m, addr, control, td)
677 	struct inpcb *inp;
678 	struct mbuf *m;
679 	struct sockaddr *addr;
680 	struct mbuf *control;
681 	struct thread *td;
682 {
683 	struct udpiphdr *ui;
684 	int len = m->m_pkthdr.len;
685 	struct in_addr laddr;
686 	struct sockaddr_in *sin;
687 	int s = 0, error = 0;
688 
689 	if (control)
690 		m_freem(control);		/* XXX */
691 
692 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
693 		error = EMSGSIZE;
694 		goto release;
695 	}
696 
697 	if (addr) {
698 		sin = (struct sockaddr_in *)addr;
699 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
700 		laddr = inp->inp_laddr;
701 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
702 			error = EISCONN;
703 			goto release;
704 		}
705 		/*
706 		 * Must block input while temporarily connected.
707 		 */
708 		s = splnet();
709 		error = in_pcbconnect(inp, addr, td);
710 		if (error) {
711 			splx(s);
712 			goto release;
713 		}
714 	} else {
715 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
716 			error = ENOTCONN;
717 			goto release;
718 		}
719 	}
720 	/*
721 	 * Calculate data length and get a mbuf
722 	 * for UDP and IP headers.
723 	 */
724 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
725 	if (m == 0) {
726 		error = ENOBUFS;
727 		if (addr)
728 			splx(s);
729 		goto release;
730 	}
731 
732 	/*
733 	 * Fill in mbuf with extended UDP header
734 	 * and addresses and length put into network format.
735 	 */
736 	ui = mtod(m, struct udpiphdr *);
737 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
738 	ui->ui_pr = IPPROTO_UDP;
739 	ui->ui_src = inp->inp_laddr;
740 	ui->ui_dst = inp->inp_faddr;
741 	ui->ui_sport = inp->inp_lport;
742 	ui->ui_dport = inp->inp_fport;
743 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
744 
745 	/*
746 	 * Set up checksum and output datagram.
747 	 */
748 	if (udpcksum) {
749         	ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
750 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
751 		m->m_pkthdr.csum_flags = CSUM_UDP;
752 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
753 	} else {
754 		ui->ui_sum = 0;
755 	}
756 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
757 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
758 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
759 	udpstat.udps_opackets++;
760 
761 	error = ip_output(m, inp->inp_options, &inp->inp_route,
762 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
763 	    inp->inp_moptions, inp);
764 
765 	if (addr) {
766 		in_pcbdisconnect(inp);
767 		inp->inp_laddr = laddr;	/* XXX rehash? */
768 		splx(s);
769 	}
770 	return (error);
771 
772 release:
773 	m_freem(m);
774 	return (error);
775 }
776 
777 u_long	udp_sendspace = 9216;		/* really max datagram size */
778 					/* 40 1K datagrams */
779 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
780     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
781 
782 u_long	udp_recvspace = 40 * (1024 +
783 #ifdef INET6
784 				      sizeof(struct sockaddr_in6)
785 #else
786 				      sizeof(struct sockaddr_in)
787 #endif
788 				      );
789 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
790     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
791 
792 static int
793 udp_abort(struct socket *so)
794 {
795 	struct inpcb *inp;
796 	int s;
797 
798 	inp = sotoinpcb(so);
799 	if (inp == 0)
800 		return EINVAL;	/* ??? possible? panic instead? */
801 	soisdisconnected(so);
802 	s = splnet();
803 	in_pcbdetach(inp);
804 	splx(s);
805 	return 0;
806 }
807 
808 static int
809 udp_attach(struct socket *so, int proto, struct thread *td)
810 {
811 	struct inpcb *inp;
812 	int s, error;
813 
814 	inp = sotoinpcb(so);
815 	if (inp != 0)
816 		return EINVAL;
817 
818 	error = soreserve(so, udp_sendspace, udp_recvspace);
819 	if (error)
820 		return error;
821 	s = splnet();
822 	error = in_pcballoc(so, &udbinfo, td);
823 	splx(s);
824 	if (error)
825 		return error;
826 
827 	inp = (struct inpcb *)so->so_pcb;
828 	inp->inp_vflag |= INP_IPV4;
829 	inp->inp_ip_ttl = ip_defttl;
830 	return 0;
831 }
832 
833 static int
834 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
835 {
836 	struct inpcb *inp;
837 	int s, error;
838 
839 	inp = sotoinpcb(so);
840 	if (inp == 0)
841 		return EINVAL;
842 	s = splnet();
843 	error = in_pcbbind(inp, nam, td);
844 	splx(s);
845 	return error;
846 }
847 
848 static int
849 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
850 {
851 	struct inpcb *inp;
852 	int s, error;
853 	struct sockaddr_in *sin;
854 
855 	inp = sotoinpcb(so);
856 	if (inp == 0)
857 		return EINVAL;
858 	if (inp->inp_faddr.s_addr != INADDR_ANY)
859 		return EISCONN;
860 	error = 0;
861 	s = splnet();
862 	if (inp->inp_laddr.s_addr == INADDR_ANY &&
863 	    td->td_proc &&
864 	    td->td_proc->p_ucred->cr_prison != NULL) {
865 		error = in_pcbbind(inp, NULL, td);
866 	}
867 	if (error == 0) {
868 		sin = (struct sockaddr_in *)nam;
869 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
870 		error = in_pcbconnect(inp, nam, td);
871 	}
872 	splx(s);
873 	if (error == 0)
874 		soisconnected(so);
875 	return error;
876 }
877 
878 static int
879 udp_detach(struct socket *so)
880 {
881 	struct inpcb *inp;
882 	int s;
883 
884 	inp = sotoinpcb(so);
885 	if (inp == 0)
886 		return EINVAL;
887 	s = splnet();
888 	in_pcbdetach(inp);
889 	splx(s);
890 	return 0;
891 }
892 
893 static int
894 udp_disconnect(struct socket *so)
895 {
896 	struct inpcb *inp;
897 	int s;
898 
899 	inp = sotoinpcb(so);
900 	if (inp == 0)
901 		return EINVAL;
902 	if (inp->inp_faddr.s_addr == INADDR_ANY)
903 		return ENOTCONN;
904 
905 	s = splnet();
906 	in_pcbdisconnect(inp);
907 	inp->inp_laddr.s_addr = INADDR_ANY;
908 	splx(s);
909 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
910 	return 0;
911 }
912 
913 static int
914 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
915 	    struct mbuf *control, struct thread *td)
916 {
917 	struct inpcb *inp;
918 
919 	inp = sotoinpcb(so);
920 	if (inp == 0) {
921 		m_freem(m);
922 		return EINVAL;
923 	}
924 	return udp_output(inp, m, addr, control, td);
925 }
926 
927 int
928 udp_shutdown(struct socket *so)
929 {
930 	struct inpcb *inp;
931 
932 	inp = sotoinpcb(so);
933 	if (inp == 0)
934 		return EINVAL;
935 	socantsendmore(so);
936 	return 0;
937 }
938 
939 struct pr_usrreqs udp_usrreqs = {
940 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
941 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
942 	pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
943 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
944 	in_setsockaddr, sosend, soreceive, sopoll
945 };
946 
947