xref: /dragonfly/sys/netinet/udp_usrreq.c (revision 2d8a3be7)
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 1.7 2003/10/28 03:51:51 dillon 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 static int	strict_mcast_mship = 1;
107 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
108 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
109 
110 struct	inpcbhead udb;		/* from udp_var.h */
111 #define	udb6	udb  /* for KAME src sync over BSD*'s */
112 struct	inpcbinfo udbinfo;
113 
114 #ifndef UDBHASHSIZE
115 #define UDBHASHSIZE 16
116 #endif
117 
118 struct	udpstat udpstat;	/* from udp_var.h */
119 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
120     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
121 
122 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
123 #ifdef INET6
124 struct udp_in6 {
125 	struct sockaddr_in6	uin6_sin;
126 	u_char			uin6_init_done : 1;
127 } udp_in6 = {
128 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
129 	0
130 };
131 struct udp_ip6 {
132 	struct ip6_hdr		uip6_ip6;
133 	u_char			uip6_init_done : 1;
134 } udp_ip6;
135 #endif /* INET6 */
136 
137 static void udp_append (struct inpcb *last, struct ip *ip,
138 			    struct mbuf *n, int off);
139 #ifdef INET6
140 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
141 #endif
142 
143 static int udp_detach (struct socket *so);
144 static	int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
145 			    struct mbuf *, struct thread *);
146 
147 void
148 udp_init()
149 {
150 	LIST_INIT(&udb);
151 	udbinfo.listhead = &udb;
152 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
153 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
154 					&udbinfo.porthashmask);
155 	udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
156 				 ZONE_INTERRUPT, 0);
157 }
158 
159 /*
160  * Check multicast packets to make sure they are only sent to sockets with
161  * multicast memberships for the packet's destination address and arrival
162  * interface.  Multicast packets to multicast-unaware sockets are also
163  * disallowed.
164  *
165  * Returns 0 if the packet is acceptable, -1 if it is not.
166  */
167 static __inline
168 int
169 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
170 {
171 	int mshipno;
172 	struct ip_moptions *mopt;
173 
174 	if (strict_mcast_mship == 0 ||
175 	    !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
176 		return(0);
177 	}
178 	mopt = inp->inp_moptions;
179 	if (mopt == NULL)
180 		return(-1);
181 	for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
182 		if (ip->ip_dst.s_addr == mopt->imo_membership[mshipno]->inm_addr.s_addr &&
183 		    m->m_pkthdr.rcvif == mopt->imo_membership[mshipno]->inm_ifp) {
184 			return(0);
185 		}
186 	}
187 	return(-1);
188 }
189 
190 void
191 udp_input(m, off, proto)
192 	struct mbuf *m;
193 	int off, proto;
194 {
195 	int iphlen = off;
196 	struct ip *ip;
197 	struct udphdr *uh;
198 	struct inpcb *inp;
199 	struct mbuf *opts = 0;
200 	int len;
201 	struct ip save_ip;
202 	struct sockaddr *append_sa;
203 
204 	udpstat.udps_ipackets++;
205 
206 	/*
207 	 * Strip IP options, if any; should skip this,
208 	 * make available to user, and use on returned packets,
209 	 * but we don't yet have a way to check the checksum
210 	 * with options still present.
211 	 */
212 	if (iphlen > sizeof (struct ip)) {
213 		ip_stripoptions(m, (struct mbuf *)0);
214 		iphlen = sizeof(struct ip);
215 	}
216 
217 	/*
218 	 * Get IP and UDP header together in first mbuf.
219 	 */
220 	ip = mtod(m, struct ip *);
221 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
222 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
223 			udpstat.udps_hdrops++;
224 			return;
225 		}
226 		ip = mtod(m, struct ip *);
227 	}
228 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
229 
230 	/* destination port of 0 is illegal, based on RFC768. */
231 	if (uh->uh_dport == 0)
232 		goto bad;
233 
234 	/*
235 	 * Make mbuf data length reflect UDP length.
236 	 * If not enough data to reflect UDP length, drop.
237 	 */
238 	len = ntohs((u_short)uh->uh_ulen);
239 	if (ip->ip_len != len) {
240 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
241 			udpstat.udps_badlen++;
242 			goto bad;
243 		}
244 		m_adj(m, len - ip->ip_len);
245 		/* ip->ip_len = len; */
246 	}
247 	/*
248 	 * Save a copy of the IP header in case we want restore it
249 	 * for sending an ICMP error message in response.
250 	 */
251 	save_ip = *ip;
252 
253 	/*
254 	 * Checksum extended UDP header and data.
255 	 */
256 	if (uh->uh_sum) {
257 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
258 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
259 				uh->uh_sum = m->m_pkthdr.csum_data;
260 			else
261 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
262 				    ip->ip_dst.s_addr, htonl((u_short)len +
263 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
264 			uh->uh_sum ^= 0xffff;
265 		} else {
266 			char b[9];
267 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
268 			bzero(((struct ipovly *)ip)->ih_x1, 9);
269 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
270 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
271 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
272 		}
273 		if (uh->uh_sum) {
274 			udpstat.udps_badsum++;
275 			m_freem(m);
276 			return;
277 		}
278 	} else
279 		udpstat.udps_nosum++;
280 
281 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
282 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
283 		struct inpcb *last;
284 		/*
285 		 * Deliver a multicast or broadcast datagram to *all* sockets
286 		 * for which the local and remote addresses and ports match
287 		 * those of the incoming datagram.  This allows more than
288 		 * one process to receive multi/broadcasts on the same port.
289 		 * (This really ought to be done for unicast datagrams as
290 		 * well, but that would cause problems with existing
291 		 * applications that open both address-specific sockets and
292 		 * a wildcard socket listening to the same port -- they would
293 		 * end up receiving duplicates of every unicast datagram.
294 		 * Those applications open the multiple sockets to overcome an
295 		 * inadequacy of the UDP socket interface, but for backwards
296 		 * compatibility we avoid the problem here rather than
297 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
298 		 */
299 
300 		/*
301 		 * Construct sockaddr format source address.
302 		 */
303 		udp_in.sin_port = uh->uh_sport;
304 		udp_in.sin_addr = ip->ip_src;
305 		/*
306 		 * Locate pcb(s) for datagram.
307 		 * (Algorithm copied from raw_intr().)
308 		 */
309 		last = NULL;
310 #ifdef INET6
311 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
312 #endif
313 		LIST_FOREACH(inp, &udb, inp_list) {
314 #ifdef INET6
315 			if ((inp->inp_vflag & INP_IPV4) == 0)
316 				continue;
317 #endif
318 			if (inp->inp_lport != uh->uh_dport)
319 				continue;
320 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
321 				if (inp->inp_laddr.s_addr !=
322 				    ip->ip_dst.s_addr)
323 					continue;
324 			}
325 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
326 				if (inp->inp_faddr.s_addr !=
327 				    ip->ip_src.s_addr ||
328 				    inp->inp_fport != uh->uh_sport)
329 					continue;
330 			}
331 
332 			if (check_multicast_membership(ip, inp, m) < 0)
333 				continue;
334 
335 			if (last != NULL) {
336 				struct mbuf *n;
337 
338 #ifdef IPSEC
339 				/* check AH/ESP integrity. */
340 				if (ipsec4_in_reject_so(m, last->inp_socket))
341 					ipsecstat.in_polvio++;
342 					/* do not inject data to pcb */
343 				else
344 #endif /*IPSEC*/
345 #ifdef FAST_IPSEC
346 				/* check AH/ESP integrity. */
347 				if (ipsec4_in_reject(m, last))
348 					;
349 				else
350 #endif /*FAST_IPSEC*/
351 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
352 					udp_append(last, ip, n,
353 						   iphlen +
354 						   sizeof(struct udphdr));
355 			}
356 			last = inp;
357 			/*
358 			 * Don't look for additional matches if this one does
359 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
360 			 * socket options set.  This heuristic avoids searching
361 			 * through all pcbs in the common case of a non-shared
362 			 * port.  It * assumes that an application will never
363 			 * clear these options after setting them.
364 			 */
365 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
366 				break;
367 		}
368 
369 		if (last == NULL) {
370 			/*
371 			 * No matching pcb found; discard datagram.
372 			 * (No need to send an ICMP Port Unreachable
373 			 * for a broadcast or multicast datgram.)
374 			 */
375 			udpstat.udps_noportbcast++;
376 			goto bad;
377 		}
378 #ifdef IPSEC
379 		/* check AH/ESP integrity. */
380 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
381 			ipsecstat.in_polvio++;
382 			goto bad;
383 		}
384 #endif /*IPSEC*/
385 #ifdef FAST_IPSEC
386 		/* check AH/ESP integrity. */
387 		if (ipsec4_in_reject(m, last))
388 			goto bad;
389 #endif /*FAST_IPSEC*/
390 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
391 		return;
392 	}
393 	/*
394 	 * Locate pcb for datagram.
395 	 */
396 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
397 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
398 	if (inp == NULL) {
399 		if (log_in_vain) {
400 			char buf[4*sizeof "123"];
401 
402 			strcpy(buf, inet_ntoa(ip->ip_dst));
403 			log(LOG_INFO,
404 			    "Connection attempt to UDP %s:%d from %s:%d\n",
405 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
406 			    ntohs(uh->uh_sport));
407 		}
408 		udpstat.udps_noport++;
409 		if (m->m_flags & (M_BCAST | M_MCAST)) {
410 			udpstat.udps_noportbcast++;
411 			goto bad;
412 		}
413 		if (blackhole)
414 			goto bad;
415 #ifdef ICMP_BANDLIM
416 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
417 			goto bad;
418 #endif
419 		*ip = save_ip;
420 		ip->ip_len += iphlen;
421 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
422 		return;
423 	}
424 #ifdef IPSEC
425 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
426 		ipsecstat.in_polvio++;
427 		goto bad;
428 	}
429 #endif /*IPSEC*/
430 #ifdef FAST_IPSEC
431 	if (ipsec4_in_reject(m, inp))
432 		goto bad;
433 #endif /*FAST_IPSEC*/
434 
435 	/*
436 	 * Construct sockaddr format source address.
437 	 * Stuff source address and datagram in user buffer.
438 	 */
439 	udp_in.sin_port = uh->uh_sport;
440 	udp_in.sin_addr = ip->ip_src;
441 	if (inp->inp_flags & INP_CONTROLOPTS
442 	    || inp->inp_socket->so_options & SO_TIMESTAMP) {
443 #ifdef INET6
444 		if (inp->inp_vflag & INP_IPV6) {
445 			int savedflags;
446 
447 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
448 			savedflags = inp->inp_flags;
449 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
450 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
451 			inp->inp_flags = savedflags;
452 		} else
453 #endif
454 		ip_savecontrol(inp, &opts, ip, m);
455 	}
456  	m_adj(m, iphlen + sizeof(struct udphdr));
457 #ifdef INET6
458 	if (inp->inp_vflag & INP_IPV6) {
459 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
460 		append_sa = (struct sockaddr *)&udp_in6;
461 	} else
462 #endif
463 	append_sa = (struct sockaddr *)&udp_in;
464 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
465 		udpstat.udps_fullsock++;
466 		goto bad;
467 	}
468 	sorwakeup(inp->inp_socket);
469 	return;
470 bad:
471 	m_freem(m);
472 	if (opts)
473 		m_freem(opts);
474 	return;
475 }
476 
477 #ifdef INET6
478 static void
479 ip_2_ip6_hdr(ip6, ip)
480 	struct ip6_hdr *ip6;
481 	struct ip *ip;
482 {
483 	bzero(ip6, sizeof(*ip6));
484 
485 	ip6->ip6_vfc = IPV6_VERSION;
486 	ip6->ip6_plen = ip->ip_len;
487 	ip6->ip6_nxt = ip->ip_p;
488 	ip6->ip6_hlim = ip->ip_ttl;
489 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
490 		IPV6_ADDR_INT32_SMP;
491 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
492 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
493 }
494 #endif
495 
496 /*
497  * subroutine of udp_input(), mainly for source code readability.
498  * caller must properly init udp_ip6 and udp_in6 beforehand.
499  */
500 static void
501 udp_append(last, ip, n, off)
502 	struct inpcb *last;
503 	struct ip *ip;
504 	struct mbuf *n;
505 	int off;
506 {
507 	struct sockaddr *append_sa;
508 	struct mbuf *opts = 0;
509 
510 	if (last->inp_flags & INP_CONTROLOPTS ||
511 	    last->inp_socket->so_options & SO_TIMESTAMP) {
512 #ifdef INET6
513 		if (last->inp_vflag & INP_IPV6) {
514 			int savedflags;
515 
516 			if (udp_ip6.uip6_init_done == 0) {
517 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
518 				udp_ip6.uip6_init_done = 1;
519 			}
520 			savedflags = last->inp_flags;
521 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
522 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
523 			last->inp_flags = savedflags;
524 		} else
525 #endif
526 		ip_savecontrol(last, &opts, ip, n);
527 	}
528 #ifdef INET6
529 	if (last->inp_vflag & INP_IPV6) {
530 		if (udp_in6.uin6_init_done == 0) {
531 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
532 			udp_in6.uin6_init_done = 1;
533 		}
534 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
535 	} else
536 #endif
537 	append_sa = (struct sockaddr *)&udp_in;
538 	m_adj(n, off);
539 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
540 		m_freem(n);
541 		if (opts)
542 			m_freem(opts);
543 		udpstat.udps_fullsock++;
544 	} else
545 		sorwakeup(last->inp_socket);
546 }
547 
548 /*
549  * Notify a udp user of an asynchronous error;
550  * just wake up so that he can collect error status.
551  */
552 void
553 udp_notify(inp, errno)
554 	struct inpcb *inp;
555 	int errno;
556 {
557 	inp->inp_socket->so_error = errno;
558 	sorwakeup(inp->inp_socket);
559 	sowwakeup(inp->inp_socket);
560 }
561 
562 void
563 udp_ctlinput(cmd, sa, vip)
564 	int cmd;
565 	struct sockaddr *sa;
566 	void *vip;
567 {
568 	struct ip *ip = vip;
569 	struct udphdr *uh;
570 	void (*notify) (struct inpcb *, int) = udp_notify;
571         struct in_addr faddr;
572 	struct inpcb *inp;
573 	int s;
574 
575 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
576 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
577         	return;
578 
579 	if (PRC_IS_REDIRECT(cmd)) {
580 		ip = 0;
581 		notify = in_rtchange;
582 	} else if (cmd == PRC_HOSTDEAD)
583 		ip = 0;
584 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
585 		return;
586 	if (ip) {
587 		s = splnet();
588 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
589 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
590                     ip->ip_src, uh->uh_sport, 0, NULL);
591 		if (inp != NULL && inp->inp_socket != NULL)
592 			(*notify)(inp, inetctlerrmap[cmd]);
593 		splx(s);
594 	} else
595 		in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
596 }
597 
598 static int
599 udp_pcblist(SYSCTL_HANDLER_ARGS)
600 {
601 	int error, i, n, s;
602 	struct inpcb *inp, **inp_list;
603 	inp_gen_t gencnt;
604 	struct xinpgen xig;
605 
606 	/*
607 	 * The process of preparing the TCB list is too time-consuming and
608 	 * resource-intensive to repeat twice on every request.
609 	 */
610 	if (req->oldptr == 0) {
611 		n = udbinfo.ipi_count;
612 		req->oldidx = 2 * (sizeof xig)
613 			+ (n + n/8) * sizeof(struct xinpcb);
614 		return 0;
615 	}
616 
617 	if (req->newptr != 0)
618 		return EPERM;
619 
620 	/*
621 	 * OK, now we're committed to doing something.
622 	 */
623 	s = splnet();
624 	gencnt = udbinfo.ipi_gencnt;
625 	n = udbinfo.ipi_count;
626 	splx(s);
627 
628 	xig.xig_len = sizeof xig;
629 	xig.xig_count = n;
630 	xig.xig_gen = gencnt;
631 	xig.xig_sogen = so_gencnt;
632 	error = SYSCTL_OUT(req, &xig, sizeof xig);
633 	if (error)
634 		return error;
635 
636 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
637 	if (inp_list == 0)
638 		return ENOMEM;
639 
640 	s = splnet();
641 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
642 	     inp = LIST_NEXT(inp, inp_list)) {
643 		if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
644 			inp_list[i++] = inp;
645 	}
646 	splx(s);
647 	n = i;
648 
649 	error = 0;
650 	for (i = 0; i < n; i++) {
651 		inp = inp_list[i];
652 		if (inp->inp_gencnt <= gencnt) {
653 			struct xinpcb xi;
654 			xi.xi_len = sizeof xi;
655 			/* XXX should avoid extra copy */
656 			bcopy(inp, &xi.xi_inp, sizeof *inp);
657 			if (inp->inp_socket)
658 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
659 			error = SYSCTL_OUT(req, &xi, sizeof xi);
660 		}
661 	}
662 	if (!error) {
663 		/*
664 		 * Give the user an updated idea of our state.
665 		 * If the generation differs from what we told
666 		 * her before, she knows that something happened
667 		 * while we were processing this request, and it
668 		 * might be necessary to retry.
669 		 */
670 		s = splnet();
671 		xig.xig_gen = udbinfo.ipi_gencnt;
672 		xig.xig_sogen = so_gencnt;
673 		xig.xig_count = udbinfo.ipi_count;
674 		splx(s);
675 		error = SYSCTL_OUT(req, &xig, sizeof xig);
676 	}
677 	free(inp_list, M_TEMP);
678 	return error;
679 }
680 
681 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
682 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
683 
684 static int
685 udp_getcred(SYSCTL_HANDLER_ARGS)
686 {
687 	struct sockaddr_in addrs[2];
688 	struct inpcb *inp;
689 	int error, s;
690 
691 	error = suser(req->td);
692 	if (error)
693 		return (error);
694 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
695 	if (error)
696 		return (error);
697 	s = splnet();
698 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
699 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
700 	if (inp == NULL || inp->inp_socket == NULL) {
701 		error = ENOENT;
702 		goto out;
703 	}
704 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
705 out:
706 	splx(s);
707 	return (error);
708 }
709 
710 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
711     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
712 
713 static int
714 udp_output(inp, m, addr, control, td)
715 	struct inpcb *inp;
716 	struct mbuf *m;
717 	struct sockaddr *addr;
718 	struct mbuf *control;
719 	struct thread *td;
720 {
721 	struct udpiphdr *ui;
722 	int len = m->m_pkthdr.len;
723 	struct in_addr laddr;
724 	struct sockaddr_in *sin;
725 	int s = 0, error = 0;
726 
727 	if (control)
728 		m_freem(control);		/* XXX */
729 
730 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
731 		error = EMSGSIZE;
732 		goto release;
733 	}
734 
735 	if (addr) {
736 		sin = (struct sockaddr_in *)addr;
737 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
738 		laddr = inp->inp_laddr;
739 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
740 			error = EISCONN;
741 			goto release;
742 		}
743 		/*
744 		 * Must block input while temporarily connected.
745 		 */
746 		s = splnet();
747 		error = in_pcbconnect(inp, addr, td);
748 		if (error) {
749 			splx(s);
750 			goto release;
751 		}
752 	} else {
753 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
754 			error = ENOTCONN;
755 			goto release;
756 		}
757 	}
758 	/*
759 	 * Calculate data length and get a mbuf
760 	 * for UDP and IP headers.
761 	 */
762 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
763 	if (m == 0) {
764 		error = ENOBUFS;
765 		if (addr)
766 			splx(s);
767 		goto release;
768 	}
769 
770 	/*
771 	 * Fill in mbuf with extended UDP header
772 	 * and addresses and length put into network format.
773 	 */
774 	ui = mtod(m, struct udpiphdr *);
775 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
776 	ui->ui_pr = IPPROTO_UDP;
777 	ui->ui_src = inp->inp_laddr;
778 	ui->ui_dst = inp->inp_faddr;
779 	ui->ui_sport = inp->inp_lport;
780 	ui->ui_dport = inp->inp_fport;
781 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
782 
783 	/*
784 	 * Set up checksum and output datagram.
785 	 */
786 	if (udpcksum) {
787         	ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
788 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
789 		m->m_pkthdr.csum_flags = CSUM_UDP;
790 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
791 	} else {
792 		ui->ui_sum = 0;
793 	}
794 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
795 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
796 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
797 	udpstat.udps_opackets++;
798 
799 	error = ip_output(m, inp->inp_options, &inp->inp_route,
800 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
801 	    inp->inp_moptions, inp);
802 
803 	if (addr) {
804 		in_pcbdisconnect(inp);
805 		inp->inp_laddr = laddr;	/* XXX rehash? */
806 		splx(s);
807 	}
808 	return (error);
809 
810 release:
811 	m_freem(m);
812 	return (error);
813 }
814 
815 u_long	udp_sendspace = 9216;		/* really max datagram size */
816 					/* 40 1K datagrams */
817 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
818     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
819 
820 u_long	udp_recvspace = 40 * (1024 +
821 #ifdef INET6
822 				      sizeof(struct sockaddr_in6)
823 #else
824 				      sizeof(struct sockaddr_in)
825 #endif
826 				      );
827 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
828     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
829 
830 static int
831 udp_abort(struct socket *so)
832 {
833 	struct inpcb *inp;
834 	int s;
835 
836 	inp = sotoinpcb(so);
837 	if (inp == 0)
838 		return EINVAL;	/* ??? possible? panic instead? */
839 	soisdisconnected(so);
840 	s = splnet();
841 	in_pcbdetach(inp);
842 	splx(s);
843 	return 0;
844 }
845 
846 static int
847 udp_attach(struct socket *so, int proto, struct thread *td)
848 {
849 	struct inpcb *inp;
850 	int s, error;
851 
852 	inp = sotoinpcb(so);
853 	if (inp != 0)
854 		return EINVAL;
855 
856 	error = soreserve(so, udp_sendspace, udp_recvspace);
857 	if (error)
858 		return error;
859 	s = splnet();
860 	error = in_pcballoc(so, &udbinfo, td);
861 	splx(s);
862 	if (error)
863 		return error;
864 
865 	inp = (struct inpcb *)so->so_pcb;
866 	inp->inp_vflag |= INP_IPV4;
867 	inp->inp_ip_ttl = ip_defttl;
868 	return 0;
869 }
870 
871 static int
872 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
873 {
874 	struct inpcb *inp;
875 	int s, error;
876 
877 	inp = sotoinpcb(so);
878 	if (inp == 0)
879 		return EINVAL;
880 	s = splnet();
881 	error = in_pcbbind(inp, nam, td);
882 	splx(s);
883 	return error;
884 }
885 
886 static int
887 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
888 {
889 	struct inpcb *inp;
890 	int s, error;
891 	struct sockaddr_in *sin;
892 
893 	inp = sotoinpcb(so);
894 	if (inp == 0)
895 		return EINVAL;
896 	if (inp->inp_faddr.s_addr != INADDR_ANY)
897 		return EISCONN;
898 	error = 0;
899 	s = splnet();
900 	if (inp->inp_laddr.s_addr == INADDR_ANY &&
901 	    td->td_proc &&
902 	    td->td_proc->p_ucred->cr_prison != NULL) {
903 		error = in_pcbbind(inp, NULL, td);
904 	}
905 	if (error == 0) {
906 		sin = (struct sockaddr_in *)nam;
907 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
908 		error = in_pcbconnect(inp, nam, td);
909 	}
910 	splx(s);
911 	if (error == 0)
912 		soisconnected(so);
913 	return error;
914 }
915 
916 static int
917 udp_detach(struct socket *so)
918 {
919 	struct inpcb *inp;
920 	int s;
921 
922 	inp = sotoinpcb(so);
923 	if (inp == 0)
924 		return EINVAL;
925 	s = splnet();
926 	in_pcbdetach(inp);
927 	splx(s);
928 	return 0;
929 }
930 
931 static int
932 udp_disconnect(struct socket *so)
933 {
934 	struct inpcb *inp;
935 	int s;
936 
937 	inp = sotoinpcb(so);
938 	if (inp == 0)
939 		return EINVAL;
940 	if (inp->inp_faddr.s_addr == INADDR_ANY)
941 		return ENOTCONN;
942 
943 	s = splnet();
944 	in_pcbdisconnect(inp);
945 	inp->inp_laddr.s_addr = INADDR_ANY;
946 	splx(s);
947 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
948 	return 0;
949 }
950 
951 static int
952 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
953 	    struct mbuf *control, struct thread *td)
954 {
955 	struct inpcb *inp;
956 
957 	inp = sotoinpcb(so);
958 	if (inp == 0) {
959 		m_freem(m);
960 		return EINVAL;
961 	}
962 	return udp_output(inp, m, addr, control, td);
963 }
964 
965 int
966 udp_shutdown(struct socket *so)
967 {
968 	struct inpcb *inp;
969 
970 	inp = sotoinpcb(so);
971 	if (inp == 0)
972 		return EINVAL;
973 	socantsendmore(so);
974 	return 0;
975 }
976 
977 struct pr_usrreqs udp_usrreqs = {
978 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
979 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
980 	pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
981 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
982 	in_setsockaddr, sosend, soreceive, sopoll
983 };
984 
985