xref: /freebsd/sys/netinet/udp_usrreq.c (revision 0f8f86b71f022b803e99151c19db81b280f245dc)
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 #include "opt_mac.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/domain.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/mac.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/signalvar.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sx.h>
56 #include <sys/sysctl.h>
57 #include <sys/syslog.h>
58 
59 #include <vm/uma.h>
60 
61 #include <net/if.h>
62 #include <net/route.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/in_var.h>
68 #include <netinet/ip.h>
69 #ifdef INET6
70 #include <netinet/ip6.h>
71 #endif
72 #include <netinet/ip_icmp.h>
73 #include <netinet/icmp_var.h>
74 #include <netinet/ip_var.h>
75 #ifdef INET6
76 #include <netinet6/ip6_var.h>
77 #endif
78 #include <netinet/udp.h>
79 #include <netinet/udp_var.h>
80 
81 #ifdef FAST_IPSEC
82 #include <netipsec/ipsec.h>
83 #endif /*FAST_IPSEC*/
84 
85 #ifdef IPSEC
86 #include <netinet6/ipsec.h>
87 #endif /*IPSEC*/
88 
89 #include <machine/in_cksum.h>
90 
91 /*
92  * UDP protocol implementation.
93  * Per RFC 768, August, 1980.
94  */
95 #ifndef	COMPAT_42
96 static int	udpcksum = 1;
97 #else
98 static int	udpcksum = 0;		/* XXX */
99 #endif
100 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
101 		&udpcksum, 0, "");
102 
103 int	log_in_vain = 0;
104 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
105     &log_in_vain, 0, "Log all incoming UDP packets");
106 
107 static int	blackhole = 0;
108 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
109 	&blackhole, 0, "Do not send port unreachables for refused connects");
110 
111 static int	strict_mcast_mship = 0;
112 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
113 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
114 
115 struct	inpcbhead udb;		/* from udp_var.h */
116 #define	udb6	udb  /* for KAME src sync over BSD*'s */
117 struct	inpcbinfo udbinfo;
118 
119 #ifndef UDBHASHSIZE
120 #define UDBHASHSIZE 16
121 #endif
122 
123 struct	udpstat udpstat;	/* from udp_var.h */
124 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
125     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
126 
127 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
128 #ifdef INET6
129 struct udp_in6 {
130 	struct sockaddr_in6	uin6_sin;
131 	u_char			uin6_init_done : 1;
132 } udp_in6 = {
133 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
134 	0
135 };
136 struct udp_ip6 {
137 	struct ip6_hdr		uip6_ip6;
138 	u_char			uip6_init_done : 1;
139 } udp_ip6;
140 #endif /* INET6 */
141 
142 static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
143 		int off);
144 #ifdef INET6
145 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
146 #endif
147 
148 static int udp_detach(struct socket *so);
149 static	int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
150 		struct mbuf *, struct thread *);
151 
152 void
153 udp_init()
154 {
155 	INP_INFO_LOCK_INIT(&udbinfo, "udp");
156 	LIST_INIT(&udb);
157 	udbinfo.listhead = &udb;
158 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
159 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
160 					&udbinfo.porthashmask);
161 	udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL,
162 	    NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
163 	uma_zone_set_max(udbinfo.ipi_zone, maxsockets);
164 }
165 
166 void
167 udp_input(m, off)
168 	register struct mbuf *m;
169 	int off;
170 {
171 	int iphlen = off;
172 	register struct ip *ip;
173 	register struct udphdr *uh;
174 	register struct inpcb *inp;
175 	struct mbuf *opts = 0;
176 	int len;
177 	struct ip save_ip;
178 
179 	udpstat.udps_ipackets++;
180 
181 	/*
182 	 * Strip IP options, if any; should skip this,
183 	 * make available to user, and use on returned packets,
184 	 * but we don't yet have a way to check the checksum
185 	 * with options still present.
186 	 */
187 	if (iphlen > sizeof (struct ip)) {
188 		ip_stripoptions(m, (struct mbuf *)0);
189 		iphlen = sizeof(struct ip);
190 	}
191 
192 	/*
193 	 * Get IP and UDP header together in first mbuf.
194 	 */
195 	ip = mtod(m, struct ip *);
196 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
197 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
198 			udpstat.udps_hdrops++;
199 			return;
200 		}
201 		ip = mtod(m, struct ip *);
202 	}
203 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
204 
205 	/* destination port of 0 is illegal, based on RFC768. */
206 	if (uh->uh_dport == 0)
207 		goto badunlocked;
208 
209 	/*
210 	 * Construct sockaddr format source address.
211 	 * Stuff source address and datagram in user buffer.
212 	 */
213 	udp_in.sin_port = uh->uh_sport;
214 	udp_in.sin_addr = ip->ip_src;
215 #ifdef INET6
216 	udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
217 #endif
218 
219 	/*
220 	 * Make mbuf data length reflect UDP length.
221 	 * If not enough data to reflect UDP length, drop.
222 	 */
223 	len = ntohs((u_short)uh->uh_ulen);
224 	if (ip->ip_len != len) {
225 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
226 			udpstat.udps_badlen++;
227 			goto badunlocked;
228 		}
229 		m_adj(m, len - ip->ip_len);
230 		/* ip->ip_len = len; */
231 	}
232 	/*
233 	 * Save a copy of the IP header in case we want restore it
234 	 * for sending an ICMP error message in response.
235 	 */
236 	if (!blackhole)
237 		save_ip = *ip;
238 
239 	/*
240 	 * Checksum extended UDP header and data.
241 	 */
242 	if (uh->uh_sum) {
243 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
244 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
245 				uh->uh_sum = m->m_pkthdr.csum_data;
246 			else
247 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
248 				    ip->ip_dst.s_addr, htonl((u_short)len +
249 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
250 			uh->uh_sum ^= 0xffff;
251 		} else {
252 			char b[9];
253 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
254 			bzero(((struct ipovly *)ip)->ih_x1, 9);
255 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
256 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
257 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
258 		}
259 		if (uh->uh_sum) {
260 			udpstat.udps_badsum++;
261 			m_freem(m);
262 			return;
263 		}
264 	} else
265 		udpstat.udps_nosum++;
266 
267 	INP_INFO_RLOCK(&udbinfo);
268 
269 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
270 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
271 		struct inpcb *last;
272 		/*
273 		 * Deliver a multicast or broadcast datagram to *all* sockets
274 		 * for which the local and remote addresses and ports match
275 		 * those of the incoming datagram.  This allows more than
276 		 * one process to receive multi/broadcasts on the same port.
277 		 * (This really ought to be done for unicast datagrams as
278 		 * well, but that would cause problems with existing
279 		 * applications that open both address-specific sockets and
280 		 * a wildcard socket listening to the same port -- they would
281 		 * end up receiving duplicates of every unicast datagram.
282 		 * Those applications open the multiple sockets to overcome an
283 		 * inadequacy of the UDP socket interface, but for backwards
284 		 * compatibility we avoid the problem here rather than
285 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
286 		 */
287 
288 		/*
289 		 * Locate pcb(s) for datagram.
290 		 * (Algorithm copied from raw_intr().)
291 		 */
292 		last = NULL;
293 		LIST_FOREACH(inp, &udb, inp_list) {
294 			INP_LOCK(inp);
295 			if (inp->inp_lport != uh->uh_dport) {
296 		docontinue:
297 				INP_UNLOCK(inp);
298 				continue;
299 			}
300 #ifdef INET6
301 			if ((inp->inp_vflag & INP_IPV4) == 0)
302 				goto docontinue;
303 #endif
304 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
305 				if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
306 					goto docontinue;
307 			}
308 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
309 				if (inp->inp_faddr.s_addr !=
310 				    ip->ip_src.s_addr ||
311 				    inp->inp_fport != uh->uh_sport)
312 					goto docontinue;
313 			}
314 
315 			/*
316 			 * Check multicast packets to make sure they are only
317 			 * sent to sockets with multicast memberships for the
318 			 * packet's destination address and arrival interface
319 			 */
320 #define MSHIP(_inp, n) ((_inp)->inp_moptions->imo_membership[(n)])
321 #define NMSHIPS(_inp) ((_inp)->inp_moptions->imo_num_memberships)
322 			if (strict_mcast_mship && inp->inp_moptions != NULL) {
323 				int mship, foundmship = 0;
324 
325 				for (mship = 0; mship < NMSHIPS(inp); mship++) {
326 					if (MSHIP(inp, mship)->inm_addr.s_addr
327 					    == ip->ip_dst.s_addr &&
328 					    MSHIP(inp, mship)->inm_ifp
329 					    == m->m_pkthdr.rcvif) {
330 						foundmship = 1;
331 						break;
332 					}
333 				}
334 				if (foundmship == 0)
335 					goto docontinue;
336 			}
337 #undef NMSHIPS
338 #undef MSHIP
339 			if (last != NULL) {
340 				struct mbuf *n;
341 
342 				n = m_copy(m, 0, M_COPYALL);
343 				if (n != NULL)
344 					udp_append(last, ip, n,
345 						   iphlen +
346 						   sizeof(struct udphdr));
347 				INP_UNLOCK(last);
348 			}
349 			last = inp;
350 			/*
351 			 * Don't look for additional matches if this one does
352 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
353 			 * socket options set.  This heuristic avoids searching
354 			 * through all pcbs in the common case of a non-shared
355 			 * port.  It * assumes that an application will never
356 			 * clear these options after setting them.
357 			 */
358 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
359 				break;
360 		}
361 
362 		if (last == NULL) {
363 			/*
364 			 * No matching pcb found; discard datagram.
365 			 * (No need to send an ICMP Port Unreachable
366 			 * for a broadcast or multicast datgram.)
367 			 */
368 			udpstat.udps_noportbcast++;
369 			goto badheadlocked;
370 		}
371 		INP_INFO_RUNLOCK(&udbinfo);
372 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
373 		INP_UNLOCK(last);
374 		return;
375 	}
376 	/*
377 	 * Locate pcb for datagram.
378 	 */
379 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
380 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
381 	if (inp == NULL) {
382 		if (log_in_vain) {
383 			char buf[4*sizeof "123"];
384 
385 			strcpy(buf, inet_ntoa(ip->ip_dst));
386 			log(LOG_INFO,
387 			    "Connection attempt to UDP %s:%d from %s:%d\n",
388 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
389 			    ntohs(uh->uh_sport));
390 		}
391 		udpstat.udps_noport++;
392 		if (m->m_flags & (M_BCAST | M_MCAST)) {
393 			udpstat.udps_noportbcast++;
394 			goto badheadlocked;
395 		}
396 		if (blackhole)
397 			goto badheadlocked;
398 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
399 			goto badheadlocked;
400 		*ip = save_ip;
401 		ip->ip_len += iphlen;
402 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
403 		INP_INFO_RUNLOCK(&udbinfo);
404 		return;
405 	}
406 	INP_LOCK(inp);
407 	INP_INFO_RUNLOCK(&udbinfo);
408 	udp_append(inp, ip, m, iphlen + sizeof(struct udphdr));
409 	INP_UNLOCK(inp);
410 	return;
411 
412 badheadlocked:
413 	INP_INFO_RUNLOCK(&udbinfo);
414 	if (inp)
415 		INP_UNLOCK(inp);
416 badunlocked:
417 	m_freem(m);
418 	if (opts)
419 		m_freem(opts);
420 	return;
421 }
422 
423 #ifdef INET6
424 static void
425 ip_2_ip6_hdr(ip6, ip)
426 	struct ip6_hdr *ip6;
427 	struct ip *ip;
428 {
429 	bzero(ip6, sizeof(*ip6));
430 
431 	ip6->ip6_vfc = IPV6_VERSION;
432 	ip6->ip6_plen = ip->ip_len;
433 	ip6->ip6_nxt = ip->ip_p;
434 	ip6->ip6_hlim = ip->ip_ttl;
435 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
436 		IPV6_ADDR_INT32_SMP;
437 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
438 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
439 }
440 #endif
441 
442 /*
443  * subroutine of udp_input(), mainly for source code readability.
444  * caller must properly init udp_ip6 and udp_in6 beforehand.
445  */
446 static void
447 udp_append(last, ip, n, off)
448 	struct inpcb *last;
449 	struct ip *ip;
450 	struct mbuf *n;
451 	int off;
452 {
453 	struct sockaddr *append_sa;
454 	struct mbuf *opts = 0;
455 
456 #if defined(IPSEC) || defined(FAST_IPSEC)
457 	/* check AH/ESP integrity. */
458 	if (ipsec4_in_reject(n, last)) {
459 #ifdef IPSEC
460 		ipsecstat.in_polvio++;
461 #endif /*IPSEC*/
462 		m_freem(n);
463 		return;
464 	}
465 #endif /*IPSEC || FAST_IPSEC*/
466 #ifdef MAC
467 	if (mac_check_inpcb_deliver(last, n) != 0) {
468 		m_freem(n);
469 		return;
470 	}
471 #endif
472 	if (last->inp_flags & INP_CONTROLOPTS ||
473 	    last->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) {
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, n, &opts);
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 struct inpcb *
515 udp_notify(inp, errno)
516 	register 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 	return inp;
523 }
524 
525 void
526 udp_ctlinput(cmd, sa, vip)
527 	int cmd;
528 	struct sockaddr *sa;
529 	void *vip;
530 {
531 	struct ip *ip = vip;
532 	struct udphdr *uh;
533 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
534         struct in_addr faddr;
535 	struct inpcb *inp;
536 	int s;
537 
538 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
539 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
540         	return;
541 
542 	/*
543 	 * Redirects don't need to be handled up here.
544 	 */
545 	if (PRC_IS_REDIRECT(cmd))
546 		return;
547 	/*
548 	 * Hostdead is ugly because it goes linearly through all PCBs.
549 	 * XXX: We never get this from ICMP, otherwise it makes an
550 	 * excellent DoS attack on machines with many connections.
551 	 */
552 	if (cmd == PRC_HOSTDEAD)
553 		ip = 0;
554 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
555 		return;
556 	if (ip) {
557 		s = splnet();
558 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
559 		INP_INFO_RLOCK(&udbinfo);
560 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
561                     ip->ip_src, uh->uh_sport, 0, NULL);
562 		if (inp != NULL) {
563 			INP_LOCK(inp);
564 			if (inp->inp_socket != NULL) {
565 				(*notify)(inp, inetctlerrmap[cmd]);
566 			}
567 			INP_UNLOCK(inp);
568 		}
569 		INP_INFO_RUNLOCK(&udbinfo);
570 		splx(s);
571 	} else
572 		in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
573 }
574 
575 static int
576 udp_pcblist(SYSCTL_HANDLER_ARGS)
577 {
578 	int error, i, n, s;
579 	struct inpcb *inp, **inp_list;
580 	inp_gen_t gencnt;
581 	struct xinpgen xig;
582 
583 	/*
584 	 * The process of preparing the TCB list is too time-consuming and
585 	 * resource-intensive to repeat twice on every request.
586 	 */
587 	if (req->oldptr == 0) {
588 		n = udbinfo.ipi_count;
589 		req->oldidx = 2 * (sizeof xig)
590 			+ (n + n/8) * sizeof(struct xinpcb);
591 		return 0;
592 	}
593 
594 	if (req->newptr != 0)
595 		return EPERM;
596 
597 	/*
598 	 * OK, now we're committed to doing something.
599 	 */
600 	s = splnet();
601 	INP_INFO_RLOCK(&udbinfo);
602 	gencnt = udbinfo.ipi_gencnt;
603 	n = udbinfo.ipi_count;
604 	INP_INFO_RUNLOCK(&udbinfo);
605 	splx(s);
606 
607 	error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
608 		+ n * sizeof(struct xinpcb));
609 	if (error != 0)
610 		return (error);
611 
612 	xig.xig_len = sizeof xig;
613 	xig.xig_count = n;
614 	xig.xig_gen = gencnt;
615 	xig.xig_sogen = so_gencnt;
616 	error = SYSCTL_OUT(req, &xig, sizeof xig);
617 	if (error)
618 		return error;
619 
620 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
621 	if (inp_list == 0)
622 		return ENOMEM;
623 
624 	s = splnet();
625 	INP_INFO_RLOCK(&udbinfo);
626 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
627 	     inp = LIST_NEXT(inp, inp_list)) {
628 		INP_LOCK(inp);
629 		if (inp->inp_gencnt <= gencnt &&
630 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
631 			inp_list[i++] = inp;
632 		INP_UNLOCK(inp);
633 	}
634 	INP_INFO_RUNLOCK(&udbinfo);
635 	splx(s);
636 	n = i;
637 
638 	error = 0;
639 	for (i = 0; i < n; i++) {
640 		inp = inp_list[i];
641 		if (inp->inp_gencnt <= gencnt) {
642 			struct xinpcb xi;
643 			xi.xi_len = sizeof xi;
644 			/* XXX should avoid extra copy */
645 			bcopy(inp, &xi.xi_inp, sizeof *inp);
646 			if (inp->inp_socket)
647 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
648 			xi.xi_inp.inp_gencnt = inp->inp_gencnt;
649 			error = SYSCTL_OUT(req, &xi, sizeof xi);
650 		}
651 	}
652 	if (!error) {
653 		/*
654 		 * Give the user an updated idea of our state.
655 		 * If the generation differs from what we told
656 		 * her before, she knows that something happened
657 		 * while we were processing this request, and it
658 		 * might be necessary to retry.
659 		 */
660 		s = splnet();
661 		INP_INFO_RLOCK(&udbinfo);
662 		xig.xig_gen = udbinfo.ipi_gencnt;
663 		xig.xig_sogen = so_gencnt;
664 		xig.xig_count = udbinfo.ipi_count;
665 		INP_INFO_RUNLOCK(&udbinfo);
666 		splx(s);
667 		error = SYSCTL_OUT(req, &xig, sizeof xig);
668 	}
669 	free(inp_list, M_TEMP);
670 	return error;
671 }
672 
673 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
674 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
675 
676 static int
677 udp_getcred(SYSCTL_HANDLER_ARGS)
678 {
679 	struct xucred xuc;
680 	struct sockaddr_in addrs[2];
681 	struct inpcb *inp;
682 	int error, s;
683 
684 	error = suser_cred(req->td->td_ucred, PRISON_ROOT);
685 	if (error)
686 		return (error);
687 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
688 	if (error)
689 		return (error);
690 	s = splnet();
691 	INP_INFO_RLOCK(&udbinfo);
692 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
693 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
694 	if (inp == NULL || inp->inp_socket == NULL) {
695 		error = ENOENT;
696 		goto out;
697 	}
698 	error = cr_canseesocket(req->td->td_ucred, inp->inp_socket);
699 	if (error)
700 		goto out;
701 	cru2x(inp->inp_socket->so_cred, &xuc);
702 out:
703 	INP_INFO_RUNLOCK(&udbinfo);
704 	splx(s);
705 	if (error == 0)
706 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
707 	return (error);
708 }
709 
710 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
711     CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
712     udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
713 
714 static int
715 udp_output(inp, m, addr, control, td)
716 	register struct inpcb *inp;
717 	struct mbuf *m;
718 	struct sockaddr *addr;
719 	struct mbuf *control;
720 	struct thread *td;
721 {
722 	register struct udpiphdr *ui;
723 	register int len = m->m_pkthdr.len;
724 	struct in_addr faddr, laddr;
725 	struct cmsghdr *cm;
726 	struct sockaddr_in *sin, src;
727 	int error = 0;
728 	int ipflags;
729 	u_short fport, lport;
730 
731 	INP_LOCK_ASSERT(inp);
732 #ifdef MAC
733 	mac_create_mbuf_from_socket(inp->inp_socket, m);
734 #endif
735 
736 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
737 		error = EMSGSIZE;
738 		if (control)
739 			m_freem(control);
740 		goto release;
741 	}
742 
743 	src.sin_addr.s_addr = INADDR_ANY;
744 	if (control != NULL) {
745 		/*
746 		 * XXX: Currently, we assume all the optional information
747 		 * is stored in a single mbuf.
748 		 */
749 		if (control->m_next) {
750 			error = EINVAL;
751 			m_freem(control);
752 			goto release;
753 		}
754 		for (; control->m_len > 0;
755 		    control->m_data += CMSG_ALIGN(cm->cmsg_len),
756 		    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
757 			cm = mtod(control, struct cmsghdr *);
758 			if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 ||
759 			    cm->cmsg_len > control->m_len) {
760 				error = EINVAL;
761 				break;
762 			}
763 			if (cm->cmsg_level != IPPROTO_IP)
764 				continue;
765 
766 			switch (cm->cmsg_type) {
767 			case IP_SENDSRCADDR:
768 				if (cm->cmsg_len !=
769 				    CMSG_LEN(sizeof(struct in_addr))) {
770 					error = EINVAL;
771 					break;
772 				}
773 				bzero(&src, sizeof(src));
774 				src.sin_family = AF_INET;
775 				src.sin_len = sizeof(src);
776 				src.sin_port = inp->inp_lport;
777 				src.sin_addr = *(struct in_addr *)CMSG_DATA(cm);
778 				break;
779 			default:
780 				error = ENOPROTOOPT;
781 				break;
782 			}
783 			if (error)
784 				break;
785 		}
786 		m_freem(control);
787 	}
788 	if (error)
789 		goto release;
790 	laddr = inp->inp_laddr;
791 	lport = inp->inp_lport;
792 	if (src.sin_addr.s_addr != INADDR_ANY) {
793 		if (lport == 0) {
794 			error = EINVAL;
795 			goto release;
796 		}
797 		error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
798 		    &laddr.s_addr, &lport, td);
799 		if (error)
800 			goto release;
801 	}
802 
803 	if (addr) {
804 		sin = (struct sockaddr_in *)addr;
805 		if (td && jailed(td->td_ucred))
806 			prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
807 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
808 			error = EISCONN;
809 			goto release;
810 		}
811 		error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport,
812 		    &faddr.s_addr, &fport, NULL, td);
813 		if (error)
814 			goto release;
815 
816 		/* Commit the local port if newly assigned. */
817 		if (inp->inp_laddr.s_addr == INADDR_ANY &&
818 		    inp->inp_lport == 0) {
819 			inp->inp_lport = lport;
820 			if (in_pcbinshash(inp) != 0) {
821 				inp->inp_lport = 0;
822 				error = EAGAIN;
823 				goto release;
824 			}
825 			inp->inp_flags |= INP_ANONPORT;
826 		}
827 	} else {
828 		faddr = inp->inp_faddr;
829 		fport = inp->inp_fport;
830 		if (faddr.s_addr == INADDR_ANY) {
831 			error = ENOTCONN;
832 			goto release;
833 		}
834 	}
835 	/*
836 	 * Calculate data length and get a mbuf
837 	 * for UDP and IP headers.
838 	 */
839 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
840 	if (m == 0) {
841 		error = ENOBUFS;
842 		goto release;
843 	}
844 
845 	/*
846 	 * Fill in mbuf with extended UDP header
847 	 * and addresses and length put into network format.
848 	 */
849 	ui = mtod(m, struct udpiphdr *);
850 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
851 	ui->ui_pr = IPPROTO_UDP;
852 	ui->ui_src = laddr;
853 	ui->ui_dst = faddr;
854 	ui->ui_sport = lport;
855 	ui->ui_dport = fport;
856 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
857 
858 	ipflags = inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST);
859 	if (inp->inp_flags & INP_ONESBCAST)
860 		ipflags |= IP_SENDONES;
861 
862 	/*
863 	 * Set up checksum and output datagram.
864 	 */
865 	if (udpcksum) {
866 		if (inp->inp_flags & INP_ONESBCAST)
867 			faddr.s_addr = INADDR_BROADCAST;
868 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
869 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
870 		m->m_pkthdr.csum_flags = CSUM_UDP;
871 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
872 	} else {
873 		ui->ui_sum = 0;
874 	}
875 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
876 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
877 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
878 	udpstat.udps_opackets++;
879 
880 	error = ip_output(m, inp->inp_options, NULL, ipflags,
881 	    inp->inp_moptions, inp);
882 	return (error);
883 
884 release:
885 	m_freem(m);
886 	return (error);
887 }
888 
889 u_long	udp_sendspace = 9216;		/* really max datagram size */
890 					/* 40 1K datagrams */
891 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
892     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
893 
894 u_long	udp_recvspace = 40 * (1024 +
895 #ifdef INET6
896 				      sizeof(struct sockaddr_in6)
897 #else
898 				      sizeof(struct sockaddr_in)
899 #endif
900 				      );
901 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
902     &udp_recvspace, 0, "Maximum space for incoming UDP datagrams");
903 
904 static int
905 udp_abort(struct socket *so)
906 {
907 	struct inpcb *inp;
908 	int s;
909 
910 	INP_INFO_WLOCK(&udbinfo);
911 	inp = sotoinpcb(so);
912 	if (inp == 0) {
913 		INP_INFO_WUNLOCK(&udbinfo);
914 		return EINVAL;	/* ??? possible? panic instead? */
915 	}
916 	INP_LOCK(inp);
917 	soisdisconnected(so);
918 	s = splnet();
919 	in_pcbdetach(inp);
920 	INP_INFO_WUNLOCK(&udbinfo);
921 	splx(s);
922 	return 0;
923 }
924 
925 static int
926 udp_attach(struct socket *so, int proto, struct thread *td)
927 {
928 	struct inpcb *inp;
929 	int s, error;
930 
931 	INP_INFO_WLOCK(&udbinfo);
932 	inp = sotoinpcb(so);
933 	if (inp != 0) {
934 		INP_INFO_WUNLOCK(&udbinfo);
935 		return EINVAL;
936 	}
937 	error = soreserve(so, udp_sendspace, udp_recvspace);
938 	if (error) {
939 		INP_INFO_WUNLOCK(&udbinfo);
940 		return error;
941 	}
942 	s = splnet();
943 	error = in_pcballoc(so, &udbinfo, td, "udpinp");
944 	splx(s);
945 	if (error) {
946 		INP_INFO_WUNLOCK(&udbinfo);
947 		return error;
948 	}
949 
950 	inp = (struct inpcb *)so->so_pcb;
951 	INP_LOCK(inp);
952 	INP_INFO_WUNLOCK(&udbinfo);
953 	inp->inp_vflag |= INP_IPV4;
954 	inp->inp_ip_ttl = ip_defttl;
955 	INP_UNLOCK(inp);
956 	return 0;
957 }
958 
959 static int
960 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
961 {
962 	struct inpcb *inp;
963 	int s, error;
964 
965 	INP_INFO_WLOCK(&udbinfo);
966 	inp = sotoinpcb(so);
967 	if (inp == 0) {
968 		INP_INFO_WUNLOCK(&udbinfo);
969 		return EINVAL;
970 	}
971 	INP_LOCK(inp);
972 	s = splnet();
973 	error = in_pcbbind(inp, nam, td);
974 	splx(s);
975 	INP_UNLOCK(inp);
976 	INP_INFO_WUNLOCK(&udbinfo);
977 	return error;
978 }
979 
980 static int
981 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
982 {
983 	struct inpcb *inp;
984 	int s, error;
985 	struct sockaddr_in *sin;
986 
987 	INP_INFO_WLOCK(&udbinfo);
988 	inp = sotoinpcb(so);
989 	if (inp == 0) {
990 		INP_INFO_WUNLOCK(&udbinfo);
991 		return EINVAL;
992 	}
993 	INP_LOCK(inp);
994 	if (inp->inp_faddr.s_addr != INADDR_ANY) {
995 		INP_UNLOCK(inp);
996 		INP_INFO_WUNLOCK(&udbinfo);
997 		return EISCONN;
998 	}
999 	s = splnet();
1000 	sin = (struct sockaddr_in *)nam;
1001 	if (td && jailed(td->td_ucred))
1002 		prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
1003 	error = in_pcbconnect(inp, nam, td);
1004 	splx(s);
1005 	if (error == 0)
1006 		soisconnected(so);
1007 	INP_UNLOCK(inp);
1008 	INP_INFO_WUNLOCK(&udbinfo);
1009 	return error;
1010 }
1011 
1012 static int
1013 udp_detach(struct socket *so)
1014 {
1015 	struct inpcb *inp;
1016 	int s;
1017 
1018 	INP_INFO_WLOCK(&udbinfo);
1019 	inp = sotoinpcb(so);
1020 	if (inp == 0) {
1021 		INP_INFO_WUNLOCK(&udbinfo);
1022 		return EINVAL;
1023 	}
1024 	INP_LOCK(inp);
1025 	s = splnet();
1026 	in_pcbdetach(inp);
1027 	INP_INFO_WUNLOCK(&udbinfo);
1028 	splx(s);
1029 	return 0;
1030 }
1031 
1032 static int
1033 udp_disconnect(struct socket *so)
1034 {
1035 	struct inpcb *inp;
1036 	int s;
1037 
1038 	INP_INFO_WLOCK(&udbinfo);
1039 	inp = sotoinpcb(so);
1040 	if (inp == 0) {
1041 		INP_INFO_WUNLOCK(&udbinfo);
1042 		return EINVAL;
1043 	}
1044 	INP_LOCK(inp);
1045 	if (inp->inp_faddr.s_addr == INADDR_ANY) {
1046 		INP_INFO_WUNLOCK(&udbinfo);
1047 		INP_UNLOCK(inp);
1048 		return ENOTCONN;
1049 	}
1050 
1051 	s = splnet();
1052 	in_pcbdisconnect(inp);
1053 	inp->inp_laddr.s_addr = INADDR_ANY;
1054 	INP_UNLOCK(inp);
1055 	INP_INFO_WUNLOCK(&udbinfo);
1056 	splx(s);
1057 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1058 	return 0;
1059 }
1060 
1061 static int
1062 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1063 	    struct mbuf *control, struct thread *td)
1064 {
1065 	struct inpcb *inp;
1066 	int ret;
1067 
1068 	INP_INFO_WLOCK(&udbinfo);
1069 	inp = sotoinpcb(so);
1070 	if (inp == 0) {
1071 		INP_INFO_WUNLOCK(&udbinfo);
1072 		m_freem(m);
1073 		return EINVAL;
1074 	}
1075 	INP_LOCK(inp);
1076 	ret = udp_output(inp, m, addr, control, td);
1077 	INP_UNLOCK(inp);
1078 	INP_INFO_WUNLOCK(&udbinfo);
1079 	return ret;
1080 }
1081 
1082 int
1083 udp_shutdown(struct socket *so)
1084 {
1085 	struct inpcb *inp;
1086 
1087 	INP_INFO_RLOCK(&udbinfo);
1088 	inp = sotoinpcb(so);
1089 	if (inp == 0) {
1090 		INP_INFO_RUNLOCK(&udbinfo);
1091 		return EINVAL;
1092 	}
1093 	INP_LOCK(inp);
1094 	INP_INFO_RUNLOCK(&udbinfo);
1095 	socantsendmore(so);
1096 	INP_UNLOCK(inp);
1097 	return 0;
1098 }
1099 
1100 /*
1101  * This is the wrapper function for in_setsockaddr.  We just pass down
1102  * the pcbinfo for in_setsockaddr to lock.  We don't want to do the locking
1103  * here because in_setsockaddr will call malloc and might block.
1104  */
1105 static int
1106 udp_sockaddr(struct socket *so, struct sockaddr **nam)
1107 {
1108 	return (in_setsockaddr(so, nam, &udbinfo));
1109 }
1110 
1111 /*
1112  * This is the wrapper function for in_setpeeraddr.  We just pass down
1113  * the pcbinfo for in_setpeeraddr to lock.
1114  */
1115 static int
1116 udp_peeraddr(struct socket *so, struct sockaddr **nam)
1117 {
1118 	return (in_setpeeraddr(so, nam, &udbinfo));
1119 }
1120 
1121 struct pr_usrreqs udp_usrreqs = {
1122 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
1123 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
1124 	pru_listen_notsupp, udp_peeraddr, pru_rcvd_notsupp,
1125 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
1126 	udp_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
1127 };
1128