xref: /freebsd/sys/netinet/udp_usrreq.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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 struct	inpcbhead udb;		/* from udp_var.h */
112 #define	udb6	udb  /* for KAME src sync over BSD*'s */
113 struct	inpcbinfo udbinfo;
114 
115 #ifndef UDBHASHSIZE
116 #define UDBHASHSIZE 16
117 #endif
118 
119 struct	udpstat udpstat;	/* from udp_var.h */
120 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
121     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
122 
123 static struct	sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
124 #ifdef INET6
125 struct udp_in6 {
126 	struct sockaddr_in6	uin6_sin;
127 	u_char			uin6_init_done : 1;
128 } udp_in6 = {
129 	{ sizeof(udp_in6.uin6_sin), AF_INET6 },
130 	0
131 };
132 struct udp_ip6 {
133 	struct ip6_hdr		uip6_ip6;
134 	u_char			uip6_init_done : 1;
135 } udp_ip6;
136 #endif /* INET6 */
137 
138 static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
139 		int off);
140 #ifdef INET6
141 static void ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip);
142 #endif
143 
144 static int udp_detach(struct socket *so);
145 static	int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
146 		struct mbuf *, struct thread *);
147 
148 void
149 udp_init()
150 {
151 	INP_INFO_LOCK_INIT(&udbinfo, "udp");
152 	LIST_INIT(&udb);
153 	udbinfo.listhead = &udb;
154 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
155 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
156 					&udbinfo.porthashmask);
157 	udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL,
158 	    NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
159 	uma_zone_set_max(udbinfo.ipi_zone, maxsockets);
160 }
161 
162 void
163 udp_input(m, off)
164 	register struct mbuf *m;
165 	int off;
166 {
167 	int iphlen = off;
168 	register struct ip *ip;
169 	register struct udphdr *uh;
170 	register struct inpcb *inp;
171 	struct mbuf *opts = 0;
172 	int len;
173 	struct ip save_ip;
174 
175 	udpstat.udps_ipackets++;
176 
177 	/*
178 	 * Strip IP options, if any; should skip this,
179 	 * make available to user, and use on returned packets,
180 	 * but we don't yet have a way to check the checksum
181 	 * with options still present.
182 	 */
183 	if (iphlen > sizeof (struct ip)) {
184 		ip_stripoptions(m, (struct mbuf *)0);
185 		iphlen = sizeof(struct ip);
186 	}
187 
188 	/*
189 	 * Get IP and UDP header together in first mbuf.
190 	 */
191 	ip = mtod(m, struct ip *);
192 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
193 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
194 			udpstat.udps_hdrops++;
195 			return;
196 		}
197 		ip = mtod(m, struct ip *);
198 	}
199 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
200 
201 	/* destination port of 0 is illegal, based on RFC768. */
202 	if (uh->uh_dport == 0)
203 		goto badunlocked;
204 
205 	/*
206 	 * Construct sockaddr format source address.
207 	 * Stuff source address and datagram in user buffer.
208 	 */
209 	udp_in.sin_port = uh->uh_sport;
210 	udp_in.sin_addr = ip->ip_src;
211 #ifdef INET6
212 	udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
213 #endif
214 
215 	/*
216 	 * Make mbuf data length reflect UDP length.
217 	 * If not enough data to reflect UDP length, drop.
218 	 */
219 	len = ntohs((u_short)uh->uh_ulen);
220 	if (ip->ip_len != len) {
221 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
222 			udpstat.udps_badlen++;
223 			goto badunlocked;
224 		}
225 		m_adj(m, len - ip->ip_len);
226 		/* ip->ip_len = len; */
227 	}
228 	/*
229 	 * Save a copy of the IP header in case we want restore it
230 	 * for sending an ICMP error message in response.
231 	 */
232 	if (!blackhole)
233 		save_ip = *ip;
234 
235 	/*
236 	 * Checksum extended UDP header and data.
237 	 */
238 	if (uh->uh_sum) {
239 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
240 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
241 				uh->uh_sum = m->m_pkthdr.csum_data;
242 			else
243 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
244 				    ip->ip_dst.s_addr, htonl((u_short)len +
245 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
246 			uh->uh_sum ^= 0xffff;
247 		} else {
248 			char b[9];
249 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
250 			bzero(((struct ipovly *)ip)->ih_x1, 9);
251 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
252 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
253 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
254 		}
255 		if (uh->uh_sum) {
256 			udpstat.udps_badsum++;
257 			m_freem(m);
258 			return;
259 		}
260 	} else
261 		udpstat.udps_nosum++;
262 
263 	INP_INFO_RLOCK(&udbinfo);
264 
265 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
266 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
267 		struct inpcb *last;
268 		/*
269 		 * Deliver a multicast or broadcast datagram to *all* sockets
270 		 * for which the local and remote addresses and ports match
271 		 * those of the incoming datagram.  This allows more than
272 		 * one process to receive multi/broadcasts on the same port.
273 		 * (This really ought to be done for unicast datagrams as
274 		 * well, but that would cause problems with existing
275 		 * applications that open both address-specific sockets and
276 		 * a wildcard socket listening to the same port -- they would
277 		 * end up receiving duplicates of every unicast datagram.
278 		 * Those applications open the multiple sockets to overcome an
279 		 * inadequacy of the UDP socket interface, but for backwards
280 		 * compatibility we avoid the problem here rather than
281 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
282 		 */
283 
284 		/*
285 		 * Locate pcb(s) for datagram.
286 		 * (Algorithm copied from raw_intr().)
287 		 */
288 		last = NULL;
289 		LIST_FOREACH(inp, &udb, inp_list) {
290 			INP_LOCK(inp);
291 			if (inp->inp_lport != uh->uh_dport) {
292 		docontinue:
293 				INP_UNLOCK(inp);
294 				continue;
295 			}
296 #ifdef INET6
297 			if ((inp->inp_vflag & INP_IPV4) == 0)
298 				goto docontinue;
299 #endif
300 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
301 				if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
302 					goto docontinue;
303 			}
304 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
305 				if (inp->inp_faddr.s_addr !=
306 				    ip->ip_src.s_addr ||
307 				    inp->inp_fport != uh->uh_sport)
308 					goto docontinue;
309 			}
310 
311 			if (last != NULL) {
312 				struct mbuf *n;
313 
314 				n = m_copy(m, 0, M_COPYALL);
315 				if (n != NULL)
316 					udp_append(last, ip, n,
317 						   iphlen +
318 						   sizeof(struct udphdr));
319 				INP_UNLOCK(last);
320 			}
321 			last = inp;
322 			/*
323 			 * Don't look for additional matches if this one does
324 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
325 			 * socket options set.  This heuristic avoids searching
326 			 * through all pcbs in the common case of a non-shared
327 			 * port.  It * assumes that an application will never
328 			 * clear these options after setting them.
329 			 */
330 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
331 				break;
332 		}
333 
334 		if (last == NULL) {
335 			/*
336 			 * No matching pcb found; discard datagram.
337 			 * (No need to send an ICMP Port Unreachable
338 			 * for a broadcast or multicast datgram.)
339 			 */
340 			udpstat.udps_noportbcast++;
341 			goto badheadlocked;
342 		}
343 		INP_INFO_RUNLOCK(&udbinfo);
344 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
345 		INP_UNLOCK(last);
346 		return;
347 	}
348 	/*
349 	 * Locate pcb for datagram.
350 	 */
351 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
352 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
353 	if (inp == NULL) {
354 		if (log_in_vain) {
355 			char buf[4*sizeof "123"];
356 
357 			strcpy(buf, inet_ntoa(ip->ip_dst));
358 			log(LOG_INFO,
359 			    "Connection attempt to UDP %s:%d from %s:%d\n",
360 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
361 			    ntohs(uh->uh_sport));
362 		}
363 		udpstat.udps_noport++;
364 		if (m->m_flags & (M_BCAST | M_MCAST)) {
365 			udpstat.udps_noportbcast++;
366 			goto badheadlocked;
367 		}
368 		if (blackhole)
369 			goto badheadlocked;
370 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
371 			goto badheadlocked;
372 		*ip = save_ip;
373 		ip->ip_len += iphlen;
374 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
375 		INP_INFO_RUNLOCK(&udbinfo);
376 		return;
377 	}
378 	INP_LOCK(inp);
379 	INP_INFO_RUNLOCK(&udbinfo);
380 	udp_append(inp, ip, m, iphlen + sizeof(struct udphdr));
381 	INP_UNLOCK(inp);
382 	return;
383 
384 badheadlocked:
385 	INP_INFO_RUNLOCK(&udbinfo);
386 	if (inp)
387 		INP_UNLOCK(inp);
388 badunlocked:
389 	m_freem(m);
390 	if (opts)
391 		m_freem(opts);
392 	return;
393 }
394 
395 #ifdef INET6
396 static void
397 ip_2_ip6_hdr(ip6, ip)
398 	struct ip6_hdr *ip6;
399 	struct ip *ip;
400 {
401 	bzero(ip6, sizeof(*ip6));
402 
403 	ip6->ip6_vfc = IPV6_VERSION;
404 	ip6->ip6_plen = ip->ip_len;
405 	ip6->ip6_nxt = ip->ip_p;
406 	ip6->ip6_hlim = ip->ip_ttl;
407 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
408 		IPV6_ADDR_INT32_SMP;
409 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
410 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
411 }
412 #endif
413 
414 /*
415  * subroutine of udp_input(), mainly for source code readability.
416  * caller must properly init udp_ip6 and udp_in6 beforehand.
417  */
418 static void
419 udp_append(last, ip, n, off)
420 	struct inpcb *last;
421 	struct ip *ip;
422 	struct mbuf *n;
423 	int off;
424 {
425 	struct sockaddr *append_sa;
426 	struct mbuf *opts = 0;
427 
428 #ifdef IPSEC
429 	/* check AH/ESP integrity. */
430 	if (ipsec4_in_reject_so(n, last->inp_socket)) {
431 		ipsecstat.in_polvio++;
432 		m_freem(n);
433 		return;
434 	}
435 #endif /*IPSEC*/
436 #ifdef FAST_IPSEC
437 	/* check AH/ESP integrity. */
438 	if (ipsec4_in_reject(n, last)) {
439 		m_freem(n);
440 		return;
441 	}
442 #endif /*FAST_IPSEC*/
443 #ifdef MAC
444 	if (mac_check_socket_deliver(last->inp_socket, n) != 0) {
445 		m_freem(n);
446 		return;
447 	}
448 #endif
449 	if (last->inp_flags & INP_CONTROLOPTS ||
450 	    last->inp_socket->so_options & SO_TIMESTAMP) {
451 #ifdef INET6
452 		if (last->inp_vflag & INP_IPV6) {
453 			int savedflags;
454 
455 			if (udp_ip6.uip6_init_done == 0) {
456 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
457 				udp_ip6.uip6_init_done = 1;
458 			}
459 			savedflags = last->inp_flags;
460 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
461 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
462 			last->inp_flags = savedflags;
463 		} else
464 #endif
465 		ip_savecontrol(last, &opts, ip, n);
466 	}
467 #ifdef INET6
468 	if (last->inp_vflag & INP_IPV6) {
469 		if (udp_in6.uin6_init_done == 0) {
470 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
471 			udp_in6.uin6_init_done = 1;
472 		}
473 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
474 	} else
475 #endif
476 	append_sa = (struct sockaddr *)&udp_in;
477 	m_adj(n, off);
478 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
479 		m_freem(n);
480 		if (opts)
481 			m_freem(opts);
482 		udpstat.udps_fullsock++;
483 	} else
484 		sorwakeup(last->inp_socket);
485 }
486 
487 /*
488  * Notify a udp user of an asynchronous error;
489  * just wake up so that he can collect error status.
490  */
491 struct inpcb *
492 udp_notify(inp, errno)
493 	register struct inpcb *inp;
494 	int errno;
495 {
496 	inp->inp_socket->so_error = errno;
497 	sorwakeup(inp->inp_socket);
498 	sowwakeup(inp->inp_socket);
499 	return inp;
500 }
501 
502 void
503 udp_ctlinput(cmd, sa, vip)
504 	int cmd;
505 	struct sockaddr *sa;
506 	void *vip;
507 {
508 	struct ip *ip = vip;
509 	struct udphdr *uh;
510 	struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
511         struct in_addr faddr;
512 	struct inpcb *inp;
513 	int s;
514 
515 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
516 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
517         	return;
518 
519 	if (PRC_IS_REDIRECT(cmd)) {
520 		ip = 0;
521 		notify = in_rtchange;
522 	} else if (cmd == PRC_HOSTDEAD)
523 		ip = 0;
524 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
525 		return;
526 	if (ip) {
527 		s = splnet();
528 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
529 		INP_INFO_RLOCK(&udbinfo);
530 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
531                     ip->ip_src, uh->uh_sport, 0, NULL);
532 		if (inp != NULL) {
533 			INP_LOCK(inp);
534 			if (inp->inp_socket != NULL) {
535 				(*notify)(inp, inetctlerrmap[cmd]);
536 			}
537 			INP_UNLOCK(inp);
538 		}
539 		INP_INFO_RUNLOCK(&udbinfo);
540 		splx(s);
541 	} else
542 		in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
543 }
544 
545 static int
546 udp_pcblist(SYSCTL_HANDLER_ARGS)
547 {
548 	int error, i, n, s;
549 	struct inpcb *inp, **inp_list;
550 	inp_gen_t gencnt;
551 	struct xinpgen xig;
552 
553 	/*
554 	 * The process of preparing the TCB list is too time-consuming and
555 	 * resource-intensive to repeat twice on every request.
556 	 */
557 	if (req->oldptr == 0) {
558 		n = udbinfo.ipi_count;
559 		req->oldidx = 2 * (sizeof xig)
560 			+ (n + n/8) * sizeof(struct xinpcb);
561 		return 0;
562 	}
563 
564 	if (req->newptr != 0)
565 		return EPERM;
566 
567 	/*
568 	 * OK, now we're committed to doing something.
569 	 */
570 	s = splnet();
571 	INP_INFO_RLOCK(&udbinfo);
572 	gencnt = udbinfo.ipi_gencnt;
573 	n = udbinfo.ipi_count;
574 	INP_INFO_RUNLOCK(&udbinfo);
575 	splx(s);
576 
577 	sysctl_wire_old_buffer(req, 2 * (sizeof xig)
578 		+ n * sizeof(struct xinpcb));
579 
580 	xig.xig_len = sizeof xig;
581 	xig.xig_count = n;
582 	xig.xig_gen = gencnt;
583 	xig.xig_sogen = so_gencnt;
584 	error = SYSCTL_OUT(req, &xig, sizeof xig);
585 	if (error)
586 		return error;
587 
588 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
589 	if (inp_list == 0)
590 		return ENOMEM;
591 
592 	s = splnet();
593 	INP_INFO_RLOCK(&udbinfo);
594 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
595 	     inp = LIST_NEXT(inp, inp_list)) {
596 		INP_LOCK(inp);
597 		if (inp->inp_gencnt <= gencnt &&
598 		    cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
599 			inp_list[i++] = inp;
600 		INP_UNLOCK(inp);
601 	}
602 	INP_INFO_RUNLOCK(&udbinfo);
603 	splx(s);
604 	n = i;
605 
606 	error = 0;
607 	for (i = 0; i < n; i++) {
608 		inp = inp_list[i];
609 		if (inp->inp_gencnt <= gencnt) {
610 			struct xinpcb xi;
611 			xi.xi_len = sizeof xi;
612 			/* XXX should avoid extra copy */
613 			bcopy(inp, &xi.xi_inp, sizeof *inp);
614 			if (inp->inp_socket)
615 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
616 			xi.xi_inp.inp_gencnt = inp->inp_gencnt;
617 			error = SYSCTL_OUT(req, &xi, sizeof xi);
618 		}
619 	}
620 	if (!error) {
621 		/*
622 		 * Give the user an updated idea of our state.
623 		 * If the generation differs from what we told
624 		 * her before, she knows that something happened
625 		 * while we were processing this request, and it
626 		 * might be necessary to retry.
627 		 */
628 		s = splnet();
629 		INP_INFO_RLOCK(&udbinfo);
630 		xig.xig_gen = udbinfo.ipi_gencnt;
631 		xig.xig_sogen = so_gencnt;
632 		xig.xig_count = udbinfo.ipi_count;
633 		INP_INFO_RUNLOCK(&udbinfo);
634 		splx(s);
635 		error = SYSCTL_OUT(req, &xig, sizeof xig);
636 	}
637 	free(inp_list, M_TEMP);
638 	return error;
639 }
640 
641 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
642 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
643 
644 static int
645 udp_getcred(SYSCTL_HANDLER_ARGS)
646 {
647 	struct xucred xuc;
648 	struct sockaddr_in addrs[2];
649 	struct inpcb *inp;
650 	int error, s;
651 
652 	error = suser_cred(req->td->td_ucred, PRISON_ROOT);
653 	if (error)
654 		return (error);
655 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
656 	if (error)
657 		return (error);
658 	s = splnet();
659 	INP_INFO_RLOCK(&udbinfo);
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 = cr_canseesocket(req->td->td_ucred, inp->inp_socket);
667 	if (error)
668 		goto out;
669 	cru2x(inp->inp_socket->so_cred, &xuc);
670 out:
671 	INP_INFO_RUNLOCK(&udbinfo);
672 	splx(s);
673 	if (error == 0)
674 		error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
675 	return (error);
676 }
677 
678 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
679     CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
680     udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
681 
682 static int
683 udp_output(inp, m, addr, control, td)
684 	register struct inpcb *inp;
685 	struct mbuf *m;
686 	struct sockaddr *addr;
687 	struct mbuf *control;
688 	struct thread *td;
689 {
690 	register struct udpiphdr *ui;
691 	register int len = m->m_pkthdr.len;
692 	struct in_addr faddr, laddr;
693 	struct cmsghdr *cm;
694 	struct sockaddr_in *sin, src;
695 	int error = 0;
696 	int ipflags;
697 	u_short fport, lport;
698 
699 #ifdef MAC
700 	mac_create_mbuf_from_socket(inp->inp_socket, m);
701 #endif
702 
703 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
704 		error = EMSGSIZE;
705 		if (control)
706 			m_freem(control);
707 		goto release;
708 	}
709 
710 	src.sin_addr.s_addr = INADDR_ANY;
711 	if (control != NULL) {
712 		/*
713 		 * XXX: Currently, we assume all the optional information
714 		 * is stored in a single mbuf.
715 		 */
716 		if (control->m_next) {
717 			error = EINVAL;
718 			m_freem(control);
719 			goto release;
720 		}
721 		for (; control->m_len > 0;
722 		    control->m_data += CMSG_ALIGN(cm->cmsg_len),
723 		    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
724 			cm = mtod(control, struct cmsghdr *);
725 			if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 ||
726 			    cm->cmsg_len > control->m_len) {
727 				error = EINVAL;
728 				break;
729 			}
730 			if (cm->cmsg_level != IPPROTO_IP)
731 				continue;
732 
733 			switch (cm->cmsg_type) {
734 			case IP_SENDSRCADDR:
735 				if (cm->cmsg_len !=
736 				    CMSG_LEN(sizeof(struct in_addr))) {
737 					error = EINVAL;
738 					break;
739 				}
740 				bzero(&src, sizeof(src));
741 				src.sin_family = AF_INET;
742 				src.sin_len = sizeof(src);
743 				src.sin_port = inp->inp_lport;
744 				src.sin_addr = *(struct in_addr *)CMSG_DATA(cm);
745 				break;
746 			default:
747 				error = ENOPROTOOPT;
748 				break;
749 			}
750 			if (error)
751 				break;
752 		}
753 		m_freem(control);
754 	}
755 	if (error)
756 		goto release;
757 	laddr = inp->inp_laddr;
758 	lport = inp->inp_lport;
759 	if (src.sin_addr.s_addr != INADDR_ANY) {
760 		if (lport == 0) {
761 			error = EINVAL;
762 			goto release;
763 		}
764 		error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
765 		    &laddr.s_addr, &lport, td);
766 		if (error)
767 			goto release;
768 	}
769 
770 	if (addr) {
771 		sin = (struct sockaddr_in *)addr;
772 		if (td && jailed(td->td_ucred))
773 			prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
774 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
775 			error = EISCONN;
776 			goto release;
777 		}
778 		error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport,
779 		    &faddr.s_addr, &fport, NULL, td);
780 		if (error)
781 			goto release;
782 
783 		/* Commit the local port if newly assigned. */
784 		if (inp->inp_laddr.s_addr == INADDR_ANY &&
785 		    inp->inp_lport == 0) {
786 			inp->inp_lport = lport;
787 			if (in_pcbinshash(inp) != 0) {
788 				inp->inp_lport = 0;
789 				error = EAGAIN;
790 				goto release;
791 			}
792 			inp->inp_flags |= INP_ANONPORT;
793 		}
794 	} else {
795 		faddr = inp->inp_faddr;
796 		fport = inp->inp_fport;
797 		if (faddr.s_addr == INADDR_ANY) {
798 			error = ENOTCONN;
799 			goto release;
800 		}
801 	}
802 	/*
803 	 * Calculate data length and get a mbuf
804 	 * for UDP and IP headers.
805 	 */
806 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
807 	if (m == 0) {
808 		error = ENOBUFS;
809 		goto release;
810 	}
811 
812 	/*
813 	 * Fill in mbuf with extended UDP header
814 	 * and addresses and length put into network format.
815 	 */
816 	ui = mtod(m, struct udpiphdr *);
817 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
818 	ui->ui_pr = IPPROTO_UDP;
819 	ui->ui_src = laddr;
820 	ui->ui_dst = faddr;
821 	ui->ui_sport = lport;
822 	ui->ui_dport = fport;
823 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
824 
825 	ipflags = inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST);
826 	if (inp->inp_flags & INP_ONESBCAST)
827 		ipflags |= IP_SENDONES;
828 
829 	/*
830 	 * Set up checksum and output datagram.
831 	 */
832 	if (udpcksum) {
833 		if (inp->inp_flags & INP_ONESBCAST)
834 			faddr.s_addr = INADDR_BROADCAST;
835 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
836 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
837 		m->m_pkthdr.csum_flags = CSUM_UDP;
838 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
839 	} else {
840 		ui->ui_sum = 0;
841 	}
842 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
843 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
844 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
845 	udpstat.udps_opackets++;
846 
847 	error = ip_output(m, inp->inp_options, &inp->inp_route, ipflags,
848 	    inp->inp_moptions, inp);
849 	return (error);
850 
851 release:
852 	m_freem(m);
853 	return (error);
854 }
855 
856 u_long	udp_sendspace = 9216;		/* really max datagram size */
857 					/* 40 1K datagrams */
858 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
859     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
860 
861 u_long	udp_recvspace = 40 * (1024 +
862 #ifdef INET6
863 				      sizeof(struct sockaddr_in6)
864 #else
865 				      sizeof(struct sockaddr_in)
866 #endif
867 				      );
868 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
869     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
870 
871 static int
872 udp_abort(struct socket *so)
873 {
874 	struct inpcb *inp;
875 	int s;
876 
877 	INP_INFO_WLOCK(&udbinfo);
878 	inp = sotoinpcb(so);
879 	if (inp == 0) {
880 		INP_INFO_WUNLOCK(&udbinfo);
881 		return EINVAL;	/* ??? possible? panic instead? */
882 	}
883 	INP_LOCK(inp);
884 	soisdisconnected(so);
885 	s = splnet();
886 	in_pcbdetach(inp);
887 	INP_INFO_WUNLOCK(&udbinfo);
888 	splx(s);
889 	return 0;
890 }
891 
892 static int
893 udp_attach(struct socket *so, int proto, struct thread *td)
894 {
895 	struct inpcb *inp;
896 	int s, error;
897 
898 	INP_INFO_WLOCK(&udbinfo);
899 	inp = sotoinpcb(so);
900 	if (inp != 0) {
901 		INP_INFO_WUNLOCK(&udbinfo);
902 		return EINVAL;
903 	}
904 	error = soreserve(so, udp_sendspace, udp_recvspace);
905 	if (error) {
906 		INP_INFO_WUNLOCK(&udbinfo);
907 		return error;
908 	}
909 	s = splnet();
910 	error = in_pcballoc(so, &udbinfo, td);
911 	splx(s);
912 	if (error) {
913 		INP_INFO_WUNLOCK(&udbinfo);
914 		return error;
915 	}
916 
917 	inp = (struct inpcb *)so->so_pcb;
918 	INP_LOCK(inp);
919 	INP_INFO_WUNLOCK(&udbinfo);
920 	inp->inp_vflag |= INP_IPV4;
921 	inp->inp_ip_ttl = ip_defttl;
922 	INP_UNLOCK(inp);
923 	return 0;
924 }
925 
926 static int
927 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
928 {
929 	struct inpcb *inp;
930 	int s, error;
931 
932 	INP_INFO_WLOCK(&udbinfo);
933 	inp = sotoinpcb(so);
934 	if (inp == 0) {
935 		INP_INFO_WUNLOCK(&udbinfo);
936 		return EINVAL;
937 	}
938 	INP_LOCK(inp);
939 	s = splnet();
940 	error = in_pcbbind(inp, nam, td);
941 	splx(s);
942 	INP_UNLOCK(inp);
943 	INP_INFO_WUNLOCK(&udbinfo);
944 	return error;
945 }
946 
947 static int
948 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
949 {
950 	struct inpcb *inp;
951 	int s, error;
952 	struct sockaddr_in *sin;
953 
954 	INP_INFO_WLOCK(&udbinfo);
955 	inp = sotoinpcb(so);
956 	if (inp == 0) {
957 		INP_INFO_WUNLOCK(&udbinfo);
958 		return EINVAL;
959 	}
960 	INP_LOCK(inp);
961 	if (inp->inp_faddr.s_addr != INADDR_ANY) {
962 		INP_UNLOCK(inp);
963 		INP_INFO_WUNLOCK(&udbinfo);
964 		return EISCONN;
965 	}
966 	s = splnet();
967 	sin = (struct sockaddr_in *)nam;
968 	if (td && jailed(td->td_ucred))
969 		prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
970 	error = in_pcbconnect(inp, nam, td);
971 	splx(s);
972 	if (error == 0)
973 		soisconnected(so);
974 	INP_UNLOCK(inp);
975 	INP_INFO_WUNLOCK(&udbinfo);
976 	return error;
977 }
978 
979 static int
980 udp_detach(struct socket *so)
981 {
982 	struct inpcb *inp;
983 	int s;
984 
985 	INP_INFO_WLOCK(&udbinfo);
986 	inp = sotoinpcb(so);
987 	if (inp == 0) {
988 		INP_INFO_WUNLOCK(&udbinfo);
989 		return EINVAL;
990 	}
991 	INP_LOCK(inp);
992 	s = splnet();
993 	in_pcbdetach(inp);
994 	INP_INFO_WUNLOCK(&udbinfo);
995 	splx(s);
996 	return 0;
997 }
998 
999 static int
1000 udp_disconnect(struct socket *so)
1001 {
1002 	struct inpcb *inp;
1003 	int s;
1004 
1005 	INP_INFO_WLOCK(&udbinfo);
1006 	inp = sotoinpcb(so);
1007 	if (inp == 0) {
1008 		INP_INFO_WUNLOCK(&udbinfo);
1009 		return EINVAL;
1010 	}
1011 	INP_LOCK(inp);
1012 	if (inp->inp_faddr.s_addr == INADDR_ANY) {
1013 		INP_INFO_WUNLOCK(&udbinfo);
1014 		INP_UNLOCK(inp);
1015 		return ENOTCONN;
1016 	}
1017 
1018 	s = splnet();
1019 	in_pcbdisconnect(inp);
1020 	inp->inp_laddr.s_addr = INADDR_ANY;
1021 	INP_UNLOCK(inp);
1022 	INP_INFO_WUNLOCK(&udbinfo);
1023 	splx(s);
1024 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1025 	return 0;
1026 }
1027 
1028 static int
1029 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1030 	    struct mbuf *control, struct thread *td)
1031 {
1032 	struct inpcb *inp;
1033 	int ret;
1034 
1035 	INP_INFO_WLOCK(&udbinfo);
1036 	inp = sotoinpcb(so);
1037 	if (inp == 0) {
1038 		INP_INFO_WUNLOCK(&udbinfo);
1039 		m_freem(m);
1040 		return EINVAL;
1041 	}
1042 	INP_LOCK(inp);
1043 	ret = udp_output(inp, m, addr, control, td);
1044 	INP_UNLOCK(inp);
1045 	INP_INFO_WUNLOCK(&udbinfo);
1046 	return ret;
1047 }
1048 
1049 int
1050 udp_shutdown(struct socket *so)
1051 {
1052 	struct inpcb *inp;
1053 
1054 	INP_INFO_RLOCK(&udbinfo);
1055 	inp = sotoinpcb(so);
1056 	if (inp == 0) {
1057 		INP_INFO_RUNLOCK(&udbinfo);
1058 		return EINVAL;
1059 	}
1060 	INP_LOCK(inp);
1061 	INP_INFO_RUNLOCK(&udbinfo);
1062 	socantsendmore(so);
1063 	INP_UNLOCK(inp);
1064 	return 0;
1065 }
1066 
1067 /*
1068  * This is the wrapper function for in_setsockaddr.  We just pass down
1069  * the pcbinfo for in_setsockaddr to lock.  We don't want to do the locking
1070  * here because in_setsockaddr will call malloc and might block.
1071  */
1072 static int
1073 udp_sockaddr(struct socket *so, struct sockaddr **nam)
1074 {
1075 	return (in_setsockaddr(so, nam, &udbinfo));
1076 }
1077 
1078 /*
1079  * This is the wrapper function for in_setpeeraddr.  We just pass down
1080  * the pcbinfo for in_setpeeraddr to lock.
1081  */
1082 static int
1083 udp_peeraddr(struct socket *so, struct sockaddr **nam)
1084 {
1085 	return (in_setpeeraddr(so, nam, &udbinfo));
1086 }
1087 
1088 struct pr_usrreqs udp_usrreqs = {
1089 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
1090 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
1091 	pru_listen_notsupp, udp_peeraddr, pru_rcvd_notsupp,
1092 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
1093 	udp_sockaddr, sosend, soreceive, sopoll
1094 };
1095