xref: /dflybsd-src/sys/netinet/udp_usrreq.c (revision 6b08710ee11d58fba7e8982f046f86f888c48dfa)
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 1.8 2003/11/08 07:57:51 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 	udp_thread_init();
158 }
159 
160 /*
161  * Check multicast packets to make sure they are only sent to sockets with
162  * multicast memberships for the packet's destination address and arrival
163  * interface.  Multicast packets to multicast-unaware sockets are also
164  * disallowed.
165  *
166  * Returns 0 if the packet is acceptable, -1 if it is not.
167  */
168 static __inline
169 int
170 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
171 {
172 	int mshipno;
173 	struct ip_moptions *mopt;
174 
175 	if (strict_mcast_mship == 0 ||
176 	    !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
177 		return(0);
178 	}
179 	mopt = inp->inp_moptions;
180 	if (mopt == NULL)
181 		return(-1);
182 	for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
183 		if (ip->ip_dst.s_addr == mopt->imo_membership[mshipno]->inm_addr.s_addr &&
184 		    m->m_pkthdr.rcvif == mopt->imo_membership[mshipno]->inm_ifp) {
185 			return(0);
186 		}
187 	}
188 	return(-1);
189 }
190 
191 void
192 udp_input(m, off, proto)
193 	struct mbuf *m;
194 	int off, proto;
195 {
196 	int iphlen = off;
197 	struct ip *ip;
198 	struct udphdr *uh;
199 	struct inpcb *inp;
200 	struct mbuf *opts = 0;
201 	int len;
202 	struct ip save_ip;
203 	struct sockaddr *append_sa;
204 
205 	udpstat.udps_ipackets++;
206 
207 	/*
208 	 * Strip IP options, if any; should skip this,
209 	 * make available to user, and use on returned packets,
210 	 * but we don't yet have a way to check the checksum
211 	 * with options still present.
212 	 */
213 	if (iphlen > sizeof (struct ip)) {
214 		ip_stripoptions(m, (struct mbuf *)0);
215 		iphlen = sizeof(struct ip);
216 	}
217 
218 	/*
219 	 * Get IP and UDP header together in first mbuf.
220 	 */
221 	ip = mtod(m, struct ip *);
222 	if (m->m_len < iphlen + sizeof(struct udphdr)) {
223 		if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
224 			udpstat.udps_hdrops++;
225 			return;
226 		}
227 		ip = mtod(m, struct ip *);
228 	}
229 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
230 
231 	/* destination port of 0 is illegal, based on RFC768. */
232 	if (uh->uh_dport == 0)
233 		goto bad;
234 
235 	/*
236 	 * Make mbuf data length reflect UDP length.
237 	 * If not enough data to reflect UDP length, drop.
238 	 */
239 	len = ntohs((u_short)uh->uh_ulen);
240 	if (ip->ip_len != len) {
241 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
242 			udpstat.udps_badlen++;
243 			goto bad;
244 		}
245 		m_adj(m, len - ip->ip_len);
246 		/* ip->ip_len = len; */
247 	}
248 	/*
249 	 * Save a copy of the IP header in case we want restore it
250 	 * for sending an ICMP error message in response.
251 	 */
252 	save_ip = *ip;
253 
254 	/*
255 	 * Checksum extended UDP header and data.
256 	 */
257 	if (uh->uh_sum) {
258 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
259 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
260 				uh->uh_sum = m->m_pkthdr.csum_data;
261 			else
262 	                	uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
263 				    ip->ip_dst.s_addr, htonl((u_short)len +
264 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
265 			uh->uh_sum ^= 0xffff;
266 		} else {
267 			char b[9];
268 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
269 			bzero(((struct ipovly *)ip)->ih_x1, 9);
270 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
271 			uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
272 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
273 		}
274 		if (uh->uh_sum) {
275 			udpstat.udps_badsum++;
276 			m_freem(m);
277 			return;
278 		}
279 	} else
280 		udpstat.udps_nosum++;
281 
282 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
283 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
284 		struct inpcb *last;
285 		/*
286 		 * Deliver a multicast or broadcast datagram to *all* sockets
287 		 * for which the local and remote addresses and ports match
288 		 * those of the incoming datagram.  This allows more than
289 		 * one process to receive multi/broadcasts on the same port.
290 		 * (This really ought to be done for unicast datagrams as
291 		 * well, but that would cause problems with existing
292 		 * applications that open both address-specific sockets and
293 		 * a wildcard socket listening to the same port -- they would
294 		 * end up receiving duplicates of every unicast datagram.
295 		 * Those applications open the multiple sockets to overcome an
296 		 * inadequacy of the UDP socket interface, but for backwards
297 		 * compatibility we avoid the problem here rather than
298 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
299 		 */
300 
301 		/*
302 		 * Construct sockaddr format source address.
303 		 */
304 		udp_in.sin_port = uh->uh_sport;
305 		udp_in.sin_addr = ip->ip_src;
306 		/*
307 		 * Locate pcb(s) for datagram.
308 		 * (Algorithm copied from raw_intr().)
309 		 */
310 		last = NULL;
311 #ifdef INET6
312 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
313 #endif
314 		LIST_FOREACH(inp, &udb, inp_list) {
315 #ifdef INET6
316 			if ((inp->inp_vflag & INP_IPV4) == 0)
317 				continue;
318 #endif
319 			if (inp->inp_lport != uh->uh_dport)
320 				continue;
321 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
322 				if (inp->inp_laddr.s_addr !=
323 				    ip->ip_dst.s_addr)
324 					continue;
325 			}
326 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
327 				if (inp->inp_faddr.s_addr !=
328 				    ip->ip_src.s_addr ||
329 				    inp->inp_fport != uh->uh_sport)
330 					continue;
331 			}
332 
333 			if (check_multicast_membership(ip, inp, m) < 0)
334 				continue;
335 
336 			if (last != NULL) {
337 				struct mbuf *n;
338 
339 #ifdef IPSEC
340 				/* check AH/ESP integrity. */
341 				if (ipsec4_in_reject_so(m, last->inp_socket))
342 					ipsecstat.in_polvio++;
343 					/* do not inject data to pcb */
344 				else
345 #endif /*IPSEC*/
346 #ifdef FAST_IPSEC
347 				/* check AH/ESP integrity. */
348 				if (ipsec4_in_reject(m, last))
349 					;
350 				else
351 #endif /*FAST_IPSEC*/
352 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
353 					udp_append(last, ip, n,
354 						   iphlen +
355 						   sizeof(struct udphdr));
356 			}
357 			last = inp;
358 			/*
359 			 * Don't look for additional matches if this one does
360 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
361 			 * socket options set.  This heuristic avoids searching
362 			 * through all pcbs in the common case of a non-shared
363 			 * port.  It * assumes that an application will never
364 			 * clear these options after setting them.
365 			 */
366 			if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
367 				break;
368 		}
369 
370 		if (last == NULL) {
371 			/*
372 			 * No matching pcb found; discard datagram.
373 			 * (No need to send an ICMP Port Unreachable
374 			 * for a broadcast or multicast datgram.)
375 			 */
376 			udpstat.udps_noportbcast++;
377 			goto bad;
378 		}
379 #ifdef IPSEC
380 		/* check AH/ESP integrity. */
381 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
382 			ipsecstat.in_polvio++;
383 			goto bad;
384 		}
385 #endif /*IPSEC*/
386 #ifdef FAST_IPSEC
387 		/* check AH/ESP integrity. */
388 		if (ipsec4_in_reject(m, last))
389 			goto bad;
390 #endif /*FAST_IPSEC*/
391 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
392 		return;
393 	}
394 	/*
395 	 * Locate pcb for datagram.
396 	 */
397 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
398 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
399 	if (inp == NULL) {
400 		if (log_in_vain) {
401 			char buf[4*sizeof "123"];
402 
403 			strcpy(buf, inet_ntoa(ip->ip_dst));
404 			log(LOG_INFO,
405 			    "Connection attempt to UDP %s:%d from %s:%d\n",
406 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
407 			    ntohs(uh->uh_sport));
408 		}
409 		udpstat.udps_noport++;
410 		if (m->m_flags & (M_BCAST | M_MCAST)) {
411 			udpstat.udps_noportbcast++;
412 			goto bad;
413 		}
414 		if (blackhole)
415 			goto bad;
416 #ifdef ICMP_BANDLIM
417 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
418 			goto bad;
419 #endif
420 		*ip = save_ip;
421 		ip->ip_len += iphlen;
422 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
423 		return;
424 	}
425 #ifdef IPSEC
426 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
427 		ipsecstat.in_polvio++;
428 		goto bad;
429 	}
430 #endif /*IPSEC*/
431 #ifdef FAST_IPSEC
432 	if (ipsec4_in_reject(m, inp))
433 		goto bad;
434 #endif /*FAST_IPSEC*/
435 
436 	/*
437 	 * Construct sockaddr format source address.
438 	 * Stuff source address and datagram in user buffer.
439 	 */
440 	udp_in.sin_port = uh->uh_sport;
441 	udp_in.sin_addr = ip->ip_src;
442 	if (inp->inp_flags & INP_CONTROLOPTS
443 	    || inp->inp_socket->so_options & SO_TIMESTAMP) {
444 #ifdef INET6
445 		if (inp->inp_vflag & INP_IPV6) {
446 			int savedflags;
447 
448 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
449 			savedflags = inp->inp_flags;
450 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
451 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
452 			inp->inp_flags = savedflags;
453 		} else
454 #endif
455 		ip_savecontrol(inp, &opts, ip, m);
456 	}
457  	m_adj(m, iphlen + sizeof(struct udphdr));
458 #ifdef INET6
459 	if (inp->inp_vflag & INP_IPV6) {
460 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
461 		append_sa = (struct sockaddr *)&udp_in6;
462 	} else
463 #endif
464 	append_sa = (struct sockaddr *)&udp_in;
465 	if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
466 		udpstat.udps_fullsock++;
467 		goto bad;
468 	}
469 	sorwakeup(inp->inp_socket);
470 	return;
471 bad:
472 	m_freem(m);
473 	if (opts)
474 		m_freem(opts);
475 	return;
476 }
477 
478 #ifdef INET6
479 static void
480 ip_2_ip6_hdr(ip6, ip)
481 	struct ip6_hdr *ip6;
482 	struct ip *ip;
483 {
484 	bzero(ip6, sizeof(*ip6));
485 
486 	ip6->ip6_vfc = IPV6_VERSION;
487 	ip6->ip6_plen = ip->ip_len;
488 	ip6->ip6_nxt = ip->ip_p;
489 	ip6->ip6_hlim = ip->ip_ttl;
490 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
491 		IPV6_ADDR_INT32_SMP;
492 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
493 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
494 }
495 #endif
496 
497 /*
498  * subroutine of udp_input(), mainly for source code readability.
499  * caller must properly init udp_ip6 and udp_in6 beforehand.
500  */
501 static void
502 udp_append(last, ip, n, off)
503 	struct inpcb *last;
504 	struct ip *ip;
505 	struct mbuf *n;
506 	int off;
507 {
508 	struct sockaddr *append_sa;
509 	struct mbuf *opts = 0;
510 
511 	if (last->inp_flags & INP_CONTROLOPTS ||
512 	    last->inp_socket->so_options & SO_TIMESTAMP) {
513 #ifdef INET6
514 		if (last->inp_vflag & INP_IPV6) {
515 			int savedflags;
516 
517 			if (udp_ip6.uip6_init_done == 0) {
518 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
519 				udp_ip6.uip6_init_done = 1;
520 			}
521 			savedflags = last->inp_flags;
522 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
523 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
524 			last->inp_flags = savedflags;
525 		} else
526 #endif
527 		ip_savecontrol(last, &opts, ip, n);
528 	}
529 #ifdef INET6
530 	if (last->inp_vflag & INP_IPV6) {
531 		if (udp_in6.uin6_init_done == 0) {
532 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
533 			udp_in6.uin6_init_done = 1;
534 		}
535 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
536 	} else
537 #endif
538 	append_sa = (struct sockaddr *)&udp_in;
539 	m_adj(n, off);
540 	if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
541 		m_freem(n);
542 		if (opts)
543 			m_freem(opts);
544 		udpstat.udps_fullsock++;
545 	} else
546 		sorwakeup(last->inp_socket);
547 }
548 
549 /*
550  * Notify a udp user of an asynchronous error;
551  * just wake up so that he can collect error status.
552  */
553 void
554 udp_notify(inp, errno)
555 	struct inpcb *inp;
556 	int errno;
557 {
558 	inp->inp_socket->so_error = errno;
559 	sorwakeup(inp->inp_socket);
560 	sowwakeup(inp->inp_socket);
561 }
562 
563 void
564 udp_ctlinput(cmd, sa, vip)
565 	int cmd;
566 	struct sockaddr *sa;
567 	void *vip;
568 {
569 	struct ip *ip = vip;
570 	struct udphdr *uh;
571 	void (*notify) (struct inpcb *, int) = udp_notify;
572         struct in_addr faddr;
573 	struct inpcb *inp;
574 	int s;
575 
576 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
577 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
578         	return;
579 
580 	if (PRC_IS_REDIRECT(cmd)) {
581 		ip = 0;
582 		notify = in_rtchange;
583 	} else if (cmd == PRC_HOSTDEAD)
584 		ip = 0;
585 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
586 		return;
587 	if (ip) {
588 		s = splnet();
589 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
590 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
591                     ip->ip_src, uh->uh_sport, 0, NULL);
592 		if (inp != NULL && inp->inp_socket != NULL)
593 			(*notify)(inp, inetctlerrmap[cmd]);
594 		splx(s);
595 	} else
596 		in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
597 }
598 
599 static int
600 udp_pcblist(SYSCTL_HANDLER_ARGS)
601 {
602 	int error, i, n, s;
603 	struct inpcb *inp, **inp_list;
604 	inp_gen_t gencnt;
605 	struct xinpgen xig;
606 
607 	/*
608 	 * The process of preparing the TCB list is too time-consuming and
609 	 * resource-intensive to repeat twice on every request.
610 	 */
611 	if (req->oldptr == 0) {
612 		n = udbinfo.ipi_count;
613 		req->oldidx = 2 * (sizeof xig)
614 			+ (n + n/8) * sizeof(struct xinpcb);
615 		return 0;
616 	}
617 
618 	if (req->newptr != 0)
619 		return EPERM;
620 
621 	/*
622 	 * OK, now we're committed to doing something.
623 	 */
624 	s = splnet();
625 	gencnt = udbinfo.ipi_gencnt;
626 	n = udbinfo.ipi_count;
627 	splx(s);
628 
629 	xig.xig_len = sizeof xig;
630 	xig.xig_count = n;
631 	xig.xig_gen = gencnt;
632 	xig.xig_sogen = so_gencnt;
633 	error = SYSCTL_OUT(req, &xig, sizeof xig);
634 	if (error)
635 		return error;
636 
637 	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
638 	if (inp_list == 0)
639 		return ENOMEM;
640 
641 	s = splnet();
642 	for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
643 	     inp = LIST_NEXT(inp, inp_list)) {
644 		if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
645 			inp_list[i++] = inp;
646 	}
647 	splx(s);
648 	n = i;
649 
650 	error = 0;
651 	for (i = 0; i < n; i++) {
652 		inp = inp_list[i];
653 		if (inp->inp_gencnt <= gencnt) {
654 			struct xinpcb xi;
655 			xi.xi_len = sizeof xi;
656 			/* XXX should avoid extra copy */
657 			bcopy(inp, &xi.xi_inp, sizeof *inp);
658 			if (inp->inp_socket)
659 				sotoxsocket(inp->inp_socket, &xi.xi_socket);
660 			error = SYSCTL_OUT(req, &xi, sizeof xi);
661 		}
662 	}
663 	if (!error) {
664 		/*
665 		 * Give the user an updated idea of our state.
666 		 * If the generation differs from what we told
667 		 * her before, she knows that something happened
668 		 * while we were processing this request, and it
669 		 * might be necessary to retry.
670 		 */
671 		s = splnet();
672 		xig.xig_gen = udbinfo.ipi_gencnt;
673 		xig.xig_sogen = so_gencnt;
674 		xig.xig_count = udbinfo.ipi_count;
675 		splx(s);
676 		error = SYSCTL_OUT(req, &xig, sizeof xig);
677 	}
678 	free(inp_list, M_TEMP);
679 	return error;
680 }
681 
682 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
683 	    udp_pcblist, "S,xinpcb", "List of active UDP sockets");
684 
685 static int
686 udp_getcred(SYSCTL_HANDLER_ARGS)
687 {
688 	struct sockaddr_in addrs[2];
689 	struct inpcb *inp;
690 	int error, s;
691 
692 	error = suser(req->td);
693 	if (error)
694 		return (error);
695 	error = SYSCTL_IN(req, addrs, sizeof(addrs));
696 	if (error)
697 		return (error);
698 	s = splnet();
699 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
700 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
701 	if (inp == NULL || inp->inp_socket == NULL) {
702 		error = ENOENT;
703 		goto out;
704 	}
705 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
706 out:
707 	splx(s);
708 	return (error);
709 }
710 
711 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
712     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
713 
714 static int
715 udp_output(inp, m, addr, control, td)
716 	struct inpcb *inp;
717 	struct mbuf *m;
718 	struct sockaddr *addr;
719 	struct mbuf *control;
720 	struct thread *td;
721 {
722 	struct udpiphdr *ui;
723 	int len = m->m_pkthdr.len;
724 	struct in_addr laddr;
725 	struct sockaddr_in *sin;
726 	int s = 0, error = 0;
727 
728 	if (control)
729 		m_freem(control);		/* XXX */
730 
731 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
732 		error = EMSGSIZE;
733 		goto release;
734 	}
735 
736 	if (addr) {
737 		sin = (struct sockaddr_in *)addr;
738 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
739 		laddr = inp->inp_laddr;
740 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
741 			error = EISCONN;
742 			goto release;
743 		}
744 		/*
745 		 * Must block input while temporarily connected.
746 		 */
747 		s = splnet();
748 		error = in_pcbconnect(inp, addr, td);
749 		if (error) {
750 			splx(s);
751 			goto release;
752 		}
753 	} else {
754 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
755 			error = ENOTCONN;
756 			goto release;
757 		}
758 	}
759 	/*
760 	 * Calculate data length and get a mbuf
761 	 * for UDP and IP headers.
762 	 */
763 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
764 	if (m == 0) {
765 		error = ENOBUFS;
766 		if (addr)
767 			splx(s);
768 		goto release;
769 	}
770 
771 	/*
772 	 * Fill in mbuf with extended UDP header
773 	 * and addresses and length put into network format.
774 	 */
775 	ui = mtod(m, struct udpiphdr *);
776 	bzero(ui->ui_x1, sizeof(ui->ui_x1));	/* XXX still needed? */
777 	ui->ui_pr = IPPROTO_UDP;
778 	ui->ui_src = inp->inp_laddr;
779 	ui->ui_dst = inp->inp_faddr;
780 	ui->ui_sport = inp->inp_lport;
781 	ui->ui_dport = inp->inp_fport;
782 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
783 
784 	/*
785 	 * Set up checksum and output datagram.
786 	 */
787 	if (udpcksum) {
788         	ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
789 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
790 		m->m_pkthdr.csum_flags = CSUM_UDP;
791 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
792 	} else {
793 		ui->ui_sum = 0;
794 	}
795 	((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
796 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
797 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
798 	udpstat.udps_opackets++;
799 
800 	error = ip_output(m, inp->inp_options, &inp->inp_route,
801 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
802 	    inp->inp_moptions, inp);
803 
804 	if (addr) {
805 		in_pcbdisconnect(inp);
806 		inp->inp_laddr = laddr;	/* XXX rehash? */
807 		splx(s);
808 	}
809 	return (error);
810 
811 release:
812 	m_freem(m);
813 	return (error);
814 }
815 
816 u_long	udp_sendspace = 9216;		/* really max datagram size */
817 					/* 40 1K datagrams */
818 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
819     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
820 
821 u_long	udp_recvspace = 40 * (1024 +
822 #ifdef INET6
823 				      sizeof(struct sockaddr_in6)
824 #else
825 				      sizeof(struct sockaddr_in)
826 #endif
827 				      );
828 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
829     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
830 
831 static int
832 udp_abort(struct socket *so)
833 {
834 	struct inpcb *inp;
835 	int s;
836 
837 	inp = sotoinpcb(so);
838 	if (inp == 0)
839 		return EINVAL;	/* ??? possible? panic instead? */
840 	soisdisconnected(so);
841 	s = splnet();
842 	in_pcbdetach(inp);
843 	splx(s);
844 	return 0;
845 }
846 
847 static int
848 udp_attach(struct socket *so, int proto, struct thread *td)
849 {
850 	struct inpcb *inp;
851 	int s, error;
852 
853 	inp = sotoinpcb(so);
854 	if (inp != 0)
855 		return EINVAL;
856 
857 	error = soreserve(so, udp_sendspace, udp_recvspace);
858 	if (error)
859 		return error;
860 	s = splnet();
861 	error = in_pcballoc(so, &udbinfo, td);
862 	splx(s);
863 	if (error)
864 		return error;
865 
866 	inp = (struct inpcb *)so->so_pcb;
867 	inp->inp_vflag |= INP_IPV4;
868 	inp->inp_ip_ttl = ip_defttl;
869 	return 0;
870 }
871 
872 static int
873 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
874 {
875 	struct inpcb *inp;
876 	int s, error;
877 
878 	inp = sotoinpcb(so);
879 	if (inp == 0)
880 		return EINVAL;
881 	s = splnet();
882 	error = in_pcbbind(inp, nam, td);
883 	splx(s);
884 	return error;
885 }
886 
887 static int
888 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
889 {
890 	struct inpcb *inp;
891 	int s, error;
892 	struct sockaddr_in *sin;
893 
894 	inp = sotoinpcb(so);
895 	if (inp == 0)
896 		return EINVAL;
897 	if (inp->inp_faddr.s_addr != INADDR_ANY)
898 		return EISCONN;
899 	error = 0;
900 	s = splnet();
901 	if (inp->inp_laddr.s_addr == INADDR_ANY &&
902 	    td->td_proc &&
903 	    td->td_proc->p_ucred->cr_prison != NULL) {
904 		error = in_pcbbind(inp, NULL, td);
905 	}
906 	if (error == 0) {
907 		sin = (struct sockaddr_in *)nam;
908 		prison_remote_ip(td, 0, &sin->sin_addr.s_addr);
909 		error = in_pcbconnect(inp, nam, td);
910 	}
911 	splx(s);
912 	if (error == 0)
913 		soisconnected(so);
914 	return error;
915 }
916 
917 static int
918 udp_detach(struct socket *so)
919 {
920 	struct inpcb *inp;
921 	int s;
922 
923 	inp = sotoinpcb(so);
924 	if (inp == 0)
925 		return EINVAL;
926 	s = splnet();
927 	in_pcbdetach(inp);
928 	splx(s);
929 	return 0;
930 }
931 
932 static int
933 udp_disconnect(struct socket *so)
934 {
935 	struct inpcb *inp;
936 	int s;
937 
938 	inp = sotoinpcb(so);
939 	if (inp == 0)
940 		return EINVAL;
941 	if (inp->inp_faddr.s_addr == INADDR_ANY)
942 		return ENOTCONN;
943 
944 	s = splnet();
945 	in_pcbdisconnect(inp);
946 	inp->inp_laddr.s_addr = INADDR_ANY;
947 	splx(s);
948 	so->so_state &= ~SS_ISCONNECTED;		/* XXX */
949 	return 0;
950 }
951 
952 static int
953 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
954 	    struct mbuf *control, struct thread *td)
955 {
956 	struct inpcb *inp;
957 
958 	inp = sotoinpcb(so);
959 	if (inp == 0) {
960 		m_freem(m);
961 		return EINVAL;
962 	}
963 	return udp_output(inp, m, addr, control, td);
964 }
965 
966 int
967 udp_shutdown(struct socket *so)
968 {
969 	struct inpcb *inp;
970 
971 	inp = sotoinpcb(so);
972 	if (inp == 0)
973 		return EINVAL;
974 	socantsendmore(so);
975 	return 0;
976 }
977 
978 struct pr_usrreqs udp_usrreqs = {
979 	udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
980 	pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
981 	pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
982 	pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
983 	in_setsockaddr, sosend, soreceive, sopoll
984 };
985 
986