xref: /dflybsd-src/sys/netinet/udp_usrreq.c (revision 6cef7136f04e2b24a6db289e78720d6d8c60274e)
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
67  * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
68  * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.47 2008/11/11 10:46:58 sephe Exp $
69  */
70 
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/proc.h>
81 #include <sys/priv.h>
82 #include <sys/protosw.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 #include <sys/in_cksum.h>
88 
89 #include <sys/thread2.h>
90 #include <sys/socketvar2.h>
91 
92 #include <machine/stdarg.h>
93 
94 #include <net/if.h>
95 #include <net/route.h>
96 #include <net/netmsg2.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/ip.h>
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #endif
104 #include <netinet/in_pcb.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip_var.h>
107 #ifdef INET6
108 #include <netinet6/ip6_var.h>
109 #endif
110 #include <netinet/ip_icmp.h>
111 #include <netinet/icmp_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
114 
115 #ifdef FAST_IPSEC
116 #include <netproto/ipsec/ipsec.h>
117 #endif
118 
119 #ifdef IPSEC
120 #include <netinet6/ipsec.h>
121 #endif
122 
123 /*
124  * UDP protocol implementation.
125  * Per RFC 768, August, 1980.
126  */
127 #ifndef	COMPAT_42
128 static int	udpcksum = 1;
129 #else
130 static int	udpcksum = 0;		/* XXX */
131 #endif
132 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
133 		&udpcksum, 0, "");
134 
135 int	log_in_vain = 0;
136 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
137     &log_in_vain, 0, "Log all incoming UDP packets");
138 
139 static int	blackhole = 0;
140 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
141 	&blackhole, 0, "Do not send port unreachables for refused connects");
142 
143 static int	strict_mcast_mship = 1;
144 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
145 	&strict_mcast_mship, 0, "Only send multicast to member sockets");
146 
147 struct	inpcbinfo udbinfo;
148 
149 #ifndef UDBHASHSIZE
150 #define UDBHASHSIZE 16
151 #endif
152 
153 struct	udpstat udpstat;	/* from udp_var.h */
154 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
155     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
156 
157 static struct	sockaddr_in udp_in = { sizeof udp_in, AF_INET };
158 #ifdef INET6
159 struct udp_in6 {
160 	struct sockaddr_in6	uin6_sin;
161 	u_char			uin6_init_done : 1;
162 } udp_in6 = {
163 	{ sizeof udp_in6.uin6_sin, AF_INET6 },
164 	0
165 };
166 struct udp_ip6 {
167 	struct ip6_hdr		uip6_ip6;
168 	u_char			uip6_init_done : 1;
169 } udp_ip6;
170 #endif /* INET6 */
171 
172 static void udp_append (struct inpcb *last, struct ip *ip,
173 			    struct mbuf *n, int off);
174 #ifdef INET6
175 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
176 #endif
177 
178 static int udp_connect_oncpu(struct socket *so, struct thread *td,
179 			struct sockaddr_in *sin, struct sockaddr_in *if_sin);
180 static int udp_detach (struct socket *so);
181 static	int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
182 			    struct mbuf *, struct thread *);
183 
184 void
185 udp_init(void)
186 {
187 	in_pcbinfo_init(&udbinfo);
188 	udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
189 	udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
190 					&udbinfo.porthashmask);
191 	udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
192 					    &udbinfo.wildcardhashmask);
193 	udbinfo.ipi_size = sizeof(struct inpcb);
194 }
195 
196 /*
197  * Check multicast packets to make sure they are only sent to sockets with
198  * multicast memberships for the packet's destination address and arrival
199  * interface.  Multicast packets to multicast-unaware sockets are also
200  * disallowed.
201  *
202  * Returns 0 if the packet is acceptable, -1 if it is not.
203  */
204 static __inline int
205 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
206 {
207 	int mshipno;
208 	struct ip_moptions *mopt;
209 
210 	if (strict_mcast_mship == 0 ||
211 	    !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
212 		return (0);
213 	}
214 	mopt = inp->inp_moptions;
215 	if (mopt == NULL)
216 		return (-1);
217 	for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
218 		struct in_multi *maddr = mopt->imo_membership[mshipno];
219 
220 		if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
221 		    m->m_pkthdr.rcvif == maddr->inm_ifp) {
222 			return (0);
223 		}
224 	}
225 	return (-1);
226 }
227 
228 void
229 udp_input(struct mbuf *m, ...)
230 {
231 	int iphlen;
232 	struct ip *ip;
233 	struct udphdr *uh;
234 	struct inpcb *inp;
235 	struct mbuf *opts = NULL;
236 	int len, off, proto;
237 	struct ip save_ip;
238 	struct sockaddr *append_sa;
239 	__va_list ap;
240 
241 	__va_start(ap, m);
242 	off = __va_arg(ap, int);
243 	proto = __va_arg(ap, int);
244 	__va_end(ap);
245 
246 	iphlen = off;
247 	udpstat.udps_ipackets++;
248 
249 	/*
250 	 * Strip IP options, if any; should skip this,
251 	 * make available to user, and use on returned packets,
252 	 * but we don't yet have a way to check the checksum
253 	 * with options still present.
254 	 */
255 	if (iphlen > sizeof(struct ip)) {
256 		ip_stripoptions(m);
257 		iphlen = sizeof(struct ip);
258 	}
259 
260 	/*
261 	 * IP and UDP headers are together in first mbuf.
262 	 * Already checked and pulled up in ip_demux().
263 	 */
264 	KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
265 	    ("UDP header not in one mbuf"));
266 
267 	ip = mtod(m, struct ip *);
268 	uh = (struct udphdr *)((caddr_t)ip + iphlen);
269 
270 	/* destination port of 0 is illegal, based on RFC768. */
271 	if (uh->uh_dport == 0)
272 		goto bad;
273 
274 	/*
275 	 * Make mbuf data length reflect UDP length.
276 	 * If not enough data to reflect UDP length, drop.
277 	 */
278 	len = ntohs((u_short)uh->uh_ulen);
279 	if (ip->ip_len != len) {
280 		if (len > ip->ip_len || len < sizeof(struct udphdr)) {
281 			udpstat.udps_badlen++;
282 			goto bad;
283 		}
284 		m_adj(m, len - ip->ip_len);
285 		/* ip->ip_len = len; */
286 	}
287 	/*
288 	 * Save a copy of the IP header in case we want restore it
289 	 * for sending an ICMP error message in response.
290 	 */
291 	save_ip = *ip;
292 
293 	/*
294 	 * Checksum extended UDP header and data.
295 	 */
296 	if (uh->uh_sum) {
297 		if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
298 			if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
299 				uh->uh_sum = m->m_pkthdr.csum_data;
300 			else
301 				uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
302 				    ip->ip_dst.s_addr, htonl((u_short)len +
303 				    m->m_pkthdr.csum_data + IPPROTO_UDP));
304 			uh->uh_sum ^= 0xffff;
305 		} else {
306 			char b[9];
307 
308 			bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
309 			bzero(((struct ipovly *)ip)->ih_x1, 9);
310 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
311 			uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
312 			bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
313 		}
314 		if (uh->uh_sum) {
315 			udpstat.udps_badsum++;
316 			m_freem(m);
317 			return;
318 		}
319 	} else
320 		udpstat.udps_nosum++;
321 
322 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
323 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
324 		struct inpcb *last;
325 
326 		/*
327 		 * Deliver a multicast or broadcast datagram to *all* sockets
328 		 * for which the local and remote addresses and ports match
329 		 * those of the incoming datagram.  This allows more than
330 		 * one process to receive multi/broadcasts on the same port.
331 		 * (This really ought to be done for unicast datagrams as
332 		 * well, but that would cause problems with existing
333 		 * applications that open both address-specific sockets and
334 		 * a wildcard socket listening to the same port -- they would
335 		 * end up receiving duplicates of every unicast datagram.
336 		 * Those applications open the multiple sockets to overcome an
337 		 * inadequacy of the UDP socket interface, but for backwards
338 		 * compatibility we avoid the problem here rather than
339 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
340 		 */
341 
342 		/*
343 		 * Construct sockaddr format source address.
344 		 */
345 		udp_in.sin_port = uh->uh_sport;
346 		udp_in.sin_addr = ip->ip_src;
347 		/*
348 		 * Locate pcb(s) for datagram.
349 		 * (Algorithm copied from raw_intr().)
350 		 */
351 		last = NULL;
352 #ifdef INET6
353 		udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
354 #endif
355 		LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
356 			if (inp->inp_flags & INP_PLACEMARKER)
357 				continue;
358 #ifdef INET6
359 			if (!(inp->inp_vflag & INP_IPV4))
360 				continue;
361 #endif
362 			if (inp->inp_lport != uh->uh_dport)
363 				continue;
364 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
365 				if (inp->inp_laddr.s_addr !=
366 				    ip->ip_dst.s_addr)
367 					continue;
368 			}
369 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
370 				if (inp->inp_faddr.s_addr !=
371 				    ip->ip_src.s_addr ||
372 				    inp->inp_fport != uh->uh_sport)
373 					continue;
374 			}
375 
376 			if (check_multicast_membership(ip, inp, m) < 0)
377 				continue;
378 
379 			if (last != NULL) {
380 				struct mbuf *n;
381 
382 #ifdef IPSEC
383 				/* check AH/ESP integrity. */
384 				if (ipsec4_in_reject_so(m, last->inp_socket))
385 					ipsecstat.in_polvio++;
386 					/* do not inject data to pcb */
387 				else
388 #endif /*IPSEC*/
389 #ifdef FAST_IPSEC
390 				/* check AH/ESP integrity. */
391 				if (ipsec4_in_reject(m, last))
392 					;
393 				else
394 #endif /*FAST_IPSEC*/
395 				if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
396 					udp_append(last, ip, n,
397 						   iphlen +
398 						   sizeof(struct udphdr));
399 			}
400 			last = inp;
401 			/*
402 			 * Don't look for additional matches if this one does
403 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
404 			 * socket options set.  This heuristic avoids searching
405 			 * through all pcbs in the common case of a non-shared
406 			 * port.  It * assumes that an application will never
407 			 * clear these options after setting them.
408 			 */
409 			if (!(last->inp_socket->so_options &
410 			    (SO_REUSEPORT | SO_REUSEADDR)))
411 				break;
412 		}
413 
414 		if (last == NULL) {
415 			/*
416 			 * No matching pcb found; discard datagram.
417 			 * (No need to send an ICMP Port Unreachable
418 			 * for a broadcast or multicast datgram.)
419 			 */
420 			udpstat.udps_noportbcast++;
421 			goto bad;
422 		}
423 #ifdef IPSEC
424 		/* check AH/ESP integrity. */
425 		if (ipsec4_in_reject_so(m, last->inp_socket)) {
426 			ipsecstat.in_polvio++;
427 			goto bad;
428 		}
429 #endif /*IPSEC*/
430 #ifdef FAST_IPSEC
431 		/* check AH/ESP integrity. */
432 		if (ipsec4_in_reject(m, last))
433 			goto bad;
434 #endif /*FAST_IPSEC*/
435 		udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
436 		return;
437 	}
438 	/*
439 	 * Locate pcb for datagram.
440 	 */
441 	inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
442 	    ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
443 	if (inp == NULL) {
444 		if (log_in_vain) {
445 			char buf[sizeof "aaa.bbb.ccc.ddd"];
446 
447 			strcpy(buf, inet_ntoa(ip->ip_dst));
448 			log(LOG_INFO,
449 			    "Connection attempt to UDP %s:%d from %s:%d\n",
450 			    buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
451 			    ntohs(uh->uh_sport));
452 		}
453 		udpstat.udps_noport++;
454 		if (m->m_flags & (M_BCAST | M_MCAST)) {
455 			udpstat.udps_noportbcast++;
456 			goto bad;
457 		}
458 		if (blackhole)
459 			goto bad;
460 #ifdef ICMP_BANDLIM
461 		if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
462 			goto bad;
463 #endif
464 		*ip = save_ip;
465 		ip->ip_len += iphlen;
466 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
467 		return;
468 	}
469 #ifdef IPSEC
470 	if (ipsec4_in_reject_so(m, inp->inp_socket)) {
471 		ipsecstat.in_polvio++;
472 		goto bad;
473 	}
474 #endif /*IPSEC*/
475 #ifdef FAST_IPSEC
476 	if (ipsec4_in_reject(m, inp))
477 		goto bad;
478 #endif /*FAST_IPSEC*/
479 	/*
480 	 * Check the minimum TTL for socket.
481 	 */
482 	if (ip->ip_ttl < inp->inp_ip_minttl)
483 		goto bad;
484 
485 	/*
486 	 * Construct sockaddr format source address.
487 	 * Stuff source address and datagram in user buffer.
488 	 */
489 	udp_in.sin_port = uh->uh_sport;
490 	udp_in.sin_addr = ip->ip_src;
491 	if ((inp->inp_flags & INP_CONTROLOPTS) ||
492 	    (inp->inp_socket->so_options & SO_TIMESTAMP)) {
493 #ifdef INET6
494 		if (inp->inp_vflag & INP_IPV6) {
495 			int savedflags;
496 
497 			ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
498 			savedflags = inp->inp_flags;
499 			inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
500 			ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
501 			inp->inp_flags = savedflags;
502 		} else
503 #endif
504 		ip_savecontrol(inp, &opts, ip, m);
505 	}
506 	m_adj(m, iphlen + sizeof(struct udphdr));
507 #ifdef INET6
508 	if (inp->inp_vflag & INP_IPV6) {
509 		in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
510 		append_sa = (struct sockaddr *)&udp_in6;
511 	} else
512 #endif
513 		append_sa = (struct sockaddr *)&udp_in;
514 	if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
515 		udpstat.udps_fullsock++;
516 		goto bad;
517 	}
518 	sorwakeup(inp->inp_socket);
519 	return;
520 bad:
521 	m_freem(m);
522 	if (opts)
523 		m_freem(opts);
524 	return;
525 }
526 
527 #ifdef INET6
528 static void
529 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
530 {
531 	bzero(ip6, sizeof *ip6);
532 
533 	ip6->ip6_vfc = IPV6_VERSION;
534 	ip6->ip6_plen = ip->ip_len;
535 	ip6->ip6_nxt = ip->ip_p;
536 	ip6->ip6_hlim = ip->ip_ttl;
537 	ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
538 		IPV6_ADDR_INT32_SMP;
539 	ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
540 	ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
541 }
542 #endif
543 
544 /*
545  * subroutine of udp_input(), mainly for source code readability.
546  * caller must properly init udp_ip6 and udp_in6 beforehand.
547  */
548 static void
549 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
550 {
551 	struct sockaddr *append_sa;
552 	struct mbuf *opts = NULL;
553 
554 	if (last->inp_flags & INP_CONTROLOPTS ||
555 	    last->inp_socket->so_options & SO_TIMESTAMP) {
556 #ifdef INET6
557 		if (last->inp_vflag & INP_IPV6) {
558 			int savedflags;
559 
560 			if (udp_ip6.uip6_init_done == 0) {
561 				ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
562 				udp_ip6.uip6_init_done = 1;
563 			}
564 			savedflags = last->inp_flags;
565 			last->inp_flags &= ~INP_UNMAPPABLEOPTS;
566 			ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
567 			last->inp_flags = savedflags;
568 		} else
569 #endif
570 		ip_savecontrol(last, &opts, ip, n);
571 	}
572 #ifdef INET6
573 	if (last->inp_vflag & INP_IPV6) {
574 		if (udp_in6.uin6_init_done == 0) {
575 			in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
576 			udp_in6.uin6_init_done = 1;
577 		}
578 		append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
579 	} else
580 #endif
581 		append_sa = (struct sockaddr *)&udp_in;
582 	m_adj(n, off);
583 	if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
584 		m_freem(n);
585 		if (opts)
586 			m_freem(opts);
587 		udpstat.udps_fullsock++;
588 	} else
589 		sorwakeup(last->inp_socket);
590 }
591 
592 /*
593  * Notify a udp user of an asynchronous error;
594  * just wake up so that he can collect error status.
595  */
596 void
597 udp_notify(struct inpcb *inp, int error)
598 {
599 	inp->inp_socket->so_error = error;
600 	sorwakeup(inp->inp_socket);
601 	sowwakeup(inp->inp_socket);
602 }
603 
604 struct netmsg_udp_notify {
605 	struct netmsg	nm_nmsg;
606 	void		(*nm_notify)(struct inpcb *, int);
607 	struct in_addr	nm_faddr;
608 	int		nm_arg;
609 };
610 
611 static void
612 udp_notifyall_oncpu(struct netmsg *netmsg)
613 {
614 	struct netmsg_udp_notify *nmsg = (struct netmsg_udp_notify *)netmsg;
615 	int nextcpu;
616 
617 	in_pcbnotifyall(&udbinfo.pcblisthead, nmsg->nm_faddr, nmsg->nm_arg,
618 			nmsg->nm_notify);
619 
620 	nextcpu = mycpuid + 1;
621 	if (nextcpu < ncpus2)
622 		lwkt_forwardmsg(cpu_portfn(nextcpu), &netmsg->nm_lmsg);
623 	else
624 		lwkt_replymsg(&netmsg->nm_lmsg, 0);
625 }
626 
627 static void
628 udp_rtchange(struct inpcb *inp, int err)
629 {
630 #ifdef SMP
631 	/* XXX Nuke this, once UDP inpcbs are CPU localized */
632 	if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
633 		rtfree(inp->inp_route.ro_rt);
634 		inp->inp_route.ro_rt = NULL;
635 		/*
636 		 * A new route can be allocated the next time
637 		 * output is attempted.
638 		 */
639 	}
640 #else
641 	in_rtchange(inp, err);
642 #endif
643 }
644 
645 void
646 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
647 {
648 	struct ip *ip = vip;
649 	struct udphdr *uh;
650 	void (*notify) (struct inpcb *, int) = udp_notify;
651 	struct in_addr faddr;
652 	struct inpcb *inp;
653 
654 	faddr = ((struct sockaddr_in *)sa)->sin_addr;
655 	if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
656 		return;
657 
658 	if (PRC_IS_REDIRECT(cmd)) {
659 		ip = NULL;
660 		notify = udp_rtchange;
661 	} else if (cmd == PRC_HOSTDEAD)
662 		ip = NULL;
663 	else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
664 		return;
665 	if (ip) {
666 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
667 		inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
668 					ip->ip_src, uh->uh_sport, 0, NULL);
669 		if (inp != NULL && inp->inp_socket != NULL)
670 			(*notify)(inp, inetctlerrmap[cmd]);
671 	} else if (PRC_IS_REDIRECT(cmd)) {
672 		struct netmsg_udp_notify nmsg;
673 
674 		KKASSERT(&curthread->td_msgport == cpu_portfn(0));
675 		netmsg_init(&nmsg.nm_nmsg, NULL, &curthread->td_msgport,
676 			    0, udp_notifyall_oncpu);
677 		nmsg.nm_faddr = faddr;
678 		nmsg.nm_arg = inetctlerrmap[cmd];
679 		nmsg.nm_notify = notify;
680 
681 		lwkt_domsg(cpu_portfn(0), &nmsg.nm_nmsg.nm_lmsg, 0);
682 	} else {
683 		/*
684 		 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
685 		 * once UDP inpcbs are CPU localized
686 		 */
687 		KKASSERT(&curthread->td_msgport == cpu_portfn(0));
688 		in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
689 				notify);
690 	}
691 }
692 
693 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
694 	    in_pcblist_global, "S,xinpcb", "List of active UDP sockets");
695 
696 static int
697 udp_getcred(SYSCTL_HANDLER_ARGS)
698 {
699 	struct sockaddr_in addrs[2];
700 	struct inpcb *inp;
701 	int error;
702 
703 	error = priv_check(req->td, PRIV_ROOT);
704 	if (error)
705 		return (error);
706 	error = SYSCTL_IN(req, addrs, sizeof addrs);
707 	if (error)
708 		return (error);
709 	inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
710 				addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
711 	if (inp == NULL || inp->inp_socket == NULL) {
712 		error = ENOENT;
713 		goto out;
714 	}
715 	error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
716 out:
717 	return (error);
718 }
719 
720 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
721     0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
722 
723 static int
724 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
725 	   struct mbuf *control, struct thread *td)
726 {
727 	struct udpiphdr *ui;
728 	int len = m->m_pkthdr.len;
729 	struct sockaddr_in *sin;	/* really is initialized before use */
730 	int error = 0, lport_any = 0;
731 
732 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
733 		error = EMSGSIZE;
734 		goto release;
735 	}
736 
737 	if (inp->inp_lport == 0) {	/* unbound socket */
738 		error = in_pcbbind(inp, NULL, td);
739 		if (error)
740 			goto release;
741 		in_pcbinswildcardhash(inp);
742 		lport_any = 1;
743 	}
744 
745 	if (dstaddr != NULL) {		/* destination address specified */
746 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
747 			/* already connected */
748 			error = EISCONN;
749 			goto release;
750 		}
751 		sin = (struct sockaddr_in *)dstaddr;
752 		if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
753 			error = EAFNOSUPPORT; /* IPv6 only jail */
754 			goto release;
755 		}
756 	} else {
757 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
758 			/* no destination specified and not already connected */
759 			error = ENOTCONN;
760 			goto release;
761 		}
762 		sin = NULL;
763 	}
764 
765 	/*
766 	 * Calculate data length and get a mbuf
767 	 * for UDP and IP headers.
768 	 */
769 	M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
770 	if (m == NULL) {
771 		error = ENOBUFS;
772 		goto release;
773 	}
774 
775 	/*
776 	 * Fill in mbuf with extended UDP header
777 	 * and addresses and length put into network format.
778 	 */
779 	ui = mtod(m, struct udpiphdr *);
780 	bzero(ui->ui_x1, sizeof ui->ui_x1);	/* XXX still needed? */
781 	ui->ui_pr = IPPROTO_UDP;
782 
783 	/*
784 	 * Set destination address.
785 	 */
786 	if (dstaddr != NULL) {			/* use specified destination */
787 		ui->ui_dst = sin->sin_addr;
788 		ui->ui_dport = sin->sin_port;
789 	} else {				/* use connected destination */
790 		ui->ui_dst = inp->inp_faddr;
791 		ui->ui_dport = inp->inp_fport;
792 	}
793 
794 	/*
795 	 * Set source address.
796 	 */
797 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
798 		struct sockaddr_in *if_sin;
799 
800 		if (dstaddr == NULL) {
801 			/*
802 			 * connect() had (or should have) failed because
803 			 * the interface had no IP address, but the
804 			 * application proceeded to call send() anyways.
805 			 */
806 			error = ENOTCONN;
807 			goto release;
808 		}
809 
810 		/* Look up outgoing interface. */
811 		if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
812 			goto release;
813 		ui->ui_src = if_sin->sin_addr;	/* use address of interface */
814 	} else {
815 		ui->ui_src = inp->inp_laddr;	/* use non-null bound address */
816 	}
817 	ui->ui_sport = inp->inp_lport;
818 	KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
819 
820 	ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
821 
822 	/*
823 	 * Set up checksum and output datagram.
824 	 */
825 	if (udpcksum) {
826 		ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
827 		    htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
828 		m->m_pkthdr.csum_flags = CSUM_UDP;
829 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
830 	} else {
831 		ui->ui_sum = 0;
832 	}
833 	((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
834 	((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;	/* XXX */
835 	((struct ip *)ui)->ip_tos = inp->inp_ip_tos;	/* XXX */
836 	udpstat.udps_opackets++;
837 
838 	error = ip_output(m, inp->inp_options, &inp->inp_route,
839 	    (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
840 	    IP_DEBUGROUTE,
841 	    inp->inp_moptions, inp);
842 
843 	/*
844 	 * If this is the first data gram sent on an unbound and unconnected
845 	 * UDP socket, lport will be changed in this function.  If target
846 	 * CPU after this lport changing is no longer the current CPU, then
847 	 * free the route entry allocated on the current CPU.
848 	 */
849 	if (lport_any) {
850 		if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
851 		    inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
852 			struct route *ro = &inp->inp_route;
853 
854 			if (ro->ro_rt != NULL)
855 				RTFREE(ro->ro_rt);
856 			bzero(ro, sizeof(*ro));
857 		}
858 	}
859 	return (error);
860 
861 release:
862 	m_freem(m);
863 	return (error);
864 }
865 
866 u_long	udp_sendspace = 9216;		/* really max datagram size */
867 					/* 40 1K datagrams */
868 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
869     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
870 
871 u_long	udp_recvspace = 40 * (1024 +
872 #ifdef INET6
873 				      sizeof(struct sockaddr_in6)
874 #else
875 				      sizeof(struct sockaddr_in)
876 #endif
877 				      );
878 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
879     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
880 
881 /*
882  * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
883  *	 will sofree() it when we return.
884  */
885 static int
886 udp_abort(struct socket *so)
887 {
888 	struct inpcb *inp;
889 	int error;
890 
891 	inp = so->so_pcb;
892 	if (inp) {
893 		soisdisconnected(so);
894 		in_pcbdetach(inp);
895 		error = 0;
896 	} else {
897 		error = EINVAL;
898 	}
899 	return error;
900 }
901 
902 static int
903 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai)
904 {
905 	struct inpcb *inp;
906 	int error;
907 
908 	inp = so->so_pcb;
909 	if (inp != NULL)
910 		return EINVAL;
911 
912 	error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
913 	if (error)
914 		return error;
915 	error = in_pcballoc(so, &udbinfo);
916 	if (error)
917 		return error;
918 	so->so_port = udp_soport_attach(so);
919 
920 	inp = (struct inpcb *)so->so_pcb;
921 	inp->inp_vflag |= INP_IPV4;
922 	inp->inp_ip_ttl = ip_defttl;
923 	return 0;
924 }
925 
926 static int
927 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
928 {
929 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
930 	struct inpcb *inp;
931 	int error;
932 
933 	inp = so->so_pcb;
934 	if (inp == NULL)
935 		return EINVAL;
936 	error = in_pcbbind(inp, nam, td);
937 	if (error == 0) {
938 		if (sin->sin_addr.s_addr != INADDR_ANY)
939 			inp->inp_flags |= INP_WASBOUND_NOTANY;
940 		in_pcbinswildcardhash(inp);
941 	}
942 	return error;
943 }
944 
945 #ifdef SMP
946 
947 struct netmsg_udp_connect {
948 	struct netmsg		nm_netmsg;
949 	struct socket		*nm_so;
950 	struct sockaddr_in	*nm_sin;
951 	struct sockaddr_in	*nm_ifsin;
952 	struct thread		*nm_td;
953 };
954 
955 static void
956 udp_connect_handler(netmsg_t netmsg)
957 {
958 	struct netmsg_udp_connect *msg = (void *)netmsg;
959 	int error;
960 
961 	error = udp_connect_oncpu(msg->nm_so, msg->nm_td,
962 				  msg->nm_sin, msg->nm_ifsin);
963 	lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error);
964 }
965 
966 #endif
967 
968 static int
969 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
970 {
971 	struct inpcb *inp;
972 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
973 	struct sockaddr_in *if_sin;
974 	lwkt_port_t port;
975 	int error;
976 
977 	inp = so->so_pcb;
978 	if (inp == NULL)
979 		return EINVAL;
980 	if (inp->inp_faddr.s_addr != INADDR_ANY)
981 		return EISCONN;
982 	error = 0;
983 
984 	/*
985 	 * Bind if we have to
986 	 */
987 	if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
988 	    inp->inp_laddr.s_addr == INADDR_ANY) {
989 		error = in_pcbbind(inp, NULL, td);
990 		if (error)
991 			return (error);
992 	}
993 
994 	/*
995 	 * Calculate the correct protocol processing thread.  The connect
996 	 * operation must run there.
997 	 */
998 	error = in_pcbladdr(inp, nam, &if_sin, td);
999 	if (error)
1000 		return(error);
1001 	if (!prison_remote_ip(td, nam))
1002 		return(EAFNOSUPPORT); /* IPv6 only jail */
1003 
1004 	port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1005 			    inp->inp_laddr.s_addr, inp->inp_lport);
1006 #ifdef SMP
1007 	if (port != &curthread->td_msgport) {
1008 		struct netmsg_udp_connect msg;
1009 		struct route *ro = &inp->inp_route;
1010 
1011 		/*
1012 		 * in_pcbladdr() may have allocated a route entry for us
1013 		 * on the current CPU, but we need a route entry on the
1014 		 * inpcb's owner CPU, so free it here.
1015 		 */
1016 		if (ro->ro_rt != NULL)
1017 			RTFREE(ro->ro_rt);
1018 		bzero(ro, sizeof(*ro));
1019 
1020 		/*
1021 		 * NOTE: We haven't set so->so_port yet do not pass so
1022 		 *       to netmsg_init() or it will be improperly forwarded.
1023 		 */
1024 		netmsg_init(&msg.nm_netmsg, NULL, &curthread->td_msgport,
1025 			    0, udp_connect_handler);
1026 		msg.nm_so = so;
1027 		msg.nm_sin = sin;
1028 		msg.nm_ifsin = if_sin;
1029 		msg.nm_td = td;
1030 		error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0);
1031 	} else {
1032 		error = udp_connect_oncpu(so, td, sin, if_sin);
1033 	}
1034 #else
1035 	error = udp_connect_oncpu(so, td, sin, if_sin);
1036 #endif
1037 	return (error);
1038 }
1039 
1040 static int
1041 udp_connect_oncpu(struct socket *so, struct thread *td,
1042 		  struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1043 {
1044 	struct inpcb *inp;
1045 	int error;
1046 
1047 	inp = so->so_pcb;
1048 	if (inp->inp_flags & INP_WILDCARD)
1049 		in_pcbremwildcardhash(inp);
1050 	error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1051 
1052 	if (error == 0) {
1053 		/*
1054 		 * No more errors can occur, finish adjusting the socket
1055 		 * and change the processing port to reflect the connected
1056 		 * socket.  Once set we can no longer safely mess with the
1057 		 * socket.
1058 		 */
1059 		soisconnected(so);
1060 		sosetport(so, &curthread->td_msgport);
1061 	} else if (error == EAFNOSUPPORT) {	/* connection dissolved */
1062 		/*
1063 		 * Follow traditional BSD behavior and retain
1064 		 * the local port binding.  But, fix the old misbehavior
1065 		 * of overwriting any previously bound local address.
1066 		 */
1067 		if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1068 			inp->inp_laddr.s_addr = INADDR_ANY;
1069 		in_pcbinswildcardhash(inp);
1070 	}
1071 	return error;
1072 }
1073 
1074 static int
1075 udp_detach(struct socket *so)
1076 {
1077 	struct inpcb *inp;
1078 
1079 	inp = so->so_pcb;
1080 	if (inp == NULL)
1081 		return EINVAL;
1082 	in_pcbdetach(inp);
1083 	return 0;
1084 }
1085 
1086 static int
1087 udp_disconnect(struct socket *so)
1088 {
1089 	struct route *ro;
1090 	struct inpcb *inp;
1091 
1092 	inp = so->so_pcb;
1093 	if (inp == NULL)
1094 		return EINVAL;
1095 	if (inp->inp_faddr.s_addr == INADDR_ANY)
1096 		return ENOTCONN;
1097 
1098 	soreference(so);
1099 	in_pcbdisconnect(inp);
1100 	soclrstate(so, SS_ISCONNECTED);		/* XXX */
1101 	sofree(so);
1102 
1103 	ro = &inp->inp_route;
1104 	if (ro->ro_rt != NULL)
1105 		RTFREE(ro->ro_rt);
1106 	bzero(ro, sizeof(*ro));
1107 
1108 	return 0;
1109 }
1110 
1111 static int
1112 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1113 	    struct mbuf *control, struct thread *td)
1114 {
1115 	struct inpcb *inp;
1116 
1117 	inp = so->so_pcb;
1118 	if (inp == NULL) {
1119 		m_freem(m);
1120 		return EINVAL;
1121 	}
1122 	return udp_output(inp, m, addr, control, td);
1123 }
1124 
1125 int
1126 udp_shutdown(struct socket *so)
1127 {
1128 	struct inpcb *inp;
1129 
1130 	inp = so->so_pcb;
1131 	if (inp == NULL)
1132 		return EINVAL;
1133 	socantsendmore(so);
1134 	return 0;
1135 }
1136 
1137 struct pr_usrreqs udp_usrreqs = {
1138 	.pru_abort = udp_abort,
1139 	.pru_accept = pru_accept_notsupp,
1140 	.pru_attach = udp_attach,
1141 	.pru_bind = udp_bind,
1142 	.pru_connect = udp_connect,
1143 	.pru_connect2 = pru_connect2_notsupp,
1144 	.pru_control = in_control,
1145 	.pru_detach = udp_detach,
1146 	.pru_disconnect = udp_disconnect,
1147 	.pru_listen = pru_listen_notsupp,
1148 	.pru_peeraddr = in_setpeeraddr,
1149 	.pru_rcvd = pru_rcvd_notsupp,
1150 	.pru_rcvoob = pru_rcvoob_notsupp,
1151 	.pru_send = udp_send,
1152 	.pru_sense = pru_sense_null,
1153 	.pru_shutdown = udp_shutdown,
1154 	.pru_sockaddr = in_setsockaddr,
1155 	.pru_sosend = sosendudp,
1156 	.pru_soreceive = soreceive
1157 };
1158 
1159