Home | History | Annotate | Line # | Download | only in netinet
udp_usrreq.c revision 1.146.2.2
      1 /*	$NetBSD: udp_usrreq.c,v 1.146.2.2 2006/02/07 04:51:49 rpaulo Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
     34  *	The Regents of the University of California.  All rights reserved.
     35  *
     36  * Redistribution and use in source and binary forms, with or without
     37  * modification, are permitted provided that the following conditions
     38  * are met:
     39  * 1. Redistributions of source code must retain the above copyright
     40  *    notice, this list of conditions and the following disclaimer.
     41  * 2. Redistributions in binary form must reproduce the above copyright
     42  *    notice, this list of conditions and the following disclaimer in the
     43  *    documentation and/or other materials provided with the distribution.
     44  * 3. Neither the name of the University nor the names of its contributors
     45  *    may be used to endorse or promote products derived from this software
     46  *    without specific prior written permission.
     47  *
     48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  *
     60  *	@(#)udp_usrreq.c	8.6 (Berkeley) 5/23/95
     61  */
     62 
     63 #include <sys/cdefs.h>
     64 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.146.2.2 2006/02/07 04:51:49 rpaulo Exp $");
     65 
     66 #include "opt_inet.h"
     67 #include "opt_ipsec.h"
     68 #include "opt_inet_csum.h"
     69 #include "opt_ipkdb.h"
     70 #include "opt_mbuftrace.h"
     71 
     72 #include <sys/param.h>
     73 #include <sys/malloc.h>
     74 #include <sys/mbuf.h>
     75 #include <sys/protosw.h>
     76 #include <sys/socket.h>
     77 #include <sys/socketvar.h>
     78 #include <sys/errno.h>
     79 #include <sys/stat.h>
     80 #include <sys/systm.h>
     81 #include <sys/proc.h>
     82 #include <sys/domain.h>
     83 #include <sys/sysctl.h>
     84 
     85 #include <net/if.h>
     86 #include <net/route.h>
     87 
     88 #include <netinet/in.h>
     89 #include <netinet/in_systm.h>
     90 #include <netinet/in_var.h>
     91 #include <netinet/ip.h>
     92 #include <netinet/in_pcb.h>
     93 #include <netinet/ip_var.h>
     94 #include <netinet/ip_icmp.h>
     95 #include <netinet/udp.h>
     96 #include <netinet/udp_var.h>
     97 
     98 #ifdef IPSEC_NAT_T
     99 #include <netinet6/ipsec.h>
    100 #include <netinet6/esp.h>
    101 #endif
    102 
    103 #ifdef INET6
    104 #include <netinet/ip6.h>
    105 #include <netinet/icmp6.h>
    106 #include <netinet6/ip6_var.h>
    107 #include <netinet6/udp6_var.h>
    108 #include <netinet6/scope6_var.h>
    109 #endif
    110 
    111 #ifndef INET6
    112 /* always need ip6.h for IP6_EXTHDR_GET */
    113 #include <netinet/ip6.h>
    114 #endif
    115 
    116 #include "faith.h"
    117 #if defined(NFAITH) && NFAITH > 0
    118 #include <net/if_faith.h>
    119 #endif
    120 
    121 #include <machine/stdarg.h>
    122 
    123 #ifdef FAST_IPSEC
    124 #include <netipsec/ipsec.h>
    125 #include <netipsec/ipsec_var.h>			/* XXX ipsecstat namespace */
    126 #ifdef INET6
    127 #include <netipsec/ipsec6.h>
    128 #endif
    129 #endif	/* FAST_IPSEC*/
    130 
    131 #ifdef IPSEC
    132 #include <netinet6/ipsec.h>
    133 #include <netkey/key.h>
    134 #endif /*IPSEC*/
    135 
    136 #ifdef IPKDB
    137 #include <ipkdb/ipkdb.h>
    138 #endif
    139 
    140 /*
    141  * UDP protocol implementation.
    142  * Per RFC 768, August, 1980.
    143  */
    144 #ifndef	COMPAT_42
    145 int	udpcksum = 1;
    146 #else
    147 int	udpcksum = 0;		/* XXX */
    148 #endif
    149 int	udp_do_loopback_cksum = 0;
    150 
    151 struct	inpcbtable udbtable;
    152 struct	udpstat udpstat;
    153 
    154 #ifdef INET
    155 #ifdef IPSEC_NAT_T
    156 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
    157 	struct socket *);
    158 #endif
    159 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
    160 	struct socket *);
    161 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
    162 	struct mbuf **, int);
    163 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    164 #endif
    165 #ifdef INET6
    166 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
    167 	struct socket *);
    168 static int udp6_realinput (int, struct sockaddr_in6 *,
    169 	struct sockaddr_in6 *, struct mbuf *, int);
    170 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    171 #endif
    172 #ifdef INET
    173 static	void udp_notify (struct inpcb *, int);
    174 #endif
    175 
    176 #ifndef UDBHASHSIZE
    177 #define	UDBHASHSIZE	128
    178 #endif
    179 int	udbhashsize = UDBHASHSIZE;
    180 
    181 #ifdef MBUFTRACE
    182 struct mowner udp_mowner = { "udp" };
    183 struct mowner udp_rx_mowner = { "udp", "rx" };
    184 struct mowner udp_tx_mowner = { "udp", "tx" };
    185 #endif
    186 
    187 #ifdef UDP_CSUM_COUNTERS
    188 #include <sys/device.h>
    189 
    190 #if defined(INET)
    191 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    192     NULL, "udp", "hwcsum bad");
    193 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    194     NULL, "udp", "hwcsum ok");
    195 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    196     NULL, "udp", "hwcsum data");
    197 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    198     NULL, "udp", "swcsum");
    199 
    200 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
    201 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
    202 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
    203 EVCNT_ATTACH_STATIC(udp_swcsum);
    204 #endif /* defined(INET) */
    205 
    206 #if defined(INET6)
    207 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    208     NULL, "udp6", "hwcsum bad");
    209 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    210     NULL, "udp6", "hwcsum ok");
    211 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    212     NULL, "udp6", "hwcsum data");
    213 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    214     NULL, "udp6", "swcsum");
    215 
    216 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
    217 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
    218 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
    219 EVCNT_ATTACH_STATIC(udp6_swcsum);
    220 #endif /* defined(INET6) */
    221 
    222 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    223 
    224 #else
    225 
    226 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
    227 
    228 #endif /* UDP_CSUM_COUNTERS */
    229 
    230 void
    231 udp_init(void)
    232 {
    233 
    234 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
    235 
    236 	MOWNER_ATTACH(&udp_tx_mowner);
    237 	MOWNER_ATTACH(&udp_rx_mowner);
    238 	MOWNER_ATTACH(&udp_mowner);
    239 }
    240 
    241 /*
    242  * Checksum extended UDP header and data.
    243  */
    244 
    245 int
    246 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
    247     int iphlen, int len)
    248 {
    249 
    250 	switch (af) {
    251 #ifdef INET
    252 	case AF_INET:
    253 		return udp4_input_checksum(m, uh, iphlen, len);
    254 #endif
    255 #ifdef INET6
    256 	case AF_INET6:
    257 		return udp6_input_checksum(m, uh, iphlen, len);
    258 #endif
    259 	}
    260 #ifdef DIAGNOSTIC
    261 	panic("udp_input_checksum: unknown af %d", af);
    262 #endif
    263 	/* NOTREACHED */
    264 	return -1;
    265 }
    266 
    267 #ifdef INET
    268 
    269 /*
    270  * Checksum extended UDP header and data.
    271  */
    272 
    273 static int
    274 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
    275     int iphlen, int len)
    276 {
    277 
    278 	/*
    279 	 * XXX it's better to record and check if this mbuf is
    280 	 * already checked.
    281 	 */
    282 
    283 	if (uh->uh_sum == 0)
    284 		return 0;
    285 
    286 	switch (m->m_pkthdr.csum_flags &
    287 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
    288 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    289 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
    290 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
    291 		goto badcsum;
    292 
    293 	case M_CSUM_UDPv4|M_CSUM_DATA: {
    294 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
    295 
    296 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
    297 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
    298 			const struct ip *ip =
    299 			    mtod(m, const struct ip *);
    300 
    301 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
    302 			    ip->ip_dst.s_addr,
    303 			    htons(hw_csum + len + IPPROTO_UDP));
    304 		}
    305 		if ((hw_csum ^ 0xffff) != 0)
    306 			goto badcsum;
    307 		break;
    308 	}
    309 
    310 	case M_CSUM_UDPv4:
    311 		/* Checksum was okay. */
    312 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
    313 		break;
    314 
    315 	default:
    316 		/*
    317 		 * Need to compute it ourselves.  Maybe skip checksum
    318 		 * on loopback interfaces.
    319 		 */
    320 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
    321 				     IFF_LOOPBACK) ||
    322 				   udp_do_loopback_cksum)) {
    323 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
    324 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
    325 				goto badcsum;
    326 		}
    327 		break;
    328 	}
    329 
    330 	return 0;
    331 
    332 badcsum:
    333 	udpstat.udps_badsum++;
    334 	return -1;
    335 }
    336 
    337 void
    338 udp_input(struct mbuf *m, ...)
    339 {
    340 	va_list ap;
    341 	struct sockaddr_in src, dst;
    342 	struct ip *ip;
    343 	struct udphdr *uh;
    344 	int iphlen;
    345 	int len;
    346 	int n;
    347 	u_int16_t ip_len;
    348 
    349 	va_start(ap, m);
    350 	iphlen = va_arg(ap, int);
    351 	(void)va_arg(ap, int);		/* ignore value, advance ap */
    352 	va_end(ap);
    353 
    354 	MCLAIM(m, &udp_rx_mowner);
    355 	udpstat.udps_ipackets++;
    356 
    357 	/*
    358 	 * Get IP and UDP header together in first mbuf.
    359 	 */
    360 	ip = mtod(m, struct ip *);
    361 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
    362 	if (uh == NULL) {
    363 		udpstat.udps_hdrops++;
    364 		return;
    365 	}
    366 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    367 
    368 	/* destination port of 0 is illegal, based on RFC768. */
    369 	if (uh->uh_dport == 0)
    370 		goto bad;
    371 
    372 	/*
    373 	 * Make mbuf data length reflect UDP length.
    374 	 * If not enough data to reflect UDP length, drop.
    375 	 */
    376 	ip_len = ntohs(ip->ip_len);
    377 	len = ntohs((u_int16_t)uh->uh_ulen);
    378 	if (ip_len != iphlen + len) {
    379 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
    380 			udpstat.udps_badlen++;
    381 			goto bad;
    382 		}
    383 		m_adj(m, iphlen + len - ip_len);
    384 	}
    385 
    386 	/*
    387 	 * Checksum extended UDP header and data.
    388 	 */
    389 	if (udp4_input_checksum(m, uh, iphlen, len))
    390 		goto badcsum;
    391 
    392 	/* construct source and dst sockaddrs. */
    393 	bzero(&src, sizeof(src));
    394 	src.sin_family = AF_INET;
    395 	src.sin_len = sizeof(struct sockaddr_in);
    396 	bcopy(&ip->ip_src, &src.sin_addr, sizeof(src.sin_addr));
    397 	src.sin_port = uh->uh_sport;
    398 	bzero(&dst, sizeof(dst));
    399 	dst.sin_family = AF_INET;
    400 	dst.sin_len = sizeof(struct sockaddr_in);
    401 	bcopy(&ip->ip_dst, &dst.sin_addr, sizeof(dst.sin_addr));
    402 	dst.sin_port = uh->uh_dport;
    403 
    404 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
    405 		udpstat.udps_hdrops++;
    406 		return;
    407 	}
    408 #ifdef INET6
    409 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
    410 		struct sockaddr_in6 src6, dst6;
    411 
    412 		bzero(&src6, sizeof(src6));
    413 		src6.sin6_family = AF_INET6;
    414 		src6.sin6_len = sizeof(struct sockaddr_in6);
    415 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
    416 		bcopy(&ip->ip_src, &src6.sin6_addr.s6_addr[12],
    417 			sizeof(ip->ip_src));
    418 		src6.sin6_port = uh->uh_sport;
    419 		bzero(&dst6, sizeof(dst6));
    420 		dst6.sin6_family = AF_INET6;
    421 		dst6.sin6_len = sizeof(struct sockaddr_in6);
    422 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
    423 		bcopy(&ip->ip_dst, &dst6.sin6_addr.s6_addr[12],
    424 			sizeof(ip->ip_dst));
    425 		dst6.sin6_port = uh->uh_dport;
    426 
    427 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
    428 	}
    429 #endif
    430 
    431 	if (n == 0) {
    432 		if (m->m_flags & (M_BCAST | M_MCAST)) {
    433 			udpstat.udps_noportbcast++;
    434 			goto bad;
    435 		}
    436 		udpstat.udps_noport++;
    437 #ifdef IPKDB
    438 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
    439 				m, iphlen + sizeof(struct udphdr),
    440 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
    441 			/*
    442 			 * It was a debugger connect packet,
    443 			 * just drop it now
    444 			 */
    445 			goto bad;
    446 		}
    447 #endif
    448 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
    449 		m = NULL;
    450 	}
    451 
    452 bad:
    453 	if (m)
    454 		m_freem(m);
    455 	return;
    456 
    457 badcsum:
    458 	m_freem(m);
    459 }
    460 #endif
    461 
    462 #ifdef INET6
    463 static int
    464 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
    465 {
    466 
    467 	/*
    468 	 * XXX it's better to record and check if this mbuf is
    469 	 * already checked.
    470 	 */
    471 
    472 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
    473 		goto good;
    474 	}
    475 	if (uh->uh_sum == 0) {
    476 		udp6stat.udp6s_nosum++;
    477 		goto bad;
    478 	}
    479 
    480 	switch (m->m_pkthdr.csum_flags &
    481 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
    482 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    483 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
    484 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
    485 		udp6stat.udp6s_badsum++;
    486 		goto bad;
    487 
    488 #if 0 /* notyet */
    489 	case M_CSUM_UDPv6|M_CSUM_DATA:
    490 #endif
    491 
    492 	case M_CSUM_UDPv6:
    493 		/* Checksum was okay. */
    494 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
    495 		break;
    496 
    497 	default:
    498 		/*
    499 		 * Need to compute it ourselves.  Maybe skip checksum
    500 		 * on loopback interfaces.
    501 		 */
    502 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
    503 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
    504 			udp6stat.udp6s_badsum++;
    505 			goto bad;
    506 		}
    507 	}
    508 
    509 good:
    510 	return 0;
    511 bad:
    512 	return -1;
    513 }
    514 
    515 int
    516 udp6_input(struct mbuf **mp, int *offp, int proto)
    517 {
    518 	struct mbuf *m = *mp;
    519 	int off = *offp;
    520 	struct sockaddr_in6 src, dst;
    521 	struct ip6_hdr *ip6;
    522 	struct udphdr *uh;
    523 	u_int32_t plen, ulen;
    524 
    525 	ip6 = mtod(m, struct ip6_hdr *);
    526 
    527 #if defined(NFAITH) && 0 < NFAITH
    528 	if (faithprefix(&ip6->ip6_dst)) {
    529 		/* send icmp6 host unreach? */
    530 		m_freem(m);
    531 		return IPPROTO_DONE;
    532 	}
    533 #endif
    534 
    535 	udp6stat.udp6s_ipackets++;
    536 
    537 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
    538 	plen = m->m_pkthdr.len - off;
    539 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
    540 	if (uh == NULL) {
    541 		ip6stat.ip6s_tooshort++;
    542 		return IPPROTO_DONE;
    543 	}
    544 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    545 	ulen = ntohs((u_short)uh->uh_ulen);
    546 	/*
    547 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
    548 	 * iff payload length > 0xffff.
    549 	 */
    550 	if (ulen == 0 && plen > 0xffff)
    551 		ulen = plen;
    552 
    553 	if (plen != ulen) {
    554 		udp6stat.udp6s_badlen++;
    555 		goto bad;
    556 	}
    557 
    558 	/* destination port of 0 is illegal, based on RFC768. */
    559 	if (uh->uh_dport == 0)
    560 		goto bad;
    561 
    562 	/* Be proactive about malicious use of IPv4 mapped address */
    563 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
    564 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
    565 		/* XXX stat */
    566 		goto bad;
    567 	}
    568 
    569 	/*
    570 	 * Checksum extended UDP header and data.  Maybe skip checksum
    571 	 * on loopback interfaces.
    572 	 */
    573 	if (udp6_input_checksum(m, uh, off, ulen))
    574 		goto bad;
    575 
    576 	/*
    577 	 * Construct source and dst sockaddrs.
    578 	 */
    579 	bzero(&src, sizeof(src));
    580 	src.sin6_family = AF_INET6;
    581 	src.sin6_len = sizeof(struct sockaddr_in6);
    582 	src.sin6_addr = ip6->ip6_src;
    583 	src.sin6_port = uh->uh_sport;
    584 	bzero(&dst, sizeof(dst));
    585 	dst.sin6_family = AF_INET6;
    586 	dst.sin6_len = sizeof(struct sockaddr_in6);
    587 	dst.sin6_addr = ip6->ip6_dst;
    588 	dst.sin6_port = uh->uh_dport;
    589 
    590 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
    591 		if (m->m_flags & M_MCAST) {
    592 			udp6stat.udp6s_noportmcast++;
    593 			goto bad;
    594 		}
    595 		udp6stat.udp6s_noport++;
    596 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
    597 		m = NULL;
    598 	}
    599 
    600 bad:
    601 	if (m)
    602 		m_freem(m);
    603 	return IPPROTO_DONE;
    604 }
    605 #endif
    606 
    607 #ifdef INET
    608 static void
    609 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
    610 	struct sockaddr *src, struct socket *so)
    611 {
    612 	struct mbuf *opts = NULL;
    613 	struct mbuf *n;
    614 	struct inpcb *inp = NULL;
    615 
    616 	if (!so)
    617 		return;
    618 	switch (so->so_proto->pr_domain->dom_family) {
    619 	case AF_INET:
    620 		inp = sotoinpcb(so);
    621 		break;
    622 #ifdef INET6
    623 	case AF_INET6:
    624 		break;
    625 #endif
    626 	default:
    627 		return;
    628 	}
    629 
    630 #if defined(IPSEC) || defined(FAST_IPSEC)
    631 	/* check AH/ESP integrity. */
    632 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
    633 		ipsecstat.in_polvio++;
    634 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
    635 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
    636 			    0, 0);
    637 		return;
    638 	}
    639 #endif /*IPSEC*/
    640 
    641 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
    642 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
    643 			 || so->so_options & SO_TIMESTAMP)) {
    644 			struct ip *ip = mtod(n, struct ip *);
    645 			ip_savecontrol(inp, &opts, ip, n);
    646 		}
    647 
    648 		m_adj(n, off);
    649 		if (sbappendaddr(&so->so_rcv, src, n,
    650 				opts) == 0) {
    651 			m_freem(n);
    652 			if (opts)
    653 				m_freem(opts);
    654 			so->so_rcv.sb_overflowed++;
    655 			udpstat.udps_fullsock++;
    656 		} else
    657 			sorwakeup(so);
    658 	}
    659 }
    660 #endif
    661 
    662 #ifdef INET6
    663 static void
    664 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
    665 	struct sockaddr *src, struct socket *so)
    666 {
    667 	struct mbuf *opts = NULL;
    668 	struct mbuf *n;
    669 	struct in6pcb *in6p = NULL;
    670 
    671 	if (!so)
    672 		return;
    673 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
    674 		return;
    675 	in6p = sotoin6pcb(so);
    676 
    677 #if defined(IPSEC) || defined(FAST_IPSEC)
    678 	/* check AH/ESP integrity. */
    679 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
    680 		ipsec6stat.in_polvio++;
    681 		if ((n = m_copy(m, 0, M_COPYALL)) != NULL)
    682 			icmp6_error(n, ICMP6_DST_UNREACH,
    683 			    ICMP6_DST_UNREACH_ADMIN, 0);
    684 		return;
    685 	}
    686 #endif /*IPSEC*/
    687 
    688 	if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
    689 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
    690 			  || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
    691 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
    692 			ip6_savecontrol(in6p, &opts, ip6, n);
    693 		}
    694 
    695 		m_adj(n, off);
    696 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
    697 			m_freem(n);
    698 			if (opts)
    699 				m_freem(opts);
    700 			so->so_rcv.sb_overflowed++;
    701 			udp6stat.udp6s_fullsock++;
    702 		} else
    703 			sorwakeup(so);
    704 	}
    705 }
    706 #endif
    707 
    708 #ifdef INET
    709 static int
    710 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
    711 	struct mbuf **mp, int off /* offset of udphdr */)
    712 {
    713 	u_int16_t *sport, *dport;
    714 	int rcvcnt;
    715 	struct in_addr *src4, *dst4;
    716 	struct inpcb *inp;
    717 	struct mbuf *m = *mp;
    718 
    719 	rcvcnt = 0;
    720 	off += sizeof(struct udphdr);	/* now, offset of payload */
    721 
    722 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
    723 		goto bad;
    724 
    725 	src4 = &src->sin_addr;
    726 	sport = &src->sin_port;
    727 	dst4 = &dst->sin_addr;
    728 	dport = &dst->sin_port;
    729 
    730 	if (IN_MULTICAST(dst4->s_addr) ||
    731 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
    732 		/*
    733 		 * Deliver a multicast or broadcast datagram to *all* sockets
    734 		 * for which the local and remote addresses and ports match
    735 		 * those of the incoming datagram.  This allows more than
    736 		 * one process to receive multi/broadcasts on the same port.
    737 		 * (This really ought to be done for unicast datagrams as
    738 		 * well, but that would cause problems with existing
    739 		 * applications that open both address-specific sockets and
    740 		 * a wildcard socket listening to the same port -- they would
    741 		 * end up receiving duplicates of every unicast datagram.
    742 		 * Those applications open the multiple sockets to overcome an
    743 		 * inadequacy of the UDP socket interface, but for backwards
    744 		 * compatibility we avoid the problem here rather than
    745 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    746 		 */
    747 
    748 		/*
    749 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    750 		 * we need udpiphdr for IPsec processing so we do that later.
    751 		 */
    752 		/*
    753 		 * Locate pcb(s) for datagram.
    754 		 */
    755 		CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
    756 			if (inp->inp_af != AF_INET)
    757 				continue;
    758 
    759 			if (inp->inp_lport != *dport)
    760 				continue;
    761 			if (!in_nullhost(inp->inp_laddr)) {
    762 				if (!in_hosteq(inp->inp_laddr, *dst4))
    763 					continue;
    764 			}
    765 			if (!in_nullhost(inp->inp_faddr)) {
    766 				if (!in_hosteq(inp->inp_faddr, *src4) ||
    767 				    inp->inp_fport != *sport)
    768 					continue;
    769 			}
    770 
    771 			udp4_sendup(m, off, (struct sockaddr *)src,
    772 				inp->inp_socket);
    773 			rcvcnt++;
    774 
    775 			/*
    776 			 * Don't look for additional matches if this one does
    777 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    778 			 * socket options set.  This heuristic avoids searching
    779 			 * through all pcbs in the common case of a non-shared
    780 			 * port.  It assumes that an application will never
    781 			 * clear these options after setting them.
    782 			 */
    783 			if ((inp->inp_socket->so_options &
    784 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    785 				break;
    786 		}
    787 	} else {
    788 		/*
    789 		 * Locate pcb for datagram.
    790 		 */
    791 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4, *dport);
    792 		if (inp == 0) {
    793 			++udpstat.udps_pcbhashmiss;
    794 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
    795 			if (inp == 0)
    796 				return rcvcnt;
    797 		}
    798 
    799 #ifdef IPSEC_NAT_T
    800 		/* Handle ESP over UDP */
    801 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
    802 			struct sockaddr *sa = (struct sockaddr *)src;
    803 
    804 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
    805 			case -1: 	/* Error, m was freeed */
    806 				rcvcnt = -1;
    807 				goto bad;
    808 				break;
    809 
    810 			case 1:		/* ESP over UDP */
    811 				rcvcnt++;
    812 				goto bad;
    813 				break;
    814 
    815 			case 0: 	/* plain UDP */
    816 			default: 	/* Unexpected */
    817 				/*
    818 				 * Normal UDP processing will take place
    819 				 * m may have changed.
    820 				 */
    821 				m = *mp;
    822 				break;
    823 			}
    824 		}
    825 #endif
    826 
    827 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
    828 		rcvcnt++;
    829 	}
    830 
    831 bad:
    832 	return rcvcnt;
    833 }
    834 #endif
    835 
    836 #ifdef INET6
    837 static int
    838 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
    839 	struct mbuf *m, int off)
    840 {
    841 	u_int16_t sport, dport;
    842 	int rcvcnt;
    843 	struct in6_addr src6, *dst6;
    844 	const struct in_addr *dst4;
    845 	struct inpcb *inp;
    846 
    847 	rcvcnt = 0;
    848 	off += sizeof(struct udphdr);	/* now, offset of payload */
    849 
    850 	if (af != AF_INET && af != AF_INET6)
    851 		goto bad;
    852 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
    853 		goto bad;
    854 
    855 	src6 = src->sin6_addr;
    856 	if (sa6_recoverscope(src) != 0) {
    857 		/* XXX: should be impossible. */
    858 		goto bad;
    859 	}
    860 	sport = src->sin6_port;
    861 
    862 	dport = dst->sin6_port;
    863 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
    864 	dst6 = &dst->sin6_addr;
    865 
    866 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
    867 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
    868 		/*
    869 		 * Deliver a multicast or broadcast datagram to *all* sockets
    870 		 * for which the local and remote addresses and ports match
    871 		 * those of the incoming datagram.  This allows more than
    872 		 * one process to receive multi/broadcasts on the same port.
    873 		 * (This really ought to be done for unicast datagrams as
    874 		 * well, but that would cause problems with existing
    875 		 * applications that open both address-specific sockets and
    876 		 * a wildcard socket listening to the same port -- they would
    877 		 * end up receiving duplicates of every unicast datagram.
    878 		 * Those applications open the multiple sockets to overcome an
    879 		 * inadequacy of the UDP socket interface, but for backwards
    880 		 * compatibility we avoid the problem here rather than
    881 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    882 		 */
    883 
    884 		/*
    885 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    886 		 * we need udpiphdr for IPsec processing so we do that later.
    887 		 */
    888 		/*
    889 		 * Locate pcb(s) for datagram.
    890 		 */
    891 		CIRCLEQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue) {
    892 			if (inp->inp_af != AF_INET6)
    893 				continue;
    894 
    895 			if (inp->inp_lport != dport)
    896 				continue;
    897 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
    898 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
    899 				    dst6))
    900 					continue;
    901 			} else {
    902 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
    903 				    (inp->inp_flags & IN6P_IPV6_V6ONLY))
    904 					continue;
    905 			}
    906 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
    907 				if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
    908 				    &src6) || inp->inp_fport != sport)
    909 					continue;
    910 			} else {
    911 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
    912 				    (inp->inp_flags & IN6P_IPV6_V6ONLY))
    913 					continue;
    914 			}
    915 
    916 			udp6_sendup(m, off, (struct sockaddr *)src,
    917 				inp->inp_socket);
    918 			rcvcnt++;
    919 
    920 			/*
    921 			 * Don't look for additional matches if this one does
    922 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    923 			 * socket options set.  This heuristic avoids searching
    924 			 * through all pcbs in the common case of a non-shared
    925 			 * port.  It assumes that an application will never
    926 			 * clear these options after setting them.
    927 			 */
    928 			if ((inp->inp_socket->so_options &
    929 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    930 				break;
    931 		}
    932 	} else {
    933 		/*
    934 		 * Locate pcb for datagram.
    935 		 */
    936 		inp = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
    937 		    dport, 0);
    938 		if (inp == 0) {
    939 			++udpstat.udps_pcbhashmiss;
    940 			inp = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
    941 			if (inp == 0)
    942 				return rcvcnt;
    943 		}
    944 
    945 		udp6_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
    946 		rcvcnt++;
    947 	}
    948 
    949 bad:
    950 	return rcvcnt;
    951 }
    952 #endif
    953 
    954 #ifdef INET
    955 /*
    956  * Notify a udp user of an asynchronous error;
    957  * just wake up so that he can collect error status.
    958  */
    959 static void
    960 udp_notify(struct inpcb *inp, int errno)
    961 {
    962 	inp->inp_socket->so_error = errno;
    963 	sorwakeup(inp->inp_socket);
    964 	sowwakeup(inp->inp_socket);
    965 }
    966 
    967 void *
    968 udp_ctlinput(int cmd, struct sockaddr *sa, void *v)
    969 {
    970 	struct ip *ip = v;
    971 	struct udphdr *uh;
    972 	void (*notify)(struct inpcb *, int) = udp_notify;
    973 	int errno;
    974 
    975 	if (sa->sa_family != AF_INET
    976 	 || sa->sa_len != sizeof(struct sockaddr_in))
    977 		return NULL;
    978 	if ((unsigned)cmd >= PRC_NCMDS)
    979 		return NULL;
    980 	errno = inetctlerrmap[cmd];
    981 	if (PRC_IS_REDIRECT(cmd))
    982 		notify = in_rtchange, ip = 0;
    983 	else if (cmd == PRC_HOSTDEAD)
    984 		ip = 0;
    985 	else if (errno == 0)
    986 		return NULL;
    987 	if (ip) {
    988 		uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
    989 		in_pcbnotify(&udbtable, satosin(sa)->sin_addr, uh->uh_dport,
    990 		    ip->ip_src, uh->uh_sport, errno, notify);
    991 
    992 		/* XXX mapped address case */
    993 	} else
    994 		in_pcbnotifyall(&udbtable, satosin(sa)->sin_addr, errno,
    995 		    notify);
    996 	return NULL;
    997 }
    998 
    999 int
   1000 udp_ctloutput(op, so, level, optname, mp)
   1001 	int op;
   1002 	struct socket *so;
   1003 	int level, optname;
   1004 	struct mbuf **mp;
   1005 {
   1006 	int s;
   1007 	int error = 0;
   1008 	struct mbuf *m;
   1009 	struct inpcb *inp;
   1010 	int family;
   1011 
   1012 	family = so->so_proto->pr_domain->dom_family;
   1013 
   1014 	s = splsoftnet();
   1015 	switch (family) {
   1016 #ifdef INET
   1017 	case PF_INET:
   1018 		if (level != IPPROTO_UDP) {
   1019 			error = ip_ctloutput(op, so, level, optname, mp);
   1020 			goto end;
   1021 		}
   1022 		break;
   1023 #endif
   1024 #ifdef INET6
   1025 	case PF_INET6:
   1026 		if (level != IPPROTO_UDP) {
   1027 			error = ip6_ctloutput(op, so, level, optname, mp);
   1028 			goto end;
   1029 		}
   1030 		break;
   1031 #endif
   1032 	default:
   1033 		error = EAFNOSUPPORT;
   1034 		goto end;
   1035 		break;
   1036 	}
   1037 
   1038 
   1039 	switch (op) {
   1040 	case PRCO_SETOPT:
   1041 		m = *mp;
   1042 		inp = sotoinpcb(so);
   1043 
   1044 		switch (optname) {
   1045 		case UDP_ENCAP:
   1046 			if (m == NULL || m->m_len < sizeof (int)) {
   1047 				error = EINVAL;
   1048 				goto end;
   1049 			}
   1050 
   1051 			switch(*mtod(m, int *)) {
   1052 #ifdef IPSEC_NAT_T
   1053 			case 0:
   1054 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1055 				break;
   1056 
   1057 			case UDP_ENCAP_ESPINUDP:
   1058 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1059 				inp->inp_flags |= INP_ESPINUDP;
   1060 				break;
   1061 
   1062 			case UDP_ENCAP_ESPINUDP_NON_IKE:
   1063 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1064 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
   1065 				break;
   1066 #endif
   1067 			default:
   1068 				error = EINVAL;
   1069 				goto end;
   1070 				break;
   1071 			}
   1072 			break;
   1073 
   1074 		default:
   1075 			error = ENOPROTOOPT;
   1076 			goto end;
   1077 			break;
   1078 		}
   1079 		break;
   1080 
   1081 	default:
   1082 		error = EINVAL;
   1083 		goto end;
   1084 		break;
   1085 	}
   1086 
   1087 end:
   1088 	splx(s);
   1089 	return error;
   1090 }
   1091 
   1092 
   1093 int
   1094 udp_output(struct mbuf *m, ...)
   1095 {
   1096 	struct inpcb *inp;
   1097 	struct udpiphdr *ui;
   1098 	struct route *ro;
   1099 	int len = m->m_pkthdr.len;
   1100 	int error = 0;
   1101 	va_list ap;
   1102 
   1103 	MCLAIM(m, &udp_tx_mowner);
   1104 	va_start(ap, m);
   1105 	inp = va_arg(ap, struct inpcb *);
   1106 	va_end(ap);
   1107 
   1108 	/*
   1109 	 * Calculate data length and get a mbuf
   1110 	 * for UDP and IP headers.
   1111 	 */
   1112 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
   1113 	if (m == 0) {
   1114 		error = ENOBUFS;
   1115 		goto release;
   1116 	}
   1117 
   1118 	/*
   1119 	 * Compute the packet length of the IP header, and
   1120 	 * punt if the length looks bogus.
   1121 	 */
   1122 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
   1123 		error = EMSGSIZE;
   1124 		goto release;
   1125 	}
   1126 
   1127 	/*
   1128 	 * Fill in mbuf with extended UDP header
   1129 	 * and addresses and length put into network format.
   1130 	 */
   1131 	ui = mtod(m, struct udpiphdr *);
   1132 	ui->ui_pr = IPPROTO_UDP;
   1133 	ui->ui_src = inp->inp_laddr;
   1134 	ui->ui_dst = inp->inp_faddr;
   1135 	ui->ui_sport = inp->inp_lport;
   1136 	ui->ui_dport = inp->inp_fport;
   1137 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
   1138 
   1139 	ro = &inp->inp_route;
   1140 
   1141 	/*
   1142 	 * Set up checksum and output datagram.
   1143 	 */
   1144 	if (udpcksum) {
   1145 		/*
   1146 		 * XXX Cache pseudo-header checksum part for
   1147 		 * XXX "connected" UDP sockets.
   1148 		 */
   1149 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
   1150 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
   1151 		    sizeof(struct udphdr) + IPPROTO_UDP));
   1152 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
   1153 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
   1154 	} else
   1155 		ui->ui_sum = 0;
   1156 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
   1157 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
   1158 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
   1159 	udpstat.udps_opackets++;
   1160 
   1161 	return (ip_output(m, inp->inp_options, ro,
   1162 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
   1163 	    inp->inp_moptions, inp->inp_socket));
   1164 
   1165 release:
   1166 	m_freem(m);
   1167 	return (error);
   1168 }
   1169 
   1170 int	udp_sendspace = 9216;		/* really max datagram size */
   1171 int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
   1172 					/* 40 1K datagrams */
   1173 
   1174 /*ARGSUSED*/
   1175 int
   1176 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
   1177 	struct mbuf *control, struct lwp *l)
   1178 {
   1179 	struct inpcb *inp;
   1180 	struct proc *p;
   1181 	int s;
   1182 	int error = 0;
   1183 
   1184 	p = l ? l->l_proc : NULL;
   1185 	if (req == PRU_CONTROL)
   1186 		return (in_control(so, (long)m, (caddr_t)nam,
   1187 		    (struct ifnet *)control, p));
   1188 
   1189 	if (req == PRU_PURGEIF) {
   1190 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
   1191 		in_purgeif((struct ifnet *)control);
   1192 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
   1193 		return (0);
   1194 	}
   1195 
   1196 	s = splsoftnet();
   1197 	inp = sotoinpcb(so);
   1198 #ifdef DIAGNOSTIC
   1199 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
   1200 		panic("udp_usrreq: unexpected control mbuf");
   1201 #endif
   1202 	if (inp == 0 && req != PRU_ATTACH) {
   1203 		error = EINVAL;
   1204 		goto release;
   1205 	}
   1206 
   1207 	/*
   1208 	 * Note: need to block udp_input while changing
   1209 	 * the udp pcb queue and/or pcb addresses.
   1210 	 */
   1211 	switch (req) {
   1212 
   1213 	case PRU_ATTACH:
   1214 		if (inp != 0) {
   1215 			error = EISCONN;
   1216 			break;
   1217 		}
   1218 #ifdef MBUFTRACE
   1219 		so->so_mowner = &udp_mowner;
   1220 		so->so_rcv.sb_mowner = &udp_rx_mowner;
   1221 		so->so_snd.sb_mowner = &udp_tx_mowner;
   1222 #endif
   1223 		if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
   1224 			error = soreserve(so, udp_sendspace, udp_recvspace);
   1225 			if (error)
   1226 				break;
   1227 		}
   1228 		error = in_pcballoc(so, &udbtable);
   1229 		if (error)
   1230 			break;
   1231 		inp = sotoinpcb(so);
   1232 		inp->inp_ip.ip_ttl = ip_defttl;
   1233 		break;
   1234 
   1235 	case PRU_DETACH:
   1236 		in_pcbdetach(inp);
   1237 		break;
   1238 
   1239 	case PRU_BIND:
   1240 		error = in_pcbbind(inp, nam, p);
   1241 		break;
   1242 
   1243 	case PRU_LISTEN:
   1244 		error = EOPNOTSUPP;
   1245 		break;
   1246 
   1247 	case PRU_CONNECT:
   1248 		error = in_pcbconnect(inp, nam, p);
   1249 		if (error)
   1250 			break;
   1251 		soisconnected(so);
   1252 		break;
   1253 
   1254 	case PRU_CONNECT2:
   1255 		error = EOPNOTSUPP;
   1256 		break;
   1257 
   1258 	case PRU_DISCONNECT:
   1259 		/*soisdisconnected(so);*/
   1260 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
   1261 		in_pcbdisconnect(inp);
   1262 		inp->inp_laddr = zeroin_addr;		/* XXX */
   1263 		in_pcbstate(inp, INP_BOUND);		/* XXX */
   1264 		break;
   1265 
   1266 	case PRU_SHUTDOWN:
   1267 		socantsendmore(so);
   1268 		break;
   1269 
   1270 	case PRU_RCVD:
   1271 		error = EOPNOTSUPP;
   1272 		break;
   1273 
   1274 	case PRU_SEND:
   1275 		if (control && control->m_len) {
   1276 			m_freem(control);
   1277 			m_freem(m);
   1278 			error = EINVAL;
   1279 			break;
   1280 		}
   1281 	{
   1282 		struct in_addr laddr;			/* XXX */
   1283 
   1284 		if (nam) {
   1285 			laddr = inp->inp_laddr;		/* XXX */
   1286 			if ((so->so_state & SS_ISCONNECTED) != 0) {
   1287 				error = EISCONN;
   1288 				goto die;
   1289 			}
   1290 			error = in_pcbconnect(inp, nam, p);
   1291 			if (error)
   1292 				goto die;
   1293 		} else {
   1294 			if ((so->so_state & SS_ISCONNECTED) == 0) {
   1295 				error = ENOTCONN;
   1296 				goto die;
   1297 			}
   1298 		}
   1299 		error = udp_output(m, inp);
   1300 		m = NULL;
   1301 		if (nam) {
   1302 			in_pcbdisconnect(inp);
   1303 			inp->inp_laddr = laddr;		/* XXX */
   1304 			in_pcbstate(inp, INP_BOUND);	/* XXX */
   1305 		}
   1306 	  die:
   1307 		if (m)
   1308 			m_freem(m);
   1309 	}
   1310 		break;
   1311 
   1312 	case PRU_SENSE:
   1313 		/*
   1314 		 * stat: don't bother with a blocksize.
   1315 		 */
   1316 		splx(s);
   1317 		return (0);
   1318 
   1319 	case PRU_RCVOOB:
   1320 		error =  EOPNOTSUPP;
   1321 		break;
   1322 
   1323 	case PRU_SENDOOB:
   1324 		m_freem(control);
   1325 		m_freem(m);
   1326 		error =  EOPNOTSUPP;
   1327 		break;
   1328 
   1329 	case PRU_SOCKADDR:
   1330 		in_setsockaddr(inp, nam);
   1331 		break;
   1332 
   1333 	case PRU_PEERADDR:
   1334 		in_setpeeraddr(inp, nam);
   1335 		break;
   1336 
   1337 	default:
   1338 		panic("udp_usrreq");
   1339 	}
   1340 
   1341 release:
   1342 	splx(s);
   1343 	return (error);
   1344 }
   1345 
   1346 /*
   1347  * Sysctl for udp variables.
   1348  */
   1349 SYSCTL_SETUP(sysctl_net_inet_udp_setup, "sysctl net.inet.udp subtree setup")
   1350 {
   1351 
   1352 	sysctl_createv(clog, 0, NULL, NULL,
   1353 		       CTLFLAG_PERMANENT,
   1354 		       CTLTYPE_NODE, "net", NULL,
   1355 		       NULL, 0, NULL, 0,
   1356 		       CTL_NET, CTL_EOL);
   1357 	sysctl_createv(clog, 0, NULL, NULL,
   1358 		       CTLFLAG_PERMANENT,
   1359 		       CTLTYPE_NODE, "inet", NULL,
   1360 		       NULL, 0, NULL, 0,
   1361 		       CTL_NET, PF_INET, CTL_EOL);
   1362 	sysctl_createv(clog, 0, NULL, NULL,
   1363 		       CTLFLAG_PERMANENT,
   1364 		       CTLTYPE_NODE, "udp",
   1365 		       SYSCTL_DESCR("UDPv4 related settings"),
   1366 		       NULL, 0, NULL, 0,
   1367 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
   1368 
   1369 	sysctl_createv(clog, 0, NULL, NULL,
   1370 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1371 		       CTLTYPE_INT, "checksum",
   1372 		       SYSCTL_DESCR("Compute UDP checksums"),
   1373 		       NULL, 0, &udpcksum, 0,
   1374 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
   1375 		       CTL_EOL);
   1376 	sysctl_createv(clog, 0, NULL, NULL,
   1377 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1378 		       CTLTYPE_INT, "sendspace",
   1379 		       SYSCTL_DESCR("Default UDP send buffer size"),
   1380 		       NULL, 0, &udp_sendspace, 0,
   1381 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
   1382 		       CTL_EOL);
   1383 	sysctl_createv(clog, 0, NULL, NULL,
   1384 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1385 		       CTLTYPE_INT, "recvspace",
   1386 		       SYSCTL_DESCR("Default UDP receive buffer size"),
   1387 		       NULL, 0, &udp_recvspace, 0,
   1388 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
   1389 		       CTL_EOL);
   1390 	sysctl_createv(clog, 0, NULL, NULL,
   1391 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1392 		       CTLTYPE_INT, "do_loopback_cksum",
   1393 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
   1394 		       NULL, 0, &udp_do_loopback_cksum, 0,
   1395 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
   1396 		       CTL_EOL);
   1397 	sysctl_createv(clog, 0, NULL, NULL,
   1398 		       CTLFLAG_PERMANENT,
   1399 		       CTLTYPE_STRUCT, "pcblist",
   1400 		       SYSCTL_DESCR("UDP protocol control block list"),
   1401 		       sysctl_inpcblist, 0, &udbtable, 0,
   1402 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
   1403 		       CTL_EOL);
   1404 	sysctl_createv(clog, 0, NULL, NULL,
   1405 		       CTLFLAG_PERMANENT,
   1406 		       CTLTYPE_STRUCT, "stats",
   1407 		       SYSCTL_DESCR("UDP statistics"),
   1408 		       NULL, 0, &udpstat, sizeof(udpstat),
   1409 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
   1410 		       CTL_EOL);
   1411 }
   1412 #endif
   1413 
   1414 #if (defined INET && defined IPSEC_NAT_T)
   1415 /*
   1416  * Returns:
   1417  * 1 if the packet was processed
   1418  * 0 if normal UDP processing should take place
   1419  * -1 if an error occurent and m was freed
   1420  */
   1421 static int
   1422 udp4_espinudp(mp, off, src, so)
   1423 	struct mbuf **mp;
   1424 	int off;
   1425 	struct sockaddr *src;
   1426 	struct socket *so;
   1427 {
   1428 	size_t len;
   1429 	caddr_t data;
   1430 	struct inpcb *inp;
   1431 	size_t skip = 0;
   1432 	size_t minlen;
   1433 	size_t iphdrlen;
   1434 	struct ip *ip;
   1435 	struct mbuf *n;
   1436 	struct m_tag *tag;
   1437 	struct udphdr *udphdr;
   1438 	u_int16_t sport, dport;
   1439 	struct mbuf *m = *mp;
   1440 
   1441 	/*
   1442 	 * Collapse the mbuf chain if the first mbuf is too short
   1443 	 * The longest case is: UDP + non ESP marker + ESP
   1444 	 */
   1445 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
   1446 	if (minlen > m->m_pkthdr.len)
   1447 		minlen = m->m_pkthdr.len;
   1448 
   1449 	if (m->m_len < minlen) {
   1450 		if ((*mp = m_pullup(m, minlen)) == NULL) {
   1451 			printf("udp4_espinudp: m_pullup failed\n");
   1452 			return -1;
   1453 		}
   1454 		m = *mp;
   1455 	}
   1456 
   1457 	len = m->m_len - off;
   1458 	data = mtod(m, caddr_t) + off;
   1459 	inp = sotoinpcb(so);
   1460 
   1461 	/* Ignore keepalive packets */
   1462 	if ((len == 1) && (data[0] == '\xff')) {
   1463 		return 1;
   1464 	}
   1465 
   1466 	/*
   1467 	 * Check that the payload is long enough to hold
   1468 	 * an ESP header and compute the length of encapsulation
   1469 	 * header to remove
   1470 	 */
   1471 	if (inp->inp_flags & INP_ESPINUDP) {
   1472 		u_int32_t *st = (u_int32_t *)data;
   1473 
   1474 		if ((len <= sizeof(struct esp)) || (*st == 0))
   1475 			return 0; /* Normal UDP processing */
   1476 
   1477 		skip = sizeof(struct udphdr);
   1478 	}
   1479 
   1480 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
   1481 		u_int32_t *st = (u_int32_t *)data;
   1482 
   1483 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
   1484 		    || ((st[0] | st[1]) != 0))
   1485 			return 0; /* Normal UDP processing */
   1486 
   1487 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
   1488 	}
   1489 
   1490 	/*
   1491 	 * Get the UDP ports. They are handled in network
   1492 	 * order everywhere in IPSEC_NAT_T code.
   1493 	 */
   1494 	udphdr = (struct udphdr *)(data - skip);
   1495 	sport = udphdr->uh_sport;
   1496 	dport = udphdr->uh_dport;
   1497 
   1498 	/*
   1499 	 * Remove the UDP header (and possibly the non ESP marker)
   1500 	 * IP header lendth is iphdrlen
   1501 	 * Before:
   1502 	 *   <--- off --->
   1503 	 *   +----+------+-----+
   1504 	 *   | IP |  UDP | ESP |
   1505 	 *   +----+------+-----+
   1506 	 *        <-skip->
   1507 	 * After:
   1508 	 *          +----+-----+
   1509 	 *          | IP | ESP |
   1510 	 *          +----+-----+
   1511 	 *   <-skip->
   1512 	 */
   1513 	iphdrlen = off - sizeof(struct udphdr);
   1514 	memmove(mtod(m, caddr_t) + skip, mtod(m, caddr_t), iphdrlen);
   1515 	m_adj(m, skip);
   1516 
   1517 	ip = mtod(m, struct ip *);
   1518 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
   1519 	ip->ip_p = IPPROTO_ESP;
   1520 
   1521 	/*
   1522 	 * Copy the mbuf to avoid multiple free, as both
   1523 	 * esp4_input (which we call) and udp_input (which
   1524 	 * called us) free the mbuf.
   1525 	 */
   1526 	if ((n = m_dup(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
   1527 		printf("udp4_espinudp: m_dup failed\n");
   1528 		return 0;
   1529 	}
   1530 
   1531 	/*
   1532 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
   1533 	 * the source UDP port. This is required if we want
   1534 	 * to select the right SPD for multiple hosts behind
   1535 	 * same NAT
   1536 	 */
   1537 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
   1538 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
   1539 		printf("udp4_espinudp: m_tag_get failed\n");
   1540 		m_freem(n);
   1541 		return 0;
   1542 	}
   1543 	((u_int16_t *)(tag + 1))[0] = sport;
   1544 	((u_int16_t *)(tag + 1))[1] = dport;
   1545 	m_tag_prepend(n, tag);
   1546 
   1547 	esp4_input(n, iphdrlen);
   1548 
   1549 	/* We handled it, it shoudln't be handled by UDP */
   1550 	return 1;
   1551 }
   1552 #endif
   1553