Home | History | Annotate | Line # | Download | only in netinet
udp_usrreq.c revision 1.196
      1 /*	$NetBSD: udp_usrreq.c,v 1.196 2014/05/19 02:51:25 rmind 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 1.196 2014/05/19 02:51:25 rmind Exp $");
     65 
     66 #include "opt_inet.h"
     67 #include "opt_compat_netbsd.h"
     68 #include "opt_ipsec.h"
     69 #include "opt_inet_csum.h"
     70 #include "opt_ipkdb.h"
     71 #include "opt_mbuftrace.h"
     72 
     73 #include <sys/param.h>
     74 #include <sys/malloc.h>
     75 #include <sys/mbuf.h>
     76 #include <sys/once.h>
     77 #include <sys/protosw.h>
     78 #include <sys/socket.h>
     79 #include <sys/socketvar.h>
     80 #include <sys/errno.h>
     81 #include <sys/stat.h>
     82 #include <sys/systm.h>
     83 #include <sys/proc.h>
     84 #include <sys/domain.h>
     85 #include <sys/sysctl.h>
     86 
     87 #include <net/if.h>
     88 #include <net/route.h>
     89 
     90 #include <netinet/in.h>
     91 #include <netinet/in_systm.h>
     92 #include <netinet/in_var.h>
     93 #include <netinet/ip.h>
     94 #include <netinet/in_pcb.h>
     95 #include <netinet/ip_var.h>
     96 #include <netinet/ip_icmp.h>
     97 #include <netinet/udp.h>
     98 #include <netinet/udp_var.h>
     99 #include <netinet/udp_private.h>
    100 
    101 #ifdef INET6
    102 #include <netinet/ip6.h>
    103 #include <netinet/icmp6.h>
    104 #include <netinet6/ip6_var.h>
    105 #include <netinet6/ip6_private.h>
    106 #include <netinet6/in6_pcb.h>
    107 #include <netinet6/udp6_var.h>
    108 #include <netinet6/udp6_private.h>
    109 #include <netinet6/scope6_var.h>
    110 #endif
    111 
    112 #ifndef INET6
    113 /* always need ip6.h for IP6_EXTHDR_GET */
    114 #include <netinet/ip6.h>
    115 #endif
    116 
    117 #include "faith.h"
    118 #if defined(NFAITH) && NFAITH > 0
    119 #include <net/if_faith.h>
    120 #endif
    121 
    122 #ifdef IPSEC
    123 #include <netipsec/ipsec.h>
    124 #include <netipsec/ipsec_var.h>
    125 #include <netipsec/ipsec_private.h>
    126 #include <netipsec/esp.h>
    127 #ifdef INET6
    128 #include <netipsec/ipsec6.h>
    129 #endif
    130 #endif	/* IPSEC */
    131 
    132 #ifdef COMPAT_50
    133 #include <compat/sys/socket.h>
    134 #endif
    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 int	udpcksum = 1;
    145 int	udp_do_loopback_cksum = 0;
    146 
    147 struct	inpcbtable udbtable;
    148 
    149 percpu_t *udpstat_percpu;
    150 
    151 #ifdef INET
    152 #ifdef IPSEC
    153 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
    154 	struct socket *);
    155 #endif
    156 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
    157 	struct socket *);
    158 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
    159 	struct mbuf **, int);
    160 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    161 #endif
    162 #ifdef INET6
    163 static void udp6_sendup (struct mbuf *, int, struct sockaddr *,
    164 	struct socket *);
    165 static int udp6_realinput (int, struct sockaddr_in6 *,
    166 	struct sockaddr_in6 *, struct mbuf *, int);
    167 static int udp6_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    168 #endif
    169 #ifdef INET
    170 static	void udp_notify (struct inpcb *, int);
    171 #endif
    172 
    173 #ifndef UDBHASHSIZE
    174 #define	UDBHASHSIZE	128
    175 #endif
    176 int	udbhashsize = UDBHASHSIZE;
    177 
    178 /*
    179  * For send - really max datagram size; for receive - 40 1K datagrams.
    180  */
    181 static int	udp_sendspace = 9216;
    182 static int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
    183 
    184 #ifdef MBUFTRACE
    185 struct mowner udp_mowner = MOWNER_INIT("udp", "");
    186 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
    187 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
    188 #endif
    189 
    190 #ifdef UDP_CSUM_COUNTERS
    191 #include <sys/device.h>
    192 
    193 #if defined(INET)
    194 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    195     NULL, "udp", "hwcsum bad");
    196 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    197     NULL, "udp", "hwcsum ok");
    198 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    199     NULL, "udp", "hwcsum data");
    200 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    201     NULL, "udp", "swcsum");
    202 
    203 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
    204 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
    205 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
    206 EVCNT_ATTACH_STATIC(udp_swcsum);
    207 #endif /* defined(INET) */
    208 
    209 #if defined(INET6)
    210 struct evcnt udp6_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    211     NULL, "udp6", "hwcsum bad");
    212 struct evcnt udp6_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    213     NULL, "udp6", "hwcsum ok");
    214 struct evcnt udp6_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    215     NULL, "udp6", "hwcsum data");
    216 struct evcnt udp6_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    217     NULL, "udp6", "swcsum");
    218 
    219 EVCNT_ATTACH_STATIC(udp6_hwcsum_bad);
    220 EVCNT_ATTACH_STATIC(udp6_hwcsum_ok);
    221 EVCNT_ATTACH_STATIC(udp6_hwcsum_data);
    222 EVCNT_ATTACH_STATIC(udp6_swcsum);
    223 #endif /* defined(INET6) */
    224 
    225 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    226 
    227 #else
    228 
    229 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
    230 
    231 #endif /* UDP_CSUM_COUNTERS */
    232 
    233 static void sysctl_net_inet_udp_setup(struct sysctllog **);
    234 
    235 static int
    236 do_udpinit(void)
    237 {
    238 
    239 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
    240 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
    241 
    242 	MOWNER_ATTACH(&udp_tx_mowner);
    243 	MOWNER_ATTACH(&udp_rx_mowner);
    244 	MOWNER_ATTACH(&udp_mowner);
    245 
    246 	return 0;
    247 }
    248 
    249 void
    250 udp_init_common(void)
    251 {
    252 	static ONCE_DECL(doudpinit);
    253 
    254 	RUN_ONCE(&doudpinit, do_udpinit);
    255 }
    256 
    257 void
    258 udp_init(void)
    259 {
    260 
    261 	sysctl_net_inet_udp_setup(NULL);
    262 
    263 	udp_init_common();
    264 }
    265 
    266 /*
    267  * Checksum extended UDP header and data.
    268  */
    269 
    270 int
    271 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
    272     int iphlen, int len)
    273 {
    274 
    275 	switch (af) {
    276 #ifdef INET
    277 	case AF_INET:
    278 		return udp4_input_checksum(m, uh, iphlen, len);
    279 #endif
    280 #ifdef INET6
    281 	case AF_INET6:
    282 		return udp6_input_checksum(m, uh, iphlen, len);
    283 #endif
    284 	}
    285 #ifdef DIAGNOSTIC
    286 	panic("udp_input_checksum: unknown af %d", af);
    287 #endif
    288 	/* NOTREACHED */
    289 	return -1;
    290 }
    291 
    292 #ifdef INET
    293 
    294 /*
    295  * Checksum extended UDP header and data.
    296  */
    297 
    298 static int
    299 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
    300     int iphlen, int len)
    301 {
    302 
    303 	/*
    304 	 * XXX it's better to record and check if this mbuf is
    305 	 * already checked.
    306 	 */
    307 
    308 	if (uh->uh_sum == 0)
    309 		return 0;
    310 
    311 	switch (m->m_pkthdr.csum_flags &
    312 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv4) |
    313 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    314 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
    315 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
    316 		goto badcsum;
    317 
    318 	case M_CSUM_UDPv4|M_CSUM_DATA: {
    319 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
    320 
    321 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
    322 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
    323 			const struct ip *ip =
    324 			    mtod(m, const struct ip *);
    325 
    326 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
    327 			    ip->ip_dst.s_addr,
    328 			    htons(hw_csum + len + IPPROTO_UDP));
    329 		}
    330 		if ((hw_csum ^ 0xffff) != 0)
    331 			goto badcsum;
    332 		break;
    333 	}
    334 
    335 	case M_CSUM_UDPv4:
    336 		/* Checksum was okay. */
    337 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
    338 		break;
    339 
    340 	default:
    341 		/*
    342 		 * Need to compute it ourselves.  Maybe skip checksum
    343 		 * on loopback interfaces.
    344 		 */
    345 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
    346 				     IFF_LOOPBACK) ||
    347 				   udp_do_loopback_cksum)) {
    348 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
    349 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
    350 				goto badcsum;
    351 		}
    352 		break;
    353 	}
    354 
    355 	return 0;
    356 
    357 badcsum:
    358 	UDP_STATINC(UDP_STAT_BADSUM);
    359 	return -1;
    360 }
    361 
    362 void
    363 udp_input(struct mbuf *m, ...)
    364 {
    365 	va_list ap;
    366 	struct sockaddr_in src, dst;
    367 	struct ip *ip;
    368 	struct udphdr *uh;
    369 	int iphlen;
    370 	int len;
    371 	int n;
    372 	u_int16_t ip_len;
    373 
    374 	va_start(ap, m);
    375 	iphlen = va_arg(ap, int);
    376 	(void)va_arg(ap, int);		/* ignore value, advance ap */
    377 	va_end(ap);
    378 
    379 	MCLAIM(m, &udp_rx_mowner);
    380 	UDP_STATINC(UDP_STAT_IPACKETS);
    381 
    382 	/*
    383 	 * Get IP and UDP header together in first mbuf.
    384 	 */
    385 	ip = mtod(m, struct ip *);
    386 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
    387 	if (uh == NULL) {
    388 		UDP_STATINC(UDP_STAT_HDROPS);
    389 		return;
    390 	}
    391 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    392 
    393 	/* destination port of 0 is illegal, based on RFC768. */
    394 	if (uh->uh_dport == 0)
    395 		goto bad;
    396 
    397 	/*
    398 	 * Make mbuf data length reflect UDP length.
    399 	 * If not enough data to reflect UDP length, drop.
    400 	 */
    401 	ip_len = ntohs(ip->ip_len);
    402 	len = ntohs((u_int16_t)uh->uh_ulen);
    403 	if (ip_len != iphlen + len) {
    404 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
    405 			UDP_STATINC(UDP_STAT_BADLEN);
    406 			goto bad;
    407 		}
    408 		m_adj(m, iphlen + len - ip_len);
    409 	}
    410 
    411 	/*
    412 	 * Checksum extended UDP header and data.
    413 	 */
    414 	if (udp4_input_checksum(m, uh, iphlen, len))
    415 		goto badcsum;
    416 
    417 	/* construct source and dst sockaddrs. */
    418 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
    419 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
    420 
    421 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
    422 		UDP_STATINC(UDP_STAT_HDROPS);
    423 		return;
    424 	}
    425 	if (m == NULL) {
    426 		/*
    427 		 * packet has been processed by ESP stuff -
    428 		 * e.g. dropped NAT-T-keep-alive-packet ...
    429 		 */
    430 		return;
    431 	}
    432 	ip = mtod(m, struct ip *);
    433 #ifdef INET6
    434 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
    435 		struct sockaddr_in6 src6, dst6;
    436 
    437 		memset(&src6, 0, sizeof(src6));
    438 		src6.sin6_family = AF_INET6;
    439 		src6.sin6_len = sizeof(struct sockaddr_in6);
    440 		src6.sin6_addr.s6_addr[10] = src6.sin6_addr.s6_addr[11] = 0xff;
    441 		memcpy(&src6.sin6_addr.s6_addr[12], &ip->ip_src,
    442 			sizeof(ip->ip_src));
    443 		src6.sin6_port = uh->uh_sport;
    444 		memset(&dst6, 0, sizeof(dst6));
    445 		dst6.sin6_family = AF_INET6;
    446 		dst6.sin6_len = sizeof(struct sockaddr_in6);
    447 		dst6.sin6_addr.s6_addr[10] = dst6.sin6_addr.s6_addr[11] = 0xff;
    448 		memcpy(&dst6.sin6_addr.s6_addr[12], &ip->ip_dst,
    449 			sizeof(ip->ip_dst));
    450 		dst6.sin6_port = uh->uh_dport;
    451 
    452 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
    453 	}
    454 #endif
    455 
    456 	if (n == 0) {
    457 		if (m->m_flags & (M_BCAST | M_MCAST)) {
    458 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
    459 			goto bad;
    460 		}
    461 		UDP_STATINC(UDP_STAT_NOPORT);
    462 #ifdef IPKDB
    463 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
    464 				m, iphlen + sizeof(struct udphdr),
    465 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
    466 			/*
    467 			 * It was a debugger connect packet,
    468 			 * just drop it now
    469 			 */
    470 			goto bad;
    471 		}
    472 #endif
    473 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
    474 		m = NULL;
    475 	}
    476 
    477 bad:
    478 	if (m)
    479 		m_freem(m);
    480 	return;
    481 
    482 badcsum:
    483 	m_freem(m);
    484 }
    485 #endif
    486 
    487 #ifdef INET6
    488 static int
    489 udp6_input_checksum(struct mbuf *m, const struct udphdr *uh, int off, int len)
    490 {
    491 
    492 	/*
    493 	 * XXX it's better to record and check if this mbuf is
    494 	 * already checked.
    495 	 */
    496 
    497 	if (__predict_false((m->m_flags & M_LOOP) && !udp_do_loopback_cksum)) {
    498 		goto good;
    499 	}
    500 	if (uh->uh_sum == 0) {
    501 		UDP6_STATINC(UDP6_STAT_NOSUM);
    502 		goto bad;
    503 	}
    504 
    505 	switch (m->m_pkthdr.csum_flags &
    506 	    ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_UDPv6) |
    507 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    508 	case M_CSUM_UDPv6|M_CSUM_TCP_UDP_BAD:
    509 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_bad);
    510 		UDP6_STATINC(UDP6_STAT_BADSUM);
    511 		goto bad;
    512 
    513 #if 0 /* notyet */
    514 	case M_CSUM_UDPv6|M_CSUM_DATA:
    515 #endif
    516 
    517 	case M_CSUM_UDPv6:
    518 		/* Checksum was okay. */
    519 		UDP_CSUM_COUNTER_INCR(&udp6_hwcsum_ok);
    520 		break;
    521 
    522 	default:
    523 		/*
    524 		 * Need to compute it ourselves.  Maybe skip checksum
    525 		 * on loopback interfaces.
    526 		 */
    527 		UDP_CSUM_COUNTER_INCR(&udp6_swcsum);
    528 		if (in6_cksum(m, IPPROTO_UDP, off, len) != 0) {
    529 			UDP6_STATINC(UDP6_STAT_BADSUM);
    530 			goto bad;
    531 		}
    532 	}
    533 
    534 good:
    535 	return 0;
    536 bad:
    537 	return -1;
    538 }
    539 
    540 int
    541 udp6_input(struct mbuf **mp, int *offp, int proto)
    542 {
    543 	struct mbuf *m = *mp;
    544 	int off = *offp;
    545 	struct sockaddr_in6 src, dst;
    546 	struct ip6_hdr *ip6;
    547 	struct udphdr *uh;
    548 	u_int32_t plen, ulen;
    549 
    550 	ip6 = mtod(m, struct ip6_hdr *);
    551 
    552 #if defined(NFAITH) && 0 < NFAITH
    553 	if (faithprefix(&ip6->ip6_dst)) {
    554 		/* send icmp6 host unreach? */
    555 		m_freem(m);
    556 		return IPPROTO_DONE;
    557 	}
    558 #endif
    559 
    560 	UDP6_STATINC(UDP6_STAT_IPACKETS);
    561 
    562 	/* check for jumbogram is done in ip6_input.  we can trust pkthdr.len */
    563 	plen = m->m_pkthdr.len - off;
    564 	IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(struct udphdr));
    565 	if (uh == NULL) {
    566 		IP6_STATINC(IP6_STAT_TOOSHORT);
    567 		return IPPROTO_DONE;
    568 	}
    569 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    570 	ulen = ntohs((u_short)uh->uh_ulen);
    571 	/*
    572 	 * RFC2675 section 4: jumbograms will have 0 in the UDP header field,
    573 	 * iff payload length > 0xffff.
    574 	 */
    575 	if (ulen == 0 && plen > 0xffff)
    576 		ulen = plen;
    577 
    578 	if (plen != ulen) {
    579 		UDP6_STATINC(UDP6_STAT_BADLEN);
    580 		goto bad;
    581 	}
    582 
    583 	/* destination port of 0 is illegal, based on RFC768. */
    584 	if (uh->uh_dport == 0)
    585 		goto bad;
    586 
    587 	/* Be proactive about malicious use of IPv4 mapped address */
    588 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
    589 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
    590 		/* XXX stat */
    591 		goto bad;
    592 	}
    593 
    594 	/*
    595 	 * Checksum extended UDP header and data.  Maybe skip checksum
    596 	 * on loopback interfaces.
    597 	 */
    598 	if (udp6_input_checksum(m, uh, off, ulen))
    599 		goto bad;
    600 
    601 	/*
    602 	 * Construct source and dst sockaddrs.
    603 	 */
    604 	memset(&src, 0, sizeof(src));
    605 	src.sin6_family = AF_INET6;
    606 	src.sin6_len = sizeof(struct sockaddr_in6);
    607 	src.sin6_addr = ip6->ip6_src;
    608 	src.sin6_port = uh->uh_sport;
    609 	memset(&dst, 0, sizeof(dst));
    610 	dst.sin6_family = AF_INET6;
    611 	dst.sin6_len = sizeof(struct sockaddr_in6);
    612 	dst.sin6_addr = ip6->ip6_dst;
    613 	dst.sin6_port = uh->uh_dport;
    614 
    615 	if (udp6_realinput(AF_INET6, &src, &dst, m, off) == 0) {
    616 		if (m->m_flags & M_MCAST) {
    617 			UDP6_STATINC(UDP6_STAT_NOPORTMCAST);
    618 			goto bad;
    619 		}
    620 		UDP6_STATINC(UDP6_STAT_NOPORT);
    621 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
    622 		m = NULL;
    623 	}
    624 
    625 bad:
    626 	if (m)
    627 		m_freem(m);
    628 	return IPPROTO_DONE;
    629 }
    630 #endif
    631 
    632 #ifdef INET
    633 static void
    634 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
    635 	struct sockaddr *src, struct socket *so)
    636 {
    637 	struct mbuf *opts = NULL;
    638 	struct mbuf *n;
    639 	struct inpcb *inp = NULL;
    640 
    641 	if (!so)
    642 		return;
    643 	switch (so->so_proto->pr_domain->dom_family) {
    644 	case AF_INET:
    645 		inp = sotoinpcb(so);
    646 		break;
    647 #ifdef INET6
    648 	case AF_INET6:
    649 		break;
    650 #endif
    651 	default:
    652 		return;
    653 	}
    654 
    655 #if defined(IPSEC)
    656 	/* check AH/ESP integrity. */
    657 	if (so != NULL && ipsec4_in_reject_so(m, so)) {
    658 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
    659 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
    660 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
    661 			    0, 0);
    662 		return;
    663 	}
    664 #endif /*IPSEC*/
    665 
    666 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
    667 		if (inp && (inp->inp_flags & INP_CONTROLOPTS
    668 #ifdef SO_OTIMESTAMP
    669 			 || so->so_options & SO_OTIMESTAMP
    670 #endif
    671 			 || so->so_options & SO_TIMESTAMP)) {
    672 			struct ip *ip = mtod(n, struct ip *);
    673 			ip_savecontrol(inp, &opts, ip, n);
    674 		}
    675 
    676 		m_adj(n, off);
    677 		if (sbappendaddr(&so->so_rcv, src, n,
    678 				opts) == 0) {
    679 			m_freem(n);
    680 			if (opts)
    681 				m_freem(opts);
    682 			so->so_rcv.sb_overflowed++;
    683 			UDP_STATINC(UDP_STAT_FULLSOCK);
    684 		} else
    685 			sorwakeup(so);
    686 	}
    687 }
    688 #endif
    689 
    690 #ifdef INET6
    691 static void
    692 udp6_sendup(struct mbuf *m, int off /* offset of data portion */,
    693 	struct sockaddr *src, struct socket *so)
    694 {
    695 	struct mbuf *opts = NULL;
    696 	struct mbuf *n;
    697 	struct in6pcb *in6p = NULL;
    698 
    699 	if (!so)
    700 		return;
    701 	if (so->so_proto->pr_domain->dom_family != AF_INET6)
    702 		return;
    703 	in6p = sotoin6pcb(so);
    704 
    705 #if defined(IPSEC)
    706 	/* check AH/ESP integrity. */
    707 	if (so != NULL && ipsec6_in_reject_so(m, so)) {
    708 		IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
    709 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
    710 			icmp6_error(n, ICMP6_DST_UNREACH,
    711 			    ICMP6_DST_UNREACH_ADMIN, 0);
    712 		return;
    713 	}
    714 #endif /*IPSEC*/
    715 
    716 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
    717 		if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS
    718 #ifdef SO_OTIMESTAMP
    719 		    || in6p->in6p_socket->so_options & SO_OTIMESTAMP
    720 #endif
    721 		    || in6p->in6p_socket->so_options & SO_TIMESTAMP)) {
    722 			struct ip6_hdr *ip6 = mtod(n, struct ip6_hdr *);
    723 			ip6_savecontrol(in6p, &opts, ip6, n);
    724 		}
    725 
    726 		m_adj(n, off);
    727 		if (sbappendaddr(&so->so_rcv, src, n, opts) == 0) {
    728 			m_freem(n);
    729 			if (opts)
    730 				m_freem(opts);
    731 			so->so_rcv.sb_overflowed++;
    732 			UDP6_STATINC(UDP6_STAT_FULLSOCK);
    733 		} else
    734 			sorwakeup(so);
    735 	}
    736 }
    737 #endif
    738 
    739 #ifdef INET
    740 static int
    741 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
    742 	struct mbuf **mp, int off /* offset of udphdr */)
    743 {
    744 	u_int16_t *sport, *dport;
    745 	int rcvcnt;
    746 	struct in_addr *src4, *dst4;
    747 	struct inpcb_hdr *inph;
    748 	struct inpcb *inp;
    749 	struct mbuf *m = *mp;
    750 
    751 	rcvcnt = 0;
    752 	off += sizeof(struct udphdr);	/* now, offset of payload */
    753 
    754 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
    755 		goto bad;
    756 
    757 	src4 = &src->sin_addr;
    758 	sport = &src->sin_port;
    759 	dst4 = &dst->sin_addr;
    760 	dport = &dst->sin_port;
    761 
    762 	if (IN_MULTICAST(dst4->s_addr) ||
    763 	    in_broadcast(*dst4, m->m_pkthdr.rcvif)) {
    764 		/*
    765 		 * Deliver a multicast or broadcast datagram to *all* sockets
    766 		 * for which the local and remote addresses and ports match
    767 		 * those of the incoming datagram.  This allows more than
    768 		 * one process to receive multi/broadcasts on the same port.
    769 		 * (This really ought to be done for unicast datagrams as
    770 		 * well, but that would cause problems with existing
    771 		 * applications that open both address-specific sockets and
    772 		 * a wildcard socket listening to the same port -- they would
    773 		 * end up receiving duplicates of every unicast datagram.
    774 		 * Those applications open the multiple sockets to overcome an
    775 		 * inadequacy of the UDP socket interface, but for backwards
    776 		 * compatibility we avoid the problem here rather than
    777 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    778 		 */
    779 
    780 		/*
    781 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    782 		 * we need udpiphdr for IPsec processing so we do that later.
    783 		 */
    784 		/*
    785 		 * Locate pcb(s) for datagram.
    786 		 */
    787 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
    788 			inp = (struct inpcb *)inph;
    789 			if (inp->inp_af != AF_INET)
    790 				continue;
    791 
    792 			if (inp->inp_lport != *dport)
    793 				continue;
    794 			if (!in_nullhost(inp->inp_laddr)) {
    795 				if (!in_hosteq(inp->inp_laddr, *dst4))
    796 					continue;
    797 			}
    798 			if (!in_nullhost(inp->inp_faddr)) {
    799 				if (!in_hosteq(inp->inp_faddr, *src4) ||
    800 				    inp->inp_fport != *sport)
    801 					continue;
    802 			}
    803 
    804 			udp4_sendup(m, off, (struct sockaddr *)src,
    805 				inp->inp_socket);
    806 			rcvcnt++;
    807 
    808 			/*
    809 			 * Don't look for additional matches if this one does
    810 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    811 			 * socket options set.  This heuristic avoids searching
    812 			 * through all pcbs in the common case of a non-shared
    813 			 * port.  It assumes that an application will never
    814 			 * clear these options after setting them.
    815 			 */
    816 			if ((inp->inp_socket->so_options &
    817 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    818 				break;
    819 		}
    820 	} else {
    821 		/*
    822 		 * Locate pcb for datagram.
    823 		 */
    824 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
    825 		    *dport, 0);
    826 		if (inp == 0) {
    827 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
    828 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
    829 			if (inp == 0)
    830 				return rcvcnt;
    831 		}
    832 
    833 #ifdef IPSEC
    834 		/* Handle ESP over UDP */
    835 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
    836 			struct sockaddr *sa = (struct sockaddr *)src;
    837 
    838 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
    839 			case -1: 	/* Error, m was freeed */
    840 				rcvcnt = -1;
    841 				goto bad;
    842 				break;
    843 
    844 			case 1:		/* ESP over UDP */
    845 				rcvcnt++;
    846 				goto bad;
    847 				break;
    848 
    849 			case 0: 	/* plain UDP */
    850 			default: 	/* Unexpected */
    851 				/*
    852 				 * Normal UDP processing will take place
    853 				 * m may have changed.
    854 				 */
    855 				m = *mp;
    856 				break;
    857 			}
    858 		}
    859 #endif
    860 
    861 		/*
    862 		 * Check the minimum TTL for socket.
    863 		 */
    864 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
    865 			goto bad;
    866 
    867 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
    868 		rcvcnt++;
    869 	}
    870 
    871 bad:
    872 	return rcvcnt;
    873 }
    874 #endif
    875 
    876 #ifdef INET6
    877 static int
    878 udp6_realinput(int af, struct sockaddr_in6 *src, struct sockaddr_in6 *dst,
    879 	struct mbuf *m, int off)
    880 {
    881 	u_int16_t sport, dport;
    882 	int rcvcnt;
    883 	struct in6_addr src6, *dst6;
    884 	const struct in_addr *dst4;
    885 	struct inpcb_hdr *inph;
    886 	struct in6pcb *in6p;
    887 
    888 	rcvcnt = 0;
    889 	off += sizeof(struct udphdr);	/* now, offset of payload */
    890 
    891 	if (af != AF_INET && af != AF_INET6)
    892 		goto bad;
    893 	if (src->sin6_family != AF_INET6 || dst->sin6_family != AF_INET6)
    894 		goto bad;
    895 
    896 	src6 = src->sin6_addr;
    897 	if (sa6_recoverscope(src) != 0) {
    898 		/* XXX: should be impossible. */
    899 		goto bad;
    900 	}
    901 	sport = src->sin6_port;
    902 
    903 	dport = dst->sin6_port;
    904 	dst4 = (struct in_addr *)&dst->sin6_addr.s6_addr[12];
    905 	dst6 = &dst->sin6_addr;
    906 
    907 	if (IN6_IS_ADDR_MULTICAST(dst6) ||
    908 	    (af == AF_INET && IN_MULTICAST(dst4->s_addr))) {
    909 		/*
    910 		 * Deliver a multicast or broadcast datagram to *all* sockets
    911 		 * for which the local and remote addresses and ports match
    912 		 * those of the incoming datagram.  This allows more than
    913 		 * one process to receive multi/broadcasts on the same port.
    914 		 * (This really ought to be done for unicast datagrams as
    915 		 * well, but that would cause problems with existing
    916 		 * applications that open both address-specific sockets and
    917 		 * a wildcard socket listening to the same port -- they would
    918 		 * end up receiving duplicates of every unicast datagram.
    919 		 * Those applications open the multiple sockets to overcome an
    920 		 * inadequacy of the UDP socket interface, but for backwards
    921 		 * compatibility we avoid the problem here rather than
    922 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    923 		 */
    924 
    925 		/*
    926 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    927 		 * we need udpiphdr for IPsec processing so we do that later.
    928 		 */
    929 		/*
    930 		 * Locate pcb(s) for datagram.
    931 		 */
    932 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
    933 			in6p = (struct in6pcb *)inph;
    934 			if (in6p->in6p_af != AF_INET6)
    935 				continue;
    936 
    937 			if (in6p->in6p_lport != dport)
    938 				continue;
    939 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
    940 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
    941 				    dst6))
    942 					continue;
    943 			} else {
    944 				if (IN6_IS_ADDR_V4MAPPED(dst6) &&
    945 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
    946 					continue;
    947 			}
    948 			if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
    949 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
    950 				    &src6) || in6p->in6p_fport != sport)
    951 					continue;
    952 			} else {
    953 				if (IN6_IS_ADDR_V4MAPPED(&src6) &&
    954 				    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
    955 					continue;
    956 			}
    957 
    958 			udp6_sendup(m, off, (struct sockaddr *)src,
    959 				in6p->in6p_socket);
    960 			rcvcnt++;
    961 
    962 			/*
    963 			 * Don't look for additional matches if this one does
    964 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    965 			 * socket options set.  This heuristic avoids searching
    966 			 * through all pcbs in the common case of a non-shared
    967 			 * port.  It assumes that an application will never
    968 			 * clear these options after setting them.
    969 			 */
    970 			if ((in6p->in6p_socket->so_options &
    971 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    972 				break;
    973 		}
    974 	} else {
    975 		/*
    976 		 * Locate pcb for datagram.
    977 		 */
    978 		in6p = in6_pcblookup_connect(&udbtable, &src6, sport, dst6,
    979 					     dport, 0, 0);
    980 		if (in6p == 0) {
    981 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
    982 			in6p = in6_pcblookup_bind(&udbtable, dst6, dport, 0);
    983 			if (in6p == 0)
    984 				return rcvcnt;
    985 		}
    986 
    987 		udp6_sendup(m, off, (struct sockaddr *)src, in6p->in6p_socket);
    988 		rcvcnt++;
    989 	}
    990 
    991 bad:
    992 	return rcvcnt;
    993 }
    994 #endif
    995 
    996 #ifdef INET
    997 /*
    998  * Notify a udp user of an asynchronous error;
    999  * just wake up so that he can collect error status.
   1000  */
   1001 static void
   1002 udp_notify(struct inpcb *inp, int errno)
   1003 {
   1004 	inp->inp_socket->so_error = errno;
   1005 	sorwakeup(inp->inp_socket);
   1006 	sowwakeup(inp->inp_socket);
   1007 }
   1008 
   1009 void *
   1010 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
   1011 {
   1012 	struct ip *ip = v;
   1013 	struct udphdr *uh;
   1014 	void (*notify)(struct inpcb *, int) = udp_notify;
   1015 	int errno;
   1016 
   1017 	if (sa->sa_family != AF_INET
   1018 	 || sa->sa_len != sizeof(struct sockaddr_in))
   1019 		return NULL;
   1020 	if ((unsigned)cmd >= PRC_NCMDS)
   1021 		return NULL;
   1022 	errno = inetctlerrmap[cmd];
   1023 	if (PRC_IS_REDIRECT(cmd))
   1024 		notify = in_rtchange, ip = 0;
   1025 	else if (cmd == PRC_HOSTDEAD)
   1026 		ip = 0;
   1027 	else if (errno == 0)
   1028 		return NULL;
   1029 	if (ip) {
   1030 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
   1031 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
   1032 		    ip->ip_src, uh->uh_sport, errno, notify);
   1033 
   1034 		/* XXX mapped address case */
   1035 	} else
   1036 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
   1037 		    notify);
   1038 	return NULL;
   1039 }
   1040 
   1041 int
   1042 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
   1043 {
   1044 	int s;
   1045 	int error = 0;
   1046 	struct inpcb *inp;
   1047 	int family;
   1048 	int optval;
   1049 
   1050 	family = so->so_proto->pr_domain->dom_family;
   1051 
   1052 	s = splsoftnet();
   1053 	switch (family) {
   1054 #ifdef INET
   1055 	case PF_INET:
   1056 		if (sopt->sopt_level != IPPROTO_UDP) {
   1057 			error = ip_ctloutput(op, so, sopt);
   1058 			goto end;
   1059 		}
   1060 		break;
   1061 #endif
   1062 #ifdef INET6
   1063 	case PF_INET6:
   1064 		if (sopt->sopt_level != IPPROTO_UDP) {
   1065 			error = ip6_ctloutput(op, so, sopt);
   1066 			goto end;
   1067 		}
   1068 		break;
   1069 #endif
   1070 	default:
   1071 		error = EAFNOSUPPORT;
   1072 		goto end;
   1073 	}
   1074 
   1075 
   1076 	switch (op) {
   1077 	case PRCO_SETOPT:
   1078 		inp = sotoinpcb(so);
   1079 
   1080 		switch (sopt->sopt_name) {
   1081 		case UDP_ENCAP:
   1082 			error = sockopt_getint(sopt, &optval);
   1083 			if (error)
   1084 				break;
   1085 
   1086 			switch(optval) {
   1087 			case 0:
   1088 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1089 				break;
   1090 
   1091 			case UDP_ENCAP_ESPINUDP:
   1092 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1093 				inp->inp_flags |= INP_ESPINUDP;
   1094 				break;
   1095 
   1096 			case UDP_ENCAP_ESPINUDP_NON_IKE:
   1097 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
   1098 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
   1099 				break;
   1100 			default:
   1101 				error = EINVAL;
   1102 				break;
   1103 			}
   1104 			break;
   1105 
   1106 		default:
   1107 			error = ENOPROTOOPT;
   1108 			break;
   1109 		}
   1110 		break;
   1111 
   1112 	default:
   1113 		error = EINVAL;
   1114 		break;
   1115 	}
   1116 
   1117 end:
   1118 	splx(s);
   1119 	return error;
   1120 }
   1121 
   1122 
   1123 int
   1124 udp_output(struct mbuf *m, ...)
   1125 {
   1126 	struct inpcb *inp;
   1127 	struct udpiphdr *ui;
   1128 	struct route *ro;
   1129 	int len = m->m_pkthdr.len;
   1130 	int error = 0;
   1131 	va_list ap;
   1132 
   1133 	MCLAIM(m, &udp_tx_mowner);
   1134 	va_start(ap, m);
   1135 	inp = va_arg(ap, struct inpcb *);
   1136 	va_end(ap);
   1137 
   1138 	/*
   1139 	 * Calculate data length and get a mbuf
   1140 	 * for UDP and IP headers.
   1141 	 */
   1142 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
   1143 	if (m == 0) {
   1144 		error = ENOBUFS;
   1145 		goto release;
   1146 	}
   1147 
   1148 	/*
   1149 	 * Compute the packet length of the IP header, and
   1150 	 * punt if the length looks bogus.
   1151 	 */
   1152 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
   1153 		error = EMSGSIZE;
   1154 		goto release;
   1155 	}
   1156 
   1157 	/*
   1158 	 * Fill in mbuf with extended UDP header
   1159 	 * and addresses and length put into network format.
   1160 	 */
   1161 	ui = mtod(m, struct udpiphdr *);
   1162 	ui->ui_pr = IPPROTO_UDP;
   1163 	ui->ui_src = inp->inp_laddr;
   1164 	ui->ui_dst = inp->inp_faddr;
   1165 	ui->ui_sport = inp->inp_lport;
   1166 	ui->ui_dport = inp->inp_fport;
   1167 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
   1168 
   1169 	ro = &inp->inp_route;
   1170 
   1171 	/*
   1172 	 * Set up checksum and output datagram.
   1173 	 */
   1174 	if (udpcksum) {
   1175 		/*
   1176 		 * XXX Cache pseudo-header checksum part for
   1177 		 * XXX "connected" UDP sockets.
   1178 		 */
   1179 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
   1180 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
   1181 		    sizeof(struct udphdr) + IPPROTO_UDP));
   1182 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
   1183 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
   1184 	} else
   1185 		ui->ui_sum = 0;
   1186 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
   1187 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
   1188 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
   1189 	UDP_STATINC(UDP_STAT_OPACKETS);
   1190 
   1191 	return (ip_output(m, inp->inp_options, ro,
   1192 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
   1193 	    inp->inp_moptions, inp->inp_socket));
   1194 
   1195 release:
   1196 	m_freem(m);
   1197 	return (error);
   1198 }
   1199 
   1200 static int
   1201 udp_attach(struct socket *so, int proto)
   1202 {
   1203 	struct inpcb *inp;
   1204 	int error;
   1205 
   1206 	KASSERT(sotoinpcb(so) == NULL);
   1207 
   1208 	/* Assign the lock (must happen even if we will error out). */
   1209 	sosetlock(so);
   1210 
   1211 #ifdef MBUFTRACE
   1212 	so->so_mowner = &udp_mowner;
   1213 	so->so_rcv.sb_mowner = &udp_rx_mowner;
   1214 	so->so_snd.sb_mowner = &udp_tx_mowner;
   1215 #endif
   1216 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
   1217 		error = soreserve(so, udp_sendspace, udp_recvspace);
   1218 		if (error) {
   1219 			return error;
   1220 		}
   1221 	}
   1222 
   1223 	error = in_pcballoc(so, &udbtable);
   1224 	if (error) {
   1225 		return error;
   1226 	}
   1227 	inp = sotoinpcb(so);
   1228 	inp->inp_ip.ip_ttl = ip_defttl;
   1229 	KASSERT(solocked(so));
   1230 
   1231 	return error;
   1232 }
   1233 
   1234 static void
   1235 udp_detach(struct socket *so)
   1236 {
   1237 	struct inpcb *inp;
   1238 
   1239 	KASSERT(solocked(so));
   1240 	inp = sotoinpcb(so);
   1241 	KASSERT(inp != NULL);
   1242 	in_pcbdetach(inp);
   1243 }
   1244 
   1245 static int
   1246 udp_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
   1247     struct mbuf *control, struct lwp *l)
   1248 {
   1249 	struct inpcb *inp;
   1250 	int s, error = 0;
   1251 
   1252 	KASSERT(req != PRU_ATTACH);
   1253 	KASSERT(req != PRU_DETACH);
   1254 
   1255 	if (req == PRU_CONTROL) {
   1256 		return in_control(so, (long)m, (void *)nam,
   1257 		    (struct ifnet *)control, l);
   1258 	}
   1259 	s = splsoftnet();
   1260 	if (req == PRU_PURGEIF) {
   1261 		mutex_enter(softnet_lock);
   1262 		in_pcbpurgeif0(&udbtable, (struct ifnet *)control);
   1263 		in_purgeif((struct ifnet *)control);
   1264 		in_pcbpurgeif(&udbtable, (struct ifnet *)control);
   1265 		mutex_exit(softnet_lock);
   1266 		splx(s);
   1267 		return 0;
   1268 	}
   1269 
   1270 	KASSERT(solocked(so));
   1271 	inp = sotoinpcb(so);
   1272 
   1273 	KASSERT(!control || (req == PRU_SEND || req == PRU_SENDOOB));
   1274 	if (inp == NULL) {
   1275 		splx(s);
   1276 		return EINVAL;
   1277 	}
   1278 
   1279 	/*
   1280 	 * Note: need to block udp_input while changing
   1281 	 * the udp pcb queue and/or pcb addresses.
   1282 	 */
   1283 	switch (req) {
   1284 	case PRU_BIND:
   1285 		error = in_pcbbind(inp, nam, l);
   1286 		break;
   1287 
   1288 	case PRU_LISTEN:
   1289 		error = EOPNOTSUPP;
   1290 		break;
   1291 
   1292 	case PRU_CONNECT:
   1293 		error = in_pcbconnect(inp, nam, l);
   1294 		if (error)
   1295 			break;
   1296 		soisconnected(so);
   1297 		break;
   1298 
   1299 	case PRU_CONNECT2:
   1300 		error = EOPNOTSUPP;
   1301 		break;
   1302 
   1303 	case PRU_DISCONNECT:
   1304 		/*soisdisconnected(so);*/
   1305 		so->so_state &= ~SS_ISCONNECTED;	/* XXX */
   1306 		in_pcbdisconnect(inp);
   1307 		inp->inp_laddr = zeroin_addr;		/* XXX */
   1308 		in_pcbstate(inp, INP_BOUND);		/* XXX */
   1309 		break;
   1310 
   1311 	case PRU_SHUTDOWN:
   1312 		socantsendmore(so);
   1313 		break;
   1314 
   1315 	case PRU_RCVD:
   1316 		error = EOPNOTSUPP;
   1317 		break;
   1318 
   1319 	case PRU_SEND:
   1320 		if (control && control->m_len) {
   1321 			m_freem(control);
   1322 			m_freem(m);
   1323 			error = EINVAL;
   1324 			break;
   1325 		}
   1326 	{
   1327 		struct in_addr laddr;			/* XXX */
   1328 
   1329 		memset(&laddr, 0, sizeof laddr);
   1330 		if (nam) {
   1331 			laddr = inp->inp_laddr;		/* XXX */
   1332 			if ((so->so_state & SS_ISCONNECTED) != 0) {
   1333 				error = EISCONN;
   1334 				goto die;
   1335 			}
   1336 			error = in_pcbconnect(inp, nam, l);
   1337 			if (error)
   1338 				goto die;
   1339 		} else {
   1340 			if ((so->so_state & SS_ISCONNECTED) == 0) {
   1341 				error = ENOTCONN;
   1342 				goto die;
   1343 			}
   1344 		}
   1345 		error = udp_output(m, inp);
   1346 		m = NULL;
   1347 		if (nam) {
   1348 			in_pcbdisconnect(inp);
   1349 			inp->inp_laddr = laddr;		/* XXX */
   1350 			in_pcbstate(inp, INP_BOUND);	/* XXX */
   1351 		}
   1352 	  die:
   1353 		if (m)
   1354 			m_freem(m);
   1355 	}
   1356 		break;
   1357 
   1358 	case PRU_SENSE:
   1359 		/*
   1360 		 * stat: don't bother with a blocksize.
   1361 		 */
   1362 		splx(s);
   1363 		return (0);
   1364 
   1365 	case PRU_RCVOOB:
   1366 		error =  EOPNOTSUPP;
   1367 		break;
   1368 
   1369 	case PRU_SENDOOB:
   1370 		m_freem(control);
   1371 		m_freem(m);
   1372 		error =  EOPNOTSUPP;
   1373 		break;
   1374 
   1375 	case PRU_SOCKADDR:
   1376 		in_setsockaddr(inp, nam);
   1377 		break;
   1378 
   1379 	case PRU_PEERADDR:
   1380 		in_setpeeraddr(inp, nam);
   1381 		break;
   1382 
   1383 	default:
   1384 		panic("udp_usrreq");
   1385 	}
   1386 	splx(s);
   1387 
   1388 	return error;
   1389 }
   1390 
   1391 static int
   1392 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
   1393 {
   1394 
   1395 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
   1396 }
   1397 
   1398 /*
   1399  * Sysctl for udp variables.
   1400  */
   1401 static void
   1402 sysctl_net_inet_udp_setup(struct sysctllog **clog)
   1403 {
   1404 
   1405 	sysctl_createv(clog, 0, NULL, NULL,
   1406 		       CTLFLAG_PERMANENT,
   1407 		       CTLTYPE_NODE, "inet", NULL,
   1408 		       NULL, 0, NULL, 0,
   1409 		       CTL_NET, PF_INET, CTL_EOL);
   1410 	sysctl_createv(clog, 0, NULL, NULL,
   1411 		       CTLFLAG_PERMANENT,
   1412 		       CTLTYPE_NODE, "udp",
   1413 		       SYSCTL_DESCR("UDPv4 related settings"),
   1414 		       NULL, 0, NULL, 0,
   1415 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
   1416 
   1417 	sysctl_createv(clog, 0, NULL, NULL,
   1418 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1419 		       CTLTYPE_INT, "checksum",
   1420 		       SYSCTL_DESCR("Compute UDP checksums"),
   1421 		       NULL, 0, &udpcksum, 0,
   1422 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
   1423 		       CTL_EOL);
   1424 	sysctl_createv(clog, 0, NULL, NULL,
   1425 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1426 		       CTLTYPE_INT, "sendspace",
   1427 		       SYSCTL_DESCR("Default UDP send buffer size"),
   1428 		       NULL, 0, &udp_sendspace, 0,
   1429 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
   1430 		       CTL_EOL);
   1431 	sysctl_createv(clog, 0, NULL, NULL,
   1432 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1433 		       CTLTYPE_INT, "recvspace",
   1434 		       SYSCTL_DESCR("Default UDP receive buffer size"),
   1435 		       NULL, 0, &udp_recvspace, 0,
   1436 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
   1437 		       CTL_EOL);
   1438 	sysctl_createv(clog, 0, NULL, NULL,
   1439 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1440 		       CTLTYPE_INT, "do_loopback_cksum",
   1441 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
   1442 		       NULL, 0, &udp_do_loopback_cksum, 0,
   1443 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
   1444 		       CTL_EOL);
   1445 	sysctl_createv(clog, 0, NULL, NULL,
   1446 		       CTLFLAG_PERMANENT,
   1447 		       CTLTYPE_STRUCT, "pcblist",
   1448 		       SYSCTL_DESCR("UDP protocol control block list"),
   1449 		       sysctl_inpcblist, 0, &udbtable, 0,
   1450 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
   1451 		       CTL_EOL);
   1452 	sysctl_createv(clog, 0, NULL, NULL,
   1453 		       CTLFLAG_PERMANENT,
   1454 		       CTLTYPE_STRUCT, "stats",
   1455 		       SYSCTL_DESCR("UDP statistics"),
   1456 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
   1457 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
   1458 		       CTL_EOL);
   1459 }
   1460 #endif
   1461 
   1462 void
   1463 udp_statinc(u_int stat)
   1464 {
   1465 
   1466 	KASSERT(stat < UDP_NSTATS);
   1467 	UDP_STATINC(stat);
   1468 }
   1469 
   1470 #if defined(INET) && defined(IPSEC)
   1471 /*
   1472  * Returns:
   1473  * 1 if the packet was processed
   1474  * 0 if normal UDP processing should take place
   1475  * -1 if an error occurent and m was freed
   1476  */
   1477 static int
   1478 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
   1479     struct socket *so)
   1480 {
   1481 	size_t len;
   1482 	void *data;
   1483 	struct inpcb *inp;
   1484 	size_t skip = 0;
   1485 	size_t minlen;
   1486 	size_t iphdrlen;
   1487 	struct ip *ip;
   1488 	struct m_tag *tag;
   1489 	struct udphdr *udphdr;
   1490 	u_int16_t sport, dport;
   1491 	struct mbuf *m = *mp;
   1492 
   1493 	/*
   1494 	 * Collapse the mbuf chain if the first mbuf is too short
   1495 	 * The longest case is: UDP + non ESP marker + ESP
   1496 	 */
   1497 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
   1498 	if (minlen > m->m_pkthdr.len)
   1499 		minlen = m->m_pkthdr.len;
   1500 
   1501 	if (m->m_len < minlen) {
   1502 		if ((*mp = m_pullup(m, minlen)) == NULL) {
   1503 			printf("udp4_espinudp: m_pullup failed\n");
   1504 			return -1;
   1505 		}
   1506 		m = *mp;
   1507 	}
   1508 
   1509 	len = m->m_len - off;
   1510 	data = mtod(m, char *) + off;
   1511 	inp = sotoinpcb(so);
   1512 
   1513 	/* Ignore keepalive packets */
   1514 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
   1515 		m_free(m);
   1516 		*mp = NULL; /* avoid any further processiong by caller ... */
   1517 		return 1;
   1518 	}
   1519 
   1520 	/*
   1521 	 * Check that the payload is long enough to hold
   1522 	 * an ESP header and compute the length of encapsulation
   1523 	 * header to remove
   1524 	 */
   1525 	if (inp->inp_flags & INP_ESPINUDP) {
   1526 		u_int32_t *st = (u_int32_t *)data;
   1527 
   1528 		if ((len <= sizeof(struct esp)) || (*st == 0))
   1529 			return 0; /* Normal UDP processing */
   1530 
   1531 		skip = sizeof(struct udphdr);
   1532 	}
   1533 
   1534 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
   1535 		u_int32_t *st = (u_int32_t *)data;
   1536 
   1537 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
   1538 		    || ((st[0] | st[1]) != 0))
   1539 			return 0; /* Normal UDP processing */
   1540 
   1541 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
   1542 	}
   1543 
   1544 	/*
   1545 	 * Get the UDP ports. They are handled in network
   1546 	 * order everywhere in IPSEC_NAT_T code.
   1547 	 */
   1548 	udphdr = (struct udphdr *)((char *)data - skip);
   1549 	sport = udphdr->uh_sport;
   1550 	dport = udphdr->uh_dport;
   1551 
   1552 	/*
   1553 	 * Remove the UDP header (and possibly the non ESP marker)
   1554 	 * IP header lendth is iphdrlen
   1555 	 * Before:
   1556 	 *   <--- off --->
   1557 	 *   +----+------+-----+
   1558 	 *   | IP |  UDP | ESP |
   1559 	 *   +----+------+-----+
   1560 	 *        <-skip->
   1561 	 * After:
   1562 	 *          +----+-----+
   1563 	 *          | IP | ESP |
   1564 	 *          +----+-----+
   1565 	 *   <-skip->
   1566 	 */
   1567 	iphdrlen = off - sizeof(struct udphdr);
   1568 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
   1569 	m_adj(m, skip);
   1570 
   1571 	ip = mtod(m, struct ip *);
   1572 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
   1573 	ip->ip_p = IPPROTO_ESP;
   1574 
   1575 	/*
   1576 	 * We have modified the packet - it is now ESP, so we should not
   1577 	 * return to UDP processing ...
   1578 	 *
   1579 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
   1580 	 * the source UDP port. This is required if we want
   1581 	 * to select the right SPD for multiple hosts behind
   1582 	 * same NAT
   1583 	 */
   1584 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
   1585 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
   1586 		printf("udp4_espinudp: m_tag_get failed\n");
   1587 		m_freem(m);
   1588 		return -1;
   1589 	}
   1590 	((u_int16_t *)(tag + 1))[0] = sport;
   1591 	((u_int16_t *)(tag + 1))[1] = dport;
   1592 	m_tag_prepend(m, tag);
   1593 
   1594 #ifdef IPSEC
   1595 	ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
   1596 #else
   1597 	esp4_input(m, iphdrlen);
   1598 #endif
   1599 
   1600 	/* We handled it, it shouldn't be handled by UDP */
   1601 	*mp = NULL; /* avoid free by caller ... */
   1602 	return 1;
   1603 }
   1604 #endif
   1605 
   1606 PR_WRAP_USRREQ(udp_usrreq)
   1607 
   1608 #define	udp_usrreq	udp_usrreq_wrapper
   1609 
   1610 const struct pr_usrreqs udp_usrreqs = {
   1611 	.pr_attach	= udp_attach,
   1612 	.pr_detach	= udp_detach,
   1613 	.pr_generic	= udp_usrreq,
   1614 };
   1615