Home | History | Annotate | Line # | Download | only in netinet
udp_usrreq.c revision 1.187.2.3
      1 /*	$NetBSD: udp_usrreq.c,v 1.187.2.3 2017/12/03 11:39:04 jdolecek 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 /*
     64  * UDP protocol implementation.
     65  * Per RFC 768, August, 1980.
     66  */
     67 
     68 #include <sys/cdefs.h>
     69 __KERNEL_RCSID(0, "$NetBSD: udp_usrreq.c,v 1.187.2.3 2017/12/03 11:39:04 jdolecek Exp $");
     70 
     71 #ifdef _KERNEL_OPT
     72 #include "opt_inet.h"
     73 #include "opt_ipsec.h"
     74 #include "opt_inet_csum.h"
     75 #include "opt_ipkdb.h"
     76 #include "opt_mbuftrace.h"
     77 #include "opt_net_mpsafe.h"
     78 #endif
     79 
     80 #include <sys/param.h>
     81 #include <sys/mbuf.h>
     82 #include <sys/once.h>
     83 #include <sys/protosw.h>
     84 #include <sys/socket.h>
     85 #include <sys/socketvar.h>
     86 #include <sys/systm.h>
     87 #include <sys/proc.h>
     88 #include <sys/domain.h>
     89 #include <sys/sysctl.h>
     90 
     91 #include <net/if.h>
     92 
     93 #include <netinet/in.h>
     94 #include <netinet/in_systm.h>
     95 #include <netinet/in_var.h>
     96 #include <netinet/ip.h>
     97 #include <netinet/in_pcb.h>
     98 #include <netinet/ip_var.h>
     99 #include <netinet/ip_icmp.h>
    100 #include <netinet/udp.h>
    101 #include <netinet/udp_var.h>
    102 #include <netinet/udp_private.h>
    103 
    104 #ifdef INET6
    105 #include <netinet/ip6.h>
    106 #include <netinet/icmp6.h>
    107 #include <netinet6/ip6_var.h>
    108 #include <netinet6/ip6_private.h>
    109 #include <netinet6/in6_pcb.h>
    110 #include <netinet6/udp6_var.h>
    111 #include <netinet6/udp6_private.h>
    112 #endif
    113 
    114 #ifndef INET6
    115 /* always need ip6.h for IP6_EXTHDR_GET */
    116 #include <netinet/ip6.h>
    117 #endif
    118 
    119 #ifdef IPSEC
    120 #include <netipsec/ipsec.h>
    121 #include <netipsec/ipsec_var.h>
    122 #include <netipsec/ipsec_private.h>
    123 #include <netipsec/esp.h>
    124 #ifdef INET6
    125 #include <netipsec/ipsec6.h>
    126 #endif
    127 #endif	/* IPSEC */
    128 
    129 #ifdef IPKDB
    130 #include <ipkdb/ipkdb.h>
    131 #endif
    132 
    133 int	udpcksum = 1;
    134 int	udp_do_loopback_cksum = 0;
    135 
    136 struct	inpcbtable udbtable;
    137 
    138 percpu_t *udpstat_percpu;
    139 
    140 #ifdef INET
    141 #ifdef IPSEC
    142 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
    143 	struct socket *);
    144 #endif
    145 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
    146 	struct socket *);
    147 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
    148 	struct mbuf **, int);
    149 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    150 #endif
    151 #ifdef INET
    152 static	void udp_notify (struct inpcb *, int);
    153 #endif
    154 
    155 #ifndef UDBHASHSIZE
    156 #define	UDBHASHSIZE	128
    157 #endif
    158 int	udbhashsize = UDBHASHSIZE;
    159 
    160 /*
    161  * For send - really max datagram size; for receive - 40 1K datagrams.
    162  */
    163 static int	udp_sendspace = 9216;
    164 static int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
    165 
    166 #ifdef MBUFTRACE
    167 struct mowner udp_mowner = MOWNER_INIT("udp", "");
    168 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
    169 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
    170 #endif
    171 
    172 #ifdef UDP_CSUM_COUNTERS
    173 #include <sys/device.h>
    174 
    175 #if defined(INET)
    176 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    177     NULL, "udp", "hwcsum bad");
    178 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    179     NULL, "udp", "hwcsum ok");
    180 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    181     NULL, "udp", "hwcsum data");
    182 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    183     NULL, "udp", "swcsum");
    184 
    185 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
    186 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
    187 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
    188 EVCNT_ATTACH_STATIC(udp_swcsum);
    189 #endif /* defined(INET) */
    190 
    191 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    192 #else
    193 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
    194 #endif /* UDP_CSUM_COUNTERS */
    195 
    196 static void sysctl_net_inet_udp_setup(struct sysctllog **);
    197 
    198 static int
    199 do_udpinit(void)
    200 {
    201 
    202 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
    203 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
    204 
    205 	MOWNER_ATTACH(&udp_tx_mowner);
    206 	MOWNER_ATTACH(&udp_rx_mowner);
    207 	MOWNER_ATTACH(&udp_mowner);
    208 
    209 	return 0;
    210 }
    211 
    212 void
    213 udp_init_common(void)
    214 {
    215 	static ONCE_DECL(doudpinit);
    216 
    217 	RUN_ONCE(&doudpinit, do_udpinit);
    218 }
    219 
    220 void
    221 udp_init(void)
    222 {
    223 
    224 	sysctl_net_inet_udp_setup(NULL);
    225 
    226 	udp_init_common();
    227 }
    228 
    229 /*
    230  * Checksum extended UDP header and data.
    231  */
    232 
    233 int
    234 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
    235     int iphlen, int len)
    236 {
    237 
    238 	switch (af) {
    239 #ifdef INET
    240 	case AF_INET:
    241 		return udp4_input_checksum(m, uh, iphlen, len);
    242 #endif
    243 #ifdef INET6
    244 	case AF_INET6:
    245 		return udp6_input_checksum(m, uh, iphlen, len);
    246 #endif
    247 	}
    248 #ifdef DIAGNOSTIC
    249 	panic("udp_input_checksum: unknown af %d", af);
    250 #endif
    251 	/* NOTREACHED */
    252 	return -1;
    253 }
    254 
    255 #ifdef INET
    256 
    257 /*
    258  * Checksum extended UDP header and data.
    259  */
    260 
    261 static int
    262 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
    263     int iphlen, int len)
    264 {
    265 
    266 	/*
    267 	 * XXX it's better to record and check if this mbuf is
    268 	 * already checked.
    269 	 */
    270 
    271 	if (uh->uh_sum == 0)
    272 		return 0;
    273 
    274 	switch (m->m_pkthdr.csum_flags &
    275 	    ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
    276 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    277 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
    278 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
    279 		goto badcsum;
    280 
    281 	case M_CSUM_UDPv4|M_CSUM_DATA: {
    282 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
    283 
    284 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
    285 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
    286 			const struct ip *ip =
    287 			    mtod(m, const struct ip *);
    288 
    289 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
    290 			    ip->ip_dst.s_addr,
    291 			    htons(hw_csum + len + IPPROTO_UDP));
    292 		}
    293 		if ((hw_csum ^ 0xffff) != 0)
    294 			goto badcsum;
    295 		break;
    296 	}
    297 
    298 	case M_CSUM_UDPv4:
    299 		/* Checksum was okay. */
    300 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
    301 		break;
    302 
    303 	default:
    304 		/*
    305 		 * Need to compute it ourselves.  Maybe skip checksum
    306 		 * on loopback interfaces.
    307 		 */
    308 		if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
    309 				     IFF_LOOPBACK) ||
    310 				   udp_do_loopback_cksum)) {
    311 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
    312 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
    313 				goto badcsum;
    314 		}
    315 		break;
    316 	}
    317 
    318 	return 0;
    319 
    320 badcsum:
    321 	UDP_STATINC(UDP_STAT_BADSUM);
    322 	return -1;
    323 }
    324 
    325 void
    326 udp_input(struct mbuf *m, ...)
    327 {
    328 	va_list ap;
    329 	struct sockaddr_in src, dst;
    330 	struct ip *ip;
    331 	struct udphdr *uh;
    332 	int iphlen;
    333 	int len;
    334 	int n;
    335 	u_int16_t ip_len;
    336 
    337 	va_start(ap, m);
    338 	iphlen = va_arg(ap, int);
    339 	(void)va_arg(ap, int);		/* ignore value, advance ap */
    340 	va_end(ap);
    341 
    342 	MCLAIM(m, &udp_rx_mowner);
    343 	UDP_STATINC(UDP_STAT_IPACKETS);
    344 
    345 	/*
    346 	 * Get IP and UDP header together in first mbuf.
    347 	 */
    348 	ip = mtod(m, struct ip *);
    349 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
    350 	if (uh == NULL) {
    351 		UDP_STATINC(UDP_STAT_HDROPS);
    352 		return;
    353 	}
    354 	/*
    355 	 * Enforce alignment requirements that are violated in
    356 	 * some cases, see kern/50766 for details.
    357 	 */
    358 	if (UDP_HDR_ALIGNED_P(uh) == 0) {
    359 		m = m_copyup(m, iphlen + sizeof(struct udphdr), 0);
    360 		if (m == NULL) {
    361 			UDP_STATINC(UDP_STAT_HDROPS);
    362 			return;
    363 		}
    364 		ip = mtod(m, struct ip *);
    365 		uh = (struct udphdr *)(mtod(m, char *) + iphlen);
    366 	}
    367 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    368 
    369 	/* destination port of 0 is illegal, based on RFC768. */
    370 	if (uh->uh_dport == 0)
    371 		goto bad;
    372 
    373 	/*
    374 	 * Make mbuf data length reflect UDP length.
    375 	 * If not enough data to reflect UDP length, drop.
    376 	 */
    377 	ip_len = ntohs(ip->ip_len);
    378 	len = ntohs((u_int16_t)uh->uh_ulen);
    379 	if (ip_len != iphlen + len) {
    380 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
    381 			UDP_STATINC(UDP_STAT_BADLEN);
    382 			goto bad;
    383 		}
    384 		m_adj(m, iphlen + len - ip_len);
    385 	}
    386 
    387 	/*
    388 	 * Checksum extended UDP header and data.
    389 	 */
    390 	if (udp4_input_checksum(m, uh, iphlen, len))
    391 		goto badcsum;
    392 
    393 	/* construct source and dst sockaddrs. */
    394 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
    395 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
    396 
    397 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
    398 		UDP_STATINC(UDP_STAT_HDROPS);
    399 		return;
    400 	}
    401 	if (m == NULL) {
    402 		/*
    403 		 * packet has been processed by ESP stuff -
    404 		 * e.g. dropped NAT-T-keep-alive-packet ...
    405 		 */
    406 		return;
    407 	}
    408 	ip = mtod(m, struct ip *);
    409 #ifdef INET6
    410 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
    411 		struct sockaddr_in6 src6, dst6;
    412 
    413 		memset(&src6, 0, sizeof(src6));
    414 		src6.sin6_family = AF_INET6;
    415 		src6.sin6_len = sizeof(struct sockaddr_in6);
    416 		in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
    417 		src6.sin6_port = uh->uh_sport;
    418 		memset(&dst6, 0, sizeof(dst6));
    419 		dst6.sin6_family = AF_INET6;
    420 		dst6.sin6_len = sizeof(struct sockaddr_in6);
    421 		in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
    422 		dst6.sin6_port = uh->uh_dport;
    423 
    424 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
    425 	}
    426 #endif
    427 
    428 	if (n == 0) {
    429 		if (m->m_flags & (M_BCAST | M_MCAST)) {
    430 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
    431 			goto bad;
    432 		}
    433 		UDP_STATINC(UDP_STAT_NOPORT);
    434 #ifdef IPKDB
    435 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
    436 				m, iphlen + sizeof(struct udphdr),
    437 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
    438 			/*
    439 			 * It was a debugger connect packet,
    440 			 * just drop it now
    441 			 */
    442 			goto bad;
    443 		}
    444 #endif
    445 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
    446 		m = NULL;
    447 	}
    448 
    449 bad:
    450 	if (m)
    451 		m_freem(m);
    452 	return;
    453 
    454 badcsum:
    455 	m_freem(m);
    456 }
    457 #endif
    458 
    459 #ifdef INET
    460 static void
    461 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
    462 	struct sockaddr *src, struct socket *so)
    463 {
    464 	struct mbuf *opts = NULL;
    465 	struct mbuf *n;
    466 	struct inpcb *inp;
    467 
    468 	KASSERT(so != NULL);
    469 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET);
    470 	inp = sotoinpcb(so);
    471 	KASSERT(inp != NULL);
    472 
    473 #if defined(IPSEC)
    474 	/* check AH/ESP integrity. */
    475 	if (ipsec_used && ipsec4_in_reject(m, inp)) {
    476 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
    477 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
    478 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
    479 			    0, 0);
    480 		return;
    481 	}
    482 #endif /*IPSEC*/
    483 
    484 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
    485 		if (inp->inp_flags & INP_CONTROLOPTS
    486 		    || SOOPT_TIMESTAMP(so->so_options)) {
    487 			struct ip *ip = mtod(n, struct ip *);
    488 			ip_savecontrol(inp, &opts, ip, n);
    489 		}
    490 
    491 		m_adj(n, off);
    492 		if (sbappendaddr(&so->so_rcv, src, n,
    493 				opts) == 0) {
    494 			m_freem(n);
    495 			if (opts)
    496 				m_freem(opts);
    497 			so->so_rcv.sb_overflowed++;
    498 			UDP_STATINC(UDP_STAT_FULLSOCK);
    499 		} else
    500 			sorwakeup(so);
    501 	}
    502 }
    503 #endif
    504 
    505 #ifdef INET
    506 static int
    507 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
    508 	struct mbuf **mp, int off /* offset of udphdr */)
    509 {
    510 	u_int16_t *sport, *dport;
    511 	int rcvcnt;
    512 	struct in_addr *src4, *dst4;
    513 	struct inpcb_hdr *inph;
    514 	struct inpcb *inp;
    515 	struct mbuf *m = *mp;
    516 
    517 	rcvcnt = 0;
    518 	off += sizeof(struct udphdr);	/* now, offset of payload */
    519 
    520 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
    521 		goto bad;
    522 
    523 	src4 = &src->sin_addr;
    524 	sport = &src->sin_port;
    525 	dst4 = &dst->sin_addr;
    526 	dport = &dst->sin_port;
    527 
    528 	if (IN_MULTICAST(dst4->s_addr) ||
    529 	    in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
    530 		/*
    531 		 * Deliver a multicast or broadcast datagram to *all* sockets
    532 		 * for which the local and remote addresses and ports match
    533 		 * those of the incoming datagram.  This allows more than
    534 		 * one process to receive multi/broadcasts on the same port.
    535 		 * (This really ought to be done for unicast datagrams as
    536 		 * well, but that would cause problems with existing
    537 		 * applications that open both address-specific sockets and
    538 		 * a wildcard socket listening to the same port -- they would
    539 		 * end up receiving duplicates of every unicast datagram.
    540 		 * Those applications open the multiple sockets to overcome an
    541 		 * inadequacy of the UDP socket interface, but for backwards
    542 		 * compatibility we avoid the problem here rather than
    543 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    544 		 */
    545 
    546 		/*
    547 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    548 		 * we need udpiphdr for IPsec processing so we do that later.
    549 		 */
    550 		/*
    551 		 * Locate pcb(s) for datagram.
    552 		 */
    553 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
    554 			inp = (struct inpcb *)inph;
    555 			if (inp->inp_af != AF_INET)
    556 				continue;
    557 
    558 			if (inp->inp_lport != *dport)
    559 				continue;
    560 			if (!in_nullhost(inp->inp_laddr)) {
    561 				if (!in_hosteq(inp->inp_laddr, *dst4))
    562 					continue;
    563 			}
    564 			if (!in_nullhost(inp->inp_faddr)) {
    565 				if (!in_hosteq(inp->inp_faddr, *src4) ||
    566 				    inp->inp_fport != *sport)
    567 					continue;
    568 			}
    569 
    570 			udp4_sendup(m, off, (struct sockaddr *)src,
    571 				inp->inp_socket);
    572 			rcvcnt++;
    573 
    574 			/*
    575 			 * Don't look for additional matches if this one does
    576 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    577 			 * socket options set.  This heuristic avoids searching
    578 			 * through all pcbs in the common case of a non-shared
    579 			 * port.  It assumes that an application will never
    580 			 * clear these options after setting them.
    581 			 */
    582 			if ((inp->inp_socket->so_options &
    583 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    584 				break;
    585 		}
    586 	} else {
    587 		/*
    588 		 * Locate pcb for datagram.
    589 		 */
    590 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
    591 		    *dport, 0);
    592 		if (inp == 0) {
    593 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
    594 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
    595 			if (inp == 0)
    596 				return rcvcnt;
    597 		}
    598 
    599 #ifdef IPSEC
    600 		/* Handle ESP over UDP */
    601 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
    602 			struct sockaddr *sa = (struct sockaddr *)src;
    603 
    604 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
    605 			case -1: 	/* Error, m was freeed */
    606 				rcvcnt = -1;
    607 				goto bad;
    608 				break;
    609 
    610 			case 1:		/* ESP over UDP */
    611 				rcvcnt++;
    612 				goto bad;
    613 				break;
    614 
    615 			case 0: 	/* plain UDP */
    616 			default: 	/* Unexpected */
    617 				/*
    618 				 * Normal UDP processing will take place
    619 				 * m may have changed.
    620 				 */
    621 				m = *mp;
    622 				break;
    623 			}
    624 		}
    625 #endif
    626 
    627 		/*
    628 		 * Check the minimum TTL for socket.
    629 		 */
    630 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
    631 			goto bad;
    632 
    633 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
    634 		rcvcnt++;
    635 	}
    636 
    637 bad:
    638 	return rcvcnt;
    639 }
    640 #endif
    641 
    642 #ifdef INET
    643 /*
    644  * Notify a udp user of an asynchronous error;
    645  * just wake up so that he can collect error status.
    646  */
    647 static void
    648 udp_notify(struct inpcb *inp, int errno)
    649 {
    650 	inp->inp_socket->so_error = errno;
    651 	sorwakeup(inp->inp_socket);
    652 	sowwakeup(inp->inp_socket);
    653 }
    654 
    655 void *
    656 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
    657 {
    658 	struct ip *ip = v;
    659 	struct udphdr *uh;
    660 	void (*notify)(struct inpcb *, int) = udp_notify;
    661 	int errno;
    662 
    663 	if (sa->sa_family != AF_INET
    664 	 || sa->sa_len != sizeof(struct sockaddr_in))
    665 		return NULL;
    666 	if ((unsigned)cmd >= PRC_NCMDS)
    667 		return NULL;
    668 	errno = inetctlerrmap[cmd];
    669 	if (PRC_IS_REDIRECT(cmd))
    670 		notify = in_rtchange, ip = 0;
    671 	else if (cmd == PRC_HOSTDEAD)
    672 		ip = 0;
    673 	else if (errno == 0)
    674 		return NULL;
    675 	if (ip) {
    676 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
    677 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
    678 		    ip->ip_src, uh->uh_sport, errno, notify);
    679 
    680 		/* XXX mapped address case */
    681 	} else
    682 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
    683 		    notify);
    684 	return NULL;
    685 }
    686 
    687 int
    688 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
    689 {
    690 	int s;
    691 	int error = 0;
    692 	struct inpcb *inp;
    693 	int family;
    694 	int optval;
    695 
    696 	family = so->so_proto->pr_domain->dom_family;
    697 
    698 	s = splsoftnet();
    699 	switch (family) {
    700 #ifdef INET
    701 	case PF_INET:
    702 		if (sopt->sopt_level != IPPROTO_UDP) {
    703 			error = ip_ctloutput(op, so, sopt);
    704 			goto end;
    705 		}
    706 		break;
    707 #endif
    708 #ifdef INET6
    709 	case PF_INET6:
    710 		if (sopt->sopt_level != IPPROTO_UDP) {
    711 			error = ip6_ctloutput(op, so, sopt);
    712 			goto end;
    713 		}
    714 		break;
    715 #endif
    716 	default:
    717 		error = EAFNOSUPPORT;
    718 		goto end;
    719 	}
    720 
    721 
    722 	switch (op) {
    723 	case PRCO_SETOPT:
    724 		inp = sotoinpcb(so);
    725 
    726 		switch (sopt->sopt_name) {
    727 		case UDP_ENCAP:
    728 			error = sockopt_getint(sopt, &optval);
    729 			if (error)
    730 				break;
    731 
    732 			switch(optval) {
    733 			case 0:
    734 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    735 				break;
    736 
    737 			case UDP_ENCAP_ESPINUDP:
    738 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    739 				inp->inp_flags |= INP_ESPINUDP;
    740 				break;
    741 
    742 			case UDP_ENCAP_ESPINUDP_NON_IKE:
    743 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    744 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
    745 				break;
    746 			default:
    747 				error = EINVAL;
    748 				break;
    749 			}
    750 			break;
    751 
    752 		default:
    753 			error = ENOPROTOOPT;
    754 			break;
    755 		}
    756 		break;
    757 
    758 	default:
    759 		error = EINVAL;
    760 		break;
    761 	}
    762 
    763 end:
    764 	splx(s);
    765 	return error;
    766 }
    767 
    768 int
    769 udp_output(struct mbuf *m, struct inpcb *inp, struct mbuf *control,
    770     struct lwp *l)
    771 {
    772 	struct udpiphdr *ui;
    773 	struct route *ro;
    774 	struct ip_pktopts pktopts;
    775 	kauth_cred_t cred;
    776 	int len = m->m_pkthdr.len;
    777 	int error, flags = 0;
    778 
    779 	MCLAIM(m, &udp_tx_mowner);
    780 
    781 	/*
    782 	 * Calculate data length and get a mbuf
    783 	 * for UDP and IP headers.
    784 	 */
    785 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
    786 	if (m == 0) {
    787 		error = ENOBUFS;
    788 		goto release;
    789 	}
    790 
    791 	/*
    792 	 * Compute the packet length of the IP header, and
    793 	 * punt if the length looks bogus.
    794 	 */
    795 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
    796 		error = EMSGSIZE;
    797 		goto release;
    798 	}
    799 
    800 	if (l == NULL)
    801 		cred = NULL;
    802 	else
    803 		cred = l->l_cred;
    804 
    805 	/* Setup IP outgoing packet options */
    806 	memset(&pktopts, 0, sizeof(pktopts));
    807 	error = ip_setpktopts(control, &pktopts, &flags, inp, cred,
    808 	    IPPROTO_UDP);
    809 	if (error != 0)
    810 		goto release;
    811 
    812 	if (control != NULL) {
    813 		m_freem(control);
    814 		control = NULL;
    815 	}
    816 
    817 	/*
    818 	 * Fill in mbuf with extended UDP header
    819 	 * and addresses and length put into network format.
    820 	 */
    821 	ui = mtod(m, struct udpiphdr *);
    822 	ui->ui_pr = IPPROTO_UDP;
    823 	ui->ui_src = pktopts.ippo_laddr.sin_addr;
    824 	ui->ui_dst = inp->inp_faddr;
    825 	ui->ui_sport = inp->inp_lport;
    826 	ui->ui_dport = inp->inp_fport;
    827 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
    828 
    829 	ro = &inp->inp_route;
    830 
    831 	/*
    832 	 * Set up checksum and output datagram.
    833 	 */
    834 	if (udpcksum) {
    835 		/*
    836 		 * XXX Cache pseudo-header checksum part for
    837 		 * XXX "connected" UDP sockets.
    838 		 */
    839 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
    840 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
    841 		    sizeof(struct udphdr) + IPPROTO_UDP));
    842 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
    843 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
    844 	} else
    845 		ui->ui_sum = 0;
    846 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
    847 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
    848 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
    849 	UDP_STATINC(UDP_STAT_OPACKETS);
    850 
    851 	flags |= inp->inp_socket->so_options & (SO_DONTROUTE|SO_BROADCAST);
    852 	return ip_output(m, inp->inp_options, ro, flags, pktopts.ippo_imo, inp);
    853 
    854  release:
    855 	if (control != NULL)
    856 		m_freem(control);
    857 	m_freem(m);
    858 	return error;
    859 }
    860 
    861 static int
    862 udp_attach(struct socket *so, int proto)
    863 {
    864 	struct inpcb *inp;
    865 	int error;
    866 
    867 	KASSERT(sotoinpcb(so) == NULL);
    868 
    869 	/* Assign the lock (must happen even if we will error out). */
    870 	sosetlock(so);
    871 
    872 #ifdef MBUFTRACE
    873 	so->so_mowner = &udp_mowner;
    874 	so->so_rcv.sb_mowner = &udp_rx_mowner;
    875 	so->so_snd.sb_mowner = &udp_tx_mowner;
    876 #endif
    877 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
    878 		error = soreserve(so, udp_sendspace, udp_recvspace);
    879 		if (error) {
    880 			return error;
    881 		}
    882 	}
    883 
    884 	error = in_pcballoc(so, &udbtable);
    885 	if (error) {
    886 		return error;
    887 	}
    888 	inp = sotoinpcb(so);
    889 	inp->inp_ip.ip_ttl = ip_defttl;
    890 	KASSERT(solocked(so));
    891 
    892 	return error;
    893 }
    894 
    895 static void
    896 udp_detach(struct socket *so)
    897 {
    898 	struct inpcb *inp;
    899 
    900 	KASSERT(solocked(so));
    901 	inp = sotoinpcb(so);
    902 	KASSERT(inp != NULL);
    903 	in_pcbdetach(inp);
    904 }
    905 
    906 static int
    907 udp_accept(struct socket *so, struct sockaddr *nam)
    908 {
    909 	KASSERT(solocked(so));
    910 
    911 	panic("udp_accept");
    912 
    913 	return EOPNOTSUPP;
    914 }
    915 
    916 static int
    917 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
    918 {
    919 	struct inpcb *inp = sotoinpcb(so);
    920 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
    921 	int error = 0;
    922 	int s;
    923 
    924 	KASSERT(solocked(so));
    925 	KASSERT(inp != NULL);
    926 	KASSERT(nam != NULL);
    927 
    928 	s = splsoftnet();
    929 	error = in_pcbbind(inp, sin, l);
    930 	splx(s);
    931 
    932 	return error;
    933 }
    934 
    935 static int
    936 udp_listen(struct socket *so, struct lwp *l)
    937 {
    938 	KASSERT(solocked(so));
    939 
    940 	return EOPNOTSUPP;
    941 }
    942 
    943 static int
    944 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
    945 {
    946 	struct inpcb *inp = sotoinpcb(so);
    947 	int error = 0;
    948 	int s;
    949 
    950 	KASSERT(solocked(so));
    951 	KASSERT(inp != NULL);
    952 	KASSERT(nam != NULL);
    953 
    954 	s = splsoftnet();
    955 	error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
    956 	if (! error)
    957 		soisconnected(so);
    958 	splx(s);
    959 	return error;
    960 }
    961 
    962 static int
    963 udp_connect2(struct socket *so, struct socket *so2)
    964 {
    965 	KASSERT(solocked(so));
    966 
    967 	return EOPNOTSUPP;
    968 }
    969 
    970 static int
    971 udp_disconnect(struct socket *so)
    972 {
    973 	struct inpcb *inp = sotoinpcb(so);
    974 	int s;
    975 
    976 	KASSERT(solocked(so));
    977 	KASSERT(inp != NULL);
    978 
    979 	s = splsoftnet();
    980 	/*soisdisconnected(so);*/
    981 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
    982 	in_pcbdisconnect(inp);
    983 	inp->inp_laddr = zeroin_addr;		/* XXX */
    984 	in_pcbstate(inp, INP_BOUND);		/* XXX */
    985 	splx(s);
    986 
    987 	return 0;
    988 }
    989 
    990 static int
    991 udp_shutdown(struct socket *so)
    992 {
    993 	int s;
    994 
    995 	KASSERT(solocked(so));
    996 
    997 	s = splsoftnet();
    998 	socantsendmore(so);
    999 	splx(s);
   1000 
   1001 	return 0;
   1002 }
   1003 
   1004 static int
   1005 udp_abort(struct socket *so)
   1006 {
   1007 	KASSERT(solocked(so));
   1008 
   1009 	panic("udp_abort");
   1010 
   1011 	return EOPNOTSUPP;
   1012 }
   1013 
   1014 static int
   1015 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
   1016 {
   1017 	return in_control(so, cmd, nam, ifp);
   1018 }
   1019 
   1020 static int
   1021 udp_stat(struct socket *so, struct stat *ub)
   1022 {
   1023 	KASSERT(solocked(so));
   1024 
   1025 	/* stat: don't bother with a blocksize. */
   1026 	return 0;
   1027 }
   1028 
   1029 static int
   1030 udp_peeraddr(struct socket *so, struct sockaddr *nam)
   1031 {
   1032 	int s;
   1033 
   1034 	KASSERT(solocked(so));
   1035 	KASSERT(sotoinpcb(so) != NULL);
   1036 	KASSERT(nam != NULL);
   1037 
   1038 	s = splsoftnet();
   1039 	in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
   1040 	splx(s);
   1041 
   1042 	return 0;
   1043 }
   1044 
   1045 static int
   1046 udp_sockaddr(struct socket *so, struct sockaddr *nam)
   1047 {
   1048 	int s;
   1049 
   1050 	KASSERT(solocked(so));
   1051 	KASSERT(sotoinpcb(so) != NULL);
   1052 	KASSERT(nam != NULL);
   1053 
   1054 	s = splsoftnet();
   1055 	in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
   1056 	splx(s);
   1057 
   1058 	return 0;
   1059 }
   1060 
   1061 static int
   1062 udp_rcvd(struct socket *so, int flags, struct lwp *l)
   1063 {
   1064 	KASSERT(solocked(so));
   1065 
   1066 	return EOPNOTSUPP;
   1067 }
   1068 
   1069 static int
   1070 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
   1071 {
   1072 	KASSERT(solocked(so));
   1073 
   1074 	return EOPNOTSUPP;
   1075 }
   1076 
   1077 static int
   1078 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
   1079     struct mbuf *control, struct lwp *l)
   1080 {
   1081 	struct inpcb *inp = sotoinpcb(so);
   1082 	int error = 0;
   1083 	struct in_addr laddr;			/* XXX */
   1084 	int s;
   1085 
   1086 	KASSERT(solocked(so));
   1087 	KASSERT(inp != NULL);
   1088 	KASSERT(m != NULL);
   1089 
   1090 	memset(&laddr, 0, sizeof laddr);
   1091 
   1092 	s = splsoftnet();
   1093 	if (nam) {
   1094 		laddr = inp->inp_laddr;		/* XXX */
   1095 		if ((so->so_state & SS_ISCONNECTED) != 0) {
   1096 			error = EISCONN;
   1097 			goto die;
   1098 		}
   1099 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
   1100 		if (error)
   1101 			goto die;
   1102 	} else {
   1103 		if ((so->so_state & SS_ISCONNECTED) == 0) {
   1104 			error = ENOTCONN;
   1105 			goto die;
   1106 		}
   1107 	}
   1108 	error = udp_output(m, inp, control, l);
   1109 	m = NULL;
   1110 	control = NULL;
   1111 	if (nam) {
   1112 		in_pcbdisconnect(inp);
   1113 		inp->inp_laddr = laddr;		/* XXX */
   1114 		in_pcbstate(inp, INP_BOUND);	/* XXX */
   1115 	}
   1116   die:
   1117 	if (m != NULL)
   1118 		m_freem(m);
   1119 	if (control != NULL)
   1120 		m_freem(control);
   1121 
   1122 	splx(s);
   1123 	return error;
   1124 }
   1125 
   1126 static int
   1127 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
   1128 {
   1129 	KASSERT(solocked(so));
   1130 
   1131 	m_freem(m);
   1132 	m_freem(control);
   1133 
   1134 	return EOPNOTSUPP;
   1135 }
   1136 
   1137 static int
   1138 udp_purgeif(struct socket *so, struct ifnet *ifp)
   1139 {
   1140 	int s;
   1141 
   1142 	s = splsoftnet();
   1143 	mutex_enter(softnet_lock);
   1144 	in_pcbpurgeif0(&udbtable, ifp);
   1145 #ifdef NET_MPSAFE
   1146 	mutex_exit(softnet_lock);
   1147 #endif
   1148 	in_purgeif(ifp);
   1149 #ifdef NET_MPSAFE
   1150 	mutex_enter(softnet_lock);
   1151 #endif
   1152 	in_pcbpurgeif(&udbtable, ifp);
   1153 	mutex_exit(softnet_lock);
   1154 	splx(s);
   1155 
   1156 	return 0;
   1157 }
   1158 
   1159 static int
   1160 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
   1161 {
   1162 
   1163 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
   1164 }
   1165 
   1166 /*
   1167  * Sysctl for udp variables.
   1168  */
   1169 static void
   1170 sysctl_net_inet_udp_setup(struct sysctllog **clog)
   1171 {
   1172 
   1173 	sysctl_createv(clog, 0, NULL, NULL,
   1174 		       CTLFLAG_PERMANENT,
   1175 		       CTLTYPE_NODE, "inet", NULL,
   1176 		       NULL, 0, NULL, 0,
   1177 		       CTL_NET, PF_INET, CTL_EOL);
   1178 	sysctl_createv(clog, 0, NULL, NULL,
   1179 		       CTLFLAG_PERMANENT,
   1180 		       CTLTYPE_NODE, "udp",
   1181 		       SYSCTL_DESCR("UDPv4 related settings"),
   1182 		       NULL, 0, NULL, 0,
   1183 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
   1184 
   1185 	sysctl_createv(clog, 0, NULL, NULL,
   1186 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1187 		       CTLTYPE_INT, "checksum",
   1188 		       SYSCTL_DESCR("Compute UDP checksums"),
   1189 		       NULL, 0, &udpcksum, 0,
   1190 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
   1191 		       CTL_EOL);
   1192 	sysctl_createv(clog, 0, NULL, NULL,
   1193 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1194 		       CTLTYPE_INT, "sendspace",
   1195 		       SYSCTL_DESCR("Default UDP send buffer size"),
   1196 		       NULL, 0, &udp_sendspace, 0,
   1197 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
   1198 		       CTL_EOL);
   1199 	sysctl_createv(clog, 0, NULL, NULL,
   1200 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1201 		       CTLTYPE_INT, "recvspace",
   1202 		       SYSCTL_DESCR("Default UDP receive buffer size"),
   1203 		       NULL, 0, &udp_recvspace, 0,
   1204 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
   1205 		       CTL_EOL);
   1206 	sysctl_createv(clog, 0, NULL, NULL,
   1207 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1208 		       CTLTYPE_INT, "do_loopback_cksum",
   1209 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
   1210 		       NULL, 0, &udp_do_loopback_cksum, 0,
   1211 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
   1212 		       CTL_EOL);
   1213 	sysctl_createv(clog, 0, NULL, NULL,
   1214 		       CTLFLAG_PERMANENT,
   1215 		       CTLTYPE_STRUCT, "pcblist",
   1216 		       SYSCTL_DESCR("UDP protocol control block list"),
   1217 		       sysctl_inpcblist, 0, &udbtable, 0,
   1218 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
   1219 		       CTL_EOL);
   1220 	sysctl_createv(clog, 0, NULL, NULL,
   1221 		       CTLFLAG_PERMANENT,
   1222 		       CTLTYPE_STRUCT, "stats",
   1223 		       SYSCTL_DESCR("UDP statistics"),
   1224 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
   1225 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
   1226 		       CTL_EOL);
   1227 }
   1228 #endif
   1229 
   1230 void
   1231 udp_statinc(u_int stat)
   1232 {
   1233 
   1234 	KASSERT(stat < UDP_NSTATS);
   1235 	UDP_STATINC(stat);
   1236 }
   1237 
   1238 #if defined(INET) && defined(IPSEC)
   1239 /*
   1240  * Returns:
   1241  * 1 if the packet was processed
   1242  * 0 if normal UDP processing should take place
   1243  * -1 if an error occurent and m was freed
   1244  */
   1245 static int
   1246 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
   1247     struct socket *so)
   1248 {
   1249 	size_t len;
   1250 	void *data;
   1251 	struct inpcb *inp;
   1252 	size_t skip = 0;
   1253 	size_t minlen;
   1254 	size_t iphdrlen;
   1255 	struct ip *ip;
   1256 	struct m_tag *tag;
   1257 	struct udphdr *udphdr;
   1258 	u_int16_t sport, dport;
   1259 	struct mbuf *m = *mp;
   1260 
   1261 	/*
   1262 	 * Collapse the mbuf chain if the first mbuf is too short
   1263 	 * The longest case is: UDP + non ESP marker + ESP
   1264 	 */
   1265 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
   1266 	if (minlen > m->m_pkthdr.len)
   1267 		minlen = m->m_pkthdr.len;
   1268 
   1269 	if (m->m_len < minlen) {
   1270 		if ((*mp = m_pullup(m, minlen)) == NULL) {
   1271 			printf("udp4_espinudp: m_pullup failed\n");
   1272 			return -1;
   1273 		}
   1274 		m = *mp;
   1275 	}
   1276 
   1277 	len = m->m_len - off;
   1278 	data = mtod(m, char *) + off;
   1279 	inp = sotoinpcb(so);
   1280 
   1281 	/* Ignore keepalive packets */
   1282 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
   1283 		m_free(m);
   1284 		*mp = NULL; /* avoid any further processiong by caller ... */
   1285 		return 1;
   1286 	}
   1287 
   1288 	/*
   1289 	 * Check that the payload is long enough to hold
   1290 	 * an ESP header and compute the length of encapsulation
   1291 	 * header to remove
   1292 	 */
   1293 	if (inp->inp_flags & INP_ESPINUDP) {
   1294 		u_int32_t *st = (u_int32_t *)data;
   1295 
   1296 		if ((len <= sizeof(struct esp)) || (*st == 0))
   1297 			return 0; /* Normal UDP processing */
   1298 
   1299 		skip = sizeof(struct udphdr);
   1300 	}
   1301 
   1302 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
   1303 		u_int32_t *st = (u_int32_t *)data;
   1304 
   1305 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
   1306 		    || ((st[0] | st[1]) != 0))
   1307 			return 0; /* Normal UDP processing */
   1308 
   1309 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
   1310 	}
   1311 
   1312 	/*
   1313 	 * Get the UDP ports. They are handled in network
   1314 	 * order everywhere in IPSEC_NAT_T code.
   1315 	 */
   1316 	udphdr = (struct udphdr *)((char *)data - skip);
   1317 	sport = udphdr->uh_sport;
   1318 	dport = udphdr->uh_dport;
   1319 
   1320 	/*
   1321 	 * Remove the UDP header (and possibly the non ESP marker)
   1322 	 * IP header lendth is iphdrlen
   1323 	 * Before:
   1324 	 *   <--- off --->
   1325 	 *   +----+------+-----+
   1326 	 *   | IP |  UDP | ESP |
   1327 	 *   +----+------+-----+
   1328 	 *        <-skip->
   1329 	 * After:
   1330 	 *          +----+-----+
   1331 	 *          | IP | ESP |
   1332 	 *          +----+-----+
   1333 	 *   <-skip->
   1334 	 */
   1335 	iphdrlen = off - sizeof(struct udphdr);
   1336 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
   1337 	m_adj(m, skip);
   1338 
   1339 	ip = mtod(m, struct ip *);
   1340 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
   1341 	ip->ip_p = IPPROTO_ESP;
   1342 
   1343 	/*
   1344 	 * We have modified the packet - it is now ESP, so we should not
   1345 	 * return to UDP processing ...
   1346 	 *
   1347 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
   1348 	 * the source UDP port. This is required if we want
   1349 	 * to select the right SPD for multiple hosts behind
   1350 	 * same NAT
   1351 	 */
   1352 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
   1353 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
   1354 		printf("udp4_espinudp: m_tag_get failed\n");
   1355 		m_freem(m);
   1356 		return -1;
   1357 	}
   1358 	((u_int16_t *)(tag + 1))[0] = sport;
   1359 	((u_int16_t *)(tag + 1))[1] = dport;
   1360 	m_tag_prepend(m, tag);
   1361 
   1362 	if (ipsec_used)
   1363 		ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
   1364 	/* XXX: else */
   1365 
   1366 	/* We handled it, it shouldn't be handled by UDP */
   1367 	*mp = NULL; /* avoid free by caller ... */
   1368 	return 1;
   1369 }
   1370 #endif
   1371 
   1372 PR_WRAP_USRREQS(udp)
   1373 #define	udp_attach	udp_attach_wrapper
   1374 #define	udp_detach	udp_detach_wrapper
   1375 #define	udp_accept	udp_accept_wrapper
   1376 #define	udp_bind	udp_bind_wrapper
   1377 #define	udp_listen	udp_listen_wrapper
   1378 #define	udp_connect	udp_connect_wrapper
   1379 #define	udp_connect2	udp_connect2_wrapper
   1380 #define	udp_disconnect	udp_disconnect_wrapper
   1381 #define	udp_shutdown	udp_shutdown_wrapper
   1382 #define	udp_abort	udp_abort_wrapper
   1383 #define	udp_ioctl	udp_ioctl_wrapper
   1384 #define	udp_stat	udp_stat_wrapper
   1385 #define	udp_peeraddr	udp_peeraddr_wrapper
   1386 #define	udp_sockaddr	udp_sockaddr_wrapper
   1387 #define	udp_rcvd	udp_rcvd_wrapper
   1388 #define	udp_recvoob	udp_recvoob_wrapper
   1389 #define	udp_send	udp_send_wrapper
   1390 #define	udp_sendoob	udp_sendoob_wrapper
   1391 #define	udp_purgeif	udp_purgeif_wrapper
   1392 
   1393 const struct pr_usrreqs udp_usrreqs = {
   1394 	.pr_attach	= udp_attach,
   1395 	.pr_detach	= udp_detach,
   1396 	.pr_accept	= udp_accept,
   1397 	.pr_bind	= udp_bind,
   1398 	.pr_listen	= udp_listen,
   1399 	.pr_connect	= udp_connect,
   1400 	.pr_connect2	= udp_connect2,
   1401 	.pr_disconnect	= udp_disconnect,
   1402 	.pr_shutdown	= udp_shutdown,
   1403 	.pr_abort	= udp_abort,
   1404 	.pr_ioctl	= udp_ioctl,
   1405 	.pr_stat	= udp_stat,
   1406 	.pr_peeraddr	= udp_peeraddr,
   1407 	.pr_sockaddr	= udp_sockaddr,
   1408 	.pr_rcvd	= udp_rcvd,
   1409 	.pr_recvoob	= udp_recvoob,
   1410 	.pr_send	= udp_send,
   1411 	.pr_sendoob	= udp_sendoob,
   1412 	.pr_purgeif	= udp_purgeif,
   1413 };
   1414