Home | History | Annotate | Line # | Download | only in netinet
udp_usrreq.c revision 1.226.2.4
      1 /*	$NetBSD: udp_usrreq.c,v 1.226.2.4 2017/04/26 02:53:29 pgoyette 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 1.226.2.4 2017/04/26 02:53:29 pgoyette Exp $");
     70 
     71 #ifdef _KERNEL_OPT
     72 #include "opt_inet.h"
     73 #include "opt_compat_netbsd.h"
     74 #include "opt_ipsec.h"
     75 #include "opt_inet_csum.h"
     76 #include "opt_ipkdb.h"
     77 #include "opt_mbuftrace.h"
     78 #include "opt_net_mpsafe.h"
     79 #endif
     80 
     81 #include <sys/param.h>
     82 #include <sys/mbuf.h>
     83 #include <sys/once.h>
     84 #include <sys/protosw.h>
     85 #include <sys/socket.h>
     86 #include <sys/socketvar.h>
     87 #include <sys/systm.h>
     88 #include <sys/proc.h>
     89 #include <sys/domain.h>
     90 #include <sys/sysctl.h>
     91 
     92 #include <net/if.h>
     93 
     94 #include <netinet/in.h>
     95 #include <netinet/in_systm.h>
     96 #include <netinet/in_var.h>
     97 #include <netinet/ip.h>
     98 #include <netinet/in_pcb.h>
     99 #include <netinet/ip_var.h>
    100 #include <netinet/ip_icmp.h>
    101 #include <netinet/udp.h>
    102 #include <netinet/udp_var.h>
    103 #include <netinet/udp_private.h>
    104 
    105 #ifdef INET6
    106 #include <netinet/ip6.h>
    107 #include <netinet/icmp6.h>
    108 #include <netinet6/ip6_var.h>
    109 #include <netinet6/ip6_private.h>
    110 #include <netinet6/in6_pcb.h>
    111 #include <netinet6/udp6_var.h>
    112 #include <netinet6/udp6_private.h>
    113 #endif
    114 
    115 #ifndef INET6
    116 /* always need ip6.h for IP6_EXTHDR_GET */
    117 #include <netinet/ip6.h>
    118 #endif
    119 
    120 #ifdef IPSEC
    121 #include <netipsec/ipsec.h>
    122 #include <netipsec/ipsec_var.h>
    123 #include <netipsec/ipsec_private.h>
    124 #include <netipsec/esp.h>
    125 #ifdef INET6
    126 #include <netipsec/ipsec6.h>
    127 #endif
    128 #endif	/* IPSEC */
    129 
    130 #ifdef COMPAT_50
    131 #include <compat/sys/socket.h>
    132 #endif
    133 
    134 #ifdef IPKDB
    135 #include <ipkdb/ipkdb.h>
    136 #endif
    137 
    138 int	udpcksum = 1;
    139 int	udp_do_loopback_cksum = 0;
    140 
    141 struct	inpcbtable udbtable;
    142 
    143 percpu_t *udpstat_percpu;
    144 
    145 #ifdef INET
    146 #ifdef IPSEC
    147 static int udp4_espinudp (struct mbuf **, int, struct sockaddr *,
    148 	struct socket *);
    149 #endif
    150 static void udp4_sendup (struct mbuf *, int, struct sockaddr *,
    151 	struct socket *);
    152 static int udp4_realinput (struct sockaddr_in *, struct sockaddr_in *,
    153 	struct mbuf **, int);
    154 static int udp4_input_checksum(struct mbuf *, const struct udphdr *, int, int);
    155 #endif
    156 #ifdef INET
    157 static	void udp_notify (struct inpcb *, int);
    158 #endif
    159 
    160 #ifndef UDBHASHSIZE
    161 #define	UDBHASHSIZE	128
    162 #endif
    163 int	udbhashsize = UDBHASHSIZE;
    164 
    165 /*
    166  * For send - really max datagram size; for receive - 40 1K datagrams.
    167  */
    168 static int	udp_sendspace = 9216;
    169 static int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
    170 
    171 #ifdef MBUFTRACE
    172 struct mowner udp_mowner = MOWNER_INIT("udp", "");
    173 struct mowner udp_rx_mowner = MOWNER_INIT("udp", "rx");
    174 struct mowner udp_tx_mowner = MOWNER_INIT("udp", "tx");
    175 #endif
    176 
    177 #ifdef UDP_CSUM_COUNTERS
    178 #include <sys/device.h>
    179 
    180 #if defined(INET)
    181 struct evcnt udp_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    182     NULL, "udp", "hwcsum bad");
    183 struct evcnt udp_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    184     NULL, "udp", "hwcsum ok");
    185 struct evcnt udp_hwcsum_data = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    186     NULL, "udp", "hwcsum data");
    187 struct evcnt udp_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    188     NULL, "udp", "swcsum");
    189 
    190 EVCNT_ATTACH_STATIC(udp_hwcsum_bad);
    191 EVCNT_ATTACH_STATIC(udp_hwcsum_ok);
    192 EVCNT_ATTACH_STATIC(udp_hwcsum_data);
    193 EVCNT_ATTACH_STATIC(udp_swcsum);
    194 #endif /* defined(INET) */
    195 
    196 #define	UDP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    197 #else
    198 #define	UDP_CSUM_COUNTER_INCR(ev)	/* nothing */
    199 #endif /* UDP_CSUM_COUNTERS */
    200 
    201 static void sysctl_net_inet_udp_setup(struct sysctllog **);
    202 
    203 static int
    204 do_udpinit(void)
    205 {
    206 
    207 	in_pcbinit(&udbtable, udbhashsize, udbhashsize);
    208 	udpstat_percpu = percpu_alloc(sizeof(uint64_t) * UDP_NSTATS);
    209 
    210 	MOWNER_ATTACH(&udp_tx_mowner);
    211 	MOWNER_ATTACH(&udp_rx_mowner);
    212 	MOWNER_ATTACH(&udp_mowner);
    213 
    214 	return 0;
    215 }
    216 
    217 void
    218 udp_init_common(void)
    219 {
    220 	static ONCE_DECL(doudpinit);
    221 
    222 	RUN_ONCE(&doudpinit, do_udpinit);
    223 }
    224 
    225 void
    226 udp_init(void)
    227 {
    228 
    229 	sysctl_net_inet_udp_setup(NULL);
    230 
    231 	udp_init_common();
    232 }
    233 
    234 /*
    235  * Checksum extended UDP header and data.
    236  */
    237 
    238 int
    239 udp_input_checksum(int af, struct mbuf *m, const struct udphdr *uh,
    240     int iphlen, int len)
    241 {
    242 
    243 	switch (af) {
    244 #ifdef INET
    245 	case AF_INET:
    246 		return udp4_input_checksum(m, uh, iphlen, len);
    247 #endif
    248 #ifdef INET6
    249 	case AF_INET6:
    250 		return udp6_input_checksum(m, uh, iphlen, len);
    251 #endif
    252 	}
    253 #ifdef DIAGNOSTIC
    254 	panic("udp_input_checksum: unknown af %d", af);
    255 #endif
    256 	/* NOTREACHED */
    257 	return -1;
    258 }
    259 
    260 #ifdef INET
    261 
    262 /*
    263  * Checksum extended UDP header and data.
    264  */
    265 
    266 static int
    267 udp4_input_checksum(struct mbuf *m, const struct udphdr *uh,
    268     int iphlen, int len)
    269 {
    270 
    271 	/*
    272 	 * XXX it's better to record and check if this mbuf is
    273 	 * already checked.
    274 	 */
    275 
    276 	if (uh->uh_sum == 0)
    277 		return 0;
    278 
    279 	switch (m->m_pkthdr.csum_flags &
    280 	    ((m_get_rcvif_NOMPSAFE(m)->if_csum_flags_rx & M_CSUM_UDPv4) |
    281 	    M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
    282 	case M_CSUM_UDPv4|M_CSUM_TCP_UDP_BAD:
    283 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_bad);
    284 		goto badcsum;
    285 
    286 	case M_CSUM_UDPv4|M_CSUM_DATA: {
    287 		u_int32_t hw_csum = m->m_pkthdr.csum_data;
    288 
    289 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_data);
    290 		if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
    291 			const struct ip *ip =
    292 			    mtod(m, const struct ip *);
    293 
    294 			hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
    295 			    ip->ip_dst.s_addr,
    296 			    htons(hw_csum + len + IPPROTO_UDP));
    297 		}
    298 		if ((hw_csum ^ 0xffff) != 0)
    299 			goto badcsum;
    300 		break;
    301 	}
    302 
    303 	case M_CSUM_UDPv4:
    304 		/* Checksum was okay. */
    305 		UDP_CSUM_COUNTER_INCR(&udp_hwcsum_ok);
    306 		break;
    307 
    308 	default:
    309 		/*
    310 		 * Need to compute it ourselves.  Maybe skip checksum
    311 		 * on loopback interfaces.
    312 		 */
    313 		if (__predict_true(!(m_get_rcvif_NOMPSAFE(m)->if_flags &
    314 				     IFF_LOOPBACK) ||
    315 				   udp_do_loopback_cksum)) {
    316 			UDP_CSUM_COUNTER_INCR(&udp_swcsum);
    317 			if (in4_cksum(m, IPPROTO_UDP, iphlen, len) != 0)
    318 				goto badcsum;
    319 		}
    320 		break;
    321 	}
    322 
    323 	return 0;
    324 
    325 badcsum:
    326 	UDP_STATINC(UDP_STAT_BADSUM);
    327 	return -1;
    328 }
    329 
    330 void
    331 udp_input(struct mbuf *m, ...)
    332 {
    333 	va_list ap;
    334 	struct sockaddr_in src, dst;
    335 	struct ip *ip;
    336 	struct udphdr *uh;
    337 	int iphlen;
    338 	int len;
    339 	int n;
    340 	u_int16_t ip_len;
    341 
    342 	va_start(ap, m);
    343 	iphlen = va_arg(ap, int);
    344 	(void)va_arg(ap, int);		/* ignore value, advance ap */
    345 	va_end(ap);
    346 
    347 	MCLAIM(m, &udp_rx_mowner);
    348 	UDP_STATINC(UDP_STAT_IPACKETS);
    349 
    350 	/*
    351 	 * Get IP and UDP header together in first mbuf.
    352 	 */
    353 	ip = mtod(m, struct ip *);
    354 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
    355 	if (uh == NULL) {
    356 		UDP_STATINC(UDP_STAT_HDROPS);
    357 		return;
    358 	}
    359 	/*
    360 	 * Enforce alignment requirements that are violated in
    361 	 * some cases, see kern/50766 for details.
    362 	 */
    363 	if (UDP_HDR_ALIGNED_P(uh) == 0) {
    364 		m = m_copyup(m, iphlen + sizeof(struct udphdr), 0);
    365 		if (m == NULL) {
    366 			UDP_STATINC(UDP_STAT_HDROPS);
    367 			return;
    368 		}
    369 		ip = mtod(m, struct ip *);
    370 		uh = (struct udphdr *)(mtod(m, char *) + iphlen);
    371 	}
    372 	KASSERT(UDP_HDR_ALIGNED_P(uh));
    373 
    374 	/* destination port of 0 is illegal, based on RFC768. */
    375 	if (uh->uh_dport == 0)
    376 		goto bad;
    377 
    378 	/*
    379 	 * Make mbuf data length reflect UDP length.
    380 	 * If not enough data to reflect UDP length, drop.
    381 	 */
    382 	ip_len = ntohs(ip->ip_len);
    383 	len = ntohs((u_int16_t)uh->uh_ulen);
    384 	if (ip_len != iphlen + len) {
    385 		if (ip_len < iphlen + len || len < sizeof(struct udphdr)) {
    386 			UDP_STATINC(UDP_STAT_BADLEN);
    387 			goto bad;
    388 		}
    389 		m_adj(m, iphlen + len - ip_len);
    390 	}
    391 
    392 	/*
    393 	 * Checksum extended UDP header and data.
    394 	 */
    395 	if (udp4_input_checksum(m, uh, iphlen, len))
    396 		goto badcsum;
    397 
    398 	/* construct source and dst sockaddrs. */
    399 	sockaddr_in_init(&src, &ip->ip_src, uh->uh_sport);
    400 	sockaddr_in_init(&dst, &ip->ip_dst, uh->uh_dport);
    401 
    402 	if ((n = udp4_realinput(&src, &dst, &m, iphlen)) == -1) {
    403 		UDP_STATINC(UDP_STAT_HDROPS);
    404 		return;
    405 	}
    406 	if (m == NULL) {
    407 		/*
    408 		 * packet has been processed by ESP stuff -
    409 		 * e.g. dropped NAT-T-keep-alive-packet ...
    410 		 */
    411 		return;
    412 	}
    413 	ip = mtod(m, struct ip *);
    414 #ifdef INET6
    415 	if (IN_MULTICAST(ip->ip_dst.s_addr) || n == 0) {
    416 		struct sockaddr_in6 src6, dst6;
    417 
    418 		memset(&src6, 0, sizeof(src6));
    419 		src6.sin6_family = AF_INET6;
    420 		src6.sin6_len = sizeof(struct sockaddr_in6);
    421 		in6_in_2_v4mapin6(&ip->ip_src, &src6.sin6_addr);
    422 		src6.sin6_port = uh->uh_sport;
    423 		memset(&dst6, 0, sizeof(dst6));
    424 		dst6.sin6_family = AF_INET6;
    425 		dst6.sin6_len = sizeof(struct sockaddr_in6);
    426 		in6_in_2_v4mapin6(&ip->ip_dst, &dst6.sin6_addr);
    427 		dst6.sin6_port = uh->uh_dport;
    428 
    429 		n += udp6_realinput(AF_INET, &src6, &dst6, m, iphlen);
    430 	}
    431 #endif
    432 
    433 	if (n == 0) {
    434 		if (m->m_flags & (M_BCAST | M_MCAST)) {
    435 			UDP_STATINC(UDP_STAT_NOPORTBCAST);
    436 			goto bad;
    437 		}
    438 		UDP_STATINC(UDP_STAT_NOPORT);
    439 #ifdef IPKDB
    440 		if (checkipkdb(&ip->ip_src, uh->uh_sport, uh->uh_dport,
    441 				m, iphlen + sizeof(struct udphdr),
    442 				m->m_pkthdr.len - iphlen - sizeof(struct udphdr))) {
    443 			/*
    444 			 * It was a debugger connect packet,
    445 			 * just drop it now
    446 			 */
    447 			goto bad;
    448 		}
    449 #endif
    450 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
    451 		m = NULL;
    452 	}
    453 
    454 bad:
    455 	if (m)
    456 		m_freem(m);
    457 	return;
    458 
    459 badcsum:
    460 	m_freem(m);
    461 }
    462 #endif
    463 
    464 #ifdef INET
    465 static void
    466 udp4_sendup(struct mbuf *m, int off /* offset of data portion */,
    467 	struct sockaddr *src, struct socket *so)
    468 {
    469 	struct mbuf *opts = NULL;
    470 	struct mbuf *n;
    471 	struct inpcb *inp;
    472 
    473 	KASSERT(so != NULL);
    474 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET);
    475 	inp = sotoinpcb(so);
    476 	KASSERT(inp != NULL);
    477 
    478 #if defined(IPSEC)
    479 	/* check AH/ESP integrity. */
    480 	if (ipsec_used && ipsec4_in_reject(m, inp)) {
    481 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
    482 		if ((n = m_copypacket(m, M_DONTWAIT)) != NULL)
    483 			icmp_error(n, ICMP_UNREACH, ICMP_UNREACH_ADMIN_PROHIBIT,
    484 			    0, 0);
    485 		return;
    486 	}
    487 #endif /*IPSEC*/
    488 
    489 	if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
    490 		if (inp->inp_flags & INP_CONTROLOPTS
    491 #ifdef SO_OTIMESTAMP
    492 			 || so->so_options & SO_OTIMESTAMP
    493 #endif
    494 			 || so->so_options & SO_TIMESTAMP) {
    495 			struct ip *ip = mtod(n, struct ip *);
    496 			ip_savecontrol(inp, &opts, ip, n);
    497 		}
    498 
    499 		m_adj(n, off);
    500 		if (sbappendaddr(&so->so_rcv, src, n,
    501 				opts) == 0) {
    502 			m_freem(n);
    503 			if (opts)
    504 				m_freem(opts);
    505 			so->so_rcv.sb_overflowed++;
    506 			UDP_STATINC(UDP_STAT_FULLSOCK);
    507 		} else
    508 			sorwakeup(so);
    509 	}
    510 }
    511 #endif
    512 
    513 #ifdef INET
    514 static int
    515 udp4_realinput(struct sockaddr_in *src, struct sockaddr_in *dst,
    516 	struct mbuf **mp, int off /* offset of udphdr */)
    517 {
    518 	u_int16_t *sport, *dport;
    519 	int rcvcnt;
    520 	struct in_addr *src4, *dst4;
    521 	struct inpcb_hdr *inph;
    522 	struct inpcb *inp;
    523 	struct mbuf *m = *mp;
    524 
    525 	rcvcnt = 0;
    526 	off += sizeof(struct udphdr);	/* now, offset of payload */
    527 
    528 	if (src->sin_family != AF_INET || dst->sin_family != AF_INET)
    529 		goto bad;
    530 
    531 	src4 = &src->sin_addr;
    532 	sport = &src->sin_port;
    533 	dst4 = &dst->sin_addr;
    534 	dport = &dst->sin_port;
    535 
    536 	if (IN_MULTICAST(dst4->s_addr) ||
    537 	    in_broadcast(*dst4, m_get_rcvif_NOMPSAFE(m))) {
    538 		/*
    539 		 * Deliver a multicast or broadcast datagram to *all* sockets
    540 		 * for which the local and remote addresses and ports match
    541 		 * those of the incoming datagram.  This allows more than
    542 		 * one process to receive multi/broadcasts on the same port.
    543 		 * (This really ought to be done for unicast datagrams as
    544 		 * well, but that would cause problems with existing
    545 		 * applications that open both address-specific sockets and
    546 		 * a wildcard socket listening to the same port -- they would
    547 		 * end up receiving duplicates of every unicast datagram.
    548 		 * Those applications open the multiple sockets to overcome an
    549 		 * inadequacy of the UDP socket interface, but for backwards
    550 		 * compatibility we avoid the problem here rather than
    551 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
    552 		 */
    553 
    554 		/*
    555 		 * KAME note: traditionally we dropped udpiphdr from mbuf here.
    556 		 * we need udpiphdr for IPsec processing so we do that later.
    557 		 */
    558 		/*
    559 		 * Locate pcb(s) for datagram.
    560 		 */
    561 		TAILQ_FOREACH(inph, &udbtable.inpt_queue, inph_queue) {
    562 			inp = (struct inpcb *)inph;
    563 			if (inp->inp_af != AF_INET)
    564 				continue;
    565 
    566 			if (inp->inp_lport != *dport)
    567 				continue;
    568 			if (!in_nullhost(inp->inp_laddr)) {
    569 				if (!in_hosteq(inp->inp_laddr, *dst4))
    570 					continue;
    571 			}
    572 			if (!in_nullhost(inp->inp_faddr)) {
    573 				if (!in_hosteq(inp->inp_faddr, *src4) ||
    574 				    inp->inp_fport != *sport)
    575 					continue;
    576 			}
    577 
    578 			udp4_sendup(m, off, (struct sockaddr *)src,
    579 				inp->inp_socket);
    580 			rcvcnt++;
    581 
    582 			/*
    583 			 * Don't look for additional matches if this one does
    584 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
    585 			 * socket options set.  This heuristic avoids searching
    586 			 * through all pcbs in the common case of a non-shared
    587 			 * port.  It assumes that an application will never
    588 			 * clear these options after setting them.
    589 			 */
    590 			if ((inp->inp_socket->so_options &
    591 			    (SO_REUSEPORT|SO_REUSEADDR)) == 0)
    592 				break;
    593 		}
    594 	} else {
    595 		/*
    596 		 * Locate pcb for datagram.
    597 		 */
    598 		inp = in_pcblookup_connect(&udbtable, *src4, *sport, *dst4,
    599 		    *dport, 0);
    600 		if (inp == 0) {
    601 			UDP_STATINC(UDP_STAT_PCBHASHMISS);
    602 			inp = in_pcblookup_bind(&udbtable, *dst4, *dport);
    603 			if (inp == 0)
    604 				return rcvcnt;
    605 		}
    606 
    607 #ifdef IPSEC
    608 		/* Handle ESP over UDP */
    609 		if (inp->inp_flags & INP_ESPINUDP_ALL) {
    610 			struct sockaddr *sa = (struct sockaddr *)src;
    611 
    612 			switch(udp4_espinudp(mp, off, sa, inp->inp_socket)) {
    613 			case -1: 	/* Error, m was freeed */
    614 				rcvcnt = -1;
    615 				goto bad;
    616 				break;
    617 
    618 			case 1:		/* ESP over UDP */
    619 				rcvcnt++;
    620 				goto bad;
    621 				break;
    622 
    623 			case 0: 	/* plain UDP */
    624 			default: 	/* Unexpected */
    625 				/*
    626 				 * Normal UDP processing will take place
    627 				 * m may have changed.
    628 				 */
    629 				m = *mp;
    630 				break;
    631 			}
    632 		}
    633 #endif
    634 
    635 		/*
    636 		 * Check the minimum TTL for socket.
    637 		 */
    638 		if (mtod(m, struct ip *)->ip_ttl < inp->inp_ip_minttl)
    639 			goto bad;
    640 
    641 		udp4_sendup(m, off, (struct sockaddr *)src, inp->inp_socket);
    642 		rcvcnt++;
    643 	}
    644 
    645 bad:
    646 	return rcvcnt;
    647 }
    648 #endif
    649 
    650 #ifdef INET
    651 /*
    652  * Notify a udp user of an asynchronous error;
    653  * just wake up so that he can collect error status.
    654  */
    655 static void
    656 udp_notify(struct inpcb *inp, int errno)
    657 {
    658 	inp->inp_socket->so_error = errno;
    659 	sorwakeup(inp->inp_socket);
    660 	sowwakeup(inp->inp_socket);
    661 }
    662 
    663 void *
    664 udp_ctlinput(int cmd, const struct sockaddr *sa, void *v)
    665 {
    666 	struct ip *ip = v;
    667 	struct udphdr *uh;
    668 	void (*notify)(struct inpcb *, int) = udp_notify;
    669 	int errno;
    670 
    671 	if (sa->sa_family != AF_INET
    672 	 || sa->sa_len != sizeof(struct sockaddr_in))
    673 		return NULL;
    674 	if ((unsigned)cmd >= PRC_NCMDS)
    675 		return NULL;
    676 	errno = inetctlerrmap[cmd];
    677 	if (PRC_IS_REDIRECT(cmd))
    678 		notify = in_rtchange, ip = 0;
    679 	else if (cmd == PRC_HOSTDEAD)
    680 		ip = 0;
    681 	else if (errno == 0)
    682 		return NULL;
    683 	if (ip) {
    684 		uh = (struct udphdr *)((char *)ip + (ip->ip_hl << 2));
    685 		in_pcbnotify(&udbtable, satocsin(sa)->sin_addr, uh->uh_dport,
    686 		    ip->ip_src, uh->uh_sport, errno, notify);
    687 
    688 		/* XXX mapped address case */
    689 	} else
    690 		in_pcbnotifyall(&udbtable, satocsin(sa)->sin_addr, errno,
    691 		    notify);
    692 	return NULL;
    693 }
    694 
    695 int
    696 udp_ctloutput(int op, struct socket *so, struct sockopt *sopt)
    697 {
    698 	int s;
    699 	int error = 0;
    700 	struct inpcb *inp;
    701 	int family;
    702 	int optval;
    703 
    704 	family = so->so_proto->pr_domain->dom_family;
    705 
    706 	s = splsoftnet();
    707 	switch (family) {
    708 #ifdef INET
    709 	case PF_INET:
    710 		if (sopt->sopt_level != IPPROTO_UDP) {
    711 			error = ip_ctloutput(op, so, sopt);
    712 			goto end;
    713 		}
    714 		break;
    715 #endif
    716 #ifdef INET6
    717 	case PF_INET6:
    718 		if (sopt->sopt_level != IPPROTO_UDP) {
    719 			error = ip6_ctloutput(op, so, sopt);
    720 			goto end;
    721 		}
    722 		break;
    723 #endif
    724 	default:
    725 		error = EAFNOSUPPORT;
    726 		goto end;
    727 	}
    728 
    729 
    730 	switch (op) {
    731 	case PRCO_SETOPT:
    732 		inp = sotoinpcb(so);
    733 
    734 		switch (sopt->sopt_name) {
    735 		case UDP_ENCAP:
    736 			error = sockopt_getint(sopt, &optval);
    737 			if (error)
    738 				break;
    739 
    740 			switch(optval) {
    741 			case 0:
    742 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    743 				break;
    744 
    745 			case UDP_ENCAP_ESPINUDP:
    746 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    747 				inp->inp_flags |= INP_ESPINUDP;
    748 				break;
    749 
    750 			case UDP_ENCAP_ESPINUDP_NON_IKE:
    751 				inp->inp_flags &= ~INP_ESPINUDP_ALL;
    752 				inp->inp_flags |= INP_ESPINUDP_NON_IKE;
    753 				break;
    754 			default:
    755 				error = EINVAL;
    756 				break;
    757 			}
    758 			break;
    759 
    760 		default:
    761 			error = ENOPROTOOPT;
    762 			break;
    763 		}
    764 		break;
    765 
    766 	default:
    767 		error = EINVAL;
    768 		break;
    769 	}
    770 
    771 end:
    772 	splx(s);
    773 	return error;
    774 }
    775 
    776 
    777 int
    778 udp_output(struct mbuf *m, struct inpcb *inp)
    779 {
    780 	struct udpiphdr *ui;
    781 	struct route *ro;
    782 	int len = m->m_pkthdr.len;
    783 	int error = 0;
    784 
    785 	MCLAIM(m, &udp_tx_mowner);
    786 
    787 	/*
    788 	 * Calculate data length and get a mbuf
    789 	 * for UDP and IP headers.
    790 	 */
    791 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
    792 	if (m == 0) {
    793 		error = ENOBUFS;
    794 		goto release;
    795 	}
    796 
    797 	/*
    798 	 * Compute the packet length of the IP header, and
    799 	 * punt if the length looks bogus.
    800 	 */
    801 	if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
    802 		error = EMSGSIZE;
    803 		goto release;
    804 	}
    805 
    806 	/*
    807 	 * Fill in mbuf with extended UDP header
    808 	 * and addresses and length put into network format.
    809 	 */
    810 	ui = mtod(m, struct udpiphdr *);
    811 	ui->ui_pr = IPPROTO_UDP;
    812 	ui->ui_src = inp->inp_laddr;
    813 	ui->ui_dst = inp->inp_faddr;
    814 	ui->ui_sport = inp->inp_lport;
    815 	ui->ui_dport = inp->inp_fport;
    816 	ui->ui_ulen = htons((u_int16_t)len + sizeof(struct udphdr));
    817 
    818 	ro = &inp->inp_route;
    819 
    820 	/*
    821 	 * Set up checksum and output datagram.
    822 	 */
    823 	if (udpcksum) {
    824 		/*
    825 		 * XXX Cache pseudo-header checksum part for
    826 		 * XXX "connected" UDP sockets.
    827 		 */
    828 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
    829 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
    830 		    sizeof(struct udphdr) + IPPROTO_UDP));
    831 		m->m_pkthdr.csum_flags = M_CSUM_UDPv4;
    832 		m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
    833 	} else
    834 		ui->ui_sum = 0;
    835 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
    836 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;	/* XXX */
    837 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;	/* XXX */
    838 	UDP_STATINC(UDP_STAT_OPACKETS);
    839 
    840 	return (ip_output(m, inp->inp_options, ro,
    841 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
    842 	    inp->inp_moptions, inp));
    843 
    844 release:
    845 	m_freem(m);
    846 	return (error);
    847 }
    848 
    849 static int
    850 udp_attach(struct socket *so, int proto)
    851 {
    852 	struct inpcb *inp;
    853 	int error;
    854 
    855 	KASSERT(sotoinpcb(so) == NULL);
    856 
    857 	/* Assign the lock (must happen even if we will error out). */
    858 	sosetlock(so);
    859 
    860 #ifdef MBUFTRACE
    861 	so->so_mowner = &udp_mowner;
    862 	so->so_rcv.sb_mowner = &udp_rx_mowner;
    863 	so->so_snd.sb_mowner = &udp_tx_mowner;
    864 #endif
    865 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
    866 		error = soreserve(so, udp_sendspace, udp_recvspace);
    867 		if (error) {
    868 			return error;
    869 		}
    870 	}
    871 
    872 	error = in_pcballoc(so, &udbtable);
    873 	if (error) {
    874 		return error;
    875 	}
    876 	inp = sotoinpcb(so);
    877 	inp->inp_ip.ip_ttl = ip_defttl;
    878 	KASSERT(solocked(so));
    879 
    880 	return error;
    881 }
    882 
    883 static void
    884 udp_detach(struct socket *so)
    885 {
    886 	struct inpcb *inp;
    887 
    888 	KASSERT(solocked(so));
    889 	inp = sotoinpcb(so);
    890 	KASSERT(inp != NULL);
    891 	in_pcbdetach(inp);
    892 }
    893 
    894 static int
    895 udp_accept(struct socket *so, struct sockaddr *nam)
    896 {
    897 	KASSERT(solocked(so));
    898 
    899 	panic("udp_accept");
    900 
    901 	return EOPNOTSUPP;
    902 }
    903 
    904 static int
    905 udp_bind(struct socket *so, struct sockaddr *nam, struct lwp *l)
    906 {
    907 	struct inpcb *inp = sotoinpcb(so);
    908 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
    909 	int error = 0;
    910 	int s;
    911 
    912 	KASSERT(solocked(so));
    913 	KASSERT(inp != NULL);
    914 	KASSERT(nam != NULL);
    915 
    916 	s = splsoftnet();
    917 	error = in_pcbbind(inp, sin, l);
    918 	splx(s);
    919 
    920 	return error;
    921 }
    922 
    923 static int
    924 udp_listen(struct socket *so, struct lwp *l)
    925 {
    926 	KASSERT(solocked(so));
    927 
    928 	return EOPNOTSUPP;
    929 }
    930 
    931 static int
    932 udp_connect(struct socket *so, struct sockaddr *nam, struct lwp *l)
    933 {
    934 	struct inpcb *inp = sotoinpcb(so);
    935 	int error = 0;
    936 	int s;
    937 
    938 	KASSERT(solocked(so));
    939 	KASSERT(inp != NULL);
    940 	KASSERT(nam != NULL);
    941 
    942 	s = splsoftnet();
    943 	error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
    944 	if (! error)
    945 		soisconnected(so);
    946 	splx(s);
    947 	return error;
    948 }
    949 
    950 static int
    951 udp_connect2(struct socket *so, struct socket *so2)
    952 {
    953 	KASSERT(solocked(so));
    954 
    955 	return EOPNOTSUPP;
    956 }
    957 
    958 static int
    959 udp_disconnect(struct socket *so)
    960 {
    961 	struct inpcb *inp = sotoinpcb(so);
    962 	int s;
    963 
    964 	KASSERT(solocked(so));
    965 	KASSERT(inp != NULL);
    966 
    967 	s = splsoftnet();
    968 	/*soisdisconnected(so);*/
    969 	so->so_state &= ~SS_ISCONNECTED;	/* XXX */
    970 	in_pcbdisconnect(inp);
    971 	inp->inp_laddr = zeroin_addr;		/* XXX */
    972 	in_pcbstate(inp, INP_BOUND);		/* XXX */
    973 	splx(s);
    974 
    975 	return 0;
    976 }
    977 
    978 static int
    979 udp_shutdown(struct socket *so)
    980 {
    981 	int s;
    982 
    983 	KASSERT(solocked(so));
    984 
    985 	s = splsoftnet();
    986 	socantsendmore(so);
    987 	splx(s);
    988 
    989 	return 0;
    990 }
    991 
    992 static int
    993 udp_abort(struct socket *so)
    994 {
    995 	KASSERT(solocked(so));
    996 
    997 	panic("udp_abort");
    998 
    999 	return EOPNOTSUPP;
   1000 }
   1001 
   1002 static int
   1003 udp_ioctl(struct socket *so, u_long cmd, void *nam, struct ifnet *ifp)
   1004 {
   1005 	return in_control(so, cmd, nam, ifp);
   1006 }
   1007 
   1008 static int
   1009 udp_stat(struct socket *so, struct stat *ub)
   1010 {
   1011 	KASSERT(solocked(so));
   1012 
   1013 	/* stat: don't bother with a blocksize. */
   1014 	return 0;
   1015 }
   1016 
   1017 static int
   1018 udp_peeraddr(struct socket *so, struct sockaddr *nam)
   1019 {
   1020 	int s;
   1021 
   1022 	KASSERT(solocked(so));
   1023 	KASSERT(sotoinpcb(so) != NULL);
   1024 	KASSERT(nam != NULL);
   1025 
   1026 	s = splsoftnet();
   1027 	in_setpeeraddr(sotoinpcb(so), (struct sockaddr_in *)nam);
   1028 	splx(s);
   1029 
   1030 	return 0;
   1031 }
   1032 
   1033 static int
   1034 udp_sockaddr(struct socket *so, struct sockaddr *nam)
   1035 {
   1036 	int s;
   1037 
   1038 	KASSERT(solocked(so));
   1039 	KASSERT(sotoinpcb(so) != NULL);
   1040 	KASSERT(nam != NULL);
   1041 
   1042 	s = splsoftnet();
   1043 	in_setsockaddr(sotoinpcb(so), (struct sockaddr_in *)nam);
   1044 	splx(s);
   1045 
   1046 	return 0;
   1047 }
   1048 
   1049 static int
   1050 udp_rcvd(struct socket *so, int flags, struct lwp *l)
   1051 {
   1052 	KASSERT(solocked(so));
   1053 
   1054 	return EOPNOTSUPP;
   1055 }
   1056 
   1057 static int
   1058 udp_recvoob(struct socket *so, struct mbuf *m, int flags)
   1059 {
   1060 	KASSERT(solocked(so));
   1061 
   1062 	return EOPNOTSUPP;
   1063 }
   1064 
   1065 static int
   1066 udp_send(struct socket *so, struct mbuf *m, struct sockaddr *nam,
   1067     struct mbuf *control, struct lwp *l)
   1068 {
   1069 	struct inpcb *inp = sotoinpcb(so);
   1070 	int error = 0;
   1071 	struct in_addr laddr;			/* XXX */
   1072 	int s;
   1073 
   1074 	KASSERT(solocked(so));
   1075 	KASSERT(inp != NULL);
   1076 	KASSERT(m != NULL);
   1077 
   1078 	if (control && control->m_len) {
   1079 		m_freem(control);
   1080 		m_freem(m);
   1081 		return EINVAL;
   1082 	}
   1083 
   1084 	memset(&laddr, 0, sizeof laddr);
   1085 
   1086 	s = splsoftnet();
   1087 	if (nam) {
   1088 		laddr = inp->inp_laddr;		/* XXX */
   1089 		if ((so->so_state & SS_ISCONNECTED) != 0) {
   1090 			error = EISCONN;
   1091 			goto die;
   1092 		}
   1093 		error = in_pcbconnect(inp, (struct sockaddr_in *)nam, l);
   1094 		if (error)
   1095 			goto die;
   1096 	} else {
   1097 		if ((so->so_state & SS_ISCONNECTED) == 0) {
   1098 			error = ENOTCONN;
   1099 			goto die;
   1100 		}
   1101 	}
   1102 	error = udp_output(m, inp);
   1103 	m = NULL;
   1104 	if (nam) {
   1105 		in_pcbdisconnect(inp);
   1106 		inp->inp_laddr = laddr;		/* XXX */
   1107 		in_pcbstate(inp, INP_BOUND);	/* XXX */
   1108 	}
   1109   die:
   1110 	if (m)
   1111 		m_freem(m);
   1112 
   1113 	splx(s);
   1114 	return error;
   1115 }
   1116 
   1117 static int
   1118 udp_sendoob(struct socket *so, struct mbuf *m, struct mbuf *control)
   1119 {
   1120 	KASSERT(solocked(so));
   1121 
   1122 	m_freem(m);
   1123 	m_freem(control);
   1124 
   1125 	return EOPNOTSUPP;
   1126 }
   1127 
   1128 static int
   1129 udp_purgeif(struct socket *so, struct ifnet *ifp)
   1130 {
   1131 	int s;
   1132 
   1133 	s = splsoftnet();
   1134 	mutex_enter(softnet_lock);
   1135 	in_pcbpurgeif0(&udbtable, ifp);
   1136 #ifdef NET_MPSAFE
   1137 	mutex_exit(softnet_lock);
   1138 #endif
   1139 	in_purgeif(ifp);
   1140 #ifdef NET_MPSAFE
   1141 	mutex_enter(softnet_lock);
   1142 #endif
   1143 	in_pcbpurgeif(&udbtable, ifp);
   1144 	mutex_exit(softnet_lock);
   1145 	splx(s);
   1146 
   1147 	return 0;
   1148 }
   1149 
   1150 static int
   1151 sysctl_net_inet_udp_stats(SYSCTLFN_ARGS)
   1152 {
   1153 
   1154 	return (NETSTAT_SYSCTL(udpstat_percpu, UDP_NSTATS));
   1155 }
   1156 
   1157 /*
   1158  * Sysctl for udp variables.
   1159  */
   1160 static void
   1161 sysctl_net_inet_udp_setup(struct sysctllog **clog)
   1162 {
   1163 
   1164 	sysctl_createv(clog, 0, NULL, NULL,
   1165 		       CTLFLAG_PERMANENT,
   1166 		       CTLTYPE_NODE, "inet", NULL,
   1167 		       NULL, 0, NULL, 0,
   1168 		       CTL_NET, PF_INET, CTL_EOL);
   1169 	sysctl_createv(clog, 0, NULL, NULL,
   1170 		       CTLFLAG_PERMANENT,
   1171 		       CTLTYPE_NODE, "udp",
   1172 		       SYSCTL_DESCR("UDPv4 related settings"),
   1173 		       NULL, 0, NULL, 0,
   1174 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_EOL);
   1175 
   1176 	sysctl_createv(clog, 0, NULL, NULL,
   1177 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1178 		       CTLTYPE_INT, "checksum",
   1179 		       SYSCTL_DESCR("Compute UDP checksums"),
   1180 		       NULL, 0, &udpcksum, 0,
   1181 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_CHECKSUM,
   1182 		       CTL_EOL);
   1183 	sysctl_createv(clog, 0, NULL, NULL,
   1184 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1185 		       CTLTYPE_INT, "sendspace",
   1186 		       SYSCTL_DESCR("Default UDP send buffer size"),
   1187 		       NULL, 0, &udp_sendspace, 0,
   1188 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_SENDSPACE,
   1189 		       CTL_EOL);
   1190 	sysctl_createv(clog, 0, NULL, NULL,
   1191 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1192 		       CTLTYPE_INT, "recvspace",
   1193 		       SYSCTL_DESCR("Default UDP receive buffer size"),
   1194 		       NULL, 0, &udp_recvspace, 0,
   1195 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_RECVSPACE,
   1196 		       CTL_EOL);
   1197 	sysctl_createv(clog, 0, NULL, NULL,
   1198 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1199 		       CTLTYPE_INT, "do_loopback_cksum",
   1200 		       SYSCTL_DESCR("Perform UDP checksum on loopback"),
   1201 		       NULL, 0, &udp_do_loopback_cksum, 0,
   1202 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_LOOPBACKCKSUM,
   1203 		       CTL_EOL);
   1204 	sysctl_createv(clog, 0, NULL, NULL,
   1205 		       CTLFLAG_PERMANENT,
   1206 		       CTLTYPE_STRUCT, "pcblist",
   1207 		       SYSCTL_DESCR("UDP protocol control block list"),
   1208 		       sysctl_inpcblist, 0, &udbtable, 0,
   1209 		       CTL_NET, PF_INET, IPPROTO_UDP, CTL_CREATE,
   1210 		       CTL_EOL);
   1211 	sysctl_createv(clog, 0, NULL, NULL,
   1212 		       CTLFLAG_PERMANENT,
   1213 		       CTLTYPE_STRUCT, "stats",
   1214 		       SYSCTL_DESCR("UDP statistics"),
   1215 		       sysctl_net_inet_udp_stats, 0, NULL, 0,
   1216 		       CTL_NET, PF_INET, IPPROTO_UDP, UDPCTL_STATS,
   1217 		       CTL_EOL);
   1218 }
   1219 #endif
   1220 
   1221 void
   1222 udp_statinc(u_int stat)
   1223 {
   1224 
   1225 	KASSERT(stat < UDP_NSTATS);
   1226 	UDP_STATINC(stat);
   1227 }
   1228 
   1229 #if defined(INET) && defined(IPSEC)
   1230 /*
   1231  * Returns:
   1232  * 1 if the packet was processed
   1233  * 0 if normal UDP processing should take place
   1234  * -1 if an error occurent and m was freed
   1235  */
   1236 static int
   1237 udp4_espinudp(struct mbuf **mp, int off, struct sockaddr *src,
   1238     struct socket *so)
   1239 {
   1240 	size_t len;
   1241 	void *data;
   1242 	struct inpcb *inp;
   1243 	size_t skip = 0;
   1244 	size_t minlen;
   1245 	size_t iphdrlen;
   1246 	struct ip *ip;
   1247 	struct m_tag *tag;
   1248 	struct udphdr *udphdr;
   1249 	u_int16_t sport, dport;
   1250 	struct mbuf *m = *mp;
   1251 
   1252 	/*
   1253 	 * Collapse the mbuf chain if the first mbuf is too short
   1254 	 * The longest case is: UDP + non ESP marker + ESP
   1255 	 */
   1256 	minlen = off + sizeof(u_int64_t) + sizeof(struct esp);
   1257 	if (minlen > m->m_pkthdr.len)
   1258 		minlen = m->m_pkthdr.len;
   1259 
   1260 	if (m->m_len < minlen) {
   1261 		if ((*mp = m_pullup(m, minlen)) == NULL) {
   1262 			printf("udp4_espinudp: m_pullup failed\n");
   1263 			return -1;
   1264 		}
   1265 		m = *mp;
   1266 	}
   1267 
   1268 	len = m->m_len - off;
   1269 	data = mtod(m, char *) + off;
   1270 	inp = sotoinpcb(so);
   1271 
   1272 	/* Ignore keepalive packets */
   1273 	if ((len == 1) && (*(unsigned char *)data == 0xff)) {
   1274 		m_free(m);
   1275 		*mp = NULL; /* avoid any further processiong by caller ... */
   1276 		return 1;
   1277 	}
   1278 
   1279 	/*
   1280 	 * Check that the payload is long enough to hold
   1281 	 * an ESP header and compute the length of encapsulation
   1282 	 * header to remove
   1283 	 */
   1284 	if (inp->inp_flags & INP_ESPINUDP) {
   1285 		u_int32_t *st = (u_int32_t *)data;
   1286 
   1287 		if ((len <= sizeof(struct esp)) || (*st == 0))
   1288 			return 0; /* Normal UDP processing */
   1289 
   1290 		skip = sizeof(struct udphdr);
   1291 	}
   1292 
   1293 	if (inp->inp_flags & INP_ESPINUDP_NON_IKE) {
   1294 		u_int32_t *st = (u_int32_t *)data;
   1295 
   1296 		if ((len <= sizeof(u_int64_t) + sizeof(struct esp))
   1297 		    || ((st[0] | st[1]) != 0))
   1298 			return 0; /* Normal UDP processing */
   1299 
   1300 		skip = sizeof(struct udphdr) + sizeof(u_int64_t);
   1301 	}
   1302 
   1303 	/*
   1304 	 * Get the UDP ports. They are handled in network
   1305 	 * order everywhere in IPSEC_NAT_T code.
   1306 	 */
   1307 	udphdr = (struct udphdr *)((char *)data - skip);
   1308 	sport = udphdr->uh_sport;
   1309 	dport = udphdr->uh_dport;
   1310 
   1311 	/*
   1312 	 * Remove the UDP header (and possibly the non ESP marker)
   1313 	 * IP header lendth is iphdrlen
   1314 	 * Before:
   1315 	 *   <--- off --->
   1316 	 *   +----+------+-----+
   1317 	 *   | IP |  UDP | ESP |
   1318 	 *   +----+------+-----+
   1319 	 *        <-skip->
   1320 	 * After:
   1321 	 *          +----+-----+
   1322 	 *          | IP | ESP |
   1323 	 *          +----+-----+
   1324 	 *   <-skip->
   1325 	 */
   1326 	iphdrlen = off - sizeof(struct udphdr);
   1327 	memmove(mtod(m, char *) + skip, mtod(m, void *), iphdrlen);
   1328 	m_adj(m, skip);
   1329 
   1330 	ip = mtod(m, struct ip *);
   1331 	ip->ip_len = htons(ntohs(ip->ip_len) - skip);
   1332 	ip->ip_p = IPPROTO_ESP;
   1333 
   1334 	/*
   1335 	 * We have modified the packet - it is now ESP, so we should not
   1336 	 * return to UDP processing ...
   1337 	 *
   1338 	 * Add a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
   1339 	 * the source UDP port. This is required if we want
   1340 	 * to select the right SPD for multiple hosts behind
   1341 	 * same NAT
   1342 	 */
   1343 	if ((tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
   1344 	    sizeof(sport) + sizeof(dport), M_DONTWAIT)) == NULL) {
   1345 		printf("udp4_espinudp: m_tag_get failed\n");
   1346 		m_freem(m);
   1347 		return -1;
   1348 	}
   1349 	((u_int16_t *)(tag + 1))[0] = sport;
   1350 	((u_int16_t *)(tag + 1))[1] = dport;
   1351 	m_tag_prepend(m, tag);
   1352 
   1353 	if (ipsec_used)
   1354 		ipsec4_common_input(m, iphdrlen, IPPROTO_ESP);
   1355 	/* XXX: else */
   1356 
   1357 	/* We handled it, it shouldn't be handled by UDP */
   1358 	*mp = NULL; /* avoid free by caller ... */
   1359 	return 1;
   1360 }
   1361 #endif
   1362 
   1363 PR_WRAP_USRREQS(udp)
   1364 #define	udp_attach	udp_attach_wrapper
   1365 #define	udp_detach	udp_detach_wrapper
   1366 #define	udp_accept	udp_accept_wrapper
   1367 #define	udp_bind	udp_bind_wrapper
   1368 #define	udp_listen	udp_listen_wrapper
   1369 #define	udp_connect	udp_connect_wrapper
   1370 #define	udp_connect2	udp_connect2_wrapper
   1371 #define	udp_disconnect	udp_disconnect_wrapper
   1372 #define	udp_shutdown	udp_shutdown_wrapper
   1373 #define	udp_abort	udp_abort_wrapper
   1374 #define	udp_ioctl	udp_ioctl_wrapper
   1375 #define	udp_stat	udp_stat_wrapper
   1376 #define	udp_peeraddr	udp_peeraddr_wrapper
   1377 #define	udp_sockaddr	udp_sockaddr_wrapper
   1378 #define	udp_rcvd	udp_rcvd_wrapper
   1379 #define	udp_recvoob	udp_recvoob_wrapper
   1380 #define	udp_send	udp_send_wrapper
   1381 #define	udp_sendoob	udp_sendoob_wrapper
   1382 #define	udp_purgeif	udp_purgeif_wrapper
   1383 
   1384 const struct pr_usrreqs udp_usrreqs = {
   1385 	.pr_attach	= udp_attach,
   1386 	.pr_detach	= udp_detach,
   1387 	.pr_accept	= udp_accept,
   1388 	.pr_bind	= udp_bind,
   1389 	.pr_listen	= udp_listen,
   1390 	.pr_connect	= udp_connect,
   1391 	.pr_connect2	= udp_connect2,
   1392 	.pr_disconnect	= udp_disconnect,
   1393 	.pr_shutdown	= udp_shutdown,
   1394 	.pr_abort	= udp_abort,
   1395 	.pr_ioctl	= udp_ioctl,
   1396 	.pr_stat	= udp_stat,
   1397 	.pr_peeraddr	= udp_peeraddr,
   1398 	.pr_sockaddr	= udp_sockaddr,
   1399 	.pr_rcvd	= udp_rcvd,
   1400 	.pr_recvoob	= udp_recvoob,
   1401 	.pr_send	= udp_send,
   1402 	.pr_sendoob	= udp_sendoob,
   1403 	.pr_purgeif	= udp_purgeif,
   1404 };
   1405