Home | History | Annotate | Line # | Download | only in netinet
ip_input.c revision 1.164
      1 /*	$NetBSD: ip_input.c,v 1.164 2003/02/26 06:31:14 matt 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) 1998 The NetBSD Foundation, Inc.
     34  * All rights reserved.
     35  *
     36  * This code is derived from software contributed to The NetBSD Foundation
     37  * by Public Access Networks Corporation ("Panix").  It was developed under
     38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     39  *
     40  * Redistribution and use in source and binary forms, with or without
     41  * modification, are permitted provided that the following conditions
     42  * are met:
     43  * 1. Redistributions of source code must retain the above copyright
     44  *    notice, this list of conditions and the following disclaimer.
     45  * 2. Redistributions in binary form must reproduce the above copyright
     46  *    notice, this list of conditions and the following disclaimer in the
     47  *    documentation and/or other materials provided with the distribution.
     48  * 3. All advertising materials mentioning features or use of this software
     49  *    must display the following acknowledgement:
     50  *	This product includes software developed by the NetBSD
     51  *	Foundation, Inc. and its contributors.
     52  * 4. Neither the name of The NetBSD Foundation nor the names of its
     53  *    contributors may be used to endorse or promote products derived
     54  *    from this software without specific prior written permission.
     55  *
     56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66  * POSSIBILITY OF SUCH DAMAGE.
     67  */
     68 
     69 /*
     70  * Copyright (c) 1982, 1986, 1988, 1993
     71  *	The Regents of the University of California.  All rights reserved.
     72  *
     73  * Redistribution and use in source and binary forms, with or without
     74  * modification, are permitted provided that the following conditions
     75  * are met:
     76  * 1. Redistributions of source code must retain the above copyright
     77  *    notice, this list of conditions and the following disclaimer.
     78  * 2. Redistributions in binary form must reproduce the above copyright
     79  *    notice, this list of conditions and the following disclaimer in the
     80  *    documentation and/or other materials provided with the distribution.
     81  * 3. All advertising materials mentioning features or use of this software
     82  *    must display the following acknowledgement:
     83  *	This product includes software developed by the University of
     84  *	California, Berkeley and its contributors.
     85  * 4. Neither the name of the University nor the names of its contributors
     86  *    may be used to endorse or promote products derived from this software
     87  *    without specific prior written permission.
     88  *
     89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     99  * SUCH DAMAGE.
    100  *
    101  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
    102  */
    103 
    104 #include <sys/cdefs.h>
    105 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.164 2003/02/26 06:31:14 matt Exp $");
    106 
    107 #include "opt_gateway.h"
    108 #include "opt_pfil_hooks.h"
    109 #include "opt_ipsec.h"
    110 #include "opt_mrouting.h"
    111 #include "opt_inet_csum.h"
    112 
    113 #include <sys/param.h>
    114 #include <sys/systm.h>
    115 #include <sys/malloc.h>
    116 #include <sys/mbuf.h>
    117 #include <sys/domain.h>
    118 #include <sys/protosw.h>
    119 #include <sys/socket.h>
    120 #include <sys/socketvar.h>
    121 #include <sys/errno.h>
    122 #include <sys/time.h>
    123 #include <sys/kernel.h>
    124 #include <sys/pool.h>
    125 #include <sys/sysctl.h>
    126 
    127 #include <net/if.h>
    128 #include <net/if_dl.h>
    129 #include <net/route.h>
    130 #include <net/pfil.h>
    131 
    132 #include <netinet/in.h>
    133 #include <netinet/in_systm.h>
    134 #include <netinet/ip.h>
    135 #include <netinet/in_pcb.h>
    136 #include <netinet/in_var.h>
    137 #include <netinet/ip_var.h>
    138 #include <netinet/ip_icmp.h>
    139 /* just for gif_ttl */
    140 #include <netinet/in_gif.h>
    141 #include "gif.h"
    142 #include <net/if_gre.h>
    143 #include "gre.h"
    144 
    145 #ifdef MROUTING
    146 #include <netinet/ip_mroute.h>
    147 #endif
    148 
    149 #ifdef IPSEC
    150 #include <netinet6/ipsec.h>
    151 #include <netkey/key.h>
    152 #endif
    153 
    154 #ifndef	IPFORWARDING
    155 #ifdef GATEWAY
    156 #define	IPFORWARDING	1	/* forward IP packets not for us */
    157 #else /* GATEWAY */
    158 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
    159 #endif /* GATEWAY */
    160 #endif /* IPFORWARDING */
    161 #ifndef	IPSENDREDIRECTS
    162 #define	IPSENDREDIRECTS	1
    163 #endif
    164 #ifndef IPFORWSRCRT
    165 #define	IPFORWSRCRT	1	/* forward source-routed packets */
    166 #endif
    167 #ifndef IPALLOWSRCRT
    168 #define	IPALLOWSRCRT	1	/* allow source-routed packets */
    169 #endif
    170 #ifndef IPMTUDISC
    171 #define IPMTUDISC	1
    172 #endif
    173 #ifndef IPMTUDISCTIMEOUT
    174 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
    175 #endif
    176 
    177 /*
    178  * Note: DIRECTED_BROADCAST is handled this way so that previous
    179  * configuration using this option will Just Work.
    180  */
    181 #ifndef IPDIRECTEDBCAST
    182 #ifdef DIRECTED_BROADCAST
    183 #define IPDIRECTEDBCAST	1
    184 #else
    185 #define	IPDIRECTEDBCAST	0
    186 #endif /* DIRECTED_BROADCAST */
    187 #endif /* IPDIRECTEDBCAST */
    188 int	ipforwarding = IPFORWARDING;
    189 int	ipsendredirects = IPSENDREDIRECTS;
    190 int	ip_defttl = IPDEFTTL;
    191 int	ip_forwsrcrt = IPFORWSRCRT;
    192 int	ip_directedbcast = IPDIRECTEDBCAST;
    193 int	ip_allowsrcrt = IPALLOWSRCRT;
    194 int	ip_mtudisc = IPMTUDISC;
    195 int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
    196 #ifdef DIAGNOSTIC
    197 int	ipprintfs = 0;
    198 #endif
    199 
    200 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
    201 
    202 extern	struct domain inetdomain;
    203 int	ipqmaxlen = IFQ_MAXLEN;
    204 u_long	in_ifaddrhash;				/* size of hash table - 1 */
    205 int	in_ifaddrentries;			/* total number of addrs */
    206 struct	in_ifaddrhead in_ifaddr;
    207 struct	in_ifaddrhashhead *in_ifaddrhashtbl;
    208 struct	ifqueue ipintrq;
    209 struct	ipstat	ipstat;
    210 u_int16_t	ip_id;
    211 
    212 #ifdef PFIL_HOOKS
    213 struct pfil_head inet_pfil_hook;
    214 #endif
    215 
    216 struct ipqhead ipq;
    217 int	ipq_locked;
    218 int	ip_nfragpackets = 0;
    219 int	ip_maxfragpackets = 200;
    220 
    221 static __inline int ipq_lock_try __P((void));
    222 static __inline void ipq_unlock __P((void));
    223 
    224 static __inline int
    225 ipq_lock_try()
    226 {
    227 	int s;
    228 
    229 	/*
    230 	 * Use splvm() -- we're blocking things that would cause
    231 	 * mbuf allocation.
    232 	 */
    233 	s = splvm();
    234 	if (ipq_locked) {
    235 		splx(s);
    236 		return (0);
    237 	}
    238 	ipq_locked = 1;
    239 	splx(s);
    240 	return (1);
    241 }
    242 
    243 static __inline void
    244 ipq_unlock()
    245 {
    246 	int s;
    247 
    248 	s = splvm();
    249 	ipq_locked = 0;
    250 	splx(s);
    251 }
    252 
    253 #ifdef DIAGNOSTIC
    254 #define	IPQ_LOCK()							\
    255 do {									\
    256 	if (ipq_lock_try() == 0) {					\
    257 		printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
    258 		panic("ipq_lock");					\
    259 	}								\
    260 } while (/*CONSTCOND*/ 0)
    261 #define	IPQ_LOCK_CHECK()						\
    262 do {									\
    263 	if (ipq_locked == 0) {						\
    264 		printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
    265 		panic("ipq lock check");				\
    266 	}								\
    267 } while (/*CONSTCOND*/ 0)
    268 #else
    269 #define	IPQ_LOCK()		(void) ipq_lock_try()
    270 #define	IPQ_LOCK_CHECK()	/* nothing */
    271 #endif
    272 
    273 #define	IPQ_UNLOCK()		ipq_unlock()
    274 
    275 struct pool ipqent_pool;
    276 
    277 #ifdef INET_CSUM_COUNTERS
    278 #include <sys/device.h>
    279 
    280 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    281     NULL, "inet", "hwcsum bad");
    282 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    283     NULL, "inet", "hwcsum ok");
    284 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    285     NULL, "inet", "swcsum");
    286 
    287 #define	INET_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    288 
    289 #else
    290 
    291 #define	INET_CSUM_COUNTER_INCR(ev)	/* nothing */
    292 
    293 #endif /* INET_CSUM_COUNTERS */
    294 
    295 /*
    296  * We need to save the IP options in case a protocol wants to respond
    297  * to an incoming packet over the same route if the packet got here
    298  * using IP source routing.  This allows connection establishment and
    299  * maintenance when the remote end is on a network that is not known
    300  * to us.
    301  */
    302 int	ip_nhops = 0;
    303 static	struct ip_srcrt {
    304 	struct	in_addr dst;			/* final destination */
    305 	char	nop;				/* one NOP to align */
    306 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
    307 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
    308 } ip_srcrt;
    309 
    310 static void save_rte __P((u_char *, struct in_addr));
    311 
    312 #ifdef MBUFTRACE
    313 struct mowner ip_rx_mowner = { "internet", "rx" };
    314 struct mowner ip_tx_mowner = { "internet", "tx" };
    315 #endif
    316 
    317 /*
    318  * IP initialization: fill in IP protocol switch table.
    319  * All protocols not implemented in kernel go to raw IP protocol handler.
    320  */
    321 void
    322 ip_init()
    323 {
    324 	struct protosw *pr;
    325 	int i;
    326 
    327 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
    328 	    NULL);
    329 
    330 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
    331 	if (pr == 0)
    332 		panic("ip_init");
    333 	for (i = 0; i < IPPROTO_MAX; i++)
    334 		ip_protox[i] = pr - inetsw;
    335 	for (pr = inetdomain.dom_protosw;
    336 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
    337 		if (pr->pr_domain->dom_family == PF_INET &&
    338 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
    339 			ip_protox[pr->pr_protocol] = pr - inetsw;
    340 	LIST_INIT(&ipq);
    341 	ip_id = time.tv_sec & 0xffff;
    342 	ipintrq.ifq_maxlen = ipqmaxlen;
    343 	TAILQ_INIT(&in_ifaddr);
    344 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
    345 	    M_WAITOK, &in_ifaddrhash);
    346 	ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
    347 #ifdef GATEWAY
    348 	ipflow_init();
    349 #endif
    350 
    351 #ifdef PFIL_HOOKS
    352 	/* Register our Packet Filter hook. */
    353 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
    354 	inet_pfil_hook.ph_af   = AF_INET;
    355 	i = pfil_head_register(&inet_pfil_hook);
    356 	if (i != 0)
    357 		printf("ip_init: WARNING: unable to register pfil hook, "
    358 		    "error %d\n", i);
    359 #endif /* PFIL_HOOKS */
    360 
    361 #ifdef INET_CSUM_COUNTERS
    362 	evcnt_attach_static(&ip_hwcsum_bad);
    363 	evcnt_attach_static(&ip_hwcsum_ok);
    364 	evcnt_attach_static(&ip_swcsum);
    365 #endif /* INET_CSUM_COUNTERS */
    366 
    367 #ifdef MBUFTRACE
    368 	MOWNER_ATTACH(&ip_tx_mowner);
    369 	MOWNER_ATTACH(&ip_rx_mowner);
    370 #endif /* MBUFTRACE */
    371 }
    372 
    373 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
    374 struct	route ipforward_rt;
    375 
    376 /*
    377  * IP software interrupt routine
    378  */
    379 void
    380 ipintr()
    381 {
    382 	int s;
    383 	struct mbuf *m;
    384 
    385 	while (1) {
    386 		s = splnet();
    387 		IF_DEQUEUE(&ipintrq, m);
    388 		splx(s);
    389 		if (m == 0)
    390 			return;
    391 		MCLAIM(m, &ip_rx_mowner);
    392 		ip_input(m);
    393 	}
    394 }
    395 
    396 /*
    397  * Ip input routine.  Checksum and byte swap header.  If fragmented
    398  * try to reassemble.  Process options.  Pass to next level.
    399  */
    400 void
    401 ip_input(struct mbuf *m)
    402 {
    403 	struct ip *ip = NULL;
    404 	struct ipq *fp;
    405 	struct in_ifaddr *ia;
    406 	struct ifaddr *ifa;
    407 	struct ipqent *ipqe;
    408 	int hlen = 0, mff, len;
    409 	int downmatch;
    410 
    411 	MCLAIM(m, &ip_rx_mowner);
    412 #ifdef	DIAGNOSTIC
    413 	if ((m->m_flags & M_PKTHDR) == 0)
    414 		panic("ipintr no HDR");
    415 #endif
    416 #ifdef IPSEC
    417 	/*
    418 	 * should the inner packet be considered authentic?
    419 	 * see comment in ah4_input().
    420 	 */
    421 	if (m) {
    422 		m->m_flags &= ~M_AUTHIPHDR;
    423 		m->m_flags &= ~M_AUTHIPDGM;
    424 	}
    425 #endif
    426 
    427 	/*
    428 	 * If no IP addresses have been set yet but the interfaces
    429 	 * are receiving, can't do anything with incoming packets yet.
    430 	 */
    431 	if (TAILQ_FIRST(&in_ifaddr) == 0)
    432 		goto bad;
    433 	ipstat.ips_total++;
    434 	/*
    435 	 * If the IP header is not aligned, slurp it up into a new
    436 	 * mbuf with space for link headers, in the event we forward
    437 	 * it.  Otherwise, if it is aligned, make sure the entire
    438 	 * base IP header is in the first mbuf of the chain.
    439 	 */
    440 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
    441 		if ((m = m_copyup(m, sizeof(struct ip),
    442 				  (max_linkhdr + 3) & ~3)) == NULL) {
    443 			/* XXXJRT new stat, please */
    444 			ipstat.ips_toosmall++;
    445 			return;
    446 		}
    447 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
    448 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
    449 			ipstat.ips_toosmall++;
    450 			return;
    451 		}
    452 	}
    453 	ip = mtod(m, struct ip *);
    454 	if (ip->ip_v != IPVERSION) {
    455 		ipstat.ips_badvers++;
    456 		goto bad;
    457 	}
    458 	hlen = ip->ip_hl << 2;
    459 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
    460 		ipstat.ips_badhlen++;
    461 		goto bad;
    462 	}
    463 	if (hlen > m->m_len) {
    464 		if ((m = m_pullup(m, hlen)) == 0) {
    465 			ipstat.ips_badhlen++;
    466 			return;
    467 		}
    468 		ip = mtod(m, struct ip *);
    469 	}
    470 
    471 	/*
    472 	 * RFC1122: packets with a multicast source address are
    473 	 * not allowed.
    474 	 */
    475 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    476 		ipstat.ips_badaddr++;
    477 		goto bad;
    478 	}
    479 
    480 	/* 127/8 must not appear on wire - RFC1122 */
    481 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
    482 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
    483 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
    484 			ipstat.ips_badaddr++;
    485 			goto bad;
    486 		}
    487 	}
    488 
    489 	switch (m->m_pkthdr.csum_flags &
    490 		((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
    491 		 M_CSUM_IPv4_BAD)) {
    492 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
    493 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
    494 		goto badcsum;
    495 
    496 	case M_CSUM_IPv4:
    497 		/* Checksum was okay. */
    498 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
    499 		break;
    500 
    501 	default:
    502 		/* Must compute it ourselves. */
    503 		INET_CSUM_COUNTER_INCR(&ip_swcsum);
    504 		if (in_cksum(m, hlen) != 0)
    505 			goto bad;
    506 		break;
    507 	}
    508 
    509 	/* Retrieve the packet length. */
    510 	len = ntohs(ip->ip_len);
    511 
    512 	/*
    513 	 * Check for additional length bogosity
    514 	 */
    515 	if (len < hlen) {
    516 	 	ipstat.ips_badlen++;
    517 		goto bad;
    518 	}
    519 
    520 	/*
    521 	 * Check that the amount of data in the buffers
    522 	 * is as at least much as the IP header would have us expect.
    523 	 * Trim mbufs if longer than we expect.
    524 	 * Drop packet if shorter than we expect.
    525 	 */
    526 	if (m->m_pkthdr.len < len) {
    527 		ipstat.ips_tooshort++;
    528 		goto bad;
    529 	}
    530 	if (m->m_pkthdr.len > len) {
    531 		if (m->m_len == m->m_pkthdr.len) {
    532 			m->m_len = len;
    533 			m->m_pkthdr.len = len;
    534 		} else
    535 			m_adj(m, len - m->m_pkthdr.len);
    536 	}
    537 
    538 #ifdef IPSEC
    539 	/* ipflow (IP fast forwarding) is not compatible with IPsec. */
    540 	m->m_flags &= ~M_CANFASTFWD;
    541 #else
    542 	/*
    543 	 * Assume that we can create a fast-forward IP flow entry
    544 	 * based on this packet.
    545 	 */
    546 	m->m_flags |= M_CANFASTFWD;
    547 #endif
    548 
    549 #ifdef PFIL_HOOKS
    550 	/*
    551 	 * Run through list of hooks for input packets.  If there are any
    552 	 * filters which require that additional packets in the flow are
    553 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
    554 	 * Note that filters must _never_ set this flag, as another filter
    555 	 * in the list may have previously cleared it.
    556 	 */
    557 	/*
    558 	 * let ipfilter look at packet on the wire,
    559 	 * not the decapsulated packet.
    560 	 */
    561 #ifdef IPSEC
    562 	if (!ipsec_getnhist(m))
    563 #else
    564 	if (1)
    565 #endif
    566 	{
    567 		if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
    568 				   PFIL_IN) != 0)
    569 		return;
    570 		if (m == NULL)
    571 			return;
    572 		ip = mtod(m, struct ip *);
    573 		hlen = ip->ip_hl << 2;
    574 	}
    575 #endif /* PFIL_HOOKS */
    576 
    577 #ifdef ALTQ
    578 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
    579 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
    580 		/* packet dropped by traffic conditioner */
    581 		return;
    582 	}
    583 #endif
    584 
    585 	/*
    586 	 * Process options and, if not destined for us,
    587 	 * ship it on.  ip_dooptions returns 1 when an
    588 	 * error was detected (causing an icmp message
    589 	 * to be sent and the original packet to be freed).
    590 	 */
    591 	ip_nhops = 0;		/* for source routed packets */
    592 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
    593 		return;
    594 
    595 	/*
    596 	 * Check our list of addresses, to see if the packet is for us.
    597 	 *
    598 	 * Traditional 4.4BSD did not consult IFF_UP at all.
    599 	 * The behavior here is to treat addresses on !IFF_UP interface
    600 	 * as not mine.
    601 	 */
    602 	downmatch = 0;
    603 	LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
    604 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
    605 			if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
    606 				break;
    607 			else
    608 				downmatch++;
    609 		}
    610 	}
    611 	if (ia != NULL)
    612 		goto ours;
    613 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
    614 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
    615 			if (ifa->ifa_addr->sa_family != AF_INET)
    616 				continue;
    617 			ia = ifatoia(ifa);
    618 			if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
    619 			    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
    620 			    /*
    621 			     * Look for all-0's host part (old broadcast addr),
    622 			     * either for subnet or net.
    623 			     */
    624 			    ip->ip_dst.s_addr == ia->ia_subnet ||
    625 			    ip->ip_dst.s_addr == ia->ia_net)
    626 				goto ours;
    627 			/*
    628 			 * An interface with IP address zero accepts
    629 			 * all packets that arrive on that interface.
    630 			 */
    631 			if (in_nullhost(ia->ia_addr.sin_addr))
    632 				goto ours;
    633 		}
    634 	}
    635 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
    636 		struct in_multi *inm;
    637 #ifdef MROUTING
    638 		extern struct socket *ip_mrouter;
    639 
    640 		if (M_READONLY(m)) {
    641 			if ((m = m_pullup(m, hlen)) == 0) {
    642 				ipstat.ips_toosmall++;
    643 				return;
    644 			}
    645 			ip = mtod(m, struct ip *);
    646 		}
    647 
    648 		if (ip_mrouter) {
    649 			/*
    650 			 * If we are acting as a multicast router, all
    651 			 * incoming multicast packets are passed to the
    652 			 * kernel-level multicast forwarding function.
    653 			 * The packet is returned (relatively) intact; if
    654 			 * ip_mforward() returns a non-zero value, the packet
    655 			 * must be discarded, else it may be accepted below.
    656 			 *
    657 			 * (The IP ident field is put in the same byte order
    658 			 * as expected when ip_mforward() is called from
    659 			 * ip_output().)
    660 			 */
    661 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
    662 				ipstat.ips_cantforward++;
    663 				m_freem(m);
    664 				return;
    665 			}
    666 
    667 			/*
    668 			 * The process-level routing demon needs to receive
    669 			 * all multicast IGMP packets, whether or not this
    670 			 * host belongs to their destination groups.
    671 			 */
    672 			if (ip->ip_p == IPPROTO_IGMP)
    673 				goto ours;
    674 			ipstat.ips_forward++;
    675 		}
    676 #endif
    677 		/*
    678 		 * See if we belong to the destination multicast group on the
    679 		 * arrival interface.
    680 		 */
    681 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
    682 		if (inm == NULL) {
    683 			ipstat.ips_cantforward++;
    684 			m_freem(m);
    685 			return;
    686 		}
    687 		goto ours;
    688 	}
    689 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    690 	    in_nullhost(ip->ip_dst))
    691 		goto ours;
    692 
    693 	/*
    694 	 * Not for us; forward if possible and desirable.
    695 	 */
    696 	if (ipforwarding == 0) {
    697 		ipstat.ips_cantforward++;
    698 		m_freem(m);
    699 	} else {
    700 		/*
    701 		 * If ip_dst matched any of my address on !IFF_UP interface,
    702 		 * and there's no IFF_UP interface that matches ip_dst,
    703 		 * send icmp unreach.  Forwarding it will result in in-kernel
    704 		 * forwarding loop till TTL goes to 0.
    705 		 */
    706 		if (downmatch) {
    707 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
    708 			ipstat.ips_cantforward++;
    709 			return;
    710 		}
    711 #ifdef IPSEC
    712 		if (ipsec4_in_reject(m, NULL)) {
    713 			ipsecstat.in_polvio++;
    714 			goto bad;
    715 		}
    716 #endif
    717 
    718 		ip_forward(m, 0);
    719 	}
    720 	return;
    721 
    722 ours:
    723 	/*
    724 	 * If offset or IP_MF are set, must reassemble.
    725 	 * Otherwise, nothing need be done.
    726 	 * (We could look in the reassembly queue to see
    727 	 * if the packet was previously fragmented,
    728 	 * but it's not worth the time; just let them time out.)
    729 	 */
    730 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
    731 		if (M_READONLY(m)) {
    732 			if ((m = m_pullup(m, hlen)) == NULL) {
    733 				ipstat.ips_toosmall++;
    734 				goto bad;
    735 			}
    736 			ip = mtod(m, struct ip *);
    737 		}
    738 
    739 		/*
    740 		 * Look for queue of fragments
    741 		 * of this datagram.
    742 		 */
    743 		IPQ_LOCK();
    744 		LIST_FOREACH(fp, &ipq, ipq_q)
    745 			if (ip->ip_id == fp->ipq_id &&
    746 			    in_hosteq(ip->ip_src, fp->ipq_src) &&
    747 			    in_hosteq(ip->ip_dst, fp->ipq_dst) &&
    748 			    ip->ip_p == fp->ipq_p)
    749 				goto found;
    750 		fp = 0;
    751 found:
    752 
    753 		/*
    754 		 * Adjust ip_len to not reflect header,
    755 		 * set ipqe_mff if more fragments are expected,
    756 		 * convert offset of this to bytes.
    757 		 */
    758 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
    759 		mff = (ip->ip_off & htons(IP_MF)) != 0;
    760 		if (mff) {
    761 		        /*
    762 		         * Make sure that fragments have a data length
    763 			 * that's a non-zero multiple of 8 bytes.
    764 		         */
    765 			if (ntohs(ip->ip_len) == 0 ||
    766 			    (ntohs(ip->ip_len) & 0x7) != 0) {
    767 				ipstat.ips_badfrags++;
    768 				IPQ_UNLOCK();
    769 				goto bad;
    770 			}
    771 		}
    772 		ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
    773 
    774 		/*
    775 		 * If datagram marked as having more fragments
    776 		 * or if this is not the first fragment,
    777 		 * attempt reassembly; if it succeeds, proceed.
    778 		 */
    779 		if (mff || ip->ip_off != htons(0)) {
    780 			ipstat.ips_fragments++;
    781 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
    782 			if (ipqe == NULL) {
    783 				ipstat.ips_rcvmemdrop++;
    784 				IPQ_UNLOCK();
    785 				goto bad;
    786 			}
    787 			ipqe->ipqe_mff = mff;
    788 			ipqe->ipqe_m = m;
    789 			ipqe->ipqe_ip = ip;
    790 			m = ip_reass(ipqe, fp);
    791 			if (m == 0) {
    792 				IPQ_UNLOCK();
    793 				return;
    794 			}
    795 			ipstat.ips_reassembled++;
    796 			ip = mtod(m, struct ip *);
    797 			hlen = ip->ip_hl << 2;
    798 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
    799 		} else
    800 			if (fp)
    801 				ip_freef(fp);
    802 		IPQ_UNLOCK();
    803 	}
    804 
    805 #ifdef IPSEC
    806 	/*
    807 	 * enforce IPsec policy checking if we are seeing last header.
    808 	 * note that we do not visit this with protocols with pcb layer
    809 	 * code - like udp/tcp/raw ip.
    810 	 */
    811 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
    812 	    ipsec4_in_reject(m, NULL)) {
    813 		ipsecstat.in_polvio++;
    814 		goto bad;
    815 	}
    816 #endif
    817 
    818 	/*
    819 	 * Switch out to protocol's input routine.
    820 	 */
    821 #if IFA_STATS
    822 	if (ia && ip)
    823 		ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
    824 #endif
    825 	ipstat.ips_delivered++;
    826     {
    827 	int off = hlen, nh = ip->ip_p;
    828 
    829 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
    830 	return;
    831     }
    832 bad:
    833 	m_freem(m);
    834 	return;
    835 
    836 badcsum:
    837 	ipstat.ips_badsum++;
    838 	m_freem(m);
    839 }
    840 
    841 /*
    842  * Take incoming datagram fragment and try to
    843  * reassemble it into whole datagram.  If a chain for
    844  * reassembly of this datagram already exists, then it
    845  * is given as fp; otherwise have to make a chain.
    846  */
    847 struct mbuf *
    848 ip_reass(ipqe, fp)
    849 	struct ipqent *ipqe;
    850 	struct ipq *fp;
    851 {
    852 	struct mbuf *m = ipqe->ipqe_m;
    853 	struct ipqent *nq, *p, *q;
    854 	struct ip *ip;
    855 	struct mbuf *t;
    856 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
    857 	int i, next;
    858 
    859 	IPQ_LOCK_CHECK();
    860 
    861 	/*
    862 	 * Presence of header sizes in mbufs
    863 	 * would confuse code below.
    864 	 */
    865 	m->m_data += hlen;
    866 	m->m_len -= hlen;
    867 
    868 	/*
    869 	 * If first fragment to arrive, create a reassembly queue.
    870 	 */
    871 	if (fp == 0) {
    872 		/*
    873 		 * Enforce upper bound on number of fragmented packets
    874 		 * for which we attempt reassembly;
    875 		 * If maxfrag is 0, never accept fragments.
    876 		 * If maxfrag is -1, accept all fragments without limitation.
    877 		 */
    878 		if (ip_maxfragpackets < 0)
    879 			;
    880 		else if (ip_nfragpackets >= ip_maxfragpackets)
    881 			goto dropfrag;
    882 		ip_nfragpackets++;
    883 		MALLOC(fp, struct ipq *, sizeof (struct ipq),
    884 		    M_FTABLE, M_NOWAIT);
    885 		if (fp == NULL)
    886 			goto dropfrag;
    887 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
    888 		fp->ipq_ttl = IPFRAGTTL;
    889 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
    890 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
    891 		TAILQ_INIT(&fp->ipq_fragq);
    892 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
    893 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
    894 		p = NULL;
    895 		goto insert;
    896 	}
    897 
    898 	/*
    899 	 * Find a segment which begins after this one does.
    900 	 */
    901 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
    902 	    p = q, q = TAILQ_NEXT(q, ipqe_q))
    903 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
    904 			break;
    905 
    906 	/*
    907 	 * If there is a preceding segment, it may provide some of
    908 	 * our data already.  If so, drop the data from the incoming
    909 	 * segment.  If it provides all of our data, drop us.
    910 	 */
    911 	if (p != NULL) {
    912 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
    913 		    ntohs(ipqe->ipqe_ip->ip_off);
    914 		if (i > 0) {
    915 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
    916 				goto dropfrag;
    917 			m_adj(ipqe->ipqe_m, i);
    918 			ipqe->ipqe_ip->ip_off =
    919 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
    920 			ipqe->ipqe_ip->ip_len =
    921 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
    922 		}
    923 	}
    924 
    925 	/*
    926 	 * While we overlap succeeding segments trim them or,
    927 	 * if they are completely covered, dequeue them.
    928 	 */
    929 	for (; q != NULL &&
    930 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
    931 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
    932 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
    933 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
    934 		if (i < ntohs(q->ipqe_ip->ip_len)) {
    935 			q->ipqe_ip->ip_len =
    936 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
    937 			q->ipqe_ip->ip_off =
    938 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
    939 			m_adj(q->ipqe_m, i);
    940 			break;
    941 		}
    942 		nq = TAILQ_NEXT(q, ipqe_q);
    943 		m_freem(q->ipqe_m);
    944 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
    945 		pool_put(&ipqent_pool, q);
    946 	}
    947 
    948 insert:
    949 	/*
    950 	 * Stick new segment in its place;
    951 	 * check for complete reassembly.
    952 	 */
    953 	if (p == NULL) {
    954 		TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
    955 	} else {
    956 		TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
    957 	}
    958 	next = 0;
    959 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
    960 	    p = q, q = TAILQ_NEXT(q, ipqe_q)) {
    961 		if (ntohs(q->ipqe_ip->ip_off) != next)
    962 			return (0);
    963 		next += ntohs(q->ipqe_ip->ip_len);
    964 	}
    965 	if (p->ipqe_mff)
    966 		return (0);
    967 
    968 	/*
    969 	 * Reassembly is complete.  Check for a bogus message size and
    970 	 * concatenate fragments.
    971 	 */
    972 	q = TAILQ_FIRST(&fp->ipq_fragq);
    973 	ip = q->ipqe_ip;
    974 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
    975 		ipstat.ips_toolong++;
    976 		ip_freef(fp);
    977 		return (0);
    978 	}
    979 	m = q->ipqe_m;
    980 	t = m->m_next;
    981 	m->m_next = 0;
    982 	m_cat(m, t);
    983 	nq = TAILQ_NEXT(q, ipqe_q);
    984 	pool_put(&ipqent_pool, q);
    985 	for (q = nq; q != NULL; q = nq) {
    986 		t = q->ipqe_m;
    987 		nq = TAILQ_NEXT(q, ipqe_q);
    988 		pool_put(&ipqent_pool, q);
    989 		m_cat(m, t);
    990 	}
    991 
    992 	/*
    993 	 * Create header for new ip packet by
    994 	 * modifying header of first packet;
    995 	 * dequeue and discard fragment reassembly header.
    996 	 * Make header visible.
    997 	 */
    998 	ip->ip_len = htons(next);
    999 	ip->ip_src = fp->ipq_src;
   1000 	ip->ip_dst = fp->ipq_dst;
   1001 	LIST_REMOVE(fp, ipq_q);
   1002 	FREE(fp, M_FTABLE);
   1003 	ip_nfragpackets--;
   1004 	m->m_len += (ip->ip_hl << 2);
   1005 	m->m_data -= (ip->ip_hl << 2);
   1006 	/* some debugging cruft by sklower, below, will go away soon */
   1007 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
   1008 		int plen = 0;
   1009 		for (t = m; t; t = t->m_next)
   1010 			plen += t->m_len;
   1011 		m->m_pkthdr.len = plen;
   1012 	}
   1013 	return (m);
   1014 
   1015 dropfrag:
   1016 	ipstat.ips_fragdropped++;
   1017 	m_freem(m);
   1018 	pool_put(&ipqent_pool, ipqe);
   1019 	return (0);
   1020 }
   1021 
   1022 /*
   1023  * Free a fragment reassembly header and all
   1024  * associated datagrams.
   1025  */
   1026 void
   1027 ip_freef(fp)
   1028 	struct ipq *fp;
   1029 {
   1030 	struct ipqent *q, *p;
   1031 
   1032 	IPQ_LOCK_CHECK();
   1033 
   1034 	for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
   1035 		p = TAILQ_NEXT(q, ipqe_q);
   1036 		m_freem(q->ipqe_m);
   1037 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
   1038 		pool_put(&ipqent_pool, q);
   1039 	}
   1040 	LIST_REMOVE(fp, ipq_q);
   1041 	FREE(fp, M_FTABLE);
   1042 	ip_nfragpackets--;
   1043 }
   1044 
   1045 /*
   1046  * IP timer processing;
   1047  * if a timer expires on a reassembly
   1048  * queue, discard it.
   1049  */
   1050 void
   1051 ip_slowtimo()
   1052 {
   1053 	struct ipq *fp, *nfp;
   1054 	int s = splsoftnet();
   1055 
   1056 	IPQ_LOCK();
   1057 	for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) {
   1058 		nfp = LIST_NEXT(fp, ipq_q);
   1059 		if (--fp->ipq_ttl == 0) {
   1060 			ipstat.ips_fragtimeout++;
   1061 			ip_freef(fp);
   1062 		}
   1063 	}
   1064 	/*
   1065 	 * If we are over the maximum number of fragments
   1066 	 * (due to the limit being lowered), drain off
   1067 	 * enough to get down to the new limit.
   1068 	 */
   1069 	if (ip_maxfragpackets < 0)
   1070 		;
   1071 	else {
   1072 		while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq))
   1073 			ip_freef(LIST_FIRST(&ipq));
   1074 	}
   1075 	IPQ_UNLOCK();
   1076 #ifdef GATEWAY
   1077 	ipflow_slowtimo();
   1078 #endif
   1079 	splx(s);
   1080 }
   1081 
   1082 /*
   1083  * Drain off all datagram fragments.
   1084  */
   1085 void
   1086 ip_drain()
   1087 {
   1088 
   1089 	/*
   1090 	 * We may be called from a device's interrupt context.  If
   1091 	 * the ipq is already busy, just bail out now.
   1092 	 */
   1093 	if (ipq_lock_try() == 0)
   1094 		return;
   1095 
   1096 	while (LIST_FIRST(&ipq) != NULL) {
   1097 		ipstat.ips_fragdropped++;
   1098 		ip_freef(LIST_FIRST(&ipq));
   1099 	}
   1100 
   1101 	IPQ_UNLOCK();
   1102 }
   1103 
   1104 /*
   1105  * Do option processing on a datagram,
   1106  * possibly discarding it if bad options are encountered,
   1107  * or forwarding it if source-routed.
   1108  * Returns 1 if packet has been forwarded/freed,
   1109  * 0 if the packet should be processed further.
   1110  */
   1111 int
   1112 ip_dooptions(m)
   1113 	struct mbuf *m;
   1114 {
   1115 	struct ip *ip = mtod(m, struct ip *);
   1116 	u_char *cp, *cp0;
   1117 	struct ip_timestamp *ipt;
   1118 	struct in_ifaddr *ia;
   1119 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
   1120 	struct in_addr dst;
   1121 	n_time ntime;
   1122 
   1123 	dst = ip->ip_dst;
   1124 	cp = (u_char *)(ip + 1);
   1125 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
   1126 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
   1127 		opt = cp[IPOPT_OPTVAL];
   1128 		if (opt == IPOPT_EOL)
   1129 			break;
   1130 		if (opt == IPOPT_NOP)
   1131 			optlen = 1;
   1132 		else {
   1133 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
   1134 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1135 				goto bad;
   1136 			}
   1137 			optlen = cp[IPOPT_OLEN];
   1138 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
   1139 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1140 				goto bad;
   1141 			}
   1142 		}
   1143 		switch (opt) {
   1144 
   1145 		default:
   1146 			break;
   1147 
   1148 		/*
   1149 		 * Source routing with record.
   1150 		 * Find interface with current destination address.
   1151 		 * If none on this machine then drop if strictly routed,
   1152 		 * or do nothing if loosely routed.
   1153 		 * Record interface address and bring up next address
   1154 		 * component.  If strictly routed make sure next
   1155 		 * address is on directly accessible net.
   1156 		 */
   1157 		case IPOPT_LSRR:
   1158 		case IPOPT_SSRR:
   1159 			if (ip_allowsrcrt == 0) {
   1160 				type = ICMP_UNREACH;
   1161 				code = ICMP_UNREACH_NET_PROHIB;
   1162 				goto bad;
   1163 			}
   1164 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1165 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1166 				goto bad;
   1167 			}
   1168 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1169 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1170 				goto bad;
   1171 			}
   1172 			ipaddr.sin_addr = ip->ip_dst;
   1173 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
   1174 			if (ia == 0) {
   1175 				if (opt == IPOPT_SSRR) {
   1176 					type = ICMP_UNREACH;
   1177 					code = ICMP_UNREACH_SRCFAIL;
   1178 					goto bad;
   1179 				}
   1180 				/*
   1181 				 * Loose routing, and not at next destination
   1182 				 * yet; nothing to do except forward.
   1183 				 */
   1184 				break;
   1185 			}
   1186 			off--;			/* 0 origin */
   1187 			if ((off + sizeof(struct in_addr)) > optlen) {
   1188 				/*
   1189 				 * End of source route.  Should be for us.
   1190 				 */
   1191 				save_rte(cp, ip->ip_src);
   1192 				break;
   1193 			}
   1194 			/*
   1195 			 * locate outgoing interface
   1196 			 */
   1197 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
   1198 			    sizeof(ipaddr.sin_addr));
   1199 			if (opt == IPOPT_SSRR)
   1200 				ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
   1201 			else
   1202 				ia = ip_rtaddr(ipaddr.sin_addr);
   1203 			if (ia == 0) {
   1204 				type = ICMP_UNREACH;
   1205 				code = ICMP_UNREACH_SRCFAIL;
   1206 				goto bad;
   1207 			}
   1208 			ip->ip_dst = ipaddr.sin_addr;
   1209 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
   1210 			    (caddr_t)(cp + off), sizeof(struct in_addr));
   1211 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1212 			/*
   1213 			 * Let ip_intr's mcast routing check handle mcast pkts
   1214 			 */
   1215 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
   1216 			break;
   1217 
   1218 		case IPOPT_RR:
   1219 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1220 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1221 				goto bad;
   1222 			}
   1223 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1224 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1225 				goto bad;
   1226 			}
   1227 			/*
   1228 			 * If no space remains, ignore.
   1229 			 */
   1230 			off--;			/* 0 origin */
   1231 			if ((off + sizeof(struct in_addr)) > optlen)
   1232 				break;
   1233 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
   1234 			    sizeof(ipaddr.sin_addr));
   1235 			/*
   1236 			 * locate outgoing interface; if we're the destination,
   1237 			 * use the incoming interface (should be same).
   1238 			 */
   1239 			if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
   1240 			    == NULL &&
   1241 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
   1242 				type = ICMP_UNREACH;
   1243 				code = ICMP_UNREACH_HOST;
   1244 				goto bad;
   1245 			}
   1246 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
   1247 			    (caddr_t)(cp + off), sizeof(struct in_addr));
   1248 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1249 			break;
   1250 
   1251 		case IPOPT_TS:
   1252 			code = cp - (u_char *)ip;
   1253 			ipt = (struct ip_timestamp *)cp;
   1254 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
   1255 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
   1256 				goto bad;
   1257 			}
   1258 			if (ipt->ipt_ptr < 5) {
   1259 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
   1260 				goto bad;
   1261 			}
   1262 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
   1263 				if (++ipt->ipt_oflw == 0) {
   1264 					code = (u_char *)&ipt->ipt_ptr -
   1265 					    (u_char *)ip;
   1266 					goto bad;
   1267 				}
   1268 				break;
   1269 			}
   1270 			cp0 = (cp + ipt->ipt_ptr - 1);
   1271 			switch (ipt->ipt_flg) {
   1272 
   1273 			case IPOPT_TS_TSONLY:
   1274 				break;
   1275 
   1276 			case IPOPT_TS_TSANDADDR:
   1277 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1278 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1279 					code = (u_char *)&ipt->ipt_ptr -
   1280 					    (u_char *)ip;
   1281 					goto bad;
   1282 				}
   1283 				ipaddr.sin_addr = dst;
   1284 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
   1285 				    m->m_pkthdr.rcvif));
   1286 				if (ia == 0)
   1287 					continue;
   1288 				bcopy(&ia->ia_addr.sin_addr,
   1289 				    cp0, sizeof(struct in_addr));
   1290 				ipt->ipt_ptr += sizeof(struct in_addr);
   1291 				break;
   1292 
   1293 			case IPOPT_TS_PRESPEC:
   1294 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1295 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1296 					code = (u_char *)&ipt->ipt_ptr -
   1297 					    (u_char *)ip;
   1298 					goto bad;
   1299 				}
   1300 				bcopy(cp0, &ipaddr.sin_addr,
   1301 				    sizeof(struct in_addr));
   1302 				if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
   1303 				    == NULL)
   1304 					continue;
   1305 				ipt->ipt_ptr += sizeof(struct in_addr);
   1306 				break;
   1307 
   1308 			default:
   1309 				/* XXX can't take &ipt->ipt_flg */
   1310 				code = (u_char *)&ipt->ipt_ptr -
   1311 				    (u_char *)ip + 1;
   1312 				goto bad;
   1313 			}
   1314 			ntime = iptime();
   1315 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
   1316 			bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
   1317 			    sizeof(n_time));
   1318 			ipt->ipt_ptr += sizeof(n_time);
   1319 		}
   1320 	}
   1321 	if (forward) {
   1322 		if (ip_forwsrcrt == 0) {
   1323 			type = ICMP_UNREACH;
   1324 			code = ICMP_UNREACH_SRCFAIL;
   1325 			goto bad;
   1326 		}
   1327 		ip_forward(m, 1);
   1328 		return (1);
   1329 	}
   1330 	return (0);
   1331 bad:
   1332 	icmp_error(m, type, code, 0, 0);
   1333 	ipstat.ips_badoptions++;
   1334 	return (1);
   1335 }
   1336 
   1337 /*
   1338  * Given address of next destination (final or next hop),
   1339  * return internet address info of interface to be used to get there.
   1340  */
   1341 struct in_ifaddr *
   1342 ip_rtaddr(dst)
   1343 	 struct in_addr dst;
   1344 {
   1345 	struct sockaddr_in *sin;
   1346 
   1347 	sin = satosin(&ipforward_rt.ro_dst);
   1348 
   1349 	if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
   1350 		if (ipforward_rt.ro_rt) {
   1351 			RTFREE(ipforward_rt.ro_rt);
   1352 			ipforward_rt.ro_rt = 0;
   1353 		}
   1354 		sin->sin_family = AF_INET;
   1355 		sin->sin_len = sizeof(*sin);
   1356 		sin->sin_addr = dst;
   1357 
   1358 		rtalloc(&ipforward_rt);
   1359 	}
   1360 	if (ipforward_rt.ro_rt == 0)
   1361 		return ((struct in_ifaddr *)0);
   1362 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
   1363 }
   1364 
   1365 /*
   1366  * Save incoming source route for use in replies,
   1367  * to be picked up later by ip_srcroute if the receiver is interested.
   1368  */
   1369 void
   1370 save_rte(option, dst)
   1371 	u_char *option;
   1372 	struct in_addr dst;
   1373 {
   1374 	unsigned olen;
   1375 
   1376 	olen = option[IPOPT_OLEN];
   1377 #ifdef DIAGNOSTIC
   1378 	if (ipprintfs)
   1379 		printf("save_rte: olen %d\n", olen);
   1380 #endif /* 0 */
   1381 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
   1382 		return;
   1383 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
   1384 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
   1385 	ip_srcrt.dst = dst;
   1386 }
   1387 
   1388 /*
   1389  * Retrieve incoming source route for use in replies,
   1390  * in the same form used by setsockopt.
   1391  * The first hop is placed before the options, will be removed later.
   1392  */
   1393 struct mbuf *
   1394 ip_srcroute()
   1395 {
   1396 	struct in_addr *p, *q;
   1397 	struct mbuf *m;
   1398 
   1399 	if (ip_nhops == 0)
   1400 		return ((struct mbuf *)0);
   1401 	m = m_get(M_DONTWAIT, MT_SOOPTS);
   1402 	if (m == 0)
   1403 		return ((struct mbuf *)0);
   1404 
   1405 	MCLAIM(m, &inetdomain.dom_mowner);
   1406 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
   1407 
   1408 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
   1409 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
   1410 	    OPTSIZ;
   1411 #ifdef DIAGNOSTIC
   1412 	if (ipprintfs)
   1413 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
   1414 #endif
   1415 
   1416 	/*
   1417 	 * First save first hop for return route
   1418 	 */
   1419 	p = &ip_srcrt.route[ip_nhops - 1];
   1420 	*(mtod(m, struct in_addr *)) = *p--;
   1421 #ifdef DIAGNOSTIC
   1422 	if (ipprintfs)
   1423 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
   1424 #endif
   1425 
   1426 	/*
   1427 	 * Copy option fields and padding (nop) to mbuf.
   1428 	 */
   1429 	ip_srcrt.nop = IPOPT_NOP;
   1430 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
   1431 	bcopy((caddr_t)&ip_srcrt.nop,
   1432 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
   1433 	q = (struct in_addr *)(mtod(m, caddr_t) +
   1434 	    sizeof(struct in_addr) + OPTSIZ);
   1435 #undef OPTSIZ
   1436 	/*
   1437 	 * Record return path as an IP source route,
   1438 	 * reversing the path (pointers are now aligned).
   1439 	 */
   1440 	while (p >= ip_srcrt.route) {
   1441 #ifdef DIAGNOSTIC
   1442 		if (ipprintfs)
   1443 			printf(" %x", ntohl(q->s_addr));
   1444 #endif
   1445 		*q++ = *p--;
   1446 	}
   1447 	/*
   1448 	 * Last hop goes to final destination.
   1449 	 */
   1450 	*q = ip_srcrt.dst;
   1451 #ifdef DIAGNOSTIC
   1452 	if (ipprintfs)
   1453 		printf(" %x\n", ntohl(q->s_addr));
   1454 #endif
   1455 	return (m);
   1456 }
   1457 
   1458 /*
   1459  * Strip out IP options, at higher
   1460  * level protocol in the kernel.
   1461  * Second argument is buffer to which options
   1462  * will be moved, and return value is their length.
   1463  * XXX should be deleted; last arg currently ignored.
   1464  */
   1465 void
   1466 ip_stripoptions(m, mopt)
   1467 	struct mbuf *m;
   1468 	struct mbuf *mopt;
   1469 {
   1470 	int i;
   1471 	struct ip *ip = mtod(m, struct ip *);
   1472 	caddr_t opts;
   1473 	int olen;
   1474 
   1475 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
   1476 	opts = (caddr_t)(ip + 1);
   1477 	i = m->m_len - (sizeof (struct ip) + olen);
   1478 	bcopy(opts  + olen, opts, (unsigned)i);
   1479 	m->m_len -= olen;
   1480 	if (m->m_flags & M_PKTHDR)
   1481 		m->m_pkthdr.len -= olen;
   1482 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
   1483 	ip->ip_hl = sizeof (struct ip) >> 2;
   1484 }
   1485 
   1486 const int inetctlerrmap[PRC_NCMDS] = {
   1487 	0,		0,		0,		0,
   1488 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
   1489 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
   1490 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
   1491 	0,		0,		0,		0,
   1492 	ENOPROTOOPT
   1493 };
   1494 
   1495 /*
   1496  * Forward a packet.  If some error occurs return the sender
   1497  * an icmp packet.  Note we can't always generate a meaningful
   1498  * icmp message because icmp doesn't have a large enough repertoire
   1499  * of codes and types.
   1500  *
   1501  * If not forwarding, just drop the packet.  This could be confusing
   1502  * if ipforwarding was zero but some routing protocol was advancing
   1503  * us as a gateway to somewhere.  However, we must let the routing
   1504  * protocol deal with that.
   1505  *
   1506  * The srcrt parameter indicates whether the packet is being forwarded
   1507  * via a source route.
   1508  */
   1509 void
   1510 ip_forward(m, srcrt)
   1511 	struct mbuf *m;
   1512 	int srcrt;
   1513 {
   1514 	struct ip *ip = mtod(m, struct ip *);
   1515 	struct sockaddr_in *sin;
   1516 	struct rtentry *rt;
   1517 	int error, type = 0, code = 0;
   1518 	struct mbuf *mcopy;
   1519 	n_long dest;
   1520 	struct ifnet *destifp;
   1521 #ifdef IPSEC
   1522 	struct ifnet dummyifp;
   1523 #endif
   1524 
   1525 	/*
   1526 	 * We are now in the output path.
   1527 	 */
   1528 	MCLAIM(m, &ip_tx_mowner);
   1529 
   1530 	/*
   1531 	 * Clear any in-bound checksum flags for this packet.
   1532 	 */
   1533 	m->m_pkthdr.csum_flags = 0;
   1534 
   1535 	dest = 0;
   1536 #ifdef DIAGNOSTIC
   1537 	if (ipprintfs)
   1538 		printf("forward: src %2.2x dst %2.2x ttl %x\n",
   1539 		    ntohl(ip->ip_src.s_addr),
   1540 		    ntohl(ip->ip_dst.s_addr), ip->ip_ttl);
   1541 #endif
   1542 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
   1543 		ipstat.ips_cantforward++;
   1544 		m_freem(m);
   1545 		return;
   1546 	}
   1547 	if (ip->ip_ttl <= IPTTLDEC) {
   1548 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
   1549 		return;
   1550 	}
   1551 	ip->ip_ttl -= IPTTLDEC;
   1552 
   1553 	sin = satosin(&ipforward_rt.ro_dst);
   1554 	if ((rt = ipforward_rt.ro_rt) == 0 ||
   1555 	    !in_hosteq(ip->ip_dst, sin->sin_addr)) {
   1556 		if (ipforward_rt.ro_rt) {
   1557 			RTFREE(ipforward_rt.ro_rt);
   1558 			ipforward_rt.ro_rt = 0;
   1559 		}
   1560 		sin->sin_family = AF_INET;
   1561 		sin->sin_len = sizeof(struct sockaddr_in);
   1562 		sin->sin_addr = ip->ip_dst;
   1563 
   1564 		rtalloc(&ipforward_rt);
   1565 		if (ipforward_rt.ro_rt == 0) {
   1566 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
   1567 			return;
   1568 		}
   1569 		rt = ipforward_rt.ro_rt;
   1570 	}
   1571 
   1572 	/*
   1573 	 * Save at most 68 bytes of the packet in case
   1574 	 * we need to generate an ICMP message to the src.
   1575 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
   1576 	 */
   1577 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
   1578 	if (mcopy)
   1579 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
   1580 
   1581 	/*
   1582 	 * If forwarding packet using same interface that it came in on,
   1583 	 * perhaps should send a redirect to sender to shortcut a hop.
   1584 	 * Only send redirect if source is sending directly to us,
   1585 	 * and if packet was not source routed (or has any options).
   1586 	 * Also, don't send redirect if forwarding using a default route
   1587 	 * or a route modified by a redirect.
   1588 	 */
   1589 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
   1590 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
   1591 	    !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
   1592 	    ipsendredirects && !srcrt) {
   1593 		if (rt->rt_ifa &&
   1594 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
   1595 		    ifatoia(rt->rt_ifa)->ia_subnet) {
   1596 			if (rt->rt_flags & RTF_GATEWAY)
   1597 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
   1598 			else
   1599 				dest = ip->ip_dst.s_addr;
   1600 			/*
   1601 			 * Router requirements says to only send host
   1602 			 * redirects.
   1603 			 */
   1604 			type = ICMP_REDIRECT;
   1605 			code = ICMP_REDIRECT_HOST;
   1606 #ifdef DIAGNOSTIC
   1607 			if (ipprintfs)
   1608 				printf("redirect (%d) to %x\n", code,
   1609 				    (u_int32_t)dest);
   1610 #endif
   1611 		}
   1612 	}
   1613 
   1614 #ifdef IPSEC
   1615 	/* Don't lookup socket in forwarding case */
   1616 	(void)ipsec_setsocket(m, NULL);
   1617 #endif
   1618 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
   1619 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0);
   1620 	if (error)
   1621 		ipstat.ips_cantforward++;
   1622 	else {
   1623 		ipstat.ips_forward++;
   1624 		if (type)
   1625 			ipstat.ips_redirectsent++;
   1626 		else {
   1627 			if (mcopy) {
   1628 #ifdef GATEWAY
   1629 				if (mcopy->m_flags & M_CANFASTFWD)
   1630 					ipflow_create(&ipforward_rt, mcopy);
   1631 #endif
   1632 				m_freem(mcopy);
   1633 			}
   1634 			return;
   1635 		}
   1636 	}
   1637 	if (mcopy == NULL)
   1638 		return;
   1639 	destifp = NULL;
   1640 
   1641 	switch (error) {
   1642 
   1643 	case 0:				/* forwarded, but need redirect */
   1644 		/* type, code set above */
   1645 		break;
   1646 
   1647 	case ENETUNREACH:		/* shouldn't happen, checked above */
   1648 	case EHOSTUNREACH:
   1649 	case ENETDOWN:
   1650 	case EHOSTDOWN:
   1651 	default:
   1652 		type = ICMP_UNREACH;
   1653 		code = ICMP_UNREACH_HOST;
   1654 		break;
   1655 
   1656 	case EMSGSIZE:
   1657 		type = ICMP_UNREACH;
   1658 		code = ICMP_UNREACH_NEEDFRAG;
   1659 #ifndef IPSEC
   1660 		if (ipforward_rt.ro_rt)
   1661 			destifp = ipforward_rt.ro_rt->rt_ifp;
   1662 #else
   1663 		/*
   1664 		 * If the packet is routed over IPsec tunnel, tell the
   1665 		 * originator the tunnel MTU.
   1666 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
   1667 		 * XXX quickhack!!!
   1668 		 */
   1669 		if (ipforward_rt.ro_rt) {
   1670 			struct secpolicy *sp;
   1671 			int ipsecerror;
   1672 			size_t ipsechdr;
   1673 			struct route *ro;
   1674 
   1675 			sp = ipsec4_getpolicybyaddr(mcopy,
   1676 			                            IPSEC_DIR_OUTBOUND,
   1677 			                            IP_FORWARDING,
   1678 			                            &ipsecerror);
   1679 
   1680 			if (sp == NULL)
   1681 				destifp = ipforward_rt.ro_rt->rt_ifp;
   1682 			else {
   1683 				/* count IPsec header size */
   1684 				ipsechdr = ipsec4_hdrsiz(mcopy,
   1685 				                         IPSEC_DIR_OUTBOUND,
   1686 				                         NULL);
   1687 
   1688 				/*
   1689 				 * find the correct route for outer IPv4
   1690 				 * header, compute tunnel MTU.
   1691 				 *
   1692 				 * XXX BUG ALERT
   1693 				 * The "dummyifp" code relies upon the fact
   1694 				 * that icmp_error() touches only ifp->if_mtu.
   1695 				 */
   1696 				/*XXX*/
   1697 				destifp = NULL;
   1698 				if (sp->req != NULL
   1699 				 && sp->req->sav != NULL
   1700 				 && sp->req->sav->sah != NULL) {
   1701 					ro = &sp->req->sav->sah->sa_route;
   1702 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
   1703 						dummyifp.if_mtu =
   1704 						    ro->ro_rt->rt_rmx.rmx_mtu ?
   1705 						    ro->ro_rt->rt_rmx.rmx_mtu :
   1706 						    ro->ro_rt->rt_ifp->if_mtu;
   1707 						dummyifp.if_mtu -= ipsechdr;
   1708 						destifp = &dummyifp;
   1709 					}
   1710 				}
   1711 
   1712 				key_freesp(sp);
   1713 			}
   1714 		}
   1715 #endif /*IPSEC*/
   1716 		ipstat.ips_cantfrag++;
   1717 		break;
   1718 
   1719 	case ENOBUFS:
   1720 #if 1
   1721 		/*
   1722 		 * a router should not generate ICMP_SOURCEQUENCH as
   1723 		 * required in RFC1812 Requirements for IP Version 4 Routers.
   1724 		 * source quench could be a big problem under DoS attacks,
   1725 		 * or if the underlying interface is rate-limited.
   1726 		 */
   1727 		if (mcopy)
   1728 			m_freem(mcopy);
   1729 		return;
   1730 #else
   1731 		type = ICMP_SOURCEQUENCH;
   1732 		code = 0;
   1733 		break;
   1734 #endif
   1735 	}
   1736 	icmp_error(mcopy, type, code, dest, destifp);
   1737 }
   1738 
   1739 void
   1740 ip_savecontrol(inp, mp, ip, m)
   1741 	struct inpcb *inp;
   1742 	struct mbuf **mp;
   1743 	struct ip *ip;
   1744 	struct mbuf *m;
   1745 {
   1746 
   1747 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
   1748 		struct timeval tv;
   1749 
   1750 		microtime(&tv);
   1751 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
   1752 		    SCM_TIMESTAMP, SOL_SOCKET);
   1753 		if (*mp)
   1754 			mp = &(*mp)->m_next;
   1755 	}
   1756 	if (inp->inp_flags & INP_RECVDSTADDR) {
   1757 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
   1758 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
   1759 		if (*mp)
   1760 			mp = &(*mp)->m_next;
   1761 	}
   1762 #ifdef notyet
   1763 	/*
   1764 	 * XXX
   1765 	 * Moving these out of udp_input() made them even more broken
   1766 	 * than they already were.
   1767 	 *	- fenner (at) parc.xerox.com
   1768 	 */
   1769 	/* options were tossed already */
   1770 	if (inp->inp_flags & INP_RECVOPTS) {
   1771 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
   1772 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
   1773 		if (*mp)
   1774 			mp = &(*mp)->m_next;
   1775 	}
   1776 	/* ip_srcroute doesn't do what we want here, need to fix */
   1777 	if (inp->inp_flags & INP_RECVRETOPTS) {
   1778 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
   1779 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
   1780 		if (*mp)
   1781 			mp = &(*mp)->m_next;
   1782 	}
   1783 #endif
   1784 	if (inp->inp_flags & INP_RECVIF) {
   1785 		struct sockaddr_dl sdl;
   1786 
   1787 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
   1788 		sdl.sdl_family = AF_LINK;
   1789 		sdl.sdl_index = m->m_pkthdr.rcvif ?
   1790 		    m->m_pkthdr.rcvif->if_index : 0;
   1791 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
   1792 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
   1793 		    IP_RECVIF, IPPROTO_IP);
   1794 		if (*mp)
   1795 			mp = &(*mp)->m_next;
   1796 	}
   1797 }
   1798 
   1799 int
   1800 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
   1801 	int *name;
   1802 	u_int namelen;
   1803 	void *oldp;
   1804 	size_t *oldlenp;
   1805 	void *newp;
   1806 	size_t newlen;
   1807 {
   1808 	extern int subnetsarelocal, hostzeroisbroadcast;
   1809 
   1810 	int error, old;
   1811 
   1812 	/* All sysctl names at this level are terminal. */
   1813 	if (namelen != 1)
   1814 		return (ENOTDIR);
   1815 
   1816 	switch (name[0]) {
   1817 	case IPCTL_FORWARDING:
   1818 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
   1819 	case IPCTL_SENDREDIRECTS:
   1820 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1821 			&ipsendredirects));
   1822 	case IPCTL_DEFTTL:
   1823 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
   1824 #ifdef notyet
   1825 	case IPCTL_DEFMTU:
   1826 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
   1827 #endif
   1828 	case IPCTL_FORWSRCRT:
   1829 		/* Don't allow this to change in a secure environment.  */
   1830 		if (securelevel > 0)
   1831 			return (sysctl_rdint(oldp, oldlenp, newp,
   1832 			    ip_forwsrcrt));
   1833 		else
   1834 			return (sysctl_int(oldp, oldlenp, newp, newlen,
   1835 			    &ip_forwsrcrt));
   1836 	case IPCTL_DIRECTEDBCAST:
   1837 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1838 		    &ip_directedbcast));
   1839 	case IPCTL_ALLOWSRCRT:
   1840 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1841 		    &ip_allowsrcrt));
   1842 	case IPCTL_SUBNETSARELOCAL:
   1843 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1844 		    &subnetsarelocal));
   1845 	case IPCTL_MTUDISC:
   1846 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1847 		    &ip_mtudisc);
   1848 		if (error == 0 && ip_mtudisc == 0)
   1849 			rt_timer_queue_remove_all(ip_mtudisc_timeout_q, TRUE);
   1850 		return error;
   1851 	case IPCTL_ANONPORTMIN:
   1852 		old = anonportmin;
   1853 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin);
   1854 		if (anonportmin >= anonportmax || anonportmin < 0
   1855 		    || anonportmin > 65535
   1856 #ifndef IPNOPRIVPORTS
   1857 		    || anonportmin < IPPORT_RESERVED
   1858 #endif
   1859 		    ) {
   1860 			anonportmin = old;
   1861 			return (EINVAL);
   1862 		}
   1863 		return (error);
   1864 	case IPCTL_ANONPORTMAX:
   1865 		old = anonportmax;
   1866 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax);
   1867 		if (anonportmin >= anonportmax || anonportmax < 0
   1868 		    || anonportmax > 65535
   1869 #ifndef IPNOPRIVPORTS
   1870 		    || anonportmax < IPPORT_RESERVED
   1871 #endif
   1872 		    ) {
   1873 			anonportmax = old;
   1874 			return (EINVAL);
   1875 		}
   1876 		return (error);
   1877 	case IPCTL_MTUDISCTIMEOUT:
   1878 		old = ip_mtudisc_timeout;
   1879 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1880 		   &ip_mtudisc_timeout);
   1881 		if (ip_mtudisc_timeout < 0) {
   1882 			ip_mtudisc_timeout = old;
   1883 			return (EINVAL);
   1884 		}
   1885 		if (error == 0)
   1886 			rt_timer_queue_change(ip_mtudisc_timeout_q,
   1887 					      ip_mtudisc_timeout);
   1888 		return (error);
   1889 #ifdef GATEWAY
   1890 	case IPCTL_MAXFLOWS:
   1891 	    {
   1892 		int s;
   1893 
   1894 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1895 		   &ip_maxflows);
   1896 		s = splsoftnet();
   1897 		ipflow_reap(0);
   1898 		splx(s);
   1899 		return (error);
   1900 	    }
   1901 #endif
   1902 	case IPCTL_HOSTZEROBROADCAST:
   1903 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1904 		    &hostzeroisbroadcast));
   1905 #if NGIF > 0
   1906 	case IPCTL_GIF_TTL:
   1907 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1908 				  &ip_gif_ttl));
   1909 #endif
   1910 
   1911 #if NGRE > 0
   1912 	case IPCTL_GRE_TTL:
   1913 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1914 				  &ip_gre_ttl));
   1915 #endif
   1916 
   1917 #ifndef IPNOPRIVPORTS
   1918 	case IPCTL_LOWPORTMIN:
   1919 		old = lowportmin;
   1920 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin);
   1921 		if (lowportmin >= lowportmax
   1922 		    || lowportmin > IPPORT_RESERVEDMAX
   1923 		    || lowportmin < IPPORT_RESERVEDMIN
   1924 		    ) {
   1925 			lowportmin = old;
   1926 			return (EINVAL);
   1927 		}
   1928 		return (error);
   1929 	case IPCTL_LOWPORTMAX:
   1930 		old = lowportmax;
   1931 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax);
   1932 		if (lowportmin >= lowportmax
   1933 		    || lowportmax > IPPORT_RESERVEDMAX
   1934 		    || lowportmax < IPPORT_RESERVEDMIN
   1935 		    ) {
   1936 			lowportmax = old;
   1937 			return (EINVAL);
   1938 		}
   1939 		return (error);
   1940 #endif
   1941 
   1942 	case IPCTL_MAXFRAGPACKETS:
   1943 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1944 		    &ip_maxfragpackets));
   1945 
   1946 	default:
   1947 		return (EOPNOTSUPP);
   1948 	}
   1949 	/* NOTREACHED */
   1950 }
   1951