Home | History | Annotate | Line # | Download | only in netinet
ip_input.c revision 1.163
      1 /*	$NetBSD: ip_input.c,v 1.163 2002/11/12 02:10:13 itojun 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 1.163 2002/11/12 02:10:13 itojun 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 /*
    313  * IP initialization: fill in IP protocol switch table.
    314  * All protocols not implemented in kernel go to raw IP protocol handler.
    315  */
    316 void
    317 ip_init()
    318 {
    319 	struct protosw *pr;
    320 	int i;
    321 
    322 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
    323 	    NULL);
    324 
    325 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
    326 	if (pr == 0)
    327 		panic("ip_init");
    328 	for (i = 0; i < IPPROTO_MAX; i++)
    329 		ip_protox[i] = pr - inetsw;
    330 	for (pr = inetdomain.dom_protosw;
    331 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
    332 		if (pr->pr_domain->dom_family == PF_INET &&
    333 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
    334 			ip_protox[pr->pr_protocol] = pr - inetsw;
    335 	LIST_INIT(&ipq);
    336 	ip_id = time.tv_sec & 0xffff;
    337 	ipintrq.ifq_maxlen = ipqmaxlen;
    338 	TAILQ_INIT(&in_ifaddr);
    339 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
    340 	    M_WAITOK, &in_ifaddrhash);
    341 	ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
    342 #ifdef GATEWAY
    343 	ipflow_init();
    344 #endif
    345 
    346 #ifdef PFIL_HOOKS
    347 	/* Register our Packet Filter hook. */
    348 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
    349 	inet_pfil_hook.ph_af   = AF_INET;
    350 	i = pfil_head_register(&inet_pfil_hook);
    351 	if (i != 0)
    352 		printf("ip_init: WARNING: unable to register pfil hook, "
    353 		    "error %d\n", i);
    354 #endif /* PFIL_HOOKS */
    355 
    356 #ifdef INET_CSUM_COUNTERS
    357 	evcnt_attach_static(&ip_hwcsum_bad);
    358 	evcnt_attach_static(&ip_hwcsum_ok);
    359 	evcnt_attach_static(&ip_swcsum);
    360 #endif /* INET_CSUM_COUNTERS */
    361 }
    362 
    363 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
    364 struct	route ipforward_rt;
    365 
    366 /*
    367  * IP software interrupt routine
    368  */
    369 void
    370 ipintr()
    371 {
    372 	int s;
    373 	struct mbuf *m;
    374 
    375 	while (1) {
    376 		s = splnet();
    377 		IF_DEQUEUE(&ipintrq, m);
    378 		splx(s);
    379 		if (m == 0)
    380 			return;
    381 		ip_input(m);
    382 	}
    383 }
    384 
    385 /*
    386  * Ip input routine.  Checksum and byte swap header.  If fragmented
    387  * try to reassemble.  Process options.  Pass to next level.
    388  */
    389 void
    390 ip_input(struct mbuf *m)
    391 {
    392 	struct ip *ip = NULL;
    393 	struct ipq *fp;
    394 	struct in_ifaddr *ia;
    395 	struct ifaddr *ifa;
    396 	struct ipqent *ipqe;
    397 	int hlen = 0, mff, len;
    398 	int downmatch;
    399 
    400 #ifdef	DIAGNOSTIC
    401 	if ((m->m_flags & M_PKTHDR) == 0)
    402 		panic("ipintr no HDR");
    403 #endif
    404 #ifdef IPSEC
    405 	/*
    406 	 * should the inner packet be considered authentic?
    407 	 * see comment in ah4_input().
    408 	 */
    409 	if (m) {
    410 		m->m_flags &= ~M_AUTHIPHDR;
    411 		m->m_flags &= ~M_AUTHIPDGM;
    412 	}
    413 #endif
    414 	/*
    415 	 * If no IP addresses have been set yet but the interfaces
    416 	 * are receiving, can't do anything with incoming packets yet.
    417 	 */
    418 	if (TAILQ_FIRST(&in_ifaddr) == 0)
    419 		goto bad;
    420 	ipstat.ips_total++;
    421 	/*
    422 	 * If the IP header is not aligned, slurp it up into a new
    423 	 * mbuf with space for link headers, in the event we forward
    424 	 * it.  Otherwise, if it is aligned, make sure the entire
    425 	 * base IP header is in the first mbuf of the chain.
    426 	 */
    427 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
    428 		if ((m = m_copyup(m, sizeof(struct ip),
    429 				  (max_linkhdr + 3) & ~3)) == NULL) {
    430 			/* XXXJRT new stat, please */
    431 			ipstat.ips_toosmall++;
    432 			return;
    433 		}
    434 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
    435 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
    436 			ipstat.ips_toosmall++;
    437 			return;
    438 		}
    439 	}
    440 	ip = mtod(m, struct ip *);
    441 	if (ip->ip_v != IPVERSION) {
    442 		ipstat.ips_badvers++;
    443 		goto bad;
    444 	}
    445 	hlen = ip->ip_hl << 2;
    446 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
    447 		ipstat.ips_badhlen++;
    448 		goto bad;
    449 	}
    450 	if (hlen > m->m_len) {
    451 		if ((m = m_pullup(m, hlen)) == 0) {
    452 			ipstat.ips_badhlen++;
    453 			return;
    454 		}
    455 		ip = mtod(m, struct ip *);
    456 	}
    457 
    458 	/*
    459 	 * RFC1122: packets with a multicast source address are
    460 	 * not allowed.
    461 	 */
    462 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    463 		ipstat.ips_badaddr++;
    464 		goto bad;
    465 	}
    466 
    467 	/* 127/8 must not appear on wire - RFC1122 */
    468 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
    469 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
    470 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
    471 			ipstat.ips_badaddr++;
    472 			goto bad;
    473 		}
    474 	}
    475 
    476 	switch (m->m_pkthdr.csum_flags &
    477 		((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
    478 		 M_CSUM_IPv4_BAD)) {
    479 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
    480 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
    481 		goto badcsum;
    482 
    483 	case M_CSUM_IPv4:
    484 		/* Checksum was okay. */
    485 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
    486 		break;
    487 
    488 	default:
    489 		/* Must compute it ourselves. */
    490 		INET_CSUM_COUNTER_INCR(&ip_swcsum);
    491 		if (in_cksum(m, hlen) != 0)
    492 			goto bad;
    493 		break;
    494 	}
    495 
    496 	/* Retrieve the packet length. */
    497 	len = ntohs(ip->ip_len);
    498 
    499 	/*
    500 	 * Check for additional length bogosity
    501 	 */
    502 	if (len < hlen) {
    503 	 	ipstat.ips_badlen++;
    504 		goto bad;
    505 	}
    506 
    507 	/*
    508 	 * Check that the amount of data in the buffers
    509 	 * is as at least much as the IP header would have us expect.
    510 	 * Trim mbufs if longer than we expect.
    511 	 * Drop packet if shorter than we expect.
    512 	 */
    513 	if (m->m_pkthdr.len < len) {
    514 		ipstat.ips_tooshort++;
    515 		goto bad;
    516 	}
    517 	if (m->m_pkthdr.len > len) {
    518 		if (m->m_len == m->m_pkthdr.len) {
    519 			m->m_len = len;
    520 			m->m_pkthdr.len = len;
    521 		} else
    522 			m_adj(m, len - m->m_pkthdr.len);
    523 	}
    524 
    525 #ifdef IPSEC
    526 	/* ipflow (IP fast forwarding) is not compatible with IPsec. */
    527 	m->m_flags &= ~M_CANFASTFWD;
    528 #else
    529 	/*
    530 	 * Assume that we can create a fast-forward IP flow entry
    531 	 * based on this packet.
    532 	 */
    533 	m->m_flags |= M_CANFASTFWD;
    534 #endif
    535 
    536 #ifdef PFIL_HOOKS
    537 	/*
    538 	 * Run through list of hooks for input packets.  If there are any
    539 	 * filters which require that additional packets in the flow are
    540 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
    541 	 * Note that filters must _never_ set this flag, as another filter
    542 	 * in the list may have previously cleared it.
    543 	 */
    544 	/*
    545 	 * let ipfilter look at packet on the wire,
    546 	 * not the decapsulated packet.
    547 	 */
    548 #ifdef IPSEC
    549 	if (!ipsec_getnhist(m))
    550 #else
    551 	if (1)
    552 #endif
    553 	{
    554 		if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
    555 				   PFIL_IN) != 0)
    556 		return;
    557 		if (m == NULL)
    558 			return;
    559 		ip = mtod(m, struct ip *);
    560 		hlen = ip->ip_hl << 2;
    561 	}
    562 #endif /* PFIL_HOOKS */
    563 
    564 #ifdef ALTQ
    565 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
    566 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
    567 		/* packet dropped by traffic conditioner */
    568 		return;
    569 	}
    570 #endif
    571 
    572 	/*
    573 	 * Process options and, if not destined for us,
    574 	 * ship it on.  ip_dooptions returns 1 when an
    575 	 * error was detected (causing an icmp message
    576 	 * to be sent and the original packet to be freed).
    577 	 */
    578 	ip_nhops = 0;		/* for source routed packets */
    579 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
    580 		return;
    581 
    582 	/*
    583 	 * Check our list of addresses, to see if the packet is for us.
    584 	 *
    585 	 * Traditional 4.4BSD did not consult IFF_UP at all.
    586 	 * The behavior here is to treat addresses on !IFF_UP interface
    587 	 * as not mine.
    588 	 */
    589 	downmatch = 0;
    590 	LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
    591 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
    592 			if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
    593 				break;
    594 			else
    595 				downmatch++;
    596 		}
    597 	}
    598 	if (ia != NULL)
    599 		goto ours;
    600 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
    601 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
    602 			if (ifa->ifa_addr->sa_family != AF_INET)
    603 				continue;
    604 			ia = ifatoia(ifa);
    605 			if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
    606 			    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
    607 			    /*
    608 			     * Look for all-0's host part (old broadcast addr),
    609 			     * either for subnet or net.
    610 			     */
    611 			    ip->ip_dst.s_addr == ia->ia_subnet ||
    612 			    ip->ip_dst.s_addr == ia->ia_net)
    613 				goto ours;
    614 			/*
    615 			 * An interface with IP address zero accepts
    616 			 * all packets that arrive on that interface.
    617 			 */
    618 			if (in_nullhost(ia->ia_addr.sin_addr))
    619 				goto ours;
    620 		}
    621 	}
    622 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
    623 		struct in_multi *inm;
    624 #ifdef MROUTING
    625 		extern struct socket *ip_mrouter;
    626 
    627 		if (M_READONLY(m)) {
    628 			if ((m = m_pullup(m, hlen)) == 0) {
    629 				ipstat.ips_toosmall++;
    630 				return;
    631 			}
    632 			ip = mtod(m, struct ip *);
    633 		}
    634 
    635 		if (ip_mrouter) {
    636 			/*
    637 			 * If we are acting as a multicast router, all
    638 			 * incoming multicast packets are passed to the
    639 			 * kernel-level multicast forwarding function.
    640 			 * The packet is returned (relatively) intact; if
    641 			 * ip_mforward() returns a non-zero value, the packet
    642 			 * must be discarded, else it may be accepted below.
    643 			 *
    644 			 * (The IP ident field is put in the same byte order
    645 			 * as expected when ip_mforward() is called from
    646 			 * ip_output().)
    647 			 */
    648 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
    649 				ipstat.ips_cantforward++;
    650 				m_freem(m);
    651 				return;
    652 			}
    653 
    654 			/*
    655 			 * The process-level routing demon needs to receive
    656 			 * all multicast IGMP packets, whether or not this
    657 			 * host belongs to their destination groups.
    658 			 */
    659 			if (ip->ip_p == IPPROTO_IGMP)
    660 				goto ours;
    661 			ipstat.ips_forward++;
    662 		}
    663 #endif
    664 		/*
    665 		 * See if we belong to the destination multicast group on the
    666 		 * arrival interface.
    667 		 */
    668 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
    669 		if (inm == NULL) {
    670 			ipstat.ips_cantforward++;
    671 			m_freem(m);
    672 			return;
    673 		}
    674 		goto ours;
    675 	}
    676 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    677 	    in_nullhost(ip->ip_dst))
    678 		goto ours;
    679 
    680 	/*
    681 	 * Not for us; forward if possible and desirable.
    682 	 */
    683 	if (ipforwarding == 0) {
    684 		ipstat.ips_cantforward++;
    685 		m_freem(m);
    686 	} else {
    687 		/*
    688 		 * If ip_dst matched any of my address on !IFF_UP interface,
    689 		 * and there's no IFF_UP interface that matches ip_dst,
    690 		 * send icmp unreach.  Forwarding it will result in in-kernel
    691 		 * forwarding loop till TTL goes to 0.
    692 		 */
    693 		if (downmatch) {
    694 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
    695 			ipstat.ips_cantforward++;
    696 			return;
    697 		}
    698 #ifdef IPSEC
    699 		if (ipsec4_in_reject(m, NULL)) {
    700 			ipsecstat.in_polvio++;
    701 			goto bad;
    702 		}
    703 #endif
    704 
    705 		ip_forward(m, 0);
    706 	}
    707 	return;
    708 
    709 ours:
    710 	/*
    711 	 * If offset or IP_MF are set, must reassemble.
    712 	 * Otherwise, nothing need be done.
    713 	 * (We could look in the reassembly queue to see
    714 	 * if the packet was previously fragmented,
    715 	 * but it's not worth the time; just let them time out.)
    716 	 */
    717 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
    718 		if (M_READONLY(m)) {
    719 			if ((m = m_pullup(m, hlen)) == NULL) {
    720 				ipstat.ips_toosmall++;
    721 				goto bad;
    722 			}
    723 			ip = mtod(m, struct ip *);
    724 		}
    725 
    726 		/*
    727 		 * Look for queue of fragments
    728 		 * of this datagram.
    729 		 */
    730 		IPQ_LOCK();
    731 		LIST_FOREACH(fp, &ipq, ipq_q)
    732 			if (ip->ip_id == fp->ipq_id &&
    733 			    in_hosteq(ip->ip_src, fp->ipq_src) &&
    734 			    in_hosteq(ip->ip_dst, fp->ipq_dst) &&
    735 			    ip->ip_p == fp->ipq_p)
    736 				goto found;
    737 		fp = 0;
    738 found:
    739 
    740 		/*
    741 		 * Adjust ip_len to not reflect header,
    742 		 * set ipqe_mff if more fragments are expected,
    743 		 * convert offset of this to bytes.
    744 		 */
    745 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
    746 		mff = (ip->ip_off & htons(IP_MF)) != 0;
    747 		if (mff) {
    748 		        /*
    749 		         * Make sure that fragments have a data length
    750 			 * that's a non-zero multiple of 8 bytes.
    751 		         */
    752 			if (ntohs(ip->ip_len) == 0 ||
    753 			    (ntohs(ip->ip_len) & 0x7) != 0) {
    754 				ipstat.ips_badfrags++;
    755 				IPQ_UNLOCK();
    756 				goto bad;
    757 			}
    758 		}
    759 		ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
    760 
    761 		/*
    762 		 * If datagram marked as having more fragments
    763 		 * or if this is not the first fragment,
    764 		 * attempt reassembly; if it succeeds, proceed.
    765 		 */
    766 		if (mff || ip->ip_off != htons(0)) {
    767 			ipstat.ips_fragments++;
    768 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
    769 			if (ipqe == NULL) {
    770 				ipstat.ips_rcvmemdrop++;
    771 				IPQ_UNLOCK();
    772 				goto bad;
    773 			}
    774 			ipqe->ipqe_mff = mff;
    775 			ipqe->ipqe_m = m;
    776 			ipqe->ipqe_ip = ip;
    777 			m = ip_reass(ipqe, fp);
    778 			if (m == 0) {
    779 				IPQ_UNLOCK();
    780 				return;
    781 			}
    782 			ipstat.ips_reassembled++;
    783 			ip = mtod(m, struct ip *);
    784 			hlen = ip->ip_hl << 2;
    785 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
    786 		} else
    787 			if (fp)
    788 				ip_freef(fp);
    789 		IPQ_UNLOCK();
    790 	}
    791 
    792 #ifdef IPSEC
    793 	/*
    794 	 * enforce IPsec policy checking if we are seeing last header.
    795 	 * note that we do not visit this with protocols with pcb layer
    796 	 * code - like udp/tcp/raw ip.
    797 	 */
    798 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
    799 	    ipsec4_in_reject(m, NULL)) {
    800 		ipsecstat.in_polvio++;
    801 		goto bad;
    802 	}
    803 #endif
    804 
    805 	/*
    806 	 * Switch out to protocol's input routine.
    807 	 */
    808 #if IFA_STATS
    809 	if (ia && ip)
    810 		ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
    811 #endif
    812 	ipstat.ips_delivered++;
    813     {
    814 	int off = hlen, nh = ip->ip_p;
    815 
    816 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
    817 	return;
    818     }
    819 bad:
    820 	m_freem(m);
    821 	return;
    822 
    823 badcsum:
    824 	ipstat.ips_badsum++;
    825 	m_freem(m);
    826 }
    827 
    828 /*
    829  * Take incoming datagram fragment and try to
    830  * reassemble it into whole datagram.  If a chain for
    831  * reassembly of this datagram already exists, then it
    832  * is given as fp; otherwise have to make a chain.
    833  */
    834 struct mbuf *
    835 ip_reass(ipqe, fp)
    836 	struct ipqent *ipqe;
    837 	struct ipq *fp;
    838 {
    839 	struct mbuf *m = ipqe->ipqe_m;
    840 	struct ipqent *nq, *p, *q;
    841 	struct ip *ip;
    842 	struct mbuf *t;
    843 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
    844 	int i, next;
    845 
    846 	IPQ_LOCK_CHECK();
    847 
    848 	/*
    849 	 * Presence of header sizes in mbufs
    850 	 * would confuse code below.
    851 	 */
    852 	m->m_data += hlen;
    853 	m->m_len -= hlen;
    854 
    855 	/*
    856 	 * If first fragment to arrive, create a reassembly queue.
    857 	 */
    858 	if (fp == 0) {
    859 		/*
    860 		 * Enforce upper bound on number of fragmented packets
    861 		 * for which we attempt reassembly;
    862 		 * If maxfrag is 0, never accept fragments.
    863 		 * If maxfrag is -1, accept all fragments without limitation.
    864 		 */
    865 		if (ip_maxfragpackets < 0)
    866 			;
    867 		else if (ip_nfragpackets >= ip_maxfragpackets)
    868 			goto dropfrag;
    869 		ip_nfragpackets++;
    870 		MALLOC(fp, struct ipq *, sizeof (struct ipq),
    871 		    M_FTABLE, M_NOWAIT);
    872 		if (fp == NULL)
    873 			goto dropfrag;
    874 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
    875 		fp->ipq_ttl = IPFRAGTTL;
    876 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
    877 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
    878 		TAILQ_INIT(&fp->ipq_fragq);
    879 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
    880 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
    881 		p = NULL;
    882 		goto insert;
    883 	}
    884 
    885 	/*
    886 	 * Find a segment which begins after this one does.
    887 	 */
    888 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
    889 	    p = q, q = TAILQ_NEXT(q, ipqe_q))
    890 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
    891 			break;
    892 
    893 	/*
    894 	 * If there is a preceding segment, it may provide some of
    895 	 * our data already.  If so, drop the data from the incoming
    896 	 * segment.  If it provides all of our data, drop us.
    897 	 */
    898 	if (p != NULL) {
    899 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
    900 		    ntohs(ipqe->ipqe_ip->ip_off);
    901 		if (i > 0) {
    902 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
    903 				goto dropfrag;
    904 			m_adj(ipqe->ipqe_m, i);
    905 			ipqe->ipqe_ip->ip_off =
    906 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
    907 			ipqe->ipqe_ip->ip_len =
    908 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
    909 		}
    910 	}
    911 
    912 	/*
    913 	 * While we overlap succeeding segments trim them or,
    914 	 * if they are completely covered, dequeue them.
    915 	 */
    916 	for (; q != NULL &&
    917 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
    918 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
    919 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
    920 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
    921 		if (i < ntohs(q->ipqe_ip->ip_len)) {
    922 			q->ipqe_ip->ip_len =
    923 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
    924 			q->ipqe_ip->ip_off =
    925 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
    926 			m_adj(q->ipqe_m, i);
    927 			break;
    928 		}
    929 		nq = TAILQ_NEXT(q, ipqe_q);
    930 		m_freem(q->ipqe_m);
    931 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
    932 		pool_put(&ipqent_pool, q);
    933 	}
    934 
    935 insert:
    936 	/*
    937 	 * Stick new segment in its place;
    938 	 * check for complete reassembly.
    939 	 */
    940 	if (p == NULL) {
    941 		TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
    942 	} else {
    943 		TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
    944 	}
    945 	next = 0;
    946 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
    947 	    p = q, q = TAILQ_NEXT(q, ipqe_q)) {
    948 		if (ntohs(q->ipqe_ip->ip_off) != next)
    949 			return (0);
    950 		next += ntohs(q->ipqe_ip->ip_len);
    951 	}
    952 	if (p->ipqe_mff)
    953 		return (0);
    954 
    955 	/*
    956 	 * Reassembly is complete.  Check for a bogus message size and
    957 	 * concatenate fragments.
    958 	 */
    959 	q = TAILQ_FIRST(&fp->ipq_fragq);
    960 	ip = q->ipqe_ip;
    961 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
    962 		ipstat.ips_toolong++;
    963 		ip_freef(fp);
    964 		return (0);
    965 	}
    966 	m = q->ipqe_m;
    967 	t = m->m_next;
    968 	m->m_next = 0;
    969 	m_cat(m, t);
    970 	nq = TAILQ_NEXT(q, ipqe_q);
    971 	pool_put(&ipqent_pool, q);
    972 	for (q = nq; q != NULL; q = nq) {
    973 		t = q->ipqe_m;
    974 		nq = TAILQ_NEXT(q, ipqe_q);
    975 		pool_put(&ipqent_pool, q);
    976 		m_cat(m, t);
    977 	}
    978 
    979 	/*
    980 	 * Create header for new ip packet by
    981 	 * modifying header of first packet;
    982 	 * dequeue and discard fragment reassembly header.
    983 	 * Make header visible.
    984 	 */
    985 	ip->ip_len = htons(next);
    986 	ip->ip_src = fp->ipq_src;
    987 	ip->ip_dst = fp->ipq_dst;
    988 	LIST_REMOVE(fp, ipq_q);
    989 	FREE(fp, M_FTABLE);
    990 	ip_nfragpackets--;
    991 	m->m_len += (ip->ip_hl << 2);
    992 	m->m_data -= (ip->ip_hl << 2);
    993 	/* some debugging cruft by sklower, below, will go away soon */
    994 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
    995 		int plen = 0;
    996 		for (t = m; t; t = t->m_next)
    997 			plen += t->m_len;
    998 		m->m_pkthdr.len = plen;
    999 	}
   1000 	return (m);
   1001 
   1002 dropfrag:
   1003 	ipstat.ips_fragdropped++;
   1004 	m_freem(m);
   1005 	pool_put(&ipqent_pool, ipqe);
   1006 	return (0);
   1007 }
   1008 
   1009 /*
   1010  * Free a fragment reassembly header and all
   1011  * associated datagrams.
   1012  */
   1013 void
   1014 ip_freef(fp)
   1015 	struct ipq *fp;
   1016 {
   1017 	struct ipqent *q, *p;
   1018 
   1019 	IPQ_LOCK_CHECK();
   1020 
   1021 	for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
   1022 		p = TAILQ_NEXT(q, ipqe_q);
   1023 		m_freem(q->ipqe_m);
   1024 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
   1025 		pool_put(&ipqent_pool, q);
   1026 	}
   1027 	LIST_REMOVE(fp, ipq_q);
   1028 	FREE(fp, M_FTABLE);
   1029 	ip_nfragpackets--;
   1030 }
   1031 
   1032 /*
   1033  * IP timer processing;
   1034  * if a timer expires on a reassembly
   1035  * queue, discard it.
   1036  */
   1037 void
   1038 ip_slowtimo()
   1039 {
   1040 	struct ipq *fp, *nfp;
   1041 	int s = splsoftnet();
   1042 
   1043 	IPQ_LOCK();
   1044 	for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) {
   1045 		nfp = LIST_NEXT(fp, ipq_q);
   1046 		if (--fp->ipq_ttl == 0) {
   1047 			ipstat.ips_fragtimeout++;
   1048 			ip_freef(fp);
   1049 		}
   1050 	}
   1051 	/*
   1052 	 * If we are over the maximum number of fragments
   1053 	 * (due to the limit being lowered), drain off
   1054 	 * enough to get down to the new limit.
   1055 	 */
   1056 	if (ip_maxfragpackets < 0)
   1057 		;
   1058 	else {
   1059 		while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq))
   1060 			ip_freef(LIST_FIRST(&ipq));
   1061 	}
   1062 	IPQ_UNLOCK();
   1063 #ifdef GATEWAY
   1064 	ipflow_slowtimo();
   1065 #endif
   1066 	splx(s);
   1067 }
   1068 
   1069 /*
   1070  * Drain off all datagram fragments.
   1071  */
   1072 void
   1073 ip_drain()
   1074 {
   1075 
   1076 	/*
   1077 	 * We may be called from a device's interrupt context.  If
   1078 	 * the ipq is already busy, just bail out now.
   1079 	 */
   1080 	if (ipq_lock_try() == 0)
   1081 		return;
   1082 
   1083 	while (LIST_FIRST(&ipq) != NULL) {
   1084 		ipstat.ips_fragdropped++;
   1085 		ip_freef(LIST_FIRST(&ipq));
   1086 	}
   1087 
   1088 	IPQ_UNLOCK();
   1089 }
   1090 
   1091 /*
   1092  * Do option processing on a datagram,
   1093  * possibly discarding it if bad options are encountered,
   1094  * or forwarding it if source-routed.
   1095  * Returns 1 if packet has been forwarded/freed,
   1096  * 0 if the packet should be processed further.
   1097  */
   1098 int
   1099 ip_dooptions(m)
   1100 	struct mbuf *m;
   1101 {
   1102 	struct ip *ip = mtod(m, struct ip *);
   1103 	u_char *cp, *cp0;
   1104 	struct ip_timestamp *ipt;
   1105 	struct in_ifaddr *ia;
   1106 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
   1107 	struct in_addr dst;
   1108 	n_time ntime;
   1109 
   1110 	dst = ip->ip_dst;
   1111 	cp = (u_char *)(ip + 1);
   1112 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
   1113 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
   1114 		opt = cp[IPOPT_OPTVAL];
   1115 		if (opt == IPOPT_EOL)
   1116 			break;
   1117 		if (opt == IPOPT_NOP)
   1118 			optlen = 1;
   1119 		else {
   1120 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
   1121 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1122 				goto bad;
   1123 			}
   1124 			optlen = cp[IPOPT_OLEN];
   1125 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
   1126 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1127 				goto bad;
   1128 			}
   1129 		}
   1130 		switch (opt) {
   1131 
   1132 		default:
   1133 			break;
   1134 
   1135 		/*
   1136 		 * Source routing with record.
   1137 		 * Find interface with current destination address.
   1138 		 * If none on this machine then drop if strictly routed,
   1139 		 * or do nothing if loosely routed.
   1140 		 * Record interface address and bring up next address
   1141 		 * component.  If strictly routed make sure next
   1142 		 * address is on directly accessible net.
   1143 		 */
   1144 		case IPOPT_LSRR:
   1145 		case IPOPT_SSRR:
   1146 			if (ip_allowsrcrt == 0) {
   1147 				type = ICMP_UNREACH;
   1148 				code = ICMP_UNREACH_NET_PROHIB;
   1149 				goto bad;
   1150 			}
   1151 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1152 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1153 				goto bad;
   1154 			}
   1155 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1156 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1157 				goto bad;
   1158 			}
   1159 			ipaddr.sin_addr = ip->ip_dst;
   1160 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
   1161 			if (ia == 0) {
   1162 				if (opt == IPOPT_SSRR) {
   1163 					type = ICMP_UNREACH;
   1164 					code = ICMP_UNREACH_SRCFAIL;
   1165 					goto bad;
   1166 				}
   1167 				/*
   1168 				 * Loose routing, and not at next destination
   1169 				 * yet; nothing to do except forward.
   1170 				 */
   1171 				break;
   1172 			}
   1173 			off--;			/* 0 origin */
   1174 			if ((off + sizeof(struct in_addr)) > optlen) {
   1175 				/*
   1176 				 * End of source route.  Should be for us.
   1177 				 */
   1178 				save_rte(cp, ip->ip_src);
   1179 				break;
   1180 			}
   1181 			/*
   1182 			 * locate outgoing interface
   1183 			 */
   1184 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
   1185 			    sizeof(ipaddr.sin_addr));
   1186 			if (opt == IPOPT_SSRR)
   1187 				ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
   1188 			else
   1189 				ia = ip_rtaddr(ipaddr.sin_addr);
   1190 			if (ia == 0) {
   1191 				type = ICMP_UNREACH;
   1192 				code = ICMP_UNREACH_SRCFAIL;
   1193 				goto bad;
   1194 			}
   1195 			ip->ip_dst = ipaddr.sin_addr;
   1196 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
   1197 			    (caddr_t)(cp + off), sizeof(struct in_addr));
   1198 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1199 			/*
   1200 			 * Let ip_intr's mcast routing check handle mcast pkts
   1201 			 */
   1202 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
   1203 			break;
   1204 
   1205 		case IPOPT_RR:
   1206 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1207 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1208 				goto bad;
   1209 			}
   1210 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1211 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1212 				goto bad;
   1213 			}
   1214 			/*
   1215 			 * If no space remains, ignore.
   1216 			 */
   1217 			off--;			/* 0 origin */
   1218 			if ((off + sizeof(struct in_addr)) > optlen)
   1219 				break;
   1220 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
   1221 			    sizeof(ipaddr.sin_addr));
   1222 			/*
   1223 			 * locate outgoing interface; if we're the destination,
   1224 			 * use the incoming interface (should be same).
   1225 			 */
   1226 			if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
   1227 			    == NULL &&
   1228 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
   1229 				type = ICMP_UNREACH;
   1230 				code = ICMP_UNREACH_HOST;
   1231 				goto bad;
   1232 			}
   1233 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
   1234 			    (caddr_t)(cp + off), sizeof(struct in_addr));
   1235 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1236 			break;
   1237 
   1238 		case IPOPT_TS:
   1239 			code = cp - (u_char *)ip;
   1240 			ipt = (struct ip_timestamp *)cp;
   1241 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
   1242 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
   1243 				goto bad;
   1244 			}
   1245 			if (ipt->ipt_ptr < 5) {
   1246 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
   1247 				goto bad;
   1248 			}
   1249 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
   1250 				if (++ipt->ipt_oflw == 0) {
   1251 					code = (u_char *)&ipt->ipt_ptr -
   1252 					    (u_char *)ip;
   1253 					goto bad;
   1254 				}
   1255 				break;
   1256 			}
   1257 			cp0 = (cp + ipt->ipt_ptr - 1);
   1258 			switch (ipt->ipt_flg) {
   1259 
   1260 			case IPOPT_TS_TSONLY:
   1261 				break;
   1262 
   1263 			case IPOPT_TS_TSANDADDR:
   1264 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1265 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1266 					code = (u_char *)&ipt->ipt_ptr -
   1267 					    (u_char *)ip;
   1268 					goto bad;
   1269 				}
   1270 				ipaddr.sin_addr = dst;
   1271 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
   1272 				    m->m_pkthdr.rcvif));
   1273 				if (ia == 0)
   1274 					continue;
   1275 				bcopy(&ia->ia_addr.sin_addr,
   1276 				    cp0, sizeof(struct in_addr));
   1277 				ipt->ipt_ptr += sizeof(struct in_addr);
   1278 				break;
   1279 
   1280 			case IPOPT_TS_PRESPEC:
   1281 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1282 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1283 					code = (u_char *)&ipt->ipt_ptr -
   1284 					    (u_char *)ip;
   1285 					goto bad;
   1286 				}
   1287 				bcopy(cp0, &ipaddr.sin_addr,
   1288 				    sizeof(struct in_addr));
   1289 				if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
   1290 				    == NULL)
   1291 					continue;
   1292 				ipt->ipt_ptr += sizeof(struct in_addr);
   1293 				break;
   1294 
   1295 			default:
   1296 				/* XXX can't take &ipt->ipt_flg */
   1297 				code = (u_char *)&ipt->ipt_ptr -
   1298 				    (u_char *)ip + 1;
   1299 				goto bad;
   1300 			}
   1301 			ntime = iptime();
   1302 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
   1303 			bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
   1304 			    sizeof(n_time));
   1305 			ipt->ipt_ptr += sizeof(n_time);
   1306 		}
   1307 	}
   1308 	if (forward) {
   1309 		if (ip_forwsrcrt == 0) {
   1310 			type = ICMP_UNREACH;
   1311 			code = ICMP_UNREACH_SRCFAIL;
   1312 			goto bad;
   1313 		}
   1314 		ip_forward(m, 1);
   1315 		return (1);
   1316 	}
   1317 	return (0);
   1318 bad:
   1319 	icmp_error(m, type, code, 0, 0);
   1320 	ipstat.ips_badoptions++;
   1321 	return (1);
   1322 }
   1323 
   1324 /*
   1325  * Given address of next destination (final or next hop),
   1326  * return internet address info of interface to be used to get there.
   1327  */
   1328 struct in_ifaddr *
   1329 ip_rtaddr(dst)
   1330 	 struct in_addr dst;
   1331 {
   1332 	struct sockaddr_in *sin;
   1333 
   1334 	sin = satosin(&ipforward_rt.ro_dst);
   1335 
   1336 	if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
   1337 		if (ipforward_rt.ro_rt) {
   1338 			RTFREE(ipforward_rt.ro_rt);
   1339 			ipforward_rt.ro_rt = 0;
   1340 		}
   1341 		sin->sin_family = AF_INET;
   1342 		sin->sin_len = sizeof(*sin);
   1343 		sin->sin_addr = dst;
   1344 
   1345 		rtalloc(&ipforward_rt);
   1346 	}
   1347 	if (ipforward_rt.ro_rt == 0)
   1348 		return ((struct in_ifaddr *)0);
   1349 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
   1350 }
   1351 
   1352 /*
   1353  * Save incoming source route for use in replies,
   1354  * to be picked up later by ip_srcroute if the receiver is interested.
   1355  */
   1356 void
   1357 save_rte(option, dst)
   1358 	u_char *option;
   1359 	struct in_addr dst;
   1360 {
   1361 	unsigned olen;
   1362 
   1363 	olen = option[IPOPT_OLEN];
   1364 #ifdef DIAGNOSTIC
   1365 	if (ipprintfs)
   1366 		printf("save_rte: olen %d\n", olen);
   1367 #endif /* 0 */
   1368 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
   1369 		return;
   1370 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
   1371 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
   1372 	ip_srcrt.dst = dst;
   1373 }
   1374 
   1375 /*
   1376  * Retrieve incoming source route for use in replies,
   1377  * in the same form used by setsockopt.
   1378  * The first hop is placed before the options, will be removed later.
   1379  */
   1380 struct mbuf *
   1381 ip_srcroute()
   1382 {
   1383 	struct in_addr *p, *q;
   1384 	struct mbuf *m;
   1385 
   1386 	if (ip_nhops == 0)
   1387 		return ((struct mbuf *)0);
   1388 	m = m_get(M_DONTWAIT, MT_SOOPTS);
   1389 	if (m == 0)
   1390 		return ((struct mbuf *)0);
   1391 
   1392 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
   1393 
   1394 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
   1395 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
   1396 	    OPTSIZ;
   1397 #ifdef DIAGNOSTIC
   1398 	if (ipprintfs)
   1399 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
   1400 #endif
   1401 
   1402 	/*
   1403 	 * First save first hop for return route
   1404 	 */
   1405 	p = &ip_srcrt.route[ip_nhops - 1];
   1406 	*(mtod(m, struct in_addr *)) = *p--;
   1407 #ifdef DIAGNOSTIC
   1408 	if (ipprintfs)
   1409 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
   1410 #endif
   1411 
   1412 	/*
   1413 	 * Copy option fields and padding (nop) to mbuf.
   1414 	 */
   1415 	ip_srcrt.nop = IPOPT_NOP;
   1416 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
   1417 	bcopy((caddr_t)&ip_srcrt.nop,
   1418 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
   1419 	q = (struct in_addr *)(mtod(m, caddr_t) +
   1420 	    sizeof(struct in_addr) + OPTSIZ);
   1421 #undef OPTSIZ
   1422 	/*
   1423 	 * Record return path as an IP source route,
   1424 	 * reversing the path (pointers are now aligned).
   1425 	 */
   1426 	while (p >= ip_srcrt.route) {
   1427 #ifdef DIAGNOSTIC
   1428 		if (ipprintfs)
   1429 			printf(" %x", ntohl(q->s_addr));
   1430 #endif
   1431 		*q++ = *p--;
   1432 	}
   1433 	/*
   1434 	 * Last hop goes to final destination.
   1435 	 */
   1436 	*q = ip_srcrt.dst;
   1437 #ifdef DIAGNOSTIC
   1438 	if (ipprintfs)
   1439 		printf(" %x\n", ntohl(q->s_addr));
   1440 #endif
   1441 	return (m);
   1442 }
   1443 
   1444 /*
   1445  * Strip out IP options, at higher
   1446  * level protocol in the kernel.
   1447  * Second argument is buffer to which options
   1448  * will be moved, and return value is their length.
   1449  * XXX should be deleted; last arg currently ignored.
   1450  */
   1451 void
   1452 ip_stripoptions(m, mopt)
   1453 	struct mbuf *m;
   1454 	struct mbuf *mopt;
   1455 {
   1456 	int i;
   1457 	struct ip *ip = mtod(m, struct ip *);
   1458 	caddr_t opts;
   1459 	int olen;
   1460 
   1461 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
   1462 	opts = (caddr_t)(ip + 1);
   1463 	i = m->m_len - (sizeof (struct ip) + olen);
   1464 	bcopy(opts  + olen, opts, (unsigned)i);
   1465 	m->m_len -= olen;
   1466 	if (m->m_flags & M_PKTHDR)
   1467 		m->m_pkthdr.len -= olen;
   1468 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
   1469 	ip->ip_hl = sizeof (struct ip) >> 2;
   1470 }
   1471 
   1472 const int inetctlerrmap[PRC_NCMDS] = {
   1473 	0,		0,		0,		0,
   1474 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
   1475 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
   1476 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
   1477 	0,		0,		0,		0,
   1478 	ENOPROTOOPT
   1479 };
   1480 
   1481 /*
   1482  * Forward a packet.  If some error occurs return the sender
   1483  * an icmp packet.  Note we can't always generate a meaningful
   1484  * icmp message because icmp doesn't have a large enough repertoire
   1485  * of codes and types.
   1486  *
   1487  * If not forwarding, just drop the packet.  This could be confusing
   1488  * if ipforwarding was zero but some routing protocol was advancing
   1489  * us as a gateway to somewhere.  However, we must let the routing
   1490  * protocol deal with that.
   1491  *
   1492  * The srcrt parameter indicates whether the packet is being forwarded
   1493  * via a source route.
   1494  */
   1495 void
   1496 ip_forward(m, srcrt)
   1497 	struct mbuf *m;
   1498 	int srcrt;
   1499 {
   1500 	struct ip *ip = mtod(m, struct ip *);
   1501 	struct sockaddr_in *sin;
   1502 	struct rtentry *rt;
   1503 	int error, type = 0, code = 0;
   1504 	struct mbuf *mcopy;
   1505 	n_long dest;
   1506 	struct ifnet *destifp;
   1507 #ifdef IPSEC
   1508 	struct ifnet dummyifp;
   1509 #endif
   1510 
   1511 	/*
   1512 	 * Clear any in-bound checksum flags for this packet.
   1513 	 */
   1514 	m->m_pkthdr.csum_flags = 0;
   1515 
   1516 	dest = 0;
   1517 #ifdef DIAGNOSTIC
   1518 	if (ipprintfs)
   1519 		printf("forward: src %2.2x dst %2.2x ttl %x\n",
   1520 		    ntohl(ip->ip_src.s_addr),
   1521 		    ntohl(ip->ip_dst.s_addr), ip->ip_ttl);
   1522 #endif
   1523 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
   1524 		ipstat.ips_cantforward++;
   1525 		m_freem(m);
   1526 		return;
   1527 	}
   1528 	if (ip->ip_ttl <= IPTTLDEC) {
   1529 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
   1530 		return;
   1531 	}
   1532 	ip->ip_ttl -= IPTTLDEC;
   1533 
   1534 	sin = satosin(&ipforward_rt.ro_dst);
   1535 	if ((rt = ipforward_rt.ro_rt) == 0 ||
   1536 	    !in_hosteq(ip->ip_dst, sin->sin_addr)) {
   1537 		if (ipforward_rt.ro_rt) {
   1538 			RTFREE(ipforward_rt.ro_rt);
   1539 			ipforward_rt.ro_rt = 0;
   1540 		}
   1541 		sin->sin_family = AF_INET;
   1542 		sin->sin_len = sizeof(struct sockaddr_in);
   1543 		sin->sin_addr = ip->ip_dst;
   1544 
   1545 		rtalloc(&ipforward_rt);
   1546 		if (ipforward_rt.ro_rt == 0) {
   1547 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
   1548 			return;
   1549 		}
   1550 		rt = ipforward_rt.ro_rt;
   1551 	}
   1552 
   1553 	/*
   1554 	 * Save at most 68 bytes of the packet in case
   1555 	 * we need to generate an ICMP message to the src.
   1556 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
   1557 	 */
   1558 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
   1559 	if (mcopy)
   1560 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
   1561 
   1562 	/*
   1563 	 * If forwarding packet using same interface that it came in on,
   1564 	 * perhaps should send a redirect to sender to shortcut a hop.
   1565 	 * Only send redirect if source is sending directly to us,
   1566 	 * and if packet was not source routed (or has any options).
   1567 	 * Also, don't send redirect if forwarding using a default route
   1568 	 * or a route modified by a redirect.
   1569 	 */
   1570 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
   1571 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
   1572 	    !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
   1573 	    ipsendredirects && !srcrt) {
   1574 		if (rt->rt_ifa &&
   1575 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
   1576 		    ifatoia(rt->rt_ifa)->ia_subnet) {
   1577 			if (rt->rt_flags & RTF_GATEWAY)
   1578 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
   1579 			else
   1580 				dest = ip->ip_dst.s_addr;
   1581 			/*
   1582 			 * Router requirements says to only send host
   1583 			 * redirects.
   1584 			 */
   1585 			type = ICMP_REDIRECT;
   1586 			code = ICMP_REDIRECT_HOST;
   1587 #ifdef DIAGNOSTIC
   1588 			if (ipprintfs)
   1589 				printf("redirect (%d) to %x\n", code,
   1590 				    (u_int32_t)dest);
   1591 #endif
   1592 		}
   1593 	}
   1594 
   1595 #ifdef IPSEC
   1596 	/* Don't lookup socket in forwarding case */
   1597 	(void)ipsec_setsocket(m, NULL);
   1598 #endif
   1599 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
   1600 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0);
   1601 	if (error)
   1602 		ipstat.ips_cantforward++;
   1603 	else {
   1604 		ipstat.ips_forward++;
   1605 		if (type)
   1606 			ipstat.ips_redirectsent++;
   1607 		else {
   1608 			if (mcopy) {
   1609 #ifdef GATEWAY
   1610 				if (mcopy->m_flags & M_CANFASTFWD)
   1611 					ipflow_create(&ipforward_rt, mcopy);
   1612 #endif
   1613 				m_freem(mcopy);
   1614 			}
   1615 			return;
   1616 		}
   1617 	}
   1618 	if (mcopy == NULL)
   1619 		return;
   1620 	destifp = NULL;
   1621 
   1622 	switch (error) {
   1623 
   1624 	case 0:				/* forwarded, but need redirect */
   1625 		/* type, code set above */
   1626 		break;
   1627 
   1628 	case ENETUNREACH:		/* shouldn't happen, checked above */
   1629 	case EHOSTUNREACH:
   1630 	case ENETDOWN:
   1631 	case EHOSTDOWN:
   1632 	default:
   1633 		type = ICMP_UNREACH;
   1634 		code = ICMP_UNREACH_HOST;
   1635 		break;
   1636 
   1637 	case EMSGSIZE:
   1638 		type = ICMP_UNREACH;
   1639 		code = ICMP_UNREACH_NEEDFRAG;
   1640 #ifndef IPSEC
   1641 		if (ipforward_rt.ro_rt)
   1642 			destifp = ipforward_rt.ro_rt->rt_ifp;
   1643 #else
   1644 		/*
   1645 		 * If the packet is routed over IPsec tunnel, tell the
   1646 		 * originator the tunnel MTU.
   1647 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
   1648 		 * XXX quickhack!!!
   1649 		 */
   1650 		if (ipforward_rt.ro_rt) {
   1651 			struct secpolicy *sp;
   1652 			int ipsecerror;
   1653 			size_t ipsechdr;
   1654 			struct route *ro;
   1655 
   1656 			sp = ipsec4_getpolicybyaddr(mcopy,
   1657 			                            IPSEC_DIR_OUTBOUND,
   1658 			                            IP_FORWARDING,
   1659 			                            &ipsecerror);
   1660 
   1661 			if (sp == NULL)
   1662 				destifp = ipforward_rt.ro_rt->rt_ifp;
   1663 			else {
   1664 				/* count IPsec header size */
   1665 				ipsechdr = ipsec4_hdrsiz(mcopy,
   1666 				                         IPSEC_DIR_OUTBOUND,
   1667 				                         NULL);
   1668 
   1669 				/*
   1670 				 * find the correct route for outer IPv4
   1671 				 * header, compute tunnel MTU.
   1672 				 *
   1673 				 * XXX BUG ALERT
   1674 				 * The "dummyifp" code relies upon the fact
   1675 				 * that icmp_error() touches only ifp->if_mtu.
   1676 				 */
   1677 				/*XXX*/
   1678 				destifp = NULL;
   1679 				if (sp->req != NULL
   1680 				 && sp->req->sav != NULL
   1681 				 && sp->req->sav->sah != NULL) {
   1682 					ro = &sp->req->sav->sah->sa_route;
   1683 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
   1684 						dummyifp.if_mtu =
   1685 						    ro->ro_rt->rt_rmx.rmx_mtu ?
   1686 						    ro->ro_rt->rt_rmx.rmx_mtu :
   1687 						    ro->ro_rt->rt_ifp->if_mtu;
   1688 						dummyifp.if_mtu -= ipsechdr;
   1689 						destifp = &dummyifp;
   1690 					}
   1691 				}
   1692 
   1693 				key_freesp(sp);
   1694 			}
   1695 		}
   1696 #endif /*IPSEC*/
   1697 		ipstat.ips_cantfrag++;
   1698 		break;
   1699 
   1700 	case ENOBUFS:
   1701 #if 1
   1702 		/*
   1703 		 * a router should not generate ICMP_SOURCEQUENCH as
   1704 		 * required in RFC1812 Requirements for IP Version 4 Routers.
   1705 		 * source quench could be a big problem under DoS attacks,
   1706 		 * or if the underlying interface is rate-limited.
   1707 		 */
   1708 		if (mcopy)
   1709 			m_freem(mcopy);
   1710 		return;
   1711 #else
   1712 		type = ICMP_SOURCEQUENCH;
   1713 		code = 0;
   1714 		break;
   1715 #endif
   1716 	}
   1717 	icmp_error(mcopy, type, code, dest, destifp);
   1718 }
   1719 
   1720 void
   1721 ip_savecontrol(inp, mp, ip, m)
   1722 	struct inpcb *inp;
   1723 	struct mbuf **mp;
   1724 	struct ip *ip;
   1725 	struct mbuf *m;
   1726 {
   1727 
   1728 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
   1729 		struct timeval tv;
   1730 
   1731 		microtime(&tv);
   1732 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
   1733 		    SCM_TIMESTAMP, SOL_SOCKET);
   1734 		if (*mp)
   1735 			mp = &(*mp)->m_next;
   1736 	}
   1737 	if (inp->inp_flags & INP_RECVDSTADDR) {
   1738 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
   1739 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
   1740 		if (*mp)
   1741 			mp = &(*mp)->m_next;
   1742 	}
   1743 #ifdef notyet
   1744 	/*
   1745 	 * XXX
   1746 	 * Moving these out of udp_input() made them even more broken
   1747 	 * than they already were.
   1748 	 *	- fenner (at) parc.xerox.com
   1749 	 */
   1750 	/* options were tossed already */
   1751 	if (inp->inp_flags & INP_RECVOPTS) {
   1752 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
   1753 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
   1754 		if (*mp)
   1755 			mp = &(*mp)->m_next;
   1756 	}
   1757 	/* ip_srcroute doesn't do what we want here, need to fix */
   1758 	if (inp->inp_flags & INP_RECVRETOPTS) {
   1759 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
   1760 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
   1761 		if (*mp)
   1762 			mp = &(*mp)->m_next;
   1763 	}
   1764 #endif
   1765 	if (inp->inp_flags & INP_RECVIF) {
   1766 		struct sockaddr_dl sdl;
   1767 
   1768 		sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
   1769 		sdl.sdl_family = AF_LINK;
   1770 		sdl.sdl_index = m->m_pkthdr.rcvif ?
   1771 		    m->m_pkthdr.rcvif->if_index : 0;
   1772 		sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
   1773 		*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
   1774 		    IP_RECVIF, IPPROTO_IP);
   1775 		if (*mp)
   1776 			mp = &(*mp)->m_next;
   1777 	}
   1778 }
   1779 
   1780 int
   1781 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
   1782 	int *name;
   1783 	u_int namelen;
   1784 	void *oldp;
   1785 	size_t *oldlenp;
   1786 	void *newp;
   1787 	size_t newlen;
   1788 {
   1789 	extern int subnetsarelocal, hostzeroisbroadcast;
   1790 
   1791 	int error, old;
   1792 
   1793 	/* All sysctl names at this level are terminal. */
   1794 	if (namelen != 1)
   1795 		return (ENOTDIR);
   1796 
   1797 	switch (name[0]) {
   1798 	case IPCTL_FORWARDING:
   1799 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
   1800 	case IPCTL_SENDREDIRECTS:
   1801 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1802 			&ipsendredirects));
   1803 	case IPCTL_DEFTTL:
   1804 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
   1805 #ifdef notyet
   1806 	case IPCTL_DEFMTU:
   1807 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
   1808 #endif
   1809 	case IPCTL_FORWSRCRT:
   1810 		/* Don't allow this to change in a secure environment.  */
   1811 		if (securelevel > 0)
   1812 			return (sysctl_rdint(oldp, oldlenp, newp,
   1813 			    ip_forwsrcrt));
   1814 		else
   1815 			return (sysctl_int(oldp, oldlenp, newp, newlen,
   1816 			    &ip_forwsrcrt));
   1817 	case IPCTL_DIRECTEDBCAST:
   1818 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1819 		    &ip_directedbcast));
   1820 	case IPCTL_ALLOWSRCRT:
   1821 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1822 		    &ip_allowsrcrt));
   1823 	case IPCTL_SUBNETSARELOCAL:
   1824 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1825 		    &subnetsarelocal));
   1826 	case IPCTL_MTUDISC:
   1827 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1828 		    &ip_mtudisc);
   1829 		if (error == 0 && ip_mtudisc == 0)
   1830 			rt_timer_queue_remove_all(ip_mtudisc_timeout_q, TRUE);
   1831 		return error;
   1832 	case IPCTL_ANONPORTMIN:
   1833 		old = anonportmin;
   1834 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin);
   1835 		if (anonportmin >= anonportmax || anonportmin < 0
   1836 		    || anonportmin > 65535
   1837 #ifndef IPNOPRIVPORTS
   1838 		    || anonportmin < IPPORT_RESERVED
   1839 #endif
   1840 		    ) {
   1841 			anonportmin = old;
   1842 			return (EINVAL);
   1843 		}
   1844 		return (error);
   1845 	case IPCTL_ANONPORTMAX:
   1846 		old = anonportmax;
   1847 		error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax);
   1848 		if (anonportmin >= anonportmax || anonportmax < 0
   1849 		    || anonportmax > 65535
   1850 #ifndef IPNOPRIVPORTS
   1851 		    || anonportmax < IPPORT_RESERVED
   1852 #endif
   1853 		    ) {
   1854 			anonportmax = old;
   1855 			return (EINVAL);
   1856 		}
   1857 		return (error);
   1858 	case IPCTL_MTUDISCTIMEOUT:
   1859 		old = ip_mtudisc_timeout;
   1860 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1861 		   &ip_mtudisc_timeout);
   1862 		if (ip_mtudisc_timeout < 0) {
   1863 			ip_mtudisc_timeout = old;
   1864 			return (EINVAL);
   1865 		}
   1866 		if (error == 0)
   1867 			rt_timer_queue_change(ip_mtudisc_timeout_q,
   1868 					      ip_mtudisc_timeout);
   1869 		return (error);
   1870 #ifdef GATEWAY
   1871 	case IPCTL_MAXFLOWS:
   1872 	    {
   1873 		int s;
   1874 
   1875 		error = sysctl_int(oldp, oldlenp, newp, newlen,
   1876 		   &ip_maxflows);
   1877 		s = splsoftnet();
   1878 		ipflow_reap(0);
   1879 		splx(s);
   1880 		return (error);
   1881 	    }
   1882 #endif
   1883 	case IPCTL_HOSTZEROBROADCAST:
   1884 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1885 		    &hostzeroisbroadcast));
   1886 #if NGIF > 0
   1887 	case IPCTL_GIF_TTL:
   1888 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1889 				  &ip_gif_ttl));
   1890 #endif
   1891 
   1892 #if NGRE > 0
   1893 	case IPCTL_GRE_TTL:
   1894 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1895 				  &ip_gre_ttl));
   1896 #endif
   1897 
   1898 #ifndef IPNOPRIVPORTS
   1899 	case IPCTL_LOWPORTMIN:
   1900 		old = lowportmin;
   1901 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin);
   1902 		if (lowportmin >= lowportmax
   1903 		    || lowportmin > IPPORT_RESERVEDMAX
   1904 		    || lowportmin < IPPORT_RESERVEDMIN
   1905 		    ) {
   1906 			lowportmin = old;
   1907 			return (EINVAL);
   1908 		}
   1909 		return (error);
   1910 	case IPCTL_LOWPORTMAX:
   1911 		old = lowportmax;
   1912 		error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax);
   1913 		if (lowportmin >= lowportmax
   1914 		    || lowportmax > IPPORT_RESERVEDMAX
   1915 		    || lowportmax < IPPORT_RESERVEDMIN
   1916 		    ) {
   1917 			lowportmax = old;
   1918 			return (EINVAL);
   1919 		}
   1920 		return (error);
   1921 #endif
   1922 
   1923 	case IPCTL_MAXFRAGPACKETS:
   1924 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1925 		    &ip_maxfragpackets));
   1926 
   1927 	default:
   1928 		return (EOPNOTSUPP);
   1929 	}
   1930 	/* NOTREACHED */
   1931 }
   1932