Home | History | Annotate | Line # | Download | only in netinet
      1 /*	$NetBSD: ip_input.c,v 1.407 2026/05/29 02:52:06 ozaki-r 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  *
     49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     59  * POSSIBILITY OF SUCH DAMAGE.
     60  */
     61 
     62 /*
     63  * Copyright (c) 1982, 1986, 1988, 1993
     64  *	The Regents of the University of California.  All rights reserved.
     65  *
     66  * Redistribution and use in source and binary forms, with or without
     67  * modification, are permitted provided that the following conditions
     68  * are met:
     69  * 1. Redistributions of source code must retain the above copyright
     70  *    notice, this list of conditions and the following disclaimer.
     71  * 2. Redistributions in binary form must reproduce the above copyright
     72  *    notice, this list of conditions and the following disclaimer in the
     73  *    documentation and/or other materials provided with the distribution.
     74  * 3. Neither the name of the University nor the names of its contributors
     75  *    may be used to endorse or promote products derived from this software
     76  *    without specific prior written permission.
     77  *
     78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     88  * SUCH DAMAGE.
     89  *
     90  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
     91  */
     92 
     93 #include <sys/cdefs.h>
     94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.407 2026/05/29 02:52:06 ozaki-r Exp $");
     95 
     96 #ifdef _KERNEL_OPT
     97 #include "opt_inet.h"
     98 #include "opt_gateway.h"
     99 #include "opt_ipsec.h"
    100 #include "opt_mrouting.h"
    101 #include "opt_mbuftrace.h"
    102 #include "opt_inet_csum.h"
    103 #include "opt_net_mpsafe.h"
    104 #endif
    105 
    106 #include "arp.h"
    107 
    108 #include <sys/param.h>
    109 #include <sys/systm.h>
    110 #include <sys/cpu.h>
    111 #include <sys/mbuf.h>
    112 #include <sys/domain.h>
    113 #include <sys/protosw.h>
    114 #include <sys/socket.h>
    115 #include <sys/socketvar.h>
    116 #include <sys/errno.h>
    117 #include <sys/time.h>
    118 #include <sys/kernel.h>
    119 #include <sys/pool.h>
    120 #include <sys/sysctl.h>
    121 #include <sys/kauth.h>
    122 
    123 #include <net/if.h>
    124 #include <net/if_dl.h>
    125 #include <net/if_types.h>
    126 #include <net/route.h>
    127 #include <net/pktqueue.h>
    128 #include <net/pfil.h>
    129 
    130 #include <netinet/in.h>
    131 #include <netinet/in_systm.h>
    132 #include <netinet/ip.h>
    133 #include <netinet/in_pcb.h>
    134 #include <netinet/in_proto.h>
    135 #include <netinet/in_var.h>
    136 #include <netinet/ip_var.h>
    137 #include <netinet/ip_private.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 #include <netinet/portalgo.h>
    149 
    150 #ifdef IPSEC
    151 #include <netipsec/ipsec.h>
    152 #endif
    153 
    154 #ifndef	IPFORWARDING
    155 #ifdef GATEWAY
    156 #define	IPFORWARDING	1	/* forward IP packets not for us */
    157 #else
    158 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
    159 #endif
    160 #endif
    161 
    162 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
    163 
    164 int ipforwarding = IPFORWARDING;
    165 int ipsendredirects = 1;
    166 int ip_defttl = IPDEFTTL;
    167 int ip_forwsrcrt = 0;
    168 int ip_directedbcast = 0;
    169 int ip_allowsrcrt = 0;
    170 int ip_mtudisc = 1;
    171 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
    172 int ip_do_randomid = 1;
    173 
    174 /*
    175  * XXX - Setting ip_checkinterface mostly implements the receive side of
    176  * the Strong ES model described in RFC 1122, but since the routing table
    177  * and transmit implementation do not implement the Strong ES model,
    178  * setting this to 1 results in an odd hybrid.
    179  *
    180  * XXX - ip_checkinterface currently must be disabled if you use NAT
    181  * to translate the destination address to another local interface.
    182  *
    183  * XXX - ip_checkinterface must be disabled if you add IP aliases
    184  * to the loopback interface instead of the interface where the
    185  * packets for those addresses are received.
    186  */
    187 static int		ip_checkinterface	__read_mostly = 0;
    188 
    189 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
    190 
    191 pktqueue_t *		ip_pktq			__read_mostly;
    192 pfil_head_t *		inet_pfil_hook		__read_mostly;
    193 percpu_t *		ipstat_percpu		__read_mostly;
    194 
    195 static percpu_t		*ipforward_rt_percpu	__cacheline_aligned;
    196 
    197 uint16_t ip_id;
    198 
    199 #ifdef INET_CSUM_COUNTERS
    200 #include <sys/device.h>
    201 
    202 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    203     NULL, "inet", "hwcsum bad");
    204 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    205     NULL, "inet", "hwcsum ok");
    206 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    207     NULL, "inet", "swcsum");
    208 
    209 #define	INET_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    210 
    211 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
    212 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
    213 EVCNT_ATTACH_STATIC(ip_swcsum);
    214 
    215 #else
    216 
    217 #define	INET_CSUM_COUNTER_INCR(ev)	/* nothing */
    218 
    219 #endif /* INET_CSUM_COUNTERS */
    220 
    221 /*
    222  * Used to save the IP options in case a protocol wants to respond
    223  * to an incoming packet over the same route if the packet got here
    224  * using IP source routing.  This allows connection establishment and
    225  * maintenance when the remote end is on a network that is not known
    226  * to us.
    227  */
    228 struct ip_srcrt {
    229 	int		isr_nhops;		   /* number of hops */
    230 	struct in_addr	isr_dst;		   /* final destination */
    231 	char		isr_nop;		   /* one NOP to align */
    232 	char		isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
    233 	struct in_addr	isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
    234 };
    235 
    236 static int ip_drainwanted;
    237 
    238 static void save_rte(struct mbuf *, u_char *, struct in_addr);
    239 
    240 #ifdef MBUFTRACE
    241 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
    242 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
    243 #endif
    244 
    245 static void		ipintr(void *);
    246 static void		ip_input(struct mbuf *, struct ifnet *);
    247 static void		ip_forward(struct mbuf *, int, struct ifnet *);
    248 static bool		ip_dooptions(struct mbuf *);
    249 static int		ip_rtaddr(struct in_addr, struct in_addr *);
    250 static void		sysctl_net_inet_ip_setup(struct sysctllog **);
    251 
    252 static struct in_ifaddr	*ip_match_our_address(struct ifnet *, struct ip *,
    253 			    int *);
    254 static struct in_ifaddr	*ip_match_our_address_broadcast(struct ifnet *,
    255 			    struct ip *);
    256 
    257 #ifdef NET_MPSAFE
    258 #define	SOFTNET_LOCK()		mutex_enter(softnet_lock)
    259 #define	SOFTNET_UNLOCK()	mutex_exit(softnet_lock)
    260 #else
    261 #define	SOFTNET_LOCK()		KASSERT(mutex_owned(softnet_lock))
    262 #define	SOFTNET_UNLOCK()	KASSERT(mutex_owned(softnet_lock))
    263 #endif
    264 
    265 /*
    266  * IP initialization: fill in IP protocol switch table.
    267  * All protocols not implemented in kernel go to raw IP protocol handler.
    268  */
    269 void
    270 ip_init(void)
    271 {
    272 	const struct protosw *pr;
    273 
    274 	ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
    275 	KASSERT(ip_pktq != NULL);
    276 
    277 	in_init();
    278 	sysctl_net_inet_ip_setup(NULL);
    279 
    280 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
    281 	KASSERT(pr != NULL);
    282 
    283 	for (u_int i = 0; i < IPPROTO_MAX; i++) {
    284 		ip_protox[i] = pr - inetsw;
    285 	}
    286 	for (pr = inetdomain.dom_protosw;
    287 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
    288 		if (pr->pr_domain->dom_family == PF_INET &&
    289 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
    290 			ip_protox[pr->pr_protocol] = pr - inetsw;
    291 
    292 	ip_reass_init();
    293 
    294 	ip_id = time_uptime & 0xfffff;
    295 
    296 #ifdef GATEWAY
    297 	ipflow_init();
    298 #endif
    299 
    300 	/* Register our Packet Filter hook. */
    301 	inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
    302 	KASSERT(inet_pfil_hook != NULL);
    303 
    304 #ifdef MBUFTRACE
    305 	MOWNER_ATTACH(&ip_tx_mowner);
    306 	MOWNER_ATTACH(&ip_rx_mowner);
    307 #endif
    308 
    309 	ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
    310 	ipforward_rt_percpu = rtcache_percpu_alloc();
    311 	ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
    312 }
    313 
    314 static struct in_ifaddr *
    315 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
    316 {
    317 	struct in_ifaddr *ia = NULL;
    318 	int checkif;
    319 
    320 	/*
    321 	 * Enable a consistency check between the destination address
    322 	 * and the arrival interface for a unicast packet (the RFC 1122
    323 	 * strong ES model) if IP forwarding is disabled and the packet
    324 	 * is not locally generated.
    325 	 *
    326 	 * XXX - We need to add a per ifaddr flag for this so that
    327 	 * we get finer grain control.
    328 	 */
    329 	checkif = ip_checkinterface && (ipforwarding == 0) &&
    330 	    (ifp->if_flags & IFF_LOOPBACK) == 0;
    331 
    332 	IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
    333 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
    334 			if (ia->ia4_flags & IN_IFF_NOTREADY)
    335 				continue;
    336 			if (checkif && ia->ia_ifp != ifp)
    337 				continue;
    338 			if ((ia->ia_ifp->if_flags & IFF_UP) == 0) {
    339 				(*downmatch)++;
    340 				continue;
    341 			}
    342 			if (ia->ia4_flags & IN_IFF_DETACHED &&
    343 			    (ifp->if_flags & IFF_LOOPBACK) == 0)
    344 				continue;
    345 			break;
    346 		}
    347 	}
    348 
    349 	return ia;
    350 }
    351 
    352 static struct in_ifaddr *
    353 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
    354 {
    355 	struct in_ifaddr *ia = NULL;
    356 	struct ifaddr *ifa;
    357 
    358 	IFADDR_READER_FOREACH(ifa, ifp) {
    359 		if (ifa->ifa_addr->sa_family != AF_INET)
    360 			continue;
    361 		ia = ifatoia(ifa);
    362 		if (ia->ia4_flags & IN_IFF_NOTREADY)
    363 			continue;
    364 		if (ia->ia4_flags & IN_IFF_DETACHED &&
    365 		    (ifp->if_flags & IFF_LOOPBACK) == 0)
    366 			continue;
    367 		if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
    368 		    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
    369 		    /*
    370 		     * Look for all-0's host part (old broadcast addr),
    371 		     * either for subnet or net.
    372 		     */
    373 		    ip->ip_dst.s_addr == ia->ia_subnet ||
    374 		    ip->ip_dst.s_addr == ia->ia_net)
    375 			goto matched;
    376 		/*
    377 		 * An interface with IP address zero accepts
    378 		 * all packets that arrive on that interface.
    379 		 */
    380 		if (in_nullhost(ia->ia_addr.sin_addr))
    381 			goto matched;
    382 	}
    383 	ia = NULL;
    384 
    385 matched:
    386 	return ia;
    387 }
    388 
    389 /*
    390  * IP software interrupt routine.
    391  */
    392 static void
    393 ipintr(void *arg __unused)
    394 {
    395 	struct mbuf *m;
    396 
    397 	KASSERT(cpu_softintr_p());
    398 
    399 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    400 	while ((m = pktq_dequeue(ip_pktq)) != NULL) {
    401 		struct ifnet *ifp;
    402 		struct psref psref;
    403 
    404 		ifp = m_get_rcvif_psref(m, &psref);
    405 		if (__predict_false(ifp == NULL)) {
    406 			IP_STATINC(IP_STAT_IFDROP);
    407 			m_freem(m);
    408 			continue;
    409 		}
    410 
    411 		ip_input(m, ifp);
    412 
    413 		m_put_rcvif_psref(ifp, &psref);
    414 	}
    415 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    416 }
    417 
    418 /*
    419  * IP input routine.  Checksum and byte swap header.  If fragmented
    420  * try to reassemble.  Process options.  Pass to next level.
    421  */
    422 static void
    423 ip_input(struct mbuf *m, struct ifnet *ifp)
    424 {
    425 	struct ip *ip = NULL;
    426 	struct in_ifaddr *ia = NULL;
    427 	int hlen = 0, len;
    428 	int downmatch;
    429 	int srcrt = 0;
    430 	int s;
    431 
    432 	KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
    433 	    "interrupt handler; synchronization assumptions violated");
    434 
    435 	MCLAIM(m, &ip_rx_mowner);
    436 	KASSERT((m->m_flags & M_PKTHDR) != 0);
    437 
    438 	/*
    439 	 * If no IP addresses have been set yet but the interfaces
    440 	 * are receiving, can't do anything with incoming packets yet.
    441 	 * Note: we pre-check without locks held.
    442 	 */
    443 	if (IN_ADDRLIST_READER_EMPTY()) {
    444 		IP_STATINC(IP_STAT_IFDROP);
    445 		goto out;
    446 	}
    447 
    448 	IP_STATINC(IP_STAT_TOTAL);
    449 
    450 	/*
    451 	 * If the IP header is not aligned, slurp it up into a new
    452 	 * mbuf with space for link headers, in the event we forward
    453 	 * it.  Otherwise, if it is aligned, make sure the entire
    454 	 * base IP header is in the first mbuf of the chain.
    455 	 */
    456 	if (M_GET_ALIGNED_HDR(&m, struct ip, true) != 0) {
    457 		/* XXXJRT new stat, please */
    458 		IP_STATINC(IP_STAT_TOOSMALL);
    459 		goto out;
    460 	}
    461 	ip = mtod(m, struct ip *);
    462 	if (ip->ip_v != IPVERSION) {
    463 		IP_STATINC(IP_STAT_BADVERS);
    464 		goto out;
    465 	}
    466 	hlen = ip->ip_hl << 2;
    467 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
    468 		IP_STATINC(IP_STAT_BADHLEN);
    469 		goto out;
    470 	}
    471 	if (hlen > m->m_len) {
    472 		if ((m = m_pullup(m, hlen)) == NULL) {
    473 			IP_STATINC(IP_STAT_BADHLEN);
    474 			goto out;
    475 		}
    476 		ip = mtod(m, struct ip *);
    477 	}
    478 
    479 	/*
    480 	 * RFC1122: packets with a multicast source address are
    481 	 * not allowed.
    482 	 */
    483 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    484 		IP_STATINC(IP_STAT_BADADDR);
    485 		goto out;
    486 	}
    487 
    488 	/* 127/8 must not appear on wire - RFC1122 */
    489 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
    490 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
    491 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
    492 			IP_STATINC(IP_STAT_BADADDR);
    493 			goto out;
    494 		}
    495 	}
    496 
    497 	switch (m->m_pkthdr.csum_flags &
    498 		((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) {
    499 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
    500 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
    501 		IP_STATINC(IP_STAT_BADSUM);
    502 		goto out;
    503 
    504 	case M_CSUM_IPv4:
    505 		/* Checksum was okay. */
    506 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
    507 		break;
    508 
    509 	default:
    510 		/*
    511 		 * Must compute it ourselves.  Maybe skip checksum on
    512 		 * loopback interfaces.
    513 		 */
    514 		if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
    515 		    ip_do_loopback_cksum)) {
    516 			INET_CSUM_COUNTER_INCR(&ip_swcsum);
    517 			if (in_cksum(m, hlen) != 0) {
    518 				IP_STATINC(IP_STAT_BADSUM);
    519 				goto out;
    520 			}
    521 		}
    522 		break;
    523 	}
    524 
    525 	/* Retrieve the packet length. */
    526 	len = ntohs(ip->ip_len);
    527 
    528 	/*
    529 	 * Check for additional length bogosity
    530 	 */
    531 	if (len < hlen) {
    532 		IP_STATINC(IP_STAT_BADLEN);
    533 		goto out;
    534 	}
    535 
    536 	/*
    537 	 * Check that the amount of data in the buffers is at least as much
    538 	 * as the IP header would have us expect. Trim mbufs if longer than
    539 	 * we expect. Drop packet if shorter than we expect.
    540 	 */
    541 	if (m->m_pkthdr.len < len) {
    542 		IP_STATINC(IP_STAT_TOOSHORT);
    543 		goto out;
    544 	}
    545 	if (m->m_pkthdr.len > len) {
    546 		if (m->m_len == m->m_pkthdr.len) {
    547 			m->m_len = len;
    548 			m->m_pkthdr.len = len;
    549 		} else
    550 			m_adj(m, len - m->m_pkthdr.len);
    551 	}
    552 
    553 	/*
    554 	 * Assume that we can create a fast-forward IP flow entry
    555 	 * based on this packet.
    556 	 */
    557 	m->m_flags |= M_CANFASTFWD;
    558 
    559 	/*
    560 	 * Run through list of hooks for input packets.  If there are any
    561 	 * filters which require that additional packets in the flow are
    562 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
    563 	 * Note that filters must _never_ set this flag, as another filter
    564 	 * in the list may have previously cleared it.
    565 	 *
    566 	 * Don't call hooks if the packet has already been processed by
    567 	 * IPsec (encapsulated, tunnel mode).
    568 	 */
    569 #if defined(IPSEC)
    570 	if (!ipsec_used || !ipsec_skip_pfil(m))
    571 #else
    572 	if (1)
    573 #endif
    574 	{
    575 		struct in_addr odst = ip->ip_dst;
    576 		bool freed;
    577 
    578 		freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
    579 		if (freed || m == NULL) {
    580 			m = NULL;
    581 			IP_STATINC(IP_STAT_PFILDROP_IN);
    582 			goto out;
    583 		}
    584 		if (__predict_false(m->m_len < sizeof(struct ip))) {
    585 			if ((m = m_pullup(m, sizeof(struct ip))) == NULL) {
    586 				IP_STATINC(IP_STAT_TOOSMALL);
    587 				goto out;
    588 			}
    589 		}
    590 		ip = mtod(m, struct ip *);
    591 		hlen = ip->ip_hl << 2;
    592 		if (hlen < sizeof(struct ip)) {	/* minimum header length */
    593 			IP_STATINC(IP_STAT_BADHLEN);
    594 			goto out;
    595 		}
    596 		if (hlen > m->m_len) {
    597 			if ((m = m_pullup(m, hlen)) == NULL) {
    598 				IP_STATINC(IP_STAT_BADHLEN);
    599 				goto out;
    600 			}
    601 			ip = mtod(m, struct ip *);
    602 		}
    603 
    604 		/*
    605 		 * XXX The setting of "srcrt" here is to prevent ip_forward()
    606 		 * from generating ICMP redirects for packets that have
    607 		 * been redirected by a hook back out on to the same LAN that
    608 		 * they came from and is not an indication that the packet
    609 		 * is being influenced by source routing options.  This
    610 		 * allows things like
    611 		 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
    612 		 * where tlp0 is both on the 1.1.1.0/24 network and is the
    613 		 * default route for hosts on 1.1.1.0/24.  Of course this
    614 		 * also requires a "map tlp0 ..." to complete the story.
    615 		 * One might argue whether or not this kind of network config.
    616 		 * should be supported in this manner...
    617 		 */
    618 		srcrt = (odst.s_addr != ip->ip_dst.s_addr);
    619 	}
    620 
    621 #ifdef ALTQ
    622 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
    623 	if (altq_input) {
    624 		SOFTNET_LOCK();
    625 		if ((*altq_input)(m, AF_INET) == 0) {
    626 			/* Packet dropped by traffic conditioner. */
    627 			SOFTNET_UNLOCK();
    628 			m = NULL;
    629 			goto out;
    630 		}
    631 		SOFTNET_UNLOCK();
    632 	}
    633 #endif
    634 
    635 	/*
    636 	 * Process options and, if not destined for us,
    637 	 * ship it on.  ip_dooptions returns 1 when an
    638 	 * error was detected (causing an icmp message
    639 	 * to be sent and the original packet to be freed).
    640 	 */
    641 	if (hlen > sizeof(struct ip) && ip_dooptions(m)) {
    642 		m = NULL;
    643 		goto out;
    644 	}
    645 
    646 	/*
    647 	 * Check our list of addresses, to see if the packet is for us.
    648 	 *
    649 	 * Traditional 4.4BSD did not consult IFF_UP at all.
    650 	 * The behavior here is to treat addresses on !IFF_UP interface
    651 	 * or IN_IFF_NOTREADY addresses as not mine.
    652 	 */
    653 	downmatch = 0;
    654 	s = pserialize_read_enter();
    655 	ia = ip_match_our_address(ifp, ip, &downmatch);
    656 	if (ia != NULL) {
    657 		pserialize_read_exit(s);
    658 		goto ours;
    659 	}
    660 
    661 	if (ifp->if_flags & IFF_BROADCAST) {
    662 		ia = ip_match_our_address_broadcast(ifp, ip);
    663 		if (ia != NULL) {
    664 			pserialize_read_exit(s);
    665 			goto ours;
    666 		}
    667 	}
    668 	pserialize_read_exit(s);
    669 
    670 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
    671 #ifdef MROUTING
    672 		extern struct socket *ip_mrouter;
    673 
    674 		if (ip_mrouter) {
    675 			/*
    676 			 * If we are acting as a multicast router, all
    677 			 * incoming multicast packets are passed to the
    678 			 * kernel-level multicast forwarding function.
    679 			 * The packet is returned (relatively) intact; if
    680 			 * ip_mforward() returns a non-zero value, the packet
    681 			 * must be discarded, else it may be accepted below.
    682 			 *
    683 			 * (The IP ident field is put in the same byte order
    684 			 * as expected when ip_mforward() is called from
    685 			 * ip_output().)
    686 			 */
    687 			SOFTNET_LOCK();
    688 			if (ip_mforward(m, ifp) != 0) {
    689 				SOFTNET_UNLOCK();
    690 				IP_STATINC(IP_STAT_CANTFORWARD);
    691 				goto out;
    692 			}
    693 			SOFTNET_UNLOCK();
    694 
    695 			/*
    696 			 * The process-level routing demon needs to receive
    697 			 * all multicast IGMP packets, whether or not this
    698 			 * host belongs to their destination groups.
    699 			 */
    700 			if (ip->ip_p == IPPROTO_IGMP) {
    701 				goto ours;
    702 			}
    703 			IP_STATINC(IP_STAT_CANTFORWARD);
    704 		}
    705 #endif
    706 		/*
    707 		 * See if we belong to the destination multicast group on the
    708 		 * arrival interface.
    709 		 */
    710 		if (!in_multi_group(ip->ip_dst, ifp, 0)) {
    711 			IP_STATINC(IP_STAT_CANTFORWARD);
    712 			goto out;
    713 		}
    714 		goto ours;
    715 	}
    716 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    717 	    in_nullhost(ip->ip_dst))
    718 		goto ours;
    719 
    720 	/*
    721 	 * Not for us; forward if possible and desirable.
    722 	 */
    723 	if (ipforwarding == 0) {
    724 		IP_STATINC(IP_STAT_CANTFORWARD);
    725 		m_freem(m);
    726 		return;
    727 	}
    728 
    729 	/*
    730 	 * If ip_dst matched any of my address on !IFF_UP interface,
    731 	 * and there's no IFF_UP interface that matches ip_dst,
    732 	 * send icmp unreach.  Forwarding it will result in in-kernel
    733 	 * forwarding loop till TTL goes to 0.
    734 	 */
    735 	if (downmatch) {
    736 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
    737 		IP_STATINC(IP_STAT_CANTFORWARD);
    738 		return;
    739 	}
    740 #ifdef IPSEC
    741 	/* Check the security policy (SP) for the packet */
    742 	if (ipsec_used) {
    743 		if (ipsec_ip_input_checkpolicy(m, true) != 0) {
    744 			IP_STATINC(IP_STAT_IPSECDROP_IN);
    745 			goto out;
    746 		}
    747 	}
    748 #endif
    749 	ip_forward(m, srcrt, ifp);
    750 	return;
    751 
    752 ours:
    753 	/*
    754 	 * If offset or IP_MF are set, must reassemble.
    755 	 */
    756 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
    757 		/*
    758 		 * Pass to IP reassembly mechanism.
    759 		 */
    760 		if (ip_reass_packet(&m) != 0) {
    761 			/* Failed; invalid fragment(s) or packet. */
    762 			goto out;
    763 		}
    764 		if (m == NULL) {
    765 			/* More fragments should come; silently return. */
    766 			goto out;
    767 		}
    768 		/*
    769 		 * Reassembly is done, we have the final packet.
    770 		 * Update cached data in local variable(s).
    771 		 */
    772 		ip = mtod(m, struct ip *);
    773 		hlen = ip->ip_hl << 2;
    774 	}
    775 
    776 	M_VERIFY_PACKET(m);
    777 
    778 #ifdef IPSEC
    779 	/*
    780 	 * Enforce IPsec policy checking if we are seeing last header.
    781 	 * Note that we do not visit this with protocols with PCB layer
    782 	 * code - like UDP/TCP/raw IP.
    783 	 */
    784 	if (ipsec_used &&
    785 	    (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
    786 		if (ipsec_ip_input_checkpolicy(m, false) != 0) {
    787 			IP_STATINC(IP_STAT_IPSECDROP_IN);
    788 			goto out;
    789 		}
    790 	}
    791 #endif
    792 
    793 	/*
    794 	 * Switch out to protocol's input routine.
    795 	 */
    796 #if IFA_STATS
    797 	if (ia) {
    798 		struct in_ifaddr *_ia;
    799 		/*
    800 		 * Keep a reference from ip_match_our_address with psref
    801 		 * is expensive, so explore ia here again.
    802 		 */
    803 		s = pserialize_read_enter();
    804 		_ia = in_get_ia(ip->ip_dst);
    805 		_ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
    806 		pserialize_read_exit(s);
    807 	}
    808 #endif
    809 	IP_STATINC(IP_STAT_DELIVERED);
    810 
    811 	const int off = hlen, nh = ip->ip_p;
    812 
    813 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
    814 	return;
    815 
    816 out:
    817 	m_freem(m);
    818 }
    819 
    820 /*
    821  * IP timer processing.
    822  */
    823 void
    824 ip_slowtimo(void)
    825 {
    826 
    827 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    828 
    829 	ip_reass_slowtimo();
    830 
    831 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    832 }
    833 
    834 /*
    835  * IP drain processing.
    836  */
    837 void
    838 ip_drain(void)
    839 {
    840 
    841 	KERNEL_LOCK(1, NULL);
    842 	ip_reass_drain();
    843 	KERNEL_UNLOCK_ONE(NULL);
    844 }
    845 
    846 /*
    847  * ip_dooptions: perform option processing on a datagram, possibly discarding
    848  * it if bad options are encountered, or forwarding it if source-routed.
    849  *
    850  * => Returns true if packet has been forwarded/freed.
    851  * => Returns false if the packet should be processed further.
    852  */
    853 static bool
    854 ip_dooptions(struct mbuf *m)
    855 {
    856 	struct ip *ip = mtod(m, struct ip *);
    857 	u_char *cp, *cp0;
    858 	struct ip_timestamp *ipt;
    859 	struct in_ifaddr *ia;
    860 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
    861 	int srr_present, rr_present, ts_present;
    862 	struct in_addr dst;
    863 	n_time ntime;
    864 	struct ifaddr *ifa = NULL;
    865 	int s;
    866 
    867 	srr_present = 0;
    868 	rr_present = 0;
    869 	ts_present = 0;
    870 
    871 	dst = ip->ip_dst;
    872 	cp = (u_char *)(ip + 1);
    873 	cnt = (ip->ip_hl << 2) - sizeof(struct ip);
    874 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
    875 		opt = cp[IPOPT_OPTVAL];
    876 		if (opt == IPOPT_EOL)
    877 			break;
    878 		if (opt == IPOPT_NOP)
    879 			optlen = 1;
    880 		else {
    881 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
    882 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    883 				goto bad;
    884 			}
    885 			optlen = cp[IPOPT_OLEN];
    886 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
    887 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    888 				goto bad;
    889 			}
    890 		}
    891 		switch (opt) {
    892 
    893 		default:
    894 			break;
    895 
    896 		/*
    897 		 * Source routing with record.
    898 		 * Find interface with current destination address.
    899 		 * If none on this machine then drop if strictly routed,
    900 		 * or do nothing if loosely routed.
    901 		 * Record interface address and bring up next address
    902 		 * component.  If strictly routed make sure next
    903 		 * address is on directly accessible net.
    904 		 */
    905 		case IPOPT_LSRR:
    906 		case IPOPT_SSRR: {
    907 			struct sockaddr_in ipaddr = {
    908 			    .sin_len = sizeof(ipaddr),
    909 			    .sin_family = AF_INET,
    910 			};
    911 
    912 			if (ip_allowsrcrt == 0) {
    913 				type = ICMP_UNREACH;
    914 				code = ICMP_UNREACH_NET_PROHIB;
    915 				goto bad;
    916 			}
    917 			if (srr_present++) {
    918 				code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
    919 				goto bad;
    920 			}
    921 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
    922 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    923 				goto bad;
    924 			}
    925 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
    926 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
    927 				goto bad;
    928 			}
    929 			ipaddr.sin_addr = ip->ip_dst;
    930 
    931 			s = pserialize_read_enter();
    932 			ifa = ifa_ifwithaddr(sintosa(&ipaddr));
    933 			if (ifa == NULL) {
    934 				pserialize_read_exit(s);
    935 				if (opt == IPOPT_SSRR) {
    936 					type = ICMP_UNREACH;
    937 					code = ICMP_UNREACH_SRCFAIL;
    938 					goto bad;
    939 				}
    940 				/*
    941 				 * Loose routing, and not at next destination
    942 				 * yet; nothing to do except forward.
    943 				 */
    944 				break;
    945 			}
    946 			pserialize_read_exit(s);
    947 
    948 			off--;			/* 0 origin */
    949 			if ((off + sizeof(struct in_addr)) > optlen) {
    950 				/*
    951 				 * End of source route.  Should be for us.
    952 				 */
    953 				save_rte(m, cp, ip->ip_src);
    954 				break;
    955 			}
    956 			/*
    957 			 * locate outgoing interface
    958 			 */
    959 			memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
    960 			    sizeof(ipaddr.sin_addr));
    961 			int error = -1;
    962 			if (opt == IPOPT_SSRR) {
    963 				s = pserialize_read_enter();
    964 				ifa = ifa_ifwithladdr(sintosa(&ipaddr));
    965 				if (ifa != NULL) {
    966 					ia = ifatoia(ifa);
    967 					memcpy(cp + off, &ia->ia_addr.sin_addr,
    968 					    sizeof(struct in_addr));
    969 					error = 0;
    970 				}
    971 				pserialize_read_exit(s);
    972 			} else {
    973 				struct in_addr addr;
    974 				error = ip_rtaddr(ipaddr.sin_addr, &addr);
    975 				if (error == 0)
    976 					memcpy(cp + off, &addr, sizeof(addr));
    977 			}
    978 			if (error != 0) {
    979 				type = ICMP_UNREACH;
    980 				code = ICMP_UNREACH_SRCFAIL;
    981 				goto bad;
    982 			}
    983 			ip->ip_dst = ipaddr.sin_addr;
    984 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
    985 			/*
    986 			 * Let ip_intr's mcast routing check handle mcast pkts
    987 			 */
    988 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
    989 			break;
    990 		    }
    991 
    992 		case IPOPT_RR: {
    993 			struct sockaddr_in ipaddr = {
    994 			    .sin_len = sizeof(ipaddr),
    995 			    .sin_family = AF_INET,
    996 			};
    997 
    998 			if (rr_present++) {
    999 				code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
   1000 				goto bad;
   1001 			}
   1002 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1003 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1004 				goto bad;
   1005 			}
   1006 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1007 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1008 				goto bad;
   1009 			}
   1010 			/*
   1011 			 * If no space remains, ignore.
   1012 			 */
   1013 			off--;			/* 0 origin */
   1014 			if ((off + sizeof(struct in_addr)) > optlen)
   1015 				break;
   1016 			memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst,
   1017 			    sizeof(ipaddr.sin_addr));
   1018 			/*
   1019 			 * locate outgoing interface; if we're the destination,
   1020 			 * use the incoming interface (should be same).
   1021 			 */
   1022 			s = pserialize_read_enter();
   1023 			ifa = ifa_ifwithaddr(sintosa(&ipaddr));
   1024 			if (ifa != NULL) {
   1025 				ia = ifatoia(ifa);
   1026 				memcpy(cp + off, &ia->ia_addr.sin_addr,
   1027 				    sizeof(struct in_addr));
   1028 				pserialize_read_exit(s);
   1029 			} else {
   1030 				struct in_addr addr;
   1031 				int error;
   1032 				pserialize_read_exit(s);
   1033 
   1034 				error = ip_rtaddr(ipaddr.sin_addr, &addr);
   1035 				if (error != 0) {
   1036 					type = ICMP_UNREACH;
   1037 					code = ICMP_UNREACH_HOST;
   1038 					goto bad;
   1039 				}
   1040 				memcpy(cp + off, &addr, sizeof(addr));
   1041 			}
   1042 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1043 			break;
   1044 		    }
   1045 
   1046 		case IPOPT_TS:
   1047 			code = cp - (u_char *)ip;
   1048 			ipt = (struct ip_timestamp *)cp;
   1049 			if (ts_present++) {
   1050 				code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
   1051 				goto bad;
   1052 			}
   1053 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
   1054 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
   1055 				goto bad;
   1056 			}
   1057 			if (ipt->ipt_ptr < 5) {
   1058 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
   1059 				goto bad;
   1060 			}
   1061 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) {
   1062 				if (++ipt->ipt_oflw == 0) {
   1063 					code = (u_char *)&ipt->ipt_ptr -
   1064 					    (u_char *)ip;
   1065 					goto bad;
   1066 				}
   1067 				break;
   1068 			}
   1069 			cp0 = (cp + ipt->ipt_ptr - 1);
   1070 			switch (ipt->ipt_flg) {
   1071 
   1072 			case IPOPT_TS_TSONLY:
   1073 				break;
   1074 
   1075 			case IPOPT_TS_TSANDADDR: {
   1076 				struct ifnet *rcvif;
   1077 				int _s, _ss;
   1078 				struct sockaddr_in ipaddr = {
   1079 				    .sin_len = sizeof(ipaddr),
   1080 				    .sin_family = AF_INET,
   1081 				};
   1082 
   1083 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1084 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1085 					code = (u_char *)&ipt->ipt_ptr -
   1086 					    (u_char *)ip;
   1087 					goto bad;
   1088 				}
   1089 				ipaddr.sin_addr = dst;
   1090 				_ss = pserialize_read_enter();
   1091 				rcvif = m_get_rcvif(m, &_s);
   1092 				if (__predict_true(rcvif != NULL)) {
   1093 					ifa = ifaof_ifpforaddr(sintosa(&ipaddr),
   1094 					    rcvif);
   1095 				}
   1096 				m_put_rcvif(rcvif, &_s);
   1097 				if (ifa == NULL) {
   1098 					pserialize_read_exit(_ss);
   1099 					break;
   1100 				}
   1101 				ia = ifatoia(ifa);
   1102 				memcpy(cp0, &ia->ia_addr.sin_addr,
   1103 				    sizeof(struct in_addr));
   1104 				pserialize_read_exit(_ss);
   1105 				ipt->ipt_ptr += sizeof(struct in_addr);
   1106 				break;
   1107 			}
   1108 
   1109 			case IPOPT_TS_PRESPEC: {
   1110 				struct sockaddr_in ipaddr = {
   1111 				    .sin_len = sizeof(ipaddr),
   1112 				    .sin_family = AF_INET,
   1113 				};
   1114 
   1115 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1116 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1117 					code = (u_char *)&ipt->ipt_ptr -
   1118 					    (u_char *)ip;
   1119 					goto bad;
   1120 				}
   1121 				memcpy(&ipaddr.sin_addr, cp0,
   1122 				    sizeof(struct in_addr));
   1123 				s = pserialize_read_enter();
   1124 				ifa = ifa_ifwithaddr(sintosa(&ipaddr));
   1125 				if (ifa == NULL) {
   1126 					pserialize_read_exit(s);
   1127 					continue;
   1128 				}
   1129 				pserialize_read_exit(s);
   1130 				ipt->ipt_ptr += sizeof(struct in_addr);
   1131 				break;
   1132 			    }
   1133 
   1134 			default:
   1135 				/* XXX can't take &ipt->ipt_flg */
   1136 				code = (u_char *)&ipt->ipt_ptr -
   1137 				    (u_char *)ip + 1;
   1138 				goto bad;
   1139 			}
   1140 			ntime = iptime();
   1141 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
   1142 			memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
   1143 			    sizeof(n_time));
   1144 			ipt->ipt_ptr += sizeof(n_time);
   1145 		}
   1146 	}
   1147 	if (forward) {
   1148 		struct ifnet *rcvif;
   1149 		struct psref _psref;
   1150 
   1151 		if (ip_forwsrcrt == 0) {
   1152 			type = ICMP_UNREACH;
   1153 			code = ICMP_UNREACH_SRCFAIL;
   1154 			goto bad;
   1155 		}
   1156 
   1157 		rcvif = m_get_rcvif_psref(m, &_psref);
   1158 		if (__predict_false(rcvif == NULL)) {
   1159 			type = ICMP_UNREACH;
   1160 			code = ICMP_UNREACH_HOST;
   1161 			goto bad;
   1162 		}
   1163 		ip_forward(m, 1, rcvif);
   1164 		m_put_rcvif_psref(rcvif, &_psref);
   1165 		return true;
   1166 	}
   1167 	return false;
   1168 bad:
   1169 	icmp_error(m, type, code, 0, 0);
   1170 	IP_STATINC(IP_STAT_BADOPTIONS);
   1171 	return true;
   1172 }
   1173 
   1174 /*
   1175  * ip_rtaddr: given address of next destination (final or next hop),
   1176  * return internet address of interface to be used to get there.
   1177  */
   1178 static int
   1179 ip_rtaddr(struct in_addr dst, struct in_addr *ret)
   1180 {
   1181 	struct rtentry *rt;
   1182 	union {
   1183 		struct sockaddr		dst;
   1184 		struct sockaddr_in	dst4;
   1185 	} u;
   1186 	struct route *ro;
   1187 	struct in_ifaddr *ia;
   1188 
   1189 	sockaddr_in_init(&u.dst4, &dst, 0);
   1190 
   1191 	ro = rtcache_percpu_getref(ipforward_rt_percpu);
   1192 	rt = rtcache_lookup(ro, &u.dst);
   1193 	if (rt == NULL) {
   1194 		rtcache_percpu_putref(ipforward_rt_percpu);
   1195 		return -1;
   1196 	}
   1197 
   1198 	ia = ifatoia(rt->rt_ifa);
   1199 	*ret = ia->ia_addr.sin_addr;
   1200 	rtcache_unref(rt, ro);
   1201 	rtcache_percpu_putref(ipforward_rt_percpu);
   1202 
   1203 	return 0;
   1204 }
   1205 
   1206 /*
   1207  * save_rte: save incoming source route for use in replies, to be picked
   1208  * up later by ip_srcroute if the receiver is interested.
   1209  */
   1210 static void
   1211 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
   1212 {
   1213 	struct ip_srcrt *isr;
   1214 	struct m_tag *mtag;
   1215 	unsigned olen;
   1216 
   1217 	olen = option[IPOPT_OLEN];
   1218 	if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
   1219 		return;
   1220 
   1221 	mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
   1222 	if (mtag == NULL)
   1223 		return;
   1224 	isr = (struct ip_srcrt *)(mtag + 1);
   1225 
   1226 	memcpy(isr->isr_hdr, option, olen);
   1227 	isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
   1228 	isr->isr_dst = dst;
   1229 	m_tag_prepend(m, mtag);
   1230 }
   1231 
   1232 /*
   1233  * Retrieve incoming source route for use in replies,
   1234  * in the same form used by setsockopt.
   1235  * The first hop is placed before the options, will be removed later.
   1236  */
   1237 struct mbuf *
   1238 ip_srcroute(struct mbuf *m0)
   1239 {
   1240 	struct in_addr *p, *q;
   1241 	struct mbuf *m;
   1242 	struct ip_srcrt *isr;
   1243 	struct m_tag *mtag;
   1244 
   1245 	mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE);
   1246 	if (mtag == NULL)
   1247 		return NULL;
   1248 	isr = (struct ip_srcrt *)(mtag + 1);
   1249 
   1250 	if (isr->isr_nhops == 0)
   1251 		return NULL;
   1252 
   1253 	m = m_get(M_DONTWAIT, MT_SOOPTS);
   1254 	if (m == NULL)
   1255 		return NULL;
   1256 
   1257 	MCLAIM(m, &inetdomain.dom_mowner);
   1258 #define OPTSIZ	(sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
   1259 
   1260 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
   1261 	m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
   1262 
   1263 	/*
   1264 	 * First save first hop for return route
   1265 	 */
   1266 	p = &(isr->isr_routes[isr->isr_nhops - 1]);
   1267 	*(mtod(m, struct in_addr *)) = *p--;
   1268 
   1269 	/*
   1270 	 * Copy option fields and padding (nop) to mbuf.
   1271 	 */
   1272 	isr->isr_nop = IPOPT_NOP;
   1273 	isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
   1274 	memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop,
   1275 	    OPTSIZ);
   1276 	q = (struct in_addr *)(mtod(m, char *) +
   1277 	    sizeof(struct in_addr) + OPTSIZ);
   1278 #undef OPTSIZ
   1279 	/*
   1280 	 * Record return path as an IP source route,
   1281 	 * reversing the path (pointers are now aligned).
   1282 	 */
   1283 	while (p >= isr->isr_routes) {
   1284 		*q++ = *p--;
   1285 	}
   1286 	/*
   1287 	 * Last hop goes to final destination.
   1288 	 */
   1289 	*q = isr->isr_dst;
   1290 	m_tag_delete(m0, mtag);
   1291 	return m;
   1292 }
   1293 
   1294 const int inetctlerrmap[PRC_NCMDS] = {
   1295 	[PRC_MSGSIZE] = EMSGSIZE,
   1296 	[PRC_HOSTDEAD] = EHOSTDOWN,
   1297 	[PRC_HOSTUNREACH] = EHOSTUNREACH,
   1298 	[PRC_UNREACH_NET] = EHOSTUNREACH,
   1299 	[PRC_UNREACH_HOST] = EHOSTUNREACH,
   1300 	[PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
   1301 	[PRC_UNREACH_PORT] = ECONNREFUSED,
   1302 	[PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
   1303 	[PRC_PARAMPROB] = ENOPROTOOPT,
   1304 };
   1305 
   1306 void
   1307 ip_fasttimo(void)
   1308 {
   1309 	if (ip_drainwanted) {
   1310 		ip_drain();
   1311 		ip_drainwanted = 0;
   1312 	}
   1313 }
   1314 
   1315 void
   1316 ip_drainstub(void)
   1317 {
   1318 	ip_drainwanted = 1;
   1319 }
   1320 
   1321 /*
   1322  * Forward a packet.  If some error occurs return the sender
   1323  * an icmp packet.  Note we can't always generate a meaningful
   1324  * icmp message because icmp doesn't have a large enough repertoire
   1325  * of codes and types.
   1326  *
   1327  * If not forwarding, just drop the packet.  This could be confusing
   1328  * if ipforwarding was zero but some routing protocol was advancing
   1329  * us as a gateway to somewhere.  However, we must let the routing
   1330  * protocol deal with that.
   1331  *
   1332  * The srcrt parameter indicates whether the packet is being forwarded
   1333  * via a source route.
   1334  */
   1335 static void
   1336 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
   1337 {
   1338 	struct ip *ip = mtod(m, struct ip *);
   1339 	struct rtentry *rt;
   1340 	int error, type = 0, code = 0, destmtu = 0;
   1341 	struct mbuf *mcopy;
   1342 	n_long dest;
   1343 	union {
   1344 		struct sockaddr		dst;
   1345 		struct sockaddr_in	dst4;
   1346 	} u;
   1347 	net_stat_ref_t ips;
   1348 	struct route *ro;
   1349 
   1350 	KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
   1351 	    "interrupt handler; synchronization assumptions violated");
   1352 
   1353 	/*
   1354 	 * We are now in the output path.
   1355 	 */
   1356 	MCLAIM(m, &ip_tx_mowner);
   1357 
   1358 	/*
   1359 	 * Clear any in-bound checksum flags for this packet.
   1360 	 */
   1361 	m->m_pkthdr.csum_flags = 0;
   1362 
   1363 	dest = 0;
   1364 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
   1365 		IP_STATINC(IP_STAT_CANTFORWARD);
   1366 		m_freem(m);
   1367 		return;
   1368 	}
   1369 
   1370 	if (ip->ip_ttl <= IPTTLDEC) {
   1371 		IP_STATINC(IP_STAT_TIMXCEED);
   1372 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
   1373 		return;
   1374 	}
   1375 
   1376 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
   1377 
   1378 	ro = rtcache_percpu_getref(ipforward_rt_percpu);
   1379 	rt = rtcache_lookup(ro, &u.dst);
   1380 	if (rt == NULL) {
   1381 		rtcache_percpu_putref(ipforward_rt_percpu);
   1382 		IP_STATINC(IP_STAT_NOROUTE);
   1383 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
   1384 		return;
   1385 	}
   1386 #ifdef NET_MPSAFE
   1387 	/*
   1388 	 * XXX workaround an inconsistency issue between address and route on
   1389 	 * address initialization to avoid packet looping.  See doc/TODO.smpnet.
   1390 	 */
   1391 	if (__predict_false(rt->rt_ifp->if_type == IFT_LOOP)) {
   1392 		rtcache_unref(rt, ro);
   1393 		rtcache_percpu_putref(ipforward_rt_percpu);
   1394 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
   1395 		IP_STATINC(IP_STAT_CANTFORWARD);
   1396 		return;
   1397 	}
   1398 #endif
   1399 
   1400 	/*
   1401 	 * Save at most 68 bytes of the packet in case
   1402 	 * we need to generate an ICMP message to the src.
   1403 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
   1404 	 */
   1405 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
   1406 	if (mcopy)
   1407 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
   1408 
   1409 	ip->ip_ttl -= IPTTLDEC;
   1410 
   1411 	/*
   1412 	 * If forwarding packet using same interface that it came in on,
   1413 	 * perhaps should send a redirect to sender to shortcut a hop.
   1414 	 * Only send redirect if source is sending directly to us,
   1415 	 * and if packet was not source routed (or has any options).
   1416 	 * Also, don't send redirect if forwarding using a default route
   1417 	 * or a route modified by a redirect.
   1418 	 */
   1419 	if (rt->rt_ifp == rcvif &&
   1420 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
   1421 	    !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
   1422 	    ipsendredirects && !srcrt) {
   1423 		if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
   1424 		    ifatoia(rt->rt_ifa)->ia_subnet) {
   1425 			if (rt->rt_flags & RTF_GATEWAY)
   1426 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
   1427 			else
   1428 				dest = ip->ip_dst.s_addr;
   1429 			/*
   1430 			 * Router requirements says to only send host
   1431 			 * redirects.
   1432 			 */
   1433 			type = ICMP_REDIRECT;
   1434 			code = ICMP_REDIRECT_HOST;
   1435 		}
   1436 	}
   1437 	rtcache_unref(rt, ro);
   1438 
   1439 	error = ip_output(m, NULL, ro,
   1440 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
   1441 	    NULL, NULL);
   1442 
   1443 	if (error) {
   1444 		IP_STATINC(IP_STAT_CANTFORWARD);
   1445 		goto error;
   1446 	}
   1447 
   1448 	ips = IP_STAT_GETREF();
   1449 	_NET_STATINC_REF(ips, IP_STAT_FORWARD);
   1450 
   1451 	if (type) {
   1452 		_NET_STATINC_REF(ips, IP_STAT_REDIRECTSENT);
   1453 		IP_STAT_PUTREF();
   1454 		goto redirect;
   1455 	}
   1456 
   1457 	IP_STAT_PUTREF();
   1458 	if (mcopy) {
   1459 #ifdef GATEWAY
   1460 		if (mcopy->m_flags & M_CANFASTFWD)
   1461 			ipflow_create(ro, mcopy);
   1462 #endif
   1463 		m_freem(mcopy);
   1464 	}
   1465 
   1466 	rtcache_percpu_putref(ipforward_rt_percpu);
   1467 	return;
   1468 
   1469 redirect:
   1470 error:
   1471 	if (mcopy == NULL) {
   1472 		rtcache_percpu_putref(ipforward_rt_percpu);
   1473 		return;
   1474 	}
   1475 
   1476 	switch (error) {
   1477 
   1478 	case 0:				/* forwarded, but need redirect */
   1479 		/* type, code set above */
   1480 		break;
   1481 
   1482 	case ENETUNREACH:		/* shouldn't happen, checked above */
   1483 	case EHOSTUNREACH:
   1484 	case ENETDOWN:
   1485 	case EHOSTDOWN:
   1486 	default:
   1487 		type = ICMP_UNREACH;
   1488 		code = ICMP_UNREACH_HOST;
   1489 		break;
   1490 
   1491 	case EMSGSIZE:
   1492 		type = ICMP_UNREACH;
   1493 		code = ICMP_UNREACH_NEEDFRAG;
   1494 
   1495 		if ((rt = rtcache_validate(ro)) != NULL) {
   1496 			destmtu = rt->rt_ifp->if_mtu;
   1497 			rtcache_unref(rt, ro);
   1498 		}
   1499 #ifdef IPSEC
   1500 		if (ipsec_used)
   1501 			ipsec_mtu(mcopy, &destmtu);
   1502 #endif
   1503 		IP_STATINC(IP_STAT_CANTFRAG);
   1504 		break;
   1505 
   1506 	case ENOBUFS:
   1507 		/*
   1508 		 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
   1509 		 * Requirements for IP Version 4 Routers.  Source quench can
   1510 		 * be a big problem under DoS attacks or if the underlying
   1511 		 * interface is rate-limited.
   1512 		 */
   1513 		m_freem(mcopy);
   1514 		rtcache_percpu_putref(ipforward_rt_percpu);
   1515 		return;
   1516 	}
   1517 	icmp_error(mcopy, type, code, dest, destmtu);
   1518 	rtcache_percpu_putref(ipforward_rt_percpu);
   1519 }
   1520 
   1521 void
   1522 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
   1523     struct mbuf *m)
   1524 {
   1525 	struct socket *so = inp->inp_socket;
   1526 	int inpflags = inp->inp_flags;
   1527 
   1528 	if (SOOPT_TIMESTAMP(so->so_options))
   1529 		mp = sbsavetimestamp(so->so_options, mp);
   1530 
   1531 	if (inpflags & INP_RECVDSTADDR) {
   1532 		*mp = sbcreatecontrol(&ip->ip_dst,
   1533 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
   1534 		if (*mp)
   1535 			mp = &(*mp)->m_next;
   1536 	}
   1537 
   1538 	if (inpflags & INP_RECVTTL) {
   1539 		*mp = sbcreatecontrol(&ip->ip_ttl,
   1540 		    sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
   1541 		if (*mp)
   1542 			mp = &(*mp)->m_next;
   1543 	}
   1544 
   1545 	struct psref psref;
   1546 	ifnet_t *ifp = m_get_rcvif_psref(m, &psref);
   1547 	if (__predict_false(ifp == NULL)) {
   1548 #ifdef DIAGNOSTIC
   1549 		printf("%s: missing receive interface\n", __func__);
   1550 #endif
   1551 		return; /* XXX should report error? */
   1552 	}
   1553 
   1554 	if (inpflags & INP_RECVPKTINFO) {
   1555 		struct in_pktinfo ipi;
   1556 		ipi.ipi_addr = ip->ip_dst;
   1557 		ipi.ipi_ifindex = ifp->if_index;
   1558 		*mp = sbcreatecontrol(&ipi,
   1559 		    sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
   1560 		if (*mp)
   1561 			mp = &(*mp)->m_next;
   1562 	}
   1563 	if (inpflags & INP_RECVIF) {
   1564 		struct sockaddr_dl sdl;
   1565 
   1566 		sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
   1567 		    NULL, 0);
   1568 		*mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
   1569 		if (*mp)
   1570 			mp = &(*mp)->m_next;
   1571 	}
   1572 	m_put_rcvif_psref(ifp, &psref);
   1573 }
   1574 
   1575 /*
   1576  * sysctl helper routine for net.inet.ip.forwsrcrt.
   1577  */
   1578 static int
   1579 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
   1580 {
   1581 	int error, tmp;
   1582 	struct sysctlnode node;
   1583 
   1584 	node = *rnode;
   1585 	tmp = ip_forwsrcrt;
   1586 	node.sysctl_data = &tmp;
   1587 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1588 	if (error || newp == NULL)
   1589 		return (error);
   1590 
   1591 	error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
   1592 	    0, NULL, NULL, NULL);
   1593 	if (error)
   1594 		return (error);
   1595 
   1596 	ip_forwsrcrt = tmp;
   1597 
   1598 	return (0);
   1599 }
   1600 
   1601 /*
   1602  * sysctl helper routine for net.inet.ip.mtudisctimeout.  checks the
   1603  * range of the new value and tweaks timers if it changes.
   1604  */
   1605 static int
   1606 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
   1607 {
   1608 	int error, tmp;
   1609 	struct sysctlnode node;
   1610 
   1611 	icmp_mtudisc_lock();
   1612 
   1613 	node = *rnode;
   1614 	tmp = ip_mtudisc_timeout;
   1615 	node.sysctl_data = &tmp;
   1616 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1617 	if (error || newp == NULL)
   1618 		goto out;
   1619 	if (tmp < 0) {
   1620 		error = EINVAL;
   1621 		goto out;
   1622 	}
   1623 
   1624 	ip_mtudisc_timeout = tmp;
   1625 	rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
   1626 	error = 0;
   1627 out:
   1628 	icmp_mtudisc_unlock();
   1629 	return error;
   1630 }
   1631 
   1632 static int
   1633 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
   1634 {
   1635 
   1636 	return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
   1637 }
   1638 
   1639 static void
   1640 sysctl_net_inet_ip_setup(struct sysctllog **clog)
   1641 {
   1642 	const struct sysctlnode *ip_node;
   1643 
   1644 	sysctl_createv(clog, 0, NULL, NULL,
   1645 		       CTLFLAG_PERMANENT,
   1646 		       CTLTYPE_NODE, "inet",
   1647 		       SYSCTL_DESCR("PF_INET related settings"),
   1648 		       NULL, 0, NULL, 0,
   1649 		       CTL_NET, PF_INET, CTL_EOL);
   1650 	sysctl_createv(clog, 0, NULL, &ip_node,
   1651 		       CTLFLAG_PERMANENT,
   1652 		       CTLTYPE_NODE, "ip",
   1653 		       SYSCTL_DESCR("IPv4 related settings"),
   1654 		       NULL, 0, NULL, 0,
   1655 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
   1656 
   1657 	sysctl_createv(clog, 0, NULL, NULL,
   1658 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1659 		       CTLTYPE_INT, "forwarding",
   1660 		       SYSCTL_DESCR("Enable forwarding of INET datagrams"),
   1661 		       NULL, 0, &ipforwarding, 0,
   1662 		       CTL_NET, PF_INET, IPPROTO_IP,
   1663 		       IPCTL_FORWARDING, CTL_EOL);
   1664 	sysctl_createv(clog, 0, NULL, NULL,
   1665 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1666 		       CTLTYPE_INT, "redirect",
   1667 		       SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
   1668 		       NULL, 0, &ipsendredirects, 0,
   1669 		       CTL_NET, PF_INET, IPPROTO_IP,
   1670 		       IPCTL_SENDREDIRECTS, CTL_EOL);
   1671 	sysctl_createv(clog, 0, NULL, NULL,
   1672 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1673 		       CTLTYPE_INT, "ttl",
   1674 		       SYSCTL_DESCR("Default TTL for an INET datagram"),
   1675 		       NULL, 0, &ip_defttl, 0,
   1676 		       CTL_NET, PF_INET, IPPROTO_IP,
   1677 		       IPCTL_DEFTTL, CTL_EOL);
   1678 	sysctl_createv(clog, 0, NULL, NULL,
   1679 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1680 		       CTLTYPE_INT, "forwsrcrt",
   1681 		       SYSCTL_DESCR("Enable forwarding of source-routed "
   1682 				    "datagrams"),
   1683 		       sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
   1684 		       CTL_NET, PF_INET, IPPROTO_IP,
   1685 		       IPCTL_FORWSRCRT, CTL_EOL);
   1686 	sysctl_createv(clog, 0, NULL, NULL,
   1687 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1688 		       CTLTYPE_INT, "directed-broadcast",
   1689 		       SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
   1690 		       NULL, 0, &ip_directedbcast, 0,
   1691 		       CTL_NET, PF_INET, IPPROTO_IP,
   1692 		       IPCTL_DIRECTEDBCAST, CTL_EOL);
   1693 	sysctl_createv(clog, 0, NULL, NULL,
   1694 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1695 		       CTLTYPE_INT, "allowsrcrt",
   1696 		       SYSCTL_DESCR("Accept source-routed datagrams"),
   1697 		       NULL, 0, &ip_allowsrcrt, 0,
   1698 		       CTL_NET, PF_INET, IPPROTO_IP,
   1699 		       IPCTL_ALLOWSRCRT, CTL_EOL);
   1700 
   1701 	sysctl_createv(clog, 0, NULL, NULL,
   1702 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1703 		       CTLTYPE_INT, "mtudisc",
   1704 		       SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
   1705 		       NULL, 0, &ip_mtudisc, 0,
   1706 		       CTL_NET, PF_INET, IPPROTO_IP,
   1707 		       IPCTL_MTUDISC, CTL_EOL);
   1708 	sysctl_createv(clog, 0, NULL, NULL,
   1709 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1710 		       CTLTYPE_INT, "anonportmin",
   1711 		       SYSCTL_DESCR("Lowest ephemeral port number to assign"),
   1712 		       sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
   1713 		       CTL_NET, PF_INET, IPPROTO_IP,
   1714 		       IPCTL_ANONPORTMIN, CTL_EOL);
   1715 	sysctl_createv(clog, 0, NULL, NULL,
   1716 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1717 		       CTLTYPE_INT, "anonportmax",
   1718 		       SYSCTL_DESCR("Highest ephemeral port number to assign"),
   1719 		       sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
   1720 		       CTL_NET, PF_INET, IPPROTO_IP,
   1721 		       IPCTL_ANONPORTMAX, CTL_EOL);
   1722 	sysctl_createv(clog, 0, NULL, NULL,
   1723 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1724 		       CTLTYPE_INT, "mtudisctimeout",
   1725 		       SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
   1726 		       sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
   1727 		       CTL_NET, PF_INET, IPPROTO_IP,
   1728 		       IPCTL_MTUDISCTIMEOUT, CTL_EOL);
   1729 #ifndef IPNOPRIVPORTS
   1730 	sysctl_createv(clog, 0, NULL, NULL,
   1731 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1732 		       CTLTYPE_INT, "lowportmin",
   1733 		       SYSCTL_DESCR("Lowest privileged ephemeral port number "
   1734 				    "to assign"),
   1735 		       sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
   1736 		       CTL_NET, PF_INET, IPPROTO_IP,
   1737 		       IPCTL_LOWPORTMIN, CTL_EOL);
   1738 	sysctl_createv(clog, 0, NULL, NULL,
   1739 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1740 		       CTLTYPE_INT, "lowportmax",
   1741 		       SYSCTL_DESCR("Highest privileged ephemeral port number "
   1742 				    "to assign"),
   1743 		       sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
   1744 		       CTL_NET, PF_INET, IPPROTO_IP,
   1745 		       IPCTL_LOWPORTMAX, CTL_EOL);
   1746 #endif /* IPNOPRIVPORTS */
   1747 #if NGRE > 0
   1748 	sysctl_createv(clog, 0, NULL, NULL,
   1749 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1750 		       CTLTYPE_INT, "grettl",
   1751 		       SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
   1752 		       NULL, 0, &ip_gre_ttl, 0,
   1753 		       CTL_NET, PF_INET, IPPROTO_IP,
   1754 		       IPCTL_GRE_TTL, CTL_EOL);
   1755 #endif /* NGRE */
   1756 	sysctl_createv(clog, 0, NULL, NULL,
   1757 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1758 		       CTLTYPE_INT, "checkinterface",
   1759 		       SYSCTL_DESCR("Enable receive side of Strong ES model "
   1760 				    "from RFC1122"),
   1761 		       NULL, 0, &ip_checkinterface, 0,
   1762 		       CTL_NET, PF_INET, IPPROTO_IP,
   1763 		       IPCTL_CHECKINTERFACE, CTL_EOL);
   1764 
   1765 	pktq_sysctl_setup(ip_pktq, clog, ip_node, IPCTL_IFQ);
   1766 
   1767 	sysctl_createv(clog, 0, NULL, NULL,
   1768 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1769 		       CTLTYPE_INT, "random_id",
   1770 		       SYSCTL_DESCR("Assign random ip_id values"),
   1771 		       NULL, 0, &ip_do_randomid, 0,
   1772 		       CTL_NET, PF_INET, IPPROTO_IP,
   1773 		       IPCTL_RANDOMID, CTL_EOL);
   1774 	sysctl_createv(clog, 0, NULL, NULL,
   1775 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1776 		       CTLTYPE_INT, "do_loopback_cksum",
   1777 		       SYSCTL_DESCR("Perform IP checksum on loopback"),
   1778 		       NULL, 0, &ip_do_loopback_cksum, 0,
   1779 		       CTL_NET, PF_INET, IPPROTO_IP,
   1780 		       IPCTL_LOOPBACKCKSUM, CTL_EOL);
   1781 	sysctl_createv(clog, 0, NULL, NULL,
   1782 		       CTLFLAG_PERMANENT,
   1783 		       CTLTYPE_STRUCT, "stats",
   1784 		       SYSCTL_DESCR("IP statistics"),
   1785 		       sysctl_net_inet_ip_stats, 0, NULL, 0,
   1786 		       CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
   1787 		       CTL_EOL);
   1788 #if NARP
   1789 	sysctl_createv(clog, 0, NULL, NULL,
   1790 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1791 		       CTLTYPE_INT, "dad_count",
   1792 		       SYSCTL_DESCR("Number of Duplicate Address Detection "
   1793 				    "probes to send"),
   1794 		       NULL, 0, &ip_dad_count, 0,
   1795 		       CTL_NET, PF_INET, IPPROTO_IP,
   1796 		       IPCTL_DAD_COUNT, CTL_EOL);
   1797 #endif
   1798 
   1799 	/* anonportalgo RFC6056 subtree */
   1800 	const struct sysctlnode *portalgo_node;
   1801 	sysctl_createv(clog, 0, NULL, &portalgo_node,
   1802 		       CTLFLAG_PERMANENT,
   1803 		       CTLTYPE_NODE, "anonportalgo",
   1804 		       SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
   1805 	    	       NULL, 0, NULL, 0,
   1806 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
   1807 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1808 		       CTLFLAG_PERMANENT,
   1809 		       CTLTYPE_STRING, "available",
   1810 		       SYSCTL_DESCR("available algorithms"),
   1811 		       sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
   1812 		       CTL_CREATE, CTL_EOL);
   1813 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1814 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1815 		       CTLTYPE_STRING, "selected",
   1816 		       SYSCTL_DESCR("selected algorithm"),
   1817 		       sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
   1818 		       CTL_CREATE, CTL_EOL);
   1819 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1820 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1821 		       CTLTYPE_STRUCT, "reserve",
   1822 		       SYSCTL_DESCR("bitmap of reserved ports"),
   1823 		       sysctl_portalgo_reserve4, 0, NULL, 0,
   1824 		       CTL_CREATE, CTL_EOL);
   1825 }
   1826 
   1827 void
   1828 ip_statinc(u_int stat)
   1829 {
   1830 
   1831 	KASSERT(stat < IP_NSTATS);
   1832 	IP_STATINC(stat);
   1833 }
   1834