Home | History | Annotate | Line # | Download | only in netinet
ip_input.c revision 1.355.2.5
      1 /*	$NetBSD: ip_input.c,v 1.355.2.5 2018/02/26 13:32:01 martin 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 1.355.2.5 2018/02/26 13:32:01 martin Exp $");
     95 
     96 #ifdef _KERNEL_OPT
     97 #include "opt_inet.h"
     98 #include "opt_compat_netbsd.h"
     99 #include "opt_gateway.h"
    100 #include "opt_ipsec.h"
    101 #include "opt_mrouting.h"
    102 #include "opt_mbuftrace.h"
    103 #include "opt_inet_csum.h"
    104 #include "opt_net_mpsafe.h"
    105 #endif
    106 
    107 #include "arp.h"
    108 
    109 #include <sys/param.h>
    110 #include <sys/systm.h>
    111 #include <sys/cpu.h>
    112 #include <sys/mbuf.h>
    113 #include <sys/domain.h>
    114 #include <sys/protosw.h>
    115 #include <sys/socket.h>
    116 #include <sys/socketvar.h>
    117 #include <sys/errno.h>
    118 #include <sys/time.h>
    119 #include <sys/kernel.h>
    120 #include <sys/pool.h>
    121 #include <sys/sysctl.h>
    122 #include <sys/kauth.h>
    123 
    124 #include <net/if.h>
    125 #include <net/if_dl.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 /* 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	0	/* forward source-routed packets */
    166 #endif
    167 #ifndef IPALLOWSRCRT
    168 #define	IPALLOWSRCRT	0	/* 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 #ifdef COMPAT_50
    178 #include <compat/sys/time.h>
    179 #include <compat/sys/socket.h>
    180 #endif
    181 
    182 /*
    183  * Note: DIRECTED_BROADCAST is handled this way so that previous
    184  * configuration using this option will Just Work.
    185  */
    186 #ifndef IPDIRECTEDBCAST
    187 #ifdef DIRECTED_BROADCAST
    188 #define IPDIRECTEDBCAST	1
    189 #else
    190 #define	IPDIRECTEDBCAST	0
    191 #endif /* DIRECTED_BROADCAST */
    192 #endif /* IPDIRECTEDBCAST */
    193 int	ipforwarding = IPFORWARDING;
    194 int	ipsendredirects = IPSENDREDIRECTS;
    195 int	ip_defttl = IPDEFTTL;
    196 int	ip_forwsrcrt = IPFORWSRCRT;
    197 int	ip_directedbcast = IPDIRECTEDBCAST;
    198 int	ip_allowsrcrt = IPALLOWSRCRT;
    199 int	ip_mtudisc = IPMTUDISC;
    200 int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
    201 #ifdef DIAGNOSTIC
    202 int	ipprintfs = 0;
    203 #endif
    204 
    205 int	ip_do_randomid = 0;
    206 
    207 /*
    208  * XXX - Setting ip_checkinterface mostly implements the receive side of
    209  * the Strong ES model described in RFC 1122, but since the routing table
    210  * and transmit implementation do not implement the Strong ES model,
    211  * setting this to 1 results in an odd hybrid.
    212  *
    213  * XXX - ip_checkinterface currently must be disabled if you use ipnat
    214  * to translate the destination address to another local interface.
    215  *
    216  * XXX - ip_checkinterface must be disabled if you add IP aliases
    217  * to the loopback interface instead of the interface where the
    218  * packets for those addresses are received.
    219  */
    220 static int		ip_checkinterface	__read_mostly = 0;
    221 
    222 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
    223 
    224 pktqueue_t *		ip_pktq			__read_mostly;
    225 pfil_head_t *		inet_pfil_hook		__read_mostly;
    226 ipid_state_t *		ip_ids			__read_mostly;
    227 percpu_t *		ipstat_percpu		__read_mostly;
    228 
    229 static percpu_t		*ipforward_rt_percpu	__cacheline_aligned;
    230 
    231 uint16_t ip_id;
    232 
    233 #ifdef INET_CSUM_COUNTERS
    234 #include <sys/device.h>
    235 
    236 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    237     NULL, "inet", "hwcsum bad");
    238 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    239     NULL, "inet", "hwcsum ok");
    240 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
    241     NULL, "inet", "swcsum");
    242 
    243 #define	INET_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
    244 
    245 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
    246 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
    247 EVCNT_ATTACH_STATIC(ip_swcsum);
    248 
    249 #else
    250 
    251 #define	INET_CSUM_COUNTER_INCR(ev)	/* nothing */
    252 
    253 #endif /* INET_CSUM_COUNTERS */
    254 
    255 /*
    256  * Used to save the IP options in case a protocol wants to respond
    257  * to an incoming packet over the same route if the packet got here
    258  * using IP source routing.  This allows connection establishment and
    259  * maintenance when the remote end is on a network that is not known
    260  * to us.
    261  */
    262 struct ip_srcrt {
    263 	int		isr_nhops;		   /* number of hops */
    264 	struct in_addr	isr_dst;		   /* final destination */
    265 	char		isr_nop;		   /* one NOP to align */
    266 	char		isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
    267 	struct in_addr	isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
    268 };
    269 
    270 static int ip_drainwanted;
    271 
    272 static void save_rte(struct mbuf *, u_char *, struct in_addr);
    273 
    274 #ifdef MBUFTRACE
    275 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
    276 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
    277 #endif
    278 
    279 static void		ipintr(void *);
    280 static void		ip_input(struct mbuf *);
    281 static void		ip_forward(struct mbuf *, int, struct ifnet *);
    282 static bool		ip_dooptions(struct mbuf *);
    283 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *);
    284 static void		sysctl_net_inet_ip_setup(struct sysctllog **);
    285 
    286 static struct in_ifaddr	*ip_match_our_address(struct ifnet *, struct ip *,
    287 			    int *);
    288 static struct in_ifaddr	*ip_match_our_address_broadcast(struct ifnet *,
    289 			    struct ip *);
    290 
    291 #ifdef NET_MPSAFE
    292 #define	SOFTNET_LOCK()		mutex_enter(softnet_lock)
    293 #define	SOFTNET_UNLOCK()	mutex_exit(softnet_lock)
    294 #else
    295 #define	SOFTNET_LOCK()		KASSERT(mutex_owned(softnet_lock))
    296 #define	SOFTNET_UNLOCK()	KASSERT(mutex_owned(softnet_lock))
    297 #endif
    298 
    299 /*
    300  * IP initialization: fill in IP protocol switch table.
    301  * All protocols not implemented in kernel go to raw IP protocol handler.
    302  */
    303 void
    304 ip_init(void)
    305 {
    306 	const struct protosw *pr;
    307 
    308 	in_init();
    309 	sysctl_net_inet_ip_setup(NULL);
    310 
    311 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
    312 	KASSERT(pr != NULL);
    313 
    314 	ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
    315 	KASSERT(ip_pktq != NULL);
    316 
    317 	for (u_int i = 0; i < IPPROTO_MAX; i++) {
    318 		ip_protox[i] = pr - inetsw;
    319 	}
    320 	for (pr = inetdomain.dom_protosw;
    321 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
    322 		if (pr->pr_domain->dom_family == PF_INET &&
    323 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
    324 			ip_protox[pr->pr_protocol] = pr - inetsw;
    325 
    326 	ip_reass_init();
    327 
    328 	ip_ids = ip_id_init();
    329 	ip_id = time_uptime & 0xfffff;
    330 
    331 #ifdef GATEWAY
    332 	ipflow_init();
    333 #endif
    334 
    335 	/* Register our Packet Filter hook. */
    336 	inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
    337 	KASSERT(inet_pfil_hook != NULL);
    338 
    339 #ifdef MBUFTRACE
    340 	MOWNER_ATTACH(&ip_tx_mowner);
    341 	MOWNER_ATTACH(&ip_rx_mowner);
    342 #endif /* MBUFTRACE */
    343 
    344 	ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
    345 	ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
    346 	ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
    347 }
    348 
    349 static struct in_ifaddr *
    350 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
    351 {
    352 	struct in_ifaddr *ia = NULL;
    353 	int checkif;
    354 
    355 	/*
    356 	 * Enable a consistency check between the destination address
    357 	 * and the arrival interface for a unicast packet (the RFC 1122
    358 	 * strong ES model) if IP forwarding is disabled and the packet
    359 	 * is not locally generated.
    360 	 *
    361 	 * XXX - Checking also should be disabled if the destination
    362 	 * address is ipnat'ed to a different interface.
    363 	 *
    364 	 * XXX - Checking is incompatible with IP aliases added
    365 	 * to the loopback interface instead of the interface where
    366 	 * the packets are received.
    367 	 *
    368 	 * XXX - We need to add a per ifaddr flag for this so that
    369 	 * we get finer grain control.
    370 	 */
    371 	checkif = ip_checkinterface && (ipforwarding == 0) &&
    372 	    (ifp->if_flags & IFF_LOOPBACK) == 0;
    373 
    374 	IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
    375 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
    376 			if (ia->ia4_flags & IN_IFF_NOTREADY)
    377 				continue;
    378 			if (checkif && ia->ia_ifp != ifp)
    379 				continue;
    380 			if ((ia->ia_ifp->if_flags & IFF_UP) == 0) {
    381 				(*downmatch)++;
    382 				continue;
    383 			}
    384 			if (ia->ia4_flags & IN_IFF_DETACHED &&
    385 			    (ifp->if_flags & IFF_LOOPBACK) == 0)
    386 				continue;
    387 			break;
    388 		}
    389 	}
    390 
    391 	return ia;
    392 }
    393 
    394 static struct in_ifaddr *
    395 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
    396 {
    397 	struct in_ifaddr *ia = NULL;
    398 	struct ifaddr *ifa;
    399 
    400 	IFADDR_READER_FOREACH(ifa, ifp) {
    401 		if (ifa->ifa_addr->sa_family != AF_INET)
    402 			continue;
    403 		ia = ifatoia(ifa);
    404 		if (ia->ia4_flags & IN_IFF_NOTREADY)
    405 			continue;
    406 		if (ia->ia4_flags & IN_IFF_DETACHED &&
    407 		    (ifp->if_flags & IFF_LOOPBACK) == 0)
    408 			continue;
    409 		if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
    410 		    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
    411 		    /*
    412 		     * Look for all-0's host part (old broadcast addr),
    413 		     * either for subnet or net.
    414 		     */
    415 		    ip->ip_dst.s_addr == ia->ia_subnet ||
    416 		    ip->ip_dst.s_addr == ia->ia_net)
    417 			goto matched;
    418 		/*
    419 		 * An interface with IP address zero accepts
    420 		 * all packets that arrive on that interface.
    421 		 */
    422 		if (in_nullhost(ia->ia_addr.sin_addr))
    423 			goto matched;
    424 	}
    425 	ia = NULL;
    426 
    427 matched:
    428 	return ia;
    429 }
    430 
    431 /*
    432  * IP software interrupt routine.
    433  */
    434 static void
    435 ipintr(void *arg __unused)
    436 {
    437 	struct mbuf *m;
    438 
    439 	KASSERT(cpu_softintr_p());
    440 
    441 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    442 	while ((m = pktq_dequeue(ip_pktq)) != NULL) {
    443 		ip_input(m);
    444 	}
    445 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    446 }
    447 
    448 /*
    449  * IP input routine.  Checksum and byte swap header.  If fragmented
    450  * try to reassemble.  Process options.  Pass to next level.
    451  */
    452 static void
    453 ip_input(struct mbuf *m)
    454 {
    455 	struct ip *ip = NULL;
    456 	struct in_ifaddr *ia = NULL;
    457 	int hlen = 0, len;
    458 	int downmatch;
    459 	int srcrt = 0;
    460 	ifnet_t *ifp;
    461 	struct psref psref;
    462 	int s;
    463 
    464 	KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
    465 	    "interrupt handler; synchronization assumptions violated");
    466 
    467 	MCLAIM(m, &ip_rx_mowner);
    468 	KASSERT((m->m_flags & M_PKTHDR) != 0);
    469 
    470 	ifp = m_get_rcvif_psref(m, &psref);
    471 	if (__predict_false(ifp == NULL))
    472 		goto out;
    473 
    474 	/*
    475 	 * If no IP addresses have been set yet but the interfaces
    476 	 * are receiving, can't do anything with incoming packets yet.
    477 	 * Note: we pre-check without locks held.
    478 	 */
    479 	if (IN_ADDRLIST_READER_EMPTY())
    480 		goto out;
    481 	IP_STATINC(IP_STAT_TOTAL);
    482 
    483 	/*
    484 	 * If the IP header is not aligned, slurp it up into a new
    485 	 * mbuf with space for link headers, in the event we forward
    486 	 * it.  Otherwise, if it is aligned, make sure the entire
    487 	 * base IP header is in the first mbuf of the chain.
    488 	 */
    489 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
    490 		if ((m = m_copyup(m, sizeof(struct ip),
    491 				  (max_linkhdr + 3) & ~3)) == NULL) {
    492 			/* XXXJRT new stat, please */
    493 			IP_STATINC(IP_STAT_TOOSMALL);
    494 			goto out;
    495 		}
    496 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
    497 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
    498 			IP_STATINC(IP_STAT_TOOSMALL);
    499 			goto out;
    500 		}
    501 	}
    502 	ip = mtod(m, struct ip *);
    503 	if (ip->ip_v != IPVERSION) {
    504 		IP_STATINC(IP_STAT_BADVERS);
    505 		goto out;
    506 	}
    507 	hlen = ip->ip_hl << 2;
    508 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
    509 		IP_STATINC(IP_STAT_BADHLEN);
    510 		goto out;
    511 	}
    512 	if (hlen > m->m_len) {
    513 		if ((m = m_pullup(m, hlen)) == NULL) {
    514 			IP_STATINC(IP_STAT_BADHLEN);
    515 			goto out;
    516 		}
    517 		ip = mtod(m, struct ip *);
    518 	}
    519 
    520 	/*
    521 	 * RFC1122: packets with a multicast source address are
    522 	 * not allowed.
    523 	 */
    524 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    525 		IP_STATINC(IP_STAT_BADADDR);
    526 		goto out;
    527 	}
    528 
    529 	/* 127/8 must not appear on wire - RFC1122 */
    530 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
    531 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
    532 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
    533 			IP_STATINC(IP_STAT_BADADDR);
    534 			goto out;
    535 		}
    536 	}
    537 
    538 	switch (m->m_pkthdr.csum_flags &
    539 		((ifp->if_csum_flags_rx & M_CSUM_IPv4) |
    540 		 M_CSUM_IPv4_BAD)) {
    541 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
    542 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
    543 		IP_STATINC(IP_STAT_BADSUM);
    544 		goto out;
    545 
    546 	case M_CSUM_IPv4:
    547 		/* Checksum was okay. */
    548 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
    549 		break;
    550 
    551 	default:
    552 		/*
    553 		 * Must compute it ourselves.  Maybe skip checksum on
    554 		 * loopback interfaces.
    555 		 */
    556 		if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
    557 		    ip_do_loopback_cksum)) {
    558 			INET_CSUM_COUNTER_INCR(&ip_swcsum);
    559 			if (in_cksum(m, hlen) != 0) {
    560 				IP_STATINC(IP_STAT_BADSUM);
    561 				goto out;
    562 			}
    563 		}
    564 		break;
    565 	}
    566 
    567 	/* Retrieve the packet length. */
    568 	len = ntohs(ip->ip_len);
    569 
    570 	/*
    571 	 * Check for additional length bogosity
    572 	 */
    573 	if (len < hlen) {
    574 		IP_STATINC(IP_STAT_BADLEN);
    575 		goto out;
    576 	}
    577 
    578 	/*
    579 	 * Check that the amount of data in the buffers
    580 	 * is as at least much as the IP header would have us expect.
    581 	 * Trim mbufs if longer than we expect.
    582 	 * Drop packet if shorter than we expect.
    583 	 */
    584 	if (m->m_pkthdr.len < len) {
    585 		IP_STATINC(IP_STAT_TOOSHORT);
    586 		goto out;
    587 	}
    588 	if (m->m_pkthdr.len > len) {
    589 		if (m->m_len == m->m_pkthdr.len) {
    590 			m->m_len = len;
    591 			m->m_pkthdr.len = len;
    592 		} else
    593 			m_adj(m, len - m->m_pkthdr.len);
    594 	}
    595 
    596 	/*
    597 	 * Assume that we can create a fast-forward IP flow entry
    598 	 * based on this packet.
    599 	 */
    600 	m->m_flags |= M_CANFASTFWD;
    601 
    602 	/*
    603 	 * Run through list of hooks for input packets.  If there are any
    604 	 * filters which require that additional packets in the flow are
    605 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
    606 	 * Note that filters must _never_ set this flag, as another filter
    607 	 * in the list may have previously cleared it.
    608 	 */
    609 #if defined(IPSEC)
    610 	if (!ipsec_used || !ipsec_indone(m))
    611 #else
    612 	if (1)
    613 #endif
    614 	{
    615 		struct in_addr odst = ip->ip_dst;
    616 		bool freed;
    617 
    618 		freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
    619 		if (freed || m == NULL) {
    620 			m = NULL;
    621 			goto out;
    622 		}
    623 		ip = mtod(m, struct ip *);
    624 		hlen = ip->ip_hl << 2;
    625 
    626 		/*
    627 		 * XXX The setting of "srcrt" here is to prevent ip_forward()
    628 		 * from generating ICMP redirects for packets that have
    629 		 * been redirected by a hook back out on to the same LAN that
    630 		 * they came from and is not an indication that the packet
    631 		 * is being inffluenced by source routing options.  This
    632 		 * allows things like
    633 		 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
    634 		 * where tlp0 is both on the 1.1.1.0/24 network and is the
    635 		 * default route for hosts on 1.1.1.0/24.  Of course this
    636 		 * also requires a "map tlp0 ..." to complete the story.
    637 		 * One might argue whether or not this kind of network config.
    638 		 * should be supported in this manner...
    639 		 */
    640 		srcrt = (odst.s_addr != ip->ip_dst.s_addr);
    641 	}
    642 
    643 #ifdef ALTQ
    644 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
    645 	if (altq_input) {
    646 		SOFTNET_LOCK();
    647 		if ((*altq_input)(m, AF_INET) == 0) {
    648 			/* Packet dropped by traffic conditioner. */
    649 			SOFTNET_UNLOCK();
    650 			m = NULL;
    651 			goto out;
    652 		}
    653 		SOFTNET_UNLOCK();
    654 	}
    655 #endif
    656 
    657 	/*
    658 	 * Process options and, if not destined for us,
    659 	 * ship it on.  ip_dooptions returns 1 when an
    660 	 * error was detected (causing an icmp message
    661 	 * to be sent and the original packet to be freed).
    662 	 */
    663 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
    664 		m = NULL;
    665 		goto out;
    666 	}
    667 
    668 	/*
    669 	 * Check our list of addresses, to see if the packet is for us.
    670 	 *
    671 	 * Traditional 4.4BSD did not consult IFF_UP at all.
    672 	 * The behavior here is to treat addresses on !IFF_UP interface
    673 	 * or IN_IFF_NOTREADY addresses as not mine.
    674 	 */
    675 	downmatch = 0;
    676 	s = pserialize_read_enter();
    677 	ia = ip_match_our_address(ifp, ip, &downmatch);
    678 	if (ia != NULL) {
    679 		pserialize_read_exit(s);
    680 		goto ours;
    681 	}
    682 
    683 	if (ifp->if_flags & IFF_BROADCAST) {
    684 		ia = ip_match_our_address_broadcast(ifp, ip);
    685 		if (ia != NULL) {
    686 			pserialize_read_exit(s);
    687 			goto ours;
    688 		}
    689 	}
    690 	pserialize_read_exit(s);
    691 
    692 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
    693 #ifdef MROUTING
    694 		extern struct socket *ip_mrouter;
    695 
    696 		if (ip_mrouter) {
    697 			/*
    698 			 * If we are acting as a multicast router, all
    699 			 * incoming multicast packets are passed to the
    700 			 * kernel-level multicast forwarding function.
    701 			 * The packet is returned (relatively) intact; if
    702 			 * ip_mforward() returns a non-zero value, the packet
    703 			 * must be discarded, else it may be accepted below.
    704 			 *
    705 			 * (The IP ident field is put in the same byte order
    706 			 * as expected when ip_mforward() is called from
    707 			 * ip_output().)
    708 			 */
    709 			SOFTNET_LOCK();
    710 			if (ip_mforward(m, ifp) != 0) {
    711 				SOFTNET_UNLOCK();
    712 				IP_STATINC(IP_STAT_CANTFORWARD);
    713 				goto out;
    714 			}
    715 			SOFTNET_UNLOCK();
    716 
    717 			/*
    718 			 * The process-level routing demon needs to receive
    719 			 * all multicast IGMP packets, whether or not this
    720 			 * host belongs to their destination groups.
    721 			 */
    722 			if (ip->ip_p == IPPROTO_IGMP) {
    723 				goto ours;
    724 			}
    725 			IP_STATINC(IP_STAT_CANTFORWARD);
    726 		}
    727 #endif
    728 		/*
    729 		 * See if we belong to the destination multicast group on the
    730 		 * arrival interface.
    731 		 */
    732 		if (!in_multi_group(ip->ip_dst, ifp, 0)) {
    733 			IP_STATINC(IP_STAT_CANTFORWARD);
    734 			goto out;
    735 		}
    736 		goto ours;
    737 	}
    738 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    739 	    in_nullhost(ip->ip_dst))
    740 		goto ours;
    741 
    742 	/*
    743 	 * Not for us; forward if possible and desirable.
    744 	 */
    745 	if (ipforwarding == 0) {
    746 		m_put_rcvif_psref(ifp, &psref);
    747 		IP_STATINC(IP_STAT_CANTFORWARD);
    748 		m_freem(m);
    749 	} else {
    750 		/*
    751 		 * If ip_dst matched any of my address on !IFF_UP interface,
    752 		 * and there's no IFF_UP interface that matches ip_dst,
    753 		 * send icmp unreach.  Forwarding it will result in in-kernel
    754 		 * forwarding loop till TTL goes to 0.
    755 		 */
    756 		if (downmatch) {
    757 			m_put_rcvif_psref(ifp, &psref);
    758 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
    759 			IP_STATINC(IP_STAT_CANTFORWARD);
    760 			return;
    761 		}
    762 #ifdef IPSEC
    763 		/* Check the security policy (SP) for the packet */
    764 		if (ipsec_used) {
    765 			if (ipsec4_input(m, IP_FORWARDING |
    766 			    (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
    767 				goto out;
    768 			}
    769 		}
    770 #endif
    771 		ip_forward(m, srcrt, ifp);
    772 		m_put_rcvif_psref(ifp, &psref);
    773 	}
    774 	return;
    775 
    776 ours:
    777 	m_put_rcvif_psref(ifp, &psref);
    778 	ifp = NULL;
    779 
    780 	/*
    781 	 * If offset or IP_MF are set, must reassemble.
    782 	 */
    783 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
    784 		/*
    785 		 * Pass to IP reassembly mechanism.
    786 		 */
    787 		if (ip_reass_packet(&m, ip) != 0) {
    788 			/* Failed; invalid fragment(s) or packet. */
    789 			goto out;
    790 		}
    791 		if (m == NULL) {
    792 			/* More fragments should come; silently return. */
    793 			goto out;
    794 		}
    795 		/*
    796 		 * Reassembly is done, we have the final packet.
    797 		 * Updated cached data in local variable(s).
    798 		 */
    799 		ip = mtod(m, struct ip *);
    800 		hlen = ip->ip_hl << 2;
    801 	}
    802 
    803 #ifdef IPSEC
    804 	/*
    805 	 * Enforce IPsec policy checking if we are seeing last header.
    806 	 * Note that we do not visit this with protocols with PCB layer
    807 	 * code - like UDP/TCP/raw IP.
    808 	 */
    809 	if (ipsec_used &&
    810 	    (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
    811 		if (ipsec4_input(m, 0) != 0) {
    812 			goto out;
    813 		}
    814 	}
    815 #endif
    816 
    817 	/*
    818 	 * Switch out to protocol's input routine.
    819 	 */
    820 #if IFA_STATS
    821 	if (ia && ip) {
    822 		struct in_ifaddr *_ia;
    823 		/*
    824 		 * Keep a reference from ip_match_our_address with psref
    825 		 * is expensive, so explore ia here again.
    826 		 */
    827 		s = pserialize_read_enter();
    828 		_ia = in_get_ia(ip->ip_dst);
    829 		_ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
    830 		pserialize_read_exit(s);
    831 	}
    832 #endif
    833 	IP_STATINC(IP_STAT_DELIVERED);
    834 
    835 	const int off = hlen, nh = ip->ip_p;
    836 
    837 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
    838 	return;
    839 
    840 out:
    841 	m_put_rcvif_psref(ifp, &psref);
    842 	if (m != NULL)
    843 		m_freem(m);
    844 }
    845 
    846 /*
    847  * IP timer processing.
    848  */
    849 void
    850 ip_slowtimo(void)
    851 {
    852 
    853 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
    854 
    855 	ip_reass_slowtimo();
    856 
    857 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
    858 }
    859 
    860 /*
    861  * IP drain processing.
    862  */
    863 void
    864 ip_drain(void)
    865 {
    866 
    867 	KERNEL_LOCK(1, NULL);
    868 	ip_reass_drain();
    869 	KERNEL_UNLOCK_ONE(NULL);
    870 }
    871 
    872 /*
    873  * ip_dooptions: perform option processing on a datagram, possibly discarding
    874  * it if bad options are encountered, or forwarding it if source-routed.
    875  *
    876  * => Returns true if packet has been forwarded/freed.
    877  * => Returns false if the packet should be processed further.
    878  */
    879 static bool
    880 ip_dooptions(struct mbuf *m)
    881 {
    882 	struct ip *ip = mtod(m, struct ip *);
    883 	u_char *cp, *cp0;
    884 	struct ip_timestamp *ipt;
    885 	struct in_ifaddr *ia;
    886 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
    887 	struct in_addr dst;
    888 	n_time ntime;
    889 	struct ifaddr *ifa = NULL;
    890 	int s;
    891 
    892 	dst = ip->ip_dst;
    893 	cp = (u_char *)(ip + 1);
    894 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
    895 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
    896 		opt = cp[IPOPT_OPTVAL];
    897 		if (opt == IPOPT_EOL)
    898 			break;
    899 		if (opt == IPOPT_NOP)
    900 			optlen = 1;
    901 		else {
    902 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
    903 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    904 				goto bad;
    905 			}
    906 			optlen = cp[IPOPT_OLEN];
    907 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
    908 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    909 				goto bad;
    910 			}
    911 		}
    912 		switch (opt) {
    913 
    914 		default:
    915 			break;
    916 
    917 		/*
    918 		 * Source routing with record.
    919 		 * Find interface with current destination address.
    920 		 * If none on this machine then drop if strictly routed,
    921 		 * or do nothing if loosely routed.
    922 		 * Record interface address and bring up next address
    923 		 * component.  If strictly routed make sure next
    924 		 * address is on directly accessible net.
    925 		 */
    926 		case IPOPT_LSRR:
    927 		case IPOPT_SSRR: {
    928 			struct psref psref;
    929 			struct sockaddr_in ipaddr = {
    930 			    .sin_len = sizeof(ipaddr),
    931 			    .sin_family = AF_INET,
    932 			};
    933 
    934 			if (ip_allowsrcrt == 0) {
    935 				type = ICMP_UNREACH;
    936 				code = ICMP_UNREACH_NET_PROHIB;
    937 				goto bad;
    938 			}
    939 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
    940 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    941 				goto bad;
    942 			}
    943 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
    944 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
    945 				goto bad;
    946 			}
    947 			ipaddr.sin_addr = ip->ip_dst;
    948 
    949 			s = pserialize_read_enter();
    950 			ifa = ifa_ifwithaddr(sintosa(&ipaddr));
    951 			if (ifa == NULL) {
    952 				pserialize_read_exit(s);
    953 				if (opt == IPOPT_SSRR) {
    954 					type = ICMP_UNREACH;
    955 					code = ICMP_UNREACH_SRCFAIL;
    956 					goto bad;
    957 				}
    958 				/*
    959 				 * Loose routing, and not at next destination
    960 				 * yet; nothing to do except forward.
    961 				 */
    962 				break;
    963 			}
    964 			pserialize_read_exit(s);
    965 
    966 			off--;			/* 0 origin */
    967 			if ((off + sizeof(struct in_addr)) > optlen) {
    968 				/*
    969 				 * End of source route.  Should be for us.
    970 				 */
    971 				save_rte(m, cp, ip->ip_src);
    972 				break;
    973 			}
    974 			/*
    975 			 * locate outgoing interface
    976 			 */
    977 			memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
    978 			    sizeof(ipaddr.sin_addr));
    979 			if (opt == IPOPT_SSRR) {
    980 				ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr),
    981 				    &psref);
    982 				if (ifa != NULL)
    983 					ia = ifatoia(ifa);
    984 				else
    985 					ia = NULL;
    986 			} else {
    987 				ia = ip_rtaddr(ipaddr.sin_addr, &psref);
    988 			}
    989 			if (ia == NULL) {
    990 				type = ICMP_UNREACH;
    991 				code = ICMP_UNREACH_SRCFAIL;
    992 				goto bad;
    993 			}
    994 			ip->ip_dst = ipaddr.sin_addr;
    995 			bcopy((void *)&ia->ia_addr.sin_addr,
    996 			    (void *)(cp + off), sizeof(struct in_addr));
    997 			ia4_release(ia, &psref);
    998 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
    999 			/*
   1000 			 * Let ip_intr's mcast routing check handle mcast pkts
   1001 			 */
   1002 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
   1003 			break;
   1004 		    }
   1005 
   1006 		case IPOPT_RR: {
   1007 			struct psref psref;
   1008 			struct sockaddr_in ipaddr = {
   1009 			    .sin_len = sizeof(ipaddr),
   1010 			    .sin_family = AF_INET,
   1011 			};
   1012 
   1013 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
   1014 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
   1015 				goto bad;
   1016 			}
   1017 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
   1018 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
   1019 				goto bad;
   1020 			}
   1021 			/*
   1022 			 * If no space remains, ignore.
   1023 			 */
   1024 			off--;			/* 0 origin */
   1025 			if ((off + sizeof(struct in_addr)) > optlen)
   1026 				break;
   1027 			memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
   1028 			    sizeof(ipaddr.sin_addr));
   1029 			/*
   1030 			 * locate outgoing interface; if we're the destination,
   1031 			 * use the incoming interface (should be same).
   1032 			 */
   1033 			ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref);
   1034 			if (ifa == NULL) {
   1035 				ia = ip_rtaddr(ipaddr.sin_addr, &psref);
   1036 				if (ia == NULL) {
   1037 					type = ICMP_UNREACH;
   1038 					code = ICMP_UNREACH_HOST;
   1039 					goto bad;
   1040 				}
   1041 			} else {
   1042 				ia = ifatoia(ifa);
   1043 			}
   1044 			bcopy((void *)&ia->ia_addr.sin_addr,
   1045 			    (void *)(cp + off), sizeof(struct in_addr));
   1046 			ia4_release(ia, &psref);
   1047 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
   1048 			break;
   1049 		    }
   1050 
   1051 		case IPOPT_TS:
   1052 			code = cp - (u_char *)ip;
   1053 			ipt = (struct ip_timestamp *)cp;
   1054 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
   1055 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
   1056 				goto bad;
   1057 			}
   1058 			if (ipt->ipt_ptr < 5) {
   1059 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
   1060 				goto bad;
   1061 			}
   1062 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
   1063 				if (++ipt->ipt_oflw == 0) {
   1064 					code = (u_char *)&ipt->ipt_ptr -
   1065 					    (u_char *)ip;
   1066 					goto bad;
   1067 				}
   1068 				break;
   1069 			}
   1070 			cp0 = (cp + ipt->ipt_ptr - 1);
   1071 			switch (ipt->ipt_flg) {
   1072 
   1073 			case IPOPT_TS_TSONLY:
   1074 				break;
   1075 
   1076 			case IPOPT_TS_TSANDADDR: {
   1077 				struct ifnet *rcvif;
   1078 				int _s, _ss;
   1079 				struct sockaddr_in ipaddr = {
   1080 				    .sin_len = sizeof(ipaddr),
   1081 				    .sin_family = AF_INET,
   1082 				};
   1083 
   1084 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1085 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1086 					code = (u_char *)&ipt->ipt_ptr -
   1087 					    (u_char *)ip;
   1088 					goto bad;
   1089 				}
   1090 				ipaddr.sin_addr = dst;
   1091 				_ss = pserialize_read_enter();
   1092 				rcvif = m_get_rcvif(m, &_s);
   1093 				if (__predict_true(rcvif != NULL)) {
   1094 					ifa = ifaof_ifpforaddr(sintosa(&ipaddr),
   1095 					    rcvif);
   1096 				}
   1097 				m_put_rcvif(rcvif, &_s);
   1098 				if (ifa == NULL) {
   1099 					pserialize_read_exit(_ss);
   1100 					break;
   1101 				}
   1102 				ia = ifatoia(ifa);
   1103 				bcopy(&ia->ia_addr.sin_addr,
   1104 				    cp0, sizeof(struct in_addr));
   1105 				pserialize_read_exit(_ss);
   1106 				ipt->ipt_ptr += sizeof(struct in_addr);
   1107 				break;
   1108 			}
   1109 
   1110 			case IPOPT_TS_PRESPEC: {
   1111 				struct sockaddr_in ipaddr = {
   1112 				    .sin_len = sizeof(ipaddr),
   1113 				    .sin_family = AF_INET,
   1114 				};
   1115 
   1116 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
   1117 				    sizeof(struct in_addr) > ipt->ipt_len) {
   1118 					code = (u_char *)&ipt->ipt_ptr -
   1119 					    (u_char *)ip;
   1120 					goto bad;
   1121 				}
   1122 				memcpy(&ipaddr.sin_addr, cp0,
   1123 				    sizeof(struct in_addr));
   1124 				s = pserialize_read_enter();
   1125 				ifa = ifa_ifwithaddr(sintosa(&ipaddr));
   1126 				if (ifa == NULL) {
   1127 					pserialize_read_exit(s);
   1128 					continue;
   1129 				}
   1130 				pserialize_read_exit(s);
   1131 				ipt->ipt_ptr += sizeof(struct in_addr);
   1132 				break;
   1133 			    }
   1134 
   1135 			default:
   1136 				/* XXX can't take &ipt->ipt_flg */
   1137 				code = (u_char *)&ipt->ipt_ptr -
   1138 				    (u_char *)ip + 1;
   1139 				goto bad;
   1140 			}
   1141 			ntime = iptime();
   1142 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
   1143 			memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
   1144 			    sizeof(n_time));
   1145 			ipt->ipt_ptr += sizeof(n_time);
   1146 		}
   1147 	}
   1148 	if (forward) {
   1149 		struct ifnet *rcvif;
   1150 		struct psref _psref;
   1151 
   1152 		if (ip_forwsrcrt == 0) {
   1153 			type = ICMP_UNREACH;
   1154 			code = ICMP_UNREACH_SRCFAIL;
   1155 			goto bad;
   1156 		}
   1157 
   1158 		rcvif = m_get_rcvif_psref(m, &_psref);
   1159 		if (__predict_false(rcvif == NULL)) {
   1160 			type = ICMP_UNREACH;
   1161 			code = ICMP_UNREACH_HOST;
   1162 			goto bad;
   1163 		}
   1164 		ip_forward(m, 1, rcvif);
   1165 		m_put_rcvif_psref(rcvif, &_psref);
   1166 		return true;
   1167 	}
   1168 	return false;
   1169 bad:
   1170 	icmp_error(m, type, code, 0, 0);
   1171 	IP_STATINC(IP_STAT_BADOPTIONS);
   1172 	return true;
   1173 }
   1174 
   1175 /*
   1176  * ip_rtaddr: given address of next destination (final or next hop),
   1177  * return internet address info of interface to be used to get there.
   1178  */
   1179 static struct in_ifaddr *
   1180 ip_rtaddr(struct in_addr dst, struct psref *psref)
   1181 {
   1182 	struct rtentry *rt;
   1183 	union {
   1184 		struct sockaddr		dst;
   1185 		struct sockaddr_in	dst4;
   1186 	} u;
   1187 	struct route *ro;
   1188 
   1189 	sockaddr_in_init(&u.dst4, &dst, 0);
   1190 
   1191 	ro = percpu_getref(ipforward_rt_percpu);
   1192 	rt = rtcache_lookup(ro, &u.dst);
   1193 	if (rt == NULL) {
   1194 		percpu_putref(ipforward_rt_percpu);
   1195 		return NULL;
   1196 	}
   1197 
   1198 	ia4_acquire(ifatoia(rt->rt_ifa), psref);
   1199 	rtcache_unref(rt, ro);
   1200 	percpu_putref(ipforward_rt_percpu);
   1201 
   1202 	return ifatoia(rt->rt_ifa);
   1203 }
   1204 
   1205 /*
   1206  * save_rte: save incoming source route for use in replies, to be picked
   1207  * up later by ip_srcroute if the receiver is interested.
   1208  */
   1209 static void
   1210 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
   1211 {
   1212 	struct ip_srcrt *isr;
   1213 	struct m_tag *mtag;
   1214 	unsigned olen;
   1215 
   1216 	olen = option[IPOPT_OLEN];
   1217 	if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
   1218 		return;
   1219 
   1220 	mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
   1221 	if (mtag == NULL)
   1222 		return;
   1223 	isr = (struct ip_srcrt *)(mtag + 1);
   1224 
   1225 	memcpy(isr->isr_hdr, option, olen);
   1226 	isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
   1227 	isr->isr_dst = dst;
   1228 	m_tag_prepend(m, mtag);
   1229 }
   1230 
   1231 /*
   1232  * Retrieve incoming source route for use in replies,
   1233  * in the same form used by setsockopt.
   1234  * The first hop is placed before the options, will be removed later.
   1235  */
   1236 struct mbuf *
   1237 ip_srcroute(struct mbuf *m0)
   1238 {
   1239 	struct in_addr *p, *q;
   1240 	struct mbuf *m;
   1241 	struct ip_srcrt *isr;
   1242 	struct m_tag *mtag;
   1243 
   1244 	mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
   1245 	if (mtag == NULL)
   1246 		return NULL;
   1247 	isr = (struct ip_srcrt *)(mtag + 1);
   1248 
   1249 	if (isr->isr_nhops == 0)
   1250 		return NULL;
   1251 
   1252 	m = m_get(M_DONTWAIT, MT_SOOPTS);
   1253 	if (m == NULL)
   1254 		return NULL;
   1255 
   1256 	MCLAIM(m, &inetdomain.dom_mowner);
   1257 #define OPTSIZ	(sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
   1258 
   1259 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
   1260 	m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
   1261 
   1262 	/*
   1263 	 * First save first hop for return route
   1264 	 */
   1265 	p = &(isr->isr_routes[isr->isr_nhops - 1]);
   1266 	*(mtod(m, struct in_addr *)) = *p--;
   1267 
   1268 	/*
   1269 	 * Copy option fields and padding (nop) to mbuf.
   1270 	 */
   1271 	isr->isr_nop = IPOPT_NOP;
   1272 	isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
   1273 	memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop,
   1274 	    OPTSIZ);
   1275 	q = (struct in_addr *)(mtod(m, char *) +
   1276 	    sizeof(struct in_addr) + OPTSIZ);
   1277 #undef OPTSIZ
   1278 	/*
   1279 	 * Record return path as an IP source route,
   1280 	 * reversing the path (pointers are now aligned).
   1281 	 */
   1282 	while (p >= isr->isr_routes) {
   1283 		*q++ = *p--;
   1284 	}
   1285 	/*
   1286 	 * Last hop goes to final destination.
   1287 	 */
   1288 	*q = isr->isr_dst;
   1289 	m_tag_delete(m0, mtag);
   1290 	return m;
   1291 }
   1292 
   1293 const int inetctlerrmap[PRC_NCMDS] = {
   1294 	[PRC_MSGSIZE] = EMSGSIZE,
   1295 	[PRC_HOSTDEAD] = EHOSTDOWN,
   1296 	[PRC_HOSTUNREACH] = EHOSTUNREACH,
   1297 	[PRC_UNREACH_NET] = EHOSTUNREACH,
   1298 	[PRC_UNREACH_HOST] = EHOSTUNREACH,
   1299 	[PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
   1300 	[PRC_UNREACH_PORT] = ECONNREFUSED,
   1301 	[PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
   1302 	[PRC_PARAMPROB] = ENOPROTOOPT,
   1303 };
   1304 
   1305 void
   1306 ip_fasttimo(void)
   1307 {
   1308 	if (ip_drainwanted) {
   1309 		ip_drain();
   1310 		ip_drainwanted = 0;
   1311 	}
   1312 }
   1313 
   1314 void
   1315 ip_drainstub(void)
   1316 {
   1317 	ip_drainwanted = 1;
   1318 }
   1319 
   1320 /*
   1321  * Forward a packet.  If some error occurs return the sender
   1322  * an icmp packet.  Note we can't always generate a meaningful
   1323  * icmp message because icmp doesn't have a large enough repertoire
   1324  * of codes and types.
   1325  *
   1326  * If not forwarding, just drop the packet.  This could be confusing
   1327  * if ipforwarding was zero but some routing protocol was advancing
   1328  * us as a gateway to somewhere.  However, we must let the routing
   1329  * protocol deal with that.
   1330  *
   1331  * The srcrt parameter indicates whether the packet is being forwarded
   1332  * via a source route.
   1333  */
   1334 static void
   1335 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
   1336 {
   1337 	struct ip *ip = mtod(m, struct ip *);
   1338 	struct rtentry *rt;
   1339 	int error, type = 0, code = 0, destmtu = 0;
   1340 	struct mbuf *mcopy;
   1341 	n_long dest;
   1342 	union {
   1343 		struct sockaddr		dst;
   1344 		struct sockaddr_in	dst4;
   1345 	} u;
   1346 	uint64_t *ips;
   1347 	struct route *ro;
   1348 
   1349 	KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
   1350 	    "interrupt handler; synchronization assumptions violated");
   1351 
   1352 	/*
   1353 	 * We are now in the output path.
   1354 	 */
   1355 	MCLAIM(m, &ip_tx_mowner);
   1356 
   1357 	/*
   1358 	 * Clear any in-bound checksum flags for this packet.
   1359 	 */
   1360 	m->m_pkthdr.csum_flags = 0;
   1361 
   1362 	dest = 0;
   1363 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
   1364 		IP_STATINC(IP_STAT_CANTFORWARD);
   1365 		m_freem(m);
   1366 		return;
   1367 	}
   1368 
   1369 	if (ip->ip_ttl <= IPTTLDEC) {
   1370 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
   1371 		return;
   1372 	}
   1373 
   1374 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
   1375 
   1376 	ro = percpu_getref(ipforward_rt_percpu);
   1377 	rt = rtcache_lookup(ro, &u.dst);
   1378 	if (rt == NULL) {
   1379 		percpu_putref(ipforward_rt_percpu);
   1380 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
   1381 		return;
   1382 	}
   1383 
   1384 	/*
   1385 	 * Save at most 68 bytes of the packet in case
   1386 	 * we need to generate an ICMP message to the src.
   1387 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
   1388 	 */
   1389 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
   1390 	if (mcopy)
   1391 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
   1392 
   1393 	ip->ip_ttl -= IPTTLDEC;
   1394 
   1395 	/*
   1396 	 * If forwarding packet using same interface that it came in on,
   1397 	 * perhaps should send a redirect to sender to shortcut a hop.
   1398 	 * Only send redirect if source is sending directly to us,
   1399 	 * and if packet was not source routed (or has any options).
   1400 	 * Also, don't send redirect if forwarding using a default route
   1401 	 * or a route modified by a redirect.
   1402 	 */
   1403 	if (rt->rt_ifp == rcvif &&
   1404 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
   1405 	    !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
   1406 	    ipsendredirects && !srcrt) {
   1407 		if (rt->rt_ifa &&
   1408 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
   1409 		    ifatoia(rt->rt_ifa)->ia_subnet) {
   1410 			if (rt->rt_flags & RTF_GATEWAY)
   1411 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
   1412 			else
   1413 				dest = ip->ip_dst.s_addr;
   1414 			/*
   1415 			 * Router requirements says to only send host
   1416 			 * redirects.
   1417 			 */
   1418 			type = ICMP_REDIRECT;
   1419 			code = ICMP_REDIRECT_HOST;
   1420 		}
   1421 	}
   1422 	rtcache_unref(rt, ro);
   1423 
   1424 	error = ip_output(m, NULL, ro,
   1425 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
   1426 	    NULL, NULL);
   1427 
   1428 	if (error) {
   1429 		IP_STATINC(IP_STAT_CANTFORWARD);
   1430 		goto error;
   1431 	}
   1432 
   1433 	ips = IP_STAT_GETREF();
   1434 	ips[IP_STAT_FORWARD]++;
   1435 
   1436 	if (type) {
   1437 		ips[IP_STAT_REDIRECTSENT]++;
   1438 		IP_STAT_PUTREF();
   1439 		goto redirect;
   1440 	}
   1441 
   1442 	IP_STAT_PUTREF();
   1443 	if (mcopy) {
   1444 #ifdef GATEWAY
   1445 		if (mcopy->m_flags & M_CANFASTFWD)
   1446 			ipflow_create(ro, mcopy);
   1447 #endif
   1448 		m_freem(mcopy);
   1449 	}
   1450 
   1451 	percpu_putref(ipforward_rt_percpu);
   1452 	return;
   1453 
   1454 redirect:
   1455 error:
   1456 	if (mcopy == NULL) {
   1457 		percpu_putref(ipforward_rt_percpu);
   1458 		return;
   1459 	}
   1460 
   1461 	switch (error) {
   1462 
   1463 	case 0:				/* forwarded, but need redirect */
   1464 		/* type, code set above */
   1465 		break;
   1466 
   1467 	case ENETUNREACH:		/* shouldn't happen, checked above */
   1468 	case EHOSTUNREACH:
   1469 	case ENETDOWN:
   1470 	case EHOSTDOWN:
   1471 	default:
   1472 		type = ICMP_UNREACH;
   1473 		code = ICMP_UNREACH_HOST;
   1474 		break;
   1475 
   1476 	case EMSGSIZE:
   1477 		type = ICMP_UNREACH;
   1478 		code = ICMP_UNREACH_NEEDFRAG;
   1479 
   1480 		if ((rt = rtcache_validate(ro)) != NULL) {
   1481 			destmtu = rt->rt_ifp->if_mtu;
   1482 			rtcache_unref(rt, ro);
   1483 		}
   1484 #ifdef IPSEC
   1485 		if (ipsec_used)
   1486 			(void)ipsec4_forward(mcopy, &destmtu);
   1487 #endif
   1488 		IP_STATINC(IP_STAT_CANTFRAG);
   1489 		break;
   1490 
   1491 	case ENOBUFS:
   1492 		/*
   1493 		 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
   1494 		 * Requirements for IP Version 4 Routers.  Source quench can
   1495 		 * big problem under DoS attacks or if the underlying
   1496 		 * interface is rate-limited.
   1497 		 */
   1498 		if (mcopy)
   1499 			m_freem(mcopy);
   1500 		percpu_putref(ipforward_rt_percpu);
   1501 		return;
   1502 	}
   1503 	icmp_error(mcopy, type, code, dest, destmtu);
   1504 	percpu_putref(ipforward_rt_percpu);
   1505 }
   1506 
   1507 void
   1508 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
   1509     struct mbuf *m)
   1510 {
   1511 	struct socket *so = inp->inp_socket;
   1512 	ifnet_t *ifp;
   1513 	int inpflags = inp->inp_flags;
   1514 	struct psref psref;
   1515 
   1516 	ifp = m_get_rcvif_psref(m, &psref);
   1517 	if (__predict_false(ifp == NULL))
   1518 		return; /* XXX should report error? */
   1519 
   1520 	if (so->so_options & SO_TIMESTAMP
   1521 #ifdef SO_OTIMESTAMP
   1522 	    || so->so_options & SO_OTIMESTAMP
   1523 #endif
   1524 	    ) {
   1525 		struct timeval tv;
   1526 
   1527 		microtime(&tv);
   1528 #ifdef SO_OTIMESTAMP
   1529 		if (so->so_options & SO_OTIMESTAMP) {
   1530 			struct timeval50 tv50;
   1531 			timeval_to_timeval50(&tv, &tv50);
   1532 			*mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
   1533 			    SCM_OTIMESTAMP, SOL_SOCKET);
   1534 		} else
   1535 #endif
   1536 		*mp = sbcreatecontrol((void *) &tv, sizeof(tv),
   1537 		    SCM_TIMESTAMP, SOL_SOCKET);
   1538 		if (*mp)
   1539 			mp = &(*mp)->m_next;
   1540 	}
   1541 	if (inpflags & INP_RECVDSTADDR) {
   1542 		*mp = sbcreatecontrol((void *) &ip->ip_dst,
   1543 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
   1544 		if (*mp)
   1545 			mp = &(*mp)->m_next;
   1546 	}
   1547 	if (inpflags & INP_RECVPKTINFO) {
   1548 		struct in_pktinfo ipi;
   1549 		ipi.ipi_addr = ip->ip_src;
   1550 		ipi.ipi_ifindex = ifp->if_index;
   1551 		*mp = sbcreatecontrol((void *) &ipi,
   1552 		    sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP);
   1553 		if (*mp)
   1554 			mp = &(*mp)->m_next;
   1555 	}
   1556 	if (inpflags & INP_PKTINFO) {
   1557 		struct in_pktinfo ipi;
   1558 		ipi.ipi_addr = ip->ip_dst;
   1559 		ipi.ipi_ifindex = ifp->if_index;
   1560 		*mp = sbcreatecontrol((void *) &ipi,
   1561 		    sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
   1562 		if (*mp)
   1563 			mp = &(*mp)->m_next;
   1564 	}
   1565 	if (inpflags & INP_RECVIF) {
   1566 		struct sockaddr_dl sdl;
   1567 
   1568 		sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
   1569 		    NULL, 0);
   1570 		*mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
   1571 		if (*mp)
   1572 			mp = &(*mp)->m_next;
   1573 	}
   1574 	if (inpflags & INP_RECVTTL) {
   1575 		*mp = sbcreatecontrol((void *) &ip->ip_ttl,
   1576 		    sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
   1577 		if (*mp)
   1578 			mp = &(*mp)->m_next;
   1579 	}
   1580 	m_put_rcvif_psref(ifp, &psref);
   1581 }
   1582 
   1583 /*
   1584  * sysctl helper routine for net.inet.ip.forwsrcrt.
   1585  */
   1586 static int
   1587 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
   1588 {
   1589 	int error, tmp;
   1590 	struct sysctlnode node;
   1591 
   1592 	node = *rnode;
   1593 	tmp = ip_forwsrcrt;
   1594 	node.sysctl_data = &tmp;
   1595 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1596 	if (error || newp == NULL)
   1597 		return (error);
   1598 
   1599 	error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
   1600 	    0, NULL, NULL, NULL);
   1601 	if (error)
   1602 		return (error);
   1603 
   1604 	ip_forwsrcrt = tmp;
   1605 
   1606 	return (0);
   1607 }
   1608 
   1609 /*
   1610  * sysctl helper routine for net.inet.ip.mtudisctimeout.  checks the
   1611  * range of the new value and tweaks timers if it changes.
   1612  */
   1613 static int
   1614 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
   1615 {
   1616 	int error, tmp;
   1617 	struct sysctlnode node;
   1618 
   1619 	icmp_mtudisc_lock();
   1620 
   1621 	node = *rnode;
   1622 	tmp = ip_mtudisc_timeout;
   1623 	node.sysctl_data = &tmp;
   1624 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
   1625 	if (error || newp == NULL)
   1626 		goto out;
   1627 	if (tmp < 0) {
   1628 		error = EINVAL;
   1629 		goto out;
   1630 	}
   1631 
   1632 	ip_mtudisc_timeout = tmp;
   1633 	rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
   1634 	error = 0;
   1635 out:
   1636 	icmp_mtudisc_unlock();
   1637 	return error;
   1638 }
   1639 
   1640 static int
   1641 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
   1642 {
   1643 
   1644 	return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
   1645 }
   1646 
   1647 static void
   1648 sysctl_net_inet_ip_setup(struct sysctllog **clog)
   1649 {
   1650 	sysctl_createv(clog, 0, NULL, NULL,
   1651 		       CTLFLAG_PERMANENT,
   1652 		       CTLTYPE_NODE, "inet",
   1653 		       SYSCTL_DESCR("PF_INET related settings"),
   1654 		       NULL, 0, NULL, 0,
   1655 		       CTL_NET, PF_INET, CTL_EOL);
   1656 	sysctl_createv(clog, 0, NULL, NULL,
   1657 		       CTLFLAG_PERMANENT,
   1658 		       CTLTYPE_NODE, "ip",
   1659 		       SYSCTL_DESCR("IPv4 related settings"),
   1660 		       NULL, 0, NULL, 0,
   1661 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
   1662 
   1663 	sysctl_createv(clog, 0, NULL, NULL,
   1664 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1665 		       CTLTYPE_INT, "forwarding",
   1666 		       SYSCTL_DESCR("Enable forwarding of INET datagrams"),
   1667 		       NULL, 0, &ipforwarding, 0,
   1668 		       CTL_NET, PF_INET, IPPROTO_IP,
   1669 		       IPCTL_FORWARDING, CTL_EOL);
   1670 	sysctl_createv(clog, 0, NULL, NULL,
   1671 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1672 		       CTLTYPE_INT, "redirect",
   1673 		       SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
   1674 		       NULL, 0, &ipsendredirects, 0,
   1675 		       CTL_NET, PF_INET, IPPROTO_IP,
   1676 		       IPCTL_SENDREDIRECTS, CTL_EOL);
   1677 	sysctl_createv(clog, 0, NULL, NULL,
   1678 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1679 		       CTLTYPE_INT, "ttl",
   1680 		       SYSCTL_DESCR("Default TTL for an INET datagram"),
   1681 		       NULL, 0, &ip_defttl, 0,
   1682 		       CTL_NET, PF_INET, IPPROTO_IP,
   1683 		       IPCTL_DEFTTL, CTL_EOL);
   1684 #ifdef IPCTL_DEFMTU
   1685 	sysctl_createv(clog, 0, NULL, NULL,
   1686 		       CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
   1687 		       CTLTYPE_INT, "mtu",
   1688 		       SYSCTL_DESCR("Default MTA for an INET route"),
   1689 		       NULL, 0, &ip_mtu, 0,
   1690 		       CTL_NET, PF_INET, IPPROTO_IP,
   1691 		       IPCTL_DEFMTU, CTL_EOL);
   1692 #endif /* IPCTL_DEFMTU */
   1693 	sysctl_createv(clog, 0, NULL, NULL,
   1694 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1695 		       CTLTYPE_INT, "forwsrcrt",
   1696 		       SYSCTL_DESCR("Enable forwarding of source-routed "
   1697 				    "datagrams"),
   1698 		       sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
   1699 		       CTL_NET, PF_INET, IPPROTO_IP,
   1700 		       IPCTL_FORWSRCRT, CTL_EOL);
   1701 	sysctl_createv(clog, 0, NULL, NULL,
   1702 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1703 		       CTLTYPE_INT, "directed-broadcast",
   1704 		       SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
   1705 		       NULL, 0, &ip_directedbcast, 0,
   1706 		       CTL_NET, PF_INET, IPPROTO_IP,
   1707 		       IPCTL_DIRECTEDBCAST, CTL_EOL);
   1708 	sysctl_createv(clog, 0, NULL, NULL,
   1709 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1710 		       CTLTYPE_INT, "allowsrcrt",
   1711 		       SYSCTL_DESCR("Accept source-routed datagrams"),
   1712 		       NULL, 0, &ip_allowsrcrt, 0,
   1713 		       CTL_NET, PF_INET, IPPROTO_IP,
   1714 		       IPCTL_ALLOWSRCRT, CTL_EOL);
   1715 
   1716 	sysctl_createv(clog, 0, NULL, NULL,
   1717 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1718 		       CTLTYPE_INT, "mtudisc",
   1719 		       SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
   1720 		       NULL, 0, &ip_mtudisc, 0,
   1721 		       CTL_NET, PF_INET, IPPROTO_IP,
   1722 		       IPCTL_MTUDISC, CTL_EOL);
   1723 	sysctl_createv(clog, 0, NULL, NULL,
   1724 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1725 		       CTLTYPE_INT, "anonportmin",
   1726 		       SYSCTL_DESCR("Lowest ephemeral port number to assign"),
   1727 		       sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
   1728 		       CTL_NET, PF_INET, IPPROTO_IP,
   1729 		       IPCTL_ANONPORTMIN, CTL_EOL);
   1730 	sysctl_createv(clog, 0, NULL, NULL,
   1731 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1732 		       CTLTYPE_INT, "anonportmax",
   1733 		       SYSCTL_DESCR("Highest ephemeral port number to assign"),
   1734 		       sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
   1735 		       CTL_NET, PF_INET, IPPROTO_IP,
   1736 		       IPCTL_ANONPORTMAX, CTL_EOL);
   1737 	sysctl_createv(clog, 0, NULL, NULL,
   1738 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1739 		       CTLTYPE_INT, "mtudisctimeout",
   1740 		       SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
   1741 		       sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
   1742 		       CTL_NET, PF_INET, IPPROTO_IP,
   1743 		       IPCTL_MTUDISCTIMEOUT, CTL_EOL);
   1744 #ifndef IPNOPRIVPORTS
   1745 	sysctl_createv(clog, 0, NULL, NULL,
   1746 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1747 		       CTLTYPE_INT, "lowportmin",
   1748 		       SYSCTL_DESCR("Lowest privileged ephemeral port number "
   1749 				    "to assign"),
   1750 		       sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
   1751 		       CTL_NET, PF_INET, IPPROTO_IP,
   1752 		       IPCTL_LOWPORTMIN, CTL_EOL);
   1753 	sysctl_createv(clog, 0, NULL, NULL,
   1754 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1755 		       CTLTYPE_INT, "lowportmax",
   1756 		       SYSCTL_DESCR("Highest privileged ephemeral port number "
   1757 				    "to assign"),
   1758 		       sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
   1759 		       CTL_NET, PF_INET, IPPROTO_IP,
   1760 		       IPCTL_LOWPORTMAX, CTL_EOL);
   1761 #endif /* IPNOPRIVPORTS */
   1762 #if NGRE > 0
   1763 	sysctl_createv(clog, 0, NULL, NULL,
   1764 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1765 		       CTLTYPE_INT, "grettl",
   1766 		       SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
   1767 		       NULL, 0, &ip_gre_ttl, 0,
   1768 		       CTL_NET, PF_INET, IPPROTO_IP,
   1769 		       IPCTL_GRE_TTL, CTL_EOL);
   1770 #endif /* NGRE */
   1771 	sysctl_createv(clog, 0, NULL, NULL,
   1772 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1773 		       CTLTYPE_INT, "checkinterface",
   1774 		       SYSCTL_DESCR("Enable receive side of Strong ES model "
   1775 				    "from RFC1122"),
   1776 		       NULL, 0, &ip_checkinterface, 0,
   1777 		       CTL_NET, PF_INET, IPPROTO_IP,
   1778 		       IPCTL_CHECKINTERFACE, CTL_EOL);
   1779 	sysctl_createv(clog, 0, NULL, NULL,
   1780 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1781 		       CTLTYPE_INT, "random_id",
   1782 		       SYSCTL_DESCR("Assign random ip_id values"),
   1783 		       NULL, 0, &ip_do_randomid, 0,
   1784 		       CTL_NET, PF_INET, IPPROTO_IP,
   1785 		       IPCTL_RANDOMID, CTL_EOL);
   1786 	sysctl_createv(clog, 0, NULL, NULL,
   1787 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1788 		       CTLTYPE_INT, "do_loopback_cksum",
   1789 		       SYSCTL_DESCR("Perform IP checksum on loopback"),
   1790 		       NULL, 0, &ip_do_loopback_cksum, 0,
   1791 		       CTL_NET, PF_INET, IPPROTO_IP,
   1792 		       IPCTL_LOOPBACKCKSUM, CTL_EOL);
   1793 	sysctl_createv(clog, 0, NULL, NULL,
   1794 		       CTLFLAG_PERMANENT,
   1795 		       CTLTYPE_STRUCT, "stats",
   1796 		       SYSCTL_DESCR("IP statistics"),
   1797 		       sysctl_net_inet_ip_stats, 0, NULL, 0,
   1798 		       CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
   1799 		       CTL_EOL);
   1800 #if NARP
   1801 	sysctl_createv(clog, 0, NULL, NULL,
   1802 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1803 		       CTLTYPE_INT, "dad_count",
   1804 		       SYSCTL_DESCR("Number of Duplicate Address Detection "
   1805 				    "probes to send"),
   1806 		       NULL, 0, &ip_dad_count, 0,
   1807 		       CTL_NET, PF_INET, IPPROTO_IP,
   1808 		       IPCTL_DAD_COUNT, CTL_EOL);
   1809 #endif
   1810 
   1811 	/* anonportalgo RFC6056 subtree */
   1812 	const struct sysctlnode *portalgo_node;
   1813 	sysctl_createv(clog, 0, NULL, &portalgo_node,
   1814 		       CTLFLAG_PERMANENT,
   1815 		       CTLTYPE_NODE, "anonportalgo",
   1816 		       SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
   1817 	    	       NULL, 0, NULL, 0,
   1818 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
   1819 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1820 		       CTLFLAG_PERMANENT,
   1821 		       CTLTYPE_STRING, "available",
   1822 		       SYSCTL_DESCR("available algorithms"),
   1823 		       sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
   1824 		       CTL_CREATE, CTL_EOL);
   1825 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1826 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1827 		       CTLTYPE_STRING, "selected",
   1828 		       SYSCTL_DESCR("selected algorithm"),
   1829 		       sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
   1830 		       CTL_CREATE, CTL_EOL);
   1831 	sysctl_createv(clog, 0, &portalgo_node, NULL,
   1832 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1833 		       CTLTYPE_STRUCT, "reserve",
   1834 		       SYSCTL_DESCR("bitmap of reserved ports"),
   1835 		       sysctl_portalgo_reserve4, 0, NULL, 0,
   1836 		       CTL_CREATE, CTL_EOL);
   1837 }
   1838 
   1839 void
   1840 ip_statinc(u_int stat)
   1841 {
   1842 
   1843 	KASSERT(stat < IP_NSTATS);
   1844 	IP_STATINC(stat);
   1845 }
   1846