Home | History | Annotate | Line # | Download | only in netinet
ip_output.c revision 1.174
      1  1.174     joerg /*	$NetBSD: ip_output.c,v 1.174 2007/01/13 23:13:46 joerg Exp $	*/
      2   1.61    itojun 
      3   1.61    itojun /*
      4   1.61    itojun  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5   1.61    itojun  * All rights reserved.
      6   1.97    itojun  *
      7   1.61    itojun  * Redistribution and use in source and binary forms, with or without
      8   1.61    itojun  * modification, are permitted provided that the following conditions
      9   1.61    itojun  * are met:
     10   1.61    itojun  * 1. Redistributions of source code must retain the above copyright
     11   1.61    itojun  *    notice, this list of conditions and the following disclaimer.
     12   1.61    itojun  * 2. Redistributions in binary form must reproduce the above copyright
     13   1.61    itojun  *    notice, this list of conditions and the following disclaimer in the
     14   1.61    itojun  *    documentation and/or other materials provided with the distribution.
     15   1.61    itojun  * 3. Neither the name of the project nor the names of its contributors
     16   1.61    itojun  *    may be used to endorse or promote products derived from this software
     17   1.61    itojun  *    without specific prior written permission.
     18   1.97    itojun  *
     19   1.61    itojun  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20   1.61    itojun  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21   1.61    itojun  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22   1.61    itojun  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23   1.61    itojun  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24   1.61    itojun  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25   1.61    itojun  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26   1.61    itojun  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27   1.61    itojun  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28   1.61    itojun  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29   1.61    itojun  * SUCH DAMAGE.
     30   1.61    itojun  */
     31   1.54   thorpej 
     32   1.54   thorpej /*-
     33   1.54   thorpej  * Copyright (c) 1998 The NetBSD Foundation, Inc.
     34   1.54   thorpej  * All rights reserved.
     35   1.54   thorpej  *
     36   1.54   thorpej  * This code is derived from software contributed to The NetBSD Foundation
     37   1.54   thorpej  * by Public Access Networks Corporation ("Panix").  It was developed under
     38   1.54   thorpej  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     39   1.54   thorpej  *
     40   1.54   thorpej  * Redistribution and use in source and binary forms, with or without
     41   1.54   thorpej  * modification, are permitted provided that the following conditions
     42   1.54   thorpej  * are met:
     43   1.54   thorpej  * 1. Redistributions of source code must retain the above copyright
     44   1.54   thorpej  *    notice, this list of conditions and the following disclaimer.
     45   1.54   thorpej  * 2. Redistributions in binary form must reproduce the above copyright
     46   1.54   thorpej  *    notice, this list of conditions and the following disclaimer in the
     47   1.54   thorpej  *    documentation and/or other materials provided with the distribution.
     48   1.54   thorpej  * 3. All advertising materials mentioning features or use of this software
     49   1.54   thorpej  *    must display the following acknowledgement:
     50   1.54   thorpej  *	This product includes software developed by the NetBSD
     51   1.54   thorpej  *	Foundation, Inc. and its contributors.
     52   1.54   thorpej  * 4. Neither the name of The NetBSD Foundation nor the names of its
     53   1.54   thorpej  *    contributors may be used to endorse or promote products derived
     54   1.54   thorpej  *    from this software without specific prior written permission.
     55   1.54   thorpej  *
     56   1.54   thorpej  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     57   1.54   thorpej  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     58   1.54   thorpej  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     59   1.54   thorpej  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     60   1.54   thorpej  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     61   1.54   thorpej  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     62   1.54   thorpej  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     63   1.54   thorpej  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     64   1.54   thorpej  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     65   1.54   thorpej  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66   1.54   thorpej  * POSSIBILITY OF SUCH DAMAGE.
     67   1.54   thorpej  */
     68   1.19       cgd 
     69    1.1       cgd /*
     70   1.18   mycroft  * Copyright (c) 1982, 1986, 1988, 1990, 1993
     71   1.18   mycroft  *	The Regents of the University of California.  All rights reserved.
     72    1.1       cgd  *
     73    1.1       cgd  * Redistribution and use in source and binary forms, with or without
     74    1.1       cgd  * modification, are permitted provided that the following conditions
     75    1.1       cgd  * are met:
     76    1.1       cgd  * 1. Redistributions of source code must retain the above copyright
     77    1.1       cgd  *    notice, this list of conditions and the following disclaimer.
     78    1.1       cgd  * 2. Redistributions in binary form must reproduce the above copyright
     79    1.1       cgd  *    notice, this list of conditions and the following disclaimer in the
     80    1.1       cgd  *    documentation and/or other materials provided with the distribution.
     81  1.108       agc  * 3. Neither the name of the University nor the names of its contributors
     82    1.1       cgd  *    may be used to endorse or promote products derived from this software
     83    1.1       cgd  *    without specific prior written permission.
     84    1.1       cgd  *
     85    1.1       cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     86    1.1       cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     87    1.1       cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     88    1.1       cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     89    1.1       cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     90    1.1       cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     91    1.1       cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     92    1.1       cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     93    1.1       cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     94    1.1       cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     95    1.1       cgd  * SUCH DAMAGE.
     96    1.1       cgd  *
     97   1.19       cgd  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
     98    1.1       cgd  */
     99   1.89     lukem 
    100   1.89     lukem #include <sys/cdefs.h>
    101  1.174     joerg __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.174 2007/01/13 23:13:46 joerg Exp $");
    102   1.42    scottr 
    103   1.50       mrg #include "opt_pfil_hooks.h"
    104  1.128  jonathan #include "opt_inet.h"
    105   1.62   thorpej #include "opt_ipsec.h"
    106   1.42    scottr #include "opt_mrouting.h"
    107    1.1       cgd 
    108    1.8   mycroft #include <sys/param.h>
    109    1.8   mycroft #include <sys/malloc.h>
    110    1.8   mycroft #include <sys/mbuf.h>
    111    1.8   mycroft #include <sys/errno.h>
    112    1.8   mycroft #include <sys/protosw.h>
    113    1.8   mycroft #include <sys/socket.h>
    114    1.8   mycroft #include <sys/socketvar.h>
    115  1.162  christos #include <sys/kauth.h>
    116  1.118    itojun #ifdef FAST_IPSEC
    117  1.118    itojun #include <sys/domain.h>
    118  1.118    itojun #endif
    119   1.28  christos #include <sys/systm.h>
    120   1.61    itojun #include <sys/proc.h>
    121   1.61    itojun 
    122    1.8   mycroft #include <net/if.h>
    123    1.8   mycroft #include <net/route.h>
    124   1.38       mrg #include <net/pfil.h>
    125    1.1       cgd 
    126    1.8   mycroft #include <netinet/in.h>
    127    1.8   mycroft #include <netinet/in_systm.h>
    128    1.8   mycroft #include <netinet/ip.h>
    129    1.8   mycroft #include <netinet/in_pcb.h>
    130    1.8   mycroft #include <netinet/in_var.h>
    131    1.8   mycroft #include <netinet/ip_var.h>
    132  1.152      yamt #include <netinet/in_offload.h>
    133   1.72  jdolecek 
    134   1.72  jdolecek #ifdef MROUTING
    135   1.72  jdolecek #include <netinet/ip_mroute.h>
    136   1.72  jdolecek #endif
    137   1.32       mrg 
    138   1.28  christos #include <machine/stdarg.h>
    139   1.28  christos 
    140   1.61    itojun #ifdef IPSEC
    141   1.61    itojun #include <netinet6/ipsec.h>
    142   1.61    itojun #include <netkey/key.h>
    143   1.61    itojun #include <netkey/key_debug.h>
    144   1.61    itojun #endif /*IPSEC*/
    145   1.61    itojun 
    146  1.109  jonathan #ifdef FAST_IPSEC
    147  1.109  jonathan #include <netipsec/ipsec.h>
    148  1.109  jonathan #include <netipsec/key.h>
    149  1.109  jonathan #include <netipsec/xform.h>
    150  1.109  jonathan #endif	/* FAST_IPSEC*/
    151  1.109  jonathan 
    152  1.160  christos #ifdef IPSEC_NAT_T
    153  1.160  christos #include <netinet/udp.h>
    154  1.160  christos #endif
    155  1.160  christos 
    156  1.139     perry static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
    157  1.139     perry static struct ifnet *ip_multicast_if(struct in_addr *, int *);
    158  1.139     perry static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
    159  1.156  christos static int ip_getoptval(struct mbuf *, u_int8_t *, u_int);
    160    1.1       cgd 
    161   1.78   thorpej #ifdef PFIL_HOOKS
    162   1.78   thorpej extern struct pfil_head inet_pfil_hook;			/* XXX */
    163   1.78   thorpej #endif
    164   1.78   thorpej 
    165  1.151      yamt int	ip_do_loopback_cksum = 0;
    166  1.151      yamt 
    167  1.151      yamt #define	IN_NEED_CHECKSUM(ifp, csum_flags) \
    168  1.151      yamt 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
    169  1.151      yamt 	(((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \
    170  1.151      yamt 	(((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \
    171  1.151      yamt 	(((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum)))
    172  1.151      yamt 
    173    1.1       cgd /*
    174    1.1       cgd  * IP output.  The packet in mbuf chain m contains a skeletal IP
    175    1.1       cgd  * header (with len, off, ttl, proto, tos, src, dst).
    176    1.1       cgd  * The mbuf chain containing the packet will be freed.
    177    1.1       cgd  * The mbuf opt, if present, will not be freed.
    178    1.1       cgd  */
    179   1.12   mycroft int
    180   1.28  christos ip_output(struct mbuf *m0, ...)
    181    1.1       cgd {
    182  1.110    itojun 	struct ip *ip;
    183   1.71  augustss 	struct ifnet *ifp;
    184   1.71  augustss 	struct mbuf *m = m0;
    185   1.71  augustss 	int hlen = sizeof (struct ip);
    186  1.110    itojun 	int len, error = 0;
    187    1.1       cgd 	struct route iproute;
    188    1.1       cgd 	struct sockaddr_in *dst;
    189    1.1       cgd 	struct in_ifaddr *ia;
    190  1.166    dyoung 	struct ifaddr *xifa;
    191   1.28  christos 	struct mbuf *opt;
    192   1.28  christos 	struct route *ro;
    193   1.86   thorpej 	int flags, sw_csum;
    194   1.40      matt 	int *mtu_p;
    195   1.96    itojun 	u_long mtu;
    196   1.28  christos 	struct ip_moptions *imo;
    197  1.116    itojun 	struct socket *so;
    198   1.28  christos 	va_list ap;
    199  1.141      manu #ifdef IPSEC_NAT_T
    200  1.141      manu 	int natt_frag = 0;
    201  1.141      manu #endif
    202  1.160  christos #ifdef IPSEC
    203  1.160  christos 	struct secpolicy *sp = NULL;
    204   1.61    itojun #endif /*IPSEC*/
    205  1.109  jonathan #ifdef FAST_IPSEC
    206  1.116    itojun 	struct inpcb *inp;
    207  1.109  jonathan 	struct m_tag *mtag;
    208  1.109  jonathan 	struct secpolicy *sp = NULL;
    209  1.109  jonathan 	struct tdb_ident *tdbi;
    210  1.109  jonathan 	int s;
    211  1.109  jonathan #endif
    212   1.79   thorpej 	u_int16_t ip_len;
    213   1.28  christos 
    214  1.102   darrenr 	len = 0;
    215   1.28  christos 	va_start(ap, m0);
    216   1.28  christos 	opt = va_arg(ap, struct mbuf *);
    217   1.28  christos 	ro = va_arg(ap, struct route *);
    218   1.28  christos 	flags = va_arg(ap, int);
    219   1.28  christos 	imo = va_arg(ap, struct ip_moptions *);
    220  1.116    itojun 	so = va_arg(ap, struct socket *);
    221   1.40      matt 	if (flags & IP_RETURNMTU)
    222   1.40      matt 		mtu_p = va_arg(ap, int *);
    223   1.40      matt 	else
    224   1.40      matt 		mtu_p = NULL;
    225   1.28  christos 	va_end(ap);
    226   1.28  christos 
    227  1.103      matt 	MCLAIM(m, &ip_tx_mowner);
    228  1.116    itojun #ifdef FAST_IPSEC
    229  1.121  jonathan 	if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET)
    230  1.116    itojun 		inp = (struct inpcb *)so->so_pcb;
    231  1.116    itojun 	else
    232  1.116    itojun 		inp = NULL;
    233  1.130   thorpej #endif /* FAST_IPSEC */
    234   1.61    itojun 
    235    1.1       cgd #ifdef	DIAGNOSTIC
    236    1.1       cgd 	if ((m->m_flags & M_PKTHDR) == 0)
    237  1.163      tron 		panic("ip_output: no HDR");
    238  1.163      tron 
    239  1.164      tron 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) {
    240  1.163      tron 		panic("ip_output: IPv6 checksum offload flags: %d",
    241  1.163      tron 		    m->m_pkthdr.csum_flags);
    242  1.163      tron 	}
    243  1.163      tron 
    244  1.163      tron 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) ==
    245  1.163      tron 	    (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
    246  1.163      tron 		panic("ip_output: conflicting checksum offload flags: %d",
    247  1.163      tron 		    m->m_pkthdr.csum_flags);
    248  1.163      tron 	}
    249    1.1       cgd #endif
    250    1.1       cgd 	if (opt) {
    251    1.1       cgd 		m = ip_insertoptions(m, opt, &len);
    252  1.102   darrenr 		if (len >= sizeof(struct ip))
    253  1.102   darrenr 			hlen = len;
    254    1.1       cgd 	}
    255    1.1       cgd 	ip = mtod(m, struct ip *);
    256    1.1       cgd 	/*
    257    1.1       cgd 	 * Fill in IP header.
    258    1.1       cgd 	 */
    259   1.18   mycroft 	if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
    260    1.1       cgd 		ip->ip_v = IPVERSION;
    261  1.100    itojun 		ip->ip_off = htons(0);
    262  1.150      yamt 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
    263  1.150      yamt 			ip->ip_id = ip_newid();
    264  1.150      yamt 		} else {
    265  1.150      yamt 
    266  1.150      yamt 			/*
    267  1.150      yamt 			 * TSO capable interfaces (typically?) increment
    268  1.150      yamt 			 * ip_id for each segment.
    269  1.150      yamt 			 * "allocate" enough ids here to increase the chance
    270  1.150      yamt 			 * for them to be unique.
    271  1.150      yamt 			 *
    272  1.150      yamt 			 * note that the following calculation is not
    273  1.150      yamt 			 * needed to be precise.  wasting some ip_id is fine.
    274  1.150      yamt 			 */
    275  1.150      yamt 
    276  1.150      yamt 			unsigned int segsz = m->m_pkthdr.segsz;
    277  1.150      yamt 			unsigned int datasz = ntohs(ip->ip_len) - hlen;
    278  1.150      yamt 			unsigned int num = howmany(datasz, segsz);
    279  1.150      yamt 
    280  1.150      yamt 			ip->ip_id = ip_newid_range(num);
    281  1.150      yamt 		}
    282    1.1       cgd 		ip->ip_hl = hlen >> 2;
    283   1.18   mycroft 		ipstat.ips_localout++;
    284    1.1       cgd 	} else {
    285    1.1       cgd 		hlen = ip->ip_hl << 2;
    286    1.1       cgd 	}
    287    1.1       cgd 	/*
    288    1.1       cgd 	 * Route packet.
    289    1.1       cgd 	 */
    290  1.174     joerg 	bzero(&iproute, sizeof(iproute));
    291  1.174     joerg 	if (ro == NULL)
    292    1.1       cgd 		ro = &iproute;
    293   1.24   mycroft 	dst = satosin(&ro->ro_dst);
    294    1.1       cgd 	/*
    295    1.1       cgd 	 * If there is a cached route,
    296    1.1       cgd 	 * check that it is to the same destination
    297    1.1       cgd 	 * and is still up.  If not, free it and try again.
    298   1.92    itojun 	 * The address family should also be checked in case of sharing the
    299   1.92    itojun 	 * cache with IPv6.
    300    1.1       cgd 	 */
    301  1.171     joerg 	if (dst->sin_family != AF_INET ||
    302  1.171     joerg 	    !in_hosteq(dst->sin_addr, ip->ip_dst))
    303  1.171     joerg 		rtcache_free(ro);
    304  1.171     joerg 	else
    305  1.171     joerg 		rtcache_check(ro);
    306  1.170    dyoung 	if (ro->ro_rt == NULL) {
    307   1.92    itojun 		bzero(dst, sizeof(*dst));
    308    1.1       cgd 		dst->sin_family = AF_INET;
    309    1.1       cgd 		dst->sin_len = sizeof(*dst);
    310    1.1       cgd 		dst->sin_addr = ip->ip_dst;
    311    1.1       cgd 	}
    312    1.1       cgd 	/*
    313    1.1       cgd 	 * If routing to interface only,
    314    1.1       cgd 	 * short circuit routing lookup.
    315    1.1       cgd 	 */
    316    1.1       cgd 	if (flags & IP_ROUTETOIF) {
    317   1.29       mrg 		if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
    318   1.18   mycroft 			ipstat.ips_noroute++;
    319    1.1       cgd 			error = ENETUNREACH;
    320    1.1       cgd 			goto bad;
    321    1.1       cgd 		}
    322    1.1       cgd 		ifp = ia->ia_ifp;
    323   1.48      matt 		mtu = ifp->if_mtu;
    324   1.18   mycroft 		ip->ip_ttl = 1;
    325   1.98    itojun 	} else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
    326   1.98    itojun 	    ip->ip_dst.s_addr == INADDR_BROADCAST) &&
    327   1.98    itojun 	    imo != NULL && imo->imo_multicast_ifp != NULL) {
    328   1.98    itojun 		ifp = imo->imo_multicast_ifp;
    329   1.98    itojun 		mtu = ifp->if_mtu;
    330   1.99    itojun 		IFP_TO_IA(ifp, ia);
    331    1.1       cgd 	} else {
    332  1.169    dyoung 		if (ro->ro_rt == NULL)
    333  1.171     joerg 			rtcache_init(ro);
    334  1.168    dyoung 		if (ro->ro_rt == NULL) {
    335   1.18   mycroft 			ipstat.ips_noroute++;
    336    1.1       cgd 			error = EHOSTUNREACH;
    337    1.1       cgd 			goto bad;
    338    1.1       cgd 		}
    339   1.18   mycroft 		ia = ifatoia(ro->ro_rt->rt_ifa);
    340    1.1       cgd 		ifp = ro->ro_rt->rt_ifp;
    341   1.48      matt 		if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
    342   1.48      matt 			mtu = ifp->if_mtu;
    343    1.1       cgd 		ro->ro_rt->rt_use++;
    344    1.1       cgd 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    345   1.24   mycroft 			dst = satosin(ro->ro_rt->rt_gateway);
    346    1.1       cgd 	}
    347   1.64        is 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
    348   1.64        is 	    (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
    349    1.5   hpeyerl 		struct in_multi *inm;
    350    1.5   hpeyerl 
    351   1.64        is 		m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
    352   1.64        is 			M_BCAST : M_MCAST;
    353    1.5   hpeyerl 		/*
    354    1.5   hpeyerl 		 * IP destination address is multicast.  Make sure "dst"
    355    1.5   hpeyerl 		 * still points to the address in "ro".  (It may have been
    356    1.5   hpeyerl 		 * changed to point to a gateway address, above.)
    357    1.5   hpeyerl 		 */
    358   1.24   mycroft 		dst = satosin(&ro->ro_dst);
    359    1.5   hpeyerl 		/*
    360    1.5   hpeyerl 		 * See if the caller provided any multicast options
    361    1.5   hpeyerl 		 */
    362   1.98    itojun 		if (imo != NULL)
    363    1.5   hpeyerl 			ip->ip_ttl = imo->imo_multicast_ttl;
    364   1.98    itojun 		else
    365    1.5   hpeyerl 			ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
    366   1.98    itojun 
    367   1.98    itojun 		/*
    368   1.98    itojun 		 * if we don't know the outgoing ifp yet, we can't generate
    369   1.98    itojun 		 * output
    370   1.98    itojun 		 */
    371   1.98    itojun 		if (!ifp) {
    372   1.98    itojun 			ipstat.ips_noroute++;
    373   1.98    itojun 			error = ENETUNREACH;
    374   1.98    itojun 			goto bad;
    375   1.98    itojun 		}
    376   1.98    itojun 
    377    1.5   hpeyerl 		/*
    378   1.95   thorpej 		 * If the packet is multicast or broadcast, confirm that
    379   1.95   thorpej 		 * the outgoing interface can transmit it.
    380    1.5   hpeyerl 		 */
    381   1.64        is 		if (((m->m_flags & M_MCAST) &&
    382   1.64        is 		     (ifp->if_flags & IFF_MULTICAST) == 0) ||
    383   1.97    itojun 		    ((m->m_flags & M_BCAST) &&
    384   1.95   thorpej 		     (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0))  {
    385   1.18   mycroft 			ipstat.ips_noroute++;
    386    1.5   hpeyerl 			error = ENETUNREACH;
    387    1.5   hpeyerl 			goto bad;
    388    1.5   hpeyerl 		}
    389    1.5   hpeyerl 		/*
    390   1.44       tls 		 * If source address not specified yet, use an address
    391    1.5   hpeyerl 		 * of outgoing interface.
    392    1.5   hpeyerl 		 */
    393   1.31   mycroft 		if (in_nullhost(ip->ip_src)) {
    394  1.153  christos 			struct in_ifaddr *xia;
    395    1.5   hpeyerl 
    396  1.153  christos 			IFP_TO_IA(ifp, xia);
    397  1.153  christos 			if (!xia) {
    398   1.91    itojun 				error = EADDRNOTAVAIL;
    399   1.91    itojun 				goto bad;
    400   1.91    itojun 			}
    401  1.166    dyoung 			xifa = &xia->ia_ifa;
    402  1.166    dyoung 			if (xifa->ifa_getifa != NULL) {
    403  1.166    dyoung 				xia = ifatoia((*xifa->ifa_getifa)(xifa,
    404  1.166    dyoung 				    &ro->ro_dst));
    405  1.166    dyoung 			}
    406  1.153  christos 			ip->ip_src = xia->ia_addr.sin_addr;
    407    1.5   hpeyerl 		}
    408    1.5   hpeyerl 
    409    1.5   hpeyerl 		IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
    410    1.5   hpeyerl 		if (inm != NULL &&
    411    1.5   hpeyerl 		   (imo == NULL || imo->imo_multicast_loop)) {
    412    1.5   hpeyerl 			/*
    413   1.11   mycroft 			 * If we belong to the destination multicast group
    414    1.5   hpeyerl 			 * on the outgoing interface, and the caller did not
    415    1.5   hpeyerl 			 * forbid loopback, loop back a copy.
    416    1.5   hpeyerl 			 */
    417    1.5   hpeyerl 			ip_mloopback(ifp, m, dst);
    418    1.5   hpeyerl 		}
    419    1.5   hpeyerl #ifdef MROUTING
    420   1.18   mycroft 		else {
    421    1.5   hpeyerl 			/*
    422    1.5   hpeyerl 			 * If we are acting as a multicast router, perform
    423    1.5   hpeyerl 			 * multicast forwarding as if the packet had just
    424    1.5   hpeyerl 			 * arrived on the interface to which we are about
    425    1.5   hpeyerl 			 * to send.  The multicast forwarding function
    426    1.5   hpeyerl 			 * recursively calls this function, using the
    427    1.5   hpeyerl 			 * IP_FORWARDING flag to prevent infinite recursion.
    428    1.5   hpeyerl 			 *
    429    1.5   hpeyerl 			 * Multicasts that are looped back by ip_mloopback(),
    430    1.5   hpeyerl 			 * above, will be forwarded by the ip_input() routine,
    431    1.5   hpeyerl 			 * if necessary.
    432    1.5   hpeyerl 			 */
    433   1.18   mycroft 			extern struct socket *ip_mrouter;
    434   1.22       cgd 
    435   1.18   mycroft 			if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
    436   1.18   mycroft 				if (ip_mforward(m, ifp) != 0) {
    437   1.18   mycroft 					m_freem(m);
    438   1.18   mycroft 					goto done;
    439   1.18   mycroft 				}
    440    1.5   hpeyerl 			}
    441    1.5   hpeyerl 		}
    442    1.5   hpeyerl #endif
    443    1.5   hpeyerl 		/*
    444    1.5   hpeyerl 		 * Multicasts with a time-to-live of zero may be looped-
    445    1.5   hpeyerl 		 * back, above, but must not be transmitted on a network.
    446    1.5   hpeyerl 		 * Also, multicasts addressed to the loopback interface
    447    1.5   hpeyerl 		 * are not sent -- the above call to ip_mloopback() will
    448    1.5   hpeyerl 		 * loop back a copy if this host actually belongs to the
    449    1.5   hpeyerl 		 * destination group on the loopback interface.
    450    1.5   hpeyerl 		 */
    451   1.20   mycroft 		if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
    452    1.5   hpeyerl 			m_freem(m);
    453    1.5   hpeyerl 			goto done;
    454    1.5   hpeyerl 		}
    455    1.5   hpeyerl 
    456    1.5   hpeyerl 		goto sendit;
    457    1.5   hpeyerl 	}
    458    1.1       cgd 	/*
    459    1.1       cgd 	 * If source address not specified yet, use address
    460    1.1       cgd 	 * of outgoing interface.
    461    1.1       cgd 	 */
    462  1.166    dyoung 	if (in_nullhost(ip->ip_src)) {
    463  1.166    dyoung 		xifa = &ia->ia_ifa;
    464  1.166    dyoung 		if (xifa->ifa_getifa != NULL)
    465  1.166    dyoung 			ia = ifatoia((*xifa->ifa_getifa)(xifa, &ro->ro_dst));
    466   1.25   mycroft 		ip->ip_src = ia->ia_addr.sin_addr;
    467  1.166    dyoung 	}
    468   1.59       hwr 
    469   1.59       hwr 	/*
    470   1.97    itojun 	 * packets with Class-D address as source are not valid per
    471   1.59       hwr 	 * RFC 1112
    472   1.59       hwr 	 */
    473   1.59       hwr 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
    474   1.59       hwr 		ipstat.ips_odropped++;
    475   1.59       hwr 		error = EADDRNOTAVAIL;
    476   1.59       hwr 		goto bad;
    477   1.59       hwr 	}
    478   1.59       hwr 
    479    1.1       cgd 	/*
    480    1.1       cgd 	 * Look for broadcast address and
    481    1.1       cgd 	 * and verify user is allowed to send
    482    1.1       cgd 	 * such a packet.
    483    1.1       cgd 	 */
    484   1.18   mycroft 	if (in_broadcast(dst->sin_addr, ifp)) {
    485    1.1       cgd 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
    486    1.1       cgd 			error = EADDRNOTAVAIL;
    487    1.1       cgd 			goto bad;
    488    1.1       cgd 		}
    489    1.1       cgd 		if ((flags & IP_ALLOWBROADCAST) == 0) {
    490    1.1       cgd 			error = EACCES;
    491    1.1       cgd 			goto bad;
    492    1.1       cgd 		}
    493    1.1       cgd 		/* don't allow broadcast messages to be fragmented */
    494  1.100    itojun 		if (ntohs(ip->ip_len) > ifp->if_mtu) {
    495    1.1       cgd 			error = EMSGSIZE;
    496    1.1       cgd 			goto bad;
    497    1.1       cgd 		}
    498    1.1       cgd 		m->m_flags |= M_BCAST;
    499   1.18   mycroft 	} else
    500   1.18   mycroft 		m->m_flags &= ~M_BCAST;
    501   1.18   mycroft 
    502   1.60       mrg sendit:
    503   1.76   thorpej 	/*
    504   1.76   thorpej 	 * If we're doing Path MTU Discovery, we need to set DF unless
    505   1.76   thorpej 	 * the route's MTU is locked.
    506   1.76   thorpej 	 */
    507   1.76   thorpej 	if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
    508   1.76   thorpej 	    (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
    509  1.100    itojun 		ip->ip_off |= htons(IP_DF);
    510   1.76   thorpej 
    511  1.100    itojun 	/* Remember the current ip_len */
    512  1.100    itojun 	ip_len = ntohs(ip->ip_len);
    513   1.78   thorpej 
    514   1.61    itojun #ifdef IPSEC
    515   1.61    itojun 	/* get SP for this packet */
    516   1.61    itojun 	if (so == NULL)
    517  1.110    itojun 		sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
    518  1.110    itojun 		    flags, &error);
    519  1.130   thorpej 	else {
    520  1.130   thorpej 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
    521  1.130   thorpej 					 IPSEC_DIR_OUTBOUND))
    522  1.130   thorpej 			goto skip_ipsec;
    523   1.66    itojun 		sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    524  1.130   thorpej 	}
    525   1.61    itojun 
    526   1.61    itojun 	if (sp == NULL) {
    527   1.61    itojun 		ipsecstat.out_inval++;
    528   1.61    itojun 		goto bad;
    529   1.61    itojun 	}
    530   1.61    itojun 
    531   1.61    itojun 	error = 0;
    532   1.61    itojun 
    533   1.61    itojun 	/* check policy */
    534   1.61    itojun 	switch (sp->policy) {
    535   1.61    itojun 	case IPSEC_POLICY_DISCARD:
    536   1.61    itojun 		/*
    537   1.61    itojun 		 * This packet is just discarded.
    538   1.61    itojun 		 */
    539   1.61    itojun 		ipsecstat.out_polvio++;
    540   1.61    itojun 		goto bad;
    541   1.61    itojun 
    542   1.61    itojun 	case IPSEC_POLICY_BYPASS:
    543   1.61    itojun 	case IPSEC_POLICY_NONE:
    544   1.61    itojun 		/* no need to do IPsec. */
    545   1.61    itojun 		goto skip_ipsec;
    546   1.97    itojun 
    547   1.61    itojun 	case IPSEC_POLICY_IPSEC:
    548   1.61    itojun 		if (sp->req == NULL) {
    549   1.61    itojun 			/* XXX should be panic ? */
    550   1.61    itojun 			printf("ip_output: No IPsec request specified.\n");
    551   1.61    itojun 			error = EINVAL;
    552   1.61    itojun 			goto bad;
    553   1.61    itojun 		}
    554   1.61    itojun 		break;
    555   1.61    itojun 
    556   1.61    itojun 	case IPSEC_POLICY_ENTRUST:
    557   1.61    itojun 	default:
    558   1.61    itojun 		printf("ip_output: Invalid policy found. %d\n", sp->policy);
    559   1.61    itojun 	}
    560   1.61    itojun 
    561  1.141      manu #ifdef IPSEC_NAT_T
    562  1.141      manu 	/*
    563  1.144     perry 	 * NAT-T ESP fragmentation: don't do IPSec processing now,
    564  1.144     perry 	 * we'll do it on each fragmented packet.
    565  1.141      manu 	 */
    566  1.141      manu 	if (sp->req->sav &&
    567  1.141      manu 	    ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
    568  1.141      manu 	     (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
    569  1.141      manu 		if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
    570  1.141      manu 			natt_frag = 1;
    571  1.141      manu 			mtu = sp->req->sav->esp_frag;
    572  1.141      manu 			goto skip_ipsec;
    573  1.141      manu 		}
    574  1.141      manu 	}
    575  1.141      manu #endif /* IPSEC_NAT_T */
    576  1.141      manu 
    577   1.78   thorpej 	/*
    578   1.78   thorpej 	 * ipsec4_output() expects ip_len and ip_off in network
    579   1.78   thorpej 	 * order.  They have been set to network order above.
    580   1.78   thorpej 	 */
    581   1.61    itojun 
    582   1.61    itojun     {
    583   1.61    itojun 	struct ipsec_output_state state;
    584   1.61    itojun 	bzero(&state, sizeof(state));
    585   1.61    itojun 	state.m = m;
    586   1.61    itojun 	if (flags & IP_ROUTETOIF) {
    587   1.61    itojun 		state.ro = &iproute;
    588   1.61    itojun 		bzero(&iproute, sizeof(iproute));
    589   1.61    itojun 	} else
    590   1.61    itojun 		state.ro = ro;
    591   1.61    itojun 	state.dst = (struct sockaddr *)dst;
    592   1.61    itojun 
    593   1.86   thorpej 	/*
    594   1.86   thorpej 	 * We can't defer the checksum of payload data if
    595   1.86   thorpej 	 * we're about to encrypt/authenticate it.
    596   1.86   thorpej 	 *
    597   1.86   thorpej 	 * XXX When we support crypto offloading functions of
    598   1.86   thorpej 	 * XXX network interfaces, we need to reconsider this,
    599   1.86   thorpej 	 * XXX since it's likely that they'll support checksumming,
    600   1.86   thorpej 	 * XXX as well.
    601   1.86   thorpej 	 */
    602   1.86   thorpej 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
    603   1.86   thorpej 		in_delayed_cksum(m);
    604   1.86   thorpej 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
    605   1.86   thorpej 	}
    606   1.86   thorpej 
    607   1.61    itojun 	error = ipsec4_output(&state, sp, flags);
    608   1.61    itojun 
    609   1.61    itojun 	m = state.m;
    610   1.61    itojun 	if (flags & IP_ROUTETOIF) {
    611   1.61    itojun 		/*
    612   1.61    itojun 		 * if we have tunnel mode SA, we may need to ignore
    613   1.61    itojun 		 * IP_ROUTETOIF.
    614   1.61    itojun 		 */
    615   1.61    itojun 		if (state.ro != &iproute || state.ro->ro_rt != NULL) {
    616   1.61    itojun 			flags &= ~IP_ROUTETOIF;
    617   1.61    itojun 			ro = state.ro;
    618   1.61    itojun 		}
    619   1.61    itojun 	} else
    620   1.61    itojun 		ro = state.ro;
    621   1.61    itojun 	dst = (struct sockaddr_in *)state.dst;
    622   1.61    itojun 	if (error) {
    623   1.61    itojun 		/* mbuf is already reclaimed in ipsec4_output. */
    624   1.61    itojun 		m0 = NULL;
    625   1.61    itojun 		switch (error) {
    626   1.61    itojun 		case EHOSTUNREACH:
    627   1.61    itojun 		case ENETUNREACH:
    628   1.61    itojun 		case EMSGSIZE:
    629   1.61    itojun 		case ENOBUFS:
    630   1.61    itojun 		case ENOMEM:
    631   1.61    itojun 			break;
    632   1.61    itojun 		default:
    633   1.61    itojun 			printf("ip4_output (ipsec): error code %d\n", error);
    634   1.61    itojun 			/*fall through*/
    635   1.61    itojun 		case ENOENT:
    636   1.61    itojun 			/* don't show these error codes to the user */
    637   1.61    itojun 			error = 0;
    638   1.61    itojun 			break;
    639   1.61    itojun 		}
    640   1.61    itojun 		goto bad;
    641   1.61    itojun 	}
    642   1.61    itojun 
    643   1.61    itojun 	/* be sure to update variables that are affected by ipsec4_output() */
    644   1.61    itojun 	ip = mtod(m, struct ip *);
    645   1.61    itojun 	hlen = ip->ip_hl << 2;
    646   1.78   thorpej 	ip_len = ntohs(ip->ip_len);
    647   1.78   thorpej 
    648   1.61    itojun 	if (ro->ro_rt == NULL) {
    649   1.61    itojun 		if ((flags & IP_ROUTETOIF) == 0) {
    650   1.61    itojun 			printf("ip_output: "
    651   1.61    itojun 				"can't update route after IPsec processing\n");
    652   1.61    itojun 			error = EHOSTUNREACH;	/*XXX*/
    653   1.61    itojun 			goto bad;
    654   1.61    itojun 		}
    655   1.61    itojun 	} else {
    656   1.61    itojun 		/* nobody uses ia beyond here */
    657  1.133    itojun 		if (state.encap) {
    658   1.90    itojun 			ifp = ro->ro_rt->rt_ifp;
    659  1.133    itojun 			if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
    660  1.133    itojun 				mtu = ifp->if_mtu;
    661  1.133    itojun 		}
    662   1.61    itojun 	}
    663   1.90    itojun     }
    664   1.61    itojun skip_ipsec:
    665   1.61    itojun #endif /*IPSEC*/
    666  1.109  jonathan #ifdef FAST_IPSEC
    667  1.109  jonathan 	/*
    668  1.109  jonathan 	 * Check the security policy (SP) for the packet and, if
    669  1.109  jonathan 	 * required, do IPsec-related processing.  There are two
    670  1.109  jonathan 	 * cases here; the first time a packet is sent through
    671  1.109  jonathan 	 * it will be untagged and handled by ipsec4_checkpolicy.
    672  1.109  jonathan 	 * If the packet is resubmitted to ip_output (e.g. after
    673  1.109  jonathan 	 * AH, ESP, etc. processing), there will be a tag to bypass
    674  1.109  jonathan 	 * the lookup and related policy checking.
    675  1.109  jonathan 	 */
    676  1.109  jonathan 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
    677  1.109  jonathan 	s = splsoftnet();
    678  1.109  jonathan 	if (mtag != NULL) {
    679  1.109  jonathan 		tdbi = (struct tdb_ident *)(mtag + 1);
    680  1.109  jonathan 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
    681  1.109  jonathan 		if (sp == NULL)
    682  1.109  jonathan 			error = -EINVAL;	/* force silent drop */
    683  1.109  jonathan 		m_tag_delete(m, mtag);
    684  1.109  jonathan 	} else {
    685  1.130   thorpej 		if (inp != NULL &&
    686  1.130   thorpej 		    IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND))
    687  1.130   thorpej 			goto spd_done;
    688  1.109  jonathan 		sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
    689  1.109  jonathan 					&error, inp);
    690  1.109  jonathan 	}
    691  1.109  jonathan 	/*
    692  1.109  jonathan 	 * There are four return cases:
    693  1.109  jonathan 	 *    sp != NULL	 	    apply IPsec policy
    694  1.109  jonathan 	 *    sp == NULL, error == 0	    no IPsec handling needed
    695  1.109  jonathan 	 *    sp == NULL, error == -EINVAL  discard packet w/o error
    696  1.109  jonathan 	 *    sp == NULL, error != 0	    discard packet, report error
    697  1.109  jonathan 	 */
    698  1.109  jonathan 	if (sp != NULL) {
    699  1.160  christos #ifdef IPSEC_NAT_T
    700  1.160  christos 		/*
    701  1.160  christos 		 * NAT-T ESP fragmentation: don't do IPSec processing now,
    702  1.160  christos 		 * we'll do it on each fragmented packet.
    703  1.160  christos 		 */
    704  1.160  christos 		if (sp->req->sav &&
    705  1.160  christos 		    ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
    706  1.160  christos 		     (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
    707  1.160  christos 			if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
    708  1.160  christos 				natt_frag = 1;
    709  1.160  christos 				mtu = sp->req->sav->esp_frag;
    710  1.160  christos 				goto spd_done;
    711  1.160  christos 			}
    712  1.160  christos 		}
    713  1.160  christos #endif /* IPSEC_NAT_T */
    714  1.109  jonathan 		/* Loop detection, check if ipsec processing already done */
    715  1.109  jonathan 		IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
    716  1.109  jonathan 		for (mtag = m_tag_first(m); mtag != NULL;
    717  1.109  jonathan 		     mtag = m_tag_next(m, mtag)) {
    718  1.109  jonathan #ifdef MTAG_ABI_COMPAT
    719  1.109  jonathan 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
    720  1.109  jonathan 				continue;
    721  1.109  jonathan #endif
    722  1.109  jonathan 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
    723  1.109  jonathan 			    mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
    724  1.109  jonathan 				continue;
    725  1.109  jonathan 			/*
    726  1.109  jonathan 			 * Check if policy has an SA associated with it.
    727  1.109  jonathan 			 * This can happen when an SP has yet to acquire
    728  1.109  jonathan 			 * an SA; e.g. on first reference.  If it occurs,
    729  1.109  jonathan 			 * then we let ipsec4_process_packet do its thing.
    730  1.109  jonathan 			 */
    731  1.109  jonathan 			if (sp->req->sav == NULL)
    732  1.109  jonathan 				break;
    733  1.109  jonathan 			tdbi = (struct tdb_ident *)(mtag + 1);
    734  1.109  jonathan 			if (tdbi->spi == sp->req->sav->spi &&
    735  1.109  jonathan 			    tdbi->proto == sp->req->sav->sah->saidx.proto &&
    736  1.109  jonathan 			    bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
    737  1.109  jonathan 				 sizeof (union sockaddr_union)) == 0) {
    738  1.109  jonathan 				/*
    739  1.109  jonathan 				 * No IPsec processing is needed, free
    740  1.109  jonathan 				 * reference to SP.
    741  1.109  jonathan 				 *
    742  1.109  jonathan 				 * NB: null pointer to avoid free at
    743  1.109  jonathan 				 *     done: below.
    744  1.109  jonathan 				 */
    745  1.109  jonathan 				KEY_FREESP(&sp), sp = NULL;
    746  1.109  jonathan 				splx(s);
    747  1.109  jonathan 				goto spd_done;
    748  1.109  jonathan 			}
    749  1.109  jonathan 		}
    750  1.109  jonathan 
    751  1.109  jonathan 		/*
    752  1.109  jonathan 		 * Do delayed checksums now because we send before
    753  1.109  jonathan 		 * this is done in the normal processing path.
    754  1.109  jonathan 		 */
    755  1.109  jonathan 		if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
    756  1.109  jonathan 			in_delayed_cksum(m);
    757  1.109  jonathan 			m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
    758  1.109  jonathan 		}
    759  1.109  jonathan 
    760  1.109  jonathan #ifdef __FreeBSD__
    761  1.109  jonathan 		ip->ip_len = htons(ip->ip_len);
    762  1.109  jonathan 		ip->ip_off = htons(ip->ip_off);
    763  1.109  jonathan #endif
    764  1.109  jonathan 
    765  1.109  jonathan 		/* NB: callee frees mbuf */
    766  1.109  jonathan 		error = ipsec4_process_packet(m, sp->req, flags, 0);
    767  1.109  jonathan 		/*
    768  1.109  jonathan 		 * Preserve KAME behaviour: ENOENT can be returned
    769  1.109  jonathan 		 * when an SA acquire is in progress.  Don't propagate
    770  1.109  jonathan 		 * this to user-level; it confuses applications.
    771  1.109  jonathan 		 *
    772  1.109  jonathan 		 * XXX this will go away when the SADB is redone.
    773  1.109  jonathan 		 */
    774  1.109  jonathan 		if (error == ENOENT)
    775  1.109  jonathan 			error = 0;
    776  1.109  jonathan 		splx(s);
    777  1.109  jonathan 		goto done;
    778  1.109  jonathan 	} else {
    779  1.109  jonathan 		splx(s);
    780  1.109  jonathan 
    781  1.109  jonathan 		if (error != 0) {
    782  1.109  jonathan 			/*
    783  1.109  jonathan 			 * Hack: -EINVAL is used to signal that a packet
    784  1.109  jonathan 			 * should be silently discarded.  This is typically
    785  1.109  jonathan 			 * because we asked key management for an SA and
    786  1.109  jonathan 			 * it was delayed (e.g. kicked up to IKE).
    787  1.109  jonathan 			 */
    788  1.109  jonathan 			if (error == -EINVAL)
    789  1.109  jonathan 				error = 0;
    790  1.109  jonathan 			goto bad;
    791  1.109  jonathan 		} else {
    792  1.109  jonathan 			/* No IPsec processing for this packet. */
    793  1.109  jonathan 		}
    794  1.109  jonathan #ifdef notyet
    795  1.109  jonathan 		/*
    796  1.109  jonathan 		 * If deferred crypto processing is needed, check that
    797  1.109  jonathan 		 * the interface supports it.
    798  1.144     perry 		 */
    799  1.109  jonathan 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
    800  1.109  jonathan 		if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
    801  1.109  jonathan 			/* notify IPsec to do its own crypto */
    802  1.109  jonathan 			ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
    803  1.109  jonathan 			error = EHOSTUNREACH;
    804  1.109  jonathan 			goto bad;
    805  1.109  jonathan 		}
    806  1.109  jonathan #endif
    807  1.109  jonathan 	}
    808  1.109  jonathan spd_done:
    809  1.109  jonathan #endif /* FAST_IPSEC */
    810   1.61    itojun 
    811   1.82    itojun #ifdef PFIL_HOOKS
    812   1.82    itojun 	/*
    813   1.82    itojun 	 * Run through list of hooks for output packets.
    814   1.82    itojun 	 */
    815  1.106    itojun 	if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
    816   1.82    itojun 		goto done;
    817   1.82    itojun 	if (m == NULL)
    818   1.82    itojun 		goto done;
    819   1.82    itojun 
    820   1.82    itojun 	ip = mtod(m, struct ip *);
    821  1.106    itojun 	hlen = ip->ip_hl << 2;
    822  1.173      yamt 	ip_len = ntohs(ip->ip_len);
    823   1.82    itojun #endif /* PFIL_HOOKS */
    824   1.82    itojun 
    825  1.146      matt 	m->m_pkthdr.csum_data |= hlen << 16;
    826  1.146      matt 
    827  1.136   thorpej #if IFA_STATS
    828  1.136   thorpej 	/*
    829  1.136   thorpej 	 * search for the source address structure to
    830  1.136   thorpej 	 * maintain output statistics.
    831  1.136   thorpej 	 */
    832  1.136   thorpej 	INADDR_TO_IA(ip->ip_src, ia);
    833  1.136   thorpej #endif
    834  1.136   thorpej 
    835  1.138   thorpej 	/* Maybe skip checksums on loopback interfaces. */
    836  1.151      yamt 	if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
    837  1.138   thorpej 		m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
    838  1.151      yamt 	}
    839  1.104      yamt 	sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
    840    1.1       cgd 	/*
    841  1.147      matt 	 * If small enough for mtu of path, or if using TCP segmentation
    842  1.147      matt 	 * offload, can just send directly.
    843    1.1       cgd 	 */
    844  1.147      matt 	if (ip_len <= mtu ||
    845  1.147      matt 	    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
    846   1.63    itojun #if IFA_STATS
    847   1.63    itojun 		if (ia)
    848   1.78   thorpej 			ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
    849   1.63    itojun #endif
    850   1.86   thorpej 		/*
    851   1.86   thorpej 		 * Always initialize the sum to 0!  Some HW assisted
    852   1.86   thorpej 		 * checksumming requires this.
    853   1.86   thorpej 		 */
    854    1.1       cgd 		ip->ip_sum = 0;
    855   1.86   thorpej 
    856  1.149      matt 		if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
    857  1.147      matt 			/*
    858  1.147      matt 			 * Perform any checksums that the hardware can't do
    859  1.147      matt 			 * for us.
    860  1.147      matt 			 *
    861  1.147      matt 			 * XXX Does any hardware require the {th,uh}_sum
    862  1.147      matt 			 * XXX fields to be 0?
    863  1.147      matt 			 */
    864  1.147      matt 			if (sw_csum & M_CSUM_IPv4) {
    865  1.151      yamt 				KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
    866  1.147      matt 				ip->ip_sum = in_cksum(m, hlen);
    867  1.147      matt 				m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
    868  1.147      matt 			}
    869  1.147      matt 			if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
    870  1.151      yamt 				if (IN_NEED_CHECKSUM(ifp,
    871  1.151      yamt 				    sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
    872  1.151      yamt 					in_delayed_cksum(m);
    873  1.151      yamt 				}
    874  1.147      matt 				m->m_pkthdr.csum_flags &=
    875  1.147      matt 				    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
    876  1.147      matt 			}
    877  1.146      matt 		}
    878   1.86   thorpej 
    879   1.82    itojun #ifdef IPSEC
    880   1.82    itojun 		/* clean ipsec history once it goes out of the node */
    881   1.82    itojun 		ipsec_delaux(m);
    882   1.82    itojun #endif
    883  1.152      yamt 
    884  1.152      yamt 		if (__predict_true(
    885  1.152      yamt 		    (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 ||
    886  1.152      yamt 		    (ifp->if_capenable & IFCAP_TSOv4) != 0)) {
    887  1.152      yamt 			error =
    888  1.152      yamt 			    (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
    889  1.152      yamt 		} else {
    890  1.152      yamt 			error =
    891  1.152      yamt 			    ip_tso_output(ifp, m, sintosa(dst), ro->ro_rt);
    892  1.152      yamt 		}
    893    1.1       cgd 		goto done;
    894    1.1       cgd 	}
    895   1.61    itojun 
    896    1.1       cgd 	/*
    897   1.86   thorpej 	 * We can't use HW checksumming if we're about to
    898   1.86   thorpej 	 * to fragment the packet.
    899   1.86   thorpej 	 *
    900   1.86   thorpej 	 * XXX Some hardware can do this.
    901   1.86   thorpej 	 */
    902   1.86   thorpej 	if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
    903  1.151      yamt 		if (IN_NEED_CHECKSUM(ifp,
    904  1.151      yamt 		    m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
    905  1.151      yamt 			in_delayed_cksum(m);
    906  1.151      yamt 		}
    907   1.86   thorpej 		m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
    908   1.86   thorpej 	}
    909   1.86   thorpej 
    910   1.86   thorpej 	/*
    911    1.1       cgd 	 * Too large for interface; fragment if possible.
    912    1.1       cgd 	 * Must be able to put at least 8 bytes per fragment.
    913    1.1       cgd 	 */
    914  1.100    itojun 	if (ntohs(ip->ip_off) & IP_DF) {
    915   1.40      matt 		if (flags & IP_RETURNMTU)
    916   1.48      matt 			*mtu_p = mtu;
    917    1.1       cgd 		error = EMSGSIZE;
    918   1.18   mycroft 		ipstat.ips_cantfrag++;
    919    1.1       cgd 		goto bad;
    920    1.1       cgd 	}
    921  1.110    itojun 
    922  1.110    itojun 	error = ip_fragment(m, ifp, mtu);
    923  1.124    itojun 	if (error) {
    924  1.124    itojun 		m = NULL;
    925    1.1       cgd 		goto bad;
    926  1.124    itojun 	}
    927  1.110    itojun 
    928  1.119    itojun 	for (; m; m = m0) {
    929  1.110    itojun 		m0 = m->m_nextpkt;
    930  1.110    itojun 		m->m_nextpkt = 0;
    931  1.110    itojun 		if (error == 0) {
    932  1.110    itojun #if IFA_STATS
    933  1.136   thorpej 			if (ia)
    934  1.110    itojun 				ia->ia_ifa.ifa_data.ifad_outbytes +=
    935  1.110    itojun 				    ntohs(ip->ip_len);
    936  1.110    itojun #endif
    937  1.110    itojun #ifdef IPSEC
    938  1.110    itojun 			/* clean ipsec history once it goes out of the node */
    939  1.110    itojun 			ipsec_delaux(m);
    940  1.160  christos #endif /* IPSEC */
    941  1.141      manu 
    942  1.141      manu #ifdef IPSEC_NAT_T
    943  1.144     perry 			/*
    944  1.141      manu 			 * If we get there, the packet has not been handeld by
    945  1.144     perry 			 * IPSec whereas it should have. Now that it has been
    946  1.141      manu 			 * fragmented, re-inject it in ip_output so that IPsec
    947  1.141      manu 			 * processing can occur.
    948  1.141      manu 			 */
    949  1.141      manu 			if (natt_frag) {
    950  1.144     perry 				error = ip_output(m, opt,
    951  1.141      manu 				    ro, flags, imo, so, mtu_p);
    952  1.144     perry 			} else
    953  1.141      manu #endif /* IPSEC_NAT_T */
    954  1.141      manu 			{
    955  1.141      manu 				KASSERT((m->m_pkthdr.csum_flags &
    956  1.141      manu 				    (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
    957  1.141      manu 				error = (*ifp->if_output)(ifp, m, sintosa(dst),
    958  1.141      manu 				    ro->ro_rt);
    959  1.141      manu 			}
    960  1.110    itojun 		} else
    961  1.110    itojun 			m_freem(m);
    962    1.1       cgd 	}
    963    1.1       cgd 
    964  1.110    itojun 	if (error == 0)
    965  1.110    itojun 		ipstat.ips_fragmented++;
    966  1.110    itojun done:
    967  1.174     joerg 	rtcache_free(&iproute);
    968  1.110    itojun 
    969  1.110    itojun #ifdef IPSEC
    970  1.110    itojun 	if (sp != NULL) {
    971  1.110    itojun 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
    972  1.110    itojun 			printf("DP ip_output call free SP:%p\n", sp));
    973  1.110    itojun 		key_freesp(sp);
    974  1.110    itojun 	}
    975  1.110    itojun #endif /* IPSEC */
    976  1.110    itojun #ifdef FAST_IPSEC
    977  1.110    itojun 	if (sp != NULL)
    978  1.110    itojun 		KEY_FREESP(&sp);
    979  1.110    itojun #endif /* FAST_IPSEC */
    980  1.110    itojun 
    981  1.110    itojun 	return (error);
    982  1.110    itojun bad:
    983  1.110    itojun 	m_freem(m);
    984  1.110    itojun 	goto done;
    985  1.110    itojun }
    986  1.110    itojun 
    987  1.113    itojun int
    988  1.110    itojun ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
    989  1.110    itojun {
    990  1.110    itojun 	struct ip *ip, *mhip;
    991  1.110    itojun 	struct mbuf *m0;
    992  1.110    itojun 	int len, hlen, off;
    993  1.110    itojun 	int mhlen, firstlen;
    994  1.110    itojun 	struct mbuf **mnext;
    995  1.135      manu 	int sw_csum = m->m_pkthdr.csum_flags;
    996   1.48      matt 	int fragments = 0;
    997   1.48      matt 	int s;
    998  1.110    itojun 	int error = 0;
    999  1.110    itojun 
   1000  1.110    itojun 	ip = mtod(m, struct ip *);
   1001  1.110    itojun 	hlen = ip->ip_hl << 2;
   1002  1.135      manu 	if (ifp != NULL)
   1003  1.135      manu 		sw_csum &= ~ifp->if_csum_flags_tx;
   1004  1.110    itojun 
   1005  1.110    itojun 	len = (mtu - hlen) &~ 7;
   1006  1.124    itojun 	if (len < 8) {
   1007  1.124    itojun 		m_freem(m);
   1008  1.110    itojun 		return (EMSGSIZE);
   1009  1.124    itojun 	}
   1010  1.110    itojun 
   1011  1.110    itojun 	firstlen = len;
   1012  1.110    itojun 	mnext = &m->m_nextpkt;
   1013    1.1       cgd 
   1014    1.1       cgd 	/*
   1015    1.1       cgd 	 * Loop through length of segment after first fragment,
   1016    1.1       cgd 	 * make new header and copy data of each part and link onto chain.
   1017    1.1       cgd 	 */
   1018    1.1       cgd 	m0 = m;
   1019    1.1       cgd 	mhlen = sizeof (struct ip);
   1020  1.100    itojun 	for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
   1021    1.1       cgd 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
   1022    1.1       cgd 		if (m == 0) {
   1023    1.1       cgd 			error = ENOBUFS;
   1024   1.18   mycroft 			ipstat.ips_odropped++;
   1025    1.1       cgd 			goto sendorfree;
   1026    1.1       cgd 		}
   1027  1.103      matt 		MCLAIM(m, m0->m_owner);
   1028   1.22       cgd 		*mnext = m;
   1029   1.22       cgd 		mnext = &m->m_nextpkt;
   1030    1.1       cgd 		m->m_data += max_linkhdr;
   1031    1.1       cgd 		mhip = mtod(m, struct ip *);
   1032    1.1       cgd 		*mhip = *ip;
   1033   1.73        is 		/* we must inherit MCAST and BCAST flags */
   1034   1.73        is 		m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
   1035    1.1       cgd 		if (hlen > sizeof (struct ip)) {
   1036    1.1       cgd 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
   1037    1.1       cgd 			mhip->ip_hl = mhlen >> 2;
   1038    1.1       cgd 		}
   1039    1.1       cgd 		m->m_len = mhlen;
   1040  1.122    itojun 		mhip->ip_off = ((off - hlen) >> 3) +
   1041  1.122    itojun 		    (ntohs(ip->ip_off) & ~IP_MF);
   1042  1.122    itojun 		if (ip->ip_off & htons(IP_MF))
   1043    1.1       cgd 			mhip->ip_off |= IP_MF;
   1044  1.100    itojun 		if (off + len >= ntohs(ip->ip_len))
   1045  1.100    itojun 			len = ntohs(ip->ip_len) - off;
   1046    1.1       cgd 		else
   1047    1.1       cgd 			mhip->ip_off |= IP_MF;
   1048  1.100    itojun 		HTONS(mhip->ip_off);
   1049   1.21       cgd 		mhip->ip_len = htons((u_int16_t)(len + mhlen));
   1050    1.1       cgd 		m->m_next = m_copy(m0, off, len);
   1051    1.1       cgd 		if (m->m_next == 0) {
   1052    1.1       cgd 			error = ENOBUFS;	/* ??? */
   1053   1.18   mycroft 			ipstat.ips_odropped++;
   1054    1.1       cgd 			goto sendorfree;
   1055    1.1       cgd 		}
   1056    1.1       cgd 		m->m_pkthdr.len = mhlen + len;
   1057    1.1       cgd 		m->m_pkthdr.rcvif = (struct ifnet *)0;
   1058    1.1       cgd 		mhip->ip_sum = 0;
   1059  1.104      yamt 		if (sw_csum & M_CSUM_IPv4) {
   1060  1.104      yamt 			mhip->ip_sum = in_cksum(m, mhlen);
   1061  1.104      yamt 			KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
   1062  1.104      yamt 		} else {
   1063  1.104      yamt 			m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
   1064  1.148   thorpej 			m->m_pkthdr.csum_data |= mhlen << 16;
   1065  1.104      yamt 		}
   1066    1.1       cgd 		ipstat.ips_ofragments++;
   1067   1.48      matt 		fragments++;
   1068    1.1       cgd 	}
   1069    1.1       cgd 	/*
   1070    1.1       cgd 	 * Update first fragment by trimming what's been copied out
   1071    1.1       cgd 	 * and updating header, then send each fragment (in order).
   1072    1.1       cgd 	 */
   1073    1.1       cgd 	m = m0;
   1074  1.100    itojun 	m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
   1075    1.1       cgd 	m->m_pkthdr.len = hlen + firstlen;
   1076   1.21       cgd 	ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
   1077  1.100    itojun 	ip->ip_off |= htons(IP_MF);
   1078    1.1       cgd 	ip->ip_sum = 0;
   1079  1.104      yamt 	if (sw_csum & M_CSUM_IPv4) {
   1080  1.104      yamt 		ip->ip_sum = in_cksum(m, hlen);
   1081  1.104      yamt 		m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
   1082  1.104      yamt 	} else {
   1083  1.104      yamt 		KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
   1084  1.148   thorpej 		KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
   1085  1.148   thorpej 			sizeof(struct ip));
   1086  1.104      yamt 	}
   1087    1.1       cgd sendorfree:
   1088   1.48      matt 	/*
   1089   1.48      matt 	 * If there is no room for all the fragments, don't queue
   1090   1.48      matt 	 * any of them.
   1091   1.48      matt 	 */
   1092  1.135      manu 	if (ifp != NULL) {
   1093  1.135      manu 		s = splnet();
   1094  1.135      manu 		if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
   1095  1.135      manu 		    error == 0) {
   1096  1.135      manu 			error = ENOBUFS;
   1097  1.135      manu 			ipstat.ips_odropped++;
   1098  1.135      manu 			IFQ_INC_DROPS(&ifp->if_snd);
   1099  1.135      manu 		}
   1100  1.135      manu 		splx(s);
   1101  1.126     enami 	}
   1102  1.124    itojun 	if (error) {
   1103  1.125    itojun 		for (m = m0; m; m = m0) {
   1104  1.124    itojun 			m0 = m->m_nextpkt;
   1105  1.124    itojun 			m->m_nextpkt = NULL;
   1106  1.124    itojun 			m_freem(m);
   1107  1.124    itojun 		}
   1108  1.124    itojun 	}
   1109    1.1       cgd 	return (error);
   1110   1.86   thorpej }
   1111   1.86   thorpej 
   1112   1.86   thorpej /*
   1113   1.86   thorpej  * Process a delayed payload checksum calculation.
   1114   1.86   thorpej  */
   1115   1.86   thorpej void
   1116   1.86   thorpej in_delayed_cksum(struct mbuf *m)
   1117   1.86   thorpej {
   1118   1.86   thorpej 	struct ip *ip;
   1119   1.86   thorpej 	u_int16_t csum, offset;
   1120   1.86   thorpej 
   1121   1.86   thorpej 	ip = mtod(m, struct ip *);
   1122   1.86   thorpej 	offset = ip->ip_hl << 2;
   1123   1.86   thorpej 	csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
   1124   1.86   thorpej 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
   1125   1.86   thorpej 		csum = 0xffff;
   1126   1.86   thorpej 
   1127  1.145    briggs 	offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
   1128   1.86   thorpej 
   1129   1.86   thorpej 	if ((offset + sizeof(u_int16_t)) > m->m_len) {
   1130   1.87      yamt 		/* This happen when ip options were inserted
   1131   1.86   thorpej 		printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
   1132   1.86   thorpej 		    m->m_len, offset, ip->ip_p);
   1133   1.87      yamt 		 */
   1134   1.86   thorpej 		m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
   1135   1.86   thorpej 	} else
   1136   1.86   thorpej 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
   1137    1.1       cgd }
   1138   1.47       kml 
   1139   1.47       kml /*
   1140   1.47       kml  * Determine the maximum length of the options to be inserted;
   1141   1.47       kml  * we would far rather allocate too much space rather than too little.
   1142   1.47       kml  */
   1143   1.47       kml 
   1144   1.47       kml u_int
   1145  1.140     perry ip_optlen(struct inpcb *inp)
   1146   1.47       kml {
   1147   1.47       kml 	struct mbuf *m = inp->inp_options;
   1148   1.47       kml 
   1149   1.47       kml 	if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
   1150  1.101    itojun 		return (m->m_len - offsetof(struct ipoption, ipopt_dst));
   1151   1.47       kml 	else
   1152   1.47       kml 		return 0;
   1153   1.47       kml }
   1154   1.47       kml 
   1155    1.1       cgd 
   1156    1.1       cgd /*
   1157    1.1       cgd  * Insert IP options into preformed packet.
   1158    1.1       cgd  * Adjust IP destination as required for IP source routing,
   1159    1.1       cgd  * as indicated by a non-zero in_addr at the start of the options.
   1160    1.1       cgd  */
   1161   1.12   mycroft static struct mbuf *
   1162  1.140     perry ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
   1163    1.1       cgd {
   1164   1.71  augustss 	struct ipoption *p = mtod(opt, struct ipoption *);
   1165    1.1       cgd 	struct mbuf *n;
   1166   1.71  augustss 	struct ip *ip = mtod(m, struct ip *);
   1167    1.1       cgd 	unsigned optlen;
   1168    1.1       cgd 
   1169    1.1       cgd 	optlen = opt->m_len - sizeof(p->ipopt_dst);
   1170  1.100    itojun 	if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
   1171    1.1       cgd 		return (m);		/* XXX should fail */
   1172   1.31   mycroft 	if (!in_nullhost(p->ipopt_dst))
   1173    1.1       cgd 		ip->ip_dst = p->ipopt_dst;
   1174  1.123    itojun 	if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
   1175    1.1       cgd 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
   1176    1.1       cgd 		if (n == 0)
   1177    1.1       cgd 			return (m);
   1178  1.103      matt 		MCLAIM(n, m->m_owner);
   1179  1.155      yamt 		M_MOVE_PKTHDR(n, m);
   1180    1.1       cgd 		m->m_len -= sizeof(struct ip);
   1181    1.1       cgd 		m->m_data += sizeof(struct ip);
   1182    1.1       cgd 		n->m_next = m;
   1183    1.1       cgd 		m = n;
   1184    1.1       cgd 		m->m_len = optlen + sizeof(struct ip);
   1185    1.1       cgd 		m->m_data += max_linkhdr;
   1186    1.1       cgd 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
   1187    1.1       cgd 	} else {
   1188    1.1       cgd 		m->m_data -= optlen;
   1189    1.1       cgd 		m->m_len += optlen;
   1190   1.57     perry 		memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
   1191    1.1       cgd 	}
   1192   1.87      yamt 	m->m_pkthdr.len += optlen;
   1193    1.1       cgd 	ip = mtod(m, struct ip *);
   1194    1.1       cgd 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
   1195    1.1       cgd 	*phlen = sizeof(struct ip) + optlen;
   1196  1.100    itojun 	ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
   1197    1.1       cgd 	return (m);
   1198    1.1       cgd }
   1199    1.1       cgd 
   1200    1.1       cgd /*
   1201    1.1       cgd  * Copy options from ip to jp,
   1202    1.1       cgd  * omitting those not copied during fragmentation.
   1203    1.1       cgd  */
   1204   1.12   mycroft int
   1205  1.140     perry ip_optcopy(struct ip *ip, struct ip *jp)
   1206    1.1       cgd {
   1207   1.71  augustss 	u_char *cp, *dp;
   1208    1.1       cgd 	int opt, optlen, cnt;
   1209    1.1       cgd 
   1210    1.1       cgd 	cp = (u_char *)(ip + 1);
   1211    1.1       cgd 	dp = (u_char *)(jp + 1);
   1212    1.1       cgd 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
   1213    1.1       cgd 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
   1214    1.1       cgd 		opt = cp[0];
   1215    1.1       cgd 		if (opt == IPOPT_EOL)
   1216    1.1       cgd 			break;
   1217   1.18   mycroft 		if (opt == IPOPT_NOP) {
   1218   1.18   mycroft 			/* Preserve for IP mcast tunnel's LSRR alignment. */
   1219   1.18   mycroft 			*dp++ = IPOPT_NOP;
   1220    1.1       cgd 			optlen = 1;
   1221   1.18   mycroft 			continue;
   1222   1.74    itojun 		}
   1223   1.74    itojun #ifdef DIAGNOSTIC
   1224   1.74    itojun 		if (cnt < IPOPT_OLEN + sizeof(*cp))
   1225   1.74    itojun 			panic("malformed IPv4 option passed to ip_optcopy");
   1226   1.74    itojun #endif
   1227   1.74    itojun 		optlen = cp[IPOPT_OLEN];
   1228   1.74    itojun #ifdef DIAGNOSTIC
   1229   1.74    itojun 		if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
   1230   1.74    itojun 			panic("malformed IPv4 option passed to ip_optcopy");
   1231   1.74    itojun #endif
   1232    1.1       cgd 		/* bogus lengths should have been caught by ip_dooptions */
   1233    1.1       cgd 		if (optlen > cnt)
   1234    1.1       cgd 			optlen = cnt;
   1235    1.1       cgd 		if (IPOPT_COPIED(opt)) {
   1236    1.1       cgd 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
   1237    1.1       cgd 			dp += optlen;
   1238    1.1       cgd 		}
   1239    1.1       cgd 	}
   1240    1.1       cgd 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
   1241    1.1       cgd 		*dp++ = IPOPT_EOL;
   1242    1.1       cgd 	return (optlen);
   1243    1.1       cgd }
   1244    1.1       cgd 
   1245    1.1       cgd /*
   1246    1.1       cgd  * IP socket option processing.
   1247    1.1       cgd  */
   1248   1.12   mycroft int
   1249  1.140     perry ip_ctloutput(int op, struct socket *so, int level, int optname,
   1250  1.140     perry     struct mbuf **mp)
   1251    1.1       cgd {
   1252   1.71  augustss 	struct inpcb *inp = sotoinpcb(so);
   1253   1.71  augustss 	struct mbuf *m = *mp;
   1254   1.71  augustss 	int optval = 0;
   1255    1.1       cgd 	int error = 0;
   1256  1.109  jonathan #if defined(IPSEC) || defined(FAST_IPSEC)
   1257  1.165        ad 	struct lwp *l = curlwp;	/*XXX*/
   1258   1.61    itojun #endif
   1259    1.1       cgd 
   1260   1.18   mycroft 	if (level != IPPROTO_IP) {
   1261    1.1       cgd 		error = EINVAL;
   1262   1.18   mycroft 		if (op == PRCO_SETOPT && *mp)
   1263   1.18   mycroft 			(void) m_free(*mp);
   1264   1.18   mycroft 	} else switch (op) {
   1265    1.1       cgd 
   1266    1.1       cgd 	case PRCO_SETOPT:
   1267    1.1       cgd 		switch (optname) {
   1268    1.1       cgd 		case IP_OPTIONS:
   1269    1.1       cgd #ifdef notyet
   1270    1.1       cgd 		case IP_RETOPTS:
   1271    1.1       cgd 			return (ip_pcbopts(optname, &inp->inp_options, m));
   1272    1.1       cgd #else
   1273    1.1       cgd 			return (ip_pcbopts(&inp->inp_options, m));
   1274    1.1       cgd #endif
   1275    1.1       cgd 
   1276    1.1       cgd 		case IP_TOS:
   1277    1.1       cgd 		case IP_TTL:
   1278    1.1       cgd 		case IP_RECVOPTS:
   1279    1.1       cgd 		case IP_RECVRETOPTS:
   1280    1.1       cgd 		case IP_RECVDSTADDR:
   1281   1.37   thorpej 		case IP_RECVIF:
   1282   1.27       cgd 			if (m == NULL || m->m_len != sizeof(int))
   1283    1.1       cgd 				error = EINVAL;
   1284    1.1       cgd 			else {
   1285    1.1       cgd 				optval = *mtod(m, int *);
   1286    1.1       cgd 				switch (optname) {
   1287    1.1       cgd 
   1288    1.1       cgd 				case IP_TOS:
   1289    1.1       cgd 					inp->inp_ip.ip_tos = optval;
   1290    1.1       cgd 					break;
   1291    1.1       cgd 
   1292    1.1       cgd 				case IP_TTL:
   1293    1.1       cgd 					inp->inp_ip.ip_ttl = optval;
   1294    1.1       cgd 					break;
   1295    1.1       cgd #define	OPTSET(bit) \
   1296    1.1       cgd 	if (optval) \
   1297    1.1       cgd 		inp->inp_flags |= bit; \
   1298    1.1       cgd 	else \
   1299    1.1       cgd 		inp->inp_flags &= ~bit;
   1300    1.1       cgd 
   1301    1.1       cgd 				case IP_RECVOPTS:
   1302    1.1       cgd 					OPTSET(INP_RECVOPTS);
   1303    1.1       cgd 					break;
   1304    1.1       cgd 
   1305    1.1       cgd 				case IP_RECVRETOPTS:
   1306    1.1       cgd 					OPTSET(INP_RECVRETOPTS);
   1307    1.1       cgd 					break;
   1308    1.1       cgd 
   1309    1.1       cgd 				case IP_RECVDSTADDR:
   1310    1.1       cgd 					OPTSET(INP_RECVDSTADDR);
   1311    1.1       cgd 					break;
   1312   1.37   thorpej 
   1313   1.37   thorpej 				case IP_RECVIF:
   1314   1.37   thorpej 					OPTSET(INP_RECVIF);
   1315   1.37   thorpej 					break;
   1316    1.1       cgd 				}
   1317    1.1       cgd 			}
   1318    1.1       cgd 			break;
   1319    1.1       cgd #undef OPTSET
   1320   1.18   mycroft 
   1321   1.18   mycroft 		case IP_MULTICAST_IF:
   1322   1.18   mycroft 		case IP_MULTICAST_TTL:
   1323   1.18   mycroft 		case IP_MULTICAST_LOOP:
   1324   1.18   mycroft 		case IP_ADD_MEMBERSHIP:
   1325   1.18   mycroft 		case IP_DROP_MEMBERSHIP:
   1326   1.18   mycroft 			error = ip_setmoptions(optname, &inp->inp_moptions, m);
   1327   1.18   mycroft 			break;
   1328    1.1       cgd 
   1329   1.41     lukem 		case IP_PORTRANGE:
   1330   1.41     lukem 			if (m == 0 || m->m_len != sizeof(int))
   1331   1.41     lukem 				error = EINVAL;
   1332   1.41     lukem 			else {
   1333   1.41     lukem 				optval = *mtod(m, int *);
   1334   1.41     lukem 
   1335   1.41     lukem 				switch (optval) {
   1336   1.41     lukem 
   1337   1.41     lukem 				case IP_PORTRANGE_DEFAULT:
   1338   1.41     lukem 				case IP_PORTRANGE_HIGH:
   1339   1.41     lukem 					inp->inp_flags &= ~(INP_LOWPORT);
   1340   1.41     lukem 					break;
   1341   1.41     lukem 
   1342   1.41     lukem 				case IP_PORTRANGE_LOW:
   1343   1.41     lukem 					inp->inp_flags |= INP_LOWPORT;
   1344   1.41     lukem 					break;
   1345   1.41     lukem 
   1346   1.41     lukem 				default:
   1347   1.41     lukem 					error = EINVAL;
   1348   1.41     lukem 					break;
   1349   1.41     lukem 				}
   1350   1.41     lukem 			}
   1351   1.41     lukem 			break;
   1352   1.41     lukem 
   1353  1.109  jonathan #if defined(IPSEC) || defined(FAST_IPSEC)
   1354   1.61    itojun 		case IP_IPSEC_POLICY:
   1355   1.66    itojun 		{
   1356   1.61    itojun 			caddr_t req = NULL;
   1357   1.66    itojun 			size_t len = 0;
   1358   1.61    itojun 			int priv = 0;
   1359   1.66    itojun 
   1360   1.61    itojun #ifdef __NetBSD__
   1361  1.165        ad 			if (l == 0 || kauth_authorize_generic(l->l_cred,
   1362  1.172      elad 			    KAUTH_GENERIC_ISSUSER, NULL))
   1363   1.61    itojun 				priv = 0;
   1364   1.61    itojun 			else
   1365   1.61    itojun 				priv = 1;
   1366   1.61    itojun #else
   1367   1.61    itojun 			priv = (in6p->in6p_socket->so_state & SS_PRIV);
   1368   1.61    itojun #endif
   1369   1.66    itojun 			if (m) {
   1370   1.61    itojun 				req = mtod(m, caddr_t);
   1371   1.61    itojun 				len = m->m_len;
   1372   1.61    itojun 			}
   1373   1.66    itojun 			error = ipsec4_set_policy(inp, optname, req, len, priv);
   1374   1.61    itojun 			break;
   1375   1.61    itojun 		    }
   1376   1.61    itojun #endif /*IPSEC*/
   1377   1.61    itojun 
   1378    1.1       cgd 		default:
   1379   1.18   mycroft 			error = ENOPROTOOPT;
   1380    1.1       cgd 			break;
   1381    1.1       cgd 		}
   1382    1.1       cgd 		if (m)
   1383    1.1       cgd 			(void)m_free(m);
   1384    1.1       cgd 		break;
   1385    1.1       cgd 
   1386    1.1       cgd 	case PRCO_GETOPT:
   1387    1.1       cgd 		switch (optname) {
   1388    1.1       cgd 		case IP_OPTIONS:
   1389    1.1       cgd 		case IP_RETOPTS:
   1390    1.1       cgd 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1391  1.103      matt 			MCLAIM(m, so->so_mowner);
   1392    1.1       cgd 			if (inp->inp_options) {
   1393    1.1       cgd 				m->m_len = inp->inp_options->m_len;
   1394    1.1       cgd 				bcopy(mtod(inp->inp_options, caddr_t),
   1395    1.1       cgd 				    mtod(m, caddr_t), (unsigned)m->m_len);
   1396    1.1       cgd 			} else
   1397    1.1       cgd 				m->m_len = 0;
   1398    1.1       cgd 			break;
   1399    1.1       cgd 
   1400    1.1       cgd 		case IP_TOS:
   1401    1.1       cgd 		case IP_TTL:
   1402    1.1       cgd 		case IP_RECVOPTS:
   1403    1.1       cgd 		case IP_RECVRETOPTS:
   1404    1.1       cgd 		case IP_RECVDSTADDR:
   1405   1.37   thorpej 		case IP_RECVIF:
   1406   1.40      matt 		case IP_ERRORMTU:
   1407    1.1       cgd 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1408  1.103      matt 			MCLAIM(m, so->so_mowner);
   1409    1.1       cgd 			m->m_len = sizeof(int);
   1410    1.1       cgd 			switch (optname) {
   1411    1.1       cgd 
   1412    1.1       cgd 			case IP_TOS:
   1413    1.1       cgd 				optval = inp->inp_ip.ip_tos;
   1414    1.1       cgd 				break;
   1415    1.1       cgd 
   1416    1.1       cgd 			case IP_TTL:
   1417    1.1       cgd 				optval = inp->inp_ip.ip_ttl;
   1418   1.40      matt 				break;
   1419   1.40      matt 
   1420   1.40      matt 			case IP_ERRORMTU:
   1421   1.40      matt 				optval = inp->inp_errormtu;
   1422    1.1       cgd 				break;
   1423    1.1       cgd 
   1424    1.1       cgd #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
   1425    1.1       cgd 
   1426    1.1       cgd 			case IP_RECVOPTS:
   1427    1.1       cgd 				optval = OPTBIT(INP_RECVOPTS);
   1428    1.1       cgd 				break;
   1429    1.1       cgd 
   1430    1.1       cgd 			case IP_RECVRETOPTS:
   1431    1.1       cgd 				optval = OPTBIT(INP_RECVRETOPTS);
   1432    1.1       cgd 				break;
   1433    1.1       cgd 
   1434    1.1       cgd 			case IP_RECVDSTADDR:
   1435    1.1       cgd 				optval = OPTBIT(INP_RECVDSTADDR);
   1436   1.37   thorpej 				break;
   1437   1.37   thorpej 
   1438   1.37   thorpej 			case IP_RECVIF:
   1439   1.37   thorpej 				optval = OPTBIT(INP_RECVIF);
   1440    1.1       cgd 				break;
   1441    1.1       cgd 			}
   1442    1.1       cgd 			*mtod(m, int *) = optval;
   1443    1.1       cgd 			break;
   1444   1.61    itojun 
   1445  1.134   minoura #if 0	/* defined(IPSEC) || defined(FAST_IPSEC) */
   1446  1.134   minoura 		/* XXX: code broken */
   1447   1.61    itojun 		case IP_IPSEC_POLICY:
   1448   1.66    itojun 		{
   1449   1.66    itojun 			caddr_t req = NULL;
   1450   1.80    itojun 			size_t len = 0;
   1451   1.66    itojun 
   1452   1.66    itojun 			if (m) {
   1453   1.66    itojun 				req = mtod(m, caddr_t);
   1454   1.66    itojun 				len = m->m_len;
   1455   1.66    itojun 			}
   1456   1.66    itojun 			error = ipsec4_get_policy(inp, req, len, mp);
   1457   1.61    itojun 			break;
   1458   1.66    itojun 		}
   1459   1.61    itojun #endif /*IPSEC*/
   1460   1.18   mycroft 
   1461   1.18   mycroft 		case IP_MULTICAST_IF:
   1462   1.18   mycroft 		case IP_MULTICAST_TTL:
   1463   1.18   mycroft 		case IP_MULTICAST_LOOP:
   1464   1.18   mycroft 		case IP_ADD_MEMBERSHIP:
   1465   1.18   mycroft 		case IP_DROP_MEMBERSHIP:
   1466   1.18   mycroft 			error = ip_getmoptions(optname, inp->inp_moptions, mp);
   1467  1.103      matt 			if (*mp)
   1468  1.103      matt 				MCLAIM(*mp, so->so_mowner);
   1469   1.41     lukem 			break;
   1470   1.41     lukem 
   1471   1.41     lukem 		case IP_PORTRANGE:
   1472   1.41     lukem 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1473  1.103      matt 			MCLAIM(m, so->so_mowner);
   1474   1.41     lukem 			m->m_len = sizeof(int);
   1475   1.41     lukem 
   1476   1.41     lukem 			if (inp->inp_flags & INP_LOWPORT)
   1477   1.41     lukem 				optval = IP_PORTRANGE_LOW;
   1478   1.41     lukem 			else
   1479   1.41     lukem 				optval = IP_PORTRANGE_DEFAULT;
   1480   1.41     lukem 
   1481   1.41     lukem 			*mtod(m, int *) = optval;
   1482   1.18   mycroft 			break;
   1483    1.1       cgd 
   1484    1.1       cgd 		default:
   1485   1.18   mycroft 			error = ENOPROTOOPT;
   1486    1.1       cgd 			break;
   1487    1.1       cgd 		}
   1488    1.1       cgd 		break;
   1489    1.1       cgd 	}
   1490    1.1       cgd 	return (error);
   1491    1.1       cgd }
   1492    1.1       cgd 
   1493    1.1       cgd /*
   1494    1.1       cgd  * Set up IP options in pcb for insertion in output packets.
   1495    1.1       cgd  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1496    1.1       cgd  * with destination address if source routed.
   1497    1.1       cgd  */
   1498   1.12   mycroft int
   1499    1.1       cgd #ifdef notyet
   1500  1.140     perry ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
   1501    1.1       cgd #else
   1502  1.140     perry ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
   1503    1.1       cgd #endif
   1504    1.1       cgd {
   1505   1.71  augustss 	int cnt, optlen;
   1506   1.71  augustss 	u_char *cp;
   1507    1.1       cgd 	u_char opt;
   1508    1.1       cgd 
   1509    1.1       cgd 	/* turn off any old options */
   1510    1.1       cgd 	if (*pcbopt)
   1511    1.1       cgd 		(void)m_free(*pcbopt);
   1512    1.1       cgd 	*pcbopt = 0;
   1513    1.1       cgd 	if (m == (struct mbuf *)0 || m->m_len == 0) {
   1514    1.1       cgd 		/*
   1515    1.1       cgd 		 * Only turning off any previous options.
   1516    1.1       cgd 		 */
   1517    1.1       cgd 		if (m)
   1518    1.1       cgd 			(void)m_free(m);
   1519    1.1       cgd 		return (0);
   1520    1.1       cgd 	}
   1521    1.1       cgd 
   1522   1.85     ragge #ifndef	__vax__
   1523   1.21       cgd 	if (m->m_len % sizeof(int32_t))
   1524    1.1       cgd 		goto bad;
   1525    1.1       cgd #endif
   1526    1.1       cgd 	/*
   1527    1.1       cgd 	 * IP first-hop destination address will be stored before
   1528    1.1       cgd 	 * actual options; move other options back
   1529    1.1       cgd 	 * and clear it when none present.
   1530    1.1       cgd 	 */
   1531    1.1       cgd 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
   1532    1.1       cgd 		goto bad;
   1533    1.1       cgd 	cnt = m->m_len;
   1534    1.1       cgd 	m->m_len += sizeof(struct in_addr);
   1535    1.1       cgd 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
   1536   1.57     perry 	memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
   1537    1.1       cgd 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
   1538    1.1       cgd 
   1539    1.1       cgd 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
   1540    1.1       cgd 		opt = cp[IPOPT_OPTVAL];
   1541    1.1       cgd 		if (opt == IPOPT_EOL)
   1542    1.1       cgd 			break;
   1543    1.1       cgd 		if (opt == IPOPT_NOP)
   1544    1.1       cgd 			optlen = 1;
   1545    1.1       cgd 		else {
   1546   1.74    itojun 			if (cnt < IPOPT_OLEN + sizeof(*cp))
   1547   1.74    itojun 				goto bad;
   1548    1.1       cgd 			optlen = cp[IPOPT_OLEN];
   1549   1.74    itojun 			if (optlen < IPOPT_OLEN  + sizeof(*cp) || optlen > cnt)
   1550    1.1       cgd 				goto bad;
   1551    1.1       cgd 		}
   1552    1.1       cgd 		switch (opt) {
   1553    1.1       cgd 
   1554    1.1       cgd 		default:
   1555    1.1       cgd 			break;
   1556    1.1       cgd 
   1557    1.1       cgd 		case IPOPT_LSRR:
   1558    1.1       cgd 		case IPOPT_SSRR:
   1559    1.1       cgd 			/*
   1560    1.1       cgd 			 * user process specifies route as:
   1561    1.1       cgd 			 *	->A->B->C->D
   1562    1.1       cgd 			 * D must be our final destination (but we can't
   1563    1.1       cgd 			 * check that since we may not have connected yet).
   1564    1.1       cgd 			 * A is first hop destination, which doesn't appear in
   1565    1.1       cgd 			 * actual IP option, but is stored before the options.
   1566    1.1       cgd 			 */
   1567    1.1       cgd 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
   1568    1.1       cgd 				goto bad;
   1569    1.1       cgd 			m->m_len -= sizeof(struct in_addr);
   1570    1.1       cgd 			cnt -= sizeof(struct in_addr);
   1571    1.1       cgd 			optlen -= sizeof(struct in_addr);
   1572    1.1       cgd 			cp[IPOPT_OLEN] = optlen;
   1573    1.1       cgd 			/*
   1574    1.1       cgd 			 * Move first hop before start of options.
   1575    1.1       cgd 			 */
   1576    1.1       cgd 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
   1577    1.1       cgd 			    sizeof(struct in_addr));
   1578    1.1       cgd 			/*
   1579    1.1       cgd 			 * Then copy rest of options back
   1580    1.1       cgd 			 * to close up the deleted entry.
   1581    1.1       cgd 			 */
   1582  1.132  christos 			(void)memmove(&cp[IPOPT_OFFSET+1],
   1583  1.132  christos 			    &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
   1584  1.132  christos 			    (unsigned)cnt - (IPOPT_MINOFF - 1));
   1585    1.1       cgd 			break;
   1586    1.1       cgd 		}
   1587    1.1       cgd 	}
   1588    1.1       cgd 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
   1589    1.1       cgd 		goto bad;
   1590    1.1       cgd 	*pcbopt = m;
   1591    1.1       cgd 	return (0);
   1592    1.1       cgd 
   1593    1.1       cgd bad:
   1594    1.1       cgd 	(void)m_free(m);
   1595    1.1       cgd 	return (EINVAL);
   1596    1.1       cgd }
   1597    1.5   hpeyerl 
   1598    1.5   hpeyerl /*
   1599   1.81    itojun  * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
   1600   1.81    itojun  */
   1601   1.81    itojun static struct ifnet *
   1602  1.140     perry ip_multicast_if(struct in_addr *a, int *ifindexp)
   1603   1.81    itojun {
   1604   1.81    itojun 	int ifindex;
   1605  1.111    itojun 	struct ifnet *ifp = NULL;
   1606  1.110    itojun 	struct in_ifaddr *ia;
   1607   1.81    itojun 
   1608   1.81    itojun 	if (ifindexp)
   1609   1.81    itojun 		*ifindexp = 0;
   1610   1.81    itojun 	if (ntohl(a->s_addr) >> 24 == 0) {
   1611   1.81    itojun 		ifindex = ntohl(a->s_addr) & 0xffffff;
   1612  1.129    itojun 		if (ifindex < 0 || if_indexlim <= ifindex)
   1613   1.81    itojun 			return NULL;
   1614   1.81    itojun 		ifp = ifindex2ifnet[ifindex];
   1615  1.129    itojun 		if (!ifp)
   1616  1.129    itojun 			return NULL;
   1617   1.81    itojun 		if (ifindexp)
   1618   1.81    itojun 			*ifindexp = ifindex;
   1619   1.81    itojun 	} else {
   1620  1.110    itojun 		LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
   1621  1.110    itojun 			if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
   1622  1.111    itojun 			    (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
   1623  1.111    itojun 				ifp = ia->ia_ifp;
   1624  1.110    itojun 				break;
   1625  1.111    itojun 			}
   1626  1.110    itojun 		}
   1627   1.81    itojun 	}
   1628   1.81    itojun 	return ifp;
   1629   1.81    itojun }
   1630   1.81    itojun 
   1631  1.156  christos static int
   1632  1.156  christos ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval)
   1633  1.156  christos {
   1634  1.156  christos 	u_int tval;
   1635  1.156  christos 
   1636  1.156  christos 	if (m == NULL)
   1637  1.156  christos 		return EINVAL;
   1638  1.156  christos 
   1639  1.156  christos 	switch (m->m_len) {
   1640  1.156  christos 	case sizeof(u_char):
   1641  1.156  christos 		tval = *(mtod(m, u_char *));
   1642  1.156  christos 		break;
   1643  1.156  christos 	case sizeof(u_int):
   1644  1.156  christos 		tval = *(mtod(m, u_int *));
   1645  1.156  christos 		break;
   1646  1.156  christos 	default:
   1647  1.156  christos 		return EINVAL;
   1648  1.156  christos 	}
   1649  1.156  christos 
   1650  1.156  christos 	if (tval > maxval)
   1651  1.156  christos 		return EINVAL;
   1652  1.156  christos 
   1653  1.156  christos 	*val = tval;
   1654  1.156  christos 	return 0;
   1655  1.156  christos }
   1656  1.156  christos 
   1657   1.81    itojun /*
   1658    1.5   hpeyerl  * Set the IP multicast options in response to user setsockopt().
   1659    1.5   hpeyerl  */
   1660    1.5   hpeyerl int
   1661  1.140     perry ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m)
   1662    1.5   hpeyerl {
   1663   1.71  augustss 	int error = 0;
   1664   1.71  augustss 	int i;
   1665    1.5   hpeyerl 	struct in_addr addr;
   1666   1.71  augustss 	struct ip_mreq *mreq;
   1667   1.71  augustss 	struct ifnet *ifp;
   1668   1.71  augustss 	struct ip_moptions *imo = *imop;
   1669    1.5   hpeyerl 	struct route ro;
   1670   1.71  augustss 	struct sockaddr_in *dst;
   1671   1.81    itojun 	int ifindex;
   1672    1.5   hpeyerl 
   1673    1.5   hpeyerl 	if (imo == NULL) {
   1674    1.5   hpeyerl 		/*
   1675    1.5   hpeyerl 		 * No multicast option buffer attached to the pcb;
   1676    1.5   hpeyerl 		 * allocate one and initialize to default values.
   1677    1.5   hpeyerl 		 */
   1678   1.22       cgd 		imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
   1679    1.5   hpeyerl 		    M_WAITOK);
   1680    1.5   hpeyerl 
   1681    1.5   hpeyerl 		if (imo == NULL)
   1682    1.5   hpeyerl 			return (ENOBUFS);
   1683    1.5   hpeyerl 		*imop = imo;
   1684    1.5   hpeyerl 		imo->imo_multicast_ifp = NULL;
   1685   1.81    itojun 		imo->imo_multicast_addr.s_addr = INADDR_ANY;
   1686    1.5   hpeyerl 		imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
   1687    1.5   hpeyerl 		imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
   1688    1.5   hpeyerl 		imo->imo_num_memberships = 0;
   1689    1.5   hpeyerl 	}
   1690    1.5   hpeyerl 
   1691    1.5   hpeyerl 	switch (optname) {
   1692    1.5   hpeyerl 
   1693    1.5   hpeyerl 	case IP_MULTICAST_IF:
   1694    1.5   hpeyerl 		/*
   1695    1.5   hpeyerl 		 * Select the interface for outgoing multicast packets.
   1696    1.5   hpeyerl 		 */
   1697    1.5   hpeyerl 		if (m == NULL || m->m_len != sizeof(struct in_addr)) {
   1698    1.5   hpeyerl 			error = EINVAL;
   1699    1.5   hpeyerl 			break;
   1700    1.5   hpeyerl 		}
   1701    1.5   hpeyerl 		addr = *(mtod(m, struct in_addr *));
   1702    1.5   hpeyerl 		/*
   1703    1.5   hpeyerl 		 * INADDR_ANY is used to remove a previous selection.
   1704   1.11   mycroft 		 * When no interface is selected, a default one is
   1705    1.5   hpeyerl 		 * chosen every time a multicast packet is sent.
   1706    1.5   hpeyerl 		 */
   1707   1.31   mycroft 		if (in_nullhost(addr)) {
   1708    1.5   hpeyerl 			imo->imo_multicast_ifp = NULL;
   1709    1.5   hpeyerl 			break;
   1710    1.5   hpeyerl 		}
   1711    1.5   hpeyerl 		/*
   1712    1.5   hpeyerl 		 * The selected interface is identified by its local
   1713    1.5   hpeyerl 		 * IP address.  Find the interface and confirm that
   1714   1.11   mycroft 		 * it supports multicasting.
   1715    1.5   hpeyerl 		 */
   1716   1.81    itojun 		ifp = ip_multicast_if(&addr, &ifindex);
   1717    1.5   hpeyerl 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1718    1.5   hpeyerl 			error = EADDRNOTAVAIL;
   1719    1.5   hpeyerl 			break;
   1720    1.5   hpeyerl 		}
   1721    1.5   hpeyerl 		imo->imo_multicast_ifp = ifp;
   1722   1.81    itojun 		if (ifindex)
   1723   1.81    itojun 			imo->imo_multicast_addr = addr;
   1724   1.81    itojun 		else
   1725   1.81    itojun 			imo->imo_multicast_addr.s_addr = INADDR_ANY;
   1726    1.5   hpeyerl 		break;
   1727    1.5   hpeyerl 
   1728    1.5   hpeyerl 	case IP_MULTICAST_TTL:
   1729    1.5   hpeyerl 		/*
   1730    1.5   hpeyerl 		 * Set the IP time-to-live for outgoing multicast packets.
   1731    1.5   hpeyerl 		 */
   1732  1.157       seb 		error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL);
   1733    1.5   hpeyerl 		break;
   1734   1.11   mycroft 
   1735    1.5   hpeyerl 	case IP_MULTICAST_LOOP:
   1736    1.5   hpeyerl 		/*
   1737    1.5   hpeyerl 		 * Set the loopback flag for outgoing multicast packets.
   1738    1.5   hpeyerl 		 * Must be zero or one.
   1739    1.5   hpeyerl 		 */
   1740  1.156  christos 		error = ip_getoptval(m, &imo->imo_multicast_loop, 1);
   1741    1.5   hpeyerl 		break;
   1742    1.5   hpeyerl 
   1743    1.5   hpeyerl 	case IP_ADD_MEMBERSHIP:
   1744    1.5   hpeyerl 		/*
   1745    1.5   hpeyerl 		 * Add a multicast group membership.
   1746    1.5   hpeyerl 		 * Group must be a valid IP multicast address.
   1747    1.5   hpeyerl 		 */
   1748    1.5   hpeyerl 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
   1749    1.5   hpeyerl 			error = EINVAL;
   1750    1.5   hpeyerl 			break;
   1751    1.5   hpeyerl 		}
   1752    1.5   hpeyerl 		mreq = mtod(m, struct ip_mreq *);
   1753   1.23   mycroft 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
   1754    1.5   hpeyerl 			error = EINVAL;
   1755    1.5   hpeyerl 			break;
   1756    1.5   hpeyerl 		}
   1757    1.5   hpeyerl 		/*
   1758    1.5   hpeyerl 		 * If no interface address was provided, use the interface of
   1759    1.5   hpeyerl 		 * the route to the given multicast address.
   1760    1.5   hpeyerl 		 */
   1761   1.31   mycroft 		if (in_nullhost(mreq->imr_interface)) {
   1762   1.53        ws 			bzero((caddr_t)&ro, sizeof(ro));
   1763   1.24   mycroft 			dst = satosin(&ro.ro_dst);
   1764    1.5   hpeyerl 			dst->sin_len = sizeof(*dst);
   1765    1.5   hpeyerl 			dst->sin_family = AF_INET;
   1766    1.5   hpeyerl 			dst->sin_addr = mreq->imr_multiaddr;
   1767  1.171     joerg 			rtcache_init(&ro);
   1768    1.5   hpeyerl 			if (ro.ro_rt == NULL) {
   1769    1.5   hpeyerl 				error = EADDRNOTAVAIL;
   1770    1.5   hpeyerl 				break;
   1771    1.5   hpeyerl 			}
   1772    1.5   hpeyerl 			ifp = ro.ro_rt->rt_ifp;
   1773  1.171     joerg 			rtcache_free(&ro);
   1774   1.23   mycroft 		} else {
   1775   1.81    itojun 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
   1776    1.5   hpeyerl 		}
   1777    1.5   hpeyerl 		/*
   1778    1.5   hpeyerl 		 * See if we found an interface, and confirm that it
   1779    1.5   hpeyerl 		 * supports multicast.
   1780    1.5   hpeyerl 		 */
   1781   1.11   mycroft 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1782    1.5   hpeyerl 			error = EADDRNOTAVAIL;
   1783    1.5   hpeyerl 			break;
   1784    1.5   hpeyerl 		}
   1785    1.5   hpeyerl 		/*
   1786    1.5   hpeyerl 		 * See if the membership already exists or if all the
   1787    1.5   hpeyerl 		 * membership slots are full.
   1788   1.11   mycroft 		 */
   1789    1.5   hpeyerl 		for (i = 0; i < imo->imo_num_memberships; ++i) {
   1790    1.5   hpeyerl 			if (imo->imo_membership[i]->inm_ifp == ifp &&
   1791   1.31   mycroft 			    in_hosteq(imo->imo_membership[i]->inm_addr,
   1792   1.31   mycroft 				      mreq->imr_multiaddr))
   1793    1.5   hpeyerl 				break;
   1794   1.11   mycroft 		}
   1795    1.5   hpeyerl 		if (i < imo->imo_num_memberships) {
   1796    1.5   hpeyerl 			error = EADDRINUSE;
   1797    1.5   hpeyerl 			break;
   1798    1.5   hpeyerl 		}
   1799    1.5   hpeyerl 		if (i == IP_MAX_MEMBERSHIPS) {
   1800   1.11   mycroft 			error = ETOOMANYREFS;
   1801    1.5   hpeyerl 			break;
   1802    1.5   hpeyerl 		}
   1803    1.5   hpeyerl 		/*
   1804    1.5   hpeyerl 		 * Everything looks good; add a new record to the multicast
   1805    1.5   hpeyerl 		 * address list for the given interface.
   1806    1.5   hpeyerl 		 */
   1807    1.5   hpeyerl 		if ((imo->imo_membership[i] =
   1808    1.5   hpeyerl 		    in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
   1809    1.5   hpeyerl 			error = ENOBUFS;
   1810    1.5   hpeyerl 			break;
   1811    1.5   hpeyerl 		}
   1812    1.5   hpeyerl 		++imo->imo_num_memberships;
   1813    1.5   hpeyerl 		break;
   1814    1.5   hpeyerl 
   1815    1.5   hpeyerl 	case IP_DROP_MEMBERSHIP:
   1816    1.5   hpeyerl 		/*
   1817    1.5   hpeyerl 		 * Drop a multicast group membership.
   1818    1.5   hpeyerl 		 * Group must be a valid IP multicast address.
   1819    1.5   hpeyerl 		 */
   1820    1.5   hpeyerl 		if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
   1821    1.5   hpeyerl 			error = EINVAL;
   1822    1.5   hpeyerl 			break;
   1823    1.5   hpeyerl 		}
   1824    1.5   hpeyerl 		mreq = mtod(m, struct ip_mreq *);
   1825   1.23   mycroft 		if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
   1826    1.5   hpeyerl 			error = EINVAL;
   1827    1.5   hpeyerl 			break;
   1828    1.5   hpeyerl 		}
   1829    1.5   hpeyerl 		/*
   1830    1.5   hpeyerl 		 * If an interface address was specified, get a pointer
   1831    1.5   hpeyerl 		 * to its ifnet structure.
   1832    1.5   hpeyerl 		 */
   1833   1.31   mycroft 		if (in_nullhost(mreq->imr_interface))
   1834    1.5   hpeyerl 			ifp = NULL;
   1835    1.5   hpeyerl 		else {
   1836   1.81    itojun 			ifp = ip_multicast_if(&mreq->imr_interface, NULL);
   1837    1.5   hpeyerl 			if (ifp == NULL) {
   1838    1.5   hpeyerl 				error = EADDRNOTAVAIL;
   1839    1.5   hpeyerl 				break;
   1840    1.5   hpeyerl 			}
   1841    1.5   hpeyerl 		}
   1842    1.5   hpeyerl 		/*
   1843    1.5   hpeyerl 		 * Find the membership in the membership array.
   1844    1.5   hpeyerl 		 */
   1845    1.5   hpeyerl 		for (i = 0; i < imo->imo_num_memberships; ++i) {
   1846    1.5   hpeyerl 			if ((ifp == NULL ||
   1847    1.5   hpeyerl 			     imo->imo_membership[i]->inm_ifp == ifp) &&
   1848   1.31   mycroft 			     in_hosteq(imo->imo_membership[i]->inm_addr,
   1849   1.31   mycroft 				       mreq->imr_multiaddr))
   1850    1.5   hpeyerl 				break;
   1851    1.5   hpeyerl 		}
   1852    1.5   hpeyerl 		if (i == imo->imo_num_memberships) {
   1853    1.5   hpeyerl 			error = EADDRNOTAVAIL;
   1854    1.5   hpeyerl 			break;
   1855    1.5   hpeyerl 		}
   1856    1.5   hpeyerl 		/*
   1857    1.5   hpeyerl 		 * Give up the multicast address record to which the
   1858    1.5   hpeyerl 		 * membership points.
   1859    1.5   hpeyerl 		 */
   1860   1.11   mycroft 		in_delmulti(imo->imo_membership[i]);
   1861    1.5   hpeyerl 		/*
   1862    1.5   hpeyerl 		 * Remove the gap in the membership array.
   1863    1.5   hpeyerl 		 */
   1864    1.5   hpeyerl 		for (++i; i < imo->imo_num_memberships; ++i)
   1865    1.5   hpeyerl 			imo->imo_membership[i-1] = imo->imo_membership[i];
   1866    1.5   hpeyerl 		--imo->imo_num_memberships;
   1867    1.5   hpeyerl 		break;
   1868    1.5   hpeyerl 
   1869    1.5   hpeyerl 	default:
   1870    1.5   hpeyerl 		error = EOPNOTSUPP;
   1871    1.5   hpeyerl 		break;
   1872    1.5   hpeyerl 	}
   1873    1.5   hpeyerl 
   1874    1.5   hpeyerl 	/*
   1875    1.5   hpeyerl 	 * If all options have default values, no need to keep the mbuf.
   1876    1.5   hpeyerl 	 */
   1877    1.5   hpeyerl 	if (imo->imo_multicast_ifp == NULL &&
   1878    1.5   hpeyerl 	    imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
   1879    1.5   hpeyerl 	    imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
   1880    1.5   hpeyerl 	    imo->imo_num_memberships == 0) {
   1881    1.5   hpeyerl 		free(*imop, M_IPMOPTS);
   1882    1.5   hpeyerl 		*imop = NULL;
   1883    1.5   hpeyerl 	}
   1884    1.5   hpeyerl 
   1885    1.5   hpeyerl 	return (error);
   1886    1.5   hpeyerl }
   1887    1.5   hpeyerl 
   1888    1.5   hpeyerl /*
   1889    1.5   hpeyerl  * Return the IP multicast options in response to user getsockopt().
   1890    1.5   hpeyerl  */
   1891    1.5   hpeyerl int
   1892  1.140     perry ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
   1893    1.5   hpeyerl {
   1894    1.5   hpeyerl 	u_char *ttl;
   1895    1.5   hpeyerl 	u_char *loop;
   1896    1.5   hpeyerl 	struct in_addr *addr;
   1897    1.5   hpeyerl 	struct in_ifaddr *ia;
   1898    1.5   hpeyerl 
   1899    1.5   hpeyerl 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1900    1.5   hpeyerl 
   1901    1.5   hpeyerl 	switch (optname) {
   1902    1.5   hpeyerl 
   1903    1.5   hpeyerl 	case IP_MULTICAST_IF:
   1904    1.5   hpeyerl 		addr = mtod(*mp, struct in_addr *);
   1905    1.5   hpeyerl 		(*mp)->m_len = sizeof(struct in_addr);
   1906    1.5   hpeyerl 		if (imo == NULL || imo->imo_multicast_ifp == NULL)
   1907   1.31   mycroft 			*addr = zeroin_addr;
   1908   1.81    itojun 		else if (imo->imo_multicast_addr.s_addr) {
   1909   1.81    itojun 			/* return the value user has set */
   1910   1.81    itojun 			*addr = imo->imo_multicast_addr;
   1911   1.81    itojun 		} else {
   1912    1.5   hpeyerl 			IFP_TO_IA(imo->imo_multicast_ifp, ia);
   1913   1.31   mycroft 			*addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
   1914    1.5   hpeyerl 		}
   1915    1.5   hpeyerl 		return (0);
   1916    1.5   hpeyerl 
   1917    1.5   hpeyerl 	case IP_MULTICAST_TTL:
   1918    1.5   hpeyerl 		ttl = mtod(*mp, u_char *);
   1919    1.5   hpeyerl 		(*mp)->m_len = 1;
   1920   1.31   mycroft 		*ttl = imo ? imo->imo_multicast_ttl
   1921   1.31   mycroft 			   : IP_DEFAULT_MULTICAST_TTL;
   1922    1.5   hpeyerl 		return (0);
   1923    1.5   hpeyerl 
   1924    1.5   hpeyerl 	case IP_MULTICAST_LOOP:
   1925    1.5   hpeyerl 		loop = mtod(*mp, u_char *);
   1926    1.5   hpeyerl 		(*mp)->m_len = 1;
   1927   1.31   mycroft 		*loop = imo ? imo->imo_multicast_loop
   1928   1.31   mycroft 			    : IP_DEFAULT_MULTICAST_LOOP;
   1929    1.5   hpeyerl 		return (0);
   1930    1.5   hpeyerl 
   1931    1.5   hpeyerl 	default:
   1932    1.5   hpeyerl 		return (EOPNOTSUPP);
   1933    1.5   hpeyerl 	}
   1934    1.5   hpeyerl }
   1935    1.5   hpeyerl 
   1936    1.5   hpeyerl /*
   1937    1.5   hpeyerl  * Discard the IP multicast options.
   1938    1.5   hpeyerl  */
   1939    1.5   hpeyerl void
   1940  1.140     perry ip_freemoptions(struct ip_moptions *imo)
   1941    1.5   hpeyerl {
   1942   1.71  augustss 	int i;
   1943    1.5   hpeyerl 
   1944    1.5   hpeyerl 	if (imo != NULL) {
   1945    1.5   hpeyerl 		for (i = 0; i < imo->imo_num_memberships; ++i)
   1946    1.5   hpeyerl 			in_delmulti(imo->imo_membership[i]);
   1947    1.5   hpeyerl 		free(imo, M_IPMOPTS);
   1948    1.5   hpeyerl 	}
   1949    1.5   hpeyerl }
   1950    1.5   hpeyerl 
   1951    1.5   hpeyerl /*
   1952    1.5   hpeyerl  * Routine called from ip_output() to loop back a copy of an IP multicast
   1953    1.5   hpeyerl  * packet to the input queue of a specified interface.  Note that this
   1954    1.5   hpeyerl  * calls the output routine of the loopback "driver", but with an interface
   1955  1.137     peter  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
   1956    1.5   hpeyerl  */
   1957   1.12   mycroft static void
   1958  1.140     perry ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
   1959    1.5   hpeyerl {
   1960   1.71  augustss 	struct ip *ip;
   1961    1.5   hpeyerl 	struct mbuf *copym;
   1962    1.5   hpeyerl 
   1963    1.5   hpeyerl 	copym = m_copy(m, 0, M_COPYALL);
   1964   1.70    itojun 	if (copym != NULL
   1965   1.65    itojun 	 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
   1966   1.65    itojun 		copym = m_pullup(copym, sizeof(struct ip));
   1967    1.5   hpeyerl 	if (copym != NULL) {
   1968    1.5   hpeyerl 		/*
   1969    1.5   hpeyerl 		 * We don't bother to fragment if the IP length is greater
   1970    1.5   hpeyerl 		 * than the interface's MTU.  Can this possibly matter?
   1971    1.5   hpeyerl 		 */
   1972    1.5   hpeyerl 		ip = mtod(copym, struct ip *);
   1973   1.93    itojun 
   1974   1.93    itojun 		if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
   1975   1.94   thorpej 			in_delayed_cksum(copym);
   1976   1.93    itojun 			copym->m_pkthdr.csum_flags &=
   1977   1.93    itojun 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
   1978   1.93    itojun 		}
   1979   1.93    itojun 
   1980    1.5   hpeyerl 		ip->ip_sum = 0;
   1981    1.5   hpeyerl 		ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
   1982   1.24   mycroft 		(void) looutput(ifp, copym, sintosa(dst), NULL);
   1983    1.5   hpeyerl 	}
   1984    1.5   hpeyerl }
   1985