Home | History | Annotate | Line # | Download | only in netinet
in_gif.c revision 1.86
      1 /*	$NetBSD: in_gif.c,v 1.86 2016/12/14 11:19:15 knakahara Exp $	*/
      2 /*	$KAME: in_gif.c,v 1.66 2001/07/29 04:46:09 itojun Exp $	*/
      3 
      4 /*
      5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6  * All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. Neither the name of the project nor the names of its contributors
     17  *    may be used to endorse or promote products derived from this software
     18  *    without specific prior written permission.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 #include <sys/cdefs.h>
     34 __KERNEL_RCSID(0, "$NetBSD: in_gif.c,v 1.86 2016/12/14 11:19:15 knakahara Exp $");
     35 
     36 #ifdef _KERNEL_OPT
     37 #include "opt_inet.h"
     38 #endif
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/socket.h>
     43 #include <sys/sockio.h>
     44 #include <sys/mbuf.h>
     45 #include <sys/errno.h>
     46 #include <sys/ioctl.h>
     47 #include <sys/syslog.h>
     48 #include <sys/kernel.h>
     49 
     50 #include <net/if.h>
     51 #include <net/route.h>
     52 
     53 #include <netinet/in.h>
     54 #include <netinet/in_systm.h>
     55 #include <netinet/ip.h>
     56 #include <netinet/ip_var.h>
     57 #include <netinet/in_gif.h>
     58 #include <netinet/in_var.h>
     59 #include <netinet/ip_encap.h>
     60 #include <netinet/ip_ecn.h>
     61 
     62 #ifdef INET6
     63 #include <netinet/ip6.h>
     64 #endif
     65 
     66 #include <net/if_gif.h>
     67 
     68 #include <net/net_osdep.h>
     69 
     70 static int gif_validate4(const struct ip *, struct gif_softc *,
     71 	struct ifnet *);
     72 
     73 int ip_gif_ttl = GIF_TTL;
     74 
     75 static const struct encapsw in_gif_encapsw = {
     76 	.encapsw4 = {
     77 		.pr_input	= in_gif_input,
     78 		.pr_ctlinput	= NULL,
     79 	}
     80 };
     81 
     82 int
     83 in_gif_output(struct ifnet *ifp, int family, struct mbuf *m)
     84 {
     85 	struct rtentry *rt;
     86 	struct route *ro;
     87 	struct gif_softc *sc = ifp->if_softc;
     88 	struct sockaddr_in *sin_src = satosin(sc->gif_psrc);
     89 	struct sockaddr_in *sin_dst = satosin(sc->gif_pdst);
     90 	struct ip iphdr;	/* capsule IP header, host byte ordered */
     91 	int proto, error;
     92 	u_int8_t tos;
     93 	union {
     94 		struct sockaddr		dst;
     95 		struct sockaddr_in	dst4;
     96 	} u;
     97 
     98 	if (sin_src == NULL || sin_dst == NULL ||
     99 	    sin_src->sin_family != AF_INET ||
    100 	    sin_dst->sin_family != AF_INET) {
    101 		m_freem(m);
    102 		return EAFNOSUPPORT;
    103 	}
    104 
    105 	switch (family) {
    106 #ifdef INET
    107 	case AF_INET:
    108 	    {
    109 		const struct ip *ip;
    110 
    111 		proto = IPPROTO_IPV4;
    112 		if (m->m_len < sizeof(*ip)) {
    113 			m = m_pullup(m, sizeof(*ip));
    114 			if (m == NULL)
    115 				return ENOBUFS;
    116 		}
    117 		ip = mtod(m, const struct ip *);
    118 		tos = ip->ip_tos;
    119 		break;
    120 	    }
    121 #endif /* INET */
    122 #ifdef INET6
    123 	case AF_INET6:
    124 	    {
    125 		const struct ip6_hdr *ip6;
    126 		proto = IPPROTO_IPV6;
    127 		if (m->m_len < sizeof(*ip6)) {
    128 			m = m_pullup(m, sizeof(*ip6));
    129 			if (m == NULL)
    130 				return ENOBUFS;
    131 		}
    132 		ip6 = mtod(m, const struct ip6_hdr *);
    133 		tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
    134 		break;
    135 	    }
    136 #endif /* INET6 */
    137 	default:
    138 #ifdef DEBUG
    139 		printf("in_gif_output: warning: unknown family %d passed\n",
    140 			family);
    141 #endif
    142 		m_freem(m);
    143 		return EAFNOSUPPORT;
    144 	}
    145 
    146 	memset(&iphdr, 0, sizeof(iphdr));
    147 	iphdr.ip_src = sin_src->sin_addr;
    148 	/* bidirectional configured tunnel mode */
    149 	if (sin_dst->sin_addr.s_addr != INADDR_ANY)
    150 		iphdr.ip_dst = sin_dst->sin_addr;
    151 	else {
    152 		m_freem(m);
    153 		return ENETUNREACH;
    154 	}
    155 	iphdr.ip_p = proto;
    156 	/* version will be set in ip_output() */
    157 	iphdr.ip_ttl = ip_gif_ttl;
    158 	iphdr.ip_len = htons(m->m_pkthdr.len + sizeof(struct ip));
    159 	if (ifp->if_flags & IFF_LINK1)
    160 		ip_ecn_ingress(ECN_ALLOWED, &iphdr.ip_tos, &tos);
    161 	else
    162 		ip_ecn_ingress(ECN_NOCARE, &iphdr.ip_tos, &tos);
    163 
    164 	/* prepend new IP header */
    165 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
    166 	/* XXX Is m_pullup really necessary after M_PREPEND? */
    167 	if (m != NULL && M_UNWRITABLE(m, sizeof(struct ip)))
    168 		m = m_pullup(m, sizeof(struct ip));
    169 	if (m == NULL)
    170 		return ENOBUFS;
    171 	bcopy(&iphdr, mtod(m, struct ip *), sizeof(struct ip));
    172 
    173 	sockaddr_in_init(&u.dst4, &sin_dst->sin_addr, 0);
    174 
    175 	ro = percpu_getref(sc->gif_ro_percpu);
    176 	if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
    177 		percpu_putref(sc->gif_ro_percpu);
    178 		m_freem(m);
    179 		return ENETUNREACH;
    180 	}
    181 
    182 	/* If the route constitutes infinite encapsulation, punt. */
    183 	if (rt->rt_ifp == ifp) {
    184 		rtcache_unref(rt, ro);
    185 		rtcache_free(ro);
    186 		percpu_putref(sc->gif_ro_percpu);
    187 		m_freem(m);
    188 		return ENETUNREACH;	/*XXX*/
    189 	}
    190 	rtcache_unref(rt, ro);
    191 
    192 	error = ip_output(m, NULL, ro, 0, NULL, NULL);
    193 	percpu_putref(sc->gif_ro_percpu);
    194 	return (error);
    195 }
    196 
    197 void
    198 in_gif_input(struct mbuf *m, int off, int proto)
    199 {
    200 	struct ifnet *gifp = NULL;
    201 	const struct ip *ip;
    202 	int af;
    203 	u_int8_t otos;
    204 
    205 	ip = mtod(m, const struct ip *);
    206 
    207 	gifp = (struct ifnet *)encap_getarg(m);
    208 
    209 	if (gifp == NULL || (gifp->if_flags & (IFF_UP|IFF_RUNNING))
    210 		!= (IFF_UP|IFF_RUNNING)) {
    211 		m_freem(m);
    212 		ip_statinc(IP_STAT_NOGIF);
    213 		return;
    214 	}
    215 #ifndef GIF_ENCAPCHECK
    216 	struct gif_softc *sc = (struct gif_softc *)gifp->if_softc;
    217 	/* other CPU do delete_tunnel */
    218 	if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
    219 		m_freem(m);
    220 		ip_statinc(IP_STAT_NOGIF);
    221 		return;
    222 	}
    223 
    224 	struct ifnet *rcvif;
    225 	struct psref psref;
    226 	rcvif = m_get_rcvif_psref(m, &psref);
    227 	if (!gif_validate4(ip, sc, rcvif)) {
    228 		m_put_rcvif_psref(rcvif, &psref);
    229 		m_freem(m);
    230 		ip_statinc(IP_STAT_NOGIF);
    231 		return;
    232 	}
    233 	m_put_rcvif_psref(rcvif, &psref);
    234 #endif
    235 	otos = ip->ip_tos;
    236 	m_adj(m, off);
    237 
    238 	switch (proto) {
    239 #ifdef INET
    240 	case IPPROTO_IPV4:
    241 	    {
    242 		struct ip *xip;
    243 		af = AF_INET;
    244 		if (M_UNWRITABLE(m, sizeof(*xip))) {
    245 			if ((m = m_pullup(m, sizeof(*xip))) == NULL)
    246 				return;
    247 		}
    248 		xip = mtod(m, struct ip *);
    249 		if (gifp->if_flags & IFF_LINK1)
    250 			ip_ecn_egress(ECN_ALLOWED, &otos, &xip->ip_tos);
    251 		else
    252 			ip_ecn_egress(ECN_NOCARE, &otos, &xip->ip_tos);
    253 		break;
    254 	    }
    255 #endif
    256 #ifdef INET6
    257 	case IPPROTO_IPV6:
    258 	    {
    259 		struct ip6_hdr *ip6;
    260 		u_int8_t itos;
    261 		af = AF_INET6;
    262 		if (M_UNWRITABLE(m, sizeof(*ip6))) {
    263 			if ((m = m_pullup(m, sizeof(*ip6))) == NULL)
    264 				return;
    265 		}
    266 		ip6 = mtod(m, struct ip6_hdr *);
    267 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
    268 		if (gifp->if_flags & IFF_LINK1)
    269 			ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
    270 		else
    271 			ip_ecn_egress(ECN_NOCARE, &otos, &itos);
    272 		ip6->ip6_flow &= ~htonl(0xff << 20);
    273 		ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
    274 		break;
    275 	    }
    276 #endif /* INET6 */
    277 	default:
    278 		ip_statinc(IP_STAT_NOGIF);
    279 		m_freem(m);
    280 		return;
    281 	}
    282 	gif_input(m, af, gifp);
    283 	return;
    284 }
    285 
    286 /*
    287  * validate outer address.
    288  */
    289 static int
    290 gif_validate4(const struct ip *ip, struct gif_softc *sc, struct ifnet *ifp)
    291 {
    292 	struct sockaddr_in *src, *dst;
    293 	struct in_ifaddr *ia4;
    294 	int s;
    295 
    296 	src = satosin(sc->gif_psrc);
    297 	dst = satosin(sc->gif_pdst);
    298 
    299 	/* check for address match */
    300 	if (src->sin_addr.s_addr != ip->ip_dst.s_addr ||
    301 	    dst->sin_addr.s_addr != ip->ip_src.s_addr)
    302 		return 0;
    303 
    304 	/* martian filters on outer source - NOT done in ip_input! */
    305 	if (IN_MULTICAST(ip->ip_src.s_addr))
    306 		return 0;
    307 	switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) {
    308 	case 0: case 127: case 255:
    309 		return 0;
    310 	}
    311 	/* reject packets with broadcast on source */
    312 	s = pserialize_read_enter();
    313 	IN_ADDRLIST_READER_FOREACH(ia4) {
    314 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
    315 			continue;
    316 		if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
    317 			pserialize_read_exit(s);
    318 			return 0;
    319 		}
    320 	}
    321 	pserialize_read_exit(s);
    322 
    323 	/* ingress filters on outer source */
    324 	if ((sc->gif_if.if_flags & IFF_LINK2) == 0 && ifp) {
    325 		union {
    326 			struct sockaddr sa;
    327 			struct sockaddr_in sin;
    328 		} u;
    329 		struct rtentry *rt;
    330 
    331 		sockaddr_in_init(&u.sin, &ip->ip_src, 0);
    332 		rt = rtalloc1(&u.sa, 0);
    333 		if (rt == NULL || rt->rt_ifp != ifp) {
    334 #if 0
    335 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
    336 			    "due to ingress filter\n", if_name(&sc->gif_if),
    337 			    (u_int32_t)ntohl(u.sin.sin_addr.s_addr));
    338 #endif
    339 			if (rt != NULL)
    340 				rt_unref(rt);
    341 			return 0;
    342 		}
    343 		rt_unref(rt);
    344 	}
    345 
    346 	return 32 * 2;
    347 }
    348 
    349 #ifdef GIF_ENCAPCHECK
    350 /*
    351  * we know that we are in IFF_UP, outer address available, and outer family
    352  * matched the physical addr family.  see gif_encapcheck().
    353  */
    354 int
    355 gif_encapcheck4(struct mbuf *m, int off, int proto, void *arg)
    356 {
    357 	struct ip ip;
    358 	struct gif_softc *sc;
    359 	struct ifnet *ifp = NULL;
    360 	int r;
    361 	struct psref psref;
    362 
    363 	/* sanity check done in caller */
    364 	sc = arg;
    365 
    366 	m_copydata(m, 0, sizeof(ip), &ip);
    367 	if ((m->m_flags & M_PKTHDR) != 0)
    368 		ifp = m_get_rcvif_psref(m, &psref);
    369 
    370 	r = gif_validate4(&ip, sc, ifp);
    371 
    372 	m_put_rcvif_psref(ifp, &psref);
    373 	return r;
    374 }
    375 #endif
    376 
    377 int
    378 in_gif_attach(struct gif_softc *sc)
    379 {
    380 #ifndef GIF_ENCAPCHECK
    381 	struct sockaddr_in mask4;
    382 
    383 	memset(&mask4, 0, sizeof(mask4));
    384 	mask4.sin_len = sizeof(struct sockaddr_in);
    385 	mask4.sin_addr.s_addr = ~0;
    386 
    387 	if (!sc->gif_psrc || !sc->gif_pdst)
    388 		return EINVAL;
    389 	sc->encap_cookie4 = encap_attach(AF_INET, -1, sc->gif_psrc,
    390 	    (struct sockaddr *)&mask4, sc->gif_pdst, (struct sockaddr *)&mask4,
    391 	    &in_gif_encapsw, sc);
    392 #else
    393 	sc->encap_cookie4 = encap_attach_func(AF_INET, -1, gif_encapcheck,
    394 	    &in_gif_encapsw, sc);
    395 #endif
    396 	if (sc->encap_cookie4 == NULL)
    397 		return EEXIST;
    398 	return 0;
    399 }
    400 
    401 int
    402 in_gif_detach(struct gif_softc *sc)
    403 {
    404 	int error;
    405 
    406 	error = in_gif_pause(sc);
    407 
    408 	percpu_foreach(sc->gif_ro_percpu, gif_rtcache_free_pc, NULL);
    409 
    410 	return error;
    411 }
    412 
    413 int
    414 in_gif_pause(struct gif_softc *sc)
    415 {
    416 	int error;
    417 
    418 	error = encap_detach(sc->encap_cookie4);
    419 	if (error == 0)
    420 		sc->encap_cookie4 = NULL;
    421 
    422 	return error;
    423 }
    424