if_stf.c revision 151266
1/*	$FreeBSD: head/sys/net/if_stf.c 151266 2005-10-12 19:52:16Z thompsa $	*/
2/*	$KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $	*/
3
4/*-
5 * Copyright (C) 2000 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/*
34 * 6to4 interface, based on RFC3056.
35 *
36 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
37 * There is no address mapping defined from IPv6 multicast address to IPv4
38 * address.  Therefore, we do not have IFF_MULTICAST on the interface.
39 *
40 * Due to the lack of address mapping for link-local addresses, we cannot
41 * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
42 * packets to link-local multicast addresses (ff02::x).
43 *
44 * Here are interesting symptoms due to the lack of link-local address:
45 *
46 * Unicast routing exchange:
47 * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
48 *   and link-local addresses as nexthop.
49 * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
50 *   assigned to the link, and makes use of them.  Also, HELLO packets use
51 *   link-local multicast addresses (ff02::5 and ff02::6).
52 * - BGP4+: Maybe.  You can only use global address as nexthop, and global
53 *   address as TCP endpoint address.
54 *
55 * Multicast routing protocols:
56 * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
57 *   Adjacent PIM routers must be configured manually (is it really spec-wise
58 *   correct thing to do?).
59 *
60 * ICMPv6:
61 * - Redirects cannot be used due to the lack of link-local address.
62 *
63 * stf interface does not have, and will not need, a link-local address.
64 * It seems to have no real benefit and does not help the above symptoms much.
65 * Even if we assign link-locals to interface, we cannot really
66 * use link-local unicast/multicast on top of 6to4 cloud (since there's no
67 * encapsulation defined for link-local address), and the above analysis does
68 * not change.  RFC3056 does not mandate the assignment of link-local address
69 * either.
70 *
71 * 6to4 interface has security issues.  Refer to
72 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
73 * for details.  The code tries to filter out some of malicious packets.
74 * Note that there is no way to be 100% secure.
75 */
76
77#include "opt_inet.h"
78#include "opt_inet6.h"
79#include "opt_mac.h"
80
81#include <sys/param.h>
82#include <sys/systm.h>
83#include <sys/socket.h>
84#include <sys/sockio.h>
85#include <sys/mac.h>
86#include <sys/mbuf.h>
87#include <sys/errno.h>
88#include <sys/kernel.h>
89#include <sys/module.h>
90#include <sys/protosw.h>
91#include <sys/queue.h>
92#include <machine/cpu.h>
93
94#include <sys/malloc.h>
95
96#include <net/if.h>
97#include <net/if_clone.h>
98#include <net/route.h>
99#include <net/netisr.h>
100#include <net/if_types.h>
101#include <net/if_stf.h>
102
103#include <netinet/in.h>
104#include <netinet/in_systm.h>
105#include <netinet/ip.h>
106#include <netinet/ip_var.h>
107#include <netinet/in_var.h>
108
109#include <netinet/ip6.h>
110#include <netinet6/ip6_var.h>
111#include <netinet6/in6_var.h>
112#include <netinet/ip_ecn.h>
113
114#include <netinet/ip_encap.h>
115
116#include <machine/stdarg.h>
117
118#include <net/net_osdep.h>
119
120#include <net/bpf.h>
121
122#define STFNAME		"stf"
123#define STFUNIT		0
124
125#define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
126
127/*
128 * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
129 * struct in_addr *; use bcopy() instead.
130 */
131#define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
132
133struct stf_softc {
134	struct ifnet	*sc_ifp;
135	union {
136		struct route  __sc_ro4;
137		struct route_in6 __sc_ro6; /* just for safety */
138	} __sc_ro46;
139#define sc_ro	__sc_ro46.__sc_ro4
140	const struct encaptab *encap_cookie;
141	LIST_ENTRY(stf_softc) sc_list;	/* all stf's are linked */
142};
143#define STF2IFP(sc)	((sc)->sc_ifp)
144
145/*
146 * All mutable global variables in if_stf.c are protected by stf_mtx.
147 * XXXRW: Note that mutable fields in the softc are not currently locked:
148 * in particular, sc_ro needs to be protected from concurrent entrance
149 * of stf_output().
150 */
151static struct mtx stf_mtx;
152static LIST_HEAD(, stf_softc) stf_softc_list;
153
154static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
155static const int ip_stf_ttl = 40;
156
157extern  struct domain inetdomain;
158struct protosw in_stf_protosw =
159{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
160  in_stf_input,	(pr_output_t*)rip_output, 0,	rip_ctloutput,
161  0,
162  0,            0,              0,              0,
163  &rip_usrreqs
164};
165
166static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
167
168static int stfmodevent(module_t, int, void *);
169static int stf_encapcheck(const struct mbuf *, int, int, void *);
170static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
171static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
172	struct rtentry *);
173static int isrfc1918addr(struct in_addr *);
174static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
175	struct ifnet *);
176static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
177	struct ifnet *);
178static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
179static int stf_ioctl(struct ifnet *, u_long, caddr_t);
180
181static int stf_clone_match(struct if_clone *, const char *);
182static int stf_clone_create(struct if_clone *, char *, size_t);
183static int stf_clone_destroy(struct if_clone *, struct ifnet *);
184struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
185    NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
186
187static int
188stf_clone_match(struct if_clone *ifc, const char *name)
189{
190	int i;
191
192	for(i = 0; stfnames[i] != NULL; i++) {
193		if (strcmp(stfnames[i], name) == 0)
194			return (1);
195	}
196
197	return (0);
198}
199
200static int
201stf_clone_create(struct if_clone *ifc, char *name, size_t len)
202{
203	int err, unit;
204	struct stf_softc *sc;
205	struct ifnet *ifp;
206
207	/*
208	 * We can only have one unit, but since unit allocation is
209	 * already locked, we use it to keep from allocating extra
210	 * interfaces.
211	 */
212	unit = STFUNIT;
213	err = ifc_alloc_unit(ifc, &unit);
214	if (err != 0)
215		return (err);
216
217	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
218	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
219	if (ifp == NULL) {
220		free(sc, M_STF);
221		ifc_free_unit(ifc, unit);
222		return (ENOSPC);
223	}
224	ifp->if_softc = sc;
225
226	/*
227	 * Set the name manually rather then using if_initname because
228	 * we don't conform to the default naming convention for interfaces.
229	 */
230	strlcpy(ifp->if_xname, name, IFNAMSIZ);
231	ifp->if_dname = ifc->ifc_name;
232	ifp->if_dunit = IF_DUNIT_NONE;
233
234	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
235	    stf_encapcheck, &in_stf_protosw, sc);
236	if (sc->encap_cookie == NULL) {
237		if_printf(ifp, "attach failed\n");
238		free(sc, M_STF);
239		ifc_free_unit(ifc, unit);
240		return (ENOMEM);
241	}
242
243	ifp->if_mtu    = IPV6_MMTU;
244	ifp->if_ioctl  = stf_ioctl;
245	ifp->if_output = stf_output;
246	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
247	if_attach(ifp);
248	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
249	mtx_lock(&stf_mtx);
250	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
251	mtx_unlock(&stf_mtx);
252	return (0);
253}
254
255static int
256stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
257{
258	struct stf_softc *sc = ifp->if_softc;
259	int err;
260
261	mtx_lock(&stf_mtx);
262	LIST_REMOVE(sc, sc_list);
263	mtx_unlock(&stf_mtx);
264
265	err = encap_detach(sc->encap_cookie);
266	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
267	bpfdetach(ifp);
268	if_detach(ifp);
269	if_free(ifp);
270
271	free(sc, M_STF);
272	ifc_free_unit(ifc, STFUNIT);
273
274	return (0);
275}
276
277static int
278stfmodevent(mod, type, data)
279	module_t mod;
280	int type;
281	void *data;
282{
283	struct stf_softc *sc;
284
285	switch (type) {
286	case MOD_LOAD:
287		mtx_init(&stf_mtx, "stf_mtx", NULL, MTX_DEF);
288		LIST_INIT(&stf_softc_list);
289		if_clone_attach(&stf_cloner);
290
291		break;
292	case MOD_UNLOAD:
293		if_clone_detach(&stf_cloner);
294
295		mtx_lock(&stf_mtx);
296		while ((sc = LIST_FIRST(&stf_softc_list)) != NULL) {
297			mtx_unlock(&stf_mtx);
298			stf_clone_destroy(&stf_cloner, STF2IFP(sc));
299			mtx_lock(&stf_mtx);
300		}
301		mtx_unlock(&stf_mtx);
302		mtx_destroy(&stf_mtx);
303		break;
304	default:
305		return (EOPNOTSUPP);
306	}
307
308	return (0);
309}
310
311static moduledata_t stf_mod = {
312	"if_stf",
313	stfmodevent,
314	0
315};
316
317DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
318
319static int
320stf_encapcheck(m, off, proto, arg)
321	const struct mbuf *m;
322	int off;
323	int proto;
324	void *arg;
325{
326	struct ip ip;
327	struct in6_ifaddr *ia6;
328	struct stf_softc *sc;
329	struct in_addr a, b, mask;
330
331	sc = (struct stf_softc *)arg;
332	if (sc == NULL)
333		return 0;
334
335	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
336		return 0;
337
338	/* IFF_LINK0 means "no decapsulation" */
339	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
340		return 0;
341
342	if (proto != IPPROTO_IPV6)
343		return 0;
344
345	/* LINTED const cast */
346	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
347
348	if (ip.ip_v != 4)
349		return 0;
350
351	ia6 = stf_getsrcifa6(STF2IFP(sc));
352	if (ia6 == NULL)
353		return 0;
354
355	/*
356	 * check if IPv4 dst matches the IPv4 address derived from the
357	 * local 6to4 address.
358	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
359	 */
360	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
361	    sizeof(ip.ip_dst)) != 0)
362		return 0;
363
364	/*
365	 * check if IPv4 src matches the IPv4 address derived from the
366	 * local 6to4 address masked by prefixmask.
367	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
368	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
369	 */
370	bzero(&a, sizeof(a));
371	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
372	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
373	a.s_addr &= mask.s_addr;
374	b = ip.ip_src;
375	b.s_addr &= mask.s_addr;
376	if (a.s_addr != b.s_addr)
377		return 0;
378
379	/* stf interface makes single side match only */
380	return 32;
381}
382
383static struct in6_ifaddr *
384stf_getsrcifa6(ifp)
385	struct ifnet *ifp;
386{
387	struct ifaddr *ia;
388	struct in_ifaddr *ia4;
389	struct sockaddr_in6 *sin6;
390	struct in_addr in;
391
392	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
393	     ia;
394	     ia = TAILQ_NEXT(ia, ifa_list))
395	{
396		if (ia->ifa_addr == NULL)
397			continue;
398		if (ia->ifa_addr->sa_family != AF_INET6)
399			continue;
400		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
401		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
402			continue;
403
404		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
405		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
406			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
407				break;
408		if (ia4 == NULL)
409			continue;
410
411		return (struct in6_ifaddr *)ia;
412	}
413
414	return NULL;
415}
416
417static int
418stf_output(ifp, m, dst, rt)
419	struct ifnet *ifp;
420	struct mbuf *m;
421	struct sockaddr *dst;
422	struct rtentry *rt;
423{
424	struct stf_softc *sc;
425	struct sockaddr_in6 *dst6;
426	struct in_addr in4;
427	caddr_t ptr;
428	struct sockaddr_in *dst4;
429	u_int8_t tos;
430	struct ip *ip;
431	struct ip6_hdr *ip6;
432	struct in6_ifaddr *ia6;
433	u_int32_t af;
434#ifdef MAC
435	int error;
436
437	error = mac_check_ifnet_transmit(ifp, m);
438	if (error) {
439		m_freem(m);
440		return (error);
441	}
442#endif
443
444	sc = ifp->if_softc;
445	dst6 = (struct sockaddr_in6 *)dst;
446
447	/* just in case */
448	if ((ifp->if_flags & IFF_UP) == 0) {
449		m_freem(m);
450		ifp->if_oerrors++;
451		return ENETDOWN;
452	}
453
454	/*
455	 * If we don't have an ip4 address that match my inner ip6 address,
456	 * we shouldn't generate output.  Without this check, we'll end up
457	 * using wrong IPv4 source.
458	 */
459	ia6 = stf_getsrcifa6(ifp);
460	if (ia6 == NULL) {
461		m_freem(m);
462		ifp->if_oerrors++;
463		return ENETDOWN;
464	}
465
466	if (m->m_len < sizeof(*ip6)) {
467		m = m_pullup(m, sizeof(*ip6));
468		if (!m) {
469			ifp->if_oerrors++;
470			return ENOBUFS;
471		}
472	}
473	ip6 = mtod(m, struct ip6_hdr *);
474	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
475
476	/*
477	 * BPF writes need to be handled specially.
478	 * This is a null operation, nothing here checks dst->sa_family.
479	 */
480	if (dst->sa_family == AF_UNSPEC) {
481		bcopy(dst->sa_data, &af, sizeof(af));
482		dst->sa_family = af;
483	}
484
485	/*
486	 * Pickup the right outer dst addr from the list of candidates.
487	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
488	 */
489	ptr = NULL;
490	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
491		ptr = GET_V4(&ip6->ip6_dst);
492	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
493		ptr = GET_V4(&dst6->sin6_addr);
494	else {
495		m_freem(m);
496		ifp->if_oerrors++;
497		return ENETUNREACH;
498	}
499	bcopy(ptr, &in4, sizeof(in4));
500
501	if (ifp->if_bpf) {
502		/*
503		 * We need to prepend the address family as
504		 * a four byte field.  Cons up a dummy header
505		 * to pacify bpf.  This is safe because bpf
506		 * will only read from the mbuf (i.e., it won't
507		 * try to free it or keep a pointer a to it).
508		 */
509		af = AF_INET6;
510		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
511	}
512
513	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
514	if (m && m->m_len < sizeof(struct ip))
515		m = m_pullup(m, sizeof(struct ip));
516	if (m == NULL) {
517		ifp->if_oerrors++;
518		return ENOBUFS;
519	}
520	ip = mtod(m, struct ip *);
521
522	bzero(ip, sizeof(*ip));
523
524	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
525	    &ip->ip_src, sizeof(ip->ip_src));
526	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
527	ip->ip_p = IPPROTO_IPV6;
528	ip->ip_ttl = ip_stf_ttl;
529	ip->ip_len = m->m_pkthdr.len;	/*host order*/
530	if (ifp->if_flags & IFF_LINK1)
531		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
532	else
533		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
534
535	/*
536	 * XXXRW: Locking of sc_ro required.
537	 */
538	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
539	if (dst4->sin_family != AF_INET ||
540	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
541		/* cache route doesn't match */
542		dst4->sin_family = AF_INET;
543		dst4->sin_len = sizeof(struct sockaddr_in);
544		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
545		if (sc->sc_ro.ro_rt) {
546			RTFREE(sc->sc_ro.ro_rt);
547			sc->sc_ro.ro_rt = NULL;
548		}
549	}
550
551	if (sc->sc_ro.ro_rt == NULL) {
552		rtalloc(&sc->sc_ro);
553		if (sc->sc_ro.ro_rt == NULL) {
554			m_freem(m);
555			ifp->if_oerrors++;
556			return ENETUNREACH;
557		}
558	}
559
560	ifp->if_opackets++;
561	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
562}
563
564static int
565isrfc1918addr(in)
566	struct in_addr *in;
567{
568	/*
569	 * returns 1 if private address range:
570	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
571	 */
572	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
573	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
574	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
575		return 1;
576
577	return 0;
578}
579
580static int
581stf_checkaddr4(sc, in, inifp)
582	struct stf_softc *sc;
583	struct in_addr *in;
584	struct ifnet *inifp;	/* incoming interface */
585{
586	struct in_ifaddr *ia4;
587
588	/*
589	 * reject packets with the following address:
590	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
591	 */
592	if (IN_MULTICAST(ntohl(in->s_addr)))
593		return -1;
594	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
595	case 0: case 127: case 255:
596		return -1;
597	}
598
599	/*
600	 * reject packets with private address range.
601	 * (requirement from RFC3056 section 2 1st paragraph)
602	 */
603	if (isrfc1918addr(in))
604		return -1;
605
606	/*
607	 * reject packets with broadcast
608	 */
609	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
610	     ia4;
611	     ia4 = TAILQ_NEXT(ia4, ia_link))
612	{
613		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
614			continue;
615		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
616			return -1;
617	}
618
619	/*
620	 * perform ingress filter
621	 */
622	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
623		struct sockaddr_in sin;
624		struct rtentry *rt;
625
626		bzero(&sin, sizeof(sin));
627		sin.sin_family = AF_INET;
628		sin.sin_len = sizeof(struct sockaddr_in);
629		sin.sin_addr = *in;
630		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
631		if (!rt || rt->rt_ifp != inifp) {
632#if 0
633			log(LOG_WARNING, "%s: packet from 0x%x dropped "
634			    "due to ingress filter\n", if_name(STF2IFP(sc)),
635			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
636#endif
637			if (rt)
638				rtfree(rt);
639			return -1;
640		}
641		rtfree(rt);
642	}
643
644	return 0;
645}
646
647static int
648stf_checkaddr6(sc, in6, inifp)
649	struct stf_softc *sc;
650	struct in6_addr *in6;
651	struct ifnet *inifp;	/* incoming interface */
652{
653	/*
654	 * check 6to4 addresses
655	 */
656	if (IN6_IS_ADDR_6TO4(in6)) {
657		struct in_addr in4;
658		bcopy(GET_V4(in6), &in4, sizeof(in4));
659		return stf_checkaddr4(sc, &in4, inifp);
660	}
661
662	/*
663	 * reject anything that look suspicious.  the test is implemented
664	 * in ip6_input too, but we check here as well to
665	 * (1) reject bad packets earlier, and
666	 * (2) to be safe against future ip6_input change.
667	 */
668	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
669		return -1;
670
671	return 0;
672}
673
674void
675in_stf_input(m, off)
676	struct mbuf *m;
677	int off;
678{
679	int proto;
680	struct stf_softc *sc;
681	struct ip *ip;
682	struct ip6_hdr *ip6;
683	u_int8_t otos, itos;
684	struct ifnet *ifp;
685
686	proto = mtod(m, struct ip *)->ip_p;
687
688	if (proto != IPPROTO_IPV6) {
689		m_freem(m);
690		return;
691	}
692
693	ip = mtod(m, struct ip *);
694
695	sc = (struct stf_softc *)encap_getarg(m);
696
697	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
698		m_freem(m);
699		return;
700	}
701
702	ifp = STF2IFP(sc);
703
704#ifdef MAC
705	mac_create_mbuf_from_ifnet(ifp, m);
706#endif
707
708	/*
709	 * perform sanity check against outer src/dst.
710	 * for source, perform ingress filter as well.
711	 */
712	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
713	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
714		m_freem(m);
715		return;
716	}
717
718	otos = ip->ip_tos;
719	m_adj(m, off);
720
721	if (m->m_len < sizeof(*ip6)) {
722		m = m_pullup(m, sizeof(*ip6));
723		if (!m)
724			return;
725	}
726	ip6 = mtod(m, struct ip6_hdr *);
727
728	/*
729	 * perform sanity check against inner src/dst.
730	 * for source, perform ingress filter as well.
731	 */
732	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
733	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
734		m_freem(m);
735		return;
736	}
737
738	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
739	if ((ifp->if_flags & IFF_LINK1) != 0)
740		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
741	else
742		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
743	ip6->ip6_flow &= ~htonl(0xff << 20);
744	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
745
746	m->m_pkthdr.rcvif = ifp;
747
748	if (ifp->if_bpf) {
749		/*
750		 * We need to prepend the address family as
751		 * a four byte field.  Cons up a dummy header
752		 * to pacify bpf.  This is safe because bpf
753		 * will only read from the mbuf (i.e., it won't
754		 * try to free it or keep a pointer a to it).
755		 */
756		u_int32_t af = AF_INET6;
757		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
758	}
759
760	/*
761	 * Put the packet to the network layer input queue according to the
762	 * specified address family.
763	 * See net/if_gif.c for possible issues with packet processing
764	 * reorder due to extra queueing.
765	 */
766	ifp->if_ipackets++;
767	ifp->if_ibytes += m->m_pkthdr.len;
768	netisr_dispatch(NETISR_IPV6, m);
769}
770
771/* ARGSUSED */
772static void
773stf_rtrequest(cmd, rt, info)
774	int cmd;
775	struct rtentry *rt;
776	struct rt_addrinfo *info;
777{
778	RT_LOCK_ASSERT(rt);
779	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
780}
781
782static int
783stf_ioctl(ifp, cmd, data)
784	struct ifnet *ifp;
785	u_long cmd;
786	caddr_t data;
787{
788	struct ifaddr *ifa;
789	struct ifreq *ifr;
790	struct sockaddr_in6 *sin6;
791	struct in_addr addr;
792	int error;
793
794	error = 0;
795	switch (cmd) {
796	case SIOCSIFADDR:
797		ifa = (struct ifaddr *)data;
798		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
799			error = EAFNOSUPPORT;
800			break;
801		}
802		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
803		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
804			error = EINVAL;
805			break;
806		}
807		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
808		if (isrfc1918addr(&addr)) {
809			error = EINVAL;
810			break;
811		}
812
813		ifa->ifa_rtrequest = stf_rtrequest;
814		ifp->if_flags |= IFF_UP;
815		break;
816
817	case SIOCADDMULTI:
818	case SIOCDELMULTI:
819		ifr = (struct ifreq *)data;
820		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
821			;
822		else
823			error = EAFNOSUPPORT;
824		break;
825
826	default:
827		error = EINVAL;
828		break;
829	}
830
831	return error;
832}
833