1/*	$FreeBSD: stable/11/sys/net/if_stf.c 340951 2018-11-26 11:17:12Z eugen $	*/
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 <sys/param.h>
78#include <sys/systm.h>
79#include <sys/socket.h>
80#include <sys/sockio.h>
81#include <sys/mbuf.h>
82#include <sys/errno.h>
83#include <sys/kernel.h>
84#include <sys/lock.h>
85#include <sys/module.h>
86#include <sys/protosw.h>
87#include <sys/proc.h>
88#include <sys/queue.h>
89#include <sys/rmlock.h>
90#include <sys/sysctl.h>
91#include <machine/cpu.h>
92
93#include <sys/malloc.h>
94
95#include <net/if.h>
96#include <net/if_var.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/vnet.h>
102
103#include <netinet/in.h>
104#include <netinet/in_fib.h>
105#include <netinet/in_systm.h>
106#include <netinet/ip.h>
107#include <netinet/ip_var.h>
108#include <netinet/in_var.h>
109
110#include <netinet/ip6.h>
111#include <netinet6/ip6_var.h>
112#include <netinet6/in6_var.h>
113#include <netinet/ip_ecn.h>
114
115#include <netinet/ip_encap.h>
116
117#include <machine/stdarg.h>
118
119#include <net/bpf.h>
120
121#include <security/mac/mac_framework.h>
122
123SYSCTL_DECL(_net_link);
124static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
125
126static int stf_permit_rfc1918 = 0;
127SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
128    &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
129
130#define STFUNIT		0
131
132#define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
133
134/*
135 * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
136 * struct in_addr *; use bcopy() instead.
137 */
138#define GET_V4(x)	(&(x)->s6_addr16[1])
139
140struct stf_softc {
141	struct ifnet	*sc_ifp;
142	u_int	sc_fibnum;
143	const struct encaptab *encap_cookie;
144};
145#define STF2IFP(sc)	((sc)->sc_ifp)
146
147static const char stfname[] = "stf";
148
149static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
150static const int ip_stf_ttl = 40;
151
152extern  struct domain inetdomain;
153static int in_stf_input(struct mbuf **, int *, int);
154static struct protosw in_stf_protosw = {
155	.pr_type =		SOCK_RAW,
156	.pr_domain =		&inetdomain,
157	.pr_protocol =		IPPROTO_IPV6,
158	.pr_flags =		PR_ATOMIC|PR_ADDR,
159	.pr_input =		in_stf_input,
160	.pr_output =		rip_output,
161	.pr_ctloutput =		rip_ctloutput,
162	.pr_usrreqs =		&rip_usrreqs
163};
164
165static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
166
167static int stfmodevent(module_t, int, void *);
168static int stf_encapcheck(const struct mbuf *, int, int, void *);
169static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
170static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
171	struct route *);
172static int isrfc1918addr(struct in_addr *);
173static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
174	struct ifnet *);
175static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
176	struct ifnet *);
177static int stf_ioctl(struct ifnet *, u_long, caddr_t);
178
179static int stf_clone_match(struct if_clone *, const char *);
180static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
181static int stf_clone_destroy(struct if_clone *, struct ifnet *);
182static struct if_clone *stf_cloner;
183
184static int
185stf_clone_match(struct if_clone *ifc, const char *name)
186{
187	int i;
188
189	for(i = 0; stfnames[i] != NULL; i++) {
190		if (strcmp(stfnames[i], name) == 0)
191			return (1);
192	}
193
194	return (0);
195}
196
197static int
198stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
199{
200	char *dp;
201	int err, unit, wildcard;
202	struct stf_softc *sc;
203	struct ifnet *ifp;
204
205	err = ifc_name2unit(name, &unit);
206	if (err != 0)
207		return (err);
208	wildcard = (unit < 0);
209
210	/*
211	 * We can only have one unit, but since unit allocation is
212	 * already locked, we use it to keep from allocating extra
213	 * interfaces.
214	 */
215	unit = STFUNIT;
216	err = ifc_alloc_unit(ifc, &unit);
217	if (err != 0)
218		return (err);
219
220	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
221	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
222	if (ifp == NULL) {
223		free(sc, M_STF);
224		ifc_free_unit(ifc, unit);
225		return (ENOSPC);
226	}
227	ifp->if_softc = sc;
228	sc->sc_fibnum = curthread->td_proc->p_fibnum;
229
230	/*
231	 * Set the name manually rather then using if_initname because
232	 * we don't conform to the default naming convention for interfaces.
233	 * In the wildcard case, we need to update the name.
234	 */
235	if (wildcard) {
236		for (dp = name; *dp != '\0'; dp++);
237		if (snprintf(dp, len - (dp-name), "%d", unit) >
238		    len - (dp-name) - 1) {
239			/*
240			 * This can only be a programmer error and
241			 * there's no straightforward way to recover if
242			 * it happens.
243			 */
244			panic("if_clone_create(): interface name too long");
245		}
246	}
247	strlcpy(ifp->if_xname, name, IFNAMSIZ);
248	ifp->if_dname = stfname;
249	ifp->if_dunit = IF_DUNIT_NONE;
250
251	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
252	    stf_encapcheck, &in_stf_protosw, sc);
253	if (sc->encap_cookie == NULL) {
254		if_printf(ifp, "attach failed\n");
255		free(sc, M_STF);
256		ifc_free_unit(ifc, unit);
257		return (ENOMEM);
258	}
259
260	ifp->if_mtu    = IPV6_MMTU;
261	ifp->if_ioctl  = stf_ioctl;
262	ifp->if_output = stf_output;
263	ifp->if_snd.ifq_maxlen = ifqmaxlen;
264	if_attach(ifp);
265	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
266	return (0);
267}
268
269static int
270stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
271{
272	struct stf_softc *sc = ifp->if_softc;
273	int err;
274
275	err = encap_detach(sc->encap_cookie);
276	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
277	bpfdetach(ifp);
278	if_detach(ifp);
279	if_free(ifp);
280
281	free(sc, M_STF);
282	ifc_free_unit(ifc, STFUNIT);
283
284	return (0);
285}
286
287static int
288stfmodevent(module_t mod, int type, void *data)
289{
290
291	switch (type) {
292	case MOD_LOAD:
293		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
294		    stf_clone_create, stf_clone_destroy);
295		break;
296	case MOD_UNLOAD:
297		if_clone_detach(stf_cloner);
298		break;
299	default:
300		return (EOPNOTSUPP);
301	}
302
303	return (0);
304}
305
306static moduledata_t stf_mod = {
307	"if_stf",
308	stfmodevent,
309	0
310};
311
312DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
313
314static int
315stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
316{
317	struct ip ip;
318	struct stf_softc *sc;
319	struct in_addr a, b, mask;
320	struct in6_addr addr6, mask6;
321
322	sc = (struct stf_softc *)arg;
323	if (sc == NULL)
324		return 0;
325
326	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
327		return 0;
328
329	/* IFF_LINK0 means "no decapsulation" */
330	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
331		return 0;
332
333	if (proto != IPPROTO_IPV6)
334		return 0;
335
336	/* LINTED const cast */
337	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
338
339	if (ip.ip_v != 4)
340		return 0;
341
342	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
343		return (0);
344
345	/*
346	 * check if IPv4 dst matches the IPv4 address derived from the
347	 * local 6to4 address.
348	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
349	 */
350	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
351		return 0;
352
353	/*
354	 * check if IPv4 src matches the IPv4 address derived from the
355	 * local 6to4 address masked by prefixmask.
356	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
357	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
358	 */
359	bzero(&a, sizeof(a));
360	bcopy(GET_V4(&addr6), &a, sizeof(a));
361	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
362	a.s_addr &= mask.s_addr;
363	b = ip.ip_src;
364	b.s_addr &= mask.s_addr;
365	if (a.s_addr != b.s_addr)
366		return 0;
367
368	/* stf interface makes single side match only */
369	return 32;
370}
371
372static int
373stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
374{
375	struct rm_priotracker in_ifa_tracker;
376	struct ifaddr *ia;
377	struct in_ifaddr *ia4;
378	struct in6_ifaddr *ia6;
379	struct sockaddr_in6 *sin6;
380	struct in_addr in;
381
382	if_addr_rlock(ifp);
383	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
384		if (ia->ifa_addr->sa_family != AF_INET6)
385			continue;
386		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
387		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
388			continue;
389
390		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
391		IN_IFADDR_RLOCK(&in_ifa_tracker);
392		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
393			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
394				break;
395		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
396		if (ia4 == NULL)
397			continue;
398
399		ia6 = (struct in6_ifaddr *)ia;
400
401		*addr = sin6->sin6_addr;
402		*mask = ia6->ia_prefixmask.sin6_addr;
403		if_addr_runlock(ifp);
404		return (0);
405	}
406	if_addr_runlock(ifp);
407
408	return (ENOENT);
409}
410
411static int
412stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
413    struct route *ro)
414{
415	struct stf_softc *sc;
416	const struct sockaddr_in6 *dst6;
417	struct in_addr in4;
418	const void *ptr;
419	u_int8_t tos;
420	struct ip *ip;
421	struct ip6_hdr *ip6;
422	struct in6_addr addr6, mask6;
423	int error;
424
425#ifdef MAC
426	error = mac_ifnet_check_transmit(ifp, m);
427	if (error) {
428		m_freem(m);
429		return (error);
430	}
431#endif
432
433	sc = ifp->if_softc;
434	dst6 = (const struct sockaddr_in6 *)dst;
435
436	/* just in case */
437	if ((ifp->if_flags & IFF_UP) == 0) {
438		m_freem(m);
439		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
440		return ENETDOWN;
441	}
442
443	/*
444	 * If we don't have an ip4 address that match my inner ip6 address,
445	 * we shouldn't generate output.  Without this check, we'll end up
446	 * using wrong IPv4 source.
447	 */
448	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
449		m_freem(m);
450		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
451		return ENETDOWN;
452	}
453
454	if (m->m_len < sizeof(*ip6)) {
455		m = m_pullup(m, sizeof(*ip6));
456		if (!m) {
457			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
458			return ENOBUFS;
459		}
460	}
461	ip6 = mtod(m, struct ip6_hdr *);
462	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
463
464	/*
465	 * Pickup the right outer dst addr from the list of candidates.
466	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
467	 */
468	ptr = NULL;
469	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
470		ptr = GET_V4(&ip6->ip6_dst);
471	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
472		ptr = GET_V4(&dst6->sin6_addr);
473	else {
474		m_freem(m);
475		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
476		return ENETUNREACH;
477	}
478	bcopy(ptr, &in4, sizeof(in4));
479
480	if (bpf_peers_present(ifp->if_bpf)) {
481		/*
482		 * We need to prepend the address family as
483		 * a four byte field.  Cons up a dummy header
484		 * to pacify bpf.  This is safe because bpf
485		 * will only read from the mbuf (i.e., it won't
486		 * try to free it or keep a pointer a to it).
487		 */
488		u_int af = AF_INET6;
489		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
490	}
491
492	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
493	if (m == NULL) {
494		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
495		return ENOBUFS;
496	}
497	ip = mtod(m, struct ip *);
498
499	bzero(ip, sizeof(*ip));
500
501	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
502	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
503	ip->ip_p = IPPROTO_IPV6;
504	ip->ip_ttl = ip_stf_ttl;
505	ip->ip_len = htons(m->m_pkthdr.len);
506	if (ifp->if_flags & IFF_LINK1)
507		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
508	else
509		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
510
511	M_SETFIB(m, sc->sc_fibnum);
512	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
513	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
514
515	return error;
516}
517
518static int
519isrfc1918addr(struct in_addr *in)
520{
521	/*
522	 * returns 1 if private address range:
523	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
524	 */
525	if (stf_permit_rfc1918 == 0 && (
526	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
527	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
528	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
529		return 1;
530
531	return 0;
532}
533
534static int
535stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
536{
537	struct rm_priotracker in_ifa_tracker;
538	struct in_ifaddr *ia4;
539
540	/*
541	 * reject packets with the following address:
542	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
543	 */
544	if (IN_MULTICAST(ntohl(in->s_addr)))
545		return -1;
546	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
547	case 0: case 127: case 255:
548		return -1;
549	}
550
551	/*
552	 * reject packets with private address range.
553	 * (requirement from RFC3056 section 2 1st paragraph)
554	 */
555	if (isrfc1918addr(in))
556		return -1;
557
558	/*
559	 * reject packets with broadcast
560	 */
561	IN_IFADDR_RLOCK(&in_ifa_tracker);
562	TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
563		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
564			continue;
565		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
566			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
567			return -1;
568		}
569	}
570	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
571
572	/*
573	 * perform ingress filter
574	 */
575	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
576		struct nhop4_basic nh4;
577
578		if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
579			return (-1);
580
581		if (nh4.nh_ifp != inifp)
582			return (-1);
583	}
584
585	return 0;
586}
587
588static int
589stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
590{
591	/*
592	 * check 6to4 addresses
593	 */
594	if (IN6_IS_ADDR_6TO4(in6)) {
595		struct in_addr in4;
596		bcopy(GET_V4(in6), &in4, sizeof(in4));
597		return stf_checkaddr4(sc, &in4, inifp);
598	}
599
600	/*
601	 * reject anything that look suspicious.  the test is implemented
602	 * in ip6_input too, but we check here as well to
603	 * (1) reject bad packets earlier, and
604	 * (2) to be safe against future ip6_input change.
605	 */
606	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
607		return -1;
608
609	return 0;
610}
611
612static int
613in_stf_input(struct mbuf **mp, int *offp, int proto)
614{
615	struct stf_softc *sc;
616	struct ip *ip;
617	struct ip6_hdr *ip6;
618	struct mbuf *m;
619	u_int8_t otos, itos;
620	struct ifnet *ifp;
621	int off;
622
623	m = *mp;
624	off = *offp;
625
626	if (proto != IPPROTO_IPV6) {
627		m_freem(m);
628		return (IPPROTO_DONE);
629	}
630
631	ip = mtod(m, struct ip *);
632
633	sc = (struct stf_softc *)encap_getarg(m);
634
635	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
636		m_freem(m);
637		return (IPPROTO_DONE);
638	}
639
640	ifp = STF2IFP(sc);
641
642#ifdef MAC
643	mac_ifnet_create_mbuf(ifp, m);
644#endif
645
646	/*
647	 * perform sanity check against outer src/dst.
648	 * for source, perform ingress filter as well.
649	 */
650	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
651	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
652		m_freem(m);
653		return (IPPROTO_DONE);
654	}
655
656	otos = ip->ip_tos;
657	m_adj(m, off);
658
659	if (m->m_len < sizeof(*ip6)) {
660		m = m_pullup(m, sizeof(*ip6));
661		if (!m)
662			return (IPPROTO_DONE);
663	}
664	ip6 = mtod(m, struct ip6_hdr *);
665
666	/*
667	 * perform sanity check against inner src/dst.
668	 * for source, perform ingress filter as well.
669	 */
670	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
671	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
672		m_freem(m);
673		return (IPPROTO_DONE);
674	}
675
676	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
677	if ((ifp->if_flags & IFF_LINK1) != 0)
678		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
679	else
680		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
681	ip6->ip6_flow &= ~htonl(0xff << 20);
682	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
683
684	m->m_pkthdr.rcvif = ifp;
685
686	if (bpf_peers_present(ifp->if_bpf)) {
687		/*
688		 * We need to prepend the address family as
689		 * a four byte field.  Cons up a dummy header
690		 * to pacify bpf.  This is safe because bpf
691		 * will only read from the mbuf (i.e., it won't
692		 * try to free it or keep a pointer a to it).
693		 */
694		u_int32_t af = AF_INET6;
695		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
696	}
697
698	/*
699	 * Put the packet to the network layer input queue according to the
700	 * specified address family.
701	 * See net/if_gif.c for possible issues with packet processing
702	 * reorder due to extra queueing.
703	 */
704	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
705	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
706	M_SETFIB(m, ifp->if_fib);
707	netisr_dispatch(NETISR_IPV6, m);
708	return (IPPROTO_DONE);
709}
710
711static int
712stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
713{
714	struct ifaddr *ifa;
715	struct ifreq *ifr;
716	struct sockaddr_in6 *sin6;
717	struct in_addr addr;
718	int error, mtu;
719
720	error = 0;
721	switch (cmd) {
722	case SIOCSIFADDR:
723		ifa = (struct ifaddr *)data;
724		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
725			error = EAFNOSUPPORT;
726			break;
727		}
728		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
729		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
730			error = EINVAL;
731			break;
732		}
733		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
734		if (isrfc1918addr(&addr)) {
735			error = EINVAL;
736			break;
737		}
738
739		ifp->if_flags |= IFF_UP;
740		break;
741
742	case SIOCADDMULTI:
743	case SIOCDELMULTI:
744		ifr = (struct ifreq *)data;
745		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
746			;
747		else
748			error = EAFNOSUPPORT;
749		break;
750
751	case SIOCGIFMTU:
752		break;
753
754	case SIOCSIFMTU:
755		ifr = (struct ifreq *)data;
756		mtu = ifr->ifr_mtu;
757		/* RFC 4213 3.2 ideal world MTU */
758		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
759			return (EINVAL);
760		ifp->if_mtu = mtu;
761		break;
762
763	default:
764		error = EINVAL;
765		break;
766	}
767
768	return error;
769}
770