1/* $FreeBSD: head/sys/net/if_stf.c 241610 2012-10-16 13:37:54Z glebius $ */
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
80#include <sys/param.h>
81#include <sys/systm.h>
82#include <sys/socket.h>
83#include <sys/sockio.h>
84#include <sys/mbuf.h>
85#include <sys/errno.h>
86#include <sys/kernel.h>
87#include <sys/module.h>
88#include <sys/protosw.h>
89#include <sys/proc.h>
90#include <sys/queue.h>
91#include <sys/sysctl.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#include <net/vnet.h>
103
104#include <netinet/in.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_route_cache = 1;
127SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
128 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
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) ((caddr_t)(&(x)->s6_addr16[1]))
139
140struct stf_softc {
141 struct ifnet *sc_ifp;
142 union {
143 struct route __sc_ro4;
144 struct route_in6 __sc_ro6; /* just for safety */
145 } __sc_ro46;
146#define sc_ro __sc_ro46.__sc_ro4
147 struct mtx sc_ro_mtx;
148 u_int sc_fibnum;
149 const struct encaptab *encap_cookie;
150};
151#define STF2IFP(sc) ((sc)->sc_ifp)
152
153static const char stfname[] = "stf";
154
155/*
156 * Note that mutable fields in the softc are not currently locked.
157 * We do lock sc_ro in stf_output though.
158 */
159static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
160static const int ip_stf_ttl = 40;
161
162extern struct domain inetdomain;
163struct protosw in_stf_protosw = {
164 .pr_type = SOCK_RAW,
165 .pr_domain = &inetdomain,
166 .pr_protocol = IPPROTO_IPV6,
167 .pr_flags = PR_ATOMIC|PR_ADDR,
168 .pr_input = in_stf_input,
169 .pr_output = (pr_output_t *)rip_output,
170 .pr_ctloutput = rip_ctloutput,
171 .pr_usrreqs = &rip_usrreqs
172};
173
174static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
175
176static int stfmodevent(module_t, int, void *);
177static int stf_encapcheck(const struct mbuf *, int, int, void *);
178static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
179static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
180 struct route *);
181static int isrfc1918addr(struct in_addr *);
182static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
183 struct ifnet *);
184static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
185 struct ifnet *);
186static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
187static int stf_ioctl(struct ifnet *, u_long, caddr_t);
188
189static int stf_clone_match(struct if_clone *, const char *);
190static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
191static int stf_clone_destroy(struct if_clone *, struct ifnet *);
192static struct if_clone *stf_cloner;
193
194static int
195stf_clone_match(struct if_clone *ifc, const char *name)
196{
197 int i;
198
199 for(i = 0; stfnames[i] != NULL; i++) {
200 if (strcmp(stfnames[i], name) == 0)
201 return (1);
202 }
203
204 return (0);
205}
206
207static int
208stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
209{
210 int err, unit;
211 struct stf_softc *sc;
212 struct ifnet *ifp;
213
214 /*
215 * We can only have one unit, but since unit allocation is
216 * already locked, we use it to keep from allocating extra
217 * interfaces.
218 */
219 unit = STFUNIT;
220 err = ifc_alloc_unit(ifc, &unit);
221 if (err != 0)
222 return (err);
223
224 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
225 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
226 if (ifp == NULL) {
227 free(sc, M_STF);
228 ifc_free_unit(ifc, unit);
229 return (ENOSPC);
230 }
231 ifp->if_softc = sc;
232 sc->sc_fibnum = curthread->td_proc->p_fibnum;
233
234 /*
235 * Set the name manually rather then using if_initname because
236 * we don't conform to the default naming convention for interfaces.
237 */
238 strlcpy(ifp->if_xname, name, IFNAMSIZ);
239 ifp->if_dname = stfname;
240 ifp->if_dunit = IF_DUNIT_NONE;
241
242 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
243 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
244 stf_encapcheck, &in_stf_protosw, sc);
245 if (sc->encap_cookie == NULL) {
246 if_printf(ifp, "attach failed\n");
247 free(sc, M_STF);
248 ifc_free_unit(ifc, unit);
249 return (ENOMEM);
250 }
251
252 ifp->if_mtu = IPV6_MMTU;
253 ifp->if_ioctl = stf_ioctl;
254 ifp->if_output = stf_output;
255 ifp->if_snd.ifq_maxlen = ifqmaxlen;
256 if_attach(ifp);
257 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
258 return (0);
259}
260
261static int
262stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
263{
264 struct stf_softc *sc = ifp->if_softc;
265 int err;
266
267 err = encap_detach(sc->encap_cookie);
268 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
269 mtx_destroy(&(sc)->sc_ro_mtx);
270 bpfdetach(ifp);
271 if_detach(ifp);
272 if_free(ifp);
273
274 free(sc, M_STF);
275 ifc_free_unit(ifc, STFUNIT);
276
277 return (0);
278}
279
280static int
281stfmodevent(mod, type, data)
282 module_t mod;
283 int type;
284 void *data;
285{
286
287 switch (type) {
288 case MOD_LOAD:
289 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
290 stf_clone_create, stf_clone_destroy);
291 break;
292 case MOD_UNLOAD:
293 if_clone_detach(stf_cloner);
294 break;
295 default:
296 return (EOPNOTSUPP);
297 }
298
299 return (0);
300}
301
302static moduledata_t stf_mod = {
303 "if_stf",
304 stfmodevent,
305 0
306};
307
308DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
309
310static int
311stf_encapcheck(m, off, proto, arg)
312 const struct mbuf *m;
313 int off;
314 int proto;
315 void *arg;
316{
317 struct ip ip;
318 struct in6_ifaddr *ia6;
319 struct stf_softc *sc;
320 struct in_addr a, b, mask;
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 ia6 = stf_getsrcifa6(STF2IFP(sc));
343 if (ia6 == NULL)
344 return 0;
345
346 /*
347 * check if IPv4 dst matches the IPv4 address derived from the
348 * local 6to4 address.
349 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
350 */
351 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
352 sizeof(ip.ip_dst)) != 0) {
353 ifa_free(&ia6->ia_ifa);
354 return 0;
355 }
356
357 /*
358 * check if IPv4 src matches the IPv4 address derived from the
359 * local 6to4 address masked by prefixmask.
360 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
361 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
362 */
363 bzero(&a, sizeof(a));
364 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
365 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
366 ifa_free(&ia6->ia_ifa);
367 a.s_addr &= mask.s_addr;
368 b = ip.ip_src;
369 b.s_addr &= mask.s_addr;
370 if (a.s_addr != b.s_addr)
371 return 0;
372
373 /* stf interface makes single side match only */
374 return 32;
375}
376
377static struct in6_ifaddr *
378stf_getsrcifa6(ifp)
379 struct ifnet *ifp;
380{
381 struct ifaddr *ia;
382 struct in_ifaddr *ia4;
383 struct sockaddr_in6 *sin6;
384 struct in_addr in;
385
386 if_addr_rlock(ifp);
387 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
388 if (ia->ifa_addr->sa_family != AF_INET6)
389 continue;
390 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
391 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
392 continue;
393
394 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
395 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
396 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
397 break;
398 if (ia4 == NULL)
399 continue;
400
401 ifa_ref(ia);
402 if_addr_runlock(ifp);
403 return (struct in6_ifaddr *)ia;
404 }
405 if_addr_runlock(ifp);
406
407 return NULL;
408}
409
410static int
411stf_output(ifp, m, dst, ro)
412 struct ifnet *ifp;
413 struct mbuf *m;
414 struct sockaddr *dst;
415 struct route *ro;
416{
417 struct stf_softc *sc;
418 struct sockaddr_in6 *dst6;
419 struct route *cached_route;
420 struct in_addr in4;
421 caddr_t ptr;
422 struct sockaddr_in *dst4;
423 u_int8_t tos;
424 struct ip *ip;
425 struct ip6_hdr *ip6;
426 struct in6_ifaddr *ia6;
427 u_int32_t af;
428 int error;
429
430#ifdef MAC
431 error = mac_ifnet_check_transmit(ifp, m);
432 if (error) {
433 m_freem(m);
434 return (error);
435 }
436#endif
437
438 sc = ifp->if_softc;
439 dst6 = (struct sockaddr_in6 *)dst;
440
441 /* just in case */
442 if ((ifp->if_flags & IFF_UP) == 0) {
443 m_freem(m);
444 ifp->if_oerrors++;
445 return ENETDOWN;
446 }
447
448 /*
449 * If we don't have an ip4 address that match my inner ip6 address,
450 * we shouldn't generate output. Without this check, we'll end up
451 * using wrong IPv4 source.
452 */
453 ia6 = stf_getsrcifa6(ifp);
454 if (ia6 == NULL) {
455 m_freem(m);
456 ifp->if_oerrors++;
457 return ENETDOWN;
458 }
459
460 if (m->m_len < sizeof(*ip6)) {
461 m = m_pullup(m, sizeof(*ip6));
462 if (!m) {
463 ifa_free(&ia6->ia_ifa);
464 ifp->if_oerrors++;
465 return ENOBUFS;
466 }
467 }
468 ip6 = mtod(m, struct ip6_hdr *);
469 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
470
471 /*
472 * BPF writes need to be handled specially.
473 * This is a null operation, nothing here checks dst->sa_family.
474 */
475 if (dst->sa_family == AF_UNSPEC) {
476 bcopy(dst->sa_data, &af, sizeof(af));
477 dst->sa_family = af;
478 }
479
480 /*
481 * Pickup the right outer dst addr from the list of candidates.
482 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
483 */
484 ptr = NULL;
485 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
486 ptr = GET_V4(&ip6->ip6_dst);
487 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
488 ptr = GET_V4(&dst6->sin6_addr);
489 else {
490 ifa_free(&ia6->ia_ifa);
491 m_freem(m);
492 ifp->if_oerrors++;
493 return ENETUNREACH;
494 }
495 bcopy(ptr, &in4, sizeof(in4));
496
497 if (bpf_peers_present(ifp->if_bpf)) {
498 /*
499 * We need to prepend the address family as
500 * a four byte field. Cons up a dummy header
501 * to pacify bpf. This is safe because bpf
502 * will only read from the mbuf (i.e., it won't
503 * try to free it or keep a pointer a to it).
504 */
505 af = AF_INET6;
506 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
507 }
508
509 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
510 if (m && m->m_len < sizeof(struct ip))
511 m = m_pullup(m, sizeof(struct ip));
512 if (m == NULL) {
513 ifa_free(&ia6->ia_ifa);
514 ifp->if_oerrors++;
515 return ENOBUFS;
516 }
517 ip = mtod(m, struct ip *);
518
519 bzero(ip, sizeof(*ip));
520
521 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
522 &ip->ip_src, sizeof(ip->ip_src));
523 ifa_free(&ia6->ia_ifa);
524 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
525 ip->ip_p = IPPROTO_IPV6;
526 ip->ip_ttl = ip_stf_ttl;
527 ip->ip_len = m->m_pkthdr.len; /*host order*/
528 if (ifp->if_flags & IFF_LINK1)
529 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
530 else
531 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
532
533 if (!stf_route_cache) {
534 cached_route = NULL;
535 goto sendit;
536 }
537
538 /*
539 * Do we have a cached route?
540 */
541 mtx_lock(&(sc)->sc_ro_mtx);
542 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
543 if (dst4->sin_family != AF_INET ||
544 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
545 /* cache route doesn't match */
546 dst4->sin_family = AF_INET;
547 dst4->sin_len = sizeof(struct sockaddr_in);
548 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
549 if (sc->sc_ro.ro_rt) {
550 RTFREE(sc->sc_ro.ro_rt);
551 sc->sc_ro.ro_rt = NULL;
552 }
553 }
554
555 if (sc->sc_ro.ro_rt == NULL) {
556 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
557 if (sc->sc_ro.ro_rt == NULL) {
558 m_freem(m);
559 mtx_unlock(&(sc)->sc_ro_mtx);
560 ifp->if_oerrors++;
561 return ENETUNREACH;
562 }
563 }
564 cached_route = &sc->sc_ro;
565
566sendit:
567 M_SETFIB(m, sc->sc_fibnum);
568 ifp->if_opackets++;
569 error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
570
571 if (cached_route != NULL)
572 mtx_unlock(&(sc)->sc_ro_mtx);
573 return error;
574}
575
576static int
577isrfc1918addr(in)
578 struct in_addr *in;
579{
580 /*
581 * returns 1 if private address range:
582 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
583 */
584 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
585 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
586 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
587 return 1;
588
589 return 0;
590}
591
592static int
593stf_checkaddr4(sc, in, inifp)
594 struct stf_softc *sc;
595 struct in_addr *in;
596 struct ifnet *inifp; /* incoming interface */
597{
598 struct in_ifaddr *ia4;
599
600 /*
601 * reject packets with the following address:
602 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
603 */
604 if (IN_MULTICAST(ntohl(in->s_addr)))
605 return -1;
606 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
607 case 0: case 127: case 255:
608 return -1;
609 }
610
611 /*
612 * reject packets with private address range.
613 * (requirement from RFC3056 section 2 1st paragraph)
614 */
615 if (isrfc1918addr(in))
616 return -1;
617
618 /*
619 * reject packets with broadcast
620 */
621 IN_IFADDR_RLOCK();
622 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
623 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
624 continue;
625 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
626 IN_IFADDR_RUNLOCK();
627 return -1;
628 }
629 }
630 IN_IFADDR_RUNLOCK();
631
632 /*
633 * perform ingress filter
634 */
635 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
636 struct sockaddr_in sin;
637 struct rtentry *rt;
638
639 bzero(&sin, sizeof(sin));
640 sin.sin_family = AF_INET;
641 sin.sin_len = sizeof(struct sockaddr_in);
642 sin.sin_addr = *in;
643 rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
644 0UL, sc->sc_fibnum);
645 if (!rt || rt->rt_ifp != inifp) {
646#if 0
647 log(LOG_WARNING, "%s: packet from 0x%x dropped "
648 "due to ingress filter\n", if_name(STF2IFP(sc)),
649 (u_int32_t)ntohl(sin.sin_addr.s_addr));
650#endif
651 if (rt)
652 RTFREE_LOCKED(rt);
653 return -1;
654 }
655 RTFREE_LOCKED(rt);
656 }
657
658 return 0;
659}
660
661static int
662stf_checkaddr6(sc, in6, inifp)
663 struct stf_softc *sc;
664 struct in6_addr *in6;
665 struct ifnet *inifp; /* incoming interface */
666{
667 /*
668 * check 6to4 addresses
669 */
670 if (IN6_IS_ADDR_6TO4(in6)) {
671 struct in_addr in4;
672 bcopy(GET_V4(in6), &in4, sizeof(in4));
673 return stf_checkaddr4(sc, &in4, inifp);
674 }
675
676 /*
677 * reject anything that look suspicious. the test is implemented
678 * in ip6_input too, but we check here as well to
679 * (1) reject bad packets earlier, and
680 * (2) to be safe against future ip6_input change.
681 */
682 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
683 return -1;
684
685 return 0;
686}
687
688void
689in_stf_input(m, off)
690 struct mbuf *m;
691 int off;
692{
693 int proto;
694 struct stf_softc *sc;
695 struct ip *ip;
696 struct ip6_hdr *ip6;
697 u_int8_t otos, itos;
698 struct ifnet *ifp;
699
700 proto = mtod(m, struct ip *)->ip_p;
701
702 if (proto != IPPROTO_IPV6) {
703 m_freem(m);
704 return;
705 }
706
707 ip = mtod(m, struct ip *);
708
709 sc = (struct stf_softc *)encap_getarg(m);
710
711 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
712 m_freem(m);
713 return;
714 }
715
716 ifp = STF2IFP(sc);
717
718#ifdef MAC
719 mac_ifnet_create_mbuf(ifp, m);
720#endif
721
722 /*
723 * perform sanity check against outer src/dst.
724 * for source, perform ingress filter as well.
725 */
726 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
727 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
728 m_freem(m);
729 return;
730 }
731
732 otos = ip->ip_tos;
733 m_adj(m, off);
734
735 if (m->m_len < sizeof(*ip6)) {
736 m = m_pullup(m, sizeof(*ip6));
737 if (!m)
738 return;
739 }
740 ip6 = mtod(m, struct ip6_hdr *);
741
742 /*
743 * perform sanity check against inner src/dst.
744 * for source, perform ingress filter as well.
745 */
746 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
747 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
748 m_freem(m);
749 return;
750 }
751
752 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
753 if ((ifp->if_flags & IFF_LINK1) != 0)
754 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
755 else
756 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
757 ip6->ip6_flow &= ~htonl(0xff << 20);
758 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
759
760 m->m_pkthdr.rcvif = ifp;
761
762 if (bpf_peers_present(ifp->if_bpf)) {
763 /*
764 * We need to prepend the address family as
765 * a four byte field. Cons up a dummy header
766 * to pacify bpf. This is safe because bpf
767 * will only read from the mbuf (i.e., it won't
768 * try to free it or keep a pointer a to it).
769 */
770 u_int32_t af = AF_INET6;
771 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
772 }
773
774 /*
775 * Put the packet to the network layer input queue according to the
776 * specified address family.
777 * See net/if_gif.c for possible issues with packet processing
778 * reorder due to extra queueing.
779 */
780 ifp->if_ipackets++;
781 ifp->if_ibytes += m->m_pkthdr.len;
782 M_SETFIB(m, ifp->if_fib);
783 netisr_dispatch(NETISR_IPV6, m);
784}
785
786/* ARGSUSED */
787static void
788stf_rtrequest(cmd, rt, info)
789 int cmd;
790 struct rtentry *rt;
791 struct rt_addrinfo *info;
792{
793 RT_LOCK_ASSERT(rt);
794 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
795}
796
797static int
798stf_ioctl(ifp, cmd, data)
799 struct ifnet *ifp;
800 u_long cmd;
801 caddr_t data;
802{
803 struct ifaddr *ifa;
804 struct ifreq *ifr;
805 struct sockaddr_in6 *sin6;
806 struct in_addr addr;
807 int error, mtu;
808
809 error = 0;
810 switch (cmd) {
811 case SIOCSIFADDR:
812 ifa = (struct ifaddr *)data;
813 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
814 error = EAFNOSUPPORT;
815 break;
816 }
817 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
818 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
819 error = EINVAL;
820 break;
821 }
822 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
823 if (isrfc1918addr(&addr)) {
824 error = EINVAL;
825 break;
826 }
827
828 ifa->ifa_rtrequest = stf_rtrequest;
829 ifp->if_flags |= IFF_UP;
830 break;
831
832 case SIOCADDMULTI:
833 case SIOCDELMULTI:
834 ifr = (struct ifreq *)data;
835 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
836 ;
837 else
838 error = EAFNOSUPPORT;
839 break;
840
841 case SIOCGIFMTU:
842 break;
843
844 case SIOCSIFMTU:
845 ifr = (struct ifreq *)data;
846 mtu = ifr->ifr_mtu;
847 /* RFC 4213 3.2 ideal world MTU */
848 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
849 return (EINVAL);
850 ifp->if_mtu = mtu;
851 break;
852
853 default:
854 error = EINVAL;
855 break;
856 }
857
858 return error;
859}
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
80#include <sys/param.h>
81#include <sys/systm.h>
82#include <sys/socket.h>
83#include <sys/sockio.h>
84#include <sys/mbuf.h>
85#include <sys/errno.h>
86#include <sys/kernel.h>
87#include <sys/module.h>
88#include <sys/protosw.h>
89#include <sys/proc.h>
90#include <sys/queue.h>
91#include <sys/sysctl.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#include <net/vnet.h>
103
104#include <netinet/in.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_route_cache = 1;
127SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
128 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
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) ((caddr_t)(&(x)->s6_addr16[1]))
139
140struct stf_softc {
141 struct ifnet *sc_ifp;
142 union {
143 struct route __sc_ro4;
144 struct route_in6 __sc_ro6; /* just for safety */
145 } __sc_ro46;
146#define sc_ro __sc_ro46.__sc_ro4
147 struct mtx sc_ro_mtx;
148 u_int sc_fibnum;
149 const struct encaptab *encap_cookie;
150};
151#define STF2IFP(sc) ((sc)->sc_ifp)
152
153static const char stfname[] = "stf";
154
155/*
156 * Note that mutable fields in the softc are not currently locked.
157 * We do lock sc_ro in stf_output though.
158 */
159static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
160static const int ip_stf_ttl = 40;
161
162extern struct domain inetdomain;
163struct protosw in_stf_protosw = {
164 .pr_type = SOCK_RAW,
165 .pr_domain = &inetdomain,
166 .pr_protocol = IPPROTO_IPV6,
167 .pr_flags = PR_ATOMIC|PR_ADDR,
168 .pr_input = in_stf_input,
169 .pr_output = (pr_output_t *)rip_output,
170 .pr_ctloutput = rip_ctloutput,
171 .pr_usrreqs = &rip_usrreqs
172};
173
174static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
175
176static int stfmodevent(module_t, int, void *);
177static int stf_encapcheck(const struct mbuf *, int, int, void *);
178static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
179static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
180 struct route *);
181static int isrfc1918addr(struct in_addr *);
182static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
183 struct ifnet *);
184static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
185 struct ifnet *);
186static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
187static int stf_ioctl(struct ifnet *, u_long, caddr_t);
188
189static int stf_clone_match(struct if_clone *, const char *);
190static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
191static int stf_clone_destroy(struct if_clone *, struct ifnet *);
192static struct if_clone *stf_cloner;
193
194static int
195stf_clone_match(struct if_clone *ifc, const char *name)
196{
197 int i;
198
199 for(i = 0; stfnames[i] != NULL; i++) {
200 if (strcmp(stfnames[i], name) == 0)
201 return (1);
202 }
203
204 return (0);
205}
206
207static int
208stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
209{
210 int err, unit;
211 struct stf_softc *sc;
212 struct ifnet *ifp;
213
214 /*
215 * We can only have one unit, but since unit allocation is
216 * already locked, we use it to keep from allocating extra
217 * interfaces.
218 */
219 unit = STFUNIT;
220 err = ifc_alloc_unit(ifc, &unit);
221 if (err != 0)
222 return (err);
223
224 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
225 ifp = STF2IFP(sc) = if_alloc(IFT_STF);
226 if (ifp == NULL) {
227 free(sc, M_STF);
228 ifc_free_unit(ifc, unit);
229 return (ENOSPC);
230 }
231 ifp->if_softc = sc;
232 sc->sc_fibnum = curthread->td_proc->p_fibnum;
233
234 /*
235 * Set the name manually rather then using if_initname because
236 * we don't conform to the default naming convention for interfaces.
237 */
238 strlcpy(ifp->if_xname, name, IFNAMSIZ);
239 ifp->if_dname = stfname;
240 ifp->if_dunit = IF_DUNIT_NONE;
241
242 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
243 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
244 stf_encapcheck, &in_stf_protosw, sc);
245 if (sc->encap_cookie == NULL) {
246 if_printf(ifp, "attach failed\n");
247 free(sc, M_STF);
248 ifc_free_unit(ifc, unit);
249 return (ENOMEM);
250 }
251
252 ifp->if_mtu = IPV6_MMTU;
253 ifp->if_ioctl = stf_ioctl;
254 ifp->if_output = stf_output;
255 ifp->if_snd.ifq_maxlen = ifqmaxlen;
256 if_attach(ifp);
257 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
258 return (0);
259}
260
261static int
262stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
263{
264 struct stf_softc *sc = ifp->if_softc;
265 int err;
266
267 err = encap_detach(sc->encap_cookie);
268 KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
269 mtx_destroy(&(sc)->sc_ro_mtx);
270 bpfdetach(ifp);
271 if_detach(ifp);
272 if_free(ifp);
273
274 free(sc, M_STF);
275 ifc_free_unit(ifc, STFUNIT);
276
277 return (0);
278}
279
280static int
281stfmodevent(mod, type, data)
282 module_t mod;
283 int type;
284 void *data;
285{
286
287 switch (type) {
288 case MOD_LOAD:
289 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
290 stf_clone_create, stf_clone_destroy);
291 break;
292 case MOD_UNLOAD:
293 if_clone_detach(stf_cloner);
294 break;
295 default:
296 return (EOPNOTSUPP);
297 }
298
299 return (0);
300}
301
302static moduledata_t stf_mod = {
303 "if_stf",
304 stfmodevent,
305 0
306};
307
308DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
309
310static int
311stf_encapcheck(m, off, proto, arg)
312 const struct mbuf *m;
313 int off;
314 int proto;
315 void *arg;
316{
317 struct ip ip;
318 struct in6_ifaddr *ia6;
319 struct stf_softc *sc;
320 struct in_addr a, b, mask;
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 ia6 = stf_getsrcifa6(STF2IFP(sc));
343 if (ia6 == NULL)
344 return 0;
345
346 /*
347 * check if IPv4 dst matches the IPv4 address derived from the
348 * local 6to4 address.
349 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
350 */
351 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
352 sizeof(ip.ip_dst)) != 0) {
353 ifa_free(&ia6->ia_ifa);
354 return 0;
355 }
356
357 /*
358 * check if IPv4 src matches the IPv4 address derived from the
359 * local 6to4 address masked by prefixmask.
360 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
361 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
362 */
363 bzero(&a, sizeof(a));
364 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
365 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
366 ifa_free(&ia6->ia_ifa);
367 a.s_addr &= mask.s_addr;
368 b = ip.ip_src;
369 b.s_addr &= mask.s_addr;
370 if (a.s_addr != b.s_addr)
371 return 0;
372
373 /* stf interface makes single side match only */
374 return 32;
375}
376
377static struct in6_ifaddr *
378stf_getsrcifa6(ifp)
379 struct ifnet *ifp;
380{
381 struct ifaddr *ia;
382 struct in_ifaddr *ia4;
383 struct sockaddr_in6 *sin6;
384 struct in_addr in;
385
386 if_addr_rlock(ifp);
387 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
388 if (ia->ifa_addr->sa_family != AF_INET6)
389 continue;
390 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
391 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
392 continue;
393
394 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
395 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
396 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
397 break;
398 if (ia4 == NULL)
399 continue;
400
401 ifa_ref(ia);
402 if_addr_runlock(ifp);
403 return (struct in6_ifaddr *)ia;
404 }
405 if_addr_runlock(ifp);
406
407 return NULL;
408}
409
410static int
411stf_output(ifp, m, dst, ro)
412 struct ifnet *ifp;
413 struct mbuf *m;
414 struct sockaddr *dst;
415 struct route *ro;
416{
417 struct stf_softc *sc;
418 struct sockaddr_in6 *dst6;
419 struct route *cached_route;
420 struct in_addr in4;
421 caddr_t ptr;
422 struct sockaddr_in *dst4;
423 u_int8_t tos;
424 struct ip *ip;
425 struct ip6_hdr *ip6;
426 struct in6_ifaddr *ia6;
427 u_int32_t af;
428 int error;
429
430#ifdef MAC
431 error = mac_ifnet_check_transmit(ifp, m);
432 if (error) {
433 m_freem(m);
434 return (error);
435 }
436#endif
437
438 sc = ifp->if_softc;
439 dst6 = (struct sockaddr_in6 *)dst;
440
441 /* just in case */
442 if ((ifp->if_flags & IFF_UP) == 0) {
443 m_freem(m);
444 ifp->if_oerrors++;
445 return ENETDOWN;
446 }
447
448 /*
449 * If we don't have an ip4 address that match my inner ip6 address,
450 * we shouldn't generate output. Without this check, we'll end up
451 * using wrong IPv4 source.
452 */
453 ia6 = stf_getsrcifa6(ifp);
454 if (ia6 == NULL) {
455 m_freem(m);
456 ifp->if_oerrors++;
457 return ENETDOWN;
458 }
459
460 if (m->m_len < sizeof(*ip6)) {
461 m = m_pullup(m, sizeof(*ip6));
462 if (!m) {
463 ifa_free(&ia6->ia_ifa);
464 ifp->if_oerrors++;
465 return ENOBUFS;
466 }
467 }
468 ip6 = mtod(m, struct ip6_hdr *);
469 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
470
471 /*
472 * BPF writes need to be handled specially.
473 * This is a null operation, nothing here checks dst->sa_family.
474 */
475 if (dst->sa_family == AF_UNSPEC) {
476 bcopy(dst->sa_data, &af, sizeof(af));
477 dst->sa_family = af;
478 }
479
480 /*
481 * Pickup the right outer dst addr from the list of candidates.
482 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
483 */
484 ptr = NULL;
485 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
486 ptr = GET_V4(&ip6->ip6_dst);
487 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
488 ptr = GET_V4(&dst6->sin6_addr);
489 else {
490 ifa_free(&ia6->ia_ifa);
491 m_freem(m);
492 ifp->if_oerrors++;
493 return ENETUNREACH;
494 }
495 bcopy(ptr, &in4, sizeof(in4));
496
497 if (bpf_peers_present(ifp->if_bpf)) {
498 /*
499 * We need to prepend the address family as
500 * a four byte field. Cons up a dummy header
501 * to pacify bpf. This is safe because bpf
502 * will only read from the mbuf (i.e., it won't
503 * try to free it or keep a pointer a to it).
504 */
505 af = AF_INET6;
506 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
507 }
508
509 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
510 if (m && m->m_len < sizeof(struct ip))
511 m = m_pullup(m, sizeof(struct ip));
512 if (m == NULL) {
513 ifa_free(&ia6->ia_ifa);
514 ifp->if_oerrors++;
515 return ENOBUFS;
516 }
517 ip = mtod(m, struct ip *);
518
519 bzero(ip, sizeof(*ip));
520
521 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
522 &ip->ip_src, sizeof(ip->ip_src));
523 ifa_free(&ia6->ia_ifa);
524 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
525 ip->ip_p = IPPROTO_IPV6;
526 ip->ip_ttl = ip_stf_ttl;
527 ip->ip_len = m->m_pkthdr.len; /*host order*/
528 if (ifp->if_flags & IFF_LINK1)
529 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
530 else
531 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
532
533 if (!stf_route_cache) {
534 cached_route = NULL;
535 goto sendit;
536 }
537
538 /*
539 * Do we have a cached route?
540 */
541 mtx_lock(&(sc)->sc_ro_mtx);
542 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
543 if (dst4->sin_family != AF_INET ||
544 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
545 /* cache route doesn't match */
546 dst4->sin_family = AF_INET;
547 dst4->sin_len = sizeof(struct sockaddr_in);
548 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
549 if (sc->sc_ro.ro_rt) {
550 RTFREE(sc->sc_ro.ro_rt);
551 sc->sc_ro.ro_rt = NULL;
552 }
553 }
554
555 if (sc->sc_ro.ro_rt == NULL) {
556 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
557 if (sc->sc_ro.ro_rt == NULL) {
558 m_freem(m);
559 mtx_unlock(&(sc)->sc_ro_mtx);
560 ifp->if_oerrors++;
561 return ENETUNREACH;
562 }
563 }
564 cached_route = &sc->sc_ro;
565
566sendit:
567 M_SETFIB(m, sc->sc_fibnum);
568 ifp->if_opackets++;
569 error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
570
571 if (cached_route != NULL)
572 mtx_unlock(&(sc)->sc_ro_mtx);
573 return error;
574}
575
576static int
577isrfc1918addr(in)
578 struct in_addr *in;
579{
580 /*
581 * returns 1 if private address range:
582 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
583 */
584 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
585 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
586 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
587 return 1;
588
589 return 0;
590}
591
592static int
593stf_checkaddr4(sc, in, inifp)
594 struct stf_softc *sc;
595 struct in_addr *in;
596 struct ifnet *inifp; /* incoming interface */
597{
598 struct in_ifaddr *ia4;
599
600 /*
601 * reject packets with the following address:
602 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
603 */
604 if (IN_MULTICAST(ntohl(in->s_addr)))
605 return -1;
606 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
607 case 0: case 127: case 255:
608 return -1;
609 }
610
611 /*
612 * reject packets with private address range.
613 * (requirement from RFC3056 section 2 1st paragraph)
614 */
615 if (isrfc1918addr(in))
616 return -1;
617
618 /*
619 * reject packets with broadcast
620 */
621 IN_IFADDR_RLOCK();
622 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
623 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
624 continue;
625 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
626 IN_IFADDR_RUNLOCK();
627 return -1;
628 }
629 }
630 IN_IFADDR_RUNLOCK();
631
632 /*
633 * perform ingress filter
634 */
635 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
636 struct sockaddr_in sin;
637 struct rtentry *rt;
638
639 bzero(&sin, sizeof(sin));
640 sin.sin_family = AF_INET;
641 sin.sin_len = sizeof(struct sockaddr_in);
642 sin.sin_addr = *in;
643 rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
644 0UL, sc->sc_fibnum);
645 if (!rt || rt->rt_ifp != inifp) {
646#if 0
647 log(LOG_WARNING, "%s: packet from 0x%x dropped "
648 "due to ingress filter\n", if_name(STF2IFP(sc)),
649 (u_int32_t)ntohl(sin.sin_addr.s_addr));
650#endif
651 if (rt)
652 RTFREE_LOCKED(rt);
653 return -1;
654 }
655 RTFREE_LOCKED(rt);
656 }
657
658 return 0;
659}
660
661static int
662stf_checkaddr6(sc, in6, inifp)
663 struct stf_softc *sc;
664 struct in6_addr *in6;
665 struct ifnet *inifp; /* incoming interface */
666{
667 /*
668 * check 6to4 addresses
669 */
670 if (IN6_IS_ADDR_6TO4(in6)) {
671 struct in_addr in4;
672 bcopy(GET_V4(in6), &in4, sizeof(in4));
673 return stf_checkaddr4(sc, &in4, inifp);
674 }
675
676 /*
677 * reject anything that look suspicious. the test is implemented
678 * in ip6_input too, but we check here as well to
679 * (1) reject bad packets earlier, and
680 * (2) to be safe against future ip6_input change.
681 */
682 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
683 return -1;
684
685 return 0;
686}
687
688void
689in_stf_input(m, off)
690 struct mbuf *m;
691 int off;
692{
693 int proto;
694 struct stf_softc *sc;
695 struct ip *ip;
696 struct ip6_hdr *ip6;
697 u_int8_t otos, itos;
698 struct ifnet *ifp;
699
700 proto = mtod(m, struct ip *)->ip_p;
701
702 if (proto != IPPROTO_IPV6) {
703 m_freem(m);
704 return;
705 }
706
707 ip = mtod(m, struct ip *);
708
709 sc = (struct stf_softc *)encap_getarg(m);
710
711 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
712 m_freem(m);
713 return;
714 }
715
716 ifp = STF2IFP(sc);
717
718#ifdef MAC
719 mac_ifnet_create_mbuf(ifp, m);
720#endif
721
722 /*
723 * perform sanity check against outer src/dst.
724 * for source, perform ingress filter as well.
725 */
726 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
727 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
728 m_freem(m);
729 return;
730 }
731
732 otos = ip->ip_tos;
733 m_adj(m, off);
734
735 if (m->m_len < sizeof(*ip6)) {
736 m = m_pullup(m, sizeof(*ip6));
737 if (!m)
738 return;
739 }
740 ip6 = mtod(m, struct ip6_hdr *);
741
742 /*
743 * perform sanity check against inner src/dst.
744 * for source, perform ingress filter as well.
745 */
746 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
747 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
748 m_freem(m);
749 return;
750 }
751
752 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
753 if ((ifp->if_flags & IFF_LINK1) != 0)
754 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
755 else
756 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
757 ip6->ip6_flow &= ~htonl(0xff << 20);
758 ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
759
760 m->m_pkthdr.rcvif = ifp;
761
762 if (bpf_peers_present(ifp->if_bpf)) {
763 /*
764 * We need to prepend the address family as
765 * a four byte field. Cons up a dummy header
766 * to pacify bpf. This is safe because bpf
767 * will only read from the mbuf (i.e., it won't
768 * try to free it or keep a pointer a to it).
769 */
770 u_int32_t af = AF_INET6;
771 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
772 }
773
774 /*
775 * Put the packet to the network layer input queue according to the
776 * specified address family.
777 * See net/if_gif.c for possible issues with packet processing
778 * reorder due to extra queueing.
779 */
780 ifp->if_ipackets++;
781 ifp->if_ibytes += m->m_pkthdr.len;
782 M_SETFIB(m, ifp->if_fib);
783 netisr_dispatch(NETISR_IPV6, m);
784}
785
786/* ARGSUSED */
787static void
788stf_rtrequest(cmd, rt, info)
789 int cmd;
790 struct rtentry *rt;
791 struct rt_addrinfo *info;
792{
793 RT_LOCK_ASSERT(rt);
794 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
795}
796
797static int
798stf_ioctl(ifp, cmd, data)
799 struct ifnet *ifp;
800 u_long cmd;
801 caddr_t data;
802{
803 struct ifaddr *ifa;
804 struct ifreq *ifr;
805 struct sockaddr_in6 *sin6;
806 struct in_addr addr;
807 int error, mtu;
808
809 error = 0;
810 switch (cmd) {
811 case SIOCSIFADDR:
812 ifa = (struct ifaddr *)data;
813 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
814 error = EAFNOSUPPORT;
815 break;
816 }
817 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
818 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
819 error = EINVAL;
820 break;
821 }
822 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
823 if (isrfc1918addr(&addr)) {
824 error = EINVAL;
825 break;
826 }
827
828 ifa->ifa_rtrequest = stf_rtrequest;
829 ifp->if_flags |= IFF_UP;
830 break;
831
832 case SIOCADDMULTI:
833 case SIOCDELMULTI:
834 ifr = (struct ifreq *)data;
835 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
836 ;
837 else
838 error = EAFNOSUPPORT;
839 break;
840
841 case SIOCGIFMTU:
842 break;
843
844 case SIOCSIFMTU:
845 ifr = (struct ifreq *)data;
846 mtu = ifr->ifr_mtu;
847 /* RFC 4213 3.2 ideal world MTU */
848 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
849 return (EINVAL);
850 ifp->if_mtu = mtu;
851 break;
852
853 default:
854 error = EINVAL;
855 break;
856 }
857
858 return error;
859}