xref: /freebsd/sys/net/if_gif.c (revision 1c05a6ea6b849ff95e539c31adea887c644a6a01)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/module.h>
46 #include <sys/rmlock.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/sx.h>
50 #include <sys/errno.h>
51 #include <sys/time.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/protosw.h>
57 #include <sys/conf.h>
58 #include <machine/cpu.h>
59 
60 #include <net/if.h>
61 #include <net/if_var.h>
62 #include <net/if_clone.h>
63 #include <net/if_types.h>
64 #include <net/netisr.h>
65 #include <net/route.h>
66 #include <net/bpf.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_ecn.h>
73 #ifdef	INET
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #endif	/* INET */
77 
78 #ifdef INET6
79 #ifndef INET
80 #include <netinet/in.h>
81 #endif
82 #include <netinet6/in6_var.h>
83 #include <netinet/ip6.h>
84 #include <netinet6/ip6_ecn.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet6/scope6_var.h>
87 #include <netinet6/ip6protosw.h>
88 #endif /* INET6 */
89 
90 #include <netinet/ip_encap.h>
91 #include <net/ethernet.h>
92 #include <net/if_bridgevar.h>
93 #include <net/if_gif.h>
94 
95 #include <security/mac/mac_framework.h>
96 
97 static const char gifname[] = "gif";
98 
99 /*
100  * gif_mtx protects a per-vnet gif_softc_list.
101  */
102 static VNET_DEFINE(struct mtx, gif_mtx);
103 #define	V_gif_mtx		VNET(gif_mtx)
104 static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
105 static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
106 #define	V_gif_softc_list	VNET(gif_softc_list)
107 static struct sx gif_ioctl_sx;
108 SX_SYSINIT(gif_ioctl_sx, &gif_ioctl_sx, "gif_ioctl");
109 
110 #define	GIF_LIST_LOCK_INIT(x)		mtx_init(&V_gif_mtx, "gif_mtx", \
111 					    NULL, MTX_DEF)
112 #define	GIF_LIST_LOCK_DESTROY(x)	mtx_destroy(&V_gif_mtx)
113 #define	GIF_LIST_LOCK(x)		mtx_lock(&V_gif_mtx)
114 #define	GIF_LIST_UNLOCK(x)		mtx_unlock(&V_gif_mtx)
115 
116 void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
117 void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
118 void	(*ng_gif_attach_p)(struct ifnet *ifp);
119 void	(*ng_gif_detach_p)(struct ifnet *ifp);
120 
121 static int	gif_check_nesting(struct ifnet *, struct mbuf *);
122 static int	gif_set_tunnel(struct ifnet *, struct sockaddr *,
123     struct sockaddr *);
124 static void	gif_delete_tunnel(struct ifnet *);
125 static int	gif_ioctl(struct ifnet *, u_long, caddr_t);
126 static int	gif_transmit(struct ifnet *, struct mbuf *);
127 static void	gif_qflush(struct ifnet *);
128 static int	gif_clone_create(struct if_clone *, int, caddr_t);
129 static void	gif_clone_destroy(struct ifnet *);
130 static VNET_DEFINE(struct if_clone *, gif_cloner);
131 #define	V_gif_cloner	VNET(gif_cloner)
132 
133 static int gifmodevent(module_t, int, void *);
134 
135 SYSCTL_DECL(_net_link);
136 static SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
137     "Generic Tunnel Interface");
138 #ifndef MAX_GIF_NEST
139 /*
140  * This macro controls the default upper limitation on nesting of gif tunnels.
141  * Since, setting a large value to this macro with a careless configuration
142  * may introduce system crash, we don't allow any nestings by default.
143  * If you need to configure nested gif tunnels, you can define this macro
144  * in your kernel configuration file.  However, if you do so, please be
145  * careful to configure the tunnels so that it won't make a loop.
146  */
147 #define MAX_GIF_NEST 1
148 #endif
149 static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST;
150 #define	V_max_gif_nesting	VNET(max_gif_nesting)
151 SYSCTL_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_VNET | CTLFLAG_RW,
152     &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
153 
154 /*
155  * By default, we disallow creation of multiple tunnels between the same
156  * pair of addresses.  Some applications require this functionality so
157  * we allow control over this check here.
158  */
159 #ifdef XBONEHACK
160 static VNET_DEFINE(int, parallel_tunnels) = 1;
161 #else
162 static VNET_DEFINE(int, parallel_tunnels) = 0;
163 #endif
164 #define	V_parallel_tunnels	VNET(parallel_tunnels)
165 SYSCTL_INT(_net_link_gif, OID_AUTO, parallel_tunnels,
166     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(parallel_tunnels), 0,
167     "Allow parallel tunnels?");
168 
169 static int
170 gif_clone_create(struct if_clone *ifc, int unit, caddr_t params)
171 {
172 	struct gif_softc *sc;
173 
174 	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
175 	sc->gif_fibnum = curthread->td_proc->p_fibnum;
176 	GIF2IFP(sc) = if_alloc(IFT_GIF);
177 	GIF_LOCK_INIT(sc);
178 	GIF2IFP(sc)->if_softc = sc;
179 	if_initname(GIF2IFP(sc), gifname, unit);
180 
181 	GIF2IFP(sc)->if_addrlen = 0;
182 	GIF2IFP(sc)->if_mtu    = GIF_MTU;
183 	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
184 #if 0
185 	/* turn off ingress filter */
186 	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
187 #endif
188 	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
189 	GIF2IFP(sc)->if_transmit  = gif_transmit;
190 	GIF2IFP(sc)->if_qflush  = gif_qflush;
191 	GIF2IFP(sc)->if_output = gif_output;
192 	GIF2IFP(sc)->if_capabilities |= IFCAP_LINKSTATE;
193 	GIF2IFP(sc)->if_capenable |= IFCAP_LINKSTATE;
194 	if_attach(GIF2IFP(sc));
195 	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
196 	if (ng_gif_attach_p != NULL)
197 		(*ng_gif_attach_p)(GIF2IFP(sc));
198 
199 	GIF_LIST_LOCK();
200 	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
201 	GIF_LIST_UNLOCK();
202 	return (0);
203 }
204 
205 static void
206 gif_clone_destroy(struct ifnet *ifp)
207 {
208 	struct gif_softc *sc;
209 
210 	sx_xlock(&gif_ioctl_sx);
211 	sc = ifp->if_softc;
212 	gif_delete_tunnel(ifp);
213 	GIF_LIST_LOCK();
214 	LIST_REMOVE(sc, gif_list);
215 	GIF_LIST_UNLOCK();
216 	if (ng_gif_detach_p != NULL)
217 		(*ng_gif_detach_p)(ifp);
218 	bpfdetach(ifp);
219 	if_detach(ifp);
220 	ifp->if_softc = NULL;
221 	sx_xunlock(&gif_ioctl_sx);
222 
223 	if_free(ifp);
224 	GIF_LOCK_DESTROY(sc);
225 	free(sc, M_GIF);
226 }
227 
228 static void
229 vnet_gif_init(const void *unused __unused)
230 {
231 
232 	LIST_INIT(&V_gif_softc_list);
233 	GIF_LIST_LOCK_INIT();
234 	V_gif_cloner = if_clone_simple(gifname, gif_clone_create,
235 	    gif_clone_destroy, 0);
236 }
237 VNET_SYSINIT(vnet_gif_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
238     vnet_gif_init, NULL);
239 
240 static void
241 vnet_gif_uninit(const void *unused __unused)
242 {
243 
244 	if_clone_detach(V_gif_cloner);
245 	GIF_LIST_LOCK_DESTROY();
246 }
247 VNET_SYSUNINIT(vnet_gif_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
248     vnet_gif_uninit, NULL);
249 
250 static int
251 gifmodevent(module_t mod, int type, void *data)
252 {
253 
254 	switch (type) {
255 	case MOD_LOAD:
256 	case MOD_UNLOAD:
257 		break;
258 	default:
259 		return (EOPNOTSUPP);
260 	}
261 	return (0);
262 }
263 
264 static moduledata_t gif_mod = {
265 	"if_gif",
266 	gifmodevent,
267 	0
268 };
269 
270 DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
271 MODULE_VERSION(if_gif, 1);
272 
273 int
274 gif_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
275 {
276 	GIF_RLOCK_TRACKER;
277 	const struct ip *ip;
278 	struct gif_softc *sc;
279 	int ret;
280 
281 	sc = (struct gif_softc *)arg;
282 	if (sc == NULL || (GIF2IFP(sc)->if_flags & IFF_UP) == 0)
283 		return (0);
284 
285 	ret = 0;
286 	GIF_RLOCK(sc);
287 
288 	/* no physical address */
289 	if (sc->gif_family == 0)
290 		goto done;
291 
292 	switch (proto) {
293 #ifdef INET
294 	case IPPROTO_IPV4:
295 #endif
296 #ifdef INET6
297 	case IPPROTO_IPV6:
298 #endif
299 	case IPPROTO_ETHERIP:
300 		break;
301 	default:
302 		goto done;
303 	}
304 
305 	/* Bail on short packets */
306 	M_ASSERTPKTHDR(m);
307 	if (m->m_pkthdr.len < sizeof(struct ip))
308 		goto done;
309 
310 	ip = mtod(m, const struct ip *);
311 	switch (ip->ip_v) {
312 #ifdef INET
313 	case 4:
314 		if (sc->gif_family != AF_INET)
315 			goto done;
316 		ret = in_gif_encapcheck(m, off, proto, arg);
317 		break;
318 #endif
319 #ifdef INET6
320 	case 6:
321 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
322 			goto done;
323 		if (sc->gif_family != AF_INET6)
324 			goto done;
325 		ret = in6_gif_encapcheck(m, off, proto, arg);
326 		break;
327 #endif
328 	}
329 done:
330 	GIF_RUNLOCK(sc);
331 	return (ret);
332 }
333 
334 static int
335 gif_transmit(struct ifnet *ifp, struct mbuf *m)
336 {
337 	struct gif_softc *sc;
338 	struct etherip_header *eth;
339 #ifdef INET
340 	struct ip *ip;
341 #endif
342 #ifdef INET6
343 	struct ip6_hdr *ip6;
344 	uint32_t t;
345 #endif
346 	uint32_t af;
347 	uint8_t proto, ecn;
348 	int error;
349 
350 #ifdef MAC
351 	error = mac_ifnet_check_transmit(ifp, m);
352 	if (error) {
353 		m_freem(m);
354 		goto err;
355 	}
356 #endif
357 	error = ENETDOWN;
358 	sc = ifp->if_softc;
359 	if ((ifp->if_flags & IFF_MONITOR) != 0 ||
360 	    (ifp->if_flags & IFF_UP) == 0 ||
361 	    sc->gif_family == 0 ||
362 	    (error = gif_check_nesting(ifp, m)) != 0) {
363 		m_freem(m);
364 		goto err;
365 	}
366 	/* Now pull back the af that we stashed in the csum_data. */
367 	if (ifp->if_bridge)
368 		af = AF_LINK;
369 	else
370 		af = m->m_pkthdr.csum_data;
371 	m->m_flags &= ~(M_BCAST|M_MCAST);
372 	M_SETFIB(m, sc->gif_fibnum);
373 	BPF_MTAP2(ifp, &af, sizeof(af), m);
374 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
375 	if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
376 	/* inner AF-specific encapsulation */
377 	ecn = 0;
378 	switch (af) {
379 #ifdef INET
380 	case AF_INET:
381 		proto = IPPROTO_IPV4;
382 		if (m->m_len < sizeof(struct ip))
383 			m = m_pullup(m, sizeof(struct ip));
384 		if (m == NULL) {
385 			error = ENOBUFS;
386 			goto err;
387 		}
388 		ip = mtod(m, struct ip *);
389 		ip_ecn_ingress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
390 		    ECN_NOCARE, &ecn, &ip->ip_tos);
391 		break;
392 #endif
393 #ifdef INET6
394 	case AF_INET6:
395 		proto = IPPROTO_IPV6;
396 		if (m->m_len < sizeof(struct ip6_hdr))
397 			m = m_pullup(m, sizeof(struct ip6_hdr));
398 		if (m == NULL) {
399 			error = ENOBUFS;
400 			goto err;
401 		}
402 		t = 0;
403 		ip6 = mtod(m, struct ip6_hdr *);
404 		ip6_ecn_ingress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
405 		    ECN_NOCARE, &t, &ip6->ip6_flow);
406 		ecn = (ntohl(t) >> 20) & 0xff;
407 		break;
408 #endif
409 	case AF_LINK:
410 		proto = IPPROTO_ETHERIP;
411 		M_PREPEND(m, sizeof(struct etherip_header), M_NOWAIT);
412 		if (m == NULL) {
413 			error = ENOBUFS;
414 			goto err;
415 		}
416 		eth = mtod(m, struct etherip_header *);
417 		eth->eip_resvh = 0;
418 		eth->eip_ver = ETHERIP_VERSION;
419 		eth->eip_resvl = 0;
420 		break;
421 	default:
422 		error = EAFNOSUPPORT;
423 		m_freem(m);
424 		goto err;
425 	}
426 	/* XXX should we check if our outer source is legal? */
427 	/* dispatch to output logic based on outer AF */
428 	switch (sc->gif_family) {
429 #ifdef INET
430 	case AF_INET:
431 		error = in_gif_output(ifp, m, proto, ecn);
432 		break;
433 #endif
434 #ifdef INET6
435 	case AF_INET6:
436 		error = in6_gif_output(ifp, m, proto, ecn);
437 		break;
438 #endif
439 	default:
440 		m_freem(m);
441 	}
442 err:
443 	if (error)
444 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
445 	return (error);
446 }
447 
448 static void
449 gif_qflush(struct ifnet *ifp __unused)
450 {
451 
452 }
453 
454 #define	MTAG_GIF	1080679712
455 static int
456 gif_check_nesting(struct ifnet *ifp, struct mbuf *m)
457 {
458 	struct m_tag *mtag;
459 	int count;
460 
461 	/*
462 	 * gif may cause infinite recursion calls when misconfigured.
463 	 * We'll prevent this by detecting loops.
464 	 *
465 	 * High nesting level may cause stack exhaustion.
466 	 * We'll prevent this by introducing upper limit.
467 	 */
468 	count = 1;
469 	mtag = NULL;
470 	while ((mtag = m_tag_locate(m, MTAG_GIF, 0, mtag)) != NULL) {
471 		if (*(struct ifnet **)(mtag + 1) == ifp) {
472 			log(LOG_NOTICE, "%s: loop detected\n", if_name(ifp));
473 			return (EIO);
474 		}
475 		count++;
476 	}
477 	if (count > V_max_gif_nesting) {
478 		log(LOG_NOTICE,
479 		    "%s: if_output recursively called too many times(%d)\n",
480 		    if_name(ifp), count);
481 		return (EIO);
482 	}
483 	mtag = m_tag_alloc(MTAG_GIF, 0, sizeof(struct ifnet *), M_NOWAIT);
484 	if (mtag == NULL)
485 		return (ENOMEM);
486 	*(struct ifnet **)(mtag + 1) = ifp;
487 	m_tag_prepend(m, mtag);
488 	return (0);
489 }
490 
491 int
492 gif_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
493 	struct route *ro)
494 {
495 	uint32_t af;
496 
497 	if (dst->sa_family == AF_UNSPEC)
498 		bcopy(dst->sa_data, &af, sizeof(af));
499 	else
500 		af = dst->sa_family;
501 	/*
502 	 * Now save the af in the inbound pkt csum data, this is a cheat since
503 	 * we are using the inbound csum_data field to carry the af over to
504 	 * the gif_transmit() routine, avoiding using yet another mtag.
505 	 */
506 	m->m_pkthdr.csum_data = af;
507 	return (ifp->if_transmit(ifp, m));
508 }
509 
510 void
511 gif_input(struct mbuf *m, struct ifnet *ifp, int proto, uint8_t ecn)
512 {
513 	struct etherip_header *eip;
514 #ifdef INET
515 	struct ip *ip;
516 #endif
517 #ifdef INET6
518 	struct ip6_hdr *ip6;
519 	uint32_t t;
520 #endif
521 	struct gif_softc *sc;
522 	struct ether_header *eh;
523 	struct ifnet *oldifp;
524 	int isr, n, af;
525 
526 	if (ifp == NULL) {
527 		/* just in case */
528 		m_freem(m);
529 		return;
530 	}
531 	sc = ifp->if_softc;
532 	m->m_pkthdr.rcvif = ifp;
533 	m_clrprotoflags(m);
534 	switch (proto) {
535 #ifdef INET
536 	case IPPROTO_IPV4:
537 		af = AF_INET;
538 		if (m->m_len < sizeof(struct ip))
539 			m = m_pullup(m, sizeof(struct ip));
540 		if (m == NULL)
541 			goto drop;
542 		ip = mtod(m, struct ip *);
543 		if (ip_ecn_egress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
544 		    ECN_NOCARE, &ecn, &ip->ip_tos) == 0) {
545 			m_freem(m);
546 			goto drop;
547 		}
548 		break;
549 #endif
550 #ifdef INET6
551 	case IPPROTO_IPV6:
552 		af = AF_INET6;
553 		if (m->m_len < sizeof(struct ip6_hdr))
554 			m = m_pullup(m, sizeof(struct ip6_hdr));
555 		if (m == NULL)
556 			goto drop;
557 		t = htonl((uint32_t)ecn << 20);
558 		ip6 = mtod(m, struct ip6_hdr *);
559 		if (ip6_ecn_egress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
560 		    ECN_NOCARE, &t, &ip6->ip6_flow) == 0) {
561 			m_freem(m);
562 			goto drop;
563 		}
564 		break;
565 #endif
566 	case IPPROTO_ETHERIP:
567 		af = AF_LINK;
568 		break;
569 	default:
570 		m_freem(m);
571 		goto drop;
572 	}
573 
574 #ifdef MAC
575 	mac_ifnet_create_mbuf(ifp, m);
576 #endif
577 
578 	if (bpf_peers_present(ifp->if_bpf)) {
579 		uint32_t af1 = af;
580 		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
581 	}
582 
583 	if ((ifp->if_flags & IFF_MONITOR) != 0) {
584 		if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
585 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
586 		m_freem(m);
587 		return;
588 	}
589 
590 	if (ng_gif_input_p != NULL) {
591 		(*ng_gif_input_p)(ifp, &m, af);
592 		if (m == NULL)
593 			goto drop;
594 	}
595 
596 	/*
597 	 * Put the packet to the network layer input queue according to the
598 	 * specified address family.
599 	 * Note: older versions of gif_input directly called network layer
600 	 * input functions, e.g. ip6_input, here.  We changed the policy to
601 	 * prevent too many recursive calls of such input functions, which
602 	 * might cause kernel panic.  But the change may introduce another
603 	 * problem; if the input queue is full, packets are discarded.
604 	 * The kernel stack overflow really happened, and we believed
605 	 * queue-full rarely occurs, so we changed the policy.
606 	 */
607 	switch (af) {
608 #ifdef INET
609 	case AF_INET:
610 		isr = NETISR_IP;
611 		break;
612 #endif
613 #ifdef INET6
614 	case AF_INET6:
615 		isr = NETISR_IPV6;
616 		break;
617 #endif
618 	case AF_LINK:
619 		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
620 		if (n > m->m_len)
621 			m = m_pullup(m, n);
622 		if (m == NULL)
623 			goto drop;
624 		eip = mtod(m, struct etherip_header *);
625 		if (eip->eip_ver != ETHERIP_VERSION) {
626 			/* discard unknown versions */
627 			m_freem(m);
628 			goto drop;
629 		}
630 		m_adj(m, sizeof(struct etherip_header));
631 
632 		m->m_flags &= ~(M_BCAST|M_MCAST);
633 		m->m_pkthdr.rcvif = ifp;
634 
635 		if (ifp->if_bridge) {
636 			oldifp = ifp;
637 			eh = mtod(m, struct ether_header *);
638 			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
639 				if (ETHER_IS_BROADCAST(eh->ether_dhost))
640 					m->m_flags |= M_BCAST;
641 				else
642 					m->m_flags |= M_MCAST;
643 				if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
644 			}
645 			BRIDGE_INPUT(ifp, m);
646 
647 			if (m != NULL && ifp != oldifp) {
648 				/*
649 				 * The bridge gave us back itself or one of the
650 				 * members for which the frame is addressed.
651 				 */
652 				ether_demux(ifp, m);
653 				return;
654 			}
655 		}
656 		if (m != NULL)
657 			m_freem(m);
658 		return;
659 
660 	default:
661 		if (ng_gif_input_orphan_p != NULL)
662 			(*ng_gif_input_orphan_p)(ifp, m, af);
663 		else
664 			m_freem(m);
665 		return;
666 	}
667 
668 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
669 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
670 	M_SETFIB(m, ifp->if_fib);
671 	netisr_dispatch(isr, m);
672 	return;
673 drop:
674 	if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
675 }
676 
677 /* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
678 int
679 gif_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
680 {
681 	GIF_RLOCK_TRACKER;
682 	struct ifreq *ifr = (struct ifreq*)data;
683 	struct sockaddr *dst, *src;
684 	struct gif_softc *sc;
685 #ifdef INET
686 	struct sockaddr_in *sin = NULL;
687 #endif
688 #ifdef INET6
689 	struct sockaddr_in6 *sin6 = NULL;
690 #endif
691 	u_int options;
692 	int error;
693 
694 	switch (cmd) {
695 	case SIOCSIFADDR:
696 		ifp->if_flags |= IFF_UP;
697 	case SIOCADDMULTI:
698 	case SIOCDELMULTI:
699 	case SIOCGIFMTU:
700 	case SIOCSIFFLAGS:
701 		return (0);
702 	case SIOCSIFMTU:
703 		if (ifr->ifr_mtu < GIF_MTU_MIN ||
704 		    ifr->ifr_mtu > GIF_MTU_MAX)
705 			return (EINVAL);
706 		else
707 			ifp->if_mtu = ifr->ifr_mtu;
708 		return (0);
709 	}
710 	sx_xlock(&gif_ioctl_sx);
711 	sc = ifp->if_softc;
712 	if (sc == NULL) {
713 		error = ENXIO;
714 		goto bad;
715 	}
716 	error = 0;
717 	switch (cmd) {
718 	case SIOCSIFPHYADDR:
719 #ifdef INET6
720 	case SIOCSIFPHYADDR_IN6:
721 #endif
722 		error = EINVAL;
723 		switch (cmd) {
724 #ifdef INET
725 		case SIOCSIFPHYADDR:
726 			src = (struct sockaddr *)
727 				&(((struct in_aliasreq *)data)->ifra_addr);
728 			dst = (struct sockaddr *)
729 				&(((struct in_aliasreq *)data)->ifra_dstaddr);
730 			break;
731 #endif
732 #ifdef INET6
733 		case SIOCSIFPHYADDR_IN6:
734 			src = (struct sockaddr *)
735 				&(((struct in6_aliasreq *)data)->ifra_addr);
736 			dst = (struct sockaddr *)
737 				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
738 			break;
739 #endif
740 		default:
741 			goto bad;
742 		}
743 		/* sa_family must be equal */
744 		if (src->sa_family != dst->sa_family ||
745 		    src->sa_len != dst->sa_len)
746 			goto bad;
747 
748 		/* validate sa_len */
749 		/* check sa_family looks sane for the cmd */
750 		switch (src->sa_family) {
751 #ifdef INET
752 		case AF_INET:
753 			if (src->sa_len != sizeof(struct sockaddr_in))
754 				goto bad;
755 			if (cmd != SIOCSIFPHYADDR) {
756 				error = EAFNOSUPPORT;
757 				goto bad;
758 			}
759 			if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
760 			    satosin(dst)->sin_addr.s_addr == INADDR_ANY) {
761 				error = EADDRNOTAVAIL;
762 				goto bad;
763 			}
764 			break;
765 #endif
766 #ifdef INET6
767 		case AF_INET6:
768 			if (src->sa_len != sizeof(struct sockaddr_in6))
769 				goto bad;
770 			if (cmd != SIOCSIFPHYADDR_IN6) {
771 				error = EAFNOSUPPORT;
772 				goto bad;
773 			}
774 			error = EADDRNOTAVAIL;
775 			if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(src)->sin6_addr)
776 			    ||
777 			    IN6_IS_ADDR_UNSPECIFIED(&satosin6(dst)->sin6_addr))
778 				goto bad;
779 			/*
780 			 * Check validity of the scope zone ID of the
781 			 * addresses, and convert it into the kernel
782 			 * internal form if necessary.
783 			 */
784 			error = sa6_embedscope(satosin6(src), 0);
785 			if (error != 0)
786 				goto bad;
787 			error = sa6_embedscope(satosin6(dst), 0);
788 			if (error != 0)
789 				goto bad;
790 			break;
791 #endif
792 		default:
793 			error = EAFNOSUPPORT;
794 			goto bad;
795 		}
796 		error = gif_set_tunnel(ifp, src, dst);
797 		break;
798 	case SIOCDIFPHYADDR:
799 		gif_delete_tunnel(ifp);
800 		break;
801 	case SIOCGIFPSRCADDR:
802 	case SIOCGIFPDSTADDR:
803 #ifdef INET6
804 	case SIOCGIFPSRCADDR_IN6:
805 	case SIOCGIFPDSTADDR_IN6:
806 #endif
807 		if (sc->gif_family == 0) {
808 			error = EADDRNOTAVAIL;
809 			break;
810 		}
811 		GIF_RLOCK(sc);
812 		switch (cmd) {
813 #ifdef INET
814 		case SIOCGIFPSRCADDR:
815 		case SIOCGIFPDSTADDR:
816 			if (sc->gif_family != AF_INET) {
817 				error = EADDRNOTAVAIL;
818 				break;
819 			}
820 			sin = (struct sockaddr_in *)&ifr->ifr_addr;
821 			memset(sin, 0, sizeof(*sin));
822 			sin->sin_family = AF_INET;
823 			sin->sin_len = sizeof(*sin);
824 			break;
825 #endif
826 #ifdef INET6
827 		case SIOCGIFPSRCADDR_IN6:
828 		case SIOCGIFPDSTADDR_IN6:
829 			if (sc->gif_family != AF_INET6) {
830 				error = EADDRNOTAVAIL;
831 				break;
832 			}
833 			sin6 = (struct sockaddr_in6 *)
834 				&(((struct in6_ifreq *)data)->ifr_addr);
835 			memset(sin6, 0, sizeof(*sin6));
836 			sin6->sin6_family = AF_INET6;
837 			sin6->sin6_len = sizeof(*sin6);
838 			break;
839 #endif
840 		default:
841 			error = EAFNOSUPPORT;
842 		}
843 		if (error == 0) {
844 			switch (cmd) {
845 #ifdef INET
846 			case SIOCGIFPSRCADDR:
847 				sin->sin_addr = sc->gif_iphdr->ip_src;
848 				break;
849 			case SIOCGIFPDSTADDR:
850 				sin->sin_addr = sc->gif_iphdr->ip_dst;
851 				break;
852 #endif
853 #ifdef INET6
854 			case SIOCGIFPSRCADDR_IN6:
855 				sin6->sin6_addr = sc->gif_ip6hdr->ip6_src;
856 				break;
857 			case SIOCGIFPDSTADDR_IN6:
858 				sin6->sin6_addr = sc->gif_ip6hdr->ip6_dst;
859 				break;
860 #endif
861 			}
862 		}
863 		GIF_RUNLOCK(sc);
864 		if (error != 0)
865 			break;
866 		switch (cmd) {
867 #ifdef INET
868 		case SIOCGIFPSRCADDR:
869 		case SIOCGIFPDSTADDR:
870 			error = prison_if(curthread->td_ucred,
871 			    (struct sockaddr *)sin);
872 			if (error != 0)
873 				memset(sin, 0, sizeof(*sin));
874 			break;
875 #endif
876 #ifdef INET6
877 		case SIOCGIFPSRCADDR_IN6:
878 		case SIOCGIFPDSTADDR_IN6:
879 			error = prison_if(curthread->td_ucred,
880 			    (struct sockaddr *)sin6);
881 			if (error == 0)
882 				error = sa6_recoverscope(sin6);
883 			if (error != 0)
884 				memset(sin6, 0, sizeof(*sin6));
885 #endif
886 		}
887 		break;
888 	case SIOCGTUNFIB:
889 		ifr->ifr_fib = sc->gif_fibnum;
890 		break;
891 	case SIOCSTUNFIB:
892 		if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
893 			break;
894 		if (ifr->ifr_fib >= rt_numfibs)
895 			error = EINVAL;
896 		else
897 			sc->gif_fibnum = ifr->ifr_fib;
898 		break;
899 	case GIFGOPTS:
900 		options = sc->gif_options;
901 		error = copyout(&options, ifr->ifr_data, sizeof(options));
902 		break;
903 	case GIFSOPTS:
904 		if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
905 			break;
906 		error = copyin(ifr->ifr_data, &options, sizeof(options));
907 		if (error)
908 			break;
909 		if (options & ~GIF_OPTMASK)
910 			error = EINVAL;
911 		else
912 			sc->gif_options = options;
913 		break;
914 	default:
915 		error = EINVAL;
916 		break;
917 	}
918 bad:
919 	sx_xunlock(&gif_ioctl_sx);
920 	return (error);
921 }
922 
923 static void
924 gif_detach(struct gif_softc *sc)
925 {
926 
927 	sx_assert(&gif_ioctl_sx, SA_XLOCKED);
928 	if (sc->gif_ecookie != NULL)
929 		encap_detach(sc->gif_ecookie);
930 	sc->gif_ecookie = NULL;
931 }
932 
933 static int
934 gif_attach(struct gif_softc *sc, int af)
935 {
936 
937 	sx_assert(&gif_ioctl_sx, SA_XLOCKED);
938 	switch (af) {
939 #ifdef INET
940 	case AF_INET:
941 		return (in_gif_attach(sc));
942 #endif
943 #ifdef INET6
944 	case AF_INET6:
945 		return (in6_gif_attach(sc));
946 #endif
947 	}
948 	return (EAFNOSUPPORT);
949 }
950 
951 static int
952 gif_set_tunnel(struct ifnet *ifp, struct sockaddr *src, struct sockaddr *dst)
953 {
954 	struct gif_softc *sc = ifp->if_softc;
955 	struct gif_softc *tsc;
956 #ifdef INET
957 	struct ip *ip;
958 #endif
959 #ifdef INET6
960 	struct ip6_hdr *ip6;
961 #endif
962 	void *hdr;
963 	int error = 0;
964 
965 	if (sc == NULL)
966 		return (ENXIO);
967 	/* Disallow parallel tunnels unless instructed otherwise. */
968 	if (V_parallel_tunnels == 0) {
969 		GIF_LIST_LOCK();
970 		LIST_FOREACH(tsc, &V_gif_softc_list, gif_list) {
971 			if (tsc == sc || tsc->gif_family != src->sa_family)
972 				continue;
973 #ifdef INET
974 			if (tsc->gif_family == AF_INET &&
975 			    tsc->gif_iphdr->ip_src.s_addr ==
976 			    satosin(src)->sin_addr.s_addr &&
977 			    tsc->gif_iphdr->ip_dst.s_addr ==
978 			    satosin(dst)->sin_addr.s_addr) {
979 				error = EADDRNOTAVAIL;
980 				GIF_LIST_UNLOCK();
981 				goto bad;
982 			}
983 #endif
984 #ifdef INET6
985 			if (tsc->gif_family == AF_INET6 &&
986 			    IN6_ARE_ADDR_EQUAL(&tsc->gif_ip6hdr->ip6_src,
987 			    &satosin6(src)->sin6_addr) &&
988 			    IN6_ARE_ADDR_EQUAL(&tsc->gif_ip6hdr->ip6_dst,
989 			    &satosin6(dst)->sin6_addr)) {
990 				error = EADDRNOTAVAIL;
991 				GIF_LIST_UNLOCK();
992 				goto bad;
993 			}
994 #endif
995 		}
996 		GIF_LIST_UNLOCK();
997 	}
998 	switch (src->sa_family) {
999 #ifdef INET
1000 	case AF_INET:
1001 		hdr = ip = malloc(sizeof(struct ip), M_GIF,
1002 		    M_WAITOK | M_ZERO);
1003 		ip->ip_src.s_addr = satosin(src)->sin_addr.s_addr;
1004 		ip->ip_dst.s_addr = satosin(dst)->sin_addr.s_addr;
1005 		break;
1006 #endif
1007 #ifdef INET6
1008 	case AF_INET6:
1009 		hdr = ip6 = malloc(sizeof(struct ip6_hdr), M_GIF,
1010 		    M_WAITOK | M_ZERO);
1011 		ip6->ip6_src = satosin6(src)->sin6_addr;
1012 		ip6->ip6_dst = satosin6(dst)->sin6_addr;
1013 		ip6->ip6_vfc = IPV6_VERSION;
1014 		break;
1015 #endif
1016 	default:
1017 		return (EAFNOSUPPORT);
1018 	}
1019 
1020 	if (sc->gif_family != src->sa_family)
1021 		gif_detach(sc);
1022 	if (sc->gif_family == 0 ||
1023 	    sc->gif_family != src->sa_family)
1024 		error = gif_attach(sc, src->sa_family);
1025 
1026 	GIF_WLOCK(sc);
1027 	if (sc->gif_family != 0)
1028 		free(sc->gif_hdr, M_GIF);
1029 	sc->gif_family = src->sa_family;
1030 	sc->gif_hdr = hdr;
1031 	GIF_WUNLOCK(sc);
1032 #if defined(INET) || defined(INET6)
1033 bad:
1034 #endif
1035 	if (error == 0 && sc->gif_family != 0) {
1036 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1037 		if_link_state_change(ifp, LINK_STATE_UP);
1038 	} else {
1039 		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1040 		if_link_state_change(ifp, LINK_STATE_DOWN);
1041 	}
1042 	return (error);
1043 }
1044 
1045 static void
1046 gif_delete_tunnel(struct ifnet *ifp)
1047 {
1048 	struct gif_softc *sc = ifp->if_softc;
1049 	int family;
1050 
1051 	if (sc == NULL)
1052 		return;
1053 
1054 	GIF_WLOCK(sc);
1055 	family = sc->gif_family;
1056 	sc->gif_family = 0;
1057 	GIF_WUNLOCK(sc);
1058 	if (family != 0) {
1059 		gif_detach(sc);
1060 		free(sc->gif_hdr, M_GIF);
1061 	}
1062 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1063 	if_link_state_change(ifp, LINK_STATE_DOWN);
1064 }
1065