xref: /freebsd/sys/net/if_gif.c (revision 891b8ed4672a213bbe6f3f10522eeadb34d01b76)
1 /*	$FreeBSD$	*/
2 /*	$KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/jail.h>
39 #include <sys/kernel.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/module.h>
43 #include <sys/socket.h>
44 #include <sys/sockio.h>
45 #include <sys/errno.h>
46 #include <sys/time.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/protosw.h>
52 #include <sys/conf.h>
53 #include <machine/cpu.h>
54 
55 #include <net/if.h>
56 #include <net/if_clone.h>
57 #include <net/if_types.h>
58 #include <net/netisr.h>
59 #include <net/route.h>
60 #include <net/bpf.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/ip.h>
66 #ifdef	INET
67 #include <netinet/in_var.h>
68 #include <netinet/in_gif.h>
69 #include <netinet/ip_var.h>
70 #endif	/* INET */
71 
72 #ifdef INET6
73 #ifndef INET
74 #include <netinet/in.h>
75 #endif
76 #include <netinet6/in6_var.h>
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/scope6_var.h>
80 #include <netinet6/in6_gif.h>
81 #include <netinet6/ip6protosw.h>
82 #endif /* INET6 */
83 
84 #include <netinet/ip_encap.h>
85 #include <net/ethernet.h>
86 #include <net/if_bridgevar.h>
87 #include <net/if_gif.h>
88 
89 #include <security/mac/mac_framework.h>
90 
91 #define GIFNAME		"gif"
92 
93 /*
94  * gif_mtx protects the global gif_softc_list.
95  */
96 static struct mtx gif_mtx;
97 static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
98 static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
99 #define	V_gif_softc_list	VNET(gif_softc_list)
100 
101 void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
102 void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
103 void	(*ng_gif_attach_p)(struct ifnet *ifp);
104 void	(*ng_gif_detach_p)(struct ifnet *ifp);
105 
106 static void	gif_start(struct ifnet *);
107 static int	gif_clone_create(struct if_clone *, int, caddr_t);
108 static void	gif_clone_destroy(struct ifnet *);
109 
110 IFC_SIMPLE_DECLARE(gif, 0);
111 
112 static int gifmodevent(module_t, int, void *);
113 
114 SYSCTL_DECL(_net_link);
115 SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
116     "Generic Tunnel Interface");
117 #ifndef MAX_GIF_NEST
118 /*
119  * This macro controls the default upper limitation on nesting of gif tunnels.
120  * Since, setting a large value to this macro with a careless configuration
121  * may introduce system crash, we don't allow any nestings by default.
122  * If you need to configure nested gif tunnels, you can define this macro
123  * in your kernel configuration file.  However, if you do so, please be
124  * careful to configure the tunnels so that it won't make a loop.
125  */
126 #define MAX_GIF_NEST 1
127 #endif
128 static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST;
129 #define	V_max_gif_nesting	VNET(max_gif_nesting)
130 SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
131     &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
132 
133 /*
134  * By default, we disallow creation of multiple tunnels between the same
135  * pair of addresses.  Some applications require this functionality so
136  * we allow control over this check here.
137  */
138 #ifdef XBONEHACK
139 static VNET_DEFINE(int, parallel_tunnels) = 1;
140 #else
141 static VNET_DEFINE(int, parallel_tunnels) = 0;
142 #endif
143 #define	V_parallel_tunnels	VNET(parallel_tunnels)
144 SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
145     &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
146 
147 /* copy from src/sys/net/if_ethersubr.c */
148 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
149 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
150 #ifndef ETHER_IS_BROADCAST
151 #define ETHER_IS_BROADCAST(addr) \
152 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
153 #endif
154 
155 static int
156 gif_clone_create(ifc, unit, params)
157 	struct if_clone *ifc;
158 	int unit;
159 	caddr_t params;
160 {
161 	struct gif_softc *sc;
162 
163 	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
164 	sc->gif_fibnum = curthread->td_proc->p_fibnum;
165 	GIF2IFP(sc) = if_alloc(IFT_GIF);
166 	if (GIF2IFP(sc) == NULL) {
167 		free(sc, M_GIF);
168 		return (ENOSPC);
169 	}
170 
171 	GIF_LOCK_INIT(sc);
172 
173 	GIF2IFP(sc)->if_softc = sc;
174 	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
175 
176 	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
177 	sc->gif_options = GIF_ACCEPT_REVETHIP;
178 
179 	GIF2IFP(sc)->if_addrlen = 0;
180 	GIF2IFP(sc)->if_mtu    = GIF_MTU;
181 	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
182 #if 0
183 	/* turn off ingress filter */
184 	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
185 #endif
186 	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
187 	GIF2IFP(sc)->if_start  = gif_start;
188 	GIF2IFP(sc)->if_output = gif_output;
189 	GIF2IFP(sc)->if_snd.ifq_maxlen = ifqmaxlen;
190 	if_attach(GIF2IFP(sc));
191 	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
192 	if (ng_gif_attach_p != NULL)
193 		(*ng_gif_attach_p)(GIF2IFP(sc));
194 
195 	mtx_lock(&gif_mtx);
196 	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
197 	mtx_unlock(&gif_mtx);
198 
199 	return (0);
200 }
201 
202 static void
203 gif_clone_destroy(ifp)
204 	struct ifnet *ifp;
205 {
206 #if defined(INET) || defined(INET6)
207 	int err;
208 #endif
209 	struct gif_softc *sc = ifp->if_softc;
210 
211 	mtx_lock(&gif_mtx);
212 	LIST_REMOVE(sc, gif_list);
213 	mtx_unlock(&gif_mtx);
214 
215 	gif_delete_tunnel(ifp);
216 #ifdef INET6
217 	if (sc->encap_cookie6 != NULL) {
218 		err = encap_detach(sc->encap_cookie6);
219 		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
220 	}
221 #endif
222 #ifdef INET
223 	if (sc->encap_cookie4 != NULL) {
224 		err = encap_detach(sc->encap_cookie4);
225 		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
226 	}
227 #endif
228 
229 	if (ng_gif_detach_p != NULL)
230 		(*ng_gif_detach_p)(ifp);
231 	bpfdetach(ifp);
232 	if_detach(ifp);
233 	if_free(ifp);
234 
235 	GIF_LOCK_DESTROY(sc);
236 
237 	free(sc, M_GIF);
238 }
239 
240 static void
241 vnet_gif_init(const void *unused __unused)
242 {
243 
244 	LIST_INIT(&V_gif_softc_list);
245 }
246 VNET_SYSINIT(vnet_gif_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_gif_init,
247     NULL);
248 
249 static int
250 gifmodevent(mod, type, data)
251 	module_t mod;
252 	int type;
253 	void *data;
254 {
255 
256 	switch (type) {
257 	case MOD_LOAD:
258 		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
259 		if_clone_attach(&gif_cloner);
260 		break;
261 
262 	case MOD_UNLOAD:
263 		if_clone_detach(&gif_cloner);
264 		mtx_destroy(&gif_mtx);
265 		break;
266 	default:
267 		return EOPNOTSUPP;
268 	}
269 	return 0;
270 }
271 
272 static moduledata_t gif_mod = {
273 	"if_gif",
274 	gifmodevent,
275 	0
276 };
277 
278 DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
279 MODULE_VERSION(if_gif, 1);
280 
281 int
282 gif_encapcheck(m, off, proto, arg)
283 	const struct mbuf *m;
284 	int off;
285 	int proto;
286 	void *arg;
287 {
288 	struct ip ip;
289 	struct gif_softc *sc;
290 
291 	sc = (struct gif_softc *)arg;
292 	if (sc == NULL)
293 		return 0;
294 
295 	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
296 		return 0;
297 
298 	/* no physical address */
299 	if (!sc->gif_psrc || !sc->gif_pdst)
300 		return 0;
301 
302 	switch (proto) {
303 #ifdef INET
304 	case IPPROTO_IPV4:
305 		break;
306 #endif
307 #ifdef INET6
308 	case IPPROTO_IPV6:
309 		break;
310 #endif
311 	case IPPROTO_ETHERIP:
312 		break;
313 
314 	default:
315 		return 0;
316 	}
317 
318 	/* Bail on short packets */
319 	if (m->m_pkthdr.len < sizeof(ip))
320 		return 0;
321 
322 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
323 
324 	switch (ip.ip_v) {
325 #ifdef INET
326 	case 4:
327 		if (sc->gif_psrc->sa_family != AF_INET ||
328 		    sc->gif_pdst->sa_family != AF_INET)
329 			return 0;
330 		return gif_encapcheck4(m, off, proto, arg);
331 #endif
332 #ifdef INET6
333 	case 6:
334 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
335 			return 0;
336 		if (sc->gif_psrc->sa_family != AF_INET6 ||
337 		    sc->gif_pdst->sa_family != AF_INET6)
338 			return 0;
339 		return gif_encapcheck6(m, off, proto, arg);
340 #endif
341 	default:
342 		return 0;
343 	}
344 }
345 
346 static void
347 gif_start(struct ifnet *ifp)
348 {
349 	struct gif_softc *sc;
350 	struct mbuf *m;
351 
352 	sc = ifp->if_softc;
353 
354 	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
355 	for (;;) {
356 		IFQ_DEQUEUE(&ifp->if_snd, m);
357 		if (m == 0)
358 			break;
359 
360 		gif_output(ifp, m, sc->gif_pdst, NULL);
361 
362 	}
363 	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
364 
365 	return;
366 }
367 
368 int
369 gif_output(ifp, m, dst, ro)
370 	struct ifnet *ifp;
371 	struct mbuf *m;
372 	struct sockaddr *dst;
373 	struct route *ro;
374 {
375 	struct gif_softc *sc = ifp->if_softc;
376 	struct m_tag *mtag;
377 	int error = 0;
378 	int gif_called;
379 	u_int32_t af;
380 
381 #ifdef MAC
382 	error = mac_ifnet_check_transmit(ifp, m);
383 	if (error) {
384 		m_freem(m);
385 		goto end;
386 	}
387 #endif
388 
389 	/*
390 	 * gif may cause infinite recursion calls when misconfigured.
391 	 * We'll prevent this by detecting loops.
392 	 *
393 	 * High nesting level may cause stack exhaustion.
394 	 * We'll prevent this by introducing upper limit.
395 	 */
396 	gif_called = 1;
397 	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
398 	while (mtag != NULL) {
399 		if (*(struct ifnet **)(mtag + 1) == ifp) {
400 			log(LOG_NOTICE,
401 			    "gif_output: loop detected on %s\n",
402 			    (*(struct ifnet **)(mtag + 1))->if_xname);
403 			m_freem(m);
404 			error = EIO;	/* is there better errno? */
405 			goto end;
406 		}
407 		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
408 		gif_called++;
409 	}
410 	if (gif_called > V_max_gif_nesting) {
411 		log(LOG_NOTICE,
412 		    "gif_output: recursively called too many times(%d)\n",
413 		    gif_called);
414 		m_freem(m);
415 		error = EIO;	/* is there better errno? */
416 		goto end;
417 	}
418 	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
419 	    M_NOWAIT);
420 	if (mtag == NULL) {
421 		m_freem(m);
422 		error = ENOMEM;
423 		goto end;
424 	}
425 	*(struct ifnet **)(mtag + 1) = ifp;
426 	m_tag_prepend(m, mtag);
427 
428 	m->m_flags &= ~(M_BCAST|M_MCAST);
429 
430 	GIF_LOCK(sc);
431 
432 	if (!(ifp->if_flags & IFF_UP) ||
433 	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
434 		GIF_UNLOCK(sc);
435 		m_freem(m);
436 		error = ENETDOWN;
437 		goto end;
438 	}
439 
440 	/* BPF writes need to be handled specially. */
441 	if (dst->sa_family == AF_UNSPEC) {
442 		bcopy(dst->sa_data, &af, sizeof(af));
443 		dst->sa_family = af;
444 	}
445 
446 	af = dst->sa_family;
447 	BPF_MTAP2(ifp, &af, sizeof(af), m);
448 	ifp->if_opackets++;
449 	ifp->if_obytes += m->m_pkthdr.len;
450 
451 	/* override to IPPROTO_ETHERIP for bridged traffic */
452 	if (ifp->if_bridge)
453 		af = AF_LINK;
454 
455 	M_SETFIB(m, sc->gif_fibnum);
456 	/* inner AF-specific encapsulation */
457 
458 	/* XXX should we check if our outer source is legal? */
459 
460 	/* dispatch to output logic based on outer AF */
461 	switch (sc->gif_psrc->sa_family) {
462 #ifdef INET
463 	case AF_INET:
464 		error = in_gif_output(ifp, af, m);
465 		break;
466 #endif
467 #ifdef INET6
468 	case AF_INET6:
469 		error = in6_gif_output(ifp, af, m);
470 		break;
471 #endif
472 	default:
473 		m_freem(m);
474 		error = ENETDOWN;
475 	}
476 
477 	GIF_UNLOCK(sc);
478   end:
479 	if (error)
480 		ifp->if_oerrors++;
481 	return (error);
482 }
483 
484 void
485 gif_input(m, af, ifp)
486 	struct mbuf *m;
487 	int af;
488 	struct ifnet *ifp;
489 {
490 	int isr, n;
491 	struct gif_softc *sc;
492 	struct etherip_header *eip;
493 	struct ether_header *eh;
494 	struct ifnet *oldifp;
495 
496 	if (ifp == NULL) {
497 		/* just in case */
498 		m_freem(m);
499 		return;
500 	}
501 	sc = ifp->if_softc;
502 	m->m_pkthdr.rcvif = ifp;
503 
504 #ifdef MAC
505 	mac_ifnet_create_mbuf(ifp, m);
506 #endif
507 
508 	if (bpf_peers_present(ifp->if_bpf)) {
509 		u_int32_t af1 = af;
510 		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
511 	}
512 
513 	if (ng_gif_input_p != NULL) {
514 		(*ng_gif_input_p)(ifp, &m, af);
515 		if (m == NULL)
516 			return;
517 	}
518 
519 	/*
520 	 * Put the packet to the network layer input queue according to the
521 	 * specified address family.
522 	 * Note: older versions of gif_input directly called network layer
523 	 * input functions, e.g. ip6_input, here.  We changed the policy to
524 	 * prevent too many recursive calls of such input functions, which
525 	 * might cause kernel panic.  But the change may introduce another
526 	 * problem; if the input queue is full, packets are discarded.
527 	 * The kernel stack overflow really happened, and we believed
528 	 * queue-full rarely occurs, so we changed the policy.
529 	 */
530 	switch (af) {
531 #ifdef INET
532 	case AF_INET:
533 		isr = NETISR_IP;
534 		break;
535 #endif
536 #ifdef INET6
537 	case AF_INET6:
538 		isr = NETISR_IPV6;
539 		break;
540 #endif
541 	case AF_LINK:
542 		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
543 		if (n > m->m_len) {
544 			m = m_pullup(m, n);
545 			if (m == NULL) {
546 				ifp->if_ierrors++;
547 				return;
548 			}
549 		}
550 
551 		eip = mtod(m, struct etherip_header *);
552 		/*
553 		 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
554 		 * accepts an EtherIP packet with revered version field in
555 		 * the header.  This is a knob for backward compatibility
556 		 * with FreeBSD 7.2R or prior.
557 		 */
558 		if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
559 			if (eip->eip_resvl != ETHERIP_VERSION
560 			    && eip->eip_ver != ETHERIP_VERSION) {
561 				/* discard unknown versions */
562 				m_freem(m);
563 				return;
564 			}
565 		} else {
566 			if (eip->eip_ver != ETHERIP_VERSION) {
567 				/* discard unknown versions */
568 				m_freem(m);
569 				return;
570 			}
571 		}
572 		m_adj(m, sizeof(struct etherip_header));
573 
574 		m->m_flags &= ~(M_BCAST|M_MCAST);
575 		m->m_pkthdr.rcvif = ifp;
576 
577 		if (ifp->if_bridge) {
578 			oldifp = ifp;
579 			eh = mtod(m, struct ether_header *);
580 			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
581 				if (ETHER_IS_BROADCAST(eh->ether_dhost))
582 					m->m_flags |= M_BCAST;
583 				else
584 					m->m_flags |= M_MCAST;
585 				ifp->if_imcasts++;
586 			}
587 			BRIDGE_INPUT(ifp, m);
588 
589 			if (m != NULL && ifp != oldifp) {
590 				/*
591 				 * The bridge gave us back itself or one of the
592 				 * members for which the frame is addressed.
593 				 */
594 				ether_demux(ifp, m);
595 				return;
596 			}
597 		}
598 		if (m != NULL)
599 			m_freem(m);
600 		return;
601 
602 	default:
603 		if (ng_gif_input_orphan_p != NULL)
604 			(*ng_gif_input_orphan_p)(ifp, m, af);
605 		else
606 			m_freem(m);
607 		return;
608 	}
609 
610 	ifp->if_ipackets++;
611 	ifp->if_ibytes += m->m_pkthdr.len;
612 	netisr_dispatch(isr, m);
613 }
614 
615 /* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
616 int
617 gif_ioctl(ifp, cmd, data)
618 	struct ifnet *ifp;
619 	u_long cmd;
620 	caddr_t data;
621 {
622 	struct gif_softc *sc  = ifp->if_softc;
623 	struct ifreq     *ifr = (struct ifreq*)data;
624 	int error = 0, size;
625 	u_int	options;
626 	struct sockaddr *dst, *src;
627 #ifdef	SIOCSIFMTU /* xxx */
628 	u_long mtu;
629 #endif
630 
631 	switch (cmd) {
632 	case SIOCSIFADDR:
633 		ifp->if_flags |= IFF_UP;
634 		break;
635 
636 	case SIOCSIFDSTADDR:
637 		break;
638 
639 	case SIOCADDMULTI:
640 	case SIOCDELMULTI:
641 		break;
642 
643 #ifdef	SIOCSIFMTU /* xxx */
644 	case SIOCGIFMTU:
645 		break;
646 
647 	case SIOCSIFMTU:
648 		mtu = ifr->ifr_mtu;
649 		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
650 			return (EINVAL);
651 		ifp->if_mtu = mtu;
652 		break;
653 #endif /* SIOCSIFMTU */
654 
655 #ifdef INET
656 	case SIOCSIFPHYADDR:
657 #endif
658 #ifdef INET6
659 	case SIOCSIFPHYADDR_IN6:
660 #endif /* INET6 */
661 	case SIOCSLIFPHYADDR:
662 		switch (cmd) {
663 #ifdef INET
664 		case SIOCSIFPHYADDR:
665 			src = (struct sockaddr *)
666 				&(((struct in_aliasreq *)data)->ifra_addr);
667 			dst = (struct sockaddr *)
668 				&(((struct in_aliasreq *)data)->ifra_dstaddr);
669 			break;
670 #endif
671 #ifdef INET6
672 		case SIOCSIFPHYADDR_IN6:
673 			src = (struct sockaddr *)
674 				&(((struct in6_aliasreq *)data)->ifra_addr);
675 			dst = (struct sockaddr *)
676 				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
677 			break;
678 #endif
679 		case SIOCSLIFPHYADDR:
680 			src = (struct sockaddr *)
681 				&(((struct if_laddrreq *)data)->addr);
682 			dst = (struct sockaddr *)
683 				&(((struct if_laddrreq *)data)->dstaddr);
684 			break;
685 		default:
686 			return EINVAL;
687 		}
688 
689 		/* sa_family must be equal */
690 		if (src->sa_family != dst->sa_family)
691 			return EINVAL;
692 
693 		/* validate sa_len */
694 		switch (src->sa_family) {
695 #ifdef INET
696 		case AF_INET:
697 			if (src->sa_len != sizeof(struct sockaddr_in))
698 				return EINVAL;
699 			break;
700 #endif
701 #ifdef INET6
702 		case AF_INET6:
703 			if (src->sa_len != sizeof(struct sockaddr_in6))
704 				return EINVAL;
705 			break;
706 #endif
707 		default:
708 			return EAFNOSUPPORT;
709 		}
710 		switch (dst->sa_family) {
711 #ifdef INET
712 		case AF_INET:
713 			if (dst->sa_len != sizeof(struct sockaddr_in))
714 				return EINVAL;
715 			break;
716 #endif
717 #ifdef INET6
718 		case AF_INET6:
719 			if (dst->sa_len != sizeof(struct sockaddr_in6))
720 				return EINVAL;
721 			break;
722 #endif
723 		default:
724 			return EAFNOSUPPORT;
725 		}
726 
727 		/* check sa_family looks sane for the cmd */
728 		switch (cmd) {
729 		case SIOCSIFPHYADDR:
730 			if (src->sa_family == AF_INET)
731 				break;
732 			return EAFNOSUPPORT;
733 #ifdef INET6
734 		case SIOCSIFPHYADDR_IN6:
735 			if (src->sa_family == AF_INET6)
736 				break;
737 			return EAFNOSUPPORT;
738 #endif /* INET6 */
739 		case SIOCSLIFPHYADDR:
740 			/* checks done in the above */
741 			break;
742 		}
743 
744 		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
745 		break;
746 
747 #ifdef SIOCDIFPHYADDR
748 	case SIOCDIFPHYADDR:
749 		gif_delete_tunnel(GIF2IFP(sc));
750 		break;
751 #endif
752 
753 	case SIOCGIFPSRCADDR:
754 #ifdef INET6
755 	case SIOCGIFPSRCADDR_IN6:
756 #endif /* INET6 */
757 		if (sc->gif_psrc == NULL) {
758 			error = EADDRNOTAVAIL;
759 			goto bad;
760 		}
761 		src = sc->gif_psrc;
762 		switch (cmd) {
763 #ifdef INET
764 		case SIOCGIFPSRCADDR:
765 			dst = &ifr->ifr_addr;
766 			size = sizeof(ifr->ifr_addr);
767 			break;
768 #endif /* INET */
769 #ifdef INET6
770 		case SIOCGIFPSRCADDR_IN6:
771 			dst = (struct sockaddr *)
772 				&(((struct in6_ifreq *)data)->ifr_addr);
773 			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
774 			break;
775 #endif /* INET6 */
776 		default:
777 			error = EADDRNOTAVAIL;
778 			goto bad;
779 		}
780 		if (src->sa_len > size)
781 			return EINVAL;
782 		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
783 #ifdef INET6
784 		if (dst->sa_family == AF_INET6) {
785 			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
786 			if (error != 0)
787 				return (error);
788 		}
789 #endif
790 		break;
791 
792 	case SIOCGIFPDSTADDR:
793 #ifdef INET6
794 	case SIOCGIFPDSTADDR_IN6:
795 #endif /* INET6 */
796 		if (sc->gif_pdst == NULL) {
797 			error = EADDRNOTAVAIL;
798 			goto bad;
799 		}
800 		src = sc->gif_pdst;
801 		switch (cmd) {
802 #ifdef INET
803 		case SIOCGIFPDSTADDR:
804 			dst = &ifr->ifr_addr;
805 			size = sizeof(ifr->ifr_addr);
806 			break;
807 #endif /* INET */
808 #ifdef INET6
809 		case SIOCGIFPDSTADDR_IN6:
810 			dst = (struct sockaddr *)
811 				&(((struct in6_ifreq *)data)->ifr_addr);
812 			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
813 			break;
814 #endif /* INET6 */
815 		default:
816 			error = EADDRNOTAVAIL;
817 			goto bad;
818 		}
819 		if (src->sa_len > size)
820 			return EINVAL;
821 		error = prison_if(curthread->td_ucred, src);
822 		if (error != 0)
823 			return (error);
824 		error = prison_if(curthread->td_ucred, dst);
825 		if (error != 0)
826 			return (error);
827 		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
828 #ifdef INET6
829 		if (dst->sa_family == AF_INET6) {
830 			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
831 			if (error != 0)
832 				return (error);
833 		}
834 #endif
835 		break;
836 
837 	case SIOCGLIFPHYADDR:
838 		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
839 			error = EADDRNOTAVAIL;
840 			goto bad;
841 		}
842 
843 		/* copy src */
844 		src = sc->gif_psrc;
845 		dst = (struct sockaddr *)
846 			&(((struct if_laddrreq *)data)->addr);
847 		size = sizeof(((struct if_laddrreq *)data)->addr);
848 		if (src->sa_len > size)
849 			return EINVAL;
850 		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
851 
852 		/* copy dst */
853 		src = sc->gif_pdst;
854 		dst = (struct sockaddr *)
855 			&(((struct if_laddrreq *)data)->dstaddr);
856 		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
857 		if (src->sa_len > size)
858 			return EINVAL;
859 		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
860 		break;
861 
862 	case SIOCSIFFLAGS:
863 		/* if_ioctl() takes care of it */
864 		break;
865 
866 	case GIFGOPTS:
867 		options = sc->gif_options;
868 		error = copyout(&options, ifr->ifr_data,
869 				sizeof(options));
870 		break;
871 
872 	case GIFSOPTS:
873 		if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
874 			break;
875 		error = copyin(ifr->ifr_data, &options, sizeof(options));
876 		if (error)
877 			break;
878 		if (options & ~GIF_OPTMASK)
879 			error = EINVAL;
880 		else
881 			sc->gif_options = options;
882 		break;
883 
884 	default:
885 		error = EINVAL;
886 		break;
887 	}
888  bad:
889 	return error;
890 }
891 
892 /*
893  * XXXRW: There's a general event-ordering issue here: the code to check
894  * if a given tunnel is already present happens before we perform a
895  * potentially blocking setup of the tunnel.  This code needs to be
896  * re-ordered so that the check and replacement can be atomic using
897  * a mutex.
898  */
899 int
900 gif_set_tunnel(ifp, src, dst)
901 	struct ifnet *ifp;
902 	struct sockaddr *src;
903 	struct sockaddr *dst;
904 {
905 	struct gif_softc *sc = ifp->if_softc;
906 	struct gif_softc *sc2;
907 	struct sockaddr *osrc, *odst, *sa;
908 	int error = 0;
909 
910 	mtx_lock(&gif_mtx);
911 	LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
912 		if (sc2 == sc)
913 			continue;
914 		if (!sc2->gif_pdst || !sc2->gif_psrc)
915 			continue;
916 		if (sc2->gif_pdst->sa_family != dst->sa_family ||
917 		    sc2->gif_pdst->sa_len != dst->sa_len ||
918 		    sc2->gif_psrc->sa_family != src->sa_family ||
919 		    sc2->gif_psrc->sa_len != src->sa_len)
920 			continue;
921 
922 		/*
923 		 * Disallow parallel tunnels unless instructed
924 		 * otherwise.
925 		 */
926 		if (!V_parallel_tunnels &&
927 		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
928 		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
929 			error = EADDRNOTAVAIL;
930 			mtx_unlock(&gif_mtx);
931 			goto bad;
932 		}
933 
934 		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
935 	}
936 	mtx_unlock(&gif_mtx);
937 
938 	/* XXX we can detach from both, but be polite just in case */
939 	if (sc->gif_psrc)
940 		switch (sc->gif_psrc->sa_family) {
941 #ifdef INET
942 		case AF_INET:
943 			(void)in_gif_detach(sc);
944 			break;
945 #endif
946 #ifdef INET6
947 		case AF_INET6:
948 			(void)in6_gif_detach(sc);
949 			break;
950 #endif
951 		}
952 
953 	osrc = sc->gif_psrc;
954 	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
955 	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
956 	sc->gif_psrc = sa;
957 
958 	odst = sc->gif_pdst;
959 	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
960 	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
961 	sc->gif_pdst = sa;
962 
963 	switch (sc->gif_psrc->sa_family) {
964 #ifdef INET
965 	case AF_INET:
966 		error = in_gif_attach(sc);
967 		break;
968 #endif
969 #ifdef INET6
970 	case AF_INET6:
971 		/*
972 		 * Check validity of the scope zone ID of the addresses, and
973 		 * convert it into the kernel internal form if necessary.
974 		 */
975 		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
976 		if (error != 0)
977 			break;
978 		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
979 		if (error != 0)
980 			break;
981 		error = in6_gif_attach(sc);
982 		break;
983 #endif
984 	}
985 	if (error) {
986 		/* rollback */
987 		free((caddr_t)sc->gif_psrc, M_IFADDR);
988 		free((caddr_t)sc->gif_pdst, M_IFADDR);
989 		sc->gif_psrc = osrc;
990 		sc->gif_pdst = odst;
991 		goto bad;
992 	}
993 
994 	if (osrc)
995 		free((caddr_t)osrc, M_IFADDR);
996 	if (odst)
997 		free((caddr_t)odst, M_IFADDR);
998 
999  bad:
1000 	if (sc->gif_psrc && sc->gif_pdst)
1001 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1002 	else
1003 		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1004 
1005 	return error;
1006 }
1007 
1008 void
1009 gif_delete_tunnel(ifp)
1010 	struct ifnet *ifp;
1011 {
1012 	struct gif_softc *sc = ifp->if_softc;
1013 
1014 	if (sc->gif_psrc) {
1015 		free((caddr_t)sc->gif_psrc, M_IFADDR);
1016 		sc->gif_psrc = NULL;
1017 	}
1018 	if (sc->gif_pdst) {
1019 		free((caddr_t)sc->gif_pdst, M_IFADDR);
1020 		sc->gif_pdst = NULL;
1021 	}
1022 	/* it is safe to detach from both */
1023 #ifdef INET
1024 	(void)in_gif_detach(sc);
1025 #endif
1026 #ifdef INET6
1027 	(void)in6_gif_detach(sc);
1028 #endif
1029 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1030 }
1031