if_gif.c revision 160769
1/*	$FreeBSD: head/sys/net/if_gif.c 160769 2006-07-27 21:01:48Z thompsa $	*/
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#include "opt_mac.h"
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/kernel.h>
40#include <sys/mac.h>
41#include <sys/malloc.h>
42#include <sys/mbuf.h>
43#include <sys/module.h>
44#include <sys/socket.h>
45#include <sys/sockio.h>
46#include <sys/errno.h>
47#include <sys/time.h>
48#include <sys/sysctl.h>
49#include <sys/syslog.h>
50#include <sys/protosw.h>
51#include <sys/conf.h>
52#include <machine/cpu.h>
53
54#include <net/if.h>
55#include <net/if_clone.h>
56#include <net/if_types.h>
57#include <net/netisr.h>
58#include <net/route.h>
59#include <net/bpf.h>
60
61#include <netinet/in.h>
62#include <netinet/in_systm.h>
63#include <netinet/ip.h>
64#ifdef	INET
65#include <netinet/in_var.h>
66#include <netinet/in_gif.h>
67#include <netinet/ip_var.h>
68#endif	/* INET */
69
70#ifdef INET6
71#ifndef INET
72#include <netinet/in.h>
73#endif
74#include <netinet6/in6_var.h>
75#include <netinet/ip6.h>
76#include <netinet6/ip6_var.h>
77#include <netinet6/scope6_var.h>
78#include <netinet6/in6_gif.h>
79#include <netinet6/ip6protosw.h>
80#endif /* INET6 */
81
82#include <netinet/ip_encap.h>
83#include <net/ethernet.h>
84#include <net/if_bridgevar.h>
85#include <net/if_gif.h>
86
87#include <net/net_osdep.h>
88
89#define GIFNAME		"gif"
90
91/*
92 * gif_mtx protects the global gif_softc_list.
93 */
94static struct mtx gif_mtx;
95static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
96static LIST_HEAD(, gif_softc) gif_softc_list;
97
98void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
99void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
100void	(*ng_gif_attach_p)(struct ifnet *ifp);
101void	(*ng_gif_detach_p)(struct ifnet *ifp);
102
103static void	gif_start(struct ifnet *);
104static int	gif_clone_create(struct if_clone *, int, caddr_t);
105static void	gif_clone_destroy(struct ifnet *);
106
107IFC_SIMPLE_DECLARE(gif, 0);
108
109static int gifmodevent(module_t, int, void *);
110
111SYSCTL_DECL(_net_link);
112SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
113    "Generic Tunnel Interface");
114#ifndef MAX_GIF_NEST
115/*
116 * This macro controls the default upper limitation on nesting of gif tunnels.
117 * Since, setting a large value to this macro with a careless configuration
118 * may introduce system crash, we don't allow any nestings by default.
119 * If you need to configure nested gif tunnels, you can define this macro
120 * in your kernel configuration file.  However, if you do so, please be
121 * careful to configure the tunnels so that it won't make a loop.
122 */
123#define MAX_GIF_NEST 1
124#endif
125static int max_gif_nesting = MAX_GIF_NEST;
126SYSCTL_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
127    &max_gif_nesting, 0, "Max nested tunnels");
128
129/*
130 * By default, we disallow creation of multiple tunnels between the same
131 * pair of addresses.  Some applications require this functionality so
132 * we allow control over this check here.
133 */
134#ifdef XBONEHACK
135static int parallel_tunnels = 1;
136#else
137static int parallel_tunnels = 0;
138#endif
139SYSCTL_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
140    &parallel_tunnels, 0, "Allow parallel tunnels?");
141
142static int
143gif_clone_create(ifc, unit, params)
144	struct if_clone *ifc;
145	int unit;
146	caddr_t params;
147{
148	struct gif_softc *sc;
149
150	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
151	GIF2IFP(sc) = if_alloc(IFT_GIF);
152	if (GIF2IFP(sc) == NULL) {
153		free(sc, M_GIF);
154		return (ENOSPC);
155	}
156
157	GIF_LOCK_INIT(sc);
158
159	GIF2IFP(sc)->if_softc = sc;
160	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
161
162	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
163
164	GIF2IFP(sc)->if_addrlen = 0;
165	GIF2IFP(sc)->if_mtu    = GIF_MTU;
166	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
167#if 0
168	/* turn off ingress filter */
169	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
170#endif
171	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
172	GIF2IFP(sc)->if_start  = gif_start;
173	GIF2IFP(sc)->if_output = gif_output;
174	GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
175	if_attach(GIF2IFP(sc));
176	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
177	if (ng_gif_attach_p != NULL)
178		(*ng_gif_attach_p)(GIF2IFP(sc));
179
180	mtx_lock(&gif_mtx);
181	LIST_INSERT_HEAD(&gif_softc_list, sc, gif_list);
182	mtx_unlock(&gif_mtx);
183
184	return (0);
185}
186
187static void
188gif_clone_destroy(ifp)
189	struct ifnet *ifp;
190{
191	int err;
192	struct gif_softc *sc = ifp->if_softc;
193
194	mtx_lock(&gif_mtx);
195	LIST_REMOVE(sc, gif_list);
196	mtx_unlock(&gif_mtx);
197
198	gif_delete_tunnel(ifp);
199#ifdef INET6
200	if (sc->encap_cookie6 != NULL) {
201		err = encap_detach(sc->encap_cookie6);
202		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
203	}
204#endif
205#ifdef INET
206	if (sc->encap_cookie4 != NULL) {
207		err = encap_detach(sc->encap_cookie4);
208		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
209	}
210#endif
211
212	if (ng_gif_detach_p != NULL)
213		(*ng_gif_detach_p)(ifp);
214	bpfdetach(ifp);
215	if_detach(ifp);
216	if_free(ifp);
217
218	GIF_LOCK_DESTROY(sc);
219
220	free(sc, M_GIF);
221}
222
223static int
224gifmodevent(mod, type, data)
225	module_t mod;
226	int type;
227	void *data;
228{
229
230	switch (type) {
231	case MOD_LOAD:
232		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
233		LIST_INIT(&gif_softc_list);
234		if_clone_attach(&gif_cloner);
235
236#ifdef INET6
237		ip6_gif_hlim = GIF_HLIM;
238#endif
239
240		break;
241	case MOD_UNLOAD:
242		if_clone_detach(&gif_cloner);
243		mtx_destroy(&gif_mtx);
244#ifdef INET6
245		ip6_gif_hlim = 0;
246#endif
247		break;
248	default:
249		return EOPNOTSUPP;
250	}
251	return 0;
252}
253
254static moduledata_t gif_mod = {
255	"if_gif",
256	gifmodevent,
257	0
258};
259
260DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
261MODULE_VERSION(if_gif, 1);
262
263int
264gif_encapcheck(m, off, proto, arg)
265	const struct mbuf *m;
266	int off;
267	int proto;
268	void *arg;
269{
270	struct ip ip;
271	struct gif_softc *sc;
272
273	sc = (struct gif_softc *)arg;
274	if (sc == NULL)
275		return 0;
276
277	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
278		return 0;
279
280	/* no physical address */
281	if (!sc->gif_psrc || !sc->gif_pdst)
282		return 0;
283
284	switch (proto) {
285#ifdef INET
286	case IPPROTO_IPV4:
287		break;
288#endif
289#ifdef INET6
290	case IPPROTO_IPV6:
291		break;
292#endif
293	case IPPROTO_ETHERIP:
294		break;
295
296	default:
297		return 0;
298	}
299
300	/* Bail on short packets */
301	if (m->m_pkthdr.len < sizeof(ip))
302		return 0;
303
304	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
305
306	switch (ip.ip_v) {
307#ifdef INET
308	case 4:
309		if (sc->gif_psrc->sa_family != AF_INET ||
310		    sc->gif_pdst->sa_family != AF_INET)
311			return 0;
312		return gif_encapcheck4(m, off, proto, arg);
313#endif
314#ifdef INET6
315	case 6:
316		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
317			return 0;
318		if (sc->gif_psrc->sa_family != AF_INET6 ||
319		    sc->gif_pdst->sa_family != AF_INET6)
320			return 0;
321		return gif_encapcheck6(m, off, proto, arg);
322#endif
323	default:
324		return 0;
325	}
326}
327
328static void
329gif_start(struct ifnet *ifp)
330{
331	struct gif_softc *sc;
332	struct mbuf *m;
333
334	sc = ifp->if_softc;
335
336	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
337	for (;;) {
338		IFQ_DEQUEUE(&ifp->if_snd, m);
339		if (m == 0)
340			break;
341
342		gif_output(ifp, m, sc->gif_pdst, NULL);
343
344	}
345	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
346
347	return;
348}
349
350int
351gif_output(ifp, m, dst, rt)
352	struct ifnet *ifp;
353	struct mbuf *m;
354	struct sockaddr *dst;
355	struct rtentry *rt;	/* added in net2 */
356{
357	struct gif_softc *sc = ifp->if_softc;
358	struct m_tag *mtag;
359	int error = 0;
360	int gif_called;
361	u_int32_t af;
362
363#ifdef MAC
364	error = mac_check_ifnet_transmit(ifp, m);
365	if (error) {
366		m_freem(m);
367		goto end;
368	}
369#endif
370
371	/*
372	 * gif may cause infinite recursion calls when misconfigured.
373	 * We'll prevent this by detecting loops.
374	 *
375	 * High nesting level may cause stack exhaustion.
376	 * We'll prevent this by introducing upper limit.
377	 */
378	gif_called = 1;
379	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
380	while (mtag != NULL) {
381		if (*(struct ifnet **)(mtag + 1) == ifp) {
382			log(LOG_NOTICE,
383			    "gif_output: loop detected on %s\n",
384			    (*(struct ifnet **)(mtag + 1))->if_xname);
385			m_freem(m);
386			error = EIO;	/* is there better errno? */
387			goto end;
388		}
389		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
390		gif_called++;
391	}
392	if (gif_called > max_gif_nesting) {
393		log(LOG_NOTICE,
394		    "gif_output: recursively called too many times(%d)\n",
395		    gif_called);
396		m_freem(m);
397		error = EIO;	/* is there better errno? */
398		goto end;
399	}
400	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
401	    M_NOWAIT);
402	if (mtag == NULL) {
403		m_freem(m);
404		error = ENOMEM;
405		goto end;
406	}
407	*(struct ifnet **)(mtag + 1) = ifp;
408	m_tag_prepend(m, mtag);
409
410	m->m_flags &= ~(M_BCAST|M_MCAST);
411
412	GIF_LOCK(sc);
413
414	if (!(ifp->if_flags & IFF_UP) ||
415	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
416		GIF_UNLOCK(sc);
417		m_freem(m);
418		error = ENETDOWN;
419		goto end;
420	}
421
422	/* BPF writes need to be handled specially. */
423	if (dst->sa_family == AF_UNSPEC) {
424		bcopy(dst->sa_data, &af, sizeof(af));
425		dst->sa_family = af;
426	}
427
428	af = dst->sa_family;
429	BPF_MTAP2(ifp, &af, sizeof(af), m);
430	ifp->if_opackets++;
431	ifp->if_obytes += m->m_pkthdr.len;
432
433	/* override to IPPROTO_ETHERIP for bridged traffic */
434	if (ifp->if_bridge)
435		af = AF_LINK;
436
437	/* inner AF-specific encapsulation */
438
439	/* XXX should we check if our outer source is legal? */
440
441	/* dispatch to output logic based on outer AF */
442	switch (sc->gif_psrc->sa_family) {
443#ifdef INET
444	case AF_INET:
445		error = in_gif_output(ifp, af, m);
446		break;
447#endif
448#ifdef INET6
449	case AF_INET6:
450		error = in6_gif_output(ifp, af, m);
451		break;
452#endif
453	default:
454		m_freem(m);
455		error = ENETDOWN;
456	}
457
458	GIF_UNLOCK(sc);
459  end:
460	if (error)
461		ifp->if_oerrors++;
462	return (error);
463}
464
465void
466gif_input(m, af, ifp)
467	struct mbuf *m;
468	int af;
469	struct ifnet *ifp;
470{
471	int isr, n;
472	struct etherip_header *eip;
473
474	if (ifp == NULL) {
475		/* just in case */
476		m_freem(m);
477		return;
478	}
479
480	m->m_pkthdr.rcvif = ifp;
481
482#ifdef MAC
483	mac_create_mbuf_from_ifnet(ifp, m);
484#endif
485
486	if (bpf_peers_present(ifp->if_bpf)) {
487		u_int32_t af1 = af;
488		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
489	}
490
491	if (ng_gif_input_p != NULL) {
492		(*ng_gif_input_p)(ifp, &m, af);
493		if (m == NULL)
494			return;
495	}
496
497	/*
498	 * Put the packet to the network layer input queue according to the
499	 * specified address family.
500	 * Note: older versions of gif_input directly called network layer
501	 * input functions, e.g. ip6_input, here.  We changed the policy to
502	 * prevent too many recursive calls of such input functions, which
503	 * might cause kernel panic.  But the change may introduce another
504	 * problem; if the input queue is full, packets are discarded.
505	 * The kernel stack overflow really happened, and we believed
506	 * queue-full rarely occurs, so we changed the policy.
507	 */
508	switch (af) {
509#ifdef INET
510	case AF_INET:
511		isr = NETISR_IP;
512		break;
513#endif
514#ifdef INET6
515	case AF_INET6:
516		isr = NETISR_IPV6;
517		break;
518#endif
519	case AF_LINK:
520		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
521		if (n > m->m_len) {
522			m = m_pullup(m, n);
523			if (m == NULL) {
524				ifp->if_ierrors++;
525				return;
526			}
527		}
528
529		eip = mtod(m, struct etherip_header *);
530 		if (eip->eip_ver !=
531		    (ETHERIP_VERSION & ETHERIP_VER_VERS_MASK)) {
532			/* discard unknown versions */
533			m_freem(m);
534			return;
535		}
536		m_adj(m, sizeof(struct etherip_header));
537
538		m->m_flags &= ~(M_BCAST|M_MCAST);
539		m->m_pkthdr.rcvif = ifp;
540
541		if (ifp->if_bridge)
542			BRIDGE_INPUT(ifp, m);
543
544		if (m != NULL)
545			m_freem(m);
546		return;
547
548	default:
549		if (ng_gif_input_orphan_p != NULL)
550			(*ng_gif_input_orphan_p)(ifp, m, af);
551		else
552			m_freem(m);
553		return;
554	}
555
556	ifp->if_ipackets++;
557	ifp->if_ibytes += m->m_pkthdr.len;
558	netisr_dispatch(isr, m);
559}
560
561/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
562int
563gif_ioctl(ifp, cmd, data)
564	struct ifnet *ifp;
565	u_long cmd;
566	caddr_t data;
567{
568	struct gif_softc *sc  = ifp->if_softc;
569	struct ifreq     *ifr = (struct ifreq*)data;
570	int error = 0, size;
571	struct sockaddr *dst, *src;
572#ifdef	SIOCSIFMTU /* xxx */
573	u_long mtu;
574#endif
575
576	switch (cmd) {
577	case SIOCSIFADDR:
578		ifp->if_flags |= IFF_UP;
579		break;
580
581	case SIOCSIFDSTADDR:
582		break;
583
584	case SIOCADDMULTI:
585	case SIOCDELMULTI:
586		break;
587
588#ifdef	SIOCSIFMTU /* xxx */
589	case SIOCGIFMTU:
590		break;
591
592	case SIOCSIFMTU:
593		mtu = ifr->ifr_mtu;
594		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
595			return (EINVAL);
596		ifp->if_mtu = mtu;
597		break;
598#endif /* SIOCSIFMTU */
599
600#ifdef INET
601	case SIOCSIFPHYADDR:
602#endif
603#ifdef INET6
604	case SIOCSIFPHYADDR_IN6:
605#endif /* INET6 */
606	case SIOCSLIFPHYADDR:
607		switch (cmd) {
608#ifdef INET
609		case SIOCSIFPHYADDR:
610			src = (struct sockaddr *)
611				&(((struct in_aliasreq *)data)->ifra_addr);
612			dst = (struct sockaddr *)
613				&(((struct in_aliasreq *)data)->ifra_dstaddr);
614			break;
615#endif
616#ifdef INET6
617		case SIOCSIFPHYADDR_IN6:
618			src = (struct sockaddr *)
619				&(((struct in6_aliasreq *)data)->ifra_addr);
620			dst = (struct sockaddr *)
621				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
622			break;
623#endif
624		case SIOCSLIFPHYADDR:
625			src = (struct sockaddr *)
626				&(((struct if_laddrreq *)data)->addr);
627			dst = (struct sockaddr *)
628				&(((struct if_laddrreq *)data)->dstaddr);
629			break;
630		default:
631			return EINVAL;
632		}
633
634		/* sa_family must be equal */
635		if (src->sa_family != dst->sa_family)
636			return EINVAL;
637
638		/* validate sa_len */
639		switch (src->sa_family) {
640#ifdef INET
641		case AF_INET:
642			if (src->sa_len != sizeof(struct sockaddr_in))
643				return EINVAL;
644			break;
645#endif
646#ifdef INET6
647		case AF_INET6:
648			if (src->sa_len != sizeof(struct sockaddr_in6))
649				return EINVAL;
650			break;
651#endif
652		default:
653			return EAFNOSUPPORT;
654		}
655		switch (dst->sa_family) {
656#ifdef INET
657		case AF_INET:
658			if (dst->sa_len != sizeof(struct sockaddr_in))
659				return EINVAL;
660			break;
661#endif
662#ifdef INET6
663		case AF_INET6:
664			if (dst->sa_len != sizeof(struct sockaddr_in6))
665				return EINVAL;
666			break;
667#endif
668		default:
669			return EAFNOSUPPORT;
670		}
671
672		/* check sa_family looks sane for the cmd */
673		switch (cmd) {
674		case SIOCSIFPHYADDR:
675			if (src->sa_family == AF_INET)
676				break;
677			return EAFNOSUPPORT;
678#ifdef INET6
679		case SIOCSIFPHYADDR_IN6:
680			if (src->sa_family == AF_INET6)
681				break;
682			return EAFNOSUPPORT;
683#endif /* INET6 */
684		case SIOCSLIFPHYADDR:
685			/* checks done in the above */
686			break;
687		}
688
689		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
690		break;
691
692#ifdef SIOCDIFPHYADDR
693	case SIOCDIFPHYADDR:
694		gif_delete_tunnel(GIF2IFP(sc));
695		break;
696#endif
697
698	case SIOCGIFPSRCADDR:
699#ifdef INET6
700	case SIOCGIFPSRCADDR_IN6:
701#endif /* INET6 */
702		if (sc->gif_psrc == NULL) {
703			error = EADDRNOTAVAIL;
704			goto bad;
705		}
706		src = sc->gif_psrc;
707		switch (cmd) {
708#ifdef INET
709		case SIOCGIFPSRCADDR:
710			dst = &ifr->ifr_addr;
711			size = sizeof(ifr->ifr_addr);
712			break;
713#endif /* INET */
714#ifdef INET6
715		case SIOCGIFPSRCADDR_IN6:
716			dst = (struct sockaddr *)
717				&(((struct in6_ifreq *)data)->ifr_addr);
718			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
719			break;
720#endif /* INET6 */
721		default:
722			error = EADDRNOTAVAIL;
723			goto bad;
724		}
725		if (src->sa_len > size)
726			return EINVAL;
727		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
728#ifdef INET6
729		if (dst->sa_family == AF_INET6) {
730			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
731			if (error != 0)
732				return (error);
733		}
734#endif
735		break;
736
737	case SIOCGIFPDSTADDR:
738#ifdef INET6
739	case SIOCGIFPDSTADDR_IN6:
740#endif /* INET6 */
741		if (sc->gif_pdst == NULL) {
742			error = EADDRNOTAVAIL;
743			goto bad;
744		}
745		src = sc->gif_pdst;
746		switch (cmd) {
747#ifdef INET
748		case SIOCGIFPDSTADDR:
749			dst = &ifr->ifr_addr;
750			size = sizeof(ifr->ifr_addr);
751			break;
752#endif /* INET */
753#ifdef INET6
754		case SIOCGIFPDSTADDR_IN6:
755			dst = (struct sockaddr *)
756				&(((struct in6_ifreq *)data)->ifr_addr);
757			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
758			break;
759#endif /* INET6 */
760		default:
761			error = EADDRNOTAVAIL;
762			goto bad;
763		}
764		if (src->sa_len > size)
765			return EINVAL;
766		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
767#ifdef INET6
768		if (dst->sa_family == AF_INET6) {
769			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
770			if (error != 0)
771				return (error);
772		}
773#endif
774		break;
775
776	case SIOCGLIFPHYADDR:
777		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
778			error = EADDRNOTAVAIL;
779			goto bad;
780		}
781
782		/* copy src */
783		src = sc->gif_psrc;
784		dst = (struct sockaddr *)
785			&(((struct if_laddrreq *)data)->addr);
786		size = sizeof(((struct if_laddrreq *)data)->addr);
787		if (src->sa_len > size)
788			return EINVAL;
789		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
790
791		/* copy dst */
792		src = sc->gif_pdst;
793		dst = (struct sockaddr *)
794			&(((struct if_laddrreq *)data)->dstaddr);
795		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
796		if (src->sa_len > size)
797			return EINVAL;
798		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
799		break;
800
801	case SIOCSIFFLAGS:
802		/* if_ioctl() takes care of it */
803		break;
804
805	default:
806		error = EINVAL;
807		break;
808	}
809 bad:
810	return error;
811}
812
813/*
814 * XXXRW: There's a general event-ordering issue here: the code to check
815 * if a given tunnel is already present happens before we perform a
816 * potentially blocking setup of the tunnel.  This code needs to be
817 * re-ordered so that the check and replacement can be atomic using
818 * a mutex.
819 */
820int
821gif_set_tunnel(ifp, src, dst)
822	struct ifnet *ifp;
823	struct sockaddr *src;
824	struct sockaddr *dst;
825{
826	struct gif_softc *sc = ifp->if_softc;
827	struct gif_softc *sc2;
828	struct sockaddr *osrc, *odst, *sa;
829	int error = 0;
830
831	mtx_lock(&gif_mtx);
832	LIST_FOREACH(sc2, &gif_softc_list, gif_list) {
833		if (sc2 == sc)
834			continue;
835		if (!sc2->gif_pdst || !sc2->gif_psrc)
836			continue;
837		if (sc2->gif_pdst->sa_family != dst->sa_family ||
838		    sc2->gif_pdst->sa_len != dst->sa_len ||
839		    sc2->gif_psrc->sa_family != src->sa_family ||
840		    sc2->gif_psrc->sa_len != src->sa_len)
841			continue;
842
843		/*
844		 * Disallow parallel tunnels unless instructed
845		 * otherwise.
846		 */
847		if (!parallel_tunnels &&
848		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
849		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
850			error = EADDRNOTAVAIL;
851			mtx_unlock(&gif_mtx);
852			goto bad;
853		}
854
855		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
856	}
857	mtx_unlock(&gif_mtx);
858
859	/* XXX we can detach from both, but be polite just in case */
860	if (sc->gif_psrc)
861		switch (sc->gif_psrc->sa_family) {
862#ifdef INET
863		case AF_INET:
864			(void)in_gif_detach(sc);
865			break;
866#endif
867#ifdef INET6
868		case AF_INET6:
869			(void)in6_gif_detach(sc);
870			break;
871#endif
872		}
873
874	osrc = sc->gif_psrc;
875	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
876	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
877	sc->gif_psrc = sa;
878
879	odst = sc->gif_pdst;
880	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
881	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
882	sc->gif_pdst = sa;
883
884	switch (sc->gif_psrc->sa_family) {
885#ifdef INET
886	case AF_INET:
887		error = in_gif_attach(sc);
888		break;
889#endif
890#ifdef INET6
891	case AF_INET6:
892		/*
893		 * Check validity of the scope zone ID of the addresses, and
894		 * convert it into the kernel internal form if necessary.
895		 */
896		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
897		if (error != 0)
898			break;
899		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
900		if (error != 0)
901			break;
902		error = in6_gif_attach(sc);
903		break;
904#endif
905	}
906	if (error) {
907		/* rollback */
908		free((caddr_t)sc->gif_psrc, M_IFADDR);
909		free((caddr_t)sc->gif_pdst, M_IFADDR);
910		sc->gif_psrc = osrc;
911		sc->gif_pdst = odst;
912		goto bad;
913	}
914
915	if (osrc)
916		free((caddr_t)osrc, M_IFADDR);
917	if (odst)
918		free((caddr_t)odst, M_IFADDR);
919
920 bad:
921	if (sc->gif_psrc && sc->gif_pdst)
922		ifp->if_drv_flags |= IFF_DRV_RUNNING;
923	else
924		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
925
926	return error;
927}
928
929void
930gif_delete_tunnel(ifp)
931	struct ifnet *ifp;
932{
933	struct gif_softc *sc = ifp->if_softc;
934
935	if (sc->gif_psrc) {
936		free((caddr_t)sc->gif_psrc, M_IFADDR);
937		sc->gif_psrc = NULL;
938	}
939	if (sc->gif_pdst) {
940		free((caddr_t)sc->gif_pdst, M_IFADDR);
941		sc->gif_pdst = NULL;
942	}
943	/* it is safe to detach from both */
944#ifdef INET
945	(void)in_gif_detach(sc);
946#endif
947#ifdef INET6
948	(void)in6_gif_detach(sc);
949#endif
950	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
951}
952