if_gif.c revision 176879 
1/*	$FreeBSD: head/sys/net/if_gif.c 176879 2008-03-06 19:02:37Z 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/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/protosw.h>
50#include <sys/conf.h>
51#include <machine/cpu.h>
52
53#include <net/if.h>
54#include <net/if_clone.h>
55#include <net/if_types.h>
56#include <net/netisr.h>
57#include <net/route.h>
58#include <net/bpf.h>
59
60#include <netinet/in.h>
61#include <netinet/in_systm.h>
62#include <netinet/ip.h>
63#ifdef	INET
64#include <netinet/in_var.h>
65#include <netinet/in_gif.h>
66#include <netinet/ip_var.h>
67#endif	/* INET */
68
69#ifdef INET6
70#ifndef INET
71#include <netinet/in.h>
72#endif
73#include <netinet6/in6_var.h>
74#include <netinet/ip6.h>
75#include <netinet6/ip6_var.h>
76#include <netinet6/scope6_var.h>
77#include <netinet6/in6_gif.h>
78#include <netinet6/ip6protosw.h>
79#endif /* INET6 */
80
81#include <netinet/ip_encap.h>
82#include <net/ethernet.h>
83#include <net/if_bridgevar.h>
84#include <net/if_gif.h>
85
86#include <security/mac/mac_framework.h>
87
88#define GIFNAME		"gif"
89
90/*
91 * gif_mtx protects the global gif_softc_list.
92 */
93static struct mtx gif_mtx;
94static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
95static LIST_HEAD(, gif_softc) gif_softc_list;
96
97void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
98void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
99void	(*ng_gif_attach_p)(struct ifnet *ifp);
100void	(*ng_gif_detach_p)(struct ifnet *ifp);
101
102static void	gif_start(struct ifnet *);
103static int	gif_clone_create(struct if_clone *, int, caddr_t);
104static void	gif_clone_destroy(struct ifnet *);
105
106IFC_SIMPLE_DECLARE(gif, 0);
107
108static int gifmodevent(module_t, int, void *);
109
110SYSCTL_DECL(_net_link);
111SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
112    "Generic Tunnel Interface");
113#ifndef MAX_GIF_NEST
114/*
115 * This macro controls the default upper limitation on nesting of gif tunnels.
116 * Since, setting a large value to this macro with a careless configuration
117 * may introduce system crash, we don't allow any nestings by default.
118 * If you need to configure nested gif tunnels, you can define this macro
119 * in your kernel configuration file.  However, if you do so, please be
120 * careful to configure the tunnels so that it won't make a loop.
121 */
122#define MAX_GIF_NEST 1
123#endif
124static int max_gif_nesting = MAX_GIF_NEST;
125SYSCTL_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
126    &max_gif_nesting, 0, "Max nested tunnels");
127
128/*
129 * By default, we disallow creation of multiple tunnels between the same
130 * pair of addresses.  Some applications require this functionality so
131 * we allow control over this check here.
132 */
133#ifdef XBONEHACK
134static int parallel_tunnels = 1;
135#else
136static int parallel_tunnels = 0;
137#endif
138SYSCTL_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
139    &parallel_tunnels, 0, "Allow parallel tunnels?");
140
141/* copy from src/sys/net/if_ethersubr.c */
142static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
143			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
144#ifndef ETHER_IS_BROADCAST
145#define ETHER_IS_BROADCAST(addr) \
146	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
147#endif
148
149static int
150gif_clone_create(ifc, unit, params)
151	struct if_clone *ifc;
152	int unit;
153	caddr_t params;
154{
155	struct gif_softc *sc;
156
157	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
158	GIF2IFP(sc) = if_alloc(IFT_GIF);
159	if (GIF2IFP(sc) == NULL) {
160		free(sc, M_GIF);
161		return (ENOSPC);
162	}
163
164	GIF_LOCK_INIT(sc);
165
166	GIF2IFP(sc)->if_softc = sc;
167	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
168
169	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
170
171	GIF2IFP(sc)->if_addrlen = 0;
172	GIF2IFP(sc)->if_mtu    = GIF_MTU;
173	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
174#if 0
175	/* turn off ingress filter */
176	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
177#endif
178	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
179	GIF2IFP(sc)->if_start  = gif_start;
180	GIF2IFP(sc)->if_output = gif_output;
181	GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
182	if_attach(GIF2IFP(sc));
183	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
184	if (ng_gif_attach_p != NULL)
185		(*ng_gif_attach_p)(GIF2IFP(sc));
186
187	mtx_lock(&gif_mtx);
188	LIST_INSERT_HEAD(&gif_softc_list, sc, gif_list);
189	mtx_unlock(&gif_mtx);
190
191	return (0);
192}
193
194static void
195gif_clone_destroy(ifp)
196	struct ifnet *ifp;
197{
198	int err;
199	struct gif_softc *sc = ifp->if_softc;
200
201	mtx_lock(&gif_mtx);
202	LIST_REMOVE(sc, gif_list);
203	mtx_unlock(&gif_mtx);
204
205	gif_delete_tunnel(ifp);
206#ifdef INET6
207	if (sc->encap_cookie6 != NULL) {
208		err = encap_detach(sc->encap_cookie6);
209		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
210	}
211#endif
212#ifdef INET
213	if (sc->encap_cookie4 != NULL) {
214		err = encap_detach(sc->encap_cookie4);
215		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
216	}
217#endif
218
219	if (ng_gif_detach_p != NULL)
220		(*ng_gif_detach_p)(ifp);
221	bpfdetach(ifp);
222	if_detach(ifp);
223	if_free(ifp);
224
225	GIF_LOCK_DESTROY(sc);
226
227	free(sc, M_GIF);
228}
229
230static int
231gifmodevent(mod, type, data)
232	module_t mod;
233	int type;
234	void *data;
235{
236
237	switch (type) {
238	case MOD_LOAD:
239		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
240		LIST_INIT(&gif_softc_list);
241		if_clone_attach(&gif_cloner);
242
243#ifdef INET6
244		ip6_gif_hlim = GIF_HLIM;
245#endif
246
247		break;
248	case MOD_UNLOAD:
249		if_clone_detach(&gif_cloner);
250		mtx_destroy(&gif_mtx);
251#ifdef INET6
252		ip6_gif_hlim = 0;
253#endif
254		break;
255	default:
256		return EOPNOTSUPP;
257	}
258	return 0;
259}
260
261static moduledata_t gif_mod = {
262	"if_gif",
263	gifmodevent,
264	0
265};
266
267DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
268MODULE_VERSION(if_gif, 1);
269
270int
271gif_encapcheck(m, off, proto, arg)
272	const struct mbuf *m;
273	int off;
274	int proto;
275	void *arg;
276{
277	struct ip ip;
278	struct gif_softc *sc;
279
280	sc = (struct gif_softc *)arg;
281	if (sc == NULL)
282		return 0;
283
284	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
285		return 0;
286
287	/* no physical address */
288	if (!sc->gif_psrc || !sc->gif_pdst)
289		return 0;
290
291	switch (proto) {
292#ifdef INET
293	case IPPROTO_IPV4:
294		break;
295#endif
296#ifdef INET6
297	case IPPROTO_IPV6:
298		break;
299#endif
300	case IPPROTO_ETHERIP:
301		break;
302
303	default:
304		return 0;
305	}
306
307	/* Bail on short packets */
308	if (m->m_pkthdr.len < sizeof(ip))
309		return 0;
310
311	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
312
313	switch (ip.ip_v) {
314#ifdef INET
315	case 4:
316		if (sc->gif_psrc->sa_family != AF_INET ||
317		    sc->gif_pdst->sa_family != AF_INET)
318			return 0;
319		return gif_encapcheck4(m, off, proto, arg);
320#endif
321#ifdef INET6
322	case 6:
323		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
324			return 0;
325		if (sc->gif_psrc->sa_family != AF_INET6 ||
326		    sc->gif_pdst->sa_family != AF_INET6)
327			return 0;
328		return gif_encapcheck6(m, off, proto, arg);
329#endif
330	default:
331		return 0;
332	}
333}
334
335static void
336gif_start(struct ifnet *ifp)
337{
338	struct gif_softc *sc;
339	struct mbuf *m;
340
341	sc = ifp->if_softc;
342
343	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
344	for (;;) {
345		IFQ_DEQUEUE(&ifp->if_snd, m);
346		if (m == 0)
347			break;
348
349		gif_output(ifp, m, sc->gif_pdst, NULL);
350
351	}
352	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
353
354	return;
355}
356
357int
358gif_output(ifp, m, dst, rt)
359	struct ifnet *ifp;
360	struct mbuf *m;
361	struct sockaddr *dst;
362	struct rtentry *rt;	/* added in net2 */
363{
364	struct gif_softc *sc = ifp->if_softc;
365	struct m_tag *mtag;
366	int error = 0;
367	int gif_called;
368	u_int32_t af;
369
370#ifdef MAC
371	error = mac_ifnet_check_transmit(ifp, m);
372	if (error) {
373		m_freem(m);
374		goto end;
375	}
376#endif
377
378	/*
379	 * gif may cause infinite recursion calls when misconfigured.
380	 * We'll prevent this by detecting loops.
381	 *
382	 * High nesting level may cause stack exhaustion.
383	 * We'll prevent this by introducing upper limit.
384	 */
385	gif_called = 1;
386	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
387	while (mtag != NULL) {
388		if (*(struct ifnet **)(mtag + 1) == ifp) {
389			log(LOG_NOTICE,
390			    "gif_output: loop detected on %s\n",
391			    (*(struct ifnet **)(mtag + 1))->if_xname);
392			m_freem(m);
393			error = EIO;	/* is there better errno? */
394			goto end;
395		}
396		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
397		gif_called++;
398	}
399	if (gif_called > max_gif_nesting) {
400		log(LOG_NOTICE,
401		    "gif_output: recursively called too many times(%d)\n",
402		    gif_called);
403		m_freem(m);
404		error = EIO;	/* is there better errno? */
405		goto end;
406	}
407	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
408	    M_NOWAIT);
409	if (mtag == NULL) {
410		m_freem(m);
411		error = ENOMEM;
412		goto end;
413	}
414	*(struct ifnet **)(mtag + 1) = ifp;
415	m_tag_prepend(m, mtag);
416
417	m->m_flags &= ~(M_BCAST|M_MCAST);
418
419	GIF_LOCK(sc);
420
421	if (!(ifp->if_flags & IFF_UP) ||
422	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
423		GIF_UNLOCK(sc);
424		m_freem(m);
425		error = ENETDOWN;
426		goto end;
427	}
428
429	/* BPF writes need to be handled specially. */
430	if (dst->sa_family == AF_UNSPEC) {
431		bcopy(dst->sa_data, &af, sizeof(af));
432		dst->sa_family = af;
433	}
434
435	af = dst->sa_family;
436	BPF_MTAP2(ifp, &af, sizeof(af), m);
437	ifp->if_opackets++;
438	ifp->if_obytes += m->m_pkthdr.len;
439
440	/* override to IPPROTO_ETHERIP for bridged traffic */
441	if (ifp->if_bridge)
442		af = AF_LINK;
443
444	/* inner AF-specific encapsulation */
445
446	/* XXX should we check if our outer source is legal? */
447
448	/* dispatch to output logic based on outer AF */
449	switch (sc->gif_psrc->sa_family) {
450#ifdef INET
451	case AF_INET:
452		error = in_gif_output(ifp, af, m);
453		break;
454#endif
455#ifdef INET6
456	case AF_INET6:
457		error = in6_gif_output(ifp, af, m);
458		break;
459#endif
460	default:
461		m_freem(m);
462		error = ENETDOWN;
463	}
464
465	GIF_UNLOCK(sc);
466  end:
467	if (error)
468		ifp->if_oerrors++;
469	return (error);
470}
471
472void
473gif_input(m, af, ifp)
474	struct mbuf *m;
475	int af;
476	struct ifnet *ifp;
477{
478	int isr, n;
479	struct etherip_header *eip;
480	struct ether_header *eh;
481	struct ifnet *oldifp;
482
483	if (ifp == NULL) {
484		/* just in case */
485		m_freem(m);
486		return;
487	}
488
489	m->m_pkthdr.rcvif = ifp;
490
491#ifdef MAC
492	mac_ifnet_create_mbuf(ifp, m);
493#endif
494
495	if (bpf_peers_present(ifp->if_bpf)) {
496		u_int32_t af1 = af;
497		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
498	}
499
500	if (ng_gif_input_p != NULL) {
501		(*ng_gif_input_p)(ifp, &m, af);
502		if (m == NULL)
503			return;
504	}
505
506	/*
507	 * Put the packet to the network layer input queue according to the
508	 * specified address family.
509	 * Note: older versions of gif_input directly called network layer
510	 * input functions, e.g. ip6_input, here.  We changed the policy to
511	 * prevent too many recursive calls of such input functions, which
512	 * might cause kernel panic.  But the change may introduce another
513	 * problem; if the input queue is full, packets are discarded.
514	 * The kernel stack overflow really happened, and we believed
515	 * queue-full rarely occurs, so we changed the policy.
516	 */
517	switch (af) {
518#ifdef INET
519	case AF_INET:
520		isr = NETISR_IP;
521		break;
522#endif
523#ifdef INET6
524	case AF_INET6:
525		isr = NETISR_IPV6;
526		break;
527#endif
528	case AF_LINK:
529		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
530		if (n > m->m_len) {
531			m = m_pullup(m, n);
532			if (m == NULL) {
533				ifp->if_ierrors++;
534				return;
535			}
536		}
537
538		eip = mtod(m, struct etherip_header *);
539 		if (eip->eip_ver !=
540		    (ETHERIP_VERSION & ETHERIP_VER_VERS_MASK)) {
541			/* discard unknown versions */
542			m_freem(m);
543			return;
544		}
545		m_adj(m, sizeof(struct etherip_header));
546
547		m->m_flags &= ~(M_BCAST|M_MCAST);
548		m->m_pkthdr.rcvif = ifp;
549
550		if (ifp->if_bridge) {
551			oldifp = ifp;
552			eh = mtod(m, struct ether_header *);
553			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
554				if (ETHER_IS_BROADCAST(eh->ether_dhost))
555					m->m_flags |= M_BCAST;
556				else
557					m->m_flags |= M_MCAST;
558				ifp->if_imcasts++;
559			}
560			BRIDGE_INPUT(ifp, m);
561
562			if (m != NULL && ifp != oldifp) {
563				/*
564				 * The bridge gave us back itself or one of the
565				 * members for which the frame is addressed.
566				 */
567				ether_demux(ifp, m);
568				return;
569			}
570		}
571		if (m != NULL)
572			m_freem(m);
573		return;
574
575	default:
576		if (ng_gif_input_orphan_p != NULL)
577			(*ng_gif_input_orphan_p)(ifp, m, af);
578		else
579			m_freem(m);
580		return;
581	}
582
583	ifp->if_ipackets++;
584	ifp->if_ibytes += m->m_pkthdr.len;
585	netisr_dispatch(isr, m);
586}
587
588/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
589int
590gif_ioctl(ifp, cmd, data)
591	struct ifnet *ifp;
592	u_long cmd;
593	caddr_t data;
594{
595	struct gif_softc *sc  = ifp->if_softc;
596	struct ifreq     *ifr = (struct ifreq*)data;
597	int error = 0, size;
598	struct sockaddr *dst, *src;
599#ifdef	SIOCSIFMTU /* xxx */
600	u_long mtu;
601#endif
602
603	switch (cmd) {
604	case SIOCSIFADDR:
605		ifp->if_flags |= IFF_UP;
606		break;
607
608	case SIOCSIFDSTADDR:
609		break;
610
611	case SIOCADDMULTI:
612	case SIOCDELMULTI:
613		break;
614
615#ifdef	SIOCSIFMTU /* xxx */
616	case SIOCGIFMTU:
617		break;
618
619	case SIOCSIFMTU:
620		mtu = ifr->ifr_mtu;
621		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
622			return (EINVAL);
623		ifp->if_mtu = mtu;
624		break;
625#endif /* SIOCSIFMTU */
626
627#ifdef INET
628	case SIOCSIFPHYADDR:
629#endif
630#ifdef INET6
631	case SIOCSIFPHYADDR_IN6:
632#endif /* INET6 */
633	case SIOCSLIFPHYADDR:
634		switch (cmd) {
635#ifdef INET
636		case SIOCSIFPHYADDR:
637			src = (struct sockaddr *)
638				&(((struct in_aliasreq *)data)->ifra_addr);
639			dst = (struct sockaddr *)
640				&(((struct in_aliasreq *)data)->ifra_dstaddr);
641			break;
642#endif
643#ifdef INET6
644		case SIOCSIFPHYADDR_IN6:
645			src = (struct sockaddr *)
646				&(((struct in6_aliasreq *)data)->ifra_addr);
647			dst = (struct sockaddr *)
648				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
649			break;
650#endif
651		case SIOCSLIFPHYADDR:
652			src = (struct sockaddr *)
653				&(((struct if_laddrreq *)data)->addr);
654			dst = (struct sockaddr *)
655				&(((struct if_laddrreq *)data)->dstaddr);
656			break;
657		default:
658			return EINVAL;
659		}
660
661		/* sa_family must be equal */
662		if (src->sa_family != dst->sa_family)
663			return EINVAL;
664
665		/* validate sa_len */
666		switch (src->sa_family) {
667#ifdef INET
668		case AF_INET:
669			if (src->sa_len != sizeof(struct sockaddr_in))
670				return EINVAL;
671			break;
672#endif
673#ifdef INET6
674		case AF_INET6:
675			if (src->sa_len != sizeof(struct sockaddr_in6))
676				return EINVAL;
677			break;
678#endif
679		default:
680			return EAFNOSUPPORT;
681		}
682		switch (dst->sa_family) {
683#ifdef INET
684		case AF_INET:
685			if (dst->sa_len != sizeof(struct sockaddr_in))
686				return EINVAL;
687			break;
688#endif
689#ifdef INET6
690		case AF_INET6:
691			if (dst->sa_len != sizeof(struct sockaddr_in6))
692				return EINVAL;
693			break;
694#endif
695		default:
696			return EAFNOSUPPORT;
697		}
698
699		/* check sa_family looks sane for the cmd */
700		switch (cmd) {
701		case SIOCSIFPHYADDR:
702			if (src->sa_family == AF_INET)
703				break;
704			return EAFNOSUPPORT;
705#ifdef INET6
706		case SIOCSIFPHYADDR_IN6:
707			if (src->sa_family == AF_INET6)
708				break;
709			return EAFNOSUPPORT;
710#endif /* INET6 */
711		case SIOCSLIFPHYADDR:
712			/* checks done in the above */
713			break;
714		}
715
716		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
717		break;
718
719#ifdef SIOCDIFPHYADDR
720	case SIOCDIFPHYADDR:
721		gif_delete_tunnel(GIF2IFP(sc));
722		break;
723#endif
724
725	case SIOCGIFPSRCADDR:
726#ifdef INET6
727	case SIOCGIFPSRCADDR_IN6:
728#endif /* INET6 */
729		if (sc->gif_psrc == NULL) {
730			error = EADDRNOTAVAIL;
731			goto bad;
732		}
733		src = sc->gif_psrc;
734		switch (cmd) {
735#ifdef INET
736		case SIOCGIFPSRCADDR:
737			dst = &ifr->ifr_addr;
738			size = sizeof(ifr->ifr_addr);
739			break;
740#endif /* INET */
741#ifdef INET6
742		case SIOCGIFPSRCADDR_IN6:
743			dst = (struct sockaddr *)
744				&(((struct in6_ifreq *)data)->ifr_addr);
745			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
746			break;
747#endif /* INET6 */
748		default:
749			error = EADDRNOTAVAIL;
750			goto bad;
751		}
752		if (src->sa_len > size)
753			return EINVAL;
754		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
755#ifdef INET6
756		if (dst->sa_family == AF_INET6) {
757			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
758			if (error != 0)
759				return (error);
760		}
761#endif
762		break;
763
764	case SIOCGIFPDSTADDR:
765#ifdef INET6
766	case SIOCGIFPDSTADDR_IN6:
767#endif /* INET6 */
768		if (sc->gif_pdst == NULL) {
769			error = EADDRNOTAVAIL;
770			goto bad;
771		}
772		src = sc->gif_pdst;
773		switch (cmd) {
774#ifdef INET
775		case SIOCGIFPDSTADDR:
776			dst = &ifr->ifr_addr;
777			size = sizeof(ifr->ifr_addr);
778			break;
779#endif /* INET */
780#ifdef INET6
781		case SIOCGIFPDSTADDR_IN6:
782			dst = (struct sockaddr *)
783				&(((struct in6_ifreq *)data)->ifr_addr);
784			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
785			break;
786#endif /* INET6 */
787		default:
788			error = EADDRNOTAVAIL;
789			goto bad;
790		}
791		if (src->sa_len > size)
792			return EINVAL;
793		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
794#ifdef INET6
795		if (dst->sa_family == AF_INET6) {
796			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
797			if (error != 0)
798				return (error);
799		}
800#endif
801		break;
802
803	case SIOCGLIFPHYADDR:
804		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
805			error = EADDRNOTAVAIL;
806			goto bad;
807		}
808
809		/* copy src */
810		src = sc->gif_psrc;
811		dst = (struct sockaddr *)
812			&(((struct if_laddrreq *)data)->addr);
813		size = sizeof(((struct if_laddrreq *)data)->addr);
814		if (src->sa_len > size)
815			return EINVAL;
816		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
817
818		/* copy dst */
819		src = sc->gif_pdst;
820		dst = (struct sockaddr *)
821			&(((struct if_laddrreq *)data)->dstaddr);
822		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
823		if (src->sa_len > size)
824			return EINVAL;
825		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
826		break;
827
828	case SIOCSIFFLAGS:
829		/* if_ioctl() takes care of it */
830		break;
831
832	default:
833		error = EINVAL;
834		break;
835	}
836 bad:
837	return error;
838}
839
840/*
841 * XXXRW: There's a general event-ordering issue here: the code to check
842 * if a given tunnel is already present happens before we perform a
843 * potentially blocking setup of the tunnel.  This code needs to be
844 * re-ordered so that the check and replacement can be atomic using
845 * a mutex.
846 */
847int
848gif_set_tunnel(ifp, src, dst)
849	struct ifnet *ifp;
850	struct sockaddr *src;
851	struct sockaddr *dst;
852{
853	struct gif_softc *sc = ifp->if_softc;
854	struct gif_softc *sc2;
855	struct sockaddr *osrc, *odst, *sa;
856	int error = 0;
857
858	mtx_lock(&gif_mtx);
859	LIST_FOREACH(sc2, &gif_softc_list, gif_list) {
860		if (sc2 == sc)
861			continue;
862		if (!sc2->gif_pdst || !sc2->gif_psrc)
863			continue;
864		if (sc2->gif_pdst->sa_family != dst->sa_family ||
865		    sc2->gif_pdst->sa_len != dst->sa_len ||
866		    sc2->gif_psrc->sa_family != src->sa_family ||
867		    sc2->gif_psrc->sa_len != src->sa_len)
868			continue;
869
870		/*
871		 * Disallow parallel tunnels unless instructed
872		 * otherwise.
873		 */
874		if (!parallel_tunnels &&
875		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
876		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
877			error = EADDRNOTAVAIL;
878			mtx_unlock(&gif_mtx);
879			goto bad;
880		}
881
882		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
883	}
884	mtx_unlock(&gif_mtx);
885
886	/* XXX we can detach from both, but be polite just in case */
887	if (sc->gif_psrc)
888		switch (sc->gif_psrc->sa_family) {
889#ifdef INET
890		case AF_INET:
891			(void)in_gif_detach(sc);
892			break;
893#endif
894#ifdef INET6
895		case AF_INET6:
896			(void)in6_gif_detach(sc);
897			break;
898#endif
899		}
900
901	osrc = sc->gif_psrc;
902	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
903	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
904	sc->gif_psrc = sa;
905
906	odst = sc->gif_pdst;
907	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
908	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
909	sc->gif_pdst = sa;
910
911	switch (sc->gif_psrc->sa_family) {
912#ifdef INET
913	case AF_INET:
914		error = in_gif_attach(sc);
915		break;
916#endif
917#ifdef INET6
918	case AF_INET6:
919		/*
920		 * Check validity of the scope zone ID of the addresses, and
921		 * convert it into the kernel internal form if necessary.
922		 */
923		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
924		if (error != 0)
925			break;
926		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
927		if (error != 0)
928			break;
929		error = in6_gif_attach(sc);
930		break;
931#endif
932	}
933	if (error) {
934		/* rollback */
935		free((caddr_t)sc->gif_psrc, M_IFADDR);
936		free((caddr_t)sc->gif_pdst, M_IFADDR);
937		sc->gif_psrc = osrc;
938		sc->gif_pdst = odst;
939		goto bad;
940	}
941
942	if (osrc)
943		free((caddr_t)osrc, M_IFADDR);
944	if (odst)
945		free((caddr_t)odst, M_IFADDR);
946
947 bad:
948	if (sc->gif_psrc && sc->gif_pdst)
949		ifp->if_drv_flags |= IFF_DRV_RUNNING;
950	else
951		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
952
953	return error;
954}
955
956void
957gif_delete_tunnel(ifp)
958	struct ifnet *ifp;
959{
960	struct gif_softc *sc = ifp->if_softc;
961
962	if (sc->gif_psrc) {
963		free((caddr_t)sc->gif_psrc, M_IFADDR);
964		sc->gif_psrc = NULL;
965	}
966	if (sc->gif_pdst) {
967		free((caddr_t)sc->gif_pdst, M_IFADDR);
968		sc->gif_pdst = NULL;
969	}
970	/* it is safe to detach from both */
971#ifdef INET
972	(void)in_gif_detach(sc);
973#endif
974#ifdef INET6
975	(void)in6_gif_detach(sc);
976#endif
977	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
978}
979