if_gif.c revision 185088 
1/*	$FreeBSD: head/sys/net/if_gif.c 185088 2008-11-19 09:39:34Z zec $	*/
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/proc.h>
50#include <sys/protosw.h>
51#include <sys/conf.h>
52#include <sys/vimage.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
62#include <netinet/in.h>
63#include <netinet/in_systm.h>
64#include <netinet/ip.h>
65#ifdef	INET
66#include <netinet/in_var.h>
67#include <netinet/in_gif.h>
68#include <netinet/ip_var.h>
69#endif	/* INET */
70
71#ifdef INET6
72#ifndef INET
73#include <netinet/in.h>
74#endif
75#include <netinet6/in6_var.h>
76#include <netinet/ip6.h>
77#include <netinet6/ip6_var.h>
78#include <netinet6/scope6_var.h>
79#include <netinet6/in6_gif.h>
80#include <netinet6/ip6protosw.h>
81#endif /* INET6 */
82
83#include <netinet/ip_encap.h>
84#include <net/ethernet.h>
85#include <net/if_bridgevar.h>
86#include <net/if_gif.h>
87
88#include <security/mac/mac_framework.h>
89
90#define GIFNAME		"gif"
91
92/*
93 * gif_mtx protects the global gif_softc_list.
94 */
95static struct mtx gif_mtx;
96static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
97
98#ifdef VIMAGE_GLOBALS
99static LIST_HEAD(, gif_softc) gif_softc_list;
100static int max_gif_nesting;
101static int parallel_tunnels;
102#ifdef INET
103int ip_gif_ttl;
104#endif
105#ifdef INET6
106int ip6_gif_hlim;
107#endif
108#endif
109
110void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
111void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
112void	(*ng_gif_attach_p)(struct ifnet *ifp);
113void	(*ng_gif_detach_p)(struct ifnet *ifp);
114
115static void	gif_start(struct ifnet *);
116static int	gif_clone_create(struct if_clone *, int, caddr_t);
117static void	gif_clone_destroy(struct ifnet *);
118
119IFC_SIMPLE_DECLARE(gif, 0);
120
121static int gifmodevent(module_t, int, void *);
122
123SYSCTL_DECL(_net_link);
124SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
125    "Generic Tunnel Interface");
126#ifndef MAX_GIF_NEST
127/*
128 * This macro controls the default upper limitation on nesting of gif tunnels.
129 * Since, setting a large value to this macro with a careless configuration
130 * may introduce system crash, we don't allow any nestings by default.
131 * If you need to configure nested gif tunnels, you can define this macro
132 * in your kernel configuration file.  However, if you do so, please be
133 * careful to configure the tunnels so that it won't make a loop.
134 */
135#define MAX_GIF_NEST 1
136#endif
137SYSCTL_V_INT(V_NET, vnet_gif, _net_link_gif, OID_AUTO, max_nesting,
138    CTLFLAG_RW, max_gif_nesting, 0, "Max nested tunnels");
139
140#ifdef INET6
141SYSCTL_DECL(_net_inet6_ip6);
142SYSCTL_V_INT(V_NET, vnet_gif, _net_inet6_ip6, IPV6CTL_GIF_HLIM,
143    gifhlim, CTLFLAG_RW, ip6_gif_hlim, 0, "");
144#endif
145
146/*
147 * By default, we disallow creation of multiple tunnels between the same
148 * pair of addresses.  Some applications require this functionality so
149 * we allow control over this check here.
150 */
151SYSCTL_V_INT(V_NET, vnet_gif, _net_link_gif, OID_AUTO, parallel_tunnels,
152    CTLFLAG_RW, parallel_tunnels, 0, "Allow parallel tunnels?");
153
154/* copy from src/sys/net/if_ethersubr.c */
155static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
156			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
157#ifndef ETHER_IS_BROADCAST
158#define ETHER_IS_BROADCAST(addr) \
159	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
160#endif
161
162static int
163gif_clone_create(ifc, unit, params)
164	struct if_clone *ifc;
165	int unit;
166	caddr_t params;
167{
168	INIT_VNET_GIF(curvnet);
169	struct gif_softc *sc;
170
171	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
172	sc->gif_fibnum = curthread->td_proc->p_fibnum;
173	GIF2IFP(sc) = if_alloc(IFT_GIF);
174	if (GIF2IFP(sc) == NULL) {
175		free(sc, M_GIF);
176		return (ENOSPC);
177	}
178
179	GIF_LOCK_INIT(sc);
180
181	GIF2IFP(sc)->if_softc = sc;
182	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
183
184	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
185
186	GIF2IFP(sc)->if_addrlen = 0;
187	GIF2IFP(sc)->if_mtu    = GIF_MTU;
188	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
189#if 0
190	/* turn off ingress filter */
191	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
192#endif
193	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
194	GIF2IFP(sc)->if_start  = gif_start;
195	GIF2IFP(sc)->if_output = gif_output;
196	GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
197	if_attach(GIF2IFP(sc));
198	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
199	if (ng_gif_attach_p != NULL)
200		(*ng_gif_attach_p)(GIF2IFP(sc));
201
202	mtx_lock(&gif_mtx);
203	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
204	mtx_unlock(&gif_mtx);
205
206	return (0);
207}
208
209static void
210gif_clone_destroy(ifp)
211	struct ifnet *ifp;
212{
213#if defined(INET) || defined(INET6)
214	int err;
215#endif
216	struct gif_softc *sc = ifp->if_softc;
217
218	mtx_lock(&gif_mtx);
219	LIST_REMOVE(sc, gif_list);
220	mtx_unlock(&gif_mtx);
221
222	gif_delete_tunnel(ifp);
223#ifdef INET6
224	if (sc->encap_cookie6 != NULL) {
225		err = encap_detach(sc->encap_cookie6);
226		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
227	}
228#endif
229#ifdef INET
230	if (sc->encap_cookie4 != NULL) {
231		err = encap_detach(sc->encap_cookie4);
232		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
233	}
234#endif
235
236	if (ng_gif_detach_p != NULL)
237		(*ng_gif_detach_p)(ifp);
238	bpfdetach(ifp);
239	if_detach(ifp);
240	if_free(ifp);
241
242	GIF_LOCK_DESTROY(sc);
243
244	free(sc, M_GIF);
245}
246
247static int
248gifmodevent(mod, type, data)
249	module_t mod;
250	int type;
251	void *data;
252{
253
254	switch (type) {
255	case MOD_LOAD:
256		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
257
258		LIST_INIT(&V_gif_softc_list);
259		V_max_gif_nesting = MAX_GIF_NEST;
260#ifdef XBONEHACK
261		V_parallel_tunnels = 1;
262#else
263		V_parallel_tunnels = 0;
264#endif
265#ifdef INET
266		V_ip_gif_ttl = GIF_TTL;
267#endif
268#ifdef INET6
269		V_ip6_gif_hlim = GIF_HLIM;
270#endif
271		if_clone_attach(&gif_cloner);
272
273		break;
274	case MOD_UNLOAD:
275		if_clone_detach(&gif_cloner);
276		mtx_destroy(&gif_mtx);
277#ifdef INET6
278		V_ip6_gif_hlim = 0;
279#endif
280		break;
281	default:
282		return EOPNOTSUPP;
283	}
284	return 0;
285}
286
287static moduledata_t gif_mod = {
288	"if_gif",
289	gifmodevent,
290	0
291};
292
293DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
294MODULE_VERSION(if_gif, 1);
295
296int
297gif_encapcheck(m, off, proto, arg)
298	const struct mbuf *m;
299	int off;
300	int proto;
301	void *arg;
302{
303	struct ip ip;
304	struct gif_softc *sc;
305
306	sc = (struct gif_softc *)arg;
307	if (sc == NULL)
308		return 0;
309
310	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
311		return 0;
312
313	/* no physical address */
314	if (!sc->gif_psrc || !sc->gif_pdst)
315		return 0;
316
317	switch (proto) {
318#ifdef INET
319	case IPPROTO_IPV4:
320		break;
321#endif
322#ifdef INET6
323	case IPPROTO_IPV6:
324		break;
325#endif
326	case IPPROTO_ETHERIP:
327		break;
328
329	default:
330		return 0;
331	}
332
333	/* Bail on short packets */
334	if (m->m_pkthdr.len < sizeof(ip))
335		return 0;
336
337	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
338
339	switch (ip.ip_v) {
340#ifdef INET
341	case 4:
342		if (sc->gif_psrc->sa_family != AF_INET ||
343		    sc->gif_pdst->sa_family != AF_INET)
344			return 0;
345		return gif_encapcheck4(m, off, proto, arg);
346#endif
347#ifdef INET6
348	case 6:
349		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
350			return 0;
351		if (sc->gif_psrc->sa_family != AF_INET6 ||
352		    sc->gif_pdst->sa_family != AF_INET6)
353			return 0;
354		return gif_encapcheck6(m, off, proto, arg);
355#endif
356	default:
357		return 0;
358	}
359}
360
361static void
362gif_start(struct ifnet *ifp)
363{
364	struct gif_softc *sc;
365	struct mbuf *m;
366
367	sc = ifp->if_softc;
368
369	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
370	for (;;) {
371		IFQ_DEQUEUE(&ifp->if_snd, m);
372		if (m == 0)
373			break;
374
375		gif_output(ifp, m, sc->gif_pdst, NULL);
376
377	}
378	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
379
380	return;
381}
382
383int
384gif_output(ifp, m, dst, rt)
385	struct ifnet *ifp;
386	struct mbuf *m;
387	struct sockaddr *dst;
388	struct rtentry *rt;	/* added in net2 */
389{
390	INIT_VNET_GIF(ifp->if_vnet);
391	struct gif_softc *sc = ifp->if_softc;
392	struct m_tag *mtag;
393	int error = 0;
394	int gif_called;
395	u_int32_t af;
396
397#ifdef MAC
398	error = mac_ifnet_check_transmit(ifp, m);
399	if (error) {
400		m_freem(m);
401		goto end;
402	}
403#endif
404
405	/*
406	 * gif may cause infinite recursion calls when misconfigured.
407	 * We'll prevent this by detecting loops.
408	 *
409	 * High nesting level may cause stack exhaustion.
410	 * We'll prevent this by introducing upper limit.
411	 */
412	gif_called = 1;
413	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
414	while (mtag != NULL) {
415		if (*(struct ifnet **)(mtag + 1) == ifp) {
416			log(LOG_NOTICE,
417			    "gif_output: loop detected on %s\n",
418			    (*(struct ifnet **)(mtag + 1))->if_xname);
419			m_freem(m);
420			error = EIO;	/* is there better errno? */
421			goto end;
422		}
423		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
424		gif_called++;
425	}
426	if (gif_called > V_max_gif_nesting) {
427		log(LOG_NOTICE,
428		    "gif_output: recursively called too many times(%d)\n",
429		    gif_called);
430		m_freem(m);
431		error = EIO;	/* is there better errno? */
432		goto end;
433	}
434	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
435	    M_NOWAIT);
436	if (mtag == NULL) {
437		m_freem(m);
438		error = ENOMEM;
439		goto end;
440	}
441	*(struct ifnet **)(mtag + 1) = ifp;
442	m_tag_prepend(m, mtag);
443
444	m->m_flags &= ~(M_BCAST|M_MCAST);
445
446	GIF_LOCK(sc);
447
448	if (!(ifp->if_flags & IFF_UP) ||
449	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
450		GIF_UNLOCK(sc);
451		m_freem(m);
452		error = ENETDOWN;
453		goto end;
454	}
455
456	/* BPF writes need to be handled specially. */
457	if (dst->sa_family == AF_UNSPEC) {
458		bcopy(dst->sa_data, &af, sizeof(af));
459		dst->sa_family = af;
460	}
461
462	af = dst->sa_family;
463	BPF_MTAP2(ifp, &af, sizeof(af), m);
464	ifp->if_opackets++;
465	ifp->if_obytes += m->m_pkthdr.len;
466
467	/* override to IPPROTO_ETHERIP for bridged traffic */
468	if (ifp->if_bridge)
469		af = AF_LINK;
470
471	M_SETFIB(m, sc->gif_fibnum);
472	/* inner AF-specific encapsulation */
473
474	/* XXX should we check if our outer source is legal? */
475
476	/* dispatch to output logic based on outer AF */
477	switch (sc->gif_psrc->sa_family) {
478#ifdef INET
479	case AF_INET:
480		error = in_gif_output(ifp, af, m);
481		break;
482#endif
483#ifdef INET6
484	case AF_INET6:
485		error = in6_gif_output(ifp, af, m);
486		break;
487#endif
488	default:
489		m_freem(m);
490		error = ENETDOWN;
491	}
492
493	GIF_UNLOCK(sc);
494  end:
495	if (error)
496		ifp->if_oerrors++;
497	return (error);
498}
499
500void
501gif_input(m, af, ifp)
502	struct mbuf *m;
503	int af;
504	struct ifnet *ifp;
505{
506	int isr, n;
507	struct etherip_header *eip;
508	struct ether_header *eh;
509	struct ifnet *oldifp;
510
511	if (ifp == NULL) {
512		/* just in case */
513		m_freem(m);
514		return;
515	}
516
517	m->m_pkthdr.rcvif = ifp;
518
519#ifdef MAC
520	mac_ifnet_create_mbuf(ifp, m);
521#endif
522
523	if (bpf_peers_present(ifp->if_bpf)) {
524		u_int32_t af1 = af;
525		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
526	}
527
528	if (ng_gif_input_p != NULL) {
529		(*ng_gif_input_p)(ifp, &m, af);
530		if (m == NULL)
531			return;
532	}
533
534	/*
535	 * Put the packet to the network layer input queue according to the
536	 * specified address family.
537	 * Note: older versions of gif_input directly called network layer
538	 * input functions, e.g. ip6_input, here.  We changed the policy to
539	 * prevent too many recursive calls of such input functions, which
540	 * might cause kernel panic.  But the change may introduce another
541	 * problem; if the input queue is full, packets are discarded.
542	 * The kernel stack overflow really happened, and we believed
543	 * queue-full rarely occurs, so we changed the policy.
544	 */
545	switch (af) {
546#ifdef INET
547	case AF_INET:
548		isr = NETISR_IP;
549		break;
550#endif
551#ifdef INET6
552	case AF_INET6:
553		isr = NETISR_IPV6;
554		break;
555#endif
556	case AF_LINK:
557		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
558		if (n > m->m_len) {
559			m = m_pullup(m, n);
560			if (m == NULL) {
561				ifp->if_ierrors++;
562				return;
563			}
564		}
565
566		eip = mtod(m, struct etherip_header *);
567 		if (eip->eip_ver !=
568		    (ETHERIP_VERSION & ETHERIP_VER_VERS_MASK)) {
569			/* discard unknown versions */
570			m_freem(m);
571			return;
572		}
573		m_adj(m, sizeof(struct etherip_header));
574
575		m->m_flags &= ~(M_BCAST|M_MCAST);
576		m->m_pkthdr.rcvif = ifp;
577
578		if (ifp->if_bridge) {
579			oldifp = ifp;
580			eh = mtod(m, struct ether_header *);
581			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
582				if (ETHER_IS_BROADCAST(eh->ether_dhost))
583					m->m_flags |= M_BCAST;
584				else
585					m->m_flags |= M_MCAST;
586				ifp->if_imcasts++;
587			}
588			BRIDGE_INPUT(ifp, m);
589
590			if (m != NULL && ifp != oldifp) {
591				/*
592				 * The bridge gave us back itself or one of the
593				 * members for which the frame is addressed.
594				 */
595				ether_demux(ifp, m);
596				return;
597			}
598		}
599		if (m != NULL)
600			m_freem(m);
601		return;
602
603	default:
604		if (ng_gif_input_orphan_p != NULL)
605			(*ng_gif_input_orphan_p)(ifp, m, af);
606		else
607			m_freem(m);
608		return;
609	}
610
611	ifp->if_ipackets++;
612	ifp->if_ibytes += m->m_pkthdr.len;
613	netisr_dispatch(isr, m);
614}
615
616/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
617int
618gif_ioctl(ifp, cmd, data)
619	struct ifnet *ifp;
620	u_long cmd;
621	caddr_t data;
622{
623	struct gif_softc *sc  = ifp->if_softc;
624	struct ifreq     *ifr = (struct ifreq*)data;
625	int error = 0, size;
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		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
822#ifdef INET6
823		if (dst->sa_family == AF_INET6) {
824			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
825			if (error != 0)
826				return (error);
827		}
828#endif
829		break;
830
831	case SIOCGLIFPHYADDR:
832		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
833			error = EADDRNOTAVAIL;
834			goto bad;
835		}
836
837		/* copy src */
838		src = sc->gif_psrc;
839		dst = (struct sockaddr *)
840			&(((struct if_laddrreq *)data)->addr);
841		size = sizeof(((struct if_laddrreq *)data)->addr);
842		if (src->sa_len > size)
843			return EINVAL;
844		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
845
846		/* copy dst */
847		src = sc->gif_pdst;
848		dst = (struct sockaddr *)
849			&(((struct if_laddrreq *)data)->dstaddr);
850		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
851		if (src->sa_len > size)
852			return EINVAL;
853		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
854		break;
855
856	case SIOCSIFFLAGS:
857		/* if_ioctl() takes care of it */
858		break;
859
860	default:
861		error = EINVAL;
862		break;
863	}
864 bad:
865	return error;
866}
867
868/*
869 * XXXRW: There's a general event-ordering issue here: the code to check
870 * if a given tunnel is already present happens before we perform a
871 * potentially blocking setup of the tunnel.  This code needs to be
872 * re-ordered so that the check and replacement can be atomic using
873 * a mutex.
874 */
875int
876gif_set_tunnel(ifp, src, dst)
877	struct ifnet *ifp;
878	struct sockaddr *src;
879	struct sockaddr *dst;
880{
881	INIT_VNET_GIF(ifp->if_vnet);
882	struct gif_softc *sc = ifp->if_softc;
883	struct gif_softc *sc2;
884	struct sockaddr *osrc, *odst, *sa;
885	int error = 0;
886
887	mtx_lock(&gif_mtx);
888	LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
889		if (sc2 == sc)
890			continue;
891		if (!sc2->gif_pdst || !sc2->gif_psrc)
892			continue;
893		if (sc2->gif_pdst->sa_family != dst->sa_family ||
894		    sc2->gif_pdst->sa_len != dst->sa_len ||
895		    sc2->gif_psrc->sa_family != src->sa_family ||
896		    sc2->gif_psrc->sa_len != src->sa_len)
897			continue;
898
899		/*
900		 * Disallow parallel tunnels unless instructed
901		 * otherwise.
902		 */
903		if (!V_parallel_tunnels &&
904		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
905		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
906			error = EADDRNOTAVAIL;
907			mtx_unlock(&gif_mtx);
908			goto bad;
909		}
910
911		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
912	}
913	mtx_unlock(&gif_mtx);
914
915	/* XXX we can detach from both, but be polite just in case */
916	if (sc->gif_psrc)
917		switch (sc->gif_psrc->sa_family) {
918#ifdef INET
919		case AF_INET:
920			(void)in_gif_detach(sc);
921			break;
922#endif
923#ifdef INET6
924		case AF_INET6:
925			(void)in6_gif_detach(sc);
926			break;
927#endif
928		}
929
930	osrc = sc->gif_psrc;
931	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
932	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
933	sc->gif_psrc = sa;
934
935	odst = sc->gif_pdst;
936	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
937	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
938	sc->gif_pdst = sa;
939
940	switch (sc->gif_psrc->sa_family) {
941#ifdef INET
942	case AF_INET:
943		error = in_gif_attach(sc);
944		break;
945#endif
946#ifdef INET6
947	case AF_INET6:
948		/*
949		 * Check validity of the scope zone ID of the addresses, and
950		 * convert it into the kernel internal form if necessary.
951		 */
952		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
953		if (error != 0)
954			break;
955		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
956		if (error != 0)
957			break;
958		error = in6_gif_attach(sc);
959		break;
960#endif
961	}
962	if (error) {
963		/* rollback */
964		free((caddr_t)sc->gif_psrc, M_IFADDR);
965		free((caddr_t)sc->gif_pdst, M_IFADDR);
966		sc->gif_psrc = osrc;
967		sc->gif_pdst = odst;
968		goto bad;
969	}
970
971	if (osrc)
972		free((caddr_t)osrc, M_IFADDR);
973	if (odst)
974		free((caddr_t)odst, M_IFADDR);
975
976 bad:
977	if (sc->gif_psrc && sc->gif_pdst)
978		ifp->if_drv_flags |= IFF_DRV_RUNNING;
979	else
980		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
981
982	return error;
983}
984
985void
986gif_delete_tunnel(ifp)
987	struct ifnet *ifp;
988{
989	struct gif_softc *sc = ifp->if_softc;
990
991	if (sc->gif_psrc) {
992		free((caddr_t)sc->gif_psrc, M_IFADDR);
993		sc->gif_psrc = NULL;
994	}
995	if (sc->gif_pdst) {
996		free((caddr_t)sc->gif_pdst, M_IFADDR);
997		sc->gif_pdst = NULL;
998	}
999	/* it is safe to detach from both */
1000#ifdef INET
1001	(void)in_gif_detach(sc);
1002#endif
1003#ifdef INET6
1004	(void)in6_gif_detach(sc);
1005#endif
1006	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1007}
1008