if_gif.c revision 193664
1/*	$FreeBSD: head/sys/net/if_gif.c 193664 2009-06-07 23:00:40Z hrs $	*/
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/kernel.h>
39#include <sys/malloc.h>
40#include <sys/mbuf.h>
41#include <sys/module.h>
42#include <sys/socket.h>
43#include <sys/sockio.h>
44#include <sys/errno.h>
45#include <sys/time.h>
46#include <sys/sysctl.h>
47#include <sys/syslog.h>
48#include <sys/priv.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#ifndef VIMAGE
99#ifndef VIMAGE_GLOBALS
100struct vnet_gif vnet_gif_0;
101#endif
102#endif
103
104#ifdef VIMAGE_GLOBALS
105static LIST_HEAD(, gif_softc) gif_softc_list;
106static int max_gif_nesting;
107static int parallel_tunnels;
108#ifdef INET
109int ip_gif_ttl;
110#endif
111#ifdef INET6
112int ip6_gif_hlim;
113#endif
114#endif
115
116void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
117void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
118void	(*ng_gif_attach_p)(struct ifnet *ifp);
119void	(*ng_gif_detach_p)(struct ifnet *ifp);
120
121static void	gif_start(struct ifnet *);
122static int	gif_clone_create(struct if_clone *, int, caddr_t);
123static void	gif_clone_destroy(struct ifnet *);
124static int	vnet_gif_iattach(const void *);
125
126#ifndef VIMAGE_GLOBALS
127static const vnet_modinfo_t vnet_gif_modinfo = {
128	.vmi_id		= VNET_MOD_GIF,
129	.vmi_name	= "gif",
130	.vmi_size	= sizeof(struct vnet_gif),
131	.vmi_dependson	= VNET_MOD_NET,
132	.vmi_iattach	= vnet_gif_iattach
133};
134#endif
135
136IFC_SIMPLE_DECLARE(gif, 0);
137
138static int gifmodevent(module_t, int, void *);
139
140SYSCTL_DECL(_net_link);
141SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
142    "Generic Tunnel Interface");
143#ifndef MAX_GIF_NEST
144/*
145 * This macro controls the default upper limitation on nesting of gif tunnels.
146 * Since, setting a large value to this macro with a careless configuration
147 * may introduce system crash, we don't allow any nestings by default.
148 * If you need to configure nested gif tunnels, you can define this macro
149 * in your kernel configuration file.  However, if you do so, please be
150 * careful to configure the tunnels so that it won't make a loop.
151 */
152#define MAX_GIF_NEST 1
153#endif
154SYSCTL_V_INT(V_NET, vnet_gif, _net_link_gif, OID_AUTO, max_nesting,
155    CTLFLAG_RW, max_gif_nesting, 0, "Max nested tunnels");
156
157#ifdef INET6
158SYSCTL_DECL(_net_inet6_ip6);
159SYSCTL_V_INT(V_NET, vnet_gif, _net_inet6_ip6, IPV6CTL_GIF_HLIM,
160    gifhlim, CTLFLAG_RW, ip6_gif_hlim, 0, "");
161#endif
162
163/*
164 * By default, we disallow creation of multiple tunnels between the same
165 * pair of addresses.  Some applications require this functionality so
166 * we allow control over this check here.
167 */
168SYSCTL_V_INT(V_NET, vnet_gif, _net_link_gif, OID_AUTO, parallel_tunnels,
169    CTLFLAG_RW, parallel_tunnels, 0, "Allow parallel tunnels?");
170
171/* copy from src/sys/net/if_ethersubr.c */
172static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
173			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
174#ifndef ETHER_IS_BROADCAST
175#define ETHER_IS_BROADCAST(addr) \
176	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
177#endif
178
179static int
180gif_clone_create(ifc, unit, params)
181	struct if_clone *ifc;
182	int unit;
183	caddr_t params;
184{
185	INIT_VNET_GIF(curvnet);
186	struct gif_softc *sc;
187
188	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
189	sc->gif_fibnum = curthread->td_proc->p_fibnum;
190	GIF2IFP(sc) = if_alloc(IFT_GIF);
191	if (GIF2IFP(sc) == NULL) {
192		free(sc, M_GIF);
193		return (ENOSPC);
194	}
195
196	GIF_LOCK_INIT(sc);
197
198	GIF2IFP(sc)->if_softc = sc;
199	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
200
201	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
202	sc->gif_options = GIF_ACCEPT_REVETHIP;
203
204	GIF2IFP(sc)->if_addrlen = 0;
205	GIF2IFP(sc)->if_mtu    = GIF_MTU;
206	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
207#if 0
208	/* turn off ingress filter */
209	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
210#endif
211	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
212	GIF2IFP(sc)->if_start  = gif_start;
213	GIF2IFP(sc)->if_output = gif_output;
214	GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
215	if_attach(GIF2IFP(sc));
216	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
217	if (ng_gif_attach_p != NULL)
218		(*ng_gif_attach_p)(GIF2IFP(sc));
219
220	mtx_lock(&gif_mtx);
221	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
222	mtx_unlock(&gif_mtx);
223
224	return (0);
225}
226
227static void
228gif_clone_destroy(ifp)
229	struct ifnet *ifp;
230{
231#if defined(INET) || defined(INET6)
232	int err;
233#endif
234	struct gif_softc *sc = ifp->if_softc;
235
236	mtx_lock(&gif_mtx);
237	LIST_REMOVE(sc, gif_list);
238	mtx_unlock(&gif_mtx);
239
240	gif_delete_tunnel(ifp);
241#ifdef INET6
242	if (sc->encap_cookie6 != NULL) {
243		err = encap_detach(sc->encap_cookie6);
244		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
245	}
246#endif
247#ifdef INET
248	if (sc->encap_cookie4 != NULL) {
249		err = encap_detach(sc->encap_cookie4);
250		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
251	}
252#endif
253
254	if (ng_gif_detach_p != NULL)
255		(*ng_gif_detach_p)(ifp);
256	bpfdetach(ifp);
257	if_detach(ifp);
258	if_free(ifp);
259
260	GIF_LOCK_DESTROY(sc);
261
262	free(sc, M_GIF);
263}
264
265static int
266vnet_gif_iattach(const void *unused __unused)
267{
268	INIT_VNET_GIF(curvnet);
269
270	LIST_INIT(&V_gif_softc_list);
271	V_max_gif_nesting = MAX_GIF_NEST;
272#ifdef XBONEHACK
273	V_parallel_tunnels = 1;
274#else
275	V_parallel_tunnels = 0;
276#endif
277	V_ip_gif_ttl = GIF_TTL;
278#ifdef INET6
279	V_ip6_gif_hlim = GIF_HLIM;
280#endif
281
282	return (0);
283}
284
285static int
286gifmodevent(mod, type, data)
287	module_t mod;
288	int type;
289	void *data;
290{
291
292	switch (type) {
293	case MOD_LOAD:
294		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
295
296#ifndef VIMAGE_GLOBALS
297		vnet_mod_register(&vnet_gif_modinfo);
298#else
299		vnet_gif_iattach(NULL);
300#endif
301		if_clone_attach(&gif_cloner);
302
303		break;
304	case MOD_UNLOAD:
305		if_clone_detach(&gif_cloner);
306		mtx_destroy(&gif_mtx);
307#ifdef INET6
308#ifndef VIMAGE
309		V_ip6_gif_hlim = 0;	/* XXX -> vnet_gif_idetach() */
310#endif
311#endif
312		break;
313	default:
314		return EOPNOTSUPP;
315	}
316	return 0;
317}
318
319static moduledata_t gif_mod = {
320	"if_gif",
321	gifmodevent,
322	0
323};
324
325DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
326MODULE_VERSION(if_gif, 1);
327
328int
329gif_encapcheck(m, off, proto, arg)
330	const struct mbuf *m;
331	int off;
332	int proto;
333	void *arg;
334{
335	struct ip ip;
336	struct gif_softc *sc;
337
338	sc = (struct gif_softc *)arg;
339	if (sc == NULL)
340		return 0;
341
342	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
343		return 0;
344
345	/* no physical address */
346	if (!sc->gif_psrc || !sc->gif_pdst)
347		return 0;
348
349	switch (proto) {
350#ifdef INET
351	case IPPROTO_IPV4:
352		break;
353#endif
354#ifdef INET6
355	case IPPROTO_IPV6:
356		break;
357#endif
358	case IPPROTO_ETHERIP:
359		break;
360
361	default:
362		return 0;
363	}
364
365	/* Bail on short packets */
366	if (m->m_pkthdr.len < sizeof(ip))
367		return 0;
368
369	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
370
371	switch (ip.ip_v) {
372#ifdef INET
373	case 4:
374		if (sc->gif_psrc->sa_family != AF_INET ||
375		    sc->gif_pdst->sa_family != AF_INET)
376			return 0;
377		return gif_encapcheck4(m, off, proto, arg);
378#endif
379#ifdef INET6
380	case 6:
381		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
382			return 0;
383		if (sc->gif_psrc->sa_family != AF_INET6 ||
384		    sc->gif_pdst->sa_family != AF_INET6)
385			return 0;
386		return gif_encapcheck6(m, off, proto, arg);
387#endif
388	default:
389		return 0;
390	}
391}
392
393static void
394gif_start(struct ifnet *ifp)
395{
396	struct gif_softc *sc;
397	struct mbuf *m;
398
399	sc = ifp->if_softc;
400
401	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
402	for (;;) {
403		IFQ_DEQUEUE(&ifp->if_snd, m);
404		if (m == 0)
405			break;
406
407		gif_output(ifp, m, sc->gif_pdst, NULL);
408
409	}
410	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
411
412	return;
413}
414
415int
416gif_output(ifp, m, dst, ro)
417	struct ifnet *ifp;
418	struct mbuf *m;
419	struct sockaddr *dst;
420	struct route *ro;
421{
422	INIT_VNET_GIF(ifp->if_vnet);
423	struct gif_softc *sc = ifp->if_softc;
424	struct m_tag *mtag;
425	int error = 0;
426	int gif_called;
427	u_int32_t af;
428
429#ifdef MAC
430	error = mac_ifnet_check_transmit(ifp, m);
431	if (error) {
432		m_freem(m);
433		goto end;
434	}
435#endif
436
437	/*
438	 * gif may cause infinite recursion calls when misconfigured.
439	 * We'll prevent this by detecting loops.
440	 *
441	 * High nesting level may cause stack exhaustion.
442	 * We'll prevent this by introducing upper limit.
443	 */
444	gif_called = 1;
445	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
446	while (mtag != NULL) {
447		if (*(struct ifnet **)(mtag + 1) == ifp) {
448			log(LOG_NOTICE,
449			    "gif_output: loop detected on %s\n",
450			    (*(struct ifnet **)(mtag + 1))->if_xname);
451			m_freem(m);
452			error = EIO;	/* is there better errno? */
453			goto end;
454		}
455		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
456		gif_called++;
457	}
458	if (gif_called > V_max_gif_nesting) {
459		log(LOG_NOTICE,
460		    "gif_output: recursively called too many times(%d)\n",
461		    gif_called);
462		m_freem(m);
463		error = EIO;	/* is there better errno? */
464		goto end;
465	}
466	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
467	    M_NOWAIT);
468	if (mtag == NULL) {
469		m_freem(m);
470		error = ENOMEM;
471		goto end;
472	}
473	*(struct ifnet **)(mtag + 1) = ifp;
474	m_tag_prepend(m, mtag);
475
476	m->m_flags &= ~(M_BCAST|M_MCAST);
477
478	GIF_LOCK(sc);
479
480	if (!(ifp->if_flags & IFF_UP) ||
481	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
482		GIF_UNLOCK(sc);
483		m_freem(m);
484		error = ENETDOWN;
485		goto end;
486	}
487
488	/* BPF writes need to be handled specially. */
489	if (dst->sa_family == AF_UNSPEC) {
490		bcopy(dst->sa_data, &af, sizeof(af));
491		dst->sa_family = af;
492	}
493
494	af = dst->sa_family;
495	BPF_MTAP2(ifp, &af, sizeof(af), m);
496	ifp->if_opackets++;
497	ifp->if_obytes += m->m_pkthdr.len;
498
499	/* override to IPPROTO_ETHERIP for bridged traffic */
500	if (ifp->if_bridge)
501		af = AF_LINK;
502
503	M_SETFIB(m, sc->gif_fibnum);
504	/* inner AF-specific encapsulation */
505
506	/* XXX should we check if our outer source is legal? */
507
508	/* dispatch to output logic based on outer AF */
509	switch (sc->gif_psrc->sa_family) {
510#ifdef INET
511	case AF_INET:
512		error = in_gif_output(ifp, af, m);
513		break;
514#endif
515#ifdef INET6
516	case AF_INET6:
517		error = in6_gif_output(ifp, af, m);
518		break;
519#endif
520	default:
521		m_freem(m);
522		error = ENETDOWN;
523	}
524
525	GIF_UNLOCK(sc);
526  end:
527	if (error)
528		ifp->if_oerrors++;
529	return (error);
530}
531
532void
533gif_input(m, af, ifp)
534	struct mbuf *m;
535	int af;
536	struct ifnet *ifp;
537{
538	int isr, n;
539	struct gif_softc *sc = ifp->if_softc;
540	struct etherip_header *eip;
541	struct ether_header *eh;
542	struct ifnet *oldifp;
543
544	if (ifp == NULL) {
545		/* just in case */
546		m_freem(m);
547		return;
548	}
549
550	m->m_pkthdr.rcvif = ifp;
551
552#ifdef MAC
553	mac_ifnet_create_mbuf(ifp, m);
554#endif
555
556	if (bpf_peers_present(ifp->if_bpf)) {
557		u_int32_t af1 = af;
558		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
559	}
560
561	if (ng_gif_input_p != NULL) {
562		(*ng_gif_input_p)(ifp, &m, af);
563		if (m == NULL)
564			return;
565	}
566
567	/*
568	 * Put the packet to the network layer input queue according to the
569	 * specified address family.
570	 * Note: older versions of gif_input directly called network layer
571	 * input functions, e.g. ip6_input, here.  We changed the policy to
572	 * prevent too many recursive calls of such input functions, which
573	 * might cause kernel panic.  But the change may introduce another
574	 * problem; if the input queue is full, packets are discarded.
575	 * The kernel stack overflow really happened, and we believed
576	 * queue-full rarely occurs, so we changed the policy.
577	 */
578	switch (af) {
579#ifdef INET
580	case AF_INET:
581		isr = NETISR_IP;
582		break;
583#endif
584#ifdef INET6
585	case AF_INET6:
586		isr = NETISR_IPV6;
587		break;
588#endif
589	case AF_LINK:
590		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
591		if (n > m->m_len) {
592			m = m_pullup(m, n);
593			if (m == NULL) {
594				ifp->if_ierrors++;
595				return;
596			}
597		}
598
599		eip = mtod(m, struct etherip_header *);
600		/*
601		 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
602		 * accepts an EtherIP packet with revered version field in
603		 * the header.  This is a knob for backward compatibility
604		 * with FreeBSD 7.2R or prior.
605		 */
606		if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
607			if (eip->eip_resvl != ETHERIP_VERSION
608			    && eip->eip_ver != ETHERIP_VERSION) {
609				/* discard unknown versions */
610				m_freem(m);
611				return;
612			}
613		} else {
614			if (eip->eip_ver != ETHERIP_VERSION) {
615				/* discard unknown versions */
616				m_freem(m);
617				return;
618			}
619		}
620		m_adj(m, sizeof(struct etherip_header));
621
622		m->m_flags &= ~(M_BCAST|M_MCAST);
623		m->m_pkthdr.rcvif = ifp;
624
625		if (ifp->if_bridge) {
626			oldifp = ifp;
627			eh = mtod(m, struct ether_header *);
628			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
629				if (ETHER_IS_BROADCAST(eh->ether_dhost))
630					m->m_flags |= M_BCAST;
631				else
632					m->m_flags |= M_MCAST;
633				ifp->if_imcasts++;
634			}
635			BRIDGE_INPUT(ifp, m);
636
637			if (m != NULL && ifp != oldifp) {
638				/*
639				 * The bridge gave us back itself or one of the
640				 * members for which the frame is addressed.
641				 */
642				ether_demux(ifp, m);
643				return;
644			}
645		}
646		if (m != NULL)
647			m_freem(m);
648		return;
649
650	default:
651		if (ng_gif_input_orphan_p != NULL)
652			(*ng_gif_input_orphan_p)(ifp, m, af);
653		else
654			m_freem(m);
655		return;
656	}
657
658	ifp->if_ipackets++;
659	ifp->if_ibytes += m->m_pkthdr.len;
660	netisr_dispatch(isr, m);
661}
662
663/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
664int
665gif_ioctl(ifp, cmd, data)
666	struct ifnet *ifp;
667	u_long cmd;
668	caddr_t data;
669{
670	struct gif_softc *sc  = ifp->if_softc;
671	struct ifreq     *ifr = (struct ifreq*)data;
672	int error = 0, size;
673	u_int	options;
674	struct sockaddr *dst, *src;
675#ifdef	SIOCSIFMTU /* xxx */
676	u_long mtu;
677#endif
678
679	switch (cmd) {
680	case SIOCSIFADDR:
681		ifp->if_flags |= IFF_UP;
682		break;
683
684	case SIOCSIFDSTADDR:
685		break;
686
687	case SIOCADDMULTI:
688	case SIOCDELMULTI:
689		break;
690
691#ifdef	SIOCSIFMTU /* xxx */
692	case SIOCGIFMTU:
693		break;
694
695	case SIOCSIFMTU:
696		mtu = ifr->ifr_mtu;
697		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
698			return (EINVAL);
699		ifp->if_mtu = mtu;
700		break;
701#endif /* SIOCSIFMTU */
702
703#ifdef INET
704	case SIOCSIFPHYADDR:
705#endif
706#ifdef INET6
707	case SIOCSIFPHYADDR_IN6:
708#endif /* INET6 */
709	case SIOCSLIFPHYADDR:
710		switch (cmd) {
711#ifdef INET
712		case SIOCSIFPHYADDR:
713			src = (struct sockaddr *)
714				&(((struct in_aliasreq *)data)->ifra_addr);
715			dst = (struct sockaddr *)
716				&(((struct in_aliasreq *)data)->ifra_dstaddr);
717			break;
718#endif
719#ifdef INET6
720		case SIOCSIFPHYADDR_IN6:
721			src = (struct sockaddr *)
722				&(((struct in6_aliasreq *)data)->ifra_addr);
723			dst = (struct sockaddr *)
724				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
725			break;
726#endif
727		case SIOCSLIFPHYADDR:
728			src = (struct sockaddr *)
729				&(((struct if_laddrreq *)data)->addr);
730			dst = (struct sockaddr *)
731				&(((struct if_laddrreq *)data)->dstaddr);
732			break;
733		default:
734			return EINVAL;
735		}
736
737		/* sa_family must be equal */
738		if (src->sa_family != dst->sa_family)
739			return EINVAL;
740
741		/* validate sa_len */
742		switch (src->sa_family) {
743#ifdef INET
744		case AF_INET:
745			if (src->sa_len != sizeof(struct sockaddr_in))
746				return EINVAL;
747			break;
748#endif
749#ifdef INET6
750		case AF_INET6:
751			if (src->sa_len != sizeof(struct sockaddr_in6))
752				return EINVAL;
753			break;
754#endif
755		default:
756			return EAFNOSUPPORT;
757		}
758		switch (dst->sa_family) {
759#ifdef INET
760		case AF_INET:
761			if (dst->sa_len != sizeof(struct sockaddr_in))
762				return EINVAL;
763			break;
764#endif
765#ifdef INET6
766		case AF_INET6:
767			if (dst->sa_len != sizeof(struct sockaddr_in6))
768				return EINVAL;
769			break;
770#endif
771		default:
772			return EAFNOSUPPORT;
773		}
774
775		/* check sa_family looks sane for the cmd */
776		switch (cmd) {
777		case SIOCSIFPHYADDR:
778			if (src->sa_family == AF_INET)
779				break;
780			return EAFNOSUPPORT;
781#ifdef INET6
782		case SIOCSIFPHYADDR_IN6:
783			if (src->sa_family == AF_INET6)
784				break;
785			return EAFNOSUPPORT;
786#endif /* INET6 */
787		case SIOCSLIFPHYADDR:
788			/* checks done in the above */
789			break;
790		}
791
792		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
793		break;
794
795#ifdef SIOCDIFPHYADDR
796	case SIOCDIFPHYADDR:
797		gif_delete_tunnel(GIF2IFP(sc));
798		break;
799#endif
800
801	case SIOCGIFPSRCADDR:
802#ifdef INET6
803	case SIOCGIFPSRCADDR_IN6:
804#endif /* INET6 */
805		if (sc->gif_psrc == NULL) {
806			error = EADDRNOTAVAIL;
807			goto bad;
808		}
809		src = sc->gif_psrc;
810		switch (cmd) {
811#ifdef INET
812		case SIOCGIFPSRCADDR:
813			dst = &ifr->ifr_addr;
814			size = sizeof(ifr->ifr_addr);
815			break;
816#endif /* INET */
817#ifdef INET6
818		case SIOCGIFPSRCADDR_IN6:
819			dst = (struct sockaddr *)
820				&(((struct in6_ifreq *)data)->ifr_addr);
821			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
822			break;
823#endif /* INET6 */
824		default:
825			error = EADDRNOTAVAIL;
826			goto bad;
827		}
828		if (src->sa_len > size)
829			return EINVAL;
830		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
831#ifdef INET6
832		if (dst->sa_family == AF_INET6) {
833			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
834			if (error != 0)
835				return (error);
836		}
837#endif
838		break;
839
840	case SIOCGIFPDSTADDR:
841#ifdef INET6
842	case SIOCGIFPDSTADDR_IN6:
843#endif /* INET6 */
844		if (sc->gif_pdst == NULL) {
845			error = EADDRNOTAVAIL;
846			goto bad;
847		}
848		src = sc->gif_pdst;
849		switch (cmd) {
850#ifdef INET
851		case SIOCGIFPDSTADDR:
852			dst = &ifr->ifr_addr;
853			size = sizeof(ifr->ifr_addr);
854			break;
855#endif /* INET */
856#ifdef INET6
857		case SIOCGIFPDSTADDR_IN6:
858			dst = (struct sockaddr *)
859				&(((struct in6_ifreq *)data)->ifr_addr);
860			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
861			break;
862#endif /* INET6 */
863		default:
864			error = EADDRNOTAVAIL;
865			goto bad;
866		}
867		if (src->sa_len > size)
868			return EINVAL;
869		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
870#ifdef INET6
871		if (dst->sa_family == AF_INET6) {
872			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
873			if (error != 0)
874				return (error);
875		}
876#endif
877		break;
878
879	case SIOCGLIFPHYADDR:
880		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
881			error = EADDRNOTAVAIL;
882			goto bad;
883		}
884
885		/* copy src */
886		src = sc->gif_psrc;
887		dst = (struct sockaddr *)
888			&(((struct if_laddrreq *)data)->addr);
889		size = sizeof(((struct if_laddrreq *)data)->addr);
890		if (src->sa_len > size)
891			return EINVAL;
892		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
893
894		/* copy dst */
895		src = sc->gif_pdst;
896		dst = (struct sockaddr *)
897			&(((struct if_laddrreq *)data)->dstaddr);
898		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
899		if (src->sa_len > size)
900			return EINVAL;
901		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
902		break;
903
904	case SIOCSIFFLAGS:
905		/* if_ioctl() takes care of it */
906		break;
907
908	case GIFGOPTS:
909		options = sc->gif_options;
910		error = copyout(&options, ifr->ifr_data,
911				sizeof(options));
912		break;
913
914	case GIFSOPTS:
915		if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
916			break;
917		if ((error = copyin(&options, &sc->gif_options,
918				sizeof(sc->gif_options)))) {
919			if ((options | GIF_FULLOPTS) == GIF_FULLOPTS)
920				ifr->ifr_data = (caddr_t)options;
921			else
922				error = EINVAL;
923		}
924		break;
925
926	default:
927		error = EINVAL;
928		break;
929	}
930 bad:
931	return error;
932}
933
934/*
935 * XXXRW: There's a general event-ordering issue here: the code to check
936 * if a given tunnel is already present happens before we perform a
937 * potentially blocking setup of the tunnel.  This code needs to be
938 * re-ordered so that the check and replacement can be atomic using
939 * a mutex.
940 */
941int
942gif_set_tunnel(ifp, src, dst)
943	struct ifnet *ifp;
944	struct sockaddr *src;
945	struct sockaddr *dst;
946{
947	INIT_VNET_GIF(ifp->if_vnet);
948	struct gif_softc *sc = ifp->if_softc;
949	struct gif_softc *sc2;
950	struct sockaddr *osrc, *odst, *sa;
951	int error = 0;
952
953	mtx_lock(&gif_mtx);
954	LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
955		if (sc2 == sc)
956			continue;
957		if (!sc2->gif_pdst || !sc2->gif_psrc)
958			continue;
959		if (sc2->gif_pdst->sa_family != dst->sa_family ||
960		    sc2->gif_pdst->sa_len != dst->sa_len ||
961		    sc2->gif_psrc->sa_family != src->sa_family ||
962		    sc2->gif_psrc->sa_len != src->sa_len)
963			continue;
964
965		/*
966		 * Disallow parallel tunnels unless instructed
967		 * otherwise.
968		 */
969		if (!V_parallel_tunnels &&
970		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
971		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
972			error = EADDRNOTAVAIL;
973			mtx_unlock(&gif_mtx);
974			goto bad;
975		}
976
977		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
978	}
979	mtx_unlock(&gif_mtx);
980
981	/* XXX we can detach from both, but be polite just in case */
982	if (sc->gif_psrc)
983		switch (sc->gif_psrc->sa_family) {
984#ifdef INET
985		case AF_INET:
986			(void)in_gif_detach(sc);
987			break;
988#endif
989#ifdef INET6
990		case AF_INET6:
991			(void)in6_gif_detach(sc);
992			break;
993#endif
994		}
995
996	osrc = sc->gif_psrc;
997	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
998	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
999	sc->gif_psrc = sa;
1000
1001	odst = sc->gif_pdst;
1002	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
1003	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
1004	sc->gif_pdst = sa;
1005
1006	switch (sc->gif_psrc->sa_family) {
1007#ifdef INET
1008	case AF_INET:
1009		error = in_gif_attach(sc);
1010		break;
1011#endif
1012#ifdef INET6
1013	case AF_INET6:
1014		/*
1015		 * Check validity of the scope zone ID of the addresses, and
1016		 * convert it into the kernel internal form if necessary.
1017		 */
1018		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
1019		if (error != 0)
1020			break;
1021		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
1022		if (error != 0)
1023			break;
1024		error = in6_gif_attach(sc);
1025		break;
1026#endif
1027	}
1028	if (error) {
1029		/* rollback */
1030		free((caddr_t)sc->gif_psrc, M_IFADDR);
1031		free((caddr_t)sc->gif_pdst, M_IFADDR);
1032		sc->gif_psrc = osrc;
1033		sc->gif_pdst = odst;
1034		goto bad;
1035	}
1036
1037	if (osrc)
1038		free((caddr_t)osrc, M_IFADDR);
1039	if (odst)
1040		free((caddr_t)odst, M_IFADDR);
1041
1042 bad:
1043	if (sc->gif_psrc && sc->gif_pdst)
1044		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1045	else
1046		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1047
1048	return error;
1049}
1050
1051void
1052gif_delete_tunnel(ifp)
1053	struct ifnet *ifp;
1054{
1055	struct gif_softc *sc = ifp->if_softc;
1056
1057	if (sc->gif_psrc) {
1058		free((caddr_t)sc->gif_psrc, M_IFADDR);
1059		sc->gif_psrc = NULL;
1060	}
1061	if (sc->gif_pdst) {
1062		free((caddr_t)sc->gif_pdst, M_IFADDR);
1063		sc->gif_pdst = NULL;
1064	}
1065	/* it is safe to detach from both */
1066#ifdef INET
1067	(void)in_gif_detach(sc);
1068#endif
1069#ifdef INET6
1070	(void)in6_gif_detach(sc);
1071#endif
1072	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1073}
1074