if_gif.c revision 249925
1/*	$FreeBSD: head/sys/net/if_gif.c 249925 2013-04-26 12:50:32Z glebius $	*/
2/*	$KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $	*/
3
4/*-
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#include "opt_inet.h"
34#include "opt_inet6.h"
35
36#include <sys/param.h>
37#include <sys/systm.h>
38#include <sys/jail.h>
39#include <sys/kernel.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/module.h>
43#include <sys/socket.h>
44#include <sys/sockio.h>
45#include <sys/errno.h>
46#include <sys/time.h>
47#include <sys/sysctl.h>
48#include <sys/syslog.h>
49#include <sys/priv.h>
50#include <sys/proc.h>
51#include <sys/protosw.h>
52#include <sys/conf.h>
53#include <machine/cpu.h>
54
55#include <net/if.h>
56#include <net/if_clone.h>
57#include <net/if_types.h>
58#include <net/netisr.h>
59#include <net/route.h>
60#include <net/bpf.h>
61#include <net/vnet.h>
62
63#include <netinet/in.h>
64#include <netinet/in_systm.h>
65#include <netinet/ip.h>
66#ifdef	INET
67#include <netinet/in_var.h>
68#include <netinet/in_gif.h>
69#include <netinet/ip_var.h>
70#endif	/* INET */
71
72#ifdef INET6
73#ifndef INET
74#include <netinet/in.h>
75#endif
76#include <netinet6/in6_var.h>
77#include <netinet/ip6.h>
78#include <netinet6/ip6_var.h>
79#include <netinet6/scope6_var.h>
80#include <netinet6/in6_gif.h>
81#include <netinet6/ip6protosw.h>
82#endif /* INET6 */
83
84#include <netinet/ip_encap.h>
85#include <net/ethernet.h>
86#include <net/if_bridgevar.h>
87#include <net/if_gif.h>
88
89#include <security/mac/mac_framework.h>
90
91static const char gifname[] = "gif";
92
93/*
94 * gif_mtx protects the global gif_softc_list.
95 */
96static struct mtx gif_mtx;
97static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
98static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
99#define	V_gif_softc_list	VNET(gif_softc_list)
100
101void	(*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
102void	(*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
103void	(*ng_gif_attach_p)(struct ifnet *ifp);
104void	(*ng_gif_detach_p)(struct ifnet *ifp);
105
106static void	gif_start(struct ifnet *);
107static int	gif_clone_create(struct if_clone *, int, caddr_t);
108static void	gif_clone_destroy(struct ifnet *);
109static struct if_clone *gif_cloner;
110
111static int gifmodevent(module_t, int, void *);
112
113SYSCTL_DECL(_net_link);
114static SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
115    "Generic Tunnel Interface");
116#ifndef MAX_GIF_NEST
117/*
118 * This macro controls the default upper limitation on nesting of gif tunnels.
119 * Since, setting a large value to this macro with a careless configuration
120 * may introduce system crash, we don't allow any nestings by default.
121 * If you need to configure nested gif tunnels, you can define this macro
122 * in your kernel configuration file.  However, if you do so, please be
123 * careful to configure the tunnels so that it won't make a loop.
124 */
125#define MAX_GIF_NEST 1
126#endif
127static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST;
128#define	V_max_gif_nesting	VNET(max_gif_nesting)
129SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
130    &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
131
132/*
133 * By default, we disallow creation of multiple tunnels between the same
134 * pair of addresses.  Some applications require this functionality so
135 * we allow control over this check here.
136 */
137#ifdef XBONEHACK
138static VNET_DEFINE(int, parallel_tunnels) = 1;
139#else
140static VNET_DEFINE(int, parallel_tunnels) = 0;
141#endif
142#define	V_parallel_tunnels	VNET(parallel_tunnels)
143SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
144    &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
145
146/* copy from src/sys/net/if_ethersubr.c */
147static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
148			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
149#ifndef ETHER_IS_BROADCAST
150#define ETHER_IS_BROADCAST(addr) \
151	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
152#endif
153
154static int
155gif_clone_create(ifc, unit, params)
156	struct if_clone *ifc;
157	int unit;
158	caddr_t params;
159{
160	struct gif_softc *sc;
161
162	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
163	sc->gif_fibnum = curthread->td_proc->p_fibnum;
164	GIF2IFP(sc) = if_alloc(IFT_GIF);
165	if (GIF2IFP(sc) == NULL) {
166		free(sc, M_GIF);
167		return (ENOSPC);
168	}
169
170	GIF_LOCK_INIT(sc);
171
172	GIF2IFP(sc)->if_softc = sc;
173	if_initname(GIF2IFP(sc), gifname, unit);
174
175	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
176	sc->gif_options = GIF_ACCEPT_REVETHIP;
177
178	GIF2IFP(sc)->if_addrlen = 0;
179	GIF2IFP(sc)->if_mtu    = GIF_MTU;
180	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
181#if 0
182	/* turn off ingress filter */
183	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
184#endif
185	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
186	GIF2IFP(sc)->if_start  = gif_start;
187	GIF2IFP(sc)->if_output = gif_output;
188	GIF2IFP(sc)->if_snd.ifq_maxlen = ifqmaxlen;
189	if_attach(GIF2IFP(sc));
190	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
191	if (ng_gif_attach_p != NULL)
192		(*ng_gif_attach_p)(GIF2IFP(sc));
193
194	mtx_lock(&gif_mtx);
195	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
196	mtx_unlock(&gif_mtx);
197
198	return (0);
199}
200
201static void
202gif_clone_destroy(ifp)
203	struct ifnet *ifp;
204{
205#if defined(INET) || defined(INET6)
206	int err;
207#endif
208	struct gif_softc *sc = ifp->if_softc;
209
210	mtx_lock(&gif_mtx);
211	LIST_REMOVE(sc, gif_list);
212	mtx_unlock(&gif_mtx);
213
214	gif_delete_tunnel(ifp);
215#ifdef INET6
216	if (sc->encap_cookie6 != NULL) {
217		err = encap_detach(sc->encap_cookie6);
218		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
219	}
220#endif
221#ifdef INET
222	if (sc->encap_cookie4 != NULL) {
223		err = encap_detach(sc->encap_cookie4);
224		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
225	}
226#endif
227
228	if (ng_gif_detach_p != NULL)
229		(*ng_gif_detach_p)(ifp);
230	bpfdetach(ifp);
231	if_detach(ifp);
232	if_free(ifp);
233
234	GIF_LOCK_DESTROY(sc);
235
236	free(sc, M_GIF);
237}
238
239static void
240vnet_gif_init(const void *unused __unused)
241{
242
243	LIST_INIT(&V_gif_softc_list);
244}
245VNET_SYSINIT(vnet_gif_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_gif_init,
246    NULL);
247
248static int
249gifmodevent(mod, type, data)
250	module_t mod;
251	int type;
252	void *data;
253{
254
255	switch (type) {
256	case MOD_LOAD:
257		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
258		gif_cloner = if_clone_simple(gifname, gif_clone_create,
259		    gif_clone_destroy, 0);
260		break;
261
262	case MOD_UNLOAD:
263		if_clone_detach(gif_cloner);
264		mtx_destroy(&gif_mtx);
265		break;
266	default:
267		return EOPNOTSUPP;
268	}
269	return 0;
270}
271
272static moduledata_t gif_mod = {
273	"if_gif",
274	gifmodevent,
275	0
276};
277
278DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
279MODULE_VERSION(if_gif, 1);
280
281int
282gif_encapcheck(m, off, proto, arg)
283	const struct mbuf *m;
284	int off;
285	int proto;
286	void *arg;
287{
288	struct ip ip;
289	struct gif_softc *sc;
290
291	sc = (struct gif_softc *)arg;
292	if (sc == NULL)
293		return 0;
294
295	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
296		return 0;
297
298	/* no physical address */
299	if (!sc->gif_psrc || !sc->gif_pdst)
300		return 0;
301
302	switch (proto) {
303#ifdef INET
304	case IPPROTO_IPV4:
305		break;
306#endif
307#ifdef INET6
308	case IPPROTO_IPV6:
309		break;
310#endif
311	case IPPROTO_ETHERIP:
312		break;
313
314	default:
315		return 0;
316	}
317
318	/* Bail on short packets */
319	if (m->m_pkthdr.len < sizeof(ip))
320		return 0;
321
322	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
323
324	switch (ip.ip_v) {
325#ifdef INET
326	case 4:
327		if (sc->gif_psrc->sa_family != AF_INET ||
328		    sc->gif_pdst->sa_family != AF_INET)
329			return 0;
330		return gif_encapcheck4(m, off, proto, arg);
331#endif
332#ifdef INET6
333	case 6:
334		if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
335			return 0;
336		if (sc->gif_psrc->sa_family != AF_INET6 ||
337		    sc->gif_pdst->sa_family != AF_INET6)
338			return 0;
339		return gif_encapcheck6(m, off, proto, arg);
340#endif
341	default:
342		return 0;
343	}
344}
345#ifdef INET
346#define GIF_HDR_LEN (ETHER_HDR_LEN + sizeof (struct ip))
347#endif
348#ifdef INET6
349#define GIF_HDR_LEN6 (ETHER_HDR_LEN + sizeof (struct ip6_hdr))
350#endif
351
352static void
353gif_start(struct ifnet *ifp)
354{
355	struct gif_softc *sc;
356	struct mbuf *m;
357	uint32_t af;
358	int error = 0;
359
360	sc = ifp->if_softc;
361	GIF_LOCK(sc);
362	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
363	while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
364
365		IFQ_DRV_DEQUEUE(&ifp->if_snd, m);
366		if (m == 0)
367			break;
368
369#ifdef ALTQ
370		/* Take out those altq bytes we add in gif_output  */
371#ifdef INET
372		if (sc->gif_psrc->sa_family == AF_INET)
373			m->m_pkthdr.len -= GIF_HDR_LEN;
374#endif
375#ifdef INET6
376		if (sc->gif_psrc->sa_family == AF_INET6)
377		    m->m_pkthdr.len -= GIF_HDR_LEN6;
378#endif
379#endif
380		/*
381		 * Now pull back the af that we
382		 * stashed in the csum_data.
383		 */
384		af = m->m_pkthdr.csum_data;
385
386		if (ifp->if_bridge)
387			af = AF_LINK;
388
389		BPF_MTAP2(ifp, &af, sizeof(af), m);
390		ifp->if_opackets++;
391
392/*              Done by IFQ_HANDOFF */
393/* 		ifp->if_obytes += m->m_pkthdr.len;*/
394		/* override to IPPROTO_ETHERIP for bridged traffic */
395
396		M_SETFIB(m, sc->gif_fibnum);
397		/* inner AF-specific encapsulation */
398		/* XXX should we check if our outer source is legal? */
399		/* dispatch to output logic based on outer AF */
400		switch (sc->gif_psrc->sa_family) {
401#ifdef INET
402		case AF_INET:
403			error = in_gif_output(ifp, af, m);
404			break;
405#endif
406#ifdef INET6
407		case AF_INET6:
408			error = in6_gif_output(ifp, af, m);
409			break;
410#endif
411		default:
412			m_freem(m);
413			error = ENETDOWN;
414		}
415		if (error)
416			ifp->if_oerrors++;
417
418	}
419	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
420	GIF_UNLOCK(sc);
421	return;
422}
423
424int
425gif_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
426	struct route *ro)
427{
428	struct gif_softc *sc = ifp->if_softc;
429	struct m_tag *mtag;
430	int error = 0;
431	int gif_called;
432	uint32_t af;
433#ifdef MAC
434	error = mac_ifnet_check_transmit(ifp, m);
435	if (error) {
436		m_freem(m);
437		goto end;
438	}
439#endif
440
441	/*
442	 * gif may cause infinite recursion calls when misconfigured.
443	 * We'll prevent this by detecting loops.
444	 *
445	 * High nesting level may cause stack exhaustion.
446	 * We'll prevent this by introducing upper limit.
447	 */
448	gif_called = 1;
449	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
450	while (mtag != NULL) {
451		if (*(struct ifnet **)(mtag + 1) == ifp) {
452			log(LOG_NOTICE,
453			    "gif_output: loop detected on %s\n",
454			    (*(struct ifnet **)(mtag + 1))->if_xname);
455			m_freem(m);
456			error = EIO;	/* is there better errno? */
457			goto end;
458		}
459		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
460		gif_called++;
461	}
462	if (gif_called > V_max_gif_nesting) {
463		log(LOG_NOTICE,
464		    "gif_output: recursively called too many times(%d)\n",
465		    gif_called);
466		m_freem(m);
467		error = EIO;	/* is there better errno? */
468		goto end;
469	}
470	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
471	    M_NOWAIT);
472	if (mtag == NULL) {
473		m_freem(m);
474		error = ENOMEM;
475		goto end;
476	}
477	*(struct ifnet **)(mtag + 1) = ifp;
478	m_tag_prepend(m, mtag);
479
480	m->m_flags &= ~(M_BCAST|M_MCAST);
481	/* BPF writes need to be handled specially. */
482	if (dst->sa_family == AF_UNSPEC)
483		bcopy(dst->sa_data, &af, sizeof(af));
484	else
485		af = dst->sa_family;
486	/*
487	 * Now save the af in the inbound pkt csum
488	 * data, this is a cheat since we are using
489	 * the inbound csum_data field to carry the
490	 * af over to the gif_start() routine, avoiding
491	 * using yet another mtag.
492	 */
493	m->m_pkthdr.csum_data = af;
494	if (!(ifp->if_flags & IFF_UP) ||
495	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
496		m_freem(m);
497		error = ENETDOWN;
498		goto end;
499	}
500#ifdef ALTQ
501	/*
502	 * Make altq aware of the bytes we will add
503	 * when we actually send it.
504	 */
505#ifdef INET
506	if (sc->gif_psrc->sa_family == AF_INET)
507		m->m_pkthdr.len += GIF_HDR_LEN;
508#endif
509#ifdef INET6
510	if (sc->gif_psrc->sa_family == AF_INET6)
511		m->m_pkthdr.len += GIF_HDR_LEN6;
512#endif
513#endif
514	/*
515	 * Queue message on interface, update output statistics if
516	 * successful, and start output if interface not yet active.
517	 */
518	IFQ_HANDOFF(ifp, m, error);
519  end:
520	if (error)
521		ifp->if_oerrors++;
522	return (error);
523}
524
525void
526gif_input(m, af, ifp)
527	struct mbuf *m;
528	int af;
529	struct ifnet *ifp;
530{
531	int isr, n;
532	struct gif_softc *sc;
533	struct etherip_header *eip;
534	struct ether_header *eh;
535	struct ifnet *oldifp;
536
537	if (ifp == NULL) {
538		/* just in case */
539		m_freem(m);
540		return;
541	}
542	sc = ifp->if_softc;
543	m->m_pkthdr.rcvif = ifp;
544
545#ifdef MAC
546	mac_ifnet_create_mbuf(ifp, m);
547#endif
548
549	if (bpf_peers_present(ifp->if_bpf)) {
550		u_int32_t af1 = af;
551		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
552	}
553
554	if (ng_gif_input_p != NULL) {
555		(*ng_gif_input_p)(ifp, &m, af);
556		if (m == NULL)
557			return;
558	}
559
560	/*
561	 * Put the packet to the network layer input queue according to the
562	 * specified address family.
563	 * Note: older versions of gif_input directly called network layer
564	 * input functions, e.g. ip6_input, here.  We changed the policy to
565	 * prevent too many recursive calls of such input functions, which
566	 * might cause kernel panic.  But the change may introduce another
567	 * problem; if the input queue is full, packets are discarded.
568	 * The kernel stack overflow really happened, and we believed
569	 * queue-full rarely occurs, so we changed the policy.
570	 */
571	switch (af) {
572#ifdef INET
573	case AF_INET:
574		isr = NETISR_IP;
575		break;
576#endif
577#ifdef INET6
578	case AF_INET6:
579		isr = NETISR_IPV6;
580		break;
581#endif
582	case AF_LINK:
583		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
584		if (n > m->m_len) {
585			m = m_pullup(m, n);
586			if (m == NULL) {
587				ifp->if_ierrors++;
588				return;
589			}
590		}
591
592		eip = mtod(m, struct etherip_header *);
593		/*
594		 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
595		 * accepts an EtherIP packet with revered version field in
596		 * the header.  This is a knob for backward compatibility
597		 * with FreeBSD 7.2R or prior.
598		 */
599		if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
600			if (eip->eip_resvl != ETHERIP_VERSION
601			    && eip->eip_ver != ETHERIP_VERSION) {
602				/* discard unknown versions */
603				m_freem(m);
604				return;
605			}
606		} else {
607			if (eip->eip_ver != ETHERIP_VERSION) {
608				/* discard unknown versions */
609				m_freem(m);
610				return;
611			}
612		}
613		m_adj(m, sizeof(struct etherip_header));
614
615		m->m_flags &= ~(M_BCAST|M_MCAST);
616		m->m_pkthdr.rcvif = ifp;
617
618		if (ifp->if_bridge) {
619			oldifp = ifp;
620			eh = mtod(m, struct ether_header *);
621			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
622				if (ETHER_IS_BROADCAST(eh->ether_dhost))
623					m->m_flags |= M_BCAST;
624				else
625					m->m_flags |= M_MCAST;
626				ifp->if_imcasts++;
627			}
628			BRIDGE_INPUT(ifp, m);
629
630			if (m != NULL && ifp != oldifp) {
631				/*
632				 * The bridge gave us back itself or one of the
633				 * members for which the frame is addressed.
634				 */
635				ether_demux(ifp, m);
636				return;
637			}
638		}
639		if (m != NULL)
640			m_freem(m);
641		return;
642
643	default:
644		if (ng_gif_input_orphan_p != NULL)
645			(*ng_gif_input_orphan_p)(ifp, m, af);
646		else
647			m_freem(m);
648		return;
649	}
650
651	ifp->if_ipackets++;
652	ifp->if_ibytes += m->m_pkthdr.len;
653	M_SETFIB(m, ifp->if_fib);
654	netisr_dispatch(isr, m);
655}
656
657/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
658int
659gif_ioctl(ifp, cmd, data)
660	struct ifnet *ifp;
661	u_long cmd;
662	caddr_t data;
663{
664	struct gif_softc *sc  = ifp->if_softc;
665	struct ifreq     *ifr = (struct ifreq*)data;
666	int error = 0, size;
667	u_int	options;
668	struct sockaddr *dst, *src;
669#ifdef	SIOCSIFMTU /* xxx */
670	u_long mtu;
671#endif
672
673	switch (cmd) {
674	case SIOCSIFADDR:
675		ifp->if_flags |= IFF_UP;
676		break;
677
678	case SIOCSIFDSTADDR:
679		break;
680
681	case SIOCADDMULTI:
682	case SIOCDELMULTI:
683		break;
684
685#ifdef	SIOCSIFMTU /* xxx */
686	case SIOCGIFMTU:
687		break;
688
689	case SIOCSIFMTU:
690		mtu = ifr->ifr_mtu;
691		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
692			return (EINVAL);
693		ifp->if_mtu = mtu;
694		break;
695#endif /* SIOCSIFMTU */
696
697#ifdef INET
698	case SIOCSIFPHYADDR:
699#endif
700#ifdef INET6
701	case SIOCSIFPHYADDR_IN6:
702#endif /* INET6 */
703	case SIOCSLIFPHYADDR:
704		switch (cmd) {
705#ifdef INET
706		case SIOCSIFPHYADDR:
707			src = (struct sockaddr *)
708				&(((struct in_aliasreq *)data)->ifra_addr);
709			dst = (struct sockaddr *)
710				&(((struct in_aliasreq *)data)->ifra_dstaddr);
711			break;
712#endif
713#ifdef INET6
714		case SIOCSIFPHYADDR_IN6:
715			src = (struct sockaddr *)
716				&(((struct in6_aliasreq *)data)->ifra_addr);
717			dst = (struct sockaddr *)
718				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
719			break;
720#endif
721		case SIOCSLIFPHYADDR:
722			src = (struct sockaddr *)
723				&(((struct if_laddrreq *)data)->addr);
724			dst = (struct sockaddr *)
725				&(((struct if_laddrreq *)data)->dstaddr);
726			break;
727		default:
728			return EINVAL;
729		}
730
731		/* sa_family must be equal */
732		if (src->sa_family != dst->sa_family)
733			return EINVAL;
734
735		/* validate sa_len */
736		switch (src->sa_family) {
737#ifdef INET
738		case AF_INET:
739			if (src->sa_len != sizeof(struct sockaddr_in))
740				return EINVAL;
741			break;
742#endif
743#ifdef INET6
744		case AF_INET6:
745			if (src->sa_len != sizeof(struct sockaddr_in6))
746				return EINVAL;
747			break;
748#endif
749		default:
750			return EAFNOSUPPORT;
751		}
752		switch (dst->sa_family) {
753#ifdef INET
754		case AF_INET:
755			if (dst->sa_len != sizeof(struct sockaddr_in))
756				return EINVAL;
757			break;
758#endif
759#ifdef INET6
760		case AF_INET6:
761			if (dst->sa_len != sizeof(struct sockaddr_in6))
762				return EINVAL;
763			break;
764#endif
765		default:
766			return EAFNOSUPPORT;
767		}
768
769		/* check sa_family looks sane for the cmd */
770		switch (cmd) {
771		case SIOCSIFPHYADDR:
772			if (src->sa_family == AF_INET)
773				break;
774			return EAFNOSUPPORT;
775#ifdef INET6
776		case SIOCSIFPHYADDR_IN6:
777			if (src->sa_family == AF_INET6)
778				break;
779			return EAFNOSUPPORT;
780#endif /* INET6 */
781		case SIOCSLIFPHYADDR:
782			/* checks done in the above */
783			break;
784		}
785
786		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
787		break;
788
789#ifdef SIOCDIFPHYADDR
790	case SIOCDIFPHYADDR:
791		gif_delete_tunnel(GIF2IFP(sc));
792		break;
793#endif
794
795	case SIOCGIFPSRCADDR:
796#ifdef INET6
797	case SIOCGIFPSRCADDR_IN6:
798#endif /* INET6 */
799		if (sc->gif_psrc == NULL) {
800			error = EADDRNOTAVAIL;
801			goto bad;
802		}
803		src = sc->gif_psrc;
804		switch (cmd) {
805#ifdef INET
806		case SIOCGIFPSRCADDR:
807			dst = &ifr->ifr_addr;
808			size = sizeof(ifr->ifr_addr);
809			break;
810#endif /* INET */
811#ifdef INET6
812		case SIOCGIFPSRCADDR_IN6:
813			dst = (struct sockaddr *)
814				&(((struct in6_ifreq *)data)->ifr_addr);
815			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
816			break;
817#endif /* INET6 */
818		default:
819			error = EADDRNOTAVAIL;
820			goto bad;
821		}
822		if (src->sa_len > size)
823			return EINVAL;
824		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
825#ifdef INET6
826		if (dst->sa_family == AF_INET6) {
827			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
828			if (error != 0)
829				return (error);
830		}
831#endif
832		break;
833
834	case SIOCGIFPDSTADDR:
835#ifdef INET6
836	case SIOCGIFPDSTADDR_IN6:
837#endif /* INET6 */
838		if (sc->gif_pdst == NULL) {
839			error = EADDRNOTAVAIL;
840			goto bad;
841		}
842		src = sc->gif_pdst;
843		switch (cmd) {
844#ifdef INET
845		case SIOCGIFPDSTADDR:
846			dst = &ifr->ifr_addr;
847			size = sizeof(ifr->ifr_addr);
848			break;
849#endif /* INET */
850#ifdef INET6
851		case SIOCGIFPDSTADDR_IN6:
852			dst = (struct sockaddr *)
853				&(((struct in6_ifreq *)data)->ifr_addr);
854			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
855			break;
856#endif /* INET6 */
857		default:
858			error = EADDRNOTAVAIL;
859			goto bad;
860		}
861		if (src->sa_len > size)
862			return EINVAL;
863		error = prison_if(curthread->td_ucred, src);
864		if (error != 0)
865			return (error);
866		error = prison_if(curthread->td_ucred, dst);
867		if (error != 0)
868			return (error);
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		error = copyin(ifr->ifr_data, &options, sizeof(options));
918		if (error)
919			break;
920		if (options & ~GIF_OPTMASK)
921			error = EINVAL;
922		else
923			sc->gif_options = options;
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	struct gif_softc *sc = ifp->if_softc;
948	struct gif_softc *sc2;
949	struct sockaddr *osrc, *odst, *sa;
950	int error = 0;
951
952	mtx_lock(&gif_mtx);
953	LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
954		if (sc2 == sc)
955			continue;
956		if (!sc2->gif_pdst || !sc2->gif_psrc)
957			continue;
958		if (sc2->gif_pdst->sa_family != dst->sa_family ||
959		    sc2->gif_pdst->sa_len != dst->sa_len ||
960		    sc2->gif_psrc->sa_family != src->sa_family ||
961		    sc2->gif_psrc->sa_len != src->sa_len)
962			continue;
963
964		/*
965		 * Disallow parallel tunnels unless instructed
966		 * otherwise.
967		 */
968		if (!V_parallel_tunnels &&
969		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
970		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
971			error = EADDRNOTAVAIL;
972			mtx_unlock(&gif_mtx);
973			goto bad;
974		}
975
976		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
977	}
978	mtx_unlock(&gif_mtx);
979
980	/* XXX we can detach from both, but be polite just in case */
981	if (sc->gif_psrc)
982		switch (sc->gif_psrc->sa_family) {
983#ifdef INET
984		case AF_INET:
985			(void)in_gif_detach(sc);
986			break;
987#endif
988#ifdef INET6
989		case AF_INET6:
990			(void)in6_gif_detach(sc);
991			break;
992#endif
993		}
994
995	osrc = sc->gif_psrc;
996	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
997	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
998	sc->gif_psrc = sa;
999
1000	odst = sc->gif_pdst;
1001	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
1002	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
1003	sc->gif_pdst = sa;
1004
1005	switch (sc->gif_psrc->sa_family) {
1006#ifdef INET
1007	case AF_INET:
1008		error = in_gif_attach(sc);
1009		break;
1010#endif
1011#ifdef INET6
1012	case AF_INET6:
1013		/*
1014		 * Check validity of the scope zone ID of the addresses, and
1015		 * convert it into the kernel internal form if necessary.
1016		 */
1017		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
1018		if (error != 0)
1019			break;
1020		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
1021		if (error != 0)
1022			break;
1023		error = in6_gif_attach(sc);
1024		break;
1025#endif
1026	}
1027	if (error) {
1028		/* rollback */
1029		free((caddr_t)sc->gif_psrc, M_IFADDR);
1030		free((caddr_t)sc->gif_pdst, M_IFADDR);
1031		sc->gif_psrc = osrc;
1032		sc->gif_pdst = odst;
1033		goto bad;
1034	}
1035
1036	if (osrc)
1037		free((caddr_t)osrc, M_IFADDR);
1038	if (odst)
1039		free((caddr_t)odst, M_IFADDR);
1040
1041 bad:
1042	if (sc->gif_psrc && sc->gif_pdst)
1043		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1044	else
1045		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1046
1047	return error;
1048}
1049
1050void
1051gif_delete_tunnel(ifp)
1052	struct ifnet *ifp;
1053{
1054	struct gif_softc *sc = ifp->if_softc;
1055
1056	if (sc->gif_psrc) {
1057		free((caddr_t)sc->gif_psrc, M_IFADDR);
1058		sc->gif_psrc = NULL;
1059	}
1060	if (sc->gif_pdst) {
1061		free((caddr_t)sc->gif_pdst, M_IFADDR);
1062		sc->gif_pdst = NULL;
1063	}
1064	/* it is safe to detach from both */
1065#ifdef INET
1066	(void)in_gif_detach(sc);
1067#endif
1068#ifdef INET6
1069	(void)in6_gif_detach(sc);
1070#endif
1071	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1072}
1073