if_gif.c revision 227309
1/*	$FreeBSD: head/sys/net/if_gif.c 227309 2011-11-07 15:43:11Z ed $	*/
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
91#define 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 *);
109
110IFC_SIMPLE_DECLARE(gif, 0);
111
112static int gifmodevent(module_t, int, void *);
113
114SYSCTL_DECL(_net_link);
115static SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
116    "Generic Tunnel Interface");
117#ifndef MAX_GIF_NEST
118/*
119 * This macro controls the default upper limitation on nesting of gif tunnels.
120 * Since, setting a large value to this macro with a careless configuration
121 * may introduce system crash, we don't allow any nestings by default.
122 * If you need to configure nested gif tunnels, you can define this macro
123 * in your kernel configuration file.  However, if you do so, please be
124 * careful to configure the tunnels so that it won't make a loop.
125 */
126#define MAX_GIF_NEST 1
127#endif
128static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST;
129#define	V_max_gif_nesting	VNET(max_gif_nesting)
130SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
131    &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
132
133/*
134 * By default, we disallow creation of multiple tunnels between the same
135 * pair of addresses.  Some applications require this functionality so
136 * we allow control over this check here.
137 */
138#ifdef XBONEHACK
139static VNET_DEFINE(int, parallel_tunnels) = 1;
140#else
141static VNET_DEFINE(int, parallel_tunnels) = 0;
142#endif
143#define	V_parallel_tunnels	VNET(parallel_tunnels)
144SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
145    &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
146
147/* copy from src/sys/net/if_ethersubr.c */
148static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
149			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
150#ifndef ETHER_IS_BROADCAST
151#define ETHER_IS_BROADCAST(addr) \
152	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
153#endif
154
155static int
156gif_clone_create(ifc, unit, params)
157	struct if_clone *ifc;
158	int unit;
159	caddr_t params;
160{
161	struct gif_softc *sc;
162
163	sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
164	sc->gif_fibnum = curthread->td_proc->p_fibnum;
165	GIF2IFP(sc) = if_alloc(IFT_GIF);
166	if (GIF2IFP(sc) == NULL) {
167		free(sc, M_GIF);
168		return (ENOSPC);
169	}
170
171	GIF_LOCK_INIT(sc);
172
173	GIF2IFP(sc)->if_softc = sc;
174	if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
175
176	sc->encap_cookie4 = sc->encap_cookie6 = NULL;
177	sc->gif_options = GIF_ACCEPT_REVETHIP;
178
179	GIF2IFP(sc)->if_addrlen = 0;
180	GIF2IFP(sc)->if_mtu    = GIF_MTU;
181	GIF2IFP(sc)->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
182#if 0
183	/* turn off ingress filter */
184	GIF2IFP(sc)->if_flags  |= IFF_LINK2;
185#endif
186	GIF2IFP(sc)->if_ioctl  = gif_ioctl;
187	GIF2IFP(sc)->if_start  = gif_start;
188	GIF2IFP(sc)->if_output = gif_output;
189	GIF2IFP(sc)->if_snd.ifq_maxlen = ifqmaxlen;
190	if_attach(GIF2IFP(sc));
191	bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
192	if (ng_gif_attach_p != NULL)
193		(*ng_gif_attach_p)(GIF2IFP(sc));
194
195	mtx_lock(&gif_mtx);
196	LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
197	mtx_unlock(&gif_mtx);
198
199	return (0);
200}
201
202static void
203gif_clone_destroy(ifp)
204	struct ifnet *ifp;
205{
206#if defined(INET) || defined(INET6)
207	int err;
208#endif
209	struct gif_softc *sc = ifp->if_softc;
210
211	mtx_lock(&gif_mtx);
212	LIST_REMOVE(sc, gif_list);
213	mtx_unlock(&gif_mtx);
214
215	gif_delete_tunnel(ifp);
216#ifdef INET6
217	if (sc->encap_cookie6 != NULL) {
218		err = encap_detach(sc->encap_cookie6);
219		KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
220	}
221#endif
222#ifdef INET
223	if (sc->encap_cookie4 != NULL) {
224		err = encap_detach(sc->encap_cookie4);
225		KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
226	}
227#endif
228
229	if (ng_gif_detach_p != NULL)
230		(*ng_gif_detach_p)(ifp);
231	bpfdetach(ifp);
232	if_detach(ifp);
233	if_free(ifp);
234
235	GIF_LOCK_DESTROY(sc);
236
237	free(sc, M_GIF);
238}
239
240static void
241vnet_gif_init(const void *unused __unused)
242{
243
244	LIST_INIT(&V_gif_softc_list);
245}
246VNET_SYSINIT(vnet_gif_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_gif_init,
247    NULL);
248
249static int
250gifmodevent(mod, type, data)
251	module_t mod;
252	int type;
253	void *data;
254{
255
256	switch (type) {
257	case MOD_LOAD:
258		mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
259		if_clone_attach(&gif_cloner);
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
346static void
347gif_start(struct ifnet *ifp)
348{
349	struct gif_softc *sc;
350	struct mbuf *m;
351
352	sc = ifp->if_softc;
353
354	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
355	for (;;) {
356		IFQ_DEQUEUE(&ifp->if_snd, m);
357		if (m == 0)
358			break;
359
360		gif_output(ifp, m, sc->gif_pdst, NULL);
361
362	}
363	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
364
365	return;
366}
367
368int
369gif_output(ifp, m, dst, ro)
370	struct ifnet *ifp;
371	struct mbuf *m;
372	struct sockaddr *dst;
373	struct route *ro;
374{
375	struct gif_softc *sc = ifp->if_softc;
376	struct m_tag *mtag;
377	int error = 0;
378	int gif_called;
379	u_int32_t af;
380
381#ifdef MAC
382	error = mac_ifnet_check_transmit(ifp, m);
383	if (error) {
384		m_freem(m);
385		goto end;
386	}
387#endif
388
389	/*
390	 * gif may cause infinite recursion calls when misconfigured.
391	 * We'll prevent this by detecting loops.
392	 *
393	 * High nesting level may cause stack exhaustion.
394	 * We'll prevent this by introducing upper limit.
395	 */
396	gif_called = 1;
397	mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
398	while (mtag != NULL) {
399		if (*(struct ifnet **)(mtag + 1) == ifp) {
400			log(LOG_NOTICE,
401			    "gif_output: loop detected on %s\n",
402			    (*(struct ifnet **)(mtag + 1))->if_xname);
403			m_freem(m);
404			error = EIO;	/* is there better errno? */
405			goto end;
406		}
407		mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
408		gif_called++;
409	}
410	if (gif_called > V_max_gif_nesting) {
411		log(LOG_NOTICE,
412		    "gif_output: recursively called too many times(%d)\n",
413		    gif_called);
414		m_freem(m);
415		error = EIO;	/* is there better errno? */
416		goto end;
417	}
418	mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
419	    M_NOWAIT);
420	if (mtag == NULL) {
421		m_freem(m);
422		error = ENOMEM;
423		goto end;
424	}
425	*(struct ifnet **)(mtag + 1) = ifp;
426	m_tag_prepend(m, mtag);
427
428	m->m_flags &= ~(M_BCAST|M_MCAST);
429
430	GIF_LOCK(sc);
431
432	if (!(ifp->if_flags & IFF_UP) ||
433	    sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
434		GIF_UNLOCK(sc);
435		m_freem(m);
436		error = ENETDOWN;
437		goto end;
438	}
439
440	/* BPF writes need to be handled specially. */
441	if (dst->sa_family == AF_UNSPEC) {
442		bcopy(dst->sa_data, &af, sizeof(af));
443		dst->sa_family = af;
444	}
445
446	af = dst->sa_family;
447	BPF_MTAP2(ifp, &af, sizeof(af), m);
448	ifp->if_opackets++;
449	ifp->if_obytes += m->m_pkthdr.len;
450
451	/* override to IPPROTO_ETHERIP for bridged traffic */
452	if (ifp->if_bridge)
453		af = AF_LINK;
454
455	M_SETFIB(m, sc->gif_fibnum);
456	/* inner AF-specific encapsulation */
457
458	/* XXX should we check if our outer source is legal? */
459
460	/* dispatch to output logic based on outer AF */
461	switch (sc->gif_psrc->sa_family) {
462#ifdef INET
463	case AF_INET:
464		error = in_gif_output(ifp, af, m);
465		break;
466#endif
467#ifdef INET6
468	case AF_INET6:
469		error = in6_gif_output(ifp, af, m);
470		break;
471#endif
472	default:
473		m_freem(m);
474		error = ENETDOWN;
475	}
476
477	GIF_UNLOCK(sc);
478  end:
479	if (error)
480		ifp->if_oerrors++;
481	return (error);
482}
483
484void
485gif_input(m, af, ifp)
486	struct mbuf *m;
487	int af;
488	struct ifnet *ifp;
489{
490	int isr, n;
491	struct gif_softc *sc;
492	struct etherip_header *eip;
493	struct ether_header *eh;
494	struct ifnet *oldifp;
495
496	if (ifp == NULL) {
497		/* just in case */
498		m_freem(m);
499		return;
500	}
501	sc = ifp->if_softc;
502	m->m_pkthdr.rcvif = ifp;
503
504#ifdef MAC
505	mac_ifnet_create_mbuf(ifp, m);
506#endif
507
508	if (bpf_peers_present(ifp->if_bpf)) {
509		u_int32_t af1 = af;
510		bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
511	}
512
513	if (ng_gif_input_p != NULL) {
514		(*ng_gif_input_p)(ifp, &m, af);
515		if (m == NULL)
516			return;
517	}
518
519	/*
520	 * Put the packet to the network layer input queue according to the
521	 * specified address family.
522	 * Note: older versions of gif_input directly called network layer
523	 * input functions, e.g. ip6_input, here.  We changed the policy to
524	 * prevent too many recursive calls of such input functions, which
525	 * might cause kernel panic.  But the change may introduce another
526	 * problem; if the input queue is full, packets are discarded.
527	 * The kernel stack overflow really happened, and we believed
528	 * queue-full rarely occurs, so we changed the policy.
529	 */
530	switch (af) {
531#ifdef INET
532	case AF_INET:
533		isr = NETISR_IP;
534		break;
535#endif
536#ifdef INET6
537	case AF_INET6:
538		isr = NETISR_IPV6;
539		break;
540#endif
541	case AF_LINK:
542		n = sizeof(struct etherip_header) + sizeof(struct ether_header);
543		if (n > m->m_len) {
544			m = m_pullup(m, n);
545			if (m == NULL) {
546				ifp->if_ierrors++;
547				return;
548			}
549		}
550
551		eip = mtod(m, struct etherip_header *);
552		/*
553		 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
554		 * accepts an EtherIP packet with revered version field in
555		 * the header.  This is a knob for backward compatibility
556		 * with FreeBSD 7.2R or prior.
557		 */
558		if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
559			if (eip->eip_resvl != ETHERIP_VERSION
560			    && eip->eip_ver != ETHERIP_VERSION) {
561				/* discard unknown versions */
562				m_freem(m);
563				return;
564			}
565		} else {
566			if (eip->eip_ver != ETHERIP_VERSION) {
567				/* discard unknown versions */
568				m_freem(m);
569				return;
570			}
571		}
572		m_adj(m, sizeof(struct etherip_header));
573
574		m->m_flags &= ~(M_BCAST|M_MCAST);
575		m->m_pkthdr.rcvif = ifp;
576
577		if (ifp->if_bridge) {
578			oldifp = ifp;
579			eh = mtod(m, struct ether_header *);
580			if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
581				if (ETHER_IS_BROADCAST(eh->ether_dhost))
582					m->m_flags |= M_BCAST;
583				else
584					m->m_flags |= M_MCAST;
585				ifp->if_imcasts++;
586			}
587			BRIDGE_INPUT(ifp, m);
588
589			if (m != NULL && ifp != oldifp) {
590				/*
591				 * The bridge gave us back itself or one of the
592				 * members for which the frame is addressed.
593				 */
594				ether_demux(ifp, m);
595				return;
596			}
597		}
598		if (m != NULL)
599			m_freem(m);
600		return;
601
602	default:
603		if (ng_gif_input_orphan_p != NULL)
604			(*ng_gif_input_orphan_p)(ifp, m, af);
605		else
606			m_freem(m);
607		return;
608	}
609
610	ifp->if_ipackets++;
611	ifp->if_ibytes += m->m_pkthdr.len;
612	M_SETFIB(m, ifp->if_fib);
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	u_int	options;
627	struct sockaddr *dst, *src;
628#ifdef	SIOCSIFMTU /* xxx */
629	u_long mtu;
630#endif
631
632	switch (cmd) {
633	case SIOCSIFADDR:
634		ifp->if_flags |= IFF_UP;
635		break;
636
637	case SIOCSIFDSTADDR:
638		break;
639
640	case SIOCADDMULTI:
641	case SIOCDELMULTI:
642		break;
643
644#ifdef	SIOCSIFMTU /* xxx */
645	case SIOCGIFMTU:
646		break;
647
648	case SIOCSIFMTU:
649		mtu = ifr->ifr_mtu;
650		if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
651			return (EINVAL);
652		ifp->if_mtu = mtu;
653		break;
654#endif /* SIOCSIFMTU */
655
656#ifdef INET
657	case SIOCSIFPHYADDR:
658#endif
659#ifdef INET6
660	case SIOCSIFPHYADDR_IN6:
661#endif /* INET6 */
662	case SIOCSLIFPHYADDR:
663		switch (cmd) {
664#ifdef INET
665		case SIOCSIFPHYADDR:
666			src = (struct sockaddr *)
667				&(((struct in_aliasreq *)data)->ifra_addr);
668			dst = (struct sockaddr *)
669				&(((struct in_aliasreq *)data)->ifra_dstaddr);
670			break;
671#endif
672#ifdef INET6
673		case SIOCSIFPHYADDR_IN6:
674			src = (struct sockaddr *)
675				&(((struct in6_aliasreq *)data)->ifra_addr);
676			dst = (struct sockaddr *)
677				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
678			break;
679#endif
680		case SIOCSLIFPHYADDR:
681			src = (struct sockaddr *)
682				&(((struct if_laddrreq *)data)->addr);
683			dst = (struct sockaddr *)
684				&(((struct if_laddrreq *)data)->dstaddr);
685			break;
686		default:
687			return EINVAL;
688		}
689
690		/* sa_family must be equal */
691		if (src->sa_family != dst->sa_family)
692			return EINVAL;
693
694		/* validate sa_len */
695		switch (src->sa_family) {
696#ifdef INET
697		case AF_INET:
698			if (src->sa_len != sizeof(struct sockaddr_in))
699				return EINVAL;
700			break;
701#endif
702#ifdef INET6
703		case AF_INET6:
704			if (src->sa_len != sizeof(struct sockaddr_in6))
705				return EINVAL;
706			break;
707#endif
708		default:
709			return EAFNOSUPPORT;
710		}
711		switch (dst->sa_family) {
712#ifdef INET
713		case AF_INET:
714			if (dst->sa_len != sizeof(struct sockaddr_in))
715				return EINVAL;
716			break;
717#endif
718#ifdef INET6
719		case AF_INET6:
720			if (dst->sa_len != sizeof(struct sockaddr_in6))
721				return EINVAL;
722			break;
723#endif
724		default:
725			return EAFNOSUPPORT;
726		}
727
728		/* check sa_family looks sane for the cmd */
729		switch (cmd) {
730		case SIOCSIFPHYADDR:
731			if (src->sa_family == AF_INET)
732				break;
733			return EAFNOSUPPORT;
734#ifdef INET6
735		case SIOCSIFPHYADDR_IN6:
736			if (src->sa_family == AF_INET6)
737				break;
738			return EAFNOSUPPORT;
739#endif /* INET6 */
740		case SIOCSLIFPHYADDR:
741			/* checks done in the above */
742			break;
743		}
744
745		error = gif_set_tunnel(GIF2IFP(sc), src, dst);
746		break;
747
748#ifdef SIOCDIFPHYADDR
749	case SIOCDIFPHYADDR:
750		gif_delete_tunnel(GIF2IFP(sc));
751		break;
752#endif
753
754	case SIOCGIFPSRCADDR:
755#ifdef INET6
756	case SIOCGIFPSRCADDR_IN6:
757#endif /* INET6 */
758		if (sc->gif_psrc == NULL) {
759			error = EADDRNOTAVAIL;
760			goto bad;
761		}
762		src = sc->gif_psrc;
763		switch (cmd) {
764#ifdef INET
765		case SIOCGIFPSRCADDR:
766			dst = &ifr->ifr_addr;
767			size = sizeof(ifr->ifr_addr);
768			break;
769#endif /* INET */
770#ifdef INET6
771		case SIOCGIFPSRCADDR_IN6:
772			dst = (struct sockaddr *)
773				&(((struct in6_ifreq *)data)->ifr_addr);
774			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
775			break;
776#endif /* INET6 */
777		default:
778			error = EADDRNOTAVAIL;
779			goto bad;
780		}
781		if (src->sa_len > size)
782			return EINVAL;
783		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
784#ifdef INET6
785		if (dst->sa_family == AF_INET6) {
786			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
787			if (error != 0)
788				return (error);
789		}
790#endif
791		break;
792
793	case SIOCGIFPDSTADDR:
794#ifdef INET6
795	case SIOCGIFPDSTADDR_IN6:
796#endif /* INET6 */
797		if (sc->gif_pdst == NULL) {
798			error = EADDRNOTAVAIL;
799			goto bad;
800		}
801		src = sc->gif_pdst;
802		switch (cmd) {
803#ifdef INET
804		case SIOCGIFPDSTADDR:
805			dst = &ifr->ifr_addr;
806			size = sizeof(ifr->ifr_addr);
807			break;
808#endif /* INET */
809#ifdef INET6
810		case SIOCGIFPDSTADDR_IN6:
811			dst = (struct sockaddr *)
812				&(((struct in6_ifreq *)data)->ifr_addr);
813			size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
814			break;
815#endif /* INET6 */
816		default:
817			error = EADDRNOTAVAIL;
818			goto bad;
819		}
820		if (src->sa_len > size)
821			return EINVAL;
822		error = prison_if(curthread->td_ucred, src);
823		if (error != 0)
824			return (error);
825		error = prison_if(curthread->td_ucred, dst);
826		if (error != 0)
827			return (error);
828		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
829#ifdef INET6
830		if (dst->sa_family == AF_INET6) {
831			error = sa6_recoverscope((struct sockaddr_in6 *)dst);
832			if (error != 0)
833				return (error);
834		}
835#endif
836		break;
837
838	case SIOCGLIFPHYADDR:
839		if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
840			error = EADDRNOTAVAIL;
841			goto bad;
842		}
843
844		/* copy src */
845		src = sc->gif_psrc;
846		dst = (struct sockaddr *)
847			&(((struct if_laddrreq *)data)->addr);
848		size = sizeof(((struct if_laddrreq *)data)->addr);
849		if (src->sa_len > size)
850			return EINVAL;
851		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
852
853		/* copy dst */
854		src = sc->gif_pdst;
855		dst = (struct sockaddr *)
856			&(((struct if_laddrreq *)data)->dstaddr);
857		size = sizeof(((struct if_laddrreq *)data)->dstaddr);
858		if (src->sa_len > size)
859			return EINVAL;
860		bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
861		break;
862
863	case SIOCSIFFLAGS:
864		/* if_ioctl() takes care of it */
865		break;
866
867	case GIFGOPTS:
868		options = sc->gif_options;
869		error = copyout(&options, ifr->ifr_data,
870				sizeof(options));
871		break;
872
873	case GIFSOPTS:
874		if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
875			break;
876		error = copyin(ifr->ifr_data, &options, sizeof(options));
877		if (error)
878			break;
879		if (options & ~GIF_OPTMASK)
880			error = EINVAL;
881		else
882			sc->gif_options = options;
883		break;
884
885	default:
886		error = EINVAL;
887		break;
888	}
889 bad:
890	return error;
891}
892
893/*
894 * XXXRW: There's a general event-ordering issue here: the code to check
895 * if a given tunnel is already present happens before we perform a
896 * potentially blocking setup of the tunnel.  This code needs to be
897 * re-ordered so that the check and replacement can be atomic using
898 * a mutex.
899 */
900int
901gif_set_tunnel(ifp, src, dst)
902	struct ifnet *ifp;
903	struct sockaddr *src;
904	struct sockaddr *dst;
905{
906	struct gif_softc *sc = ifp->if_softc;
907	struct gif_softc *sc2;
908	struct sockaddr *osrc, *odst, *sa;
909	int error = 0;
910
911	mtx_lock(&gif_mtx);
912	LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
913		if (sc2 == sc)
914			continue;
915		if (!sc2->gif_pdst || !sc2->gif_psrc)
916			continue;
917		if (sc2->gif_pdst->sa_family != dst->sa_family ||
918		    sc2->gif_pdst->sa_len != dst->sa_len ||
919		    sc2->gif_psrc->sa_family != src->sa_family ||
920		    sc2->gif_psrc->sa_len != src->sa_len)
921			continue;
922
923		/*
924		 * Disallow parallel tunnels unless instructed
925		 * otherwise.
926		 */
927		if (!V_parallel_tunnels &&
928		    bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
929		    bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
930			error = EADDRNOTAVAIL;
931			mtx_unlock(&gif_mtx);
932			goto bad;
933		}
934
935		/* XXX both end must be valid? (I mean, not 0.0.0.0) */
936	}
937	mtx_unlock(&gif_mtx);
938
939	/* XXX we can detach from both, but be polite just in case */
940	if (sc->gif_psrc)
941		switch (sc->gif_psrc->sa_family) {
942#ifdef INET
943		case AF_INET:
944			(void)in_gif_detach(sc);
945			break;
946#endif
947#ifdef INET6
948		case AF_INET6:
949			(void)in6_gif_detach(sc);
950			break;
951#endif
952		}
953
954	osrc = sc->gif_psrc;
955	sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
956	bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
957	sc->gif_psrc = sa;
958
959	odst = sc->gif_pdst;
960	sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
961	bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
962	sc->gif_pdst = sa;
963
964	switch (sc->gif_psrc->sa_family) {
965#ifdef INET
966	case AF_INET:
967		error = in_gif_attach(sc);
968		break;
969#endif
970#ifdef INET6
971	case AF_INET6:
972		/*
973		 * Check validity of the scope zone ID of the addresses, and
974		 * convert it into the kernel internal form if necessary.
975		 */
976		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
977		if (error != 0)
978			break;
979		error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
980		if (error != 0)
981			break;
982		error = in6_gif_attach(sc);
983		break;
984#endif
985	}
986	if (error) {
987		/* rollback */
988		free((caddr_t)sc->gif_psrc, M_IFADDR);
989		free((caddr_t)sc->gif_pdst, M_IFADDR);
990		sc->gif_psrc = osrc;
991		sc->gif_pdst = odst;
992		goto bad;
993	}
994
995	if (osrc)
996		free((caddr_t)osrc, M_IFADDR);
997	if (odst)
998		free((caddr_t)odst, M_IFADDR);
999
1000 bad:
1001	if (sc->gif_psrc && sc->gif_pdst)
1002		ifp->if_drv_flags |= IFF_DRV_RUNNING;
1003	else
1004		ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1005
1006	return error;
1007}
1008
1009void
1010gif_delete_tunnel(ifp)
1011	struct ifnet *ifp;
1012{
1013	struct gif_softc *sc = ifp->if_softc;
1014
1015	if (sc->gif_psrc) {
1016		free((caddr_t)sc->gif_psrc, M_IFADDR);
1017		sc->gif_psrc = NULL;
1018	}
1019	if (sc->gif_pdst) {
1020		free((caddr_t)sc->gif_pdst, M_IFADDR);
1021		sc->gif_pdst = NULL;
1022	}
1023	/* it is safe to detach from both */
1024#ifdef INET
1025	(void)in_gif_detach(sc);
1026#endif
1027#ifdef INET6
1028	(void)in6_gif_detach(sc);
1029#endif
1030	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1031}
1032