1/* $FreeBSD: head/sys/net/if_gif.c 195727 2009-07-16 21:13:04Z rwatson $ */
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
98static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
99static VNET_DEFINE(int, max_gif_nesting);
100static VNET_DEFINE(int, parallel_tunnels);
101
102#define V_gif_softc_list VNET(gif_softc_list)
103#define V_max_gif_nesting VNET(max_gif_nesting)
104#define V_parallel_tunnels VNET(parallel_tunnels)
105
106#ifdef INET
107VNET_DEFINE(int, ip_gif_ttl);
108#define V_ip_gif_ttl VNET(ip_gif_ttl)
109#endif
110#ifdef INET6
111VNET_DEFINE(int, ip6_gif_hlim);
112#define V_ip6_gif_hlim VNET(ip6_gif_hlim)
113#endif
114
115void (*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
116void (*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
117void (*ng_gif_attach_p)(struct ifnet *ifp);
118void (*ng_gif_detach_p)(struct ifnet *ifp);
119
120static void gif_start(struct ifnet *);
121static int gif_clone_create(struct if_clone *, int, caddr_t);
122static void gif_clone_destroy(struct ifnet *);
123static int vnet_gif_iattach(const void *);
124
125#ifdef VIMAGE
126static const vnet_modinfo_t vnet_gif_modinfo = {
127 .vmi_id = VNET_MOD_GIF,
128 .vmi_name = "gif",
129 .vmi_dependson = VNET_MOD_NET,
130 .vmi_iattach = vnet_gif_iattach
131};
132#endif
133
134IFC_SIMPLE_DECLARE(gif, 0);
135
136static int gifmodevent(module_t, int, void *);
137
138SYSCTL_DECL(_net_link);
139SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
140 "Generic Tunnel Interface");
141#ifndef MAX_GIF_NEST
142/*
143 * This macro controls the default upper limitation on nesting of gif tunnels.
144 * Since, setting a large value to this macro with a careless configuration
145 * may introduce system crash, we don't allow any nestings by default.
146 * If you need to configure nested gif tunnels, you can define this macro
147 * in your kernel configuration file. However, if you do so, please be
148 * careful to configure the tunnels so that it won't make a loop.
149 */
150#define MAX_GIF_NEST 1
151#endif
152SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
153 &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
154
155#ifdef INET6
156SYSCTL_DECL(_net_inet6_ip6);
157SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_GIF_HLIM, gifhlim, CTLFLAG_RW,
158 &VNET_NAME(ip6_gif_hlim), 0, "");
159#endif
160
161/*
162 * By default, we disallow creation of multiple tunnels between the same
163 * pair of addresses. Some applications require this functionality so
164 * we allow control over this check here.
165 */
166SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
167 &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
168
169/* copy from src/sys/net/if_ethersubr.c */
170static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
171 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
172#ifndef ETHER_IS_BROADCAST
173#define ETHER_IS_BROADCAST(addr) \
174 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
175#endif
176
177static int
178gif_clone_create(ifc, unit, params)
179 struct if_clone *ifc;
180 int unit;
181 caddr_t params;
182{
183 struct gif_softc *sc;
184
185 sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
186 sc->gif_fibnum = curthread->td_proc->p_fibnum;
187 GIF2IFP(sc) = if_alloc(IFT_GIF);
188 if (GIF2IFP(sc) == NULL) {
189 free(sc, M_GIF);
190 return (ENOSPC);
191 }
192
193 GIF_LOCK_INIT(sc);
194
195 GIF2IFP(sc)->if_softc = sc;
196 if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
197
198 sc->encap_cookie4 = sc->encap_cookie6 = NULL;
199 sc->gif_options = GIF_ACCEPT_REVETHIP;
200
201 GIF2IFP(sc)->if_addrlen = 0;
202 GIF2IFP(sc)->if_mtu = GIF_MTU;
203 GIF2IFP(sc)->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
204#if 0
205 /* turn off ingress filter */
206 GIF2IFP(sc)->if_flags |= IFF_LINK2;
207#endif
208 GIF2IFP(sc)->if_ioctl = gif_ioctl;
209 GIF2IFP(sc)->if_start = gif_start;
210 GIF2IFP(sc)->if_output = gif_output;
211 GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
212 if_attach(GIF2IFP(sc));
213 bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
214 if (ng_gif_attach_p != NULL)
215 (*ng_gif_attach_p)(GIF2IFP(sc));
216
217 mtx_lock(&gif_mtx);
218 LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
219 mtx_unlock(&gif_mtx);
220
221 return (0);
222}
223
224static void
225gif_clone_destroy(ifp)
226 struct ifnet *ifp;
227{
228#if defined(INET) || defined(INET6)
229 int err;
230#endif
231 struct gif_softc *sc = ifp->if_softc;
232
233 mtx_lock(&gif_mtx);
234 LIST_REMOVE(sc, gif_list);
235 mtx_unlock(&gif_mtx);
236
237 gif_delete_tunnel(ifp);
238#ifdef INET6
239 if (sc->encap_cookie6 != NULL) {
240 err = encap_detach(sc->encap_cookie6);
241 KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
242 }
243#endif
244#ifdef INET
245 if (sc->encap_cookie4 != NULL) {
246 err = encap_detach(sc->encap_cookie4);
247 KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
248 }
249#endif
250
251 if (ng_gif_detach_p != NULL)
252 (*ng_gif_detach_p)(ifp);
253 bpfdetach(ifp);
254 if_detach(ifp);
255 if_free(ifp);
256
257 GIF_LOCK_DESTROY(sc);
258
259 free(sc, M_GIF);
260}
261
262static int
263vnet_gif_iattach(const void *unused __unused)
264{
265
266 LIST_INIT(&V_gif_softc_list);
267 V_max_gif_nesting = MAX_GIF_NEST;
268#ifdef XBONEHACK
269 V_parallel_tunnels = 1;
270#else
271 V_parallel_tunnels = 0;
272#endif
273#ifdef INET
274 V_ip_gif_ttl = GIF_TTL;
275#endif
276#ifdef INET6
277 V_ip6_gif_hlim = GIF_HLIM;
278#endif
279
280 return (0);
281}
282
283static int
284gifmodevent(mod, type, data)
285 module_t mod;
286 int type;
287 void *data;
288{
289
290 switch (type) {
291 case MOD_LOAD:
292 mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
293
294#ifdef VIMAGE
295 vnet_mod_register(&vnet_gif_modinfo);
296#else
297 vnet_gif_iattach(NULL);
298#endif
299 if_clone_attach(&gif_cloner);
300
301 break;
302 case MOD_UNLOAD:
303 if_clone_detach(&gif_cloner);
304#ifdef VIMAGE
305 vnet_mod_deregister(&vnet_gif_modinfo);
306#endif
307 mtx_destroy(&gif_mtx);
308 break;
309 default:
310 return EOPNOTSUPP;
311 }
312 return 0;
313}
314
315static moduledata_t gif_mod = {
316 "if_gif",
317 gifmodevent,
318 0
319};
320
321DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
322MODULE_VERSION(if_gif, 1);
323
324int
325gif_encapcheck(m, off, proto, arg)
326 const struct mbuf *m;
327 int off;
328 int proto;
329 void *arg;
330{
331 struct ip ip;
332 struct gif_softc *sc;
333
334 sc = (struct gif_softc *)arg;
335 if (sc == NULL)
336 return 0;
337
338 if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
339 return 0;
340
341 /* no physical address */
342 if (!sc->gif_psrc || !sc->gif_pdst)
343 return 0;
344
345 switch (proto) {
346#ifdef INET
347 case IPPROTO_IPV4:
348 break;
349#endif
350#ifdef INET6
351 case IPPROTO_IPV6:
352 break;
353#endif
354 case IPPROTO_ETHERIP:
355 break;
356
357 default:
358 return 0;
359 }
360
361 /* Bail on short packets */
362 if (m->m_pkthdr.len < sizeof(ip))
363 return 0;
364
365 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
366
367 switch (ip.ip_v) {
368#ifdef INET
369 case 4:
370 if (sc->gif_psrc->sa_family != AF_INET ||
371 sc->gif_pdst->sa_family != AF_INET)
372 return 0;
373 return gif_encapcheck4(m, off, proto, arg);
374#endif
375#ifdef INET6
376 case 6:
377 if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
378 return 0;
379 if (sc->gif_psrc->sa_family != AF_INET6 ||
380 sc->gif_pdst->sa_family != AF_INET6)
381 return 0;
382 return gif_encapcheck6(m, off, proto, arg);
383#endif
384 default:
385 return 0;
386 }
387}
388
389static void
390gif_start(struct ifnet *ifp)
391{
392 struct gif_softc *sc;
393 struct mbuf *m;
394
395 sc = ifp->if_softc;
396
397 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
398 for (;;) {
399 IFQ_DEQUEUE(&ifp->if_snd, m);
400 if (m == 0)
401 break;
402
403 gif_output(ifp, m, sc->gif_pdst, NULL);
404
405 }
406 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
407
408 return;
409}
410
411int
412gif_output(ifp, m, dst, ro)
413 struct ifnet *ifp;
414 struct mbuf *m;
415 struct sockaddr *dst;
416 struct route *ro;
417{
418 struct gif_softc *sc = ifp->if_softc;
419 struct m_tag *mtag;
420 int error = 0;
421 int gif_called;
422 u_int32_t af;
423
424#ifdef MAC
425 error = mac_ifnet_check_transmit(ifp, m);
426 if (error) {
427 m_freem(m);
428 goto end;
429 }
430#endif
431
432 /*
433 * gif may cause infinite recursion calls when misconfigured.
434 * We'll prevent this by detecting loops.
435 *
436 * High nesting level may cause stack exhaustion.
437 * We'll prevent this by introducing upper limit.
438 */
439 gif_called = 1;
440 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
441 while (mtag != NULL) {
442 if (*(struct ifnet **)(mtag + 1) == ifp) {
443 log(LOG_NOTICE,
444 "gif_output: loop detected on %s\n",
445 (*(struct ifnet **)(mtag + 1))->if_xname);
446 m_freem(m);
447 error = EIO; /* is there better errno? */
448 goto end;
449 }
450 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
451 gif_called++;
452 }
453 if (gif_called > V_max_gif_nesting) {
454 log(LOG_NOTICE,
455 "gif_output: recursively called too many times(%d)\n",
456 gif_called);
457 m_freem(m);
458 error = EIO; /* is there better errno? */
459 goto end;
460 }
461 mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
462 M_NOWAIT);
463 if (mtag == NULL) {
464 m_freem(m);
465 error = ENOMEM;
466 goto end;
467 }
468 *(struct ifnet **)(mtag + 1) = ifp;
469 m_tag_prepend(m, mtag);
470
471 m->m_flags &= ~(M_BCAST|M_MCAST);
472
473 GIF_LOCK(sc);
474
475 if (!(ifp->if_flags & IFF_UP) ||
476 sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
477 GIF_UNLOCK(sc);
478 m_freem(m);
479 error = ENETDOWN;
480 goto end;
481 }
482
483 /* BPF writes need to be handled specially. */
484 if (dst->sa_family == AF_UNSPEC) {
485 bcopy(dst->sa_data, &af, sizeof(af));
486 dst->sa_family = af;
487 }
488
489 af = dst->sa_family;
490 BPF_MTAP2(ifp, &af, sizeof(af), m);
491 ifp->if_opackets++;
492 ifp->if_obytes += m->m_pkthdr.len;
493
494 /* override to IPPROTO_ETHERIP for bridged traffic */
495 if (ifp->if_bridge)
496 af = AF_LINK;
497
498 M_SETFIB(m, sc->gif_fibnum);
499 /* inner AF-specific encapsulation */
500
501 /* XXX should we check if our outer source is legal? */
502
503 /* dispatch to output logic based on outer AF */
504 switch (sc->gif_psrc->sa_family) {
505#ifdef INET
506 case AF_INET:
507 error = in_gif_output(ifp, af, m);
508 break;
509#endif
510#ifdef INET6
511 case AF_INET6:
512 error = in6_gif_output(ifp, af, m);
513 break;
514#endif
515 default:
516 m_freem(m);
517 error = ENETDOWN;
518 }
519
520 GIF_UNLOCK(sc);
521 end:
522 if (error)
523 ifp->if_oerrors++;
524 return (error);
525}
526
527void
528gif_input(m, af, ifp)
529 struct mbuf *m;
530 int af;
531 struct ifnet *ifp;
532{
533 int isr, n;
534 struct gif_softc *sc = ifp->if_softc;
535 struct etherip_header *eip;
536 struct ether_header *eh;
537 struct ifnet *oldifp;
538
539 if (ifp == NULL) {
540 /* just in case */
541 m_freem(m);
542 return;
543 }
544
545 m->m_pkthdr.rcvif = ifp;
546
547#ifdef MAC
548 mac_ifnet_create_mbuf(ifp, m);
549#endif
550
551 if (bpf_peers_present(ifp->if_bpf)) {
552 u_int32_t af1 = af;
553 bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
554 }
555
556 if (ng_gif_input_p != NULL) {
557 (*ng_gif_input_p)(ifp, &m, af);
558 if (m == NULL)
559 return;
560 }
561
562 /*
563 * Put the packet to the network layer input queue according to the
564 * specified address family.
565 * Note: older versions of gif_input directly called network layer
566 * input functions, e.g. ip6_input, here. We changed the policy to
567 * prevent too many recursive calls of such input functions, which
568 * might cause kernel panic. But the change may introduce another
569 * problem; if the input queue is full, packets are discarded.
570 * The kernel stack overflow really happened, and we believed
571 * queue-full rarely occurs, so we changed the policy.
572 */
573 switch (af) {
574#ifdef INET
575 case AF_INET:
576 isr = NETISR_IP;
577 break;
578#endif
579#ifdef INET6
580 case AF_INET6:
581 isr = NETISR_IPV6;
582 break;
583#endif
584 case AF_LINK:
585 n = sizeof(struct etherip_header) + sizeof(struct ether_header);
586 if (n > m->m_len) {
587 m = m_pullup(m, n);
588 if (m == NULL) {
589 ifp->if_ierrors++;
590 return;
591 }
592 }
593
594 eip = mtod(m, struct etherip_header *);
595 /*
596 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
597 * accepts an EtherIP packet with revered version field in
598 * the header. This is a knob for backward compatibility
599 * with FreeBSD 7.2R or prior.
600 */
601 if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
602 if (eip->eip_resvl != ETHERIP_VERSION
603 && eip->eip_ver != ETHERIP_VERSION) {
604 /* discard unknown versions */
605 m_freem(m);
606 return;
607 }
608 } else {
609 if (eip->eip_ver != ETHERIP_VERSION) {
610 /* discard unknown versions */
611 m_freem(m);
612 return;
613 }
614 }
615 m_adj(m, sizeof(struct etherip_header));
616
617 m->m_flags &= ~(M_BCAST|M_MCAST);
618 m->m_pkthdr.rcvif = ifp;
619
620 if (ifp->if_bridge) {
621 oldifp = ifp;
622 eh = mtod(m, struct ether_header *);
623 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
624 if (ETHER_IS_BROADCAST(eh->ether_dhost))
625 m->m_flags |= M_BCAST;
626 else
627 m->m_flags |= M_MCAST;
628 ifp->if_imcasts++;
629 }
630 BRIDGE_INPUT(ifp, m);
631
632 if (m != NULL && ifp != oldifp) {
633 /*
634 * The bridge gave us back itself or one of the
635 * members for which the frame is addressed.
636 */
637 ether_demux(ifp, m);
638 return;
639 }
640 }
641 if (m != NULL)
642 m_freem(m);
643 return;
644
645 default:
646 if (ng_gif_input_orphan_p != NULL)
647 (*ng_gif_input_orphan_p)(ifp, m, af);
648 else
649 m_freem(m);
650 return;
651 }
652
653 ifp->if_ipackets++;
654 ifp->if_ibytes += m->m_pkthdr.len;
655 netisr_dispatch(isr, m);
656}
657
658/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
659int
660gif_ioctl(ifp, cmd, data)
661 struct ifnet *ifp;
662 u_long cmd;
663 caddr_t data;
664{
665 struct gif_softc *sc = ifp->if_softc;
666 struct ifreq *ifr = (struct ifreq*)data;
667 int error = 0, size;
668 u_int options;
669 struct sockaddr *dst, *src;
670#ifdef SIOCSIFMTU /* xxx */
671 u_long mtu;
672#endif
673
674 switch (cmd) {
675 case SIOCSIFADDR:
676 ifp->if_flags |= IFF_UP;
677 break;
678
679 case SIOCSIFDSTADDR:
680 break;
681
682 case SIOCADDMULTI:
683 case SIOCDELMULTI:
684 break;
685
686#ifdef SIOCSIFMTU /* xxx */
687 case SIOCGIFMTU:
688 break;
689
690 case SIOCSIFMTU:
691 mtu = ifr->ifr_mtu;
692 if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
693 return (EINVAL);
694 ifp->if_mtu = mtu;
695 break;
696#endif /* SIOCSIFMTU */
697
698#ifdef INET
699 case SIOCSIFPHYADDR:
700#endif
701#ifdef INET6
702 case SIOCSIFPHYADDR_IN6:
703#endif /* INET6 */
704 case SIOCSLIFPHYADDR:
705 switch (cmd) {
706#ifdef INET
707 case SIOCSIFPHYADDR:
708 src = (struct sockaddr *)
709 &(((struct in_aliasreq *)data)->ifra_addr);
710 dst = (struct sockaddr *)
711 &(((struct in_aliasreq *)data)->ifra_dstaddr);
712 break;
713#endif
714#ifdef INET6
715 case SIOCSIFPHYADDR_IN6:
716 src = (struct sockaddr *)
717 &(((struct in6_aliasreq *)data)->ifra_addr);
718 dst = (struct sockaddr *)
719 &(((struct in6_aliasreq *)data)->ifra_dstaddr);
720 break;
721#endif
722 case SIOCSLIFPHYADDR:
723 src = (struct sockaddr *)
724 &(((struct if_laddrreq *)data)->addr);
725 dst = (struct sockaddr *)
726 &(((struct if_laddrreq *)data)->dstaddr);
727 break;
728 default:
729 return EINVAL;
730 }
731
732 /* sa_family must be equal */
733 if (src->sa_family != dst->sa_family)
734 return EINVAL;
735
736 /* validate sa_len */
737 switch (src->sa_family) {
738#ifdef INET
739 case AF_INET:
740 if (src->sa_len != sizeof(struct sockaddr_in))
741 return EINVAL;
742 break;
743#endif
744#ifdef INET6
745 case AF_INET6:
746 if (src->sa_len != sizeof(struct sockaddr_in6))
747 return EINVAL;
748 break;
749#endif
750 default:
751 return EAFNOSUPPORT;
752 }
753 switch (dst->sa_family) {
754#ifdef INET
755 case AF_INET:
756 if (dst->sa_len != sizeof(struct sockaddr_in))
757 return EINVAL;
758 break;
759#endif
760#ifdef INET6
761 case AF_INET6:
762 if (dst->sa_len != sizeof(struct sockaddr_in6))
763 return EINVAL;
764 break;
765#endif
766 default:
767 return EAFNOSUPPORT;
768 }
769
770 /* check sa_family looks sane for the cmd */
771 switch (cmd) {
772 case SIOCSIFPHYADDR:
773 if (src->sa_family == AF_INET)
774 break;
775 return EAFNOSUPPORT;
776#ifdef INET6
777 case SIOCSIFPHYADDR_IN6:
778 if (src->sa_family == AF_INET6)
779 break;
780 return EAFNOSUPPORT;
781#endif /* INET6 */
782 case SIOCSLIFPHYADDR:
783 /* checks done in the above */
784 break;
785 }
786
787 error = gif_set_tunnel(GIF2IFP(sc), src, dst);
788 break;
789
790#ifdef SIOCDIFPHYADDR
791 case SIOCDIFPHYADDR:
792 gif_delete_tunnel(GIF2IFP(sc));
793 break;
794#endif
795
796 case SIOCGIFPSRCADDR:
797#ifdef INET6
798 case SIOCGIFPSRCADDR_IN6:
799#endif /* INET6 */
800 if (sc->gif_psrc == NULL) {
801 error = EADDRNOTAVAIL;
802 goto bad;
803 }
804 src = sc->gif_psrc;
805 switch (cmd) {
806#ifdef INET
807 case SIOCGIFPSRCADDR:
808 dst = &ifr->ifr_addr;
809 size = sizeof(ifr->ifr_addr);
810 break;
811#endif /* INET */
812#ifdef INET6
813 case SIOCGIFPSRCADDR_IN6:
814 dst = (struct sockaddr *)
815 &(((struct in6_ifreq *)data)->ifr_addr);
816 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
817 break;
818#endif /* INET6 */
819 default:
820 error = EADDRNOTAVAIL;
821 goto bad;
822 }
823 if (src->sa_len > size)
824 return EINVAL;
825 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
826#ifdef INET6
827 if (dst->sa_family == AF_INET6) {
828 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
829 if (error != 0)
830 return (error);
831 }
832#endif
833 break;
834
835 case SIOCGIFPDSTADDR:
836#ifdef INET6
837 case SIOCGIFPDSTADDR_IN6:
838#endif /* INET6 */
839 if (sc->gif_pdst == NULL) {
840 error = EADDRNOTAVAIL;
841 goto bad;
842 }
843 src = sc->gif_pdst;
844 switch (cmd) {
845#ifdef INET
846 case SIOCGIFPDSTADDR:
847 dst = &ifr->ifr_addr;
848 size = sizeof(ifr->ifr_addr);
849 break;
850#endif /* INET */
851#ifdef INET6
852 case SIOCGIFPDSTADDR_IN6:
853 dst = (struct sockaddr *)
854 &(((struct in6_ifreq *)data)->ifr_addr);
855 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
856 break;
857#endif /* INET6 */
858 default:
859 error = EADDRNOTAVAIL;
860 goto bad;
861 }
862 if (src->sa_len > size)
863 return EINVAL;
864 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
865#ifdef INET6
866 if (dst->sa_family == AF_INET6) {
867 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
868 if (error != 0)
869 return (error);
870 }
871#endif
872 break;
873
874 case SIOCGLIFPHYADDR:
875 if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
876 error = EADDRNOTAVAIL;
877 goto bad;
878 }
879
880 /* copy src */
881 src = sc->gif_psrc;
882 dst = (struct sockaddr *)
883 &(((struct if_laddrreq *)data)->addr);
884 size = sizeof(((struct if_laddrreq *)data)->addr);
885 if (src->sa_len > size)
886 return EINVAL;
887 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
888
889 /* copy dst */
890 src = sc->gif_pdst;
891 dst = (struct sockaddr *)
892 &(((struct if_laddrreq *)data)->dstaddr);
893 size = sizeof(((struct if_laddrreq *)data)->dstaddr);
894 if (src->sa_len > size)
895 return EINVAL;
896 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
897 break;
898
899 case SIOCSIFFLAGS:
900 /* if_ioctl() takes care of it */
901 break;
902
903 case GIFGOPTS:
904 options = sc->gif_options;
905 error = copyout(&options, ifr->ifr_data,
906 sizeof(options));
907 break;
908
909 case GIFSOPTS:
910 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
911 break;
912 error = copyin(ifr->ifr_data, &options, sizeof(options));
913 if (error)
914 break;
915 if (options & ~GIF_OPTMASK)
916 error = EINVAL;
917 else
918 sc->gif_options = options;
919 break;
920
921 default:
922 error = EINVAL;
923 break;
924 }
925 bad:
926 return error;
927}
928
929/*
930 * XXXRW: There's a general event-ordering issue here: the code to check
931 * if a given tunnel is already present happens before we perform a
932 * potentially blocking setup of the tunnel. This code needs to be
933 * re-ordered so that the check and replacement can be atomic using
934 * a mutex.
935 */
936int
937gif_set_tunnel(ifp, src, dst)
938 struct ifnet *ifp;
939 struct sockaddr *src;
940 struct sockaddr *dst;
941{
942 struct gif_softc *sc = ifp->if_softc;
943 struct gif_softc *sc2;
944 struct sockaddr *osrc, *odst, *sa;
945 int error = 0;
946
947 mtx_lock(&gif_mtx);
948 LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
949 if (sc2 == sc)
950 continue;
951 if (!sc2->gif_pdst || !sc2->gif_psrc)
952 continue;
953 if (sc2->gif_pdst->sa_family != dst->sa_family ||
954 sc2->gif_pdst->sa_len != dst->sa_len ||
955 sc2->gif_psrc->sa_family != src->sa_family ||
956 sc2->gif_psrc->sa_len != src->sa_len)
957 continue;
958
959 /*
960 * Disallow parallel tunnels unless instructed
961 * otherwise.
962 */
963 if (!V_parallel_tunnels &&
964 bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
965 bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
966 error = EADDRNOTAVAIL;
967 mtx_unlock(&gif_mtx);
968 goto bad;
969 }
970
971 /* XXX both end must be valid? (I mean, not 0.0.0.0) */
972 }
973 mtx_unlock(&gif_mtx);
974
975 /* XXX we can detach from both, but be polite just in case */
976 if (sc->gif_psrc)
977 switch (sc->gif_psrc->sa_family) {
978#ifdef INET
979 case AF_INET:
980 (void)in_gif_detach(sc);
981 break;
982#endif
983#ifdef INET6
984 case AF_INET6:
985 (void)in6_gif_detach(sc);
986 break;
987#endif
988 }
989
990 osrc = sc->gif_psrc;
991 sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
992 bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
993 sc->gif_psrc = sa;
994
995 odst = sc->gif_pdst;
996 sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
997 bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
998 sc->gif_pdst = sa;
999
1000 switch (sc->gif_psrc->sa_family) {
1001#ifdef INET
1002 case AF_INET:
1003 error = in_gif_attach(sc);
1004 break;
1005#endif
1006#ifdef INET6
1007 case AF_INET6:
1008 /*
1009 * Check validity of the scope zone ID of the addresses, and
1010 * convert it into the kernel internal form if necessary.
1011 */
1012 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
1013 if (error != 0)
1014 break;
1015 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
1016 if (error != 0)
1017 break;
1018 error = in6_gif_attach(sc);
1019 break;
1020#endif
1021 }
1022 if (error) {
1023 /* rollback */
1024 free((caddr_t)sc->gif_psrc, M_IFADDR);
1025 free((caddr_t)sc->gif_pdst, M_IFADDR);
1026 sc->gif_psrc = osrc;
1027 sc->gif_pdst = odst;
1028 goto bad;
1029 }
1030
1031 if (osrc)
1032 free((caddr_t)osrc, M_IFADDR);
1033 if (odst)
1034 free((caddr_t)odst, M_IFADDR);
1035
1036 bad:
1037 if (sc->gif_psrc && sc->gif_pdst)
1038 ifp->if_drv_flags |= IFF_DRV_RUNNING;
1039 else
1040 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1041
1042 return error;
1043}
1044
1045void
1046gif_delete_tunnel(ifp)
1047 struct ifnet *ifp;
1048{
1049 struct gif_softc *sc = ifp->if_softc;
1050
1051 if (sc->gif_psrc) {
1052 free((caddr_t)sc->gif_psrc, M_IFADDR);
1053 sc->gif_psrc = NULL;
1054 }
1055 if (sc->gif_pdst) {
1056 free((caddr_t)sc->gif_pdst, M_IFADDR);
1057 sc->gif_pdst = NULL;
1058 }
1059 /* it is safe to detach from both */
1060#ifdef INET
1061 (void)in_gif_detach(sc);
1062#endif
1063#ifdef INET6
1064 (void)in6_gif_detach(sc);
1065#endif
1066 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1067}
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
98static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
99static VNET_DEFINE(int, max_gif_nesting);
100static VNET_DEFINE(int, parallel_tunnels);
101
102#define V_gif_softc_list VNET(gif_softc_list)
103#define V_max_gif_nesting VNET(max_gif_nesting)
104#define V_parallel_tunnels VNET(parallel_tunnels)
105
106#ifdef INET
107VNET_DEFINE(int, ip_gif_ttl);
108#define V_ip_gif_ttl VNET(ip_gif_ttl)
109#endif
110#ifdef INET6
111VNET_DEFINE(int, ip6_gif_hlim);
112#define V_ip6_gif_hlim VNET(ip6_gif_hlim)
113#endif
114
115void (*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
116void (*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
117void (*ng_gif_attach_p)(struct ifnet *ifp);
118void (*ng_gif_detach_p)(struct ifnet *ifp);
119
120static void gif_start(struct ifnet *);
121static int gif_clone_create(struct if_clone *, int, caddr_t);
122static void gif_clone_destroy(struct ifnet *);
123static int vnet_gif_iattach(const void *);
124
125#ifdef VIMAGE
126static const vnet_modinfo_t vnet_gif_modinfo = {
127 .vmi_id = VNET_MOD_GIF,
128 .vmi_name = "gif",
129 .vmi_dependson = VNET_MOD_NET,
130 .vmi_iattach = vnet_gif_iattach
131};
132#endif
133
134IFC_SIMPLE_DECLARE(gif, 0);
135
136static int gifmodevent(module_t, int, void *);
137
138SYSCTL_DECL(_net_link);
139SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
140 "Generic Tunnel Interface");
141#ifndef MAX_GIF_NEST
142/*
143 * This macro controls the default upper limitation on nesting of gif tunnels.
144 * Since, setting a large value to this macro with a careless configuration
145 * may introduce system crash, we don't allow any nestings by default.
146 * If you need to configure nested gif tunnels, you can define this macro
147 * in your kernel configuration file. However, if you do so, please be
148 * careful to configure the tunnels so that it won't make a loop.
149 */
150#define MAX_GIF_NEST 1
151#endif
152SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
153 &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
154
155#ifdef INET6
156SYSCTL_DECL(_net_inet6_ip6);
157SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_GIF_HLIM, gifhlim, CTLFLAG_RW,
158 &VNET_NAME(ip6_gif_hlim), 0, "");
159#endif
160
161/*
162 * By default, we disallow creation of multiple tunnels between the same
163 * pair of addresses. Some applications require this functionality so
164 * we allow control over this check here.
165 */
166SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
167 &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
168
169/* copy from src/sys/net/if_ethersubr.c */
170static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
171 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
172#ifndef ETHER_IS_BROADCAST
173#define ETHER_IS_BROADCAST(addr) \
174 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
175#endif
176
177static int
178gif_clone_create(ifc, unit, params)
179 struct if_clone *ifc;
180 int unit;
181 caddr_t params;
182{
183 struct gif_softc *sc;
184
185 sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
186 sc->gif_fibnum = curthread->td_proc->p_fibnum;
187 GIF2IFP(sc) = if_alloc(IFT_GIF);
188 if (GIF2IFP(sc) == NULL) {
189 free(sc, M_GIF);
190 return (ENOSPC);
191 }
192
193 GIF_LOCK_INIT(sc);
194
195 GIF2IFP(sc)->if_softc = sc;
196 if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
197
198 sc->encap_cookie4 = sc->encap_cookie6 = NULL;
199 sc->gif_options = GIF_ACCEPT_REVETHIP;
200
201 GIF2IFP(sc)->if_addrlen = 0;
202 GIF2IFP(sc)->if_mtu = GIF_MTU;
203 GIF2IFP(sc)->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
204#if 0
205 /* turn off ingress filter */
206 GIF2IFP(sc)->if_flags |= IFF_LINK2;
207#endif
208 GIF2IFP(sc)->if_ioctl = gif_ioctl;
209 GIF2IFP(sc)->if_start = gif_start;
210 GIF2IFP(sc)->if_output = gif_output;
211 GIF2IFP(sc)->if_snd.ifq_maxlen = IFQ_MAXLEN;
212 if_attach(GIF2IFP(sc));
213 bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
214 if (ng_gif_attach_p != NULL)
215 (*ng_gif_attach_p)(GIF2IFP(sc));
216
217 mtx_lock(&gif_mtx);
218 LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
219 mtx_unlock(&gif_mtx);
220
221 return (0);
222}
223
224static void
225gif_clone_destroy(ifp)
226 struct ifnet *ifp;
227{
228#if defined(INET) || defined(INET6)
229 int err;
230#endif
231 struct gif_softc *sc = ifp->if_softc;
232
233 mtx_lock(&gif_mtx);
234 LIST_REMOVE(sc, gif_list);
235 mtx_unlock(&gif_mtx);
236
237 gif_delete_tunnel(ifp);
238#ifdef INET6
239 if (sc->encap_cookie6 != NULL) {
240 err = encap_detach(sc->encap_cookie6);
241 KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
242 }
243#endif
244#ifdef INET
245 if (sc->encap_cookie4 != NULL) {
246 err = encap_detach(sc->encap_cookie4);
247 KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
248 }
249#endif
250
251 if (ng_gif_detach_p != NULL)
252 (*ng_gif_detach_p)(ifp);
253 bpfdetach(ifp);
254 if_detach(ifp);
255 if_free(ifp);
256
257 GIF_LOCK_DESTROY(sc);
258
259 free(sc, M_GIF);
260}
261
262static int
263vnet_gif_iattach(const void *unused __unused)
264{
265
266 LIST_INIT(&V_gif_softc_list);
267 V_max_gif_nesting = MAX_GIF_NEST;
268#ifdef XBONEHACK
269 V_parallel_tunnels = 1;
270#else
271 V_parallel_tunnels = 0;
272#endif
273#ifdef INET
274 V_ip_gif_ttl = GIF_TTL;
275#endif
276#ifdef INET6
277 V_ip6_gif_hlim = GIF_HLIM;
278#endif
279
280 return (0);
281}
282
283static int
284gifmodevent(mod, type, data)
285 module_t mod;
286 int type;
287 void *data;
288{
289
290 switch (type) {
291 case MOD_LOAD:
292 mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
293
294#ifdef VIMAGE
295 vnet_mod_register(&vnet_gif_modinfo);
296#else
297 vnet_gif_iattach(NULL);
298#endif
299 if_clone_attach(&gif_cloner);
300
301 break;
302 case MOD_UNLOAD:
303 if_clone_detach(&gif_cloner);
304#ifdef VIMAGE
305 vnet_mod_deregister(&vnet_gif_modinfo);
306#endif
307 mtx_destroy(&gif_mtx);
308 break;
309 default:
310 return EOPNOTSUPP;
311 }
312 return 0;
313}
314
315static moduledata_t gif_mod = {
316 "if_gif",
317 gifmodevent,
318 0
319};
320
321DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
322MODULE_VERSION(if_gif, 1);
323
324int
325gif_encapcheck(m, off, proto, arg)
326 const struct mbuf *m;
327 int off;
328 int proto;
329 void *arg;
330{
331 struct ip ip;
332 struct gif_softc *sc;
333
334 sc = (struct gif_softc *)arg;
335 if (sc == NULL)
336 return 0;
337
338 if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
339 return 0;
340
341 /* no physical address */
342 if (!sc->gif_psrc || !sc->gif_pdst)
343 return 0;
344
345 switch (proto) {
346#ifdef INET
347 case IPPROTO_IPV4:
348 break;
349#endif
350#ifdef INET6
351 case IPPROTO_IPV6:
352 break;
353#endif
354 case IPPROTO_ETHERIP:
355 break;
356
357 default:
358 return 0;
359 }
360
361 /* Bail on short packets */
362 if (m->m_pkthdr.len < sizeof(ip))
363 return 0;
364
365 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
366
367 switch (ip.ip_v) {
368#ifdef INET
369 case 4:
370 if (sc->gif_psrc->sa_family != AF_INET ||
371 sc->gif_pdst->sa_family != AF_INET)
372 return 0;
373 return gif_encapcheck4(m, off, proto, arg);
374#endif
375#ifdef INET6
376 case 6:
377 if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
378 return 0;
379 if (sc->gif_psrc->sa_family != AF_INET6 ||
380 sc->gif_pdst->sa_family != AF_INET6)
381 return 0;
382 return gif_encapcheck6(m, off, proto, arg);
383#endif
384 default:
385 return 0;
386 }
387}
388
389static void
390gif_start(struct ifnet *ifp)
391{
392 struct gif_softc *sc;
393 struct mbuf *m;
394
395 sc = ifp->if_softc;
396
397 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
398 for (;;) {
399 IFQ_DEQUEUE(&ifp->if_snd, m);
400 if (m == 0)
401 break;
402
403 gif_output(ifp, m, sc->gif_pdst, NULL);
404
405 }
406 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
407
408 return;
409}
410
411int
412gif_output(ifp, m, dst, ro)
413 struct ifnet *ifp;
414 struct mbuf *m;
415 struct sockaddr *dst;
416 struct route *ro;
417{
418 struct gif_softc *sc = ifp->if_softc;
419 struct m_tag *mtag;
420 int error = 0;
421 int gif_called;
422 u_int32_t af;
423
424#ifdef MAC
425 error = mac_ifnet_check_transmit(ifp, m);
426 if (error) {
427 m_freem(m);
428 goto end;
429 }
430#endif
431
432 /*
433 * gif may cause infinite recursion calls when misconfigured.
434 * We'll prevent this by detecting loops.
435 *
436 * High nesting level may cause stack exhaustion.
437 * We'll prevent this by introducing upper limit.
438 */
439 gif_called = 1;
440 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
441 while (mtag != NULL) {
442 if (*(struct ifnet **)(mtag + 1) == ifp) {
443 log(LOG_NOTICE,
444 "gif_output: loop detected on %s\n",
445 (*(struct ifnet **)(mtag + 1))->if_xname);
446 m_freem(m);
447 error = EIO; /* is there better errno? */
448 goto end;
449 }
450 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
451 gif_called++;
452 }
453 if (gif_called > V_max_gif_nesting) {
454 log(LOG_NOTICE,
455 "gif_output: recursively called too many times(%d)\n",
456 gif_called);
457 m_freem(m);
458 error = EIO; /* is there better errno? */
459 goto end;
460 }
461 mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
462 M_NOWAIT);
463 if (mtag == NULL) {
464 m_freem(m);
465 error = ENOMEM;
466 goto end;
467 }
468 *(struct ifnet **)(mtag + 1) = ifp;
469 m_tag_prepend(m, mtag);
470
471 m->m_flags &= ~(M_BCAST|M_MCAST);
472
473 GIF_LOCK(sc);
474
475 if (!(ifp->if_flags & IFF_UP) ||
476 sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
477 GIF_UNLOCK(sc);
478 m_freem(m);
479 error = ENETDOWN;
480 goto end;
481 }
482
483 /* BPF writes need to be handled specially. */
484 if (dst->sa_family == AF_UNSPEC) {
485 bcopy(dst->sa_data, &af, sizeof(af));
486 dst->sa_family = af;
487 }
488
489 af = dst->sa_family;
490 BPF_MTAP2(ifp, &af, sizeof(af), m);
491 ifp->if_opackets++;
492 ifp->if_obytes += m->m_pkthdr.len;
493
494 /* override to IPPROTO_ETHERIP for bridged traffic */
495 if (ifp->if_bridge)
496 af = AF_LINK;
497
498 M_SETFIB(m, sc->gif_fibnum);
499 /* inner AF-specific encapsulation */
500
501 /* XXX should we check if our outer source is legal? */
502
503 /* dispatch to output logic based on outer AF */
504 switch (sc->gif_psrc->sa_family) {
505#ifdef INET
506 case AF_INET:
507 error = in_gif_output(ifp, af, m);
508 break;
509#endif
510#ifdef INET6
511 case AF_INET6:
512 error = in6_gif_output(ifp, af, m);
513 break;
514#endif
515 default:
516 m_freem(m);
517 error = ENETDOWN;
518 }
519
520 GIF_UNLOCK(sc);
521 end:
522 if (error)
523 ifp->if_oerrors++;
524 return (error);
525}
526
527void
528gif_input(m, af, ifp)
529 struct mbuf *m;
530 int af;
531 struct ifnet *ifp;
532{
533 int isr, n;
534 struct gif_softc *sc = ifp->if_softc;
535 struct etherip_header *eip;
536 struct ether_header *eh;
537 struct ifnet *oldifp;
538
539 if (ifp == NULL) {
540 /* just in case */
541 m_freem(m);
542 return;
543 }
544
545 m->m_pkthdr.rcvif = ifp;
546
547#ifdef MAC
548 mac_ifnet_create_mbuf(ifp, m);
549#endif
550
551 if (bpf_peers_present(ifp->if_bpf)) {
552 u_int32_t af1 = af;
553 bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
554 }
555
556 if (ng_gif_input_p != NULL) {
557 (*ng_gif_input_p)(ifp, &m, af);
558 if (m == NULL)
559 return;
560 }
561
562 /*
563 * Put the packet to the network layer input queue according to the
564 * specified address family.
565 * Note: older versions of gif_input directly called network layer
566 * input functions, e.g. ip6_input, here. We changed the policy to
567 * prevent too many recursive calls of such input functions, which
568 * might cause kernel panic. But the change may introduce another
569 * problem; if the input queue is full, packets are discarded.
570 * The kernel stack overflow really happened, and we believed
571 * queue-full rarely occurs, so we changed the policy.
572 */
573 switch (af) {
574#ifdef INET
575 case AF_INET:
576 isr = NETISR_IP;
577 break;
578#endif
579#ifdef INET6
580 case AF_INET6:
581 isr = NETISR_IPV6;
582 break;
583#endif
584 case AF_LINK:
585 n = sizeof(struct etherip_header) + sizeof(struct ether_header);
586 if (n > m->m_len) {
587 m = m_pullup(m, n);
588 if (m == NULL) {
589 ifp->if_ierrors++;
590 return;
591 }
592 }
593
594 eip = mtod(m, struct etherip_header *);
595 /*
596 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
597 * accepts an EtherIP packet with revered version field in
598 * the header. This is a knob for backward compatibility
599 * with FreeBSD 7.2R or prior.
600 */
601 if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
602 if (eip->eip_resvl != ETHERIP_VERSION
603 && eip->eip_ver != ETHERIP_VERSION) {
604 /* discard unknown versions */
605 m_freem(m);
606 return;
607 }
608 } else {
609 if (eip->eip_ver != ETHERIP_VERSION) {
610 /* discard unknown versions */
611 m_freem(m);
612 return;
613 }
614 }
615 m_adj(m, sizeof(struct etherip_header));
616
617 m->m_flags &= ~(M_BCAST|M_MCAST);
618 m->m_pkthdr.rcvif = ifp;
619
620 if (ifp->if_bridge) {
621 oldifp = ifp;
622 eh = mtod(m, struct ether_header *);
623 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
624 if (ETHER_IS_BROADCAST(eh->ether_dhost))
625 m->m_flags |= M_BCAST;
626 else
627 m->m_flags |= M_MCAST;
628 ifp->if_imcasts++;
629 }
630 BRIDGE_INPUT(ifp, m);
631
632 if (m != NULL && ifp != oldifp) {
633 /*
634 * The bridge gave us back itself or one of the
635 * members for which the frame is addressed.
636 */
637 ether_demux(ifp, m);
638 return;
639 }
640 }
641 if (m != NULL)
642 m_freem(m);
643 return;
644
645 default:
646 if (ng_gif_input_orphan_p != NULL)
647 (*ng_gif_input_orphan_p)(ifp, m, af);
648 else
649 m_freem(m);
650 return;
651 }
652
653 ifp->if_ipackets++;
654 ifp->if_ibytes += m->m_pkthdr.len;
655 netisr_dispatch(isr, m);
656}
657
658/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
659int
660gif_ioctl(ifp, cmd, data)
661 struct ifnet *ifp;
662 u_long cmd;
663 caddr_t data;
664{
665 struct gif_softc *sc = ifp->if_softc;
666 struct ifreq *ifr = (struct ifreq*)data;
667 int error = 0, size;
668 u_int options;
669 struct sockaddr *dst, *src;
670#ifdef SIOCSIFMTU /* xxx */
671 u_long mtu;
672#endif
673
674 switch (cmd) {
675 case SIOCSIFADDR:
676 ifp->if_flags |= IFF_UP;
677 break;
678
679 case SIOCSIFDSTADDR:
680 break;
681
682 case SIOCADDMULTI:
683 case SIOCDELMULTI:
684 break;
685
686#ifdef SIOCSIFMTU /* xxx */
687 case SIOCGIFMTU:
688 break;
689
690 case SIOCSIFMTU:
691 mtu = ifr->ifr_mtu;
692 if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
693 return (EINVAL);
694 ifp->if_mtu = mtu;
695 break;
696#endif /* SIOCSIFMTU */
697
698#ifdef INET
699 case SIOCSIFPHYADDR:
700#endif
701#ifdef INET6
702 case SIOCSIFPHYADDR_IN6:
703#endif /* INET6 */
704 case SIOCSLIFPHYADDR:
705 switch (cmd) {
706#ifdef INET
707 case SIOCSIFPHYADDR:
708 src = (struct sockaddr *)
709 &(((struct in_aliasreq *)data)->ifra_addr);
710 dst = (struct sockaddr *)
711 &(((struct in_aliasreq *)data)->ifra_dstaddr);
712 break;
713#endif
714#ifdef INET6
715 case SIOCSIFPHYADDR_IN6:
716 src = (struct sockaddr *)
717 &(((struct in6_aliasreq *)data)->ifra_addr);
718 dst = (struct sockaddr *)
719 &(((struct in6_aliasreq *)data)->ifra_dstaddr);
720 break;
721#endif
722 case SIOCSLIFPHYADDR:
723 src = (struct sockaddr *)
724 &(((struct if_laddrreq *)data)->addr);
725 dst = (struct sockaddr *)
726 &(((struct if_laddrreq *)data)->dstaddr);
727 break;
728 default:
729 return EINVAL;
730 }
731
732 /* sa_family must be equal */
733 if (src->sa_family != dst->sa_family)
734 return EINVAL;
735
736 /* validate sa_len */
737 switch (src->sa_family) {
738#ifdef INET
739 case AF_INET:
740 if (src->sa_len != sizeof(struct sockaddr_in))
741 return EINVAL;
742 break;
743#endif
744#ifdef INET6
745 case AF_INET6:
746 if (src->sa_len != sizeof(struct sockaddr_in6))
747 return EINVAL;
748 break;
749#endif
750 default:
751 return EAFNOSUPPORT;
752 }
753 switch (dst->sa_family) {
754#ifdef INET
755 case AF_INET:
756 if (dst->sa_len != sizeof(struct sockaddr_in))
757 return EINVAL;
758 break;
759#endif
760#ifdef INET6
761 case AF_INET6:
762 if (dst->sa_len != sizeof(struct sockaddr_in6))
763 return EINVAL;
764 break;
765#endif
766 default:
767 return EAFNOSUPPORT;
768 }
769
770 /* check sa_family looks sane for the cmd */
771 switch (cmd) {
772 case SIOCSIFPHYADDR:
773 if (src->sa_family == AF_INET)
774 break;
775 return EAFNOSUPPORT;
776#ifdef INET6
777 case SIOCSIFPHYADDR_IN6:
778 if (src->sa_family == AF_INET6)
779 break;
780 return EAFNOSUPPORT;
781#endif /* INET6 */
782 case SIOCSLIFPHYADDR:
783 /* checks done in the above */
784 break;
785 }
786
787 error = gif_set_tunnel(GIF2IFP(sc), src, dst);
788 break;
789
790#ifdef SIOCDIFPHYADDR
791 case SIOCDIFPHYADDR:
792 gif_delete_tunnel(GIF2IFP(sc));
793 break;
794#endif
795
796 case SIOCGIFPSRCADDR:
797#ifdef INET6
798 case SIOCGIFPSRCADDR_IN6:
799#endif /* INET6 */
800 if (sc->gif_psrc == NULL) {
801 error = EADDRNOTAVAIL;
802 goto bad;
803 }
804 src = sc->gif_psrc;
805 switch (cmd) {
806#ifdef INET
807 case SIOCGIFPSRCADDR:
808 dst = &ifr->ifr_addr;
809 size = sizeof(ifr->ifr_addr);
810 break;
811#endif /* INET */
812#ifdef INET6
813 case SIOCGIFPSRCADDR_IN6:
814 dst = (struct sockaddr *)
815 &(((struct in6_ifreq *)data)->ifr_addr);
816 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
817 break;
818#endif /* INET6 */
819 default:
820 error = EADDRNOTAVAIL;
821 goto bad;
822 }
823 if (src->sa_len > size)
824 return EINVAL;
825 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
826#ifdef INET6
827 if (dst->sa_family == AF_INET6) {
828 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
829 if (error != 0)
830 return (error);
831 }
832#endif
833 break;
834
835 case SIOCGIFPDSTADDR:
836#ifdef INET6
837 case SIOCGIFPDSTADDR_IN6:
838#endif /* INET6 */
839 if (sc->gif_pdst == NULL) {
840 error = EADDRNOTAVAIL;
841 goto bad;
842 }
843 src = sc->gif_pdst;
844 switch (cmd) {
845#ifdef INET
846 case SIOCGIFPDSTADDR:
847 dst = &ifr->ifr_addr;
848 size = sizeof(ifr->ifr_addr);
849 break;
850#endif /* INET */
851#ifdef INET6
852 case SIOCGIFPDSTADDR_IN6:
853 dst = (struct sockaddr *)
854 &(((struct in6_ifreq *)data)->ifr_addr);
855 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
856 break;
857#endif /* INET6 */
858 default:
859 error = EADDRNOTAVAIL;
860 goto bad;
861 }
862 if (src->sa_len > size)
863 return EINVAL;
864 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
865#ifdef INET6
866 if (dst->sa_family == AF_INET6) {
867 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
868 if (error != 0)
869 return (error);
870 }
871#endif
872 break;
873
874 case SIOCGLIFPHYADDR:
875 if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
876 error = EADDRNOTAVAIL;
877 goto bad;
878 }
879
880 /* copy src */
881 src = sc->gif_psrc;
882 dst = (struct sockaddr *)
883 &(((struct if_laddrreq *)data)->addr);
884 size = sizeof(((struct if_laddrreq *)data)->addr);
885 if (src->sa_len > size)
886 return EINVAL;
887 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
888
889 /* copy dst */
890 src = sc->gif_pdst;
891 dst = (struct sockaddr *)
892 &(((struct if_laddrreq *)data)->dstaddr);
893 size = sizeof(((struct if_laddrreq *)data)->dstaddr);
894 if (src->sa_len > size)
895 return EINVAL;
896 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
897 break;
898
899 case SIOCSIFFLAGS:
900 /* if_ioctl() takes care of it */
901 break;
902
903 case GIFGOPTS:
904 options = sc->gif_options;
905 error = copyout(&options, ifr->ifr_data,
906 sizeof(options));
907 break;
908
909 case GIFSOPTS:
910 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
911 break;
912 error = copyin(ifr->ifr_data, &options, sizeof(options));
913 if (error)
914 break;
915 if (options & ~GIF_OPTMASK)
916 error = EINVAL;
917 else
918 sc->gif_options = options;
919 break;
920
921 default:
922 error = EINVAL;
923 break;
924 }
925 bad:
926 return error;
927}
928
929/*
930 * XXXRW: There's a general event-ordering issue here: the code to check
931 * if a given tunnel is already present happens before we perform a
932 * potentially blocking setup of the tunnel. This code needs to be
933 * re-ordered so that the check and replacement can be atomic using
934 * a mutex.
935 */
936int
937gif_set_tunnel(ifp, src, dst)
938 struct ifnet *ifp;
939 struct sockaddr *src;
940 struct sockaddr *dst;
941{
942 struct gif_softc *sc = ifp->if_softc;
943 struct gif_softc *sc2;
944 struct sockaddr *osrc, *odst, *sa;
945 int error = 0;
946
947 mtx_lock(&gif_mtx);
948 LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
949 if (sc2 == sc)
950 continue;
951 if (!sc2->gif_pdst || !sc2->gif_psrc)
952 continue;
953 if (sc2->gif_pdst->sa_family != dst->sa_family ||
954 sc2->gif_pdst->sa_len != dst->sa_len ||
955 sc2->gif_psrc->sa_family != src->sa_family ||
956 sc2->gif_psrc->sa_len != src->sa_len)
957 continue;
958
959 /*
960 * Disallow parallel tunnels unless instructed
961 * otherwise.
962 */
963 if (!V_parallel_tunnels &&
964 bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
965 bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
966 error = EADDRNOTAVAIL;
967 mtx_unlock(&gif_mtx);
968 goto bad;
969 }
970
971 /* XXX both end must be valid? (I mean, not 0.0.0.0) */
972 }
973 mtx_unlock(&gif_mtx);
974
975 /* XXX we can detach from both, but be polite just in case */
976 if (sc->gif_psrc)
977 switch (sc->gif_psrc->sa_family) {
978#ifdef INET
979 case AF_INET:
980 (void)in_gif_detach(sc);
981 break;
982#endif
983#ifdef INET6
984 case AF_INET6:
985 (void)in6_gif_detach(sc);
986 break;
987#endif
988 }
989
990 osrc = sc->gif_psrc;
991 sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
992 bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
993 sc->gif_psrc = sa;
994
995 odst = sc->gif_pdst;
996 sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
997 bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
998 sc->gif_pdst = sa;
999
1000 switch (sc->gif_psrc->sa_family) {
1001#ifdef INET
1002 case AF_INET:
1003 error = in_gif_attach(sc);
1004 break;
1005#endif
1006#ifdef INET6
1007 case AF_INET6:
1008 /*
1009 * Check validity of the scope zone ID of the addresses, and
1010 * convert it into the kernel internal form if necessary.
1011 */
1012 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
1013 if (error != 0)
1014 break;
1015 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
1016 if (error != 0)
1017 break;
1018 error = in6_gif_attach(sc);
1019 break;
1020#endif
1021 }
1022 if (error) {
1023 /* rollback */
1024 free((caddr_t)sc->gif_psrc, M_IFADDR);
1025 free((caddr_t)sc->gif_pdst, M_IFADDR);
1026 sc->gif_psrc = osrc;
1027 sc->gif_pdst = odst;
1028 goto bad;
1029 }
1030
1031 if (osrc)
1032 free((caddr_t)osrc, M_IFADDR);
1033 if (odst)
1034 free((caddr_t)odst, M_IFADDR);
1035
1036 bad:
1037 if (sc->gif_psrc && sc->gif_pdst)
1038 ifp->if_drv_flags |= IFF_DRV_RUNNING;
1039 else
1040 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1041
1042 return error;
1043}
1044
1045void
1046gif_delete_tunnel(ifp)
1047 struct ifnet *ifp;
1048{
1049 struct gif_softc *sc = ifp->if_softc;
1050
1051 if (sc->gif_psrc) {
1052 free((caddr_t)sc->gif_psrc, M_IFADDR);
1053 sc->gif_psrc = NULL;
1054 }
1055 if (sc->gif_pdst) {
1056 free((caddr_t)sc->gif_pdst, M_IFADDR);
1057 sc->gif_pdst = NULL;
1058 }
1059 /* it is safe to detach from both */
1060#ifdef INET
1061 (void)in_gif_detach(sc);
1062#endif
1063#ifdef INET6
1064 (void)in6_gif_detach(sc);
1065#endif
1066 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1067}