2 /* $KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $ */
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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.
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
34 #include "opt_inet6.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/kernel.h>
39 #include <sys/malloc.h>
41 #include <sys/module.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/errno.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
50 #include <sys/protosw.h>
52 #include <machine/cpu.h>
55 #include <net/if_clone.h>
56 #include <net/if_types.h>
57 #include <net/netisr.h>
58 #include <net/route.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
66 #include <netinet/in_var.h>
67 #include <netinet/in_gif.h>
68 #include <netinet/ip_var.h>
73 #include <netinet/in.h>
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>
83 #include <netinet/ip_encap.h>
84 #include <net/ethernet.h>
85 #include <net/if_bridgevar.h>
86 #include <net/if_gif.h>
88 #include <security/mac/mac_framework.h>
93 * gif_mtx protects the global gif_softc_list.
95 static struct mtx gif_mtx;
96 static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
97 static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list);
98 #define V_gif_softc_list VNET(gif_softc_list)
100 void (*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
101 void (*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
102 void (*ng_gif_attach_p)(struct ifnet *ifp);
103 void (*ng_gif_detach_p)(struct ifnet *ifp);
105 static void gif_start(struct ifnet *);
106 static int gif_clone_create(struct if_clone *, int, caddr_t);
107 static void gif_clone_destroy(struct ifnet *);
109 IFC_SIMPLE_DECLARE(gif, 0);
111 static int gifmodevent(module_t, int, void *);
113 SYSCTL_DECL(_net_link);
114 SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
115 "Generic Tunnel Interface");
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.
125 #define MAX_GIF_NEST 1
127 static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST;
128 #define V_max_gif_nesting VNET(max_gif_nesting)
129 SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW,
130 &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
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.
138 static VNET_DEFINE(int, parallel_tunnels) = 1;
140 static VNET_DEFINE(int, parallel_tunnels) = 0;
142 #define V_parallel_tunnels VNET(parallel_tunnels)
143 SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW,
144 &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?");
146 /* copy from src/sys/net/if_ethersubr.c */
147 static 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)
155 gif_clone_create(ifc, unit, params)
156 struct if_clone *ifc;
160 struct gif_softc *sc;
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) {
172 GIF2IFP(sc)->if_softc = sc;
173 if_initname(GIF2IFP(sc), ifc->ifc_name, unit);
175 sc->encap_cookie4 = sc->encap_cookie6 = NULL;
176 sc->gif_options = GIF_ACCEPT_REVETHIP;
178 GIF2IFP(sc)->if_addrlen = 0;
179 GIF2IFP(sc)->if_mtu = GIF_MTU;
180 GIF2IFP(sc)->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
182 /* turn off ingress filter */
183 GIF2IFP(sc)->if_flags |= IFF_LINK2;
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 = IFQ_MAXLEN;
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));
195 LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list);
196 mtx_unlock(&gif_mtx);
202 gif_clone_destroy(ifp)
205 #if defined(INET) || defined(INET6)
208 struct gif_softc *sc = ifp->if_softc;
211 LIST_REMOVE(sc, gif_list);
212 mtx_unlock(&gif_mtx);
214 gif_delete_tunnel(ifp);
216 if (sc->encap_cookie6 != NULL) {
217 err = encap_detach(sc->encap_cookie6);
218 KASSERT(err == 0, ("Unexpected error detaching encap_cookie6"));
222 if (sc->encap_cookie4 != NULL) {
223 err = encap_detach(sc->encap_cookie4);
224 KASSERT(err == 0, ("Unexpected error detaching encap_cookie4"));
228 if (ng_gif_detach_p != NULL)
229 (*ng_gif_detach_p)(ifp);
234 GIF_LOCK_DESTROY(sc);
240 vnet_gif_init(const void *unused __unused)
243 LIST_INIT(&V_gif_softc_list);
245 VNET_SYSINIT(vnet_gif_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_gif_init,
249 gifmodevent(mod, type, data)
257 mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF);
258 if_clone_attach(&gif_cloner);
262 if_clone_detach(&gif_cloner);
263 mtx_destroy(&gif_mtx);
271 static moduledata_t gif_mod = {
277 DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
278 MODULE_VERSION(if_gif, 1);
281 gif_encapcheck(m, off, proto, arg)
282 const struct mbuf *m;
288 struct gif_softc *sc;
290 sc = (struct gif_softc *)arg;
294 if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
297 /* no physical address */
298 if (!sc->gif_psrc || !sc->gif_pdst)
310 case IPPROTO_ETHERIP:
317 /* Bail on short packets */
318 if (m->m_pkthdr.len < sizeof(ip))
321 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
326 if (sc->gif_psrc->sa_family != AF_INET ||
327 sc->gif_pdst->sa_family != AF_INET)
329 return gif_encapcheck4(m, off, proto, arg);
333 if (m->m_pkthdr.len < sizeof(struct ip6_hdr))
335 if (sc->gif_psrc->sa_family != AF_INET6 ||
336 sc->gif_pdst->sa_family != AF_INET6)
338 return gif_encapcheck6(m, off, proto, arg);
346 gif_start(struct ifnet *ifp)
348 struct gif_softc *sc;
353 ifp->if_drv_flags |= IFF_DRV_OACTIVE;
355 IFQ_DEQUEUE(&ifp->if_snd, m);
359 gif_output(ifp, m, sc->gif_pdst, NULL);
362 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
368 gif_output(ifp, m, dst, ro)
371 struct sockaddr *dst;
374 struct gif_softc *sc = ifp->if_softc;
381 error = mac_ifnet_check_transmit(ifp, m);
389 * gif may cause infinite recursion calls when misconfigured.
390 * We'll prevent this by detecting loops.
392 * High nesting level may cause stack exhaustion.
393 * We'll prevent this by introducing upper limit.
396 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL);
397 while (mtag != NULL) {
398 if (*(struct ifnet **)(mtag + 1) == ifp) {
400 "gif_output: loop detected on %s\n",
401 (*(struct ifnet **)(mtag + 1))->if_xname);
403 error = EIO; /* is there better errno? */
406 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag);
409 if (gif_called > V_max_gif_nesting) {
411 "gif_output: recursively called too many times(%d)\n",
414 error = EIO; /* is there better errno? */
417 mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *),
424 *(struct ifnet **)(mtag + 1) = ifp;
425 m_tag_prepend(m, mtag);
427 m->m_flags &= ~(M_BCAST|M_MCAST);
431 if (!(ifp->if_flags & IFF_UP) ||
432 sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
439 /* BPF writes need to be handled specially. */
440 if (dst->sa_family == AF_UNSPEC) {
441 bcopy(dst->sa_data, &af, sizeof(af));
446 BPF_MTAP2(ifp, &af, sizeof(af), m);
448 ifp->if_obytes += m->m_pkthdr.len;
450 /* override to IPPROTO_ETHERIP for bridged traffic */
454 M_SETFIB(m, sc->gif_fibnum);
455 /* inner AF-specific encapsulation */
457 /* XXX should we check if our outer source is legal? */
459 /* dispatch to output logic based on outer AF */
460 switch (sc->gif_psrc->sa_family) {
463 error = in_gif_output(ifp, af, m);
468 error = in6_gif_output(ifp, af, m);
484 gif_input(m, af, ifp)
490 struct gif_softc *sc = ifp->if_softc;
491 struct etherip_header *eip;
492 struct ether_header *eh;
493 struct ifnet *oldifp;
501 m->m_pkthdr.rcvif = ifp;
504 mac_ifnet_create_mbuf(ifp, m);
507 if (bpf_peers_present(ifp->if_bpf)) {
509 bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
512 if (ng_gif_input_p != NULL) {
513 (*ng_gif_input_p)(ifp, &m, af);
519 * Put the packet to the network layer input queue according to the
520 * specified address family.
521 * Note: older versions of gif_input directly called network layer
522 * input functions, e.g. ip6_input, here. We changed the policy to
523 * prevent too many recursive calls of such input functions, which
524 * might cause kernel panic. But the change may introduce another
525 * problem; if the input queue is full, packets are discarded.
526 * The kernel stack overflow really happened, and we believed
527 * queue-full rarely occurs, so we changed the policy.
541 n = sizeof(struct etherip_header) + sizeof(struct ether_header);
550 eip = mtod(m, struct etherip_header *);
552 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally
553 * accepts an EtherIP packet with revered version field in
554 * the header. This is a knob for backward compatibility
555 * with FreeBSD 7.2R or prior.
557 if (sc->gif_options & GIF_ACCEPT_REVETHIP) {
558 if (eip->eip_resvl != ETHERIP_VERSION
559 && eip->eip_ver != ETHERIP_VERSION) {
560 /* discard unknown versions */
565 if (eip->eip_ver != ETHERIP_VERSION) {
566 /* discard unknown versions */
571 m_adj(m, sizeof(struct etherip_header));
573 m->m_flags &= ~(M_BCAST|M_MCAST);
574 m->m_pkthdr.rcvif = ifp;
576 if (ifp->if_bridge) {
578 eh = mtod(m, struct ether_header *);
579 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
580 if (ETHER_IS_BROADCAST(eh->ether_dhost))
581 m->m_flags |= M_BCAST;
583 m->m_flags |= M_MCAST;
586 BRIDGE_INPUT(ifp, m);
588 if (m != NULL && ifp != oldifp) {
590 * The bridge gave us back itself or one of the
591 * members for which the frame is addressed.
602 if (ng_gif_input_orphan_p != NULL)
603 (*ng_gif_input_orphan_p)(ifp, m, af);
610 ifp->if_ibytes += m->m_pkthdr.len;
611 netisr_dispatch(isr, m);
614 /* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
616 gif_ioctl(ifp, cmd, data)
621 struct gif_softc *sc = ifp->if_softc;
622 struct ifreq *ifr = (struct ifreq*)data;
625 struct sockaddr *dst, *src;
626 #ifdef SIOCSIFMTU /* xxx */
632 ifp->if_flags |= IFF_UP;
642 #ifdef SIOCSIFMTU /* xxx */
648 if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX)
652 #endif /* SIOCSIFMTU */
658 case SIOCSIFPHYADDR_IN6:
660 case SIOCSLIFPHYADDR:
664 src = (struct sockaddr *)
665 &(((struct in_aliasreq *)data)->ifra_addr);
666 dst = (struct sockaddr *)
667 &(((struct in_aliasreq *)data)->ifra_dstaddr);
671 case SIOCSIFPHYADDR_IN6:
672 src = (struct sockaddr *)
673 &(((struct in6_aliasreq *)data)->ifra_addr);
674 dst = (struct sockaddr *)
675 &(((struct in6_aliasreq *)data)->ifra_dstaddr);
678 case SIOCSLIFPHYADDR:
679 src = (struct sockaddr *)
680 &(((struct if_laddrreq *)data)->addr);
681 dst = (struct sockaddr *)
682 &(((struct if_laddrreq *)data)->dstaddr);
688 /* sa_family must be equal */
689 if (src->sa_family != dst->sa_family)
692 /* validate sa_len */
693 switch (src->sa_family) {
696 if (src->sa_len != sizeof(struct sockaddr_in))
702 if (src->sa_len != sizeof(struct sockaddr_in6))
709 switch (dst->sa_family) {
712 if (dst->sa_len != sizeof(struct sockaddr_in))
718 if (dst->sa_len != sizeof(struct sockaddr_in6))
726 /* check sa_family looks sane for the cmd */
729 if (src->sa_family == AF_INET)
733 case SIOCSIFPHYADDR_IN6:
734 if (src->sa_family == AF_INET6)
738 case SIOCSLIFPHYADDR:
739 /* checks done in the above */
743 error = gif_set_tunnel(GIF2IFP(sc), src, dst);
746 #ifdef SIOCDIFPHYADDR
748 gif_delete_tunnel(GIF2IFP(sc));
752 case SIOCGIFPSRCADDR:
754 case SIOCGIFPSRCADDR_IN6:
756 if (sc->gif_psrc == NULL) {
757 error = EADDRNOTAVAIL;
763 case SIOCGIFPSRCADDR:
764 dst = &ifr->ifr_addr;
765 size = sizeof(ifr->ifr_addr);
769 case SIOCGIFPSRCADDR_IN6:
770 dst = (struct sockaddr *)
771 &(((struct in6_ifreq *)data)->ifr_addr);
772 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
776 error = EADDRNOTAVAIL;
779 if (src->sa_len > size)
781 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
783 if (dst->sa_family == AF_INET6) {
784 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
791 case SIOCGIFPDSTADDR:
793 case SIOCGIFPDSTADDR_IN6:
795 if (sc->gif_pdst == NULL) {
796 error = EADDRNOTAVAIL;
802 case SIOCGIFPDSTADDR:
803 dst = &ifr->ifr_addr;
804 size = sizeof(ifr->ifr_addr);
808 case SIOCGIFPDSTADDR_IN6:
809 dst = (struct sockaddr *)
810 &(((struct in6_ifreq *)data)->ifr_addr);
811 size = sizeof(((struct in6_ifreq *)data)->ifr_addr);
815 error = EADDRNOTAVAIL;
818 if (src->sa_len > size)
820 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
822 if (dst->sa_family == AF_INET6) {
823 error = sa6_recoverscope((struct sockaddr_in6 *)dst);
830 case SIOCGLIFPHYADDR:
831 if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) {
832 error = EADDRNOTAVAIL;
838 dst = (struct sockaddr *)
839 &(((struct if_laddrreq *)data)->addr);
840 size = sizeof(((struct if_laddrreq *)data)->addr);
841 if (src->sa_len > size)
843 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
847 dst = (struct sockaddr *)
848 &(((struct if_laddrreq *)data)->dstaddr);
849 size = sizeof(((struct if_laddrreq *)data)->dstaddr);
850 if (src->sa_len > size)
852 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len);
856 /* if_ioctl() takes care of it */
860 options = sc->gif_options;
861 error = copyout(&options, ifr->ifr_data,
866 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
868 error = copyin(ifr->ifr_data, &options, sizeof(options));
871 if (options & ~GIF_OPTMASK)
874 sc->gif_options = options;
886 * XXXRW: There's a general event-ordering issue here: the code to check
887 * if a given tunnel is already present happens before we perform a
888 * potentially blocking setup of the tunnel. This code needs to be
889 * re-ordered so that the check and replacement can be atomic using
893 gif_set_tunnel(ifp, src, dst)
895 struct sockaddr *src;
896 struct sockaddr *dst;
898 struct gif_softc *sc = ifp->if_softc;
899 struct gif_softc *sc2;
900 struct sockaddr *osrc, *odst, *sa;
904 LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) {
907 if (!sc2->gif_pdst || !sc2->gif_psrc)
909 if (sc2->gif_pdst->sa_family != dst->sa_family ||
910 sc2->gif_pdst->sa_len != dst->sa_len ||
911 sc2->gif_psrc->sa_family != src->sa_family ||
912 sc2->gif_psrc->sa_len != src->sa_len)
916 * Disallow parallel tunnels unless instructed
919 if (!V_parallel_tunnels &&
920 bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 &&
921 bcmp(sc2->gif_psrc, src, src->sa_len) == 0) {
922 error = EADDRNOTAVAIL;
923 mtx_unlock(&gif_mtx);
927 /* XXX both end must be valid? (I mean, not 0.0.0.0) */
929 mtx_unlock(&gif_mtx);
931 /* XXX we can detach from both, but be polite just in case */
933 switch (sc->gif_psrc->sa_family) {
936 (void)in_gif_detach(sc);
941 (void)in6_gif_detach(sc);
947 sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK);
948 bcopy((caddr_t)src, (caddr_t)sa, src->sa_len);
952 sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK);
953 bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len);
956 switch (sc->gif_psrc->sa_family) {
959 error = in_gif_attach(sc);
965 * Check validity of the scope zone ID of the addresses, and
966 * convert it into the kernel internal form if necessary.
968 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0);
971 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0);
974 error = in6_gif_attach(sc);
980 free((caddr_t)sc->gif_psrc, M_IFADDR);
981 free((caddr_t)sc->gif_pdst, M_IFADDR);
988 free((caddr_t)osrc, M_IFADDR);
990 free((caddr_t)odst, M_IFADDR);
993 if (sc->gif_psrc && sc->gif_pdst)
994 ifp->if_drv_flags |= IFF_DRV_RUNNING;
996 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1002 gif_delete_tunnel(ifp)
1005 struct gif_softc *sc = ifp->if_softc;
1008 free((caddr_t)sc->gif_psrc, M_IFADDR);
1009 sc->gif_psrc = NULL;
1012 free((caddr_t)sc->gif_pdst, M_IFADDR);
1013 sc->gif_pdst = NULL;
1015 /* it is safe to detach from both */
1017 (void)in_gif_detach(sc);
1020 (void)in6_gif_detach(sc);
1022 ifp->if_drv_flags &= ~IFF_DRV_RUNNING;