2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Copyright (c) 2018 Andrey V. Elsukov <ae@FreeBSD.org>
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
32 * $KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
39 #include "opt_inet6.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
45 #include <sys/malloc.h>
47 #include <sys/module.h>
48 #include <sys/rmlock.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
52 #include <sys/errno.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
59 #include <machine/cpu.h>
62 #include <net/if_var.h>
63 #include <net/if_clone.h>
64 #include <net/if_types.h>
65 #include <net/netisr.h>
66 #include <net/route.h>
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/ip_ecn.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip_var.h>
81 #include <netinet/in.h>
83 #include <netinet6/in6_var.h>
84 #include <netinet/ip6.h>
85 #include <netinet6/ip6_ecn.h>
86 #include <netinet6/ip6_var.h>
89 #include <netinet/ip_encap.h>
90 #include <net/ethernet.h>
91 #include <net/if_bridgevar.h>
92 #include <net/if_gif.h>
94 #include <security/mac/mac_framework.h>
96 static const char gifname[] = "gif";
98 MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface");
99 static struct sx gif_ioctl_sx;
100 SX_SYSINIT(gif_ioctl_sx, &gif_ioctl_sx, "gif_ioctl");
102 void (*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af);
103 void (*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af);
104 void (*ng_gif_attach_p)(struct ifnet *ifp);
105 void (*ng_gif_detach_p)(struct ifnet *ifp);
107 static void gif_delete_tunnel(struct gif_softc *);
108 static int gif_ioctl(struct ifnet *, u_long, caddr_t);
109 static int gif_transmit(struct ifnet *, struct mbuf *);
110 static void gif_qflush(struct ifnet *);
111 static int gif_clone_create(struct if_clone *, int, caddr_t);
112 static void gif_clone_destroy(struct ifnet *);
113 VNET_DEFINE_STATIC(struct if_clone *, gif_cloner);
114 #define V_gif_cloner VNET(gif_cloner)
116 SYSCTL_DECL(_net_link);
117 static SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0,
118 "Generic Tunnel Interface");
121 * This macro controls the default upper limitation on nesting of gif tunnels.
122 * Since, setting a large value to this macro with a careless configuration
123 * may introduce system crash, we don't allow any nestings by default.
124 * If you need to configure nested gif tunnels, you can define this macro
125 * in your kernel configuration file. However, if you do so, please be
126 * careful to configure the tunnels so that it won't make a loop.
128 #define MAX_GIF_NEST 1
130 VNET_DEFINE_STATIC(int, max_gif_nesting) = MAX_GIF_NEST;
131 #define V_max_gif_nesting VNET(max_gif_nesting)
132 SYSCTL_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_VNET | CTLFLAG_RW,
133 &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels");
136 gif_clone_create(struct if_clone *ifc, int unit, caddr_t params)
138 struct gif_softc *sc;
140 sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO);
141 sc->gif_fibnum = curthread->td_proc->p_fibnum;
142 GIF2IFP(sc) = if_alloc(IFT_GIF);
143 GIF2IFP(sc)->if_softc = sc;
144 if_initname(GIF2IFP(sc), gifname, unit);
146 GIF2IFP(sc)->if_addrlen = 0;
147 GIF2IFP(sc)->if_mtu = GIF_MTU;
148 GIF2IFP(sc)->if_flags = IFF_POINTOPOINT | IFF_MULTICAST;
149 GIF2IFP(sc)->if_ioctl = gif_ioctl;
150 GIF2IFP(sc)->if_transmit = gif_transmit;
151 GIF2IFP(sc)->if_qflush = gif_qflush;
152 GIF2IFP(sc)->if_output = gif_output;
153 GIF2IFP(sc)->if_capabilities |= IFCAP_LINKSTATE;
154 GIF2IFP(sc)->if_capenable |= IFCAP_LINKSTATE;
155 if_attach(GIF2IFP(sc));
156 bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t));
157 if (ng_gif_attach_p != NULL)
158 (*ng_gif_attach_p)(GIF2IFP(sc));
164 gif_clone_destroy(struct ifnet *ifp)
166 struct gif_softc *sc;
168 sx_xlock(&gif_ioctl_sx);
170 gif_delete_tunnel(sc);
171 if (ng_gif_detach_p != NULL)
172 (*ng_gif_detach_p)(ifp);
175 ifp->if_softc = NULL;
176 sx_xunlock(&gif_ioctl_sx);
184 vnet_gif_init(const void *unused __unused)
187 V_gif_cloner = if_clone_simple(gifname, gif_clone_create,
188 gif_clone_destroy, 0);
196 VNET_SYSINIT(vnet_gif_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
197 vnet_gif_init, NULL);
200 vnet_gif_uninit(const void *unused __unused)
203 if_clone_detach(V_gif_cloner);
211 VNET_SYSUNINIT(vnet_gif_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
212 vnet_gif_uninit, NULL);
215 gifmodevent(module_t mod, int type, void *data)
228 static moduledata_t gif_mod = {
234 DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
235 MODULE_VERSION(if_gif, 1);
240 struct gif_list *hash;
243 hash = malloc(sizeof(struct gif_list) * GIF_HASH_SIZE,
245 for (i = 0; i < GIF_HASH_SIZE; i++)
246 CK_LIST_INIT(&hash[i]);
252 gif_hashdestroy(struct gif_list *hash)
258 #define MTAG_GIF 1080679712
260 gif_transmit(struct ifnet *ifp, struct mbuf *m)
262 struct gif_softc *sc;
263 struct etherip_header *eth;
277 error = mac_ifnet_check_transmit(ifp, m);
285 if ((ifp->if_flags & IFF_MONITOR) != 0 ||
286 (ifp->if_flags & IFF_UP) == 0 ||
287 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
288 sc->gif_family == 0 ||
289 (error = if_tunnel_check_nesting(ifp, m, MTAG_GIF,
290 V_max_gif_nesting)) != 0) {
294 /* Now pull back the af that we stashed in the csum_data. */
298 af = m->m_pkthdr.csum_data;
299 m->m_flags &= ~(M_BCAST|M_MCAST);
300 M_SETFIB(m, sc->gif_fibnum);
301 BPF_MTAP2(ifp, &af, sizeof(af), m);
302 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
303 if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
304 /* inner AF-specific encapsulation */
309 proto = IPPROTO_IPV4;
310 if (m->m_len < sizeof(struct ip))
311 m = m_pullup(m, sizeof(struct ip));
316 ip = mtod(m, struct ip *);
317 ip_ecn_ingress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
318 ECN_NOCARE, &ecn, &ip->ip_tos);
323 proto = IPPROTO_IPV6;
324 if (m->m_len < sizeof(struct ip6_hdr))
325 m = m_pullup(m, sizeof(struct ip6_hdr));
331 ip6 = mtod(m, struct ip6_hdr *);
332 ip6_ecn_ingress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
333 ECN_NOCARE, &t, &ip6->ip6_flow);
334 ecn = (ntohl(t) >> 20) & 0xff;
338 proto = IPPROTO_ETHERIP;
339 M_PREPEND(m, sizeof(struct etherip_header), M_NOWAIT);
344 eth = mtod(m, struct etherip_header *);
346 eth->eip_ver = ETHERIP_VERSION;
350 error = EAFNOSUPPORT;
354 /* XXX should we check if our outer source is legal? */
355 /* dispatch to output logic based on outer AF */
356 switch (sc->gif_family) {
359 error = in_gif_output(ifp, m, proto, ecn);
364 error = in6_gif_output(ifp, m, proto, ecn);
372 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
378 gif_qflush(struct ifnet *ifp __unused)
385 gif_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
390 if (dst->sa_family == AF_UNSPEC)
391 bcopy(dst->sa_data, &af, sizeof(af));
395 * Now save the af in the inbound pkt csum data, this is a cheat since
396 * we are using the inbound csum_data field to carry the af over to
397 * the gif_transmit() routine, avoiding using yet another mtag.
399 m->m_pkthdr.csum_data = af;
400 return (ifp->if_transmit(ifp, m));
404 gif_input(struct mbuf *m, struct ifnet *ifp, int proto, uint8_t ecn)
406 struct etherip_header *eip;
414 struct ether_header *eh;
415 struct ifnet *oldifp;
423 m->m_pkthdr.rcvif = ifp;
429 if (m->m_len < sizeof(struct ip))
430 m = m_pullup(m, sizeof(struct ip));
433 ip = mtod(m, struct ip *);
434 if (ip_ecn_egress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
435 ECN_NOCARE, &ecn, &ip->ip_tos) == 0) {
444 if (m->m_len < sizeof(struct ip6_hdr))
445 m = m_pullup(m, sizeof(struct ip6_hdr));
448 t = htonl((uint32_t)ecn << 20);
449 ip6 = mtod(m, struct ip6_hdr *);
450 if (ip6_ecn_egress((ifp->if_flags & IFF_LINK1) ? ECN_ALLOWED:
451 ECN_NOCARE, &t, &ip6->ip6_flow) == 0) {
457 case IPPROTO_ETHERIP:
466 mac_ifnet_create_mbuf(ifp, m);
469 if (bpf_peers_present(ifp->if_bpf)) {
471 bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m);
474 if ((ifp->if_flags & IFF_MONITOR) != 0) {
475 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
476 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
481 if (ng_gif_input_p != NULL) {
482 (*ng_gif_input_p)(ifp, &m, af);
488 * Put the packet to the network layer input queue according to the
489 * specified address family.
490 * Note: older versions of gif_input directly called network layer
491 * input functions, e.g. ip6_input, here. We changed the policy to
492 * prevent too many recursive calls of such input functions, which
493 * might cause kernel panic. But the change may introduce another
494 * problem; if the input queue is full, packets are discarded.
495 * The kernel stack overflow really happened, and we believed
496 * queue-full rarely occurs, so we changed the policy.
510 n = sizeof(struct etherip_header) +
511 sizeof(struct ether_header);
516 eip = mtod(m, struct etherip_header *);
517 if (eip->eip_ver != ETHERIP_VERSION) {
518 /* discard unknown versions */
522 m_adj(m, sizeof(struct etherip_header));
524 m->m_flags &= ~(M_BCAST|M_MCAST);
525 m->m_pkthdr.rcvif = ifp;
527 if (ifp->if_bridge) {
529 eh = mtod(m, struct ether_header *);
530 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
531 if (ETHER_IS_BROADCAST(eh->ether_dhost))
532 m->m_flags |= M_BCAST;
534 m->m_flags |= M_MCAST;
535 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
537 BRIDGE_INPUT(ifp, m);
539 if (m != NULL && ifp != oldifp) {
541 * The bridge gave us back itself or one of the
542 * members for which the frame is addressed.
553 if (ng_gif_input_orphan_p != NULL)
554 (*ng_gif_input_orphan_p)(ifp, m, af);
560 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
561 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
562 M_SETFIB(m, ifp->if_fib);
563 netisr_dispatch(isr, m);
566 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
570 gif_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
572 struct ifreq *ifr = (struct ifreq*)data;
573 struct gif_softc *sc;
579 ifp->if_flags |= IFF_UP;
586 if (ifr->ifr_mtu < GIF_MTU_MIN ||
587 ifr->ifr_mtu > GIF_MTU_MAX)
590 ifp->if_mtu = ifr->ifr_mtu;
593 sx_xlock(&gif_ioctl_sx);
602 if (sc->gif_family == 0)
604 gif_delete_tunnel(sc);
608 case SIOCGIFPSRCADDR:
609 case SIOCGIFPDSTADDR:
610 error = in_gif_ioctl(sc, cmd, data);
614 case SIOCSIFPHYADDR_IN6:
615 case SIOCGIFPSRCADDR_IN6:
616 case SIOCGIFPDSTADDR_IN6:
617 error = in6_gif_ioctl(sc, cmd, data);
621 ifr->ifr_fib = sc->gif_fibnum;
624 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
626 if (ifr->ifr_fib >= rt_numfibs)
629 sc->gif_fibnum = ifr->ifr_fib;
632 options = sc->gif_options;
633 error = copyout(&options, ifr_data_get_ptr(ifr),
637 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0)
639 error = copyin(ifr_data_get_ptr(ifr), &options,
643 if (options & ~GIF_OPTMASK) {
647 if (sc->gif_options != options) {
648 switch (sc->gif_family) {
651 error = in_gif_setopts(sc, options);
656 error = in6_gif_setopts(sc, options);
660 /* No need to invoke AF-handler */
661 sc->gif_options = options;
669 if (error == 0 && sc->gif_family != 0) {
672 cmd == SIOCSIFPHYADDR ||
675 cmd == SIOCSIFPHYADDR_IN6 ||
678 if_link_state_change(ifp, LINK_STATE_UP);
682 sx_xunlock(&gif_ioctl_sx);
687 gif_delete_tunnel(struct gif_softc *sc)
690 sx_assert(&gif_ioctl_sx, SA_XLOCKED);
691 if (sc->gif_family != 0) {
692 CK_LIST_REMOVE(sc, srchash);
693 CK_LIST_REMOVE(sc, chain);
694 /* Wait until it become safe to free gif_hdr */
696 free(sc->gif_hdr, M_GIF);
699 GIF2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
700 if_link_state_change(GIF2IFP(sc), LINK_STATE_DOWN);