xref: /freebsd/sys/netinet6/in6_gif.c (revision 190cef3d) 
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * Copyright (c) 2018 Andrey V. Elsukov <ae@FreeBSD.org>
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  *	$KAME: in6_gif.c,v 1.49 2001/05/14 14:02:17 itojun Exp $
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/jail.h>
44 #include <sys/socket.h>
45 #include <sys/sockio.h>
46 #include <sys/mbuf.h>
47 #include <sys/errno.h>
48 #include <sys/kernel.h>
49 #include <sys/syslog.h>
50 #include <sys/sysctl.h>
51 #include <sys/malloc.h>
52 #include <sys/proc.h>
53 
54 #include <net/ethernet.h>
55 #include <net/if.h>
56 #include <net/if_var.h>
57 #include <net/route.h>
58 #include <net/vnet.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_systm.h>
62 #ifdef INET
63 #include <netinet/ip.h>
64 #include <netinet/ip_ecn.h>
65 #endif
66 #include <netinet/ip_encap.h>
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/scope6_var.h>
71 #include <netinet6/ip6_ecn.h>
72 #include <netinet6/in6_fib.h>
73 
74 #include <net/if_gif.h>
75 
76 #define GIF_HLIM	30
77 VNET_DEFINE_STATIC(int, ip6_gif_hlim) = GIF_HLIM;
78 #define	V_ip6_gif_hlim			VNET(ip6_gif_hlim)
79 
80 SYSCTL_DECL(_net_inet6_ip6);
81 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_GIF_HLIM, gifhlim,
82     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_gif_hlim), 0,
83     "Default hop limit for encapsulated packets");
84 
85 /*
86  * We keep interfaces in a hash table using src+dst as key.
87  * Interfaces with GIF_IGNORE_SOURCE flag are linked into plain list.
88  */
89 VNET_DEFINE_STATIC(struct gif_list *, ipv6_hashtbl) = NULL;
90 VNET_DEFINE_STATIC(struct gif_list, ipv6_list) = CK_LIST_HEAD_INITIALIZER();
91 #define	V_ipv6_hashtbl		VNET(ipv6_hashtbl)
92 #define	V_ipv6_list		VNET(ipv6_list)
93 
94 #define	GIF_HASH(src, dst)	(V_ipv6_hashtbl[\
95     in6_gif_hashval((src), (dst)) & (GIF_HASH_SIZE - 1)])
96 #define	GIF_HASH_SC(sc)		GIF_HASH(&(sc)->gif_ip6hdr->ip6_src,\
97     &(sc)->gif_ip6hdr->ip6_dst)
98 static uint32_t
99 in6_gif_hashval(const struct in6_addr *src, const struct in6_addr *dst)
100 {
101 	uint32_t ret;
102 
103 	ret = fnv_32_buf(src, sizeof(*src), FNV1_32_INIT);
104 	return (fnv_32_buf(dst, sizeof(*dst), ret));
105 }
106 
107 static int
108 in6_gif_checkdup(const struct gif_softc *sc, const struct in6_addr *src,
109     const struct in6_addr *dst)
110 {
111 	struct gif_softc *tmp;
112 
113 	if (sc->gif_family == AF_INET6 &&
114 	    IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_src, src) &&
115 	    IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_dst, dst))
116 		return (EEXIST);
117 
118 	CK_LIST_FOREACH(tmp, &GIF_HASH(src, dst), chain) {
119 		if (tmp == sc)
120 			continue;
121 		if (IN6_ARE_ADDR_EQUAL(&tmp->gif_ip6hdr->ip6_src, src) &&
122 		    IN6_ARE_ADDR_EQUAL(&tmp->gif_ip6hdr->ip6_dst, dst))
123 			return (EADDRNOTAVAIL);
124 	}
125 	return (0);
126 }
127 
128 static void
129 in6_gif_attach(struct gif_softc *sc)
130 {
131 
132 	if (sc->gif_options & GIF_IGNORE_SOURCE)
133 		CK_LIST_INSERT_HEAD(&V_ipv6_list, sc, chain);
134 	else
135 		CK_LIST_INSERT_HEAD(&GIF_HASH_SC(sc), sc, chain);
136 }
137 
138 int
139 in6_gif_setopts(struct gif_softc *sc, u_int options)
140 {
141 
142 	/* NOTE: we are protected with gif_ioctl_sx lock */
143 	MPASS(sc->gif_family == AF_INET6);
144 	MPASS(sc->gif_options != options);
145 
146 	if ((options & GIF_IGNORE_SOURCE) !=
147 	    (sc->gif_options & GIF_IGNORE_SOURCE)) {
148 		CK_LIST_REMOVE(sc, chain);
149 		sc->gif_options = options;
150 		in6_gif_attach(sc);
151 	}
152 	return (0);
153 }
154 
155 int
156 in6_gif_ioctl(struct gif_softc *sc, u_long cmd, caddr_t data)
157 {
158 	struct in6_ifreq *ifr = (struct in6_ifreq *)data;
159 	struct sockaddr_in6 *dst, *src;
160 	struct ip6_hdr *ip6;
161 	int error;
162 
163 	/* NOTE: we are protected with gif_ioctl_sx lock */
164 	error = EINVAL;
165 	switch (cmd) {
166 	case SIOCSIFPHYADDR_IN6:
167 		src = &((struct in6_aliasreq *)data)->ifra_addr;
168 		dst = &((struct in6_aliasreq *)data)->ifra_dstaddr;
169 
170 		/* sanity checks */
171 		if (src->sin6_family != dst->sin6_family ||
172 		    src->sin6_family != AF_INET6 ||
173 		    src->sin6_len != dst->sin6_len ||
174 		    src->sin6_len != sizeof(*src))
175 			break;
176 		if (IN6_IS_ADDR_UNSPECIFIED(&src->sin6_addr) ||
177 		    IN6_IS_ADDR_UNSPECIFIED(&dst->sin6_addr)) {
178 			error = EADDRNOTAVAIL;
179 			break;
180 		}
181 		/*
182 		 * Check validity of the scope zone ID of the
183 		 * addresses, and convert it into the kernel
184 		 * internal form if necessary.
185 		 */
186 		if ((error = sa6_embedscope(src, 0)) != 0 ||
187 		    (error = sa6_embedscope(dst, 0)) != 0)
188 			break;
189 
190 		if (V_ipv6_hashtbl == NULL)
191 			V_ipv6_hashtbl = gif_hashinit();
192 		error = in6_gif_checkdup(sc, &src->sin6_addr,
193 		    &dst->sin6_addr);
194 		if (error == EADDRNOTAVAIL)
195 			break;
196 		if (error == EEXIST) {
197 			/* Addresses are the same. Just return. */
198 			error = 0;
199 			break;
200 		}
201 		ip6 = malloc(sizeof(*ip6), M_GIF, M_WAITOK | M_ZERO);
202 		ip6->ip6_src = src->sin6_addr;
203 		ip6->ip6_dst = dst->sin6_addr;
204 		ip6->ip6_vfc = IPV6_VERSION;
205 		if (sc->gif_family != 0) {
206 			/* Detach existing tunnel first */
207 			CK_LIST_REMOVE(sc, chain);
208 			GIF_WAIT();
209 			free(sc->gif_hdr, M_GIF);
210 			/* XXX: should we notify about link state change? */
211 		}
212 		sc->gif_family = AF_INET6;
213 		sc->gif_ip6hdr = ip6;
214 		in6_gif_attach(sc);
215 		break;
216 	case SIOCGIFPSRCADDR_IN6:
217 	case SIOCGIFPDSTADDR_IN6:
218 		if (sc->gif_family != AF_INET6) {
219 			error = EADDRNOTAVAIL;
220 			break;
221 		}
222 		src = (struct sockaddr_in6 *)&ifr->ifr_addr;
223 		memset(src, 0, sizeof(*src));
224 		src->sin6_family = AF_INET6;
225 		src->sin6_len = sizeof(*src);
226 		src->sin6_addr = (cmd == SIOCGIFPSRCADDR_IN6) ?
227 		    sc->gif_ip6hdr->ip6_src: sc->gif_ip6hdr->ip6_dst;
228 		error = prison_if(curthread->td_ucred, (struct sockaddr *)src);
229 		if (error == 0)
230 			error = sa6_recoverscope(src);
231 		if (error != 0)
232 			memset(src, 0, sizeof(*src));
233 		break;
234 	}
235 	return (error);
236 }
237 
238 int
239 in6_gif_output(struct ifnet *ifp, struct mbuf *m, int proto, uint8_t ecn)
240 {
241 	struct gif_softc *sc = ifp->if_softc;
242 	struct ip6_hdr *ip6;
243 	int len;
244 
245 	/* prepend new IP header */
246 	MPASS(in_epoch(net_epoch_preempt));
247 	len = sizeof(struct ip6_hdr);
248 #ifndef __NO_STRICT_ALIGNMENT
249 	if (proto == IPPROTO_ETHERIP)
250 		len += ETHERIP_ALIGN;
251 #endif
252 	M_PREPEND(m, len, M_NOWAIT);
253 	if (m == NULL)
254 		return (ENOBUFS);
255 #ifndef __NO_STRICT_ALIGNMENT
256 	if (proto == IPPROTO_ETHERIP) {
257 		len = mtod(m, vm_offset_t) & 3;
258 		KASSERT(len == 0 || len == ETHERIP_ALIGN,
259 		    ("in6_gif_output: unexpected misalignment"));
260 		m->m_data += len;
261 		m->m_len -= ETHERIP_ALIGN;
262 	}
263 #endif
264 
265 	ip6 = mtod(m, struct ip6_hdr *);
266 	MPASS(sc->gif_family == AF_INET6);
267 	bcopy(sc->gif_ip6hdr, ip6, sizeof(struct ip6_hdr));
268 
269 	ip6->ip6_flow  |= htonl((uint32_t)ecn << 20);
270 	ip6->ip6_nxt	= proto;
271 	ip6->ip6_hlim	= V_ip6_gif_hlim;
272 	/*
273 	 * force fragmentation to minimum MTU, to avoid path MTU discovery.
274 	 * it is too painful to ask for resend of inner packet, to achieve
275 	 * path MTU discovery for encapsulated packets.
276 	 */
277 	return (ip6_output(m, 0, NULL, IPV6_MINMTU, 0, NULL, NULL));
278 }
279 
280 static int
281 in6_gif_input(struct mbuf *m, int off, int proto, void *arg)
282 {
283 	struct gif_softc *sc = arg;
284 	struct ifnet *gifp;
285 	struct ip6_hdr *ip6;
286 	uint8_t ecn;
287 
288 	MPASS(in_epoch(net_epoch_preempt));
289 	if (sc == NULL) {
290 		m_freem(m);
291 		IP6STAT_INC(ip6s_nogif);
292 		return (IPPROTO_DONE);
293 	}
294 	gifp = GIF2IFP(sc);
295 	if ((gifp->if_flags & IFF_UP) != 0) {
296 		ip6 = mtod(m, struct ip6_hdr *);
297 		ecn = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
298 		m_adj(m, off);
299 		gif_input(m, gifp, proto, ecn);
300 	} else {
301 		m_freem(m);
302 		IP6STAT_INC(ip6s_nogif);
303 	}
304 	return (IPPROTO_DONE);
305 }
306 
307 static int
308 in6_gif_lookup(const struct mbuf *m, int off, int proto, void **arg)
309 {
310 	const struct ip6_hdr *ip6;
311 	struct gif_softc *sc;
312 	int ret;
313 
314 	if (V_ipv6_hashtbl == NULL)
315 		return (0);
316 
317 	MPASS(in_epoch(net_epoch_preempt));
318 	/*
319 	 * NOTE: it is safe to iterate without any locking here, because softc
320 	 * can be reclaimed only when we are not within net_epoch_preempt
321 	 * section, but ip_encap lookup+input are executed in epoch section.
322 	 */
323 	ip6 = mtod(m, const struct ip6_hdr *);
324 	ret = 0;
325 	CK_LIST_FOREACH(sc, &GIF_HASH(&ip6->ip6_dst, &ip6->ip6_src), chain) {
326 		/*
327 		 * This is an inbound packet, its ip6_dst is source address
328 		 * in softc.
329 		 */
330 		if (IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_src,
331 		    &ip6->ip6_dst) &&
332 		    IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_dst,
333 		    &ip6->ip6_src)) {
334 			ret = ENCAP_DRV_LOOKUP;
335 			goto done;
336 		}
337 	}
338 	/*
339 	 * No exact match.
340 	 * Check the list of interfaces with GIF_IGNORE_SOURCE flag.
341 	 */
342 	CK_LIST_FOREACH(sc, &V_ipv6_list, chain) {
343 		if (IN6_ARE_ADDR_EQUAL(&sc->gif_ip6hdr->ip6_src,
344 		    &ip6->ip6_dst)) {
345 			ret = 128 + 8; /* src + proto */
346 			goto done;
347 		}
348 	}
349 	return (0);
350 done:
351 	if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0)
352 		return (0);
353 	/* ingress filters on outer source */
354 	if ((GIF2IFP(sc)->if_flags & IFF_LINK2) == 0) {
355 		struct nhop6_basic nh6;
356 
357 		if (fib6_lookup_nh_basic(sc->gif_fibnum, &ip6->ip6_src,
358 		    ntohs(in6_getscope(&ip6->ip6_src)), 0, 0, &nh6) != 0)
359 			return (0);
360 
361 		if (nh6.nh_ifp != m->m_pkthdr.rcvif)
362 			return (0);
363 	}
364 	*arg = sc;
365 	return (ret);
366 }
367 
368 static struct {
369 	const struct encap_config encap;
370 	const struct encaptab *cookie;
371 } ipv6_encap_cfg[] = {
372 #ifdef INET
373 	{
374 		.encap = {
375 			.proto = IPPROTO_IPV4,
376 			.min_length = sizeof(struct ip6_hdr) +
377 			    sizeof(struct ip),
378 			.exact_match = ENCAP_DRV_LOOKUP,
379 			.lookup = in6_gif_lookup,
380 			.input = in6_gif_input
381 		},
382 	},
383 #endif
384 	{
385 		.encap = {
386 			.proto = IPPROTO_IPV6,
387 			.min_length = 2 * sizeof(struct ip6_hdr),
388 			.exact_match = ENCAP_DRV_LOOKUP,
389 			.lookup = in6_gif_lookup,
390 			.input = in6_gif_input
391 		},
392 	},
393 	{
394 		.encap = {
395 			.proto = IPPROTO_ETHERIP,
396 			.min_length = sizeof(struct ip6_hdr) +
397 			    sizeof(struct etherip_header) +
398 			    sizeof(struct ether_header),
399 			.exact_match = ENCAP_DRV_LOOKUP,
400 			.lookup = in6_gif_lookup,
401 			.input = in6_gif_input
402 		},
403 	}
404 };
405 
406 void
407 in6_gif_init(void)
408 {
409 	int i;
410 
411 	if (!IS_DEFAULT_VNET(curvnet))
412 		return;
413 	for (i = 0; i < nitems(ipv6_encap_cfg); i++)
414 		ipv6_encap_cfg[i].cookie = ip6_encap_attach(
415 		    &ipv6_encap_cfg[i].encap, NULL, M_WAITOK);
416 }
417 
418 void
419 in6_gif_uninit(void)
420 {
421 	int i;
422 
423 	if (IS_DEFAULT_VNET(curvnet)) {
424 		for (i = 0; i < nitems(ipv6_encap_cfg); i++)
425 			ip6_encap_detach(ipv6_encap_cfg[i].cookie);
426 	}
427 	if (V_ipv6_hashtbl != NULL)
428 		gif_hashdestroy(V_ipv6_hashtbl);
429 }
430