1/*-
2 * Copyright (c) 2014, 2018 Andrey V. Elsukov <ae@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD$");
29
30#include "opt_inet.h"
31#include "opt_inet6.h"
32
33#include <sys/param.h>
34#include <sys/jail.h>
35#include <sys/systm.h>
36#include <sys/socket.h>
37#include <sys/socketvar.h>
38#include <sys/sockio.h>
39#include <sys/mbuf.h>
40#include <sys/errno.h>
41#include <sys/kernel.h>
42#include <sys/sysctl.h>
43#include <sys/malloc.h>
44#include <sys/proc.h>
45
46#include <net/if.h>
47#include <net/if_var.h>
48#include <net/vnet.h>
49
50#include <netinet/in.h>
51#ifdef INET
52#include <net/ethernet.h>
53#include <netinet/ip.h>
54#endif
55#include <netinet/in_pcb.h>
56#include <netinet/ip_encap.h>
57#include <netinet/ip_var.h>
58#include <netinet/ip6.h>
59#include <netinet/udp.h>
60#include <netinet/udp_var.h>
61#include <netinet6/ip6_var.h>
62#include <netinet6/in6_var.h>
63#include <netinet6/scope6_var.h>
64#include <net/if_gre.h>
65
66VNET_DEFINE(int, ip6_gre_hlim) = IPV6_DEFHLIM;
67#define	V_ip6_gre_hlim		VNET(ip6_gre_hlim)
68
69SYSCTL_DECL(_net_inet6_ip6);
70SYSCTL_INT(_net_inet6_ip6, OID_AUTO, grehlim, CTLFLAG_VNET | CTLFLAG_RW,
71    &VNET_NAME(ip6_gre_hlim), 0, "Default hop limit for encapsulated packets");
72
73struct in6_gre_socket {
74	struct gre_socket	base;
75	struct in6_addr		addr; /* scope zone id is embedded */
76};
77VNET_DEFINE_STATIC(struct gre_sockets *, ipv6_sockets) = NULL;
78VNET_DEFINE_STATIC(struct gre_list *, ipv6_hashtbl) = NULL;
79VNET_DEFINE_STATIC(struct gre_list *, ipv6_srchashtbl) = NULL;
80#define	V_ipv6_sockets		VNET(ipv6_sockets)
81#define	V_ipv6_hashtbl		VNET(ipv6_hashtbl)
82#define	V_ipv6_srchashtbl	VNET(ipv6_srchashtbl)
83#define	GRE_HASH(src, dst)	(V_ipv6_hashtbl[\
84    in6_gre_hashval((src), (dst)) & (GRE_HASH_SIZE - 1)])
85#define	GRE_SRCHASH(src)	(V_ipv6_srchashtbl[\
86    fnv_32_buf((src), sizeof(*src), FNV1_32_INIT) & (GRE_HASH_SIZE - 1)])
87#define	GRE_SOCKHASH(src)	(V_ipv6_sockets[\
88    fnv_32_buf((src), sizeof(*src), FNV1_32_INIT) & (GRE_HASH_SIZE - 1)])
89#define	GRE_HASH_SC(sc)		GRE_HASH(&(sc)->gre_oip6.ip6_src,\
90    &(sc)->gre_oip6.ip6_dst)
91
92static uint32_t
93in6_gre_hashval(const struct in6_addr *src, const struct in6_addr *dst)
94{
95	uint32_t ret;
96
97	ret = fnv_32_buf(src, sizeof(*src), FNV1_32_INIT);
98	return (fnv_32_buf(dst, sizeof(*dst), ret));
99}
100
101static struct gre_socket*
102in6_gre_lookup_socket(const struct in6_addr *addr)
103{
104	struct gre_socket *gs;
105	struct in6_gre_socket *s;
106
107	CK_LIST_FOREACH(gs, &GRE_SOCKHASH(addr), chain) {
108		s = __containerof(gs, struct in6_gre_socket, base);
109		if (IN6_ARE_ADDR_EQUAL(&s->addr, addr))
110			break;
111	}
112	return (gs);
113}
114
115static int
116in6_gre_checkdup(const struct gre_softc *sc, const struct in6_addr *src,
117    const struct in6_addr *dst, uint32_t opts)
118{
119	struct gre_list *head;
120	struct gre_softc *tmp;
121	struct gre_socket *gs;
122
123	if (sc->gre_family == AF_INET6 &&
124	    IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src, src) &&
125	    IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst, dst) &&
126	    (sc->gre_options & GRE_UDPENCAP) == (opts & GRE_UDPENCAP))
127		return (EEXIST);
128
129	if (opts & GRE_UDPENCAP) {
130		gs = in6_gre_lookup_socket(src);
131		if (gs == NULL)
132			return (0);
133		head = &gs->list;
134	} else
135		head = &GRE_HASH(src, dst);
136
137	CK_LIST_FOREACH(tmp, head, chain) {
138		if (tmp == sc)
139			continue;
140		if (IN6_ARE_ADDR_EQUAL(&tmp->gre_oip6.ip6_src, src) &&
141		    IN6_ARE_ADDR_EQUAL(&tmp->gre_oip6.ip6_dst, dst))
142			return (EADDRNOTAVAIL);
143	}
144	return (0);
145}
146
147static int
148in6_gre_lookup(const struct mbuf *m, int off, int proto, void **arg)
149{
150	const struct ip6_hdr *ip6;
151	struct gre_softc *sc;
152
153	if (V_ipv6_hashtbl == NULL)
154		return (0);
155
156	MPASS(in_epoch(net_epoch_preempt));
157	ip6 = mtod(m, const struct ip6_hdr *);
158	CK_LIST_FOREACH(sc, &GRE_HASH(&ip6->ip6_dst, &ip6->ip6_src), chain) {
159		/*
160		 * This is an inbound packet, its ip6_dst is source address
161		 * in softc.
162		 */
163		if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src,
164		    &ip6->ip6_dst) &&
165		    IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst,
166		    &ip6->ip6_src)) {
167			if ((GRE2IFP(sc)->if_flags & IFF_UP) == 0)
168				return (0);
169			*arg = sc;
170			return (ENCAP_DRV_LOOKUP);
171		}
172	}
173	return (0);
174}
175
176/*
177 * Check that ingress address belongs to local host.
178 */
179static void
180in6_gre_set_running(struct gre_softc *sc)
181{
182
183	if (in6_localip(&sc->gre_oip6.ip6_src))
184		GRE2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING;
185	else
186		GRE2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
187}
188
189/*
190 * ifaddr_event handler.
191 * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
192 * source address spoofing.
193 */
194static void
195in6_gre_srcaddr(void *arg __unused, const struct sockaddr *sa,
196    int event __unused)
197{
198	const struct sockaddr_in6 *sin;
199	struct gre_softc *sc;
200
201	/* Check that VNET is ready */
202	if (V_ipv6_hashtbl == NULL)
203		return;
204
205	MPASS(in_epoch(net_epoch_preempt));
206	sin = (const struct sockaddr_in6 *)sa;
207	CK_LIST_FOREACH(sc, &GRE_SRCHASH(&sin->sin6_addr), srchash) {
208		if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src,
209		    &sin->sin6_addr) == 0)
210			continue;
211		in6_gre_set_running(sc);
212	}
213}
214
215static void
216in6_gre_udp_input(struct mbuf *m, int off, struct inpcb *inp,
217    const struct sockaddr *sa, void *ctx)
218{
219	struct epoch_tracker et;
220	struct gre_socket *gs;
221	struct gre_softc *sc;
222	struct sockaddr_in6 dst;
223
224	NET_EPOCH_ENTER_ET(et);
225	/*
226	 * udp_append() holds reference to inp, it is safe to check
227	 * inp_flags2 without INP_RLOCK().
228	 * If socket was closed before we have entered NET_EPOCH section,
229	 * INP_FREED flag should be set. Otherwise it should be safe to
230	 * make access to ctx data, because gre_so will be freed by
231	 * gre_sofree() via epoch_call().
232	 */
233	if (__predict_false(inp->inp_flags2 & INP_FREED)) {
234		NET_EPOCH_EXIT_ET(et);
235		m_freem(m);
236		return;
237	}
238
239	gs = (struct gre_socket *)ctx;
240	dst = *(const struct sockaddr_in6 *)sa;
241	if (sa6_embedscope(&dst, 0)) {
242		NET_EPOCH_EXIT_ET(et);
243		m_freem(m);
244		return;
245	}
246	CK_LIST_FOREACH(sc, &gs->list, chain) {
247		if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst, &dst.sin6_addr))
248			break;
249	}
250	if (sc != NULL && (GRE2IFP(sc)->if_flags & IFF_UP) != 0){
251		gre_input(m, off + sizeof(struct udphdr), IPPROTO_UDP, sc);
252		NET_EPOCH_EXIT_ET(et);
253		return;
254	}
255	m_freem(m);
256	NET_EPOCH_EXIT_ET(et);
257}
258
259static int
260in6_gre_setup_socket(struct gre_softc *sc)
261{
262	struct sockopt sopt;
263	struct sockaddr_in6 sin6;
264	struct in6_gre_socket *s;
265	struct gre_socket *gs;
266	int error, value;
267
268	/*
269	 * NOTE: we are protected with gre_ioctl_sx lock.
270	 *
271	 * First check that socket is already configured.
272	 * If so, check that source addres was not changed.
273	 * If address is different, check that there are no other tunnels
274	 * and close socket.
275	 */
276	gs = sc->gre_so;
277	if (gs != NULL) {
278		s = __containerof(gs, struct in6_gre_socket, base);
279		if (!IN6_ARE_ADDR_EQUAL(&s->addr, &sc->gre_oip6.ip6_src)) {
280			if (CK_LIST_EMPTY(&gs->list)) {
281				CK_LIST_REMOVE(gs, chain);
282				soclose(gs->so);
283				epoch_call(net_epoch_preempt, &gs->epoch_ctx,
284				    gre_sofree);
285			}
286			gs = sc->gre_so = NULL;
287		}
288	}
289
290	if (gs == NULL) {
291		/*
292		 * Check that socket for given address is already
293		 * configured.
294		 */
295		gs = in6_gre_lookup_socket(&sc->gre_oip6.ip6_src);
296		if (gs == NULL) {
297			s = malloc(sizeof(*s), M_GRE, M_WAITOK | M_ZERO);
298			s->addr = sc->gre_oip6.ip6_src;
299			gs = &s->base;
300
301			error = socreate(sc->gre_family, &gs->so,
302			    SOCK_DGRAM, IPPROTO_UDP, curthread->td_ucred,
303			    curthread);
304			if (error != 0) {
305				if_printf(GRE2IFP(sc),
306				    "cannot create socket: %d\n", error);
307				free(s, M_GRE);
308				return (error);
309			}
310
311			error = udp_set_kernel_tunneling(gs->so,
312			    in6_gre_udp_input, NULL, gs);
313			if (error != 0) {
314				if_printf(GRE2IFP(sc),
315				    "cannot set UDP tunneling: %d\n", error);
316				goto fail;
317			}
318
319			memset(&sopt, 0, sizeof(sopt));
320			sopt.sopt_dir = SOPT_SET;
321			sopt.sopt_level = IPPROTO_IPV6;
322			sopt.sopt_name = IPV6_BINDANY;
323			sopt.sopt_val = &value;
324			sopt.sopt_valsize = sizeof(value);
325			value = 1;
326			error = sosetopt(gs->so, &sopt);
327			if (error != 0) {
328				if_printf(GRE2IFP(sc),
329				    "cannot set IPV6_BINDANY opt: %d\n",
330				    error);
331				goto fail;
332			}
333
334			memset(&sin6, 0, sizeof(sin6));
335			sin6.sin6_family = AF_INET6;
336			sin6.sin6_len = sizeof(sin6);
337			sin6.sin6_addr = sc->gre_oip6.ip6_src;
338			sin6.sin6_port = htons(GRE_UDPPORT);
339			error = sa6_recoverscope(&sin6);
340			if (error != 0) {
341				if_printf(GRE2IFP(sc),
342				    "cannot determine scope zone id: %d\n",
343				    error);
344				goto fail;
345			}
346			error = sobind(gs->so, (struct sockaddr *)&sin6,
347			    curthread);
348			if (error != 0) {
349				if_printf(GRE2IFP(sc),
350				    "cannot bind socket: %d\n", error);
351				goto fail;
352			}
353			/* Add socket to the chain */
354			CK_LIST_INSERT_HEAD(
355			    &GRE_SOCKHASH(&sc->gre_oip6.ip6_src), gs, chain);
356		}
357	}
358
359	/* Add softc to the socket's list */
360	CK_LIST_INSERT_HEAD(&gs->list, sc, chain);
361	sc->gre_so = gs;
362	return (0);
363fail:
364	soclose(gs->so);
365	free(s, M_GRE);
366	return (error);
367}
368
369static int
370in6_gre_attach(struct gre_softc *sc)
371{
372	struct grehdr *gh;
373	int error;
374
375	if (sc->gre_options & GRE_UDPENCAP) {
376		sc->gre_csumflags = CSUM_UDP_IPV6;
377		sc->gre_hlen = sizeof(struct greudp6);
378		sc->gre_oip6.ip6_nxt = IPPROTO_UDP;
379		gh = &sc->gre_udp6hdr->gi6_gre;
380		gre_update_udphdr(sc, &sc->gre_udp6,
381		    in6_cksum_pseudo(&sc->gre_oip6, 0, 0, 0));
382	} else {
383		sc->gre_hlen = sizeof(struct greip6);
384		sc->gre_oip6.ip6_nxt = IPPROTO_GRE;
385		gh = &sc->gre_ip6hdr->gi6_gre;
386	}
387	sc->gre_oip6.ip6_vfc = IPV6_VERSION;
388	gre_update_hdr(sc, gh);
389
390	/*
391	 * If we return error, this means that sc is not linked,
392	 * and caller should reset gre_family and free(sc->gre_hdr).
393	 */
394	if (sc->gre_options & GRE_UDPENCAP) {
395		error = in6_gre_setup_socket(sc);
396		if (error != 0)
397			return (error);
398	} else
399		CK_LIST_INSERT_HEAD(&GRE_HASH_SC(sc), sc, chain);
400	CK_LIST_INSERT_HEAD(&GRE_SRCHASH(&sc->gre_oip6.ip6_src), sc, srchash);
401
402	/* Set IFF_DRV_RUNNING if interface is ready */
403	in6_gre_set_running(sc);
404	return (0);
405}
406
407int
408in6_gre_setopts(struct gre_softc *sc, u_long cmd, uint32_t value)
409{
410	int error;
411
412	/* NOTE: we are protected with gre_ioctl_sx lock */
413	MPASS(cmd == GRESKEY || cmd == GRESOPTS || cmd == GRESPORT);
414	MPASS(sc->gre_family == AF_INET6);
415
416	/*
417	 * If we are going to change encapsulation protocol, do check
418	 * for duplicate tunnels. Return EEXIST here to do not confuse
419	 * user.
420	 */
421	if (cmd == GRESOPTS &&
422	    (sc->gre_options & GRE_UDPENCAP) != (value & GRE_UDPENCAP) &&
423	    in6_gre_checkdup(sc, &sc->gre_oip6.ip6_src,
424		&sc->gre_oip6.ip6_dst, value) == EADDRNOTAVAIL)
425		return (EEXIST);
426
427	CK_LIST_REMOVE(sc, chain);
428	CK_LIST_REMOVE(sc, srchash);
429	GRE_WAIT();
430	switch (cmd) {
431	case GRESKEY:
432		sc->gre_key = value;
433		break;
434	case GRESOPTS:
435		sc->gre_options = value;
436		break;
437	case GRESPORT:
438		sc->gre_port = value;
439		break;
440	}
441	error = in6_gre_attach(sc);
442	if (error != 0) {
443		sc->gre_family = 0;
444		free(sc->gre_hdr, M_GRE);
445	}
446	return (error);
447}
448
449int
450in6_gre_ioctl(struct gre_softc *sc, u_long cmd, caddr_t data)
451{
452	struct in6_ifreq *ifr = (struct in6_ifreq *)data;
453	struct sockaddr_in6 *dst, *src;
454	struct ip6_hdr *ip6;
455	int error;
456
457	/* NOTE: we are protected with gre_ioctl_sx lock */
458	error = EINVAL;
459	switch (cmd) {
460	case SIOCSIFPHYADDR_IN6:
461		src = &((struct in6_aliasreq *)data)->ifra_addr;
462		dst = &((struct in6_aliasreq *)data)->ifra_dstaddr;
463
464		/* sanity checks */
465		if (src->sin6_family != dst->sin6_family ||
466		    src->sin6_family != AF_INET6 ||
467		    src->sin6_len != dst->sin6_len ||
468		    src->sin6_len != sizeof(*src))
469			break;
470		if (IN6_IS_ADDR_UNSPECIFIED(&src->sin6_addr) ||
471		    IN6_IS_ADDR_UNSPECIFIED(&dst->sin6_addr)) {
472			error = EADDRNOTAVAIL;
473			break;
474		}
475		/*
476		 * Check validity of the scope zone ID of the
477		 * addresses, and convert it into the kernel
478		 * internal form if necessary.
479		 */
480		if ((error = sa6_embedscope(src, 0)) != 0 ||
481		    (error = sa6_embedscope(dst, 0)) != 0)
482			break;
483
484		if (V_ipv6_hashtbl == NULL) {
485			V_ipv6_hashtbl = gre_hashinit();
486			V_ipv6_srchashtbl = gre_hashinit();
487			V_ipv6_sockets = (struct gre_sockets *)gre_hashinit();
488		}
489		error = in6_gre_checkdup(sc, &src->sin6_addr,
490		    &dst->sin6_addr, sc->gre_options);
491		if (error == EADDRNOTAVAIL)
492			break;
493		if (error == EEXIST) {
494			/* Addresses are the same. Just return. */
495			error = 0;
496			break;
497		}
498		ip6 = malloc(sizeof(struct greudp6) + 3 * sizeof(uint32_t),
499		    M_GRE, M_WAITOK | M_ZERO);
500		ip6->ip6_src = src->sin6_addr;
501		ip6->ip6_dst = dst->sin6_addr;
502		if (sc->gre_family != 0) {
503			/* Detach existing tunnel first */
504			CK_LIST_REMOVE(sc, chain);
505			CK_LIST_REMOVE(sc, srchash);
506			GRE_WAIT();
507			free(sc->gre_hdr, M_GRE);
508			/* XXX: should we notify about link state change? */
509		}
510		sc->gre_family = AF_INET6;
511		sc->gre_hdr = ip6;
512		sc->gre_oseq = 0;
513		sc->gre_iseq = UINT32_MAX;
514		error = in6_gre_attach(sc);
515		if (error != 0) {
516			sc->gre_family = 0;
517			free(sc->gre_hdr, M_GRE);
518		}
519		break;
520	case SIOCGIFPSRCADDR_IN6:
521	case SIOCGIFPDSTADDR_IN6:
522		if (sc->gre_family != AF_INET6) {
523			error = EADDRNOTAVAIL;
524			break;
525		}
526		src = (struct sockaddr_in6 *)&ifr->ifr_addr;
527		memset(src, 0, sizeof(*src));
528		src->sin6_family = AF_INET6;
529		src->sin6_len = sizeof(*src);
530		src->sin6_addr = (cmd == SIOCGIFPSRCADDR_IN6) ?
531		    sc->gre_oip6.ip6_src: sc->gre_oip6.ip6_dst;
532		error = prison_if(curthread->td_ucred, (struct sockaddr *)src);
533		if (error == 0)
534			error = sa6_recoverscope(src);
535		if (error != 0)
536			memset(src, 0, sizeof(*src));
537		break;
538	}
539	return (error);
540}
541
542int
543in6_gre_output(struct mbuf *m, int af __unused, int hlen __unused,
544    uint32_t flowid)
545{
546	struct greip6 *gi6;
547
548	gi6 = mtod(m, struct greip6 *);
549	gi6->gi6_ip6.ip6_hlim = V_ip6_gre_hlim;
550	gi6->gi6_ip6.ip6_flow |= flowid & IPV6_FLOWLABEL_MASK;
551	return (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, NULL));
552}
553
554static const struct srcaddrtab *ipv6_srcaddrtab = NULL;
555static const struct encaptab *ecookie = NULL;
556static const struct encap_config ipv6_encap_cfg = {
557	.proto = IPPROTO_GRE,
558	.min_length = sizeof(struct greip6) +
559#ifdef INET
560	    sizeof(struct ip),
561#else
562	    sizeof(struct ip6_hdr),
563#endif
564	.exact_match = ENCAP_DRV_LOOKUP,
565	.lookup = in6_gre_lookup,
566	.input = gre_input
567};
568
569void
570in6_gre_init(void)
571{
572
573	if (!IS_DEFAULT_VNET(curvnet))
574		return;
575	ipv6_srcaddrtab = ip6_encap_register_srcaddr(in6_gre_srcaddr,
576	    NULL, M_WAITOK);
577	ecookie = ip6_encap_attach(&ipv6_encap_cfg, NULL, M_WAITOK);
578}
579
580void
581in6_gre_uninit(void)
582{
583
584	if (IS_DEFAULT_VNET(curvnet)) {
585		ip6_encap_detach(ecookie);
586		ip6_encap_unregister_srcaddr(ipv6_srcaddrtab);
587	}
588	if (V_ipv6_hashtbl != NULL) {
589		gre_hashdestroy(V_ipv6_hashtbl);
590		V_ipv6_hashtbl = NULL;
591		GRE_WAIT();
592		gre_hashdestroy(V_ipv6_srchashtbl);
593		gre_hashdestroy((struct gre_list *)V_ipv6_sockets);
594	}
595}
596