xref: /freebsd/sys/net/if_ipsec.c (revision 8881d206f4e68b564c2c5f50fc717086fc3e827a)
1 /*-
2  * Copyright (c) 2016-2018 Yandex LLC
3  * Copyright (c) 2016-2018 Andrey V. Elsukov <ae@FreeBSD.org>
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  *
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
31 #include "opt_inet.h"
32 #include "opt_inet6.h"
33 #include "opt_ipsec.h"
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
38 #include <sys/fnv_hash.h>
39 #include <sys/jail.h>
40 #include <sys/lock.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/module.h>
44 #include <sys/socket.h>
45 #include <sys/sockio.h>
46 #include <sys/sx.h>
47 #include <sys/errno.h>
48 #include <sys/sysctl.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/conf.h>
52 
53 #include <net/if.h>
54 #include <net/if_var.h>
55 #include <net/if_clone.h>
56 #include <net/if_types.h>
57 #include <net/bpf.h>
58 #include <net/route.h>
59 #include <net/vnet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_encap.h>
65 
66 #include <netinet/ip6.h>
67 #include <netinet6/in6_var.h>
68 #include <netinet6/scope6_var.h>
69 
70 #include <netipsec/ipsec.h>
71 #ifdef INET6
72 #include <netipsec/ipsec6.h>
73 #endif
74 
75 #include <net/if_ipsec.h>
76 #include <netipsec/key.h>
77 
78 #include <security/mac/mac_framework.h>
79 
80 static MALLOC_DEFINE(M_IPSEC, "ipsec", "IPsec Virtual Tunnel Interface");
81 static const char ipsecname[] = "ipsec";
82 
83 #if defined(INET) && defined(INET6)
84 #define	IPSEC_SPCOUNT		4
85 #else
86 #define	IPSEC_SPCOUNT		2
87 #endif
88 
89 struct ipsec_softc {
90 	struct ifnet		*ifp;
91 	struct secpolicy	*sp[IPSEC_SPCOUNT];
92 	uint32_t		reqid;
93 	u_int			family;
94 	u_int			fibnum;
95 
96 	CK_LIST_ENTRY(ipsec_softc) idhash;
97 	CK_LIST_ENTRY(ipsec_softc) srchash;
98 };
99 
100 #define	IPSEC_RLOCK_TRACKER	struct epoch_tracker ipsec_et
101 #define	IPSEC_RLOCK()	epoch_enter_preempt(net_epoch_preempt, &ipsec_et)
102 #define	IPSEC_RUNLOCK()	epoch_exit_preempt(net_epoch_preempt, &ipsec_et)
103 #define	IPSEC_WAIT()	epoch_wait_preempt(net_epoch_preempt)
104 
105 #ifndef IPSEC_HASH_SIZE
106 #define	IPSEC_HASH_SIZE	(1 << 5)
107 #endif
108 
109 CK_LIST_HEAD(ipsec_iflist, ipsec_softc);
110 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec_idhtbl) = NULL;
111 #define	V_ipsec_idhtbl		VNET(ipsec_idhtbl)
112 
113 #ifdef INET
114 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec4_srchtbl) = NULL;
115 #define	V_ipsec4_srchtbl	VNET(ipsec4_srchtbl)
116 static const struct srcaddrtab *ipsec4_srctab = NULL;
117 #endif
118 
119 #ifdef INET6
120 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec6_srchtbl) = NULL;
121 #define	V_ipsec6_srchtbl	VNET(ipsec6_srchtbl)
122 static const struct srcaddrtab *ipsec6_srctab = NULL;
123 #endif
124 
125 static struct ipsec_iflist *
126 ipsec_idhash(uint32_t id)
127 {
128 
129 	return (&V_ipsec_idhtbl[fnv_32_buf(&id, sizeof(id),
130 	    FNV1_32_INIT) & (IPSEC_HASH_SIZE - 1)]);
131 }
132 
133 static struct ipsec_iflist *
134 ipsec_srchash(const struct sockaddr *sa)
135 {
136 	uint32_t hval;
137 
138 	switch (sa->sa_family) {
139 #ifdef INET
140 	case AF_INET:
141 		hval = fnv_32_buf(
142 		    &((const struct sockaddr_in *)sa)->sin_addr.s_addr,
143 		    sizeof(in_addr_t), FNV1_32_INIT);
144 		return (&V_ipsec4_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
145 #endif
146 #ifdef INET6
147 	case AF_INET6:
148 		hval = fnv_32_buf(
149 		    &((const struct sockaddr_in6 *)sa)->sin6_addr,
150 		    sizeof(struct in6_addr), FNV1_32_INIT);
151 		return (&V_ipsec6_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
152 #endif
153 	}
154 	return (NULL);
155 }
156 
157 /*
158  * ipsec_ioctl_sx protects from concurrent ioctls.
159  */
160 static struct sx ipsec_ioctl_sx;
161 SX_SYSINIT(ipsec_ioctl_sx, &ipsec_ioctl_sx, "ipsec_ioctl");
162 
163 static int	ipsec_init_reqid(struct ipsec_softc *);
164 static int	ipsec_set_tunnel(struct ipsec_softc *, struct sockaddr *,
165     struct sockaddr *, uint32_t);
166 static void	ipsec_delete_tunnel(struct ipsec_softc *);
167 
168 static int	ipsec_set_addresses(struct ifnet *, struct sockaddr *,
169     struct sockaddr *);
170 static int	ipsec_set_reqid(struct ipsec_softc *, uint32_t);
171 static void	ipsec_set_running(struct ipsec_softc *);
172 
173 #ifdef VIMAGE
174 static void	ipsec_reassign(struct ifnet *, struct vnet *, char *);
175 #endif
176 static void	ipsec_srcaddr(void *, const struct sockaddr *, int);
177 static int	ipsec_ioctl(struct ifnet *, u_long, caddr_t);
178 static int	ipsec_transmit(struct ifnet *, struct mbuf *);
179 static int	ipsec_output(struct ifnet *, struct mbuf *,
180     const struct sockaddr *, struct route *);
181 static void	ipsec_qflush(struct ifnet *);
182 static int	ipsec_clone_create(struct if_clone *, int, caddr_t);
183 static void	ipsec_clone_destroy(struct ifnet *);
184 
185 VNET_DEFINE_STATIC(struct if_clone *, ipsec_cloner);
186 #define	V_ipsec_cloner		VNET(ipsec_cloner)
187 
188 static int
189 ipsec_clone_create(struct if_clone *ifc, int unit, caddr_t params)
190 {
191 	struct ipsec_softc *sc;
192 	struct ifnet *ifp;
193 
194 	sc = malloc(sizeof(*sc), M_IPSEC, M_WAITOK | M_ZERO);
195 	sc->fibnum = curthread->td_proc->p_fibnum;
196 	sc->ifp = ifp = if_alloc(IFT_TUNNEL);
197 	ifp->if_softc = sc;
198 	if_initname(ifp, ipsecname, unit);
199 
200 	ifp->if_addrlen = 0;
201 	ifp->if_mtu = IPSEC_MTU;
202 	ifp->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
203 	ifp->if_ioctl  = ipsec_ioctl;
204 	ifp->if_transmit  = ipsec_transmit;
205 	ifp->if_qflush  = ipsec_qflush;
206 	ifp->if_output = ipsec_output;
207 #ifdef VIMAGE
208 	ifp->if_reassign = ipsec_reassign;
209 #endif
210 	if_attach(ifp);
211 	bpfattach(ifp, DLT_NULL, sizeof(uint32_t));
212 
213 	return (0);
214 }
215 
216 #ifdef VIMAGE
217 static void
218 ipsec_reassign(struct ifnet *ifp, struct vnet *new_vnet __unused,
219     char *unused __unused)
220 {
221 	struct ipsec_softc *sc;
222 
223 	sx_xlock(&ipsec_ioctl_sx);
224 	sc = ifp->if_softc;
225 	if (sc != NULL)
226 		ipsec_delete_tunnel(sc);
227 	sx_xunlock(&ipsec_ioctl_sx);
228 }
229 #endif /* VIMAGE */
230 
231 static void
232 ipsec_clone_destroy(struct ifnet *ifp)
233 {
234 	struct ipsec_softc *sc;
235 
236 	sx_xlock(&ipsec_ioctl_sx);
237 	sc = ifp->if_softc;
238 	ipsec_delete_tunnel(sc);
239 	/*
240 	 * Delete softc from idhash on interface destroy, since
241 	 * ipsec_delete_tunnel() keeps reqid unchanged.
242 	 */
243 	if (sc->reqid != 0)
244 		CK_LIST_REMOVE(sc, idhash);
245 	bpfdetach(ifp);
246 	if_detach(ifp);
247 	ifp->if_softc = NULL;
248 	sx_xunlock(&ipsec_ioctl_sx);
249 
250 	IPSEC_WAIT();
251 	if_free(ifp);
252 	free(sc, M_IPSEC);
253 }
254 
255 static struct ipsec_iflist *
256 ipsec_hashinit(void)
257 {
258 	struct ipsec_iflist *hash;
259 	int i;
260 
261 	hash = malloc(sizeof(struct ipsec_iflist) * IPSEC_HASH_SIZE,
262 	    M_IPSEC, M_WAITOK);
263 	for (i = 0; i < IPSEC_HASH_SIZE; i++)
264 		CK_LIST_INIT(&hash[i]);
265 
266 	return (hash);
267 }
268 
269 static void
270 vnet_ipsec_init(const void *unused __unused)
271 {
272 
273 	V_ipsec_idhtbl = ipsec_hashinit();
274 #ifdef INET
275 	V_ipsec4_srchtbl = ipsec_hashinit();
276 	if (IS_DEFAULT_VNET(curvnet))
277 		ipsec4_srctab = ip_encap_register_srcaddr(ipsec_srcaddr,
278 		    NULL, M_WAITOK);
279 #endif
280 #ifdef INET6
281 	V_ipsec6_srchtbl = ipsec_hashinit();
282 	if (IS_DEFAULT_VNET(curvnet))
283 		ipsec6_srctab = ip6_encap_register_srcaddr(ipsec_srcaddr,
284 		    NULL, M_WAITOK);
285 #endif
286 	V_ipsec_cloner = if_clone_simple(ipsecname, ipsec_clone_create,
287 	    ipsec_clone_destroy, 0);
288 }
289 VNET_SYSINIT(vnet_ipsec_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
290     vnet_ipsec_init, NULL);
291 
292 static void
293 vnet_ipsec_uninit(const void *unused __unused)
294 {
295 
296 	if_clone_detach(V_ipsec_cloner);
297 	free(V_ipsec_idhtbl, M_IPSEC);
298 	/*
299 	 * Use V_ipsec_idhtbl pointer as indicator that VNET is going to be
300 	 * destroyed, it is used by ipsec_srcaddr() callback.
301 	 */
302 	V_ipsec_idhtbl = NULL;
303 	IPSEC_WAIT();
304 
305 #ifdef INET
306 	if (IS_DEFAULT_VNET(curvnet))
307 		ip_encap_unregister_srcaddr(ipsec4_srctab);
308 	free(V_ipsec4_srchtbl, M_IPSEC);
309 #endif
310 #ifdef INET6
311 	if (IS_DEFAULT_VNET(curvnet))
312 		ip6_encap_unregister_srcaddr(ipsec6_srctab);
313 	free(V_ipsec6_srchtbl, M_IPSEC);
314 #endif
315 }
316 VNET_SYSUNINIT(vnet_ipsec_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
317     vnet_ipsec_uninit, NULL);
318 
319 static struct secpolicy *
320 ipsec_getpolicy(struct ipsec_softc *sc, int dir, sa_family_t af)
321 {
322 
323 	switch (af) {
324 #ifdef INET
325 	case AF_INET:
326 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)]);
327 #endif
328 #ifdef INET6
329 	case AF_INET6:
330 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)
331 #ifdef INET
332 			+ 2
333 #endif
334 		]);
335 #endif
336 	}
337 	return (NULL);
338 }
339 
340 static struct secasindex *
341 ipsec_getsaidx(struct ipsec_softc *sc, int dir, sa_family_t af)
342 {
343 	struct secpolicy *sp;
344 
345 	sp = ipsec_getpolicy(sc, dir, af);
346 	if (sp == NULL)
347 		return (NULL);
348 	return (&sp->req[0]->saidx);
349 }
350 
351 static int
352 ipsec_transmit(struct ifnet *ifp, struct mbuf *m)
353 {
354 	IPSEC_RLOCK_TRACKER;
355 	struct ipsec_softc *sc;
356 	struct secpolicy *sp;
357 	struct ip *ip;
358 	uint32_t af;
359 	int error;
360 
361 	IPSEC_RLOCK();
362 #ifdef MAC
363 	error = mac_ifnet_check_transmit(ifp, m);
364 	if (error) {
365 		m_freem(m);
366 		goto err;
367 	}
368 #endif
369 	error = ENETDOWN;
370 	sc = ifp->if_softc;
371 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
372 	    (ifp->if_flags & IFF_MONITOR) != 0 ||
373 	    (ifp->if_flags & IFF_UP) == 0 || sc->family == 0) {
374 		m_freem(m);
375 		goto err;
376 	}
377 
378 	/* Determine address family to correctly handle packet in BPF */
379 	ip = mtod(m, struct ip *);
380 	switch (ip->ip_v) {
381 #ifdef INET
382 	case IPVERSION:
383 		af = AF_INET;
384 		break;
385 #endif
386 #ifdef INET6
387 	case (IPV6_VERSION >> 4):
388 		af = AF_INET6;
389 		break;
390 #endif
391 	default:
392 		error = EAFNOSUPPORT;
393 		m_freem(m);
394 		goto err;
395 	}
396 
397 	/*
398 	 * Loop prevention.
399 	 * XXX: for now just check presence of IPSEC_OUT_DONE mbuf tag.
400 	 *      We can read full chain and compare destination address,
401 	 *      proto and mode from xform_history with values from softc.
402 	 */
403 	if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) {
404 		m_freem(m);
405 		goto err;
406 	}
407 
408 	sp = ipsec_getpolicy(sc, IPSEC_DIR_OUTBOUND, af);
409 	key_addref(sp);
410 	M_SETFIB(m, sc->fibnum);
411 
412 	BPF_MTAP2(ifp, &af, sizeof(af), m);
413 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
414 	if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
415 
416 	switch (af) {
417 #ifdef INET
418 	case AF_INET:
419 		error = ipsec4_process_packet(m, sp, NULL);
420 		break;
421 #endif
422 #ifdef INET6
423 	case AF_INET6:
424 		error = ipsec6_process_packet(m, sp, NULL);
425 		break;
426 #endif
427 	default:
428 		panic("%s: unknown address family\n", __func__);
429 	}
430 err:
431 	if (error != 0)
432 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
433 	IPSEC_RUNLOCK();
434 	return (error);
435 }
436 
437 static void
438 ipsec_qflush(struct ifnet *ifp __unused)
439 {
440 
441 }
442 
443 static int
444 ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
445 	struct route *ro)
446 {
447 
448 	return (ifp->if_transmit(ifp, m));
449 }
450 
451 int
452 ipsec_if_input(struct mbuf *m, struct secasvar *sav, uint32_t af)
453 {
454 	IPSEC_RLOCK_TRACKER;
455 	struct secasindex *saidx;
456 	struct ipsec_softc *sc;
457 	struct ifnet *ifp;
458 
459 	if (sav->state != SADB_SASTATE_MATURE &&
460 	    sav->state != SADB_SASTATE_DYING) {
461 		m_freem(m);
462 		return (ENETDOWN);
463 	}
464 
465 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL ||
466 	    sav->sah->saidx.proto != IPPROTO_ESP)
467 		return (0);
468 
469 	IPSEC_RLOCK();
470 	CK_LIST_FOREACH(sc, ipsec_idhash(sav->sah->saidx.reqid), idhash) {
471 		if (sc->family == 0)
472 			continue;
473 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_INBOUND,
474 		    sav->sah->saidx.src.sa.sa_family);
475 		/* SA's reqid should match reqid in SP */
476 		if (saidx == NULL ||
477 		    sav->sah->saidx.reqid != saidx->reqid)
478 			continue;
479 		/* SAH's addresses should match tunnel endpoints. */
480 		if (key_sockaddrcmp(&sav->sah->saidx.dst.sa,
481 		    &saidx->dst.sa, 0) != 0)
482 			continue;
483 		if (key_sockaddrcmp(&sav->sah->saidx.src.sa,
484 		    &saidx->src.sa, 0) == 0)
485 			break;
486 	}
487 	if (sc == NULL) {
488 		IPSEC_RUNLOCK();
489 		/* Tunnel was not found. Nothing to do. */
490 		return (0);
491 	}
492 	ifp = sc->ifp;
493 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
494 	    (ifp->if_flags & IFF_UP) == 0) {
495 		IPSEC_RUNLOCK();
496 		m_freem(m);
497 		return (ENETDOWN);
498 	}
499 	/*
500 	 * We found matching and working tunnel.
501 	 * Set its ifnet as receiving interface.
502 	 */
503 	m->m_pkthdr.rcvif = ifp;
504 
505 	m_clrprotoflags(m);
506 	M_SETFIB(m, ifp->if_fib);
507 	BPF_MTAP2(ifp, &af, sizeof(af), m);
508 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
509 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
510 	if ((ifp->if_flags & IFF_MONITOR) != 0) {
511 		IPSEC_RUNLOCK();
512 		m_freem(m);
513 		return (ENETDOWN);
514 	}
515 	IPSEC_RUNLOCK();
516 	return (0);
517 }
518 
519 static int
520 ipsec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
521 {
522 	struct ifreq *ifr = (struct ifreq*)data;
523 	struct sockaddr *dst, *src;
524 	struct ipsec_softc *sc;
525 	struct secasindex *saidx;
526 #ifdef INET
527 	struct sockaddr_in *sin = NULL;
528 #endif
529 #ifdef INET6
530 	struct sockaddr_in6 *sin6 = NULL;
531 #endif
532 	uint32_t reqid;
533 	int error;
534 
535 	switch (cmd) {
536 	case SIOCSIFADDR:
537 		ifp->if_flags |= IFF_UP;
538 	case SIOCADDMULTI:
539 	case SIOCDELMULTI:
540 	case SIOCGIFMTU:
541 	case SIOCSIFFLAGS:
542 		return (0);
543 	case SIOCSIFMTU:
544 		if (ifr->ifr_mtu < IPSEC_MTU_MIN ||
545 		    ifr->ifr_mtu > IPSEC_MTU_MAX)
546 			return (EINVAL);
547 		else
548 			ifp->if_mtu = ifr->ifr_mtu;
549 		return (0);
550 	}
551 	sx_xlock(&ipsec_ioctl_sx);
552 	sc = ifp->if_softc;
553 	/* Check that softc is still here */
554 	if (sc == NULL) {
555 		error = ENXIO;
556 		goto bad;
557 	}
558 	error = 0;
559 	switch (cmd) {
560 	case SIOCSIFPHYADDR:
561 #ifdef INET6
562 	case SIOCSIFPHYADDR_IN6:
563 #endif
564 		error = EINVAL;
565 		switch (cmd) {
566 #ifdef INET
567 		case SIOCSIFPHYADDR:
568 			src = (struct sockaddr *)
569 				&(((struct in_aliasreq *)data)->ifra_addr);
570 			dst = (struct sockaddr *)
571 				&(((struct in_aliasreq *)data)->ifra_dstaddr);
572 			break;
573 #endif
574 #ifdef INET6
575 		case SIOCSIFPHYADDR_IN6:
576 			src = (struct sockaddr *)
577 				&(((struct in6_aliasreq *)data)->ifra_addr);
578 			dst = (struct sockaddr *)
579 				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
580 			break;
581 #endif
582 		default:
583 			goto bad;
584 		}
585 		/* sa_family must be equal */
586 		if (src->sa_family != dst->sa_family ||
587 		    src->sa_len != dst->sa_len)
588 			goto bad;
589 
590 		/* validate sa_len */
591 		switch (src->sa_family) {
592 #ifdef INET
593 		case AF_INET:
594 			if (src->sa_len != sizeof(struct sockaddr_in))
595 				goto bad;
596 			break;
597 #endif
598 #ifdef INET6
599 		case AF_INET6:
600 			if (src->sa_len != sizeof(struct sockaddr_in6))
601 				goto bad;
602 			break;
603 #endif
604 		default:
605 			error = EAFNOSUPPORT;
606 			goto bad;
607 		}
608 		/* check sa_family looks sane for the cmd */
609 		error = EAFNOSUPPORT;
610 		switch (cmd) {
611 #ifdef INET
612 		case SIOCSIFPHYADDR:
613 			if (src->sa_family == AF_INET)
614 				break;
615 			goto bad;
616 #endif
617 #ifdef INET6
618 		case SIOCSIFPHYADDR_IN6:
619 			if (src->sa_family == AF_INET6)
620 				break;
621 			goto bad;
622 #endif
623 		}
624 		error = EADDRNOTAVAIL;
625 		switch (src->sa_family) {
626 #ifdef INET
627 		case AF_INET:
628 			if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
629 			    satosin(dst)->sin_addr.s_addr == INADDR_ANY)
630 				goto bad;
631 			break;
632 #endif
633 #ifdef INET6
634 		case AF_INET6:
635 			if (IN6_IS_ADDR_UNSPECIFIED(
636 			    &satosin6(src)->sin6_addr) ||
637 			    IN6_IS_ADDR_UNSPECIFIED(
638 			    &satosin6(dst)->sin6_addr))
639 				goto bad;
640 			/*
641 			 * Check validity of the scope zone ID of the
642 			 * addresses, and convert it into the kernel
643 			 * internal form if necessary.
644 			 */
645 			error = sa6_embedscope(satosin6(src), 0);
646 			if (error != 0)
647 				goto bad;
648 			error = sa6_embedscope(satosin6(dst), 0);
649 			if (error != 0)
650 				goto bad;
651 #endif
652 		};
653 		error = ipsec_set_addresses(ifp, src, dst);
654 		break;
655 	case SIOCDIFPHYADDR:
656 		ipsec_delete_tunnel(sc);
657 		break;
658 	case SIOCGIFPSRCADDR:
659 	case SIOCGIFPDSTADDR:
660 #ifdef INET6
661 	case SIOCGIFPSRCADDR_IN6:
662 	case SIOCGIFPDSTADDR_IN6:
663 #endif
664 		if (sc->family == 0) {
665 			error = EADDRNOTAVAIL;
666 			break;
667 		}
668 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
669 		switch (cmd) {
670 #ifdef INET
671 		case SIOCGIFPSRCADDR:
672 		case SIOCGIFPDSTADDR:
673 			if (saidx->src.sa.sa_family != AF_INET) {
674 				error = EADDRNOTAVAIL;
675 				break;
676 			}
677 			sin = (struct sockaddr_in *)&ifr->ifr_addr;
678 			memset(sin, 0, sizeof(*sin));
679 			sin->sin_family = AF_INET;
680 			sin->sin_len = sizeof(*sin);
681 			break;
682 #endif
683 #ifdef INET6
684 		case SIOCGIFPSRCADDR_IN6:
685 		case SIOCGIFPDSTADDR_IN6:
686 			if (saidx->src.sa.sa_family != AF_INET6) {
687 				error = EADDRNOTAVAIL;
688 				break;
689 			}
690 			sin6 = (struct sockaddr_in6 *)
691 				&(((struct in6_ifreq *)data)->ifr_addr);
692 			memset(sin6, 0, sizeof(*sin6));
693 			sin6->sin6_family = AF_INET6;
694 			sin6->sin6_len = sizeof(*sin6);
695 			break;
696 #endif
697 		default:
698 			error = EAFNOSUPPORT;
699 		}
700 		if (error == 0) {
701 			switch (cmd) {
702 #ifdef INET
703 			case SIOCGIFPSRCADDR:
704 				sin->sin_addr = saidx->src.sin.sin_addr;
705 				break;
706 			case SIOCGIFPDSTADDR:
707 				sin->sin_addr = saidx->dst.sin.sin_addr;
708 				break;
709 #endif
710 #ifdef INET6
711 			case SIOCGIFPSRCADDR_IN6:
712 				sin6->sin6_addr = saidx->src.sin6.sin6_addr;
713 				break;
714 			case SIOCGIFPDSTADDR_IN6:
715 				sin6->sin6_addr = saidx->dst.sin6.sin6_addr;
716 				break;
717 #endif
718 			}
719 		}
720 		if (error != 0)
721 			break;
722 		switch (cmd) {
723 #ifdef INET
724 		case SIOCGIFPSRCADDR:
725 		case SIOCGIFPDSTADDR:
726 			error = prison_if(curthread->td_ucred,
727 			    (struct sockaddr *)sin);
728 			if (error != 0)
729 				memset(sin, 0, sizeof(*sin));
730 			break;
731 #endif
732 #ifdef INET6
733 		case SIOCGIFPSRCADDR_IN6:
734 		case SIOCGIFPDSTADDR_IN6:
735 			error = prison_if(curthread->td_ucred,
736 			    (struct sockaddr *)sin6);
737 			if (error == 0)
738 				error = sa6_recoverscope(sin6);
739 			if (error != 0)
740 				memset(sin6, 0, sizeof(*sin6));
741 #endif
742 		}
743 		break;
744 	case SIOCGTUNFIB:
745 		ifr->ifr_fib = sc->fibnum;
746 		break;
747 	case SIOCSTUNFIB:
748 		if ((error = priv_check(curthread, PRIV_NET_SETIFFIB)) != 0)
749 			break;
750 		if (ifr->ifr_fib >= rt_numfibs)
751 			error = EINVAL;
752 		else
753 			sc->fibnum = ifr->ifr_fib;
754 		break;
755 	case IPSECGREQID:
756 		reqid = sc->reqid;
757 		error = copyout(&reqid, ifr_data_get_ptr(ifr), sizeof(reqid));
758 		break;
759 	case IPSECSREQID:
760 		if ((error = priv_check(curthread, PRIV_NET_SETIFCAP)) != 0)
761 			break;
762 		error = copyin(ifr_data_get_ptr(ifr), &reqid, sizeof(reqid));
763 		if (error != 0)
764 			break;
765 		error = ipsec_set_reqid(sc, reqid);
766 		break;
767 	default:
768 		error = EINVAL;
769 		break;
770 	}
771 bad:
772 	sx_xunlock(&ipsec_ioctl_sx);
773 	return (error);
774 }
775 
776 /*
777  * Check that ingress address belongs to local host.
778  */
779 static void
780 ipsec_set_running(struct ipsec_softc *sc)
781 {
782 	struct secasindex *saidx;
783 	int localip;
784 
785 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
786 	localip = 0;
787 	switch (sc->family) {
788 #ifdef INET
789 	case AF_INET:
790 		localip = in_localip(saidx->src.sin.sin_addr);
791 		break;
792 #endif
793 #ifdef INET6
794 	case AF_INET6:
795 		localip = in6_localip(&saidx->src.sin6.sin6_addr);
796 		break;
797 #endif
798 	}
799 	if (localip != 0)
800 		sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
801 	else
802 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
803 }
804 
805 /*
806  * ifaddr_event handler.
807  * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
808  * source address spoofing.
809  */
810 static void
811 ipsec_srcaddr(void *arg __unused, const struct sockaddr *sa,
812     int event __unused)
813 {
814 	struct ipsec_softc *sc;
815 	struct secasindex *saidx;
816 
817 	/* Check that VNET is ready */
818 	if (V_ipsec_idhtbl == NULL)
819 		return;
820 
821 	NET_EPOCH_ASSERT();
822 	CK_LIST_FOREACH(sc, ipsec_srchash(sa), srchash) {
823 		if (sc->family == 0)
824 			continue;
825 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sa->sa_family);
826 		if (saidx == NULL ||
827 		    key_sockaddrcmp(&saidx->src.sa, sa, 0) != 0)
828 			continue;
829 		ipsec_set_running(sc);
830 	}
831 }
832 
833 /*
834  * Allocate new private security policies for tunneling interface.
835  * Each tunneling interface has following security policies for
836  * both AF:
837  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P in \
838  *	ipsec esp/tunnel/RemoteIP-LocalIP/unique:reqid
839  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P out \
840  *	ipsec esp/tunnel/LocalIP-RemoteIP/unique:reqid
841  */
842 static int
843 ipsec_newpolicies(struct ipsec_softc *sc, struct secpolicy *sp[IPSEC_SPCOUNT],
844     const struct sockaddr *src, const struct sockaddr *dst, uint32_t reqid)
845 {
846 	struct ipsecrequest *isr;
847 	int i;
848 
849 	memset(sp, 0, sizeof(struct secpolicy *) * IPSEC_SPCOUNT);
850 	for (i = 0; i < IPSEC_SPCOUNT; i++) {
851 		if ((sp[i] = key_newsp()) == NULL)
852 			goto fail;
853 		if ((isr = ipsec_newisr()) == NULL)
854 			goto fail;
855 
856 		sp[i]->policy = IPSEC_POLICY_IPSEC;
857 		sp[i]->state = IPSEC_SPSTATE_DEAD;
858 		sp[i]->req[sp[i]->tcount++] = isr;
859 		sp[i]->created = time_second;
860 		/* Use priority field to store if_index */
861 		sp[i]->priority = sc->ifp->if_index;
862 		isr->level = IPSEC_LEVEL_UNIQUE;
863 		isr->saidx.proto = IPPROTO_ESP;
864 		isr->saidx.mode = IPSEC_MODE_TUNNEL;
865 		isr->saidx.reqid = reqid;
866 		if (i % 2 == 0) {
867 			sp[i]->spidx.dir = IPSEC_DIR_INBOUND;
868 			bcopy(src, &isr->saidx.dst, src->sa_len);
869 			bcopy(dst, &isr->saidx.src, dst->sa_len);
870 		} else {
871 			sp[i]->spidx.dir = IPSEC_DIR_OUTBOUND;
872 			bcopy(src, &isr->saidx.src, src->sa_len);
873 			bcopy(dst, &isr->saidx.dst, dst->sa_len);
874 		}
875 		sp[i]->spidx.ul_proto = IPSEC_ULPROTO_ANY;
876 #ifdef INET
877 		if (i < 2) {
878 			sp[i]->spidx.src.sa.sa_family =
879 			    sp[i]->spidx.dst.sa.sa_family = AF_INET;
880 			sp[i]->spidx.src.sa.sa_len =
881 			    sp[i]->spidx.dst.sa.sa_len =
882 			    sizeof(struct sockaddr_in);
883 			continue;
884 		}
885 #endif
886 #ifdef INET6
887 		sp[i]->spidx.src.sa.sa_family =
888 		    sp[i]->spidx.dst.sa.sa_family = AF_INET6;
889 		sp[i]->spidx.src.sa.sa_len =
890 		    sp[i]->spidx.dst.sa.sa_len = sizeof(struct sockaddr_in6);
891 #endif
892 	}
893 	return (0);
894 fail:
895 	for (i = 0; i < IPSEC_SPCOUNT; i++)
896 		key_freesp(&sp[i]);
897 	return (ENOMEM);
898 }
899 
900 static int
901 ipsec_check_reqid(uint32_t reqid)
902 {
903 	struct ipsec_softc *sc;
904 
905 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
906 	CK_LIST_FOREACH(sc, ipsec_idhash(reqid), idhash) {
907 		if (sc->reqid == reqid)
908 			return (EEXIST);
909 	}
910 	return (0);
911 }
912 
913 /*
914  * We use key_newreqid() to automatically obtain unique reqid.
915  * Then we check that given id is unique, i.e. it is not used by
916  * another if_ipsec(4) interface. This macro limits the number of
917  * tries to get unique id.
918  */
919 #define	IPSEC_REQID_TRYCNT	64
920 static int
921 ipsec_init_reqid(struct ipsec_softc *sc)
922 {
923 	uint32_t reqid;
924 	int trycount;
925 
926 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
927 	if (sc->reqid != 0) /* already initialized */
928 		return (0);
929 
930 	trycount = IPSEC_REQID_TRYCNT;
931 	while (--trycount > 0) {
932 		reqid = key_newreqid();
933 		if (ipsec_check_reqid(reqid) == 0)
934 			break;
935 	}
936 	if (trycount == 0)
937 		return (EEXIST);
938 	sc->reqid = reqid;
939 	CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
940 	return (0);
941 }
942 
943 /*
944  * Set or update reqid for given tunneling interface.
945  * When specified reqid is zero, generate new one.
946  * We are protected by ioctl_sx lock from concurrent id generation.
947  * Also softc would not disappear while we hold ioctl_sx lock.
948  */
949 static int
950 ipsec_set_reqid(struct ipsec_softc *sc, uint32_t reqid)
951 {
952 	struct secasindex *saidx;
953 
954 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
955 
956 	if (sc->reqid == reqid && reqid != 0)
957 		return (0);
958 
959 	if (reqid != 0) {
960 		/* Check that specified reqid doesn't exist */
961 		if (ipsec_check_reqid(reqid) != 0)
962 			return (EEXIST);
963 		if (sc->reqid != 0) {
964 			CK_LIST_REMOVE(sc, idhash);
965 			IPSEC_WAIT();
966 		}
967 		sc->reqid = reqid;
968 		CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
969 	} else {
970 		/* Generate new reqid */
971 		if (ipsec_init_reqid(sc) != 0)
972 			return (EEXIST);
973 	}
974 
975 	/* Tunnel isn't fully configured, just return. */
976 	if (sc->family == 0)
977 		return (0);
978 
979 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
980 	KASSERT(saidx != NULL,
981 	    ("saidx is NULL, but family is %d", sc->family));
982 	return (ipsec_set_tunnel(sc, &saidx->src.sa, &saidx->dst.sa,
983 	    sc->reqid));
984 }
985 
986 /*
987  * Set tunnel endpoints addresses.
988  */
989 static int
990 ipsec_set_addresses(struct ifnet *ifp, struct sockaddr *src,
991     struct sockaddr *dst)
992 {
993 	struct ipsec_softc *sc;
994 	struct secasindex *saidx;
995 
996 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
997 
998 	sc = ifp->if_softc;
999 	if (sc->family != 0) {
1000 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND,
1001 		    src->sa_family);
1002 		if (saidx != NULL && saidx->reqid == sc->reqid &&
1003 		    key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
1004 		    key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0)
1005 			return (0); /* Nothing has been changed. */
1006 	}
1007 	/* If reqid is not set, generate new one. */
1008 	if (ipsec_init_reqid(sc) != 0)
1009 		return (EEXIST);
1010 	return (ipsec_set_tunnel(sc, src, dst, sc->reqid));
1011 }
1012 
1013 static int
1014 ipsec_set_tunnel(struct ipsec_softc *sc, struct sockaddr *src,
1015     struct sockaddr *dst, uint32_t reqid)
1016 {
1017 	struct epoch_tracker et;
1018 	struct secpolicy *sp[IPSEC_SPCOUNT];
1019 	int i;
1020 
1021 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1022 
1023 	/* Allocate SP with new addresses. */
1024 	if (ipsec_newpolicies(sc, sp, src, dst, reqid) == 0) {
1025 		/* Add new policies to SPDB */
1026 		if (key_register_ifnet(sp, IPSEC_SPCOUNT) != 0) {
1027 			for (i = 0; i < IPSEC_SPCOUNT; i++)
1028 				key_freesp(&sp[i]);
1029 			return (EAGAIN);
1030 		}
1031 		if (sc->family != 0)
1032 			ipsec_delete_tunnel(sc);
1033 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1034 			sc->sp[i] = sp[i];
1035 		sc->family = src->sa_family;
1036 		CK_LIST_INSERT_HEAD(ipsec_srchash(src), sc, srchash);
1037 	} else {
1038 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1039 		return (ENOMEM);
1040 	}
1041 	NET_EPOCH_ENTER(et);
1042 	ipsec_set_running(sc);
1043 	NET_EPOCH_EXIT(et);
1044 	return (0);
1045 }
1046 
1047 static void
1048 ipsec_delete_tunnel(struct ipsec_softc *sc)
1049 {
1050 	int i;
1051 
1052 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1053 
1054 	sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1055 	if (sc->family != 0) {
1056 		CK_LIST_REMOVE(sc, srchash);
1057 		sc->family = 0;
1058 		/*
1059 		 * Make sure that ipsec_if_input() will not do access
1060 		 * to softc's policies.
1061 		 */
1062 		IPSEC_WAIT();
1063 
1064 		key_unregister_ifnet(sc->sp, IPSEC_SPCOUNT);
1065 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1066 			key_freesp(&sc->sp[i]);
1067 	}
1068 }
1069