xref: /freebsd/sys/net/if_ipsec.c (revision bdd1243df58e60e85101c09001d9812a789b6bc4)
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_private.h>
56 #include <net/if_clone.h>
57 #include <net/if_types.h>
58 #include <net/bpf.h>
59 #include <net/route.h>
60 #include <net/vnet.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_encap.h>
66 
67 #include <netinet/ip6.h>
68 #include <netinet6/in6_var.h>
69 #include <netinet6/scope6_var.h>
70 
71 #include <netipsec/ipsec.h>
72 #ifdef INET6
73 #include <netipsec/ipsec6.h>
74 #endif
75 
76 #include <net/if_ipsec.h>
77 #include <netipsec/key.h>
78 
79 #include <security/mac/mac_framework.h>
80 
81 static MALLOC_DEFINE(M_IPSEC, "ipsec", "IPsec Virtual Tunnel Interface");
82 static const char ipsecname[] = "ipsec";
83 
84 #if defined(INET) && defined(INET6)
85 #define	IPSEC_SPCOUNT		4
86 #else
87 #define	IPSEC_SPCOUNT		2
88 #endif
89 
90 struct ipsec_softc {
91 	struct ifnet		*ifp;
92 	struct secpolicy	*sp[IPSEC_SPCOUNT];
93 	uint32_t		reqid;
94 	u_int			family;
95 	u_int			fibnum;
96 
97 	CK_LIST_ENTRY(ipsec_softc) idhash;
98 	CK_LIST_ENTRY(ipsec_softc) srchash;
99 };
100 
101 #define	IPSEC_RLOCK_TRACKER	struct epoch_tracker ipsec_et
102 #define	IPSEC_RLOCK()	epoch_enter_preempt(net_epoch_preempt, &ipsec_et)
103 #define	IPSEC_RUNLOCK()	epoch_exit_preempt(net_epoch_preempt, &ipsec_et)
104 #define	IPSEC_WAIT()	epoch_wait_preempt(net_epoch_preempt)
105 
106 #ifndef IPSEC_HASH_SIZE
107 #define	IPSEC_HASH_SIZE	(1 << 5)
108 #endif
109 
110 CK_LIST_HEAD(ipsec_iflist, ipsec_softc);
111 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec_idhtbl) = NULL;
112 #define	V_ipsec_idhtbl		VNET(ipsec_idhtbl)
113 
114 #ifdef INET
115 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec4_srchtbl) = NULL;
116 #define	V_ipsec4_srchtbl	VNET(ipsec4_srchtbl)
117 static const struct srcaddrtab *ipsec4_srctab = NULL;
118 #endif
119 
120 #ifdef INET6
121 VNET_DEFINE_STATIC(struct ipsec_iflist *, ipsec6_srchtbl) = NULL;
122 #define	V_ipsec6_srchtbl	VNET(ipsec6_srchtbl)
123 static const struct srcaddrtab *ipsec6_srctab = NULL;
124 #endif
125 
126 static struct ipsec_iflist *
127 ipsec_idhash(uint32_t id)
128 {
129 
130 	return (&V_ipsec_idhtbl[fnv_32_buf(&id, sizeof(id),
131 	    FNV1_32_INIT) & (IPSEC_HASH_SIZE - 1)]);
132 }
133 
134 static struct ipsec_iflist *
135 ipsec_srchash(const struct sockaddr *sa)
136 {
137 	uint32_t hval;
138 
139 	switch (sa->sa_family) {
140 #ifdef INET
141 	case AF_INET:
142 		hval = fnv_32_buf(
143 		    &((const struct sockaddr_in *)sa)->sin_addr.s_addr,
144 		    sizeof(in_addr_t), FNV1_32_INIT);
145 		return (&V_ipsec4_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
146 #endif
147 #ifdef INET6
148 	case AF_INET6:
149 		hval = fnv_32_buf(
150 		    &((const struct sockaddr_in6 *)sa)->sin6_addr,
151 		    sizeof(struct in6_addr), FNV1_32_INIT);
152 		return (&V_ipsec6_srchtbl[hval & (IPSEC_HASH_SIZE - 1)]);
153 #endif
154 	}
155 	return (NULL);
156 }
157 
158 /*
159  * ipsec_ioctl_sx protects from concurrent ioctls.
160  */
161 static struct sx ipsec_ioctl_sx;
162 SX_SYSINIT(ipsec_ioctl_sx, &ipsec_ioctl_sx, "ipsec_ioctl");
163 
164 static int	ipsec_init_reqid(struct ipsec_softc *);
165 static int	ipsec_set_tunnel(struct ipsec_softc *, struct sockaddr *,
166     struct sockaddr *, uint32_t);
167 static void	ipsec_delete_tunnel(struct ipsec_softc *);
168 
169 static int	ipsec_set_addresses(struct ifnet *, struct sockaddr *,
170     struct sockaddr *);
171 static int	ipsec_set_reqid(struct ipsec_softc *, uint32_t);
172 static void	ipsec_set_running(struct ipsec_softc *);
173 
174 #ifdef VIMAGE
175 static void	ipsec_reassign(struct ifnet *, struct vnet *, char *);
176 #endif
177 static void	ipsec_srcaddr(void *, const struct sockaddr *, int);
178 static int	ipsec_ioctl(struct ifnet *, u_long, caddr_t);
179 static int	ipsec_transmit(struct ifnet *, struct mbuf *);
180 static int	ipsec_output(struct ifnet *, struct mbuf *,
181     const struct sockaddr *, struct route *);
182 static void	ipsec_qflush(struct ifnet *);
183 static int	ipsec_clone_create(struct if_clone *, int, caddr_t);
184 static void	ipsec_clone_destroy(struct ifnet *);
185 
186 VNET_DEFINE_STATIC(struct if_clone *, ipsec_cloner);
187 #define	V_ipsec_cloner		VNET(ipsec_cloner)
188 
189 static int
190 ipsec_clone_create(struct if_clone *ifc, int unit, caddr_t params)
191 {
192 	struct ipsec_softc *sc;
193 	struct ifnet *ifp;
194 
195 	sc = malloc(sizeof(*sc), M_IPSEC, M_WAITOK | M_ZERO);
196 	sc->fibnum = curthread->td_proc->p_fibnum;
197 	sc->ifp = ifp = if_alloc(IFT_TUNNEL);
198 	ifp->if_softc = sc;
199 	if_initname(ifp, ipsecname, unit);
200 
201 	ifp->if_addrlen = 0;
202 	ifp->if_mtu = IPSEC_MTU;
203 	ifp->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
204 	ifp->if_ioctl  = ipsec_ioctl;
205 	ifp->if_transmit  = ipsec_transmit;
206 	ifp->if_qflush  = ipsec_qflush;
207 	ifp->if_output = ipsec_output;
208 #ifdef VIMAGE
209 	ifp->if_reassign = ipsec_reassign;
210 #endif
211 	if_attach(ifp);
212 	bpfattach(ifp, DLT_NULL, sizeof(uint32_t));
213 
214 	return (0);
215 }
216 
217 #ifdef VIMAGE
218 static void
219 ipsec_reassign(struct ifnet *ifp, struct vnet *new_vnet __unused,
220     char *unused __unused)
221 {
222 	struct ipsec_softc *sc;
223 
224 	sx_xlock(&ipsec_ioctl_sx);
225 	sc = ifp->if_softc;
226 	if (sc != NULL)
227 		ipsec_delete_tunnel(sc);
228 	sx_xunlock(&ipsec_ioctl_sx);
229 }
230 #endif /* VIMAGE */
231 
232 static void
233 ipsec_clone_destroy(struct ifnet *ifp)
234 {
235 	struct ipsec_softc *sc;
236 
237 	sx_xlock(&ipsec_ioctl_sx);
238 	sc = ifp->if_softc;
239 	ipsec_delete_tunnel(sc);
240 	/*
241 	 * Delete softc from idhash on interface destroy, since
242 	 * ipsec_delete_tunnel() keeps reqid unchanged.
243 	 */
244 	if (sc->reqid != 0)
245 		CK_LIST_REMOVE(sc, idhash);
246 	bpfdetach(ifp);
247 	if_detach(ifp);
248 	ifp->if_softc = NULL;
249 	sx_xunlock(&ipsec_ioctl_sx);
250 
251 	IPSEC_WAIT();
252 	if_free(ifp);
253 	free(sc, M_IPSEC);
254 }
255 
256 static struct ipsec_iflist *
257 ipsec_hashinit(void)
258 {
259 	struct ipsec_iflist *hash;
260 	int i;
261 
262 	hash = malloc(sizeof(struct ipsec_iflist) * IPSEC_HASH_SIZE,
263 	    M_IPSEC, M_WAITOK);
264 	for (i = 0; i < IPSEC_HASH_SIZE; i++)
265 		CK_LIST_INIT(&hash[i]);
266 
267 	return (hash);
268 }
269 
270 static void
271 vnet_ipsec_init(const void *unused __unused)
272 {
273 
274 	V_ipsec_idhtbl = ipsec_hashinit();
275 #ifdef INET
276 	V_ipsec4_srchtbl = ipsec_hashinit();
277 	if (IS_DEFAULT_VNET(curvnet))
278 		ipsec4_srctab = ip_encap_register_srcaddr(ipsec_srcaddr,
279 		    NULL, M_WAITOK);
280 #endif
281 #ifdef INET6
282 	V_ipsec6_srchtbl = ipsec_hashinit();
283 	if (IS_DEFAULT_VNET(curvnet))
284 		ipsec6_srctab = ip6_encap_register_srcaddr(ipsec_srcaddr,
285 		    NULL, M_WAITOK);
286 #endif
287 	V_ipsec_cloner = if_clone_simple(ipsecname, ipsec_clone_create,
288 	    ipsec_clone_destroy, 0);
289 }
290 VNET_SYSINIT(vnet_ipsec_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
291     vnet_ipsec_init, NULL);
292 
293 static void
294 vnet_ipsec_uninit(const void *unused __unused)
295 {
296 
297 	if_clone_detach(V_ipsec_cloner);
298 	free(V_ipsec_idhtbl, M_IPSEC);
299 	/*
300 	 * Use V_ipsec_idhtbl pointer as indicator that VNET is going to be
301 	 * destroyed, it is used by ipsec_srcaddr() callback.
302 	 */
303 	V_ipsec_idhtbl = NULL;
304 	IPSEC_WAIT();
305 
306 #ifdef INET
307 	if (IS_DEFAULT_VNET(curvnet))
308 		ip_encap_unregister_srcaddr(ipsec4_srctab);
309 	free(V_ipsec4_srchtbl, M_IPSEC);
310 #endif
311 #ifdef INET6
312 	if (IS_DEFAULT_VNET(curvnet))
313 		ip6_encap_unregister_srcaddr(ipsec6_srctab);
314 	free(V_ipsec6_srchtbl, M_IPSEC);
315 #endif
316 }
317 VNET_SYSUNINIT(vnet_ipsec_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
318     vnet_ipsec_uninit, NULL);
319 
320 static struct secpolicy *
321 ipsec_getpolicy(struct ipsec_softc *sc, int dir, sa_family_t af)
322 {
323 
324 	switch (af) {
325 #ifdef INET
326 	case AF_INET:
327 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)]);
328 #endif
329 #ifdef INET6
330 	case AF_INET6:
331 		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)
332 #ifdef INET
333 			+ 2
334 #endif
335 		]);
336 #endif
337 	}
338 	return (NULL);
339 }
340 
341 static struct secasindex *
342 ipsec_getsaidx(struct ipsec_softc *sc, int dir, sa_family_t af)
343 {
344 	struct secpolicy *sp;
345 
346 	sp = ipsec_getpolicy(sc, dir, af);
347 	if (sp == NULL)
348 		return (NULL);
349 	return (&sp->req[0]->saidx);
350 }
351 
352 static int
353 ipsec_transmit(struct ifnet *ifp, struct mbuf *m)
354 {
355 	IPSEC_RLOCK_TRACKER;
356 	struct ipsec_softc *sc;
357 	struct secpolicy *sp;
358 	struct ip *ip;
359 	uint32_t af;
360 	int error;
361 
362 	IPSEC_RLOCK();
363 #ifdef MAC
364 	error = mac_ifnet_check_transmit(ifp, m);
365 	if (error) {
366 		m_freem(m);
367 		goto err;
368 	}
369 #endif
370 	error = ENETDOWN;
371 	sc = ifp->if_softc;
372 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
373 	    (ifp->if_flags & IFF_MONITOR) != 0 ||
374 	    (ifp->if_flags & IFF_UP) == 0 || sc->family == 0) {
375 		m_freem(m);
376 		goto err;
377 	}
378 
379 	/* Determine address family to correctly handle packet in BPF */
380 	ip = mtod(m, struct ip *);
381 	switch (ip->ip_v) {
382 #ifdef INET
383 	case IPVERSION:
384 		af = AF_INET;
385 		break;
386 #endif
387 #ifdef INET6
388 	case (IPV6_VERSION >> 4):
389 		af = AF_INET6;
390 		break;
391 #endif
392 	default:
393 		error = EAFNOSUPPORT;
394 		m_freem(m);
395 		goto err;
396 	}
397 
398 	/*
399 	 * Loop prevention.
400 	 * XXX: for now just check presence of IPSEC_OUT_DONE mbuf tag.
401 	 *      We can read full chain and compare destination address,
402 	 *      proto and mode from xform_history with values from softc.
403 	 */
404 	if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) {
405 		m_freem(m);
406 		goto err;
407 	}
408 
409 	sp = ipsec_getpolicy(sc, IPSEC_DIR_OUTBOUND, af);
410 	key_addref(sp);
411 	M_SETFIB(m, sc->fibnum);
412 
413 	BPF_MTAP2(ifp, &af, sizeof(af), m);
414 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
415 	if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
416 
417 	switch (af) {
418 #ifdef INET
419 	case AF_INET:
420 		error = ipsec4_process_packet(m, sp, NULL);
421 		break;
422 #endif
423 #ifdef INET6
424 	case AF_INET6:
425 		error = ipsec6_process_packet(m, sp, NULL);
426 		break;
427 #endif
428 	default:
429 		panic("%s: unknown address family\n", __func__);
430 	}
431 err:
432 	if (error != 0)
433 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
434 	IPSEC_RUNLOCK();
435 	return (error);
436 }
437 
438 static void
439 ipsec_qflush(struct ifnet *ifp __unused)
440 {
441 
442 }
443 
444 static int
445 ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
446 	struct route *ro)
447 {
448 
449 	return (ifp->if_transmit(ifp, m));
450 }
451 
452 int
453 ipsec_if_input(struct mbuf *m, struct secasvar *sav, uint32_t af)
454 {
455 	IPSEC_RLOCK_TRACKER;
456 	struct secasindex *saidx;
457 	struct ipsec_softc *sc;
458 	struct ifnet *ifp;
459 
460 	if (sav->state != SADB_SASTATE_MATURE &&
461 	    sav->state != SADB_SASTATE_DYING) {
462 		m_freem(m);
463 		return (ENETDOWN);
464 	}
465 
466 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL ||
467 	    sav->sah->saidx.proto != IPPROTO_ESP)
468 		return (0);
469 
470 	IPSEC_RLOCK();
471 	CK_LIST_FOREACH(sc, ipsec_idhash(sav->sah->saidx.reqid), idhash) {
472 		if (sc->family == 0)
473 			continue;
474 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_INBOUND,
475 		    sav->sah->saidx.src.sa.sa_family);
476 		/* SA's reqid should match reqid in SP */
477 		if (saidx == NULL ||
478 		    sav->sah->saidx.reqid != saidx->reqid)
479 			continue;
480 		/* SAH's addresses should match tunnel endpoints. */
481 		if (key_sockaddrcmp(&sav->sah->saidx.dst.sa,
482 		    &saidx->dst.sa, 0) != 0)
483 			continue;
484 		if (key_sockaddrcmp(&sav->sah->saidx.src.sa,
485 		    &saidx->src.sa, 0) == 0)
486 			break;
487 	}
488 	if (sc == NULL) {
489 		IPSEC_RUNLOCK();
490 		/* Tunnel was not found. Nothing to do. */
491 		return (0);
492 	}
493 	ifp = sc->ifp;
494 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
495 	    (ifp->if_flags & IFF_UP) == 0) {
496 		IPSEC_RUNLOCK();
497 		m_freem(m);
498 		return (ENETDOWN);
499 	}
500 	/*
501 	 * We found matching and working tunnel.
502 	 * Set its ifnet as receiving interface.
503 	 */
504 	m->m_pkthdr.rcvif = ifp;
505 
506 	m_clrprotoflags(m);
507 	M_SETFIB(m, ifp->if_fib);
508 	BPF_MTAP2(ifp, &af, sizeof(af), m);
509 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
510 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
511 	if ((ifp->if_flags & IFF_MONITOR) != 0) {
512 		IPSEC_RUNLOCK();
513 		m_freem(m);
514 		return (ENETDOWN);
515 	}
516 	IPSEC_RUNLOCK();
517 	return (0);
518 }
519 
520 static int
521 ipsec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
522 {
523 	struct ifreq *ifr = (struct ifreq*)data;
524 	struct sockaddr *dst, *src;
525 	struct ipsec_softc *sc;
526 	struct secasindex *saidx;
527 #ifdef INET
528 	struct sockaddr_in *sin = NULL;
529 #endif
530 #ifdef INET6
531 	struct sockaddr_in6 *sin6 = NULL;
532 #endif
533 	uint32_t reqid;
534 	int error;
535 
536 	switch (cmd) {
537 	case SIOCSIFADDR:
538 		ifp->if_flags |= IFF_UP;
539 	case SIOCADDMULTI:
540 	case SIOCDELMULTI:
541 	case SIOCGIFMTU:
542 	case SIOCSIFFLAGS:
543 		return (0);
544 	case SIOCSIFMTU:
545 		if (ifr->ifr_mtu < IPSEC_MTU_MIN ||
546 		    ifr->ifr_mtu > IPSEC_MTU_MAX)
547 			return (EINVAL);
548 		else
549 			ifp->if_mtu = ifr->ifr_mtu;
550 		return (0);
551 	}
552 	sx_xlock(&ipsec_ioctl_sx);
553 	sc = ifp->if_softc;
554 	/* Check that softc is still here */
555 	if (sc == NULL) {
556 		error = ENXIO;
557 		goto bad;
558 	}
559 	error = 0;
560 	switch (cmd) {
561 	case SIOCSIFPHYADDR:
562 #ifdef INET6
563 	case SIOCSIFPHYADDR_IN6:
564 #endif
565 		error = EINVAL;
566 		switch (cmd) {
567 #ifdef INET
568 		case SIOCSIFPHYADDR:
569 			src = (struct sockaddr *)
570 				&(((struct in_aliasreq *)data)->ifra_addr);
571 			dst = (struct sockaddr *)
572 				&(((struct in_aliasreq *)data)->ifra_dstaddr);
573 			break;
574 #endif
575 #ifdef INET6
576 		case SIOCSIFPHYADDR_IN6:
577 			src = (struct sockaddr *)
578 				&(((struct in6_aliasreq *)data)->ifra_addr);
579 			dst = (struct sockaddr *)
580 				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
581 			break;
582 #endif
583 		default:
584 			goto bad;
585 		}
586 		/* sa_family must be equal */
587 		if (src->sa_family != dst->sa_family ||
588 		    src->sa_len != dst->sa_len)
589 			goto bad;
590 
591 		/* validate sa_len */
592 		switch (src->sa_family) {
593 #ifdef INET
594 		case AF_INET:
595 			if (src->sa_len != sizeof(struct sockaddr_in))
596 				goto bad;
597 			break;
598 #endif
599 #ifdef INET6
600 		case AF_INET6:
601 			if (src->sa_len != sizeof(struct sockaddr_in6))
602 				goto bad;
603 			break;
604 #endif
605 		default:
606 			error = EAFNOSUPPORT;
607 			goto bad;
608 		}
609 		/* check sa_family looks sane for the cmd */
610 		error = EAFNOSUPPORT;
611 		switch (cmd) {
612 #ifdef INET
613 		case SIOCSIFPHYADDR:
614 			if (src->sa_family == AF_INET)
615 				break;
616 			goto bad;
617 #endif
618 #ifdef INET6
619 		case SIOCSIFPHYADDR_IN6:
620 			if (src->sa_family == AF_INET6)
621 				break;
622 			goto bad;
623 #endif
624 		}
625 		error = EADDRNOTAVAIL;
626 		switch (src->sa_family) {
627 #ifdef INET
628 		case AF_INET:
629 			if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
630 			    satosin(dst)->sin_addr.s_addr == INADDR_ANY)
631 				goto bad;
632 			break;
633 #endif
634 #ifdef INET6
635 		case AF_INET6:
636 			if (IN6_IS_ADDR_UNSPECIFIED(
637 			    &satosin6(src)->sin6_addr) ||
638 			    IN6_IS_ADDR_UNSPECIFIED(
639 			    &satosin6(dst)->sin6_addr))
640 				goto bad;
641 			/*
642 			 * Check validity of the scope zone ID of the
643 			 * addresses, and convert it into the kernel
644 			 * internal form if necessary.
645 			 */
646 			error = sa6_embedscope(satosin6(src), 0);
647 			if (error != 0)
648 				goto bad;
649 			error = sa6_embedscope(satosin6(dst), 0);
650 			if (error != 0)
651 				goto bad;
652 #endif
653 		};
654 		error = ipsec_set_addresses(ifp, src, dst);
655 		break;
656 	case SIOCDIFPHYADDR:
657 		ipsec_delete_tunnel(sc);
658 		break;
659 	case SIOCGIFPSRCADDR:
660 	case SIOCGIFPDSTADDR:
661 #ifdef INET6
662 	case SIOCGIFPSRCADDR_IN6:
663 	case SIOCGIFPDSTADDR_IN6:
664 #endif
665 		if (sc->family == 0) {
666 			error = EADDRNOTAVAIL;
667 			break;
668 		}
669 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
670 		if (saidx == NULL) {
671 			error = ENXIO;
672 			break;
673 		}
674 		switch (cmd) {
675 #ifdef INET
676 		case SIOCGIFPSRCADDR:
677 		case SIOCGIFPDSTADDR:
678 			if (saidx->src.sa.sa_family != AF_INET) {
679 				error = EADDRNOTAVAIL;
680 				break;
681 			}
682 			sin = (struct sockaddr_in *)&ifr->ifr_addr;
683 			memset(sin, 0, sizeof(*sin));
684 			sin->sin_family = AF_INET;
685 			sin->sin_len = sizeof(*sin);
686 			break;
687 #endif
688 #ifdef INET6
689 		case SIOCGIFPSRCADDR_IN6:
690 		case SIOCGIFPDSTADDR_IN6:
691 			if (saidx->src.sa.sa_family != AF_INET6) {
692 				error = EADDRNOTAVAIL;
693 				break;
694 			}
695 			sin6 = (struct sockaddr_in6 *)
696 				&(((struct in6_ifreq *)data)->ifr_addr);
697 			memset(sin6, 0, sizeof(*sin6));
698 			sin6->sin6_family = AF_INET6;
699 			sin6->sin6_len = sizeof(*sin6);
700 			break;
701 #endif
702 		default:
703 			error = EAFNOSUPPORT;
704 		}
705 		if (error == 0) {
706 			switch (cmd) {
707 #ifdef INET
708 			case SIOCGIFPSRCADDR:
709 				sin->sin_addr = saidx->src.sin.sin_addr;
710 				break;
711 			case SIOCGIFPDSTADDR:
712 				sin->sin_addr = saidx->dst.sin.sin_addr;
713 				break;
714 #endif
715 #ifdef INET6
716 			case SIOCGIFPSRCADDR_IN6:
717 				sin6->sin6_addr = saidx->src.sin6.sin6_addr;
718 				break;
719 			case SIOCGIFPDSTADDR_IN6:
720 				sin6->sin6_addr = saidx->dst.sin6.sin6_addr;
721 				break;
722 #endif
723 			}
724 		}
725 		if (error != 0)
726 			break;
727 		switch (cmd) {
728 #ifdef INET
729 		case SIOCGIFPSRCADDR:
730 		case SIOCGIFPDSTADDR:
731 			error = prison_if(curthread->td_ucred,
732 			    (struct sockaddr *)sin);
733 			if (error != 0)
734 				memset(sin, 0, sizeof(*sin));
735 			break;
736 #endif
737 #ifdef INET6
738 		case SIOCGIFPSRCADDR_IN6:
739 		case SIOCGIFPDSTADDR_IN6:
740 			error = prison_if(curthread->td_ucred,
741 			    (struct sockaddr *)sin6);
742 			if (error == 0)
743 				error = sa6_recoverscope(sin6);
744 			if (error != 0)
745 				memset(sin6, 0, sizeof(*sin6));
746 #endif
747 		}
748 		break;
749 	case SIOCGTUNFIB:
750 		ifr->ifr_fib = sc->fibnum;
751 		break;
752 	case SIOCSTUNFIB:
753 		if ((error = priv_check(curthread, PRIV_NET_SETIFFIB)) != 0)
754 			break;
755 		if (ifr->ifr_fib >= rt_numfibs)
756 			error = EINVAL;
757 		else
758 			sc->fibnum = ifr->ifr_fib;
759 		break;
760 	case IPSECGREQID:
761 		reqid = sc->reqid;
762 		error = copyout(&reqid, ifr_data_get_ptr(ifr), sizeof(reqid));
763 		break;
764 	case IPSECSREQID:
765 		if ((error = priv_check(curthread, PRIV_NET_SETIFCAP)) != 0)
766 			break;
767 		error = copyin(ifr_data_get_ptr(ifr), &reqid, sizeof(reqid));
768 		if (error != 0)
769 			break;
770 		error = ipsec_set_reqid(sc, reqid);
771 		break;
772 	default:
773 		error = EINVAL;
774 		break;
775 	}
776 bad:
777 	sx_xunlock(&ipsec_ioctl_sx);
778 	return (error);
779 }
780 
781 /*
782  * Check that ingress address belongs to local host.
783  */
784 static void
785 ipsec_set_running(struct ipsec_softc *sc)
786 {
787 	struct secasindex *saidx;
788 	int localip;
789 
790 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
791 	if (saidx == NULL)
792 		return;
793 	localip = 0;
794 	switch (sc->family) {
795 #ifdef INET
796 	case AF_INET:
797 		localip = in_localip(saidx->src.sin.sin_addr);
798 		break;
799 #endif
800 #ifdef INET6
801 	case AF_INET6:
802 		localip = in6_localip(&saidx->src.sin6.sin6_addr);
803 		break;
804 #endif
805 	}
806 	if (localip != 0)
807 		sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
808 	else
809 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
810 }
811 
812 /*
813  * ifaddr_event handler.
814  * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent
815  * source address spoofing.
816  */
817 static void
818 ipsec_srcaddr(void *arg __unused, const struct sockaddr *sa,
819     int event __unused)
820 {
821 	struct ipsec_softc *sc;
822 	struct secasindex *saidx;
823 	struct ipsec_iflist *iflist;
824 
825 	/* Check that VNET is ready */
826 	if (V_ipsec_idhtbl == NULL)
827 		return;
828 
829 	NET_EPOCH_ASSERT();
830 	iflist = ipsec_srchash(sa);
831 	if (iflist == NULL)
832 		return;
833 	CK_LIST_FOREACH(sc, iflist, srchash) {
834 		if (sc->family == 0)
835 			continue;
836 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sa->sa_family);
837 		if (saidx == NULL ||
838 		    key_sockaddrcmp(&saidx->src.sa, sa, 0) != 0)
839 			continue;
840 		ipsec_set_running(sc);
841 	}
842 }
843 
844 /*
845  * Allocate new private security policies for tunneling interface.
846  * Each tunneling interface has following security policies for
847  * both AF:
848  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P in \
849  *	ipsec esp/tunnel/RemoteIP-LocalIP/unique:reqid
850  *   0.0.0.0/0[any] 0.0.0.0/0[any] -P out \
851  *	ipsec esp/tunnel/LocalIP-RemoteIP/unique:reqid
852  */
853 static int
854 ipsec_newpolicies(struct ipsec_softc *sc, struct secpolicy *sp[IPSEC_SPCOUNT],
855     const struct sockaddr *src, const struct sockaddr *dst, uint32_t reqid)
856 {
857 	struct ipsecrequest *isr;
858 	int i;
859 
860 	memset(sp, 0, sizeof(struct secpolicy *) * IPSEC_SPCOUNT);
861 	for (i = 0; i < IPSEC_SPCOUNT; i++) {
862 		if ((sp[i] = key_newsp()) == NULL)
863 			goto fail;
864 		if ((isr = ipsec_newisr()) == NULL)
865 			goto fail;
866 
867 		sp[i]->policy = IPSEC_POLICY_IPSEC;
868 		sp[i]->state = IPSEC_SPSTATE_DEAD;
869 		sp[i]->req[sp[i]->tcount++] = isr;
870 		sp[i]->created = time_second;
871 		/* Use priority field to store if_index */
872 		sp[i]->priority = sc->ifp->if_index;
873 		isr->level = IPSEC_LEVEL_UNIQUE;
874 		isr->saidx.proto = IPPROTO_ESP;
875 		isr->saidx.mode = IPSEC_MODE_TUNNEL;
876 		isr->saidx.reqid = reqid;
877 		if (i % 2 == 0) {
878 			sp[i]->spidx.dir = IPSEC_DIR_INBOUND;
879 			bcopy(src, &isr->saidx.dst, src->sa_len);
880 			bcopy(dst, &isr->saidx.src, dst->sa_len);
881 		} else {
882 			sp[i]->spidx.dir = IPSEC_DIR_OUTBOUND;
883 			bcopy(src, &isr->saidx.src, src->sa_len);
884 			bcopy(dst, &isr->saidx.dst, dst->sa_len);
885 		}
886 		sp[i]->spidx.ul_proto = IPSEC_ULPROTO_ANY;
887 #ifdef INET
888 		if (i < 2) {
889 			sp[i]->spidx.src.sa.sa_family =
890 			    sp[i]->spidx.dst.sa.sa_family = AF_INET;
891 			sp[i]->spidx.src.sa.sa_len =
892 			    sp[i]->spidx.dst.sa.sa_len =
893 			    sizeof(struct sockaddr_in);
894 			continue;
895 		}
896 #endif
897 #ifdef INET6
898 		sp[i]->spidx.src.sa.sa_family =
899 		    sp[i]->spidx.dst.sa.sa_family = AF_INET6;
900 		sp[i]->spidx.src.sa.sa_len =
901 		    sp[i]->spidx.dst.sa.sa_len = sizeof(struct sockaddr_in6);
902 #endif
903 	}
904 	return (0);
905 fail:
906 	for (i = 0; i < IPSEC_SPCOUNT; i++)
907 		key_freesp(&sp[i]);
908 	return (ENOMEM);
909 }
910 
911 static int
912 ipsec_check_reqid(uint32_t reqid)
913 {
914 	struct ipsec_softc *sc;
915 
916 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
917 	CK_LIST_FOREACH(sc, ipsec_idhash(reqid), idhash) {
918 		if (sc->reqid == reqid)
919 			return (EEXIST);
920 	}
921 	return (0);
922 }
923 
924 /*
925  * We use key_newreqid() to automatically obtain unique reqid.
926  * Then we check that given id is unique, i.e. it is not used by
927  * another if_ipsec(4) interface. This macro limits the number of
928  * tries to get unique id.
929  */
930 #define	IPSEC_REQID_TRYCNT	64
931 static int
932 ipsec_init_reqid(struct ipsec_softc *sc)
933 {
934 	uint32_t reqid;
935 	int trycount;
936 
937 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
938 	if (sc->reqid != 0) /* already initialized */
939 		return (0);
940 
941 	trycount = IPSEC_REQID_TRYCNT;
942 	while (--trycount > 0) {
943 		reqid = key_newreqid();
944 		if (ipsec_check_reqid(reqid) == 0)
945 			break;
946 	}
947 	if (trycount == 0)
948 		return (EEXIST);
949 	sc->reqid = reqid;
950 	CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
951 	return (0);
952 }
953 
954 /*
955  * Set or update reqid for given tunneling interface.
956  * When specified reqid is zero, generate new one.
957  * We are protected by ioctl_sx lock from concurrent id generation.
958  * Also softc would not disappear while we hold ioctl_sx lock.
959  */
960 static int
961 ipsec_set_reqid(struct ipsec_softc *sc, uint32_t reqid)
962 {
963 	struct secasindex *saidx;
964 
965 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
966 
967 	if (sc->reqid == reqid && reqid != 0)
968 		return (0);
969 
970 	if (reqid != 0) {
971 		/* Check that specified reqid doesn't exist */
972 		if (ipsec_check_reqid(reqid) != 0)
973 			return (EEXIST);
974 		if (sc->reqid != 0) {
975 			CK_LIST_REMOVE(sc, idhash);
976 			IPSEC_WAIT();
977 		}
978 		sc->reqid = reqid;
979 		CK_LIST_INSERT_HEAD(ipsec_idhash(reqid), sc, idhash);
980 	} else {
981 		/* Generate new reqid */
982 		if (ipsec_init_reqid(sc) != 0)
983 			return (EEXIST);
984 	}
985 
986 	/* Tunnel isn't fully configured, just return. */
987 	if (sc->family == 0)
988 		return (0);
989 
990 	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
991 	KASSERT(saidx != NULL,
992 	    ("saidx is NULL, but family is %d", sc->family));
993 	return (ipsec_set_tunnel(sc, &saidx->src.sa, &saidx->dst.sa,
994 	    sc->reqid));
995 }
996 
997 /*
998  * Set tunnel endpoints addresses.
999  */
1000 static int
1001 ipsec_set_addresses(struct ifnet *ifp, struct sockaddr *src,
1002     struct sockaddr *dst)
1003 {
1004 	struct ipsec_softc *sc;
1005 	struct secasindex *saidx;
1006 
1007 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1008 
1009 	sc = ifp->if_softc;
1010 	if (sc->family != 0) {
1011 		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND,
1012 		    src->sa_family);
1013 		if (saidx != NULL && saidx->reqid == sc->reqid &&
1014 		    key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
1015 		    key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0)
1016 			return (0); /* Nothing has been changed. */
1017 	}
1018 	/* If reqid is not set, generate new one. */
1019 	if (ipsec_init_reqid(sc) != 0)
1020 		return (EEXIST);
1021 	return (ipsec_set_tunnel(sc, src, dst, sc->reqid));
1022 }
1023 
1024 static int
1025 ipsec_set_tunnel(struct ipsec_softc *sc, struct sockaddr *src,
1026     struct sockaddr *dst, uint32_t reqid)
1027 {
1028 	struct epoch_tracker et;
1029 	struct ipsec_iflist *iflist;
1030 	struct secpolicy *sp[IPSEC_SPCOUNT];
1031 	int i;
1032 
1033 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1034 
1035 	/* Allocate SP with new addresses. */
1036 	iflist = ipsec_srchash(src);
1037 	if (iflist == NULL) {
1038 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1039 		return (EAFNOSUPPORT);
1040 	}
1041 	if (ipsec_newpolicies(sc, sp, src, dst, reqid) == 0) {
1042 		/* Add new policies to SPDB */
1043 		if (key_register_ifnet(sp, IPSEC_SPCOUNT) != 0) {
1044 			for (i = 0; i < IPSEC_SPCOUNT; i++)
1045 				key_freesp(&sp[i]);
1046 			return (EAGAIN);
1047 		}
1048 		if (sc->family != 0)
1049 			ipsec_delete_tunnel(sc);
1050 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1051 			sc->sp[i] = sp[i];
1052 		sc->family = src->sa_family;
1053 		CK_LIST_INSERT_HEAD(iflist, sc, srchash);
1054 	} else {
1055 		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1056 		return (ENOMEM);
1057 	}
1058 	NET_EPOCH_ENTER(et);
1059 	ipsec_set_running(sc);
1060 	NET_EPOCH_EXIT(et);
1061 	return (0);
1062 }
1063 
1064 static void
1065 ipsec_delete_tunnel(struct ipsec_softc *sc)
1066 {
1067 	int i;
1068 
1069 	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
1070 
1071 	sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1072 	if (sc->family != 0) {
1073 		CK_LIST_REMOVE(sc, srchash);
1074 		sc->family = 0;
1075 		/*
1076 		 * Make sure that ipsec_if_input() will not do access
1077 		 * to softc's policies.
1078 		 */
1079 		IPSEC_WAIT();
1080 
1081 		key_unregister_ifnet(sc->sp, IPSEC_SPCOUNT);
1082 		for (i = 0; i < IPSEC_SPCOUNT; i++)
1083 			key_freesp(&sc->sp[i]);
1084 	}
1085 }
1086