if_ipsec.c revision 315514
1/*-
2 * Copyright (c) 2016 Yandex LLC
3 * Copyright (c) 2016 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: stable/11/sys/net/if_ipsec.c 315514 2017-03-18 22:04:20Z ae $");
30
31#include "opt_inet.h"
32#include "opt_inet6.h"
33
34#include <sys/param.h>
35#include <sys/systm.h>
36#include <sys/kernel.h>
37#include <sys/fnv_hash.h>
38#include <sys/jail.h>
39#include <sys/lock.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/module.h>
43#include <sys/rmlock.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
65#include <netinet/ip6.h>
66#include <netinet6/in6_var.h>
67#include <netinet6/scope6_var.h>
68
69#include <netipsec/ipsec.h>
70#ifdef INET6
71#include <netipsec/ipsec6.h>
72#endif
73
74#include <net/if_ipsec.h>
75#include <netipsec/key.h>
76
77#include <security/mac/mac_framework.h>
78
79static MALLOC_DEFINE(M_IPSEC, "ipsec", "IPsec Virtual Tunnel Interface");
80static const char ipsecname[] = "ipsec";
81
82#if defined(INET) && defined(INET6)
83#define	IPSEC_SPCOUNT		4
84#else
85#define	IPSEC_SPCOUNT		2
86#endif
87
88struct ipsec_softc {
89	struct ifnet		*ifp;
90
91	struct rmlock		lock;
92	struct secpolicy	*sp[IPSEC_SPCOUNT];
93
94	uint32_t		reqid;
95	u_int			family;
96	u_int			fibnum;
97	LIST_ENTRY(ipsec_softc)	chain;
98	LIST_ENTRY(ipsec_softc) hash;
99};
100
101#define	IPSEC_LOCK_INIT(sc)	rm_init(&(sc)->lock, "if_ipsec softc")
102#define	IPSEC_LOCK_DESTROY(sc)	rm_destroy(&(sc)->lock)
103#define	IPSEC_RLOCK_TRACKER	struct rm_priotracker ipsec_tracker
104#define	IPSEC_RLOCK(sc)		rm_rlock(&(sc)->lock, &ipsec_tracker)
105#define	IPSEC_RUNLOCK(sc)	rm_runlock(&(sc)->lock, &ipsec_tracker)
106#define	IPSEC_RLOCK_ASSERT(sc)	rm_assert(&(sc)->lock, RA_RLOCKED)
107#define	IPSEC_WLOCK(sc)		rm_wlock(&(sc)->lock)
108#define	IPSEC_WUNLOCK(sc)	rm_wunlock(&(sc)->lock)
109#define	IPSEC_WLOCK_ASSERT(sc)	rm_assert(&(sc)->lock, RA_WLOCKED)
110
111static struct rmlock ipsec_sc_lock;
112RM_SYSINIT(ipsec_sc_lock, &ipsec_sc_lock, "if_ipsec softc list");
113
114#define	IPSEC_SC_RLOCK_TRACKER	struct rm_priotracker ipsec_sc_tracker
115#define	IPSEC_SC_RLOCK()	rm_rlock(&ipsec_sc_lock, &ipsec_sc_tracker)
116#define	IPSEC_SC_RUNLOCK()	rm_runlock(&ipsec_sc_lock, &ipsec_sc_tracker)
117#define	IPSEC_SC_RLOCK_ASSERT()	rm_assert(&ipsec_sc_lock, RA_RLOCKED)
118#define	IPSEC_SC_WLOCK()	rm_wlock(&ipsec_sc_lock)
119#define	IPSEC_SC_WUNLOCK()	rm_wunlock(&ipsec_sc_lock)
120#define	IPSEC_SC_WLOCK_ASSERT()	rm_assert(&ipsec_sc_lock, RA_WLOCKED)
121
122LIST_HEAD(ipsec_iflist, ipsec_softc);
123static VNET_DEFINE(struct ipsec_iflist, ipsec_sc_list);
124static VNET_DEFINE(struct ipsec_iflist *, ipsec_sc_htbl);
125static VNET_DEFINE(u_long, ipsec_sc_hmask);
126#define	V_ipsec_sc_list		VNET(ipsec_sc_list)
127#define	V_ipsec_sc_htbl		VNET(ipsec_sc_htbl)
128#define	V_ipsec_sc_hmask	VNET(ipsec_sc_hmask)
129
130static uint32_t
131ipsec_hash(uint32_t id)
132{
133
134	return (fnv_32_buf(&id, sizeof(id), FNV1_32_INIT));
135}
136
137#define	SCHASH_NHASH_LOG2	5
138#define	SCHASH_NHASH		(1 << SCHASH_NHASH_LOG2)
139#define	SCHASH_HASHVAL(id)	(ipsec_hash((id)) & V_ipsec_sc_hmask)
140#define	SCHASH_HASH(id)		&V_ipsec_sc_htbl[SCHASH_HASHVAL(id)]
141
142/*
143 * ipsec_ioctl_sx protects from concurrent ioctls.
144 */
145static struct sx ipsec_ioctl_sx;
146SX_SYSINIT(ipsec_ioctl_sx, &ipsec_ioctl_sx, "ipsec_ioctl");
147
148static int	ipsec_init_reqid(struct ipsec_softc *);
149static int	ipsec_set_tunnel(struct ipsec_softc *, struct sockaddr *,
150    struct sockaddr *, uint32_t);
151static void	ipsec_delete_tunnel(struct ifnet *, int);
152
153static int	ipsec_set_addresses(struct ifnet *, struct sockaddr *,
154    struct sockaddr *);
155static int	ipsec_set_reqid(struct ifnet *, uint32_t);
156
157static int	ipsec_ioctl(struct ifnet *, u_long, caddr_t);
158static int	ipsec_transmit(struct ifnet *, struct mbuf *);
159static int	ipsec_output(struct ifnet *, struct mbuf *,
160    const struct sockaddr *, struct route *);
161static void	ipsec_qflush(struct ifnet *);
162static int	ipsec_clone_create(struct if_clone *, int, caddr_t);
163static void	ipsec_clone_destroy(struct ifnet *);
164
165static VNET_DEFINE(struct if_clone *, ipsec_cloner);
166#define	V_ipsec_cloner		VNET(ipsec_cloner)
167
168static int
169ipsec_clone_create(struct if_clone *ifc, int unit, caddr_t params)
170{
171	struct ipsec_softc *sc;
172	struct ifnet *ifp;
173
174	sc = malloc(sizeof(*sc), M_IPSEC, M_WAITOK | M_ZERO);
175	sc->fibnum = curthread->td_proc->p_fibnum;
176	sc->ifp = ifp = if_alloc(IFT_TUNNEL);
177	IPSEC_LOCK_INIT(sc);
178	ifp->if_softc = sc;
179	if_initname(ifp, ipsecname, unit);
180
181	ifp->if_addrlen = 0;
182	ifp->if_mtu = IPSEC_MTU;
183	ifp->if_flags  = IFF_POINTOPOINT | IFF_MULTICAST;
184	ifp->if_ioctl  = ipsec_ioctl;
185	ifp->if_transmit  = ipsec_transmit;
186	ifp->if_qflush  = ipsec_qflush;
187	ifp->if_output = ipsec_output;
188	if_attach(ifp);
189	bpfattach(ifp, DLT_NULL, sizeof(uint32_t));
190
191	IPSEC_SC_WLOCK();
192	LIST_INSERT_HEAD(&V_ipsec_sc_list, sc, chain);
193	IPSEC_SC_WUNLOCK();
194	return (0);
195}
196
197static void
198ipsec_clone_destroy(struct ifnet *ifp)
199{
200	struct ipsec_softc *sc;
201
202	sx_xlock(&ipsec_ioctl_sx);
203	sc = ifp->if_softc;
204
205	IPSEC_SC_WLOCK();
206	ipsec_delete_tunnel(ifp, 1);
207	LIST_REMOVE(sc, chain);
208	IPSEC_SC_WUNLOCK();
209
210	bpfdetach(ifp);
211	if_detach(ifp);
212	ifp->if_softc = NULL;
213	sx_xunlock(&ipsec_ioctl_sx);
214
215	if_free(ifp);
216	IPSEC_LOCK_DESTROY(sc);
217	free(sc, M_IPSEC);
218}
219
220static void
221vnet_ipsec_init(const void *unused __unused)
222{
223
224	LIST_INIT(&V_ipsec_sc_list);
225	V_ipsec_sc_htbl = hashinit(SCHASH_NHASH, M_IPSEC, &V_ipsec_sc_hmask);
226	V_ipsec_cloner = if_clone_simple(ipsecname, ipsec_clone_create,
227	    ipsec_clone_destroy, 0);
228}
229VNET_SYSINIT(vnet_ipsec_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
230    vnet_ipsec_init, NULL);
231
232static void
233vnet_ipsec_uninit(const void *unused __unused)
234{
235
236	if_clone_detach(V_ipsec_cloner);
237	hashdestroy(V_ipsec_sc_htbl, M_IPSEC, V_ipsec_sc_hmask);
238}
239VNET_SYSUNINIT(vnet_ipsec_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
240    vnet_ipsec_uninit, NULL);
241
242static struct secpolicy *
243ipsec_getpolicy(struct ipsec_softc *sc, int dir, sa_family_t af)
244{
245
246	switch (af) {
247#ifdef INET
248	case AF_INET:
249		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)]);
250#endif
251#ifdef INET6
252	case AF_INET6:
253		return (sc->sp[(dir == IPSEC_DIR_INBOUND ? 0: 1)
254#ifdef INET
255			+ 2
256#endif
257		]);
258#endif
259	}
260	return (NULL);
261}
262
263static struct secasindex *
264ipsec_getsaidx(struct ipsec_softc *sc, int dir, sa_family_t af)
265{
266	struct secpolicy *sp;
267
268	sp = ipsec_getpolicy(sc, dir, af);
269	if (sp == NULL)
270		return (NULL);
271	return (&sp->req[0]->saidx);
272}
273
274static int
275ipsec_transmit(struct ifnet *ifp, struct mbuf *m)
276{
277	IPSEC_RLOCK_TRACKER;
278	struct ipsec_softc *sc;
279	struct secpolicy *sp;
280	struct ip *ip;
281	uint32_t af;
282	int error;
283
284#ifdef MAC
285	error = mac_ifnet_check_transmit(ifp, m);
286	if (error) {
287		m_freem(m);
288		goto err;
289	}
290#endif
291	error = ENETDOWN;
292	sc = ifp->if_softc;
293	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
294	    (ifp->if_flags & IFF_MONITOR) != 0 ||
295	    (ifp->if_flags & IFF_UP) == 0) {
296		m_freem(m);
297		goto err;
298	}
299
300	/* Determine address family to correctly handle packet in BPF */
301	ip = mtod(m, struct ip *);
302	switch (ip->ip_v) {
303#ifdef INET
304	case IPVERSION:
305		af = AF_INET;
306		break;
307#endif
308#ifdef INET6
309	case (IPV6_VERSION >> 4):
310		af = AF_INET6;
311		break;
312#endif
313	default:
314		error = EAFNOSUPPORT;
315		m_freem(m);
316		goto err;
317	}
318
319	/*
320	 * Loop prevention.
321	 * XXX: for now just check presence of IPSEC_OUT_DONE mbuf tag.
322	 *      We can read full chain and compare destination address,
323	 *      proto and mode from xform_history with values from softc.
324	 */
325	if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL) {
326		m_freem(m);
327		goto err;
328	}
329
330	IPSEC_RLOCK(sc);
331	if (sc->family == 0) {
332		IPSEC_RUNLOCK(sc);
333		m_freem(m);
334		goto err;
335	}
336	sp = ipsec_getpolicy(sc, IPSEC_DIR_OUTBOUND, af);
337	key_addref(sp);
338	M_SETFIB(m, sc->fibnum);
339	IPSEC_RUNLOCK(sc);
340
341	BPF_MTAP2(ifp, &af, sizeof(af), m);
342	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
343	if_inc_counter(ifp, IFCOUNTER_OBYTES, m->m_pkthdr.len);
344
345	switch (af) {
346#ifdef INET
347	case AF_INET:
348		error = ipsec4_process_packet(m, sp, NULL);
349		break;
350#endif
351#ifdef INET6
352	case AF_INET6:
353		error = ipsec6_process_packet(m, sp, NULL);
354		break;
355#endif
356	default:
357		panic("%s: unknown address family\n", __func__);
358	}
359err:
360	if (error != 0)
361		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
362	return (error);
363}
364
365static void
366ipsec_qflush(struct ifnet *ifp __unused)
367{
368
369}
370
371static int
372ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
373	struct route *ro)
374{
375
376	return (ifp->if_transmit(ifp, m));
377}
378
379int
380ipsec_if_input(struct mbuf *m, struct secasvar *sav, uint32_t af)
381{
382	IPSEC_SC_RLOCK_TRACKER;
383	struct secasindex *saidx;
384	struct ipsec_softc *sc;
385	struct ifnet *ifp;
386
387	if (sav->state != SADB_SASTATE_MATURE &&
388	    sav->state != SADB_SASTATE_DYING) {
389		m_freem(m);
390		return (ENETDOWN);
391	}
392
393	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL ||
394	    sav->sah->saidx.proto != IPPROTO_ESP)
395		return (0);
396
397	IPSEC_SC_RLOCK();
398	/*
399	 * We only acquire SC_RLOCK() while we are doing search in
400	 * ipsec_sc_htbl. It is safe, because removing softc or changing
401	 * of reqid/addresses requires removing from hash table.
402	 */
403	LIST_FOREACH(sc, SCHASH_HASH(sav->sah->saidx.reqid), hash) {
404		saidx = ipsec_getsaidx(sc, IPSEC_DIR_INBOUND,
405		    sav->sah->saidx.src.sa.sa_family);
406		/* SA's reqid should match reqid in SP */
407		if (saidx == NULL ||
408		    sav->sah->saidx.reqid != saidx->reqid)
409			continue;
410		/* SAH's addresses should match tunnel endpoints. */
411		if (key_sockaddrcmp(&sav->sah->saidx.dst.sa,
412		    &saidx->dst.sa, 0) != 0)
413			continue;
414		if (key_sockaddrcmp(&sav->sah->saidx.src.sa,
415		    &saidx->src.sa, 0) == 0)
416			break;
417	}
418	if (sc == NULL) {
419		IPSEC_SC_RUNLOCK();
420		/* Tunnel was not found. Nothing to do. */
421		return (0);
422	}
423	ifp = sc->ifp;
424	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
425	    (ifp->if_flags & IFF_UP) == 0) {
426		IPSEC_SC_RUNLOCK();
427		m_freem(m);
428		return (ENETDOWN);
429	}
430	/*
431	 * We found matching and working tunnel.
432	 * Set its ifnet as receiving interface.
433	 */
434	m->m_pkthdr.rcvif = ifp;
435	IPSEC_SC_RUNLOCK();
436
437	/* m_clrprotoflags(m); */
438	M_SETFIB(m, ifp->if_fib);
439	BPF_MTAP2(ifp, &af, sizeof(af), m);
440	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
441	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
442	if ((ifp->if_flags & IFF_MONITOR) != 0) {
443		m_freem(m);
444		return (ENETDOWN);
445	}
446	return (0);
447}
448
449/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */
450int
451ipsec_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
452{
453	IPSEC_RLOCK_TRACKER;
454	struct ifreq *ifr = (struct ifreq*)data;
455	struct sockaddr *dst, *src;
456	struct ipsec_softc *sc;
457	struct secasindex *saidx;
458#ifdef INET
459	struct sockaddr_in *sin = NULL;
460#endif
461#ifdef INET6
462	struct sockaddr_in6 *sin6 = NULL;
463#endif
464	uint32_t reqid;
465	int error;
466
467	switch (cmd) {
468	case SIOCSIFADDR:
469		ifp->if_flags |= IFF_UP;
470	case SIOCADDMULTI:
471	case SIOCDELMULTI:
472	case SIOCGIFMTU:
473	case SIOCSIFFLAGS:
474		return (0);
475	case SIOCSIFMTU:
476		if (ifr->ifr_mtu < IPSEC_MTU_MIN ||
477		    ifr->ifr_mtu > IPSEC_MTU_MAX)
478			return (EINVAL);
479		else
480			ifp->if_mtu = ifr->ifr_mtu;
481		return (0);
482	}
483	sx_xlock(&ipsec_ioctl_sx);
484	sc = ifp->if_softc;
485	/* Check that softc is still here */
486	if (sc == NULL) {
487		error = ENXIO;
488		goto bad;
489	}
490	error = 0;
491	switch (cmd) {
492	case SIOCSIFPHYADDR:
493#ifdef INET6
494	case SIOCSIFPHYADDR_IN6:
495#endif
496		error = EINVAL;
497		switch (cmd) {
498#ifdef INET
499		case SIOCSIFPHYADDR:
500			src = (struct sockaddr *)
501				&(((struct in_aliasreq *)data)->ifra_addr);
502			dst = (struct sockaddr *)
503				&(((struct in_aliasreq *)data)->ifra_dstaddr);
504			break;
505#endif
506#ifdef INET6
507		case SIOCSIFPHYADDR_IN6:
508			src = (struct sockaddr *)
509				&(((struct in6_aliasreq *)data)->ifra_addr);
510			dst = (struct sockaddr *)
511				&(((struct in6_aliasreq *)data)->ifra_dstaddr);
512			break;
513#endif
514		default:
515			goto bad;
516		}
517		/* sa_family must be equal */
518		if (src->sa_family != dst->sa_family ||
519		    src->sa_len != dst->sa_len)
520			goto bad;
521
522		/* validate sa_len */
523		switch (src->sa_family) {
524#ifdef INET
525		case AF_INET:
526			if (src->sa_len != sizeof(struct sockaddr_in))
527				goto bad;
528			break;
529#endif
530#ifdef INET6
531		case AF_INET6:
532			if (src->sa_len != sizeof(struct sockaddr_in6))
533				goto bad;
534			break;
535#endif
536		default:
537			error = EAFNOSUPPORT;
538			goto bad;
539		}
540		/* check sa_family looks sane for the cmd */
541		error = EAFNOSUPPORT;
542		switch (cmd) {
543#ifdef INET
544		case SIOCSIFPHYADDR:
545			if (src->sa_family == AF_INET)
546				break;
547			goto bad;
548#endif
549#ifdef INET6
550		case SIOCSIFPHYADDR_IN6:
551			if (src->sa_family == AF_INET6)
552				break;
553			goto bad;
554#endif
555		}
556		error = EADDRNOTAVAIL;
557		switch (src->sa_family) {
558#ifdef INET
559		case AF_INET:
560			if (satosin(src)->sin_addr.s_addr == INADDR_ANY ||
561			    satosin(dst)->sin_addr.s_addr == INADDR_ANY)
562				goto bad;
563			break;
564#endif
565#ifdef INET6
566		case AF_INET6:
567			if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(src)->sin6_addr)
568			    ||
569			    IN6_IS_ADDR_UNSPECIFIED(&satosin6(dst)->sin6_addr))
570				goto bad;
571			/*
572			 * Check validity of the scope zone ID of the
573			 * addresses, and convert it into the kernel
574			 * internal form if necessary.
575			 */
576			error = sa6_embedscope(satosin6(src), 0);
577			if (error != 0)
578				goto bad;
579			error = sa6_embedscope(satosin6(dst), 0);
580			if (error != 0)
581				goto bad;
582#endif
583		};
584		error = ipsec_set_addresses(ifp, src, dst);
585		break;
586	case SIOCDIFPHYADDR:
587		ipsec_delete_tunnel(ifp, 0);
588		break;
589	case SIOCGIFPSRCADDR:
590	case SIOCGIFPDSTADDR:
591#ifdef INET6
592	case SIOCGIFPSRCADDR_IN6:
593	case SIOCGIFPDSTADDR_IN6:
594#endif
595		IPSEC_RLOCK(sc);
596		if (sc->family == 0) {
597			IPSEC_RUNLOCK(sc);
598			error = EADDRNOTAVAIL;
599			break;
600		}
601		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
602		switch (cmd) {
603#ifdef INET
604		case SIOCGIFPSRCADDR:
605		case SIOCGIFPDSTADDR:
606			if (saidx->src.sa.sa_family != AF_INET) {
607				error = EADDRNOTAVAIL;
608				break;
609			}
610			sin = (struct sockaddr_in *)&ifr->ifr_addr;
611			memset(sin, 0, sizeof(*sin));
612			sin->sin_family = AF_INET;
613			sin->sin_len = sizeof(*sin);
614			break;
615#endif
616#ifdef INET6
617		case SIOCGIFPSRCADDR_IN6:
618		case SIOCGIFPDSTADDR_IN6:
619			if (saidx->src.sa.sa_family != AF_INET6) {
620				error = EADDRNOTAVAIL;
621				break;
622			}
623			sin6 = (struct sockaddr_in6 *)
624				&(((struct in6_ifreq *)data)->ifr_addr);
625			memset(sin6, 0, sizeof(*sin6));
626			sin6->sin6_family = AF_INET6;
627			sin6->sin6_len = sizeof(*sin6);
628			break;
629#endif
630		default:
631			error = EAFNOSUPPORT;
632		}
633		if (error == 0) {
634			switch (cmd) {
635#ifdef INET
636			case SIOCGIFPSRCADDR:
637				sin->sin_addr = saidx->src.sin.sin_addr;
638				break;
639			case SIOCGIFPDSTADDR:
640				sin->sin_addr = saidx->dst.sin.sin_addr;
641				break;
642#endif
643#ifdef INET6
644			case SIOCGIFPSRCADDR_IN6:
645				sin6->sin6_addr = saidx->src.sin6.sin6_addr;
646				break;
647			case SIOCGIFPDSTADDR_IN6:
648				sin6->sin6_addr = saidx->dst.sin6.sin6_addr;
649				break;
650#endif
651			}
652		}
653		IPSEC_RUNLOCK(sc);
654		if (error != 0)
655			break;
656		switch (cmd) {
657#ifdef INET
658		case SIOCGIFPSRCADDR:
659		case SIOCGIFPDSTADDR:
660			error = prison_if(curthread->td_ucred,
661			    (struct sockaddr *)sin);
662			if (error != 0)
663				memset(sin, 0, sizeof(*sin));
664			break;
665#endif
666#ifdef INET6
667		case SIOCGIFPSRCADDR_IN6:
668		case SIOCGIFPDSTADDR_IN6:
669			error = prison_if(curthread->td_ucred,
670			    (struct sockaddr *)sin6);
671			if (error == 0)
672				error = sa6_recoverscope(sin6);
673			if (error != 0)
674				memset(sin6, 0, sizeof(*sin6));
675#endif
676		}
677		break;
678	case SIOCGTUNFIB:
679		ifr->ifr_fib = sc->fibnum;
680		break;
681	case SIOCSTUNFIB:
682		if ((error = priv_check(curthread, PRIV_NET_SETIFFIB)) != 0)
683			break;
684		if (ifr->ifr_fib >= rt_numfibs)
685			error = EINVAL;
686		else
687			sc->fibnum = ifr->ifr_fib;
688		break;
689	case IPSECGREQID:
690		reqid = sc->reqid;
691		error = copyout(&reqid, ifr->ifr_data, sizeof(reqid));
692		break;
693	case IPSECSREQID:
694		if ((error = priv_check(curthread, PRIV_NET_SETIFCAP)) != 0)
695			break;
696		error = copyin(ifr->ifr_data, &reqid, sizeof(reqid));
697		if (error != 0)
698			break;
699		error = ipsec_set_reqid(ifp, reqid);
700		break;
701	default:
702		error = EINVAL;
703		break;
704	}
705bad:
706	sx_xunlock(&ipsec_ioctl_sx);
707	return (error);
708}
709
710/*
711 * Allocate new private security policies for tunneling interface.
712 * Each tunneling interface has following security policies for
713 * both AF:
714 *   0.0.0.0/0[any] 0.0.0.0/0[any] -P in \
715 *	ipsec esp/tunnel/RemoteIP-LocalIP/unique:reqid
716 *   0.0.0.0/0[any] 0.0.0.0/0[any] -P out \
717 *	ipsec esp/tunnel/LocalIP-RemoteIP/unique:reqid
718 */
719static int
720ipsec_newpolicies(struct ipsec_softc *sc, struct secpolicy *sp[IPSEC_SPCOUNT],
721    const struct sockaddr *src, const struct sockaddr *dst, uint32_t reqid)
722{
723	struct ipsecrequest *isr;
724	int i;
725
726	memset(sp, 0, sizeof(struct secpolicy *) * IPSEC_SPCOUNT);
727	for (i = 0; i < IPSEC_SPCOUNT; i++) {
728		if ((sp[i] = key_newsp()) == NULL)
729			goto fail;
730		if ((isr = ipsec_newisr()) == NULL)
731			goto fail;
732
733		sp[i]->policy = IPSEC_POLICY_IPSEC;
734		sp[i]->state = IPSEC_SPSTATE_DEAD;
735		sp[i]->req[sp[i]->tcount++] = isr;
736		sp[i]->created = time_second;
737		/* Use priority field to store if_index */
738		sp[i]->priority = sc->ifp->if_index;
739		isr->level = IPSEC_LEVEL_UNIQUE;
740		isr->saidx.proto = IPPROTO_ESP;
741		isr->saidx.mode = IPSEC_MODE_TUNNEL;
742		isr->saidx.reqid = reqid;
743		if (i % 2 == 0) {
744			sp[i]->spidx.dir = IPSEC_DIR_INBOUND;
745			bcopy(src, &isr->saidx.dst, src->sa_len);
746			bcopy(dst, &isr->saidx.src, dst->sa_len);
747		} else {
748			sp[i]->spidx.dir = IPSEC_DIR_OUTBOUND;
749			bcopy(src, &isr->saidx.src, src->sa_len);
750			bcopy(dst, &isr->saidx.dst, dst->sa_len);
751		}
752		sp[i]->spidx.ul_proto = IPSEC_ULPROTO_ANY;
753#ifdef INET
754		if (i < 2) {
755			sp[i]->spidx.src.sa.sa_family =
756			    sp[i]->spidx.dst.sa.sa_family = AF_INET;
757			sp[i]->spidx.src.sa.sa_len =
758			    sp[i]->spidx.dst.sa.sa_len =
759			    sizeof(struct sockaddr_in);
760			continue;
761		}
762#endif
763#ifdef INET6
764		sp[i]->spidx.src.sa.sa_family =
765		    sp[i]->spidx.dst.sa.sa_family = AF_INET6;
766		sp[i]->spidx.src.sa.sa_len =
767		    sp[i]->spidx.dst.sa.sa_len = sizeof(struct sockaddr_in6);
768#endif
769	}
770	return (0);
771fail:
772	for (i = 0; i < IPSEC_SPCOUNT; i++)
773		key_freesp(&sp[i]);
774	return (ENOMEM);
775}
776
777static int
778ipsec_check_reqid(uint32_t reqid)
779{
780	struct ipsec_softc *sc;
781
782	IPSEC_SC_RLOCK_ASSERT();
783	LIST_FOREACH(sc, &V_ipsec_sc_list, chain) {
784		if (sc->reqid == reqid)
785			return (EEXIST);
786	}
787	return (0);
788}
789
790/*
791 * We use key_newreqid() to automatically obtain unique reqid.
792 * Then we check that given id is unique, i.e. it is not used by
793 * another if_ipsec(4) interface. This macro limits the number of
794 * tries to get unique id.
795 */
796#define	IPSEC_REQID_TRYCNT	64
797static int
798ipsec_init_reqid(struct ipsec_softc *sc)
799{
800	uint32_t reqid;
801	int trycount;
802
803	IPSEC_SC_RLOCK_ASSERT();
804
805	if (sc->reqid != 0) /* already initialized */
806		return (0);
807
808	trycount = IPSEC_REQID_TRYCNT;
809	while (--trycount > 0) {
810		reqid = key_newreqid();
811		if (ipsec_check_reqid(reqid) == 0)
812			break;
813	}
814	if (trycount == 0)
815		return (EEXIST);
816	sc->reqid = reqid;
817	return (0);
818}
819
820/*
821 * Set or update reqid for given tunneling interface.
822 * When specified reqid is zero, generate new one.
823 * We are protected by ioctl_sx lock from concurrent id generation.
824 * Also softc would not disappear while we hold ioctl_sx lock.
825 */
826static int
827ipsec_set_reqid(struct ifnet *ifp, uint32_t reqid)
828{
829	IPSEC_SC_RLOCK_TRACKER;
830	struct ipsec_softc *sc;
831	struct secasindex *saidx;
832
833	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
834
835	sc = ifp->if_softc;
836	if (sc->reqid == reqid && reqid != 0)
837		return (0);
838
839	IPSEC_SC_RLOCK();
840	if (reqid != 0) {
841		/* Check that specified reqid doesn't exist */
842		if (ipsec_check_reqid(reqid) != 0) {
843			IPSEC_SC_RUNLOCK();
844			return (EEXIST);
845		}
846		sc->reqid = reqid;
847	} else {
848		/* Generate new reqid */
849		if (ipsec_init_reqid(sc) != 0) {
850			IPSEC_SC_RUNLOCK();
851			return (EEXIST);
852		}
853	}
854	IPSEC_SC_RUNLOCK();
855
856	/* Tunnel isn't fully configured, just return. */
857	if (sc->family == 0)
858		return (0);
859
860	saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND, sc->family);
861	KASSERT(saidx != NULL,
862	    ("saidx is NULL, but family is %d", sc->family));
863	return (ipsec_set_tunnel(sc, &saidx->src.sa, &saidx->dst.sa,
864	    sc->reqid));
865}
866
867/*
868 * Set tunnel endpoints addresses.
869 */
870static int
871ipsec_set_addresses(struct ifnet *ifp, struct sockaddr *src,
872    struct sockaddr *dst)
873{
874	IPSEC_SC_RLOCK_TRACKER;
875	struct ipsec_softc *sc, *tsc;
876	struct secasindex *saidx;
877
878	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
879
880	sc = ifp->if_softc;
881	if (sc->family != 0) {
882		saidx = ipsec_getsaidx(sc, IPSEC_DIR_OUTBOUND,
883		    src->sa_family);
884		if (saidx != NULL && saidx->reqid == sc->reqid &&
885		    key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
886		    key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0)
887			return (0); /* Nothing has been changed. */
888
889	}
890	/*
891	 * We cannot service IPsec tunnel when source address is
892	 * not our own.
893	 */
894#ifdef INET
895	if (src->sa_family == AF_INET &&
896	    in_localip(satosin(src)->sin_addr) == 0)
897		return (EADDRNOTAVAIL);
898#endif
899#ifdef INET6
900	/*
901	 * NOTE: IPv6 addresses are in kernel internal form with
902	 * embedded scope zone id.
903	 */
904	if (src->sa_family == AF_INET6 &&
905	    in6_localip(&satosin6(src)->sin6_addr) == 0)
906		return (EADDRNOTAVAIL);
907#endif
908	/* Check that given addresses aren't already configured */
909	IPSEC_SC_RLOCK();
910	LIST_FOREACH(tsc, &V_ipsec_sc_list, chain) {
911		if (tsc == sc || tsc->family != src->sa_family)
912			continue;
913		saidx = ipsec_getsaidx(tsc, IPSEC_DIR_OUTBOUND, tsc->family);
914		if (key_sockaddrcmp(&saidx->src.sa, src, 0) == 0 &&
915		    key_sockaddrcmp(&saidx->dst.sa, dst, 0) == 0) {
916			/* We already have tunnel with such addresses */
917			IPSEC_SC_RUNLOCK();
918			return (EADDRNOTAVAIL);
919		}
920	}
921	/* If reqid is not set, generate new one. */
922	if (ipsec_init_reqid(sc) != 0) {
923		IPSEC_SC_RUNLOCK();
924		return (EEXIST);
925	}
926	IPSEC_SC_RUNLOCK();
927	return (ipsec_set_tunnel(sc, src, dst, sc->reqid));
928}
929
930static int
931ipsec_set_tunnel(struct ipsec_softc *sc, struct sockaddr *src,
932    struct sockaddr *dst, uint32_t reqid)
933{
934	struct secpolicy *sp[IPSEC_SPCOUNT];
935	struct secpolicy *oldsp[IPSEC_SPCOUNT];
936	int i, f;
937
938	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
939
940	/* Allocate SP with new addresses. */
941	if (ipsec_newpolicies(sc, sp, src, dst, reqid) == 0) {
942		/* Add new policies to SPDB */
943		if (key_register_ifnet(sp, IPSEC_SPCOUNT) != 0) {
944			for (i = 0; i < IPSEC_SPCOUNT; i++)
945				key_freesp(&sp[i]);
946			return (EAGAIN);
947		}
948		IPSEC_SC_WLOCK();
949		if ((f = sc->family) != 0)
950			LIST_REMOVE(sc, hash);
951		IPSEC_WLOCK(sc);
952		for (i = 0; i < IPSEC_SPCOUNT; i++) {
953			oldsp[i] = sc->sp[i];
954			sc->sp[i] = sp[i];
955		}
956		sc->family = src->sa_family;
957		IPSEC_WUNLOCK(sc);
958		LIST_INSERT_HEAD(SCHASH_HASH(sc->reqid), sc, hash);
959		IPSEC_SC_WUNLOCK();
960	} else {
961		sc->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
962		return (ENOMEM);
963	}
964
965	sc->ifp->if_drv_flags |= IFF_DRV_RUNNING;
966	if (f != 0) {
967		key_unregister_ifnet(oldsp, IPSEC_SPCOUNT);
968		for (i = 0; i < IPSEC_SPCOUNT; i++)
969			key_freesp(&oldsp[i]);
970	}
971	return (0);
972}
973
974static void
975ipsec_delete_tunnel(struct ifnet *ifp, int locked)
976{
977	struct ipsec_softc *sc = ifp->if_softc;
978	struct secpolicy *oldsp[IPSEC_SPCOUNT];
979	int i;
980
981	sx_assert(&ipsec_ioctl_sx, SA_XLOCKED);
982
983	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
984	if (sc->family != 0) {
985		if (!locked)
986			IPSEC_SC_WLOCK();
987		/* Remove from hash table */
988		LIST_REMOVE(sc, hash);
989		IPSEC_WLOCK(sc);
990		for (i = 0; i < IPSEC_SPCOUNT; i++) {
991			oldsp[i] = sc->sp[i];
992			sc->sp[i] = NULL;
993		}
994		sc->family = 0;
995		IPSEC_WUNLOCK(sc);
996		if (!locked)
997			IPSEC_SC_WUNLOCK();
998		key_unregister_ifnet(oldsp, IPSEC_SPCOUNT);
999		for (i = 0; i < IPSEC_SPCOUNT; i++)
1000			key_freesp(&oldsp[i]);
1001	}
1002}
1003