1 /* $NetBSD: if_ipsec.c,v 1.29 2020/03/13 02:43:31 knakahara Exp $ */ 2 3 /* 4 * Copyright (c) 2017 Internet Initiative Japan Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 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 NETBSD FOUNDATION, INC. AND CONTRIBUTORS 17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 * POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __KERNEL_RCSID(0, "$NetBSD: if_ipsec.c,v 1.29 2020/03/13 02:43:31 knakahara Exp $"); 31 32 #ifdef _KERNEL_OPT 33 #include "opt_inet.h" 34 #endif 35 36 #include <sys/param.h> 37 #include <sys/atomic.h> 38 #include <sys/systm.h> 39 #include <sys/kernel.h> 40 #include <sys/mbuf.h> 41 #include <sys/socket.h> 42 #include <sys/sockio.h> 43 #include <sys/errno.h> 44 #include <sys/ioctl.h> 45 #include <sys/time.h> 46 #include <sys/syslog.h> 47 #include <sys/cpu.h> 48 #include <sys/kmem.h> 49 #include <sys/mutex.h> 50 #include <sys/pserialize.h> 51 #include <sys/psref.h> 52 #include <sys/sysctl.h> 53 54 #include <net/if.h> 55 #include <net/if_types.h> 56 #include <net/route.h> 57 #include <net/bpf.h> 58 #include <net/pfkeyv2.h> 59 60 #include <netinet/in.h> 61 #include <netinet/in_systm.h> 62 #include <netinet/ip.h> 63 #ifdef INET 64 #include <netinet/in_var.h> 65 #endif /* INET */ 66 67 #ifdef INET6 68 #include <netinet6/in6_var.h> 69 #include <netinet/ip6.h> 70 #include <netinet6/ip6_var.h> 71 #endif /* INET6 */ 72 73 #include <netinet/ip_encap.h> 74 75 #include <net/if_ipsec.h> 76 77 #include <net/raw_cb.h> 78 #include <net/pfkeyv2.h> 79 80 #include <netipsec/key.h> 81 #include <netipsec/keydb.h> /* for union sockaddr_union */ 82 #include <netipsec/ipsec.h> 83 #include <netipsec/ipsecif.h> 84 85 static int if_ipsec_clone_create(struct if_clone *, int); 86 static int if_ipsec_clone_destroy(struct ifnet *); 87 88 static inline int if_ipsec_out_direct(struct ipsec_variant *, struct mbuf *, int); 89 static inline void if_ipsec_in_enqueue(struct mbuf *, int, struct ifnet *); 90 91 static int if_ipsec_encap_attach(struct ipsec_variant *); 92 static int if_ipsec_encap_detach(struct ipsec_variant *); 93 static int if_ipsec_set_tunnel(struct ifnet *, 94 struct sockaddr *, struct sockaddr *); 95 static void if_ipsec_delete_tunnel(struct ifnet *); 96 static int if_ipsec_ensure_flags(struct ifnet *, u_short); 97 static void if_ipsec_attach0(struct ipsec_softc *); 98 99 static int if_ipsec_update_variant(struct ipsec_softc *, 100 struct ipsec_variant *, struct ipsec_variant *); 101 102 /* sadb_msg */ 103 static inline void if_ipsec_add_mbuf(struct mbuf *, void *, size_t); 104 static inline void if_ipsec_add_pad(struct mbuf *, size_t); 105 static inline size_t if_ipsec_set_sadb_addr(struct sadb_address *, 106 struct sockaddr *, int, uint16_t); 107 static inline size_t if_ipsec_set_sadb_src(struct sadb_address *, 108 struct sockaddr *, int); 109 static inline size_t if_ipsec_set_sadb_dst(struct sadb_address *, 110 struct sockaddr *, int); 111 static inline size_t if_ipsec_set_sadb_x_policy(struct sadb_x_policy *, 112 struct sadb_x_ipsecrequest *, uint16_t, uint8_t, uint32_t, uint8_t, 113 struct sockaddr *, struct sockaddr *); 114 static inline void if_ipsec_set_sadb_msg(struct sadb_msg *, uint16_t, uint8_t); 115 static inline void if_ipsec_set_sadb_msg_add(struct sadb_msg *, uint16_t); 116 static inline void if_ipsec_set_sadb_msg_del(struct sadb_msg *, uint16_t); 117 /* SPD */ 118 static int if_ipsec_share_sp(struct ipsec_variant *); 119 static int if_ipsec_unshare_sp(struct ipsec_variant *); 120 static inline struct secpolicy *if_ipsec_add_sp0(struct sockaddr *, 121 in_port_t, struct sockaddr *, in_port_t, int, int, int, u_int); 122 static inline int if_ipsec_del_sp0(struct secpolicy *); 123 static int if_ipsec_add_sp(struct ipsec_variant *, 124 struct sockaddr *, in_port_t, struct sockaddr *, in_port_t); 125 static void if_ipsec_del_sp(struct ipsec_variant *); 126 static int if_ipsec_replace_sp(struct ipsec_softc *, struct ipsec_variant *, 127 struct ipsec_variant *); 128 129 static int if_ipsec_set_addr_port(struct sockaddr *, struct sockaddr *, 130 in_port_t); 131 #define IF_IPSEC_GATHER_PSRC_ADDR_PORT(var, target) \ 132 if_ipsec_set_addr_port(target, (var)->iv_psrc, (var)->iv_sport) 133 #define IF_IPSEC_GATHER_PDST_ADDR_PORT(var, target) \ 134 if_ipsec_set_addr_port(target, (var)->iv_pdst, (var)->iv_dport) 135 136 /* 137 * ipsec global variable definitions 138 */ 139 140 /* This list is used in ioctl context only. */ 141 static struct { 142 LIST_HEAD(ipsec_sclist, ipsec_softc) list; 143 kmutex_t lock; 144 } ipsec_softcs __cacheline_aligned; 145 146 struct psref_class *iv_psref_class __read_mostly; 147 148 struct if_clone ipsec_cloner = 149 IF_CLONE_INITIALIZER("ipsec", if_ipsec_clone_create, if_ipsec_clone_destroy); 150 static int max_ipsec_nesting = MAX_IPSEC_NEST; 151 152 static struct sysctllog *if_ipsec_sysctl; 153 154 #ifdef INET6 155 static int 156 sysctl_if_ipsec_pmtu_global(SYSCTLFN_ARGS) 157 { 158 int error, pmtu; 159 struct sysctlnode node = *rnode; 160 161 pmtu = ip6_ipsec_pmtu; 162 node.sysctl_data = &pmtu; 163 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 164 if (error || newp == NULL) 165 return error; 166 167 switch (pmtu) { 168 case IPSEC_PMTU_MINMTU: 169 case IPSEC_PMTU_OUTERMTU: 170 ip6_ipsec_pmtu = pmtu; 171 break; 172 default: 173 return EINVAL; 174 } 175 176 return 0; 177 } 178 179 static int 180 sysctl_if_ipsec_pmtu_perif(SYSCTLFN_ARGS) 181 { 182 int error, pmtu; 183 struct sysctlnode node = *rnode; 184 struct ipsec_softc *sc = (struct ipsec_softc *)node.sysctl_data; 185 186 pmtu = sc->ipsec_pmtu; 187 node.sysctl_data = &pmtu; 188 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 189 if (error || newp == NULL) 190 return error; 191 192 switch (pmtu) { 193 case IPSEC_PMTU_SYSDEFAULT: 194 case IPSEC_PMTU_MINMTU: 195 case IPSEC_PMTU_OUTERMTU: 196 sc->ipsec_pmtu = pmtu; 197 break; 198 default: 199 return EINVAL; 200 } 201 202 return 0; 203 } 204 #endif 205 206 static void 207 if_ipsec_sysctl_setup(void) 208 { 209 if_ipsec_sysctl = NULL; 210 211 #ifdef INET6 212 /* 213 * Previously create "net.inet6.ip6" entry to avoid sysctl_createv error. 214 */ 215 sysctl_createv(NULL, 0, NULL, NULL, 216 CTLFLAG_PERMANENT, 217 CTLTYPE_NODE, "inet6", 218 SYSCTL_DESCR("PF_INET6 related settings"), 219 NULL, 0, NULL, 0, 220 CTL_NET, PF_INET6, CTL_EOL); 221 sysctl_createv(NULL, 0, NULL, NULL, 222 CTLFLAG_PERMANENT, 223 CTLTYPE_NODE, "ip6", 224 SYSCTL_DESCR("IPv6 related settings"), 225 NULL, 0, NULL, 0, 226 CTL_NET, PF_INET6, IPPROTO_IPV6, CTL_EOL); 227 228 sysctl_createv(&if_ipsec_sysctl, 0, NULL, NULL, 229 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 230 CTLTYPE_INT, "ipsecifhlim", 231 SYSCTL_DESCR("Default hop limit for a ipsec tunnel datagram"), 232 NULL, 0, &ip6_ipsec_hlim, 0, 233 CTL_NET, PF_INET6, IPPROTO_IPV6, 234 IPV6CTL_IPSEC_HLIM, CTL_EOL); 235 236 sysctl_createv(&if_ipsec_sysctl, 0, NULL, NULL, 237 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 238 CTLTYPE_INT, "ipsecifpmtu", 239 SYSCTL_DESCR("Default Path MTU setting for ipsec tunnels"), 240 sysctl_if_ipsec_pmtu_global, 0, NULL, 0, 241 CTL_NET, PF_INET6, IPPROTO_IPV6, 242 IPV6CTL_IPSEC_PMTU, CTL_EOL); 243 #endif 244 } 245 246 static void 247 if_ipsec_perif_sysctl_setup(struct sysctllog **clog, struct ipsec_softc *sc) 248 { 249 #ifdef INET6 250 const struct sysctlnode *cnode, *rnode; 251 struct ifnet *ifp = &sc->ipsec_if; 252 const char *ifname = ifp->if_xname; 253 int rv; 254 255 /* 256 * Already created in sysctl_sndq_setup(). 257 */ 258 sysctl_createv(clog, 0, NULL, &rnode, 259 CTLFLAG_PERMANENT, 260 CTLTYPE_NODE, "interfaces", 261 SYSCTL_DESCR("Per-interface controls"), 262 NULL, 0, NULL, 0, 263 CTL_NET, CTL_CREATE, CTL_EOL); 264 sysctl_createv(clog, 0, &rnode, &rnode, 265 CTLFLAG_PERMANENT, 266 CTLTYPE_NODE, ifname, 267 SYSCTL_DESCR("Interface controls"), 268 NULL, 0, NULL, 0, 269 CTL_CREATE, CTL_EOL); 270 271 rv = sysctl_createv(clog, 0, &rnode, &cnode, 272 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 273 CTLTYPE_INT, "pmtu", 274 SYSCTL_DESCR("Path MTU setting for this ipsec tunnel"), 275 sysctl_if_ipsec_pmtu_perif, 0, (void *)sc, 0, 276 CTL_CREATE, CTL_EOL); 277 if (rv != 0) 278 log(LOG_WARNING, "%s: could not attach sysctl node pmtu\n", ifname); 279 280 sc->ipsec_pmtu = IPSEC_PMTU_SYSDEFAULT; 281 #endif 282 } 283 284 /* ARGSUSED */ 285 void 286 ipsecifattach(int count) 287 { 288 289 mutex_init(&ipsec_softcs.lock, MUTEX_DEFAULT, IPL_NONE); 290 LIST_INIT(&ipsec_softcs.list); 291 292 iv_psref_class = psref_class_create("ipsecvar", IPL_SOFTNET); 293 294 if_ipsec_sysctl_setup(); 295 296 if_clone_attach(&ipsec_cloner); 297 } 298 299 static int 300 if_ipsec_clone_create(struct if_clone *ifc, int unit) 301 { 302 struct ipsec_softc *sc; 303 struct ipsec_variant *var; 304 struct ifnet *ifp; 305 306 sc = kmem_zalloc(sizeof(*sc), KM_SLEEP); 307 308 if_initname(&sc->ipsec_if, ifc->ifc_name, unit); 309 310 if_ipsec_attach0(sc); 311 312 ifp = &sc->ipsec_if; 313 if_ipsec_perif_sysctl_setup(&ifp->if_sysctl_log, sc); 314 315 var = kmem_zalloc(sizeof(*var), KM_SLEEP); 316 var->iv_softc = sc; 317 psref_target_init(&var->iv_psref, iv_psref_class); 318 319 sc->ipsec_var = var; 320 mutex_init(&sc->ipsec_lock, MUTEX_DEFAULT, IPL_NONE); 321 sc->ipsec_psz = pserialize_create(); 322 sc->ipsec_ro_percpu = if_tunnel_alloc_ro_percpu(); 323 324 mutex_enter(&ipsec_softcs.lock); 325 LIST_INSERT_HEAD(&ipsec_softcs.list, sc, ipsec_list); 326 mutex_exit(&ipsec_softcs.lock); 327 return 0; 328 } 329 330 static void 331 if_ipsec_attach0(struct ipsec_softc *sc) 332 { 333 334 sc->ipsec_if.if_addrlen = 0; 335 sc->ipsec_if.if_mtu = IPSEC_MTU; 336 sc->ipsec_if.if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 337 /* set ipsec(4) specific default flags. */ 338 sc->ipsec_if.if_flags |= IFF_FWD_IPV6; 339 sc->ipsec_if.if_extflags = IFEF_NO_LINK_STATE_CHANGE | IFEF_MPSAFE; 340 sc->ipsec_if.if_ioctl = if_ipsec_ioctl; 341 sc->ipsec_if.if_output = if_ipsec_output; 342 sc->ipsec_if.if_type = IFT_IPSEC; 343 sc->ipsec_if.if_dlt = DLT_NULL; 344 sc->ipsec_if.if_softc = sc; 345 IFQ_SET_READY(&sc->ipsec_if.if_snd); 346 if_initialize(&sc->ipsec_if); 347 if_alloc_sadl(&sc->ipsec_if); 348 bpf_attach(&sc->ipsec_if, DLT_NULL, sizeof(u_int)); 349 if_register(&sc->ipsec_if); 350 } 351 352 static int 353 if_ipsec_clone_destroy(struct ifnet *ifp) 354 { 355 struct ipsec_softc *sc = ifp->if_softc; 356 struct ipsec_variant *var; 357 int bound; 358 359 mutex_enter(&ipsec_softcs.lock); 360 LIST_REMOVE(sc, ipsec_list); 361 mutex_exit(&ipsec_softcs.lock); 362 363 bound = curlwp_bind(); 364 if_ipsec_delete_tunnel(&sc->ipsec_if); 365 curlwp_bindx(bound); 366 367 bpf_detach(ifp); 368 if_detach(ifp); 369 370 if_tunnel_free_ro_percpu(sc->ipsec_ro_percpu); 371 372 pserialize_destroy(sc->ipsec_psz); 373 mutex_destroy(&sc->ipsec_lock); 374 375 var = sc->ipsec_var; 376 kmem_free(var, sizeof(*var)); 377 kmem_free(sc, sizeof(*sc)); 378 379 return 0; 380 } 381 382 static inline bool 383 if_ipsec_nat_t(struct ipsec_softc *sc) 384 { 385 386 return (sc->ipsec_if.if_flags & IFF_NAT_T) != 0; 387 } 388 389 static inline bool 390 if_ipsec_fwd_ipv6(struct ipsec_softc *sc) 391 { 392 393 return (sc->ipsec_if.if_flags & IFF_FWD_IPV6) != 0; 394 } 395 396 int 397 if_ipsec_encap_func(struct mbuf *m, int off, int proto, void *arg) 398 { 399 uint8_t v; 400 struct ipsec_softc *sc; 401 struct ipsec_variant *var = NULL; 402 struct psref psref; 403 int ret = 0; 404 405 sc = arg; 406 KASSERT(sc != NULL); 407 408 if ((sc->ipsec_if.if_flags & IFF_UP) == 0) 409 goto out; 410 411 var = if_ipsec_getref_variant(sc, &psref); 412 if (if_ipsec_variant_is_unconfigured(var)) 413 goto out; 414 415 switch (proto) { 416 case IPPROTO_IPV4: 417 case IPPROTO_IPV6: 418 break; 419 default: 420 goto out; 421 } 422 423 m_copydata(m, 0, sizeof(v), &v); 424 v = (v >> 4) & 0xff; /* Get the IP version number. */ 425 426 switch (v) { 427 #ifdef INET 428 case IPVERSION: { 429 struct ip ip; 430 431 if (m->m_pkthdr.len < sizeof(ip)) 432 goto out; 433 434 m_copydata(m, 0, sizeof(ip), &ip); 435 if (var->iv_psrc->sa_family != AF_INET || 436 var->iv_pdst->sa_family != AF_INET) 437 goto out; 438 ret = ipsecif4_encap_func(m, &ip, var); 439 break; 440 } 441 #endif 442 #ifdef INET6 443 case (IPV6_VERSION >> 4): { 444 struct ip6_hdr ip6; 445 446 if (m->m_pkthdr.len < sizeof(ip6)) 447 goto out; 448 449 m_copydata(m, 0, sizeof(ip6), &ip6); 450 if (var->iv_psrc->sa_family != AF_INET6 || 451 var->iv_pdst->sa_family != AF_INET6) 452 goto out; 453 ret = ipsecif6_encap_func(m, &ip6, var); 454 break; 455 } 456 #endif 457 default: 458 goto out; 459 } 460 461 out: 462 if (var != NULL) 463 if_ipsec_putref_variant(var, &psref); 464 return ret; 465 } 466 467 /* 468 * ipsec(4) I/F may cause infinite recursion calls when misconfigured. 469 * We'll prevent this by introducing upper limit. 470 */ 471 static int 472 if_ipsec_check_nesting(struct ifnet *ifp, struct mbuf *m) 473 { 474 475 return if_tunnel_check_nesting(ifp, m, max_ipsec_nesting); 476 } 477 478 int 479 if_ipsec_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, 480 const struct rtentry *rt) 481 { 482 struct ipsec_softc *sc = ifp->if_softc; 483 struct ipsec_variant *var; 484 struct psref psref; 485 int error; 486 int bound; 487 488 IFQ_CLASSIFY(&ifp->if_snd, m, dst->sa_family); 489 490 error = if_ipsec_check_nesting(ifp, m); 491 if (error) { 492 m_freem(m); 493 goto noref_end; 494 } 495 496 if ((ifp->if_flags & IFF_UP) == 0) { 497 m_freem(m); 498 error = ENETDOWN; 499 goto noref_end; 500 } 501 502 503 bound = curlwp_bind(); 504 var = if_ipsec_getref_variant(sc, &psref); 505 if (if_ipsec_variant_is_unconfigured(var)) { 506 m_freem(m); 507 error = ENETDOWN; 508 goto end; 509 } 510 511 m->m_flags &= ~(M_BCAST|M_MCAST); 512 513 /* use DLT_NULL encapsulation here to pass inner af type */ 514 M_PREPEND(m, sizeof(int), M_DONTWAIT); 515 if (!m) { 516 error = ENOBUFS; 517 goto end; 518 } 519 *mtod(m, int *) = dst->sa_family; 520 521 #if INET6 522 /* drop IPv6 packet if IFF_FWD_IPV6 is not set */ 523 if (dst->sa_family == AF_INET6 && 524 !if_ipsec_fwd_ipv6(sc)) { 525 /* 526 * IPv6 packet is not allowed to forward,that is not error. 527 */ 528 error = 0; 529 IF_DROP(&ifp->if_snd); 530 m_freem(m); 531 goto end; 532 } 533 #endif 534 535 error = if_ipsec_out_direct(var, m, dst->sa_family); 536 537 end: 538 if_ipsec_putref_variant(var, &psref); 539 curlwp_bindx(bound); 540 noref_end: 541 if (error) 542 if_statinc(ifp, if_oerrors); 543 544 return error; 545 } 546 547 static inline int 548 if_ipsec_out_direct(struct ipsec_variant *var, struct mbuf *m, int family) 549 { 550 struct ifnet *ifp = &var->iv_softc->ipsec_if; 551 int error; 552 int len; 553 554 KASSERT(if_ipsec_heldref_variant(var)); 555 KASSERT(var->iv_output != NULL); 556 557 len = m->m_pkthdr.len; 558 559 /* input DLT_NULL frame to BPF */ 560 bpf_mtap(ifp, m, BPF_D_OUT); 561 562 /* grab and chop off inner af type */ 563 /* XXX need pullup? */ 564 m_adj(m, sizeof(int)); 565 566 error = var->iv_output(var, family, m); 567 if (error) 568 return error; 569 570 if_statadd2(ifp, if_opackets, 1, if_obytes, len); 571 572 return 0; 573 } 574 575 void 576 if_ipsec_input(struct mbuf *m, int af, struct ifnet *ifp) 577 { 578 579 KASSERT(ifp != NULL); 580 581 m_set_rcvif(m, ifp); 582 583 bpf_mtap_af(ifp, af, m, BPF_D_IN); 584 585 if_ipsec_in_enqueue(m, af, ifp); 586 587 return; 588 } 589 590 static inline void 591 if_ipsec_in_enqueue(struct mbuf *m, int af, struct ifnet *ifp) 592 { 593 pktqueue_t *pktq; 594 int pktlen; 595 596 /* 597 * Put the packet to the network layer input queue according to the 598 * specified address family. 599 */ 600 switch (af) { 601 #ifdef INET 602 case AF_INET: 603 pktq = ip_pktq; 604 break; 605 #endif 606 #ifdef INET6 607 case AF_INET6: 608 pktq = ip6_pktq; 609 break; 610 #endif 611 default: 612 if_statinc(ifp, if_ierrors); 613 m_freem(m); 614 return; 615 } 616 617 #if 1 618 const u_int h = curcpu()->ci_index; 619 #else 620 const uint32_t h = pktq_rps_hash(m); 621 #endif 622 pktlen = m->m_pkthdr.len; 623 if (__predict_true(pktq_enqueue(pktq, m, h))) { 624 if_statadd2(ifp, if_ibytes, pktlen, if_ipackets, 1); 625 } else { 626 if_statinc(ifp, if_iqdrops); 627 m_freem(m); 628 } 629 630 return; 631 } 632 633 static inline int 634 if_ipsec_check_salen(struct sockaddr *addr) 635 { 636 637 switch (addr->sa_family) { 638 #ifdef INET 639 case AF_INET: 640 if (addr->sa_len != sizeof(struct sockaddr_in)) 641 return EINVAL; 642 break; 643 #endif /* INET */ 644 #ifdef INET6 645 case AF_INET6: 646 if (addr->sa_len != sizeof(struct sockaddr_in6)) 647 return EINVAL; 648 break; 649 #endif /* INET6 */ 650 default: 651 return EAFNOSUPPORT; 652 } 653 654 return 0; 655 } 656 657 /* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */ 658 int 659 if_ipsec_ioctl(struct ifnet *ifp, u_long cmd, void *data) 660 { 661 struct ipsec_softc *sc = ifp->if_softc; 662 struct ipsec_variant *var = NULL; 663 struct ifreq *ifr = (struct ifreq*)data; 664 struct ifaddr *ifa = (struct ifaddr*)data; 665 int error = 0, size; 666 struct sockaddr *dst, *src; 667 u_long mtu; 668 u_short oflags = ifp->if_flags; 669 int bound; 670 struct psref psref; 671 672 switch (cmd) { 673 case SIOCINITIFADDR: 674 ifp->if_flags |= IFF_UP; 675 ifa->ifa_rtrequest = p2p_rtrequest; 676 break; 677 678 case SIOCSIFDSTADDR: 679 break; 680 681 case SIOCADDMULTI: 682 case SIOCDELMULTI: 683 switch (ifr->ifr_addr.sa_family) { 684 #ifdef INET 685 case AF_INET: /* IP supports Multicast */ 686 break; 687 #endif /* INET */ 688 #ifdef INET6 689 case AF_INET6: /* IP6 supports Multicast */ 690 break; 691 #endif /* INET6 */ 692 default: /* Other protocols doesn't support Multicast */ 693 error = EAFNOSUPPORT; 694 break; 695 } 696 break; 697 698 case SIOCSIFMTU: 699 mtu = ifr->ifr_mtu; 700 if (mtu < IPSEC_MTU_MIN || mtu > IPSEC_MTU_MAX) 701 return EINVAL; 702 else if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 703 error = 0; 704 break; 705 706 #ifdef INET 707 case SIOCSIFPHYADDR: 708 #endif 709 #ifdef INET6 710 case SIOCSIFPHYADDR_IN6: 711 #endif /* INET6 */ 712 case SIOCSLIFPHYADDR: 713 switch (cmd) { 714 #ifdef INET 715 case SIOCSIFPHYADDR: 716 src = (struct sockaddr *) 717 &(((struct in_aliasreq *)data)->ifra_addr); 718 dst = (struct sockaddr *) 719 &(((struct in_aliasreq *)data)->ifra_dstaddr); 720 break; 721 #endif /* INET */ 722 #ifdef INET6 723 case SIOCSIFPHYADDR_IN6: 724 src = (struct sockaddr *) 725 &(((struct in6_aliasreq *)data)->ifra_addr); 726 dst = (struct sockaddr *) 727 &(((struct in6_aliasreq *)data)->ifra_dstaddr); 728 break; 729 #endif /* INET6 */ 730 case SIOCSLIFPHYADDR: 731 src = (struct sockaddr *) 732 &(((struct if_laddrreq *)data)->addr); 733 dst = (struct sockaddr *) 734 &(((struct if_laddrreq *)data)->dstaddr); 735 break; 736 default: 737 return EINVAL; 738 } 739 740 /* sa_family must be equal */ 741 if (src->sa_family != dst->sa_family) 742 return EINVAL; 743 744 error = if_ipsec_check_salen(src); 745 if (error) 746 return error; 747 error = if_ipsec_check_salen(dst); 748 if (error) 749 return error; 750 751 /* check sa_family looks sane for the cmd */ 752 switch (cmd) { 753 #ifdef INET 754 case SIOCSIFPHYADDR: 755 if (src->sa_family == AF_INET) 756 break; 757 return EAFNOSUPPORT; 758 #endif /* INET */ 759 #ifdef INET6 760 case SIOCSIFPHYADDR_IN6: 761 if (src->sa_family == AF_INET6) 762 break; 763 return EAFNOSUPPORT; 764 #endif /* INET6 */ 765 case SIOCSLIFPHYADDR: 766 /* checks done in the above */ 767 break; 768 } 769 /* 770 * calls if_ipsec_getref_variant() for other softcs to check 771 * address pair duplicattion 772 */ 773 bound = curlwp_bind(); 774 error = if_ipsec_set_tunnel(&sc->ipsec_if, src, dst); 775 if (error) 776 goto bad; 777 curlwp_bindx(bound); 778 break; 779 780 case SIOCDIFPHYADDR: 781 bound = curlwp_bind(); 782 if_ipsec_delete_tunnel(&sc->ipsec_if); 783 curlwp_bindx(bound); 784 break; 785 786 case SIOCGIFPSRCADDR: 787 #ifdef INET6 788 case SIOCGIFPSRCADDR_IN6: 789 #endif /* INET6 */ 790 bound = curlwp_bind(); 791 var = if_ipsec_getref_variant(sc, &psref); 792 if (var->iv_psrc == NULL) { 793 error = EADDRNOTAVAIL; 794 goto bad; 795 } 796 src = var->iv_psrc; 797 switch (cmd) { 798 #ifdef INET 799 case SIOCGIFPSRCADDR: 800 dst = &ifr->ifr_addr; 801 size = sizeof(ifr->ifr_addr); 802 break; 803 #endif /* INET */ 804 #ifdef INET6 805 case SIOCGIFPSRCADDR_IN6: 806 dst = (struct sockaddr *) 807 &(((struct in6_ifreq *)data)->ifr_addr); 808 size = sizeof(((struct in6_ifreq *)data)->ifr_addr); 809 break; 810 #endif /* INET6 */ 811 default: 812 error = EADDRNOTAVAIL; 813 goto bad; 814 } 815 if (src->sa_len > size) { 816 error = EINVAL; 817 goto bad; 818 } 819 error = IF_IPSEC_GATHER_PSRC_ADDR_PORT(var, dst); 820 if (error) 821 goto bad; 822 if_ipsec_putref_variant(var, &psref); 823 curlwp_bindx(bound); 824 break; 825 826 case SIOCGIFPDSTADDR: 827 #ifdef INET6 828 case SIOCGIFPDSTADDR_IN6: 829 #endif /* INET6 */ 830 bound = curlwp_bind(); 831 var = if_ipsec_getref_variant(sc, &psref); 832 if (var->iv_pdst == NULL) { 833 error = EADDRNOTAVAIL; 834 goto bad; 835 } 836 src = var->iv_pdst; 837 switch (cmd) { 838 #ifdef INET 839 case SIOCGIFPDSTADDR: 840 dst = &ifr->ifr_addr; 841 size = sizeof(ifr->ifr_addr); 842 break; 843 #endif /* INET */ 844 #ifdef INET6 845 case SIOCGIFPDSTADDR_IN6: 846 dst = (struct sockaddr *) 847 &(((struct in6_ifreq *)data)->ifr_addr); 848 size = sizeof(((struct in6_ifreq *)data)->ifr_addr); 849 break; 850 #endif /* INET6 */ 851 default: 852 error = EADDRNOTAVAIL; 853 goto bad; 854 } 855 if (src->sa_len > size) { 856 error = EINVAL; 857 goto bad; 858 } 859 error = IF_IPSEC_GATHER_PDST_ADDR_PORT(var, dst); 860 if (error) 861 goto bad; 862 if_ipsec_putref_variant(var, &psref); 863 curlwp_bindx(bound); 864 break; 865 866 case SIOCGLIFPHYADDR: 867 bound = curlwp_bind(); 868 var = if_ipsec_getref_variant(sc, &psref); 869 if (if_ipsec_variant_is_unconfigured(var)) { 870 error = EADDRNOTAVAIL; 871 goto bad; 872 } 873 874 /* copy src */ 875 src = var->iv_psrc; 876 dst = (struct sockaddr *) 877 &(((struct if_laddrreq *)data)->addr); 878 size = sizeof(((struct if_laddrreq *)data)->addr); 879 if (src->sa_len > size) { 880 error = EINVAL; 881 goto bad; 882 } 883 error = IF_IPSEC_GATHER_PSRC_ADDR_PORT(var, dst); 884 if (error) 885 goto bad; 886 887 /* copy dst */ 888 src = var->iv_pdst; 889 dst = (struct sockaddr *) 890 &(((struct if_laddrreq *)data)->dstaddr); 891 size = sizeof(((struct if_laddrreq *)data)->dstaddr); 892 if (src->sa_len > size) { 893 error = EINVAL; 894 goto bad; 895 } 896 error = IF_IPSEC_GATHER_PDST_ADDR_PORT(var, dst); 897 if (error) 898 goto bad; 899 if_ipsec_putref_variant(var, &psref); 900 curlwp_bindx(bound); 901 break; 902 903 default: 904 error = ifioctl_common(ifp, cmd, data); 905 if (!error) { 906 bound = curlwp_bind(); 907 error = if_ipsec_ensure_flags(&sc->ipsec_if, oflags); 908 if (error) 909 goto bad; 910 curlwp_bindx(bound); 911 } 912 break; 913 } 914 return error; 915 916 bad: 917 if (var != NULL) 918 if_ipsec_putref_variant(var, &psref); 919 curlwp_bindx(bound); 920 921 return error; 922 } 923 924 struct encap_funcs { 925 #ifdef INET 926 int (*ef_inet)(struct ipsec_variant *); 927 #endif 928 #ifdef INET6 929 int (*ef_inet6)(struct ipsec_variant *); 930 #endif 931 }; 932 933 static struct encap_funcs ipsec_encap_attach = { 934 #ifdef INET 935 .ef_inet = ipsecif4_attach, 936 #endif 937 #ifdef INET6 938 .ef_inet6 = &ipsecif6_attach, 939 #endif 940 }; 941 942 static struct encap_funcs ipsec_encap_detach = { 943 #ifdef INET 944 .ef_inet = ipsecif4_detach, 945 #endif 946 #ifdef INET6 947 .ef_inet6 = &ipsecif6_detach, 948 #endif 949 }; 950 951 static int 952 if_ipsec_encap_common(struct ipsec_variant *var, struct encap_funcs *funcs) 953 { 954 int error; 955 956 KASSERT(var != NULL); 957 KASSERT(if_ipsec_variant_is_configured(var)); 958 959 switch (var->iv_psrc->sa_family) { 960 #ifdef INET 961 case AF_INET: 962 error = (funcs->ef_inet)(var); 963 break; 964 #endif /* INET */ 965 #ifdef INET6 966 case AF_INET6: 967 error = (funcs->ef_inet6)(var); 968 break; 969 #endif /* INET6 */ 970 default: 971 error = EINVAL; 972 break; 973 } 974 975 return error; 976 } 977 978 static int 979 if_ipsec_encap_attach(struct ipsec_variant *var) 980 { 981 982 return if_ipsec_encap_common(var, &ipsec_encap_attach); 983 } 984 985 static int 986 if_ipsec_encap_detach(struct ipsec_variant *var) 987 { 988 989 return if_ipsec_encap_common(var, &ipsec_encap_detach); 990 } 991 992 /* 993 * Validate and set ipsec(4) I/F configurations. 994 * (1) validate 995 * (1-1) Check the argument src and dst address pair will change 996 * configuration from current src and dst address pair. 997 * (1-2) Check any ipsec(4) I/F uses duplicated src and dst address pair 998 * with argument src and dst address pair, except for NAT-T shared 999 * tunnels. 1000 * (2) set 1001 * (2-1) Create variant for new configuration. 1002 * (2-2) Create temporary "null" variant used to avoid to access 1003 * dangling variant while SPs are deleted and added. 1004 * (2-3) Swap variant include its SPs. 1005 * (2-4) Cleanup last configurations. 1006 */ 1007 static int 1008 if_ipsec_set_tunnel(struct ifnet *ifp, 1009 struct sockaddr *src, struct sockaddr *dst) 1010 { 1011 struct ipsec_softc *sc = ifp->if_softc; 1012 struct ipsec_softc *sc2; 1013 struct ipsec_variant *ovar, *nvar, *nullvar; 1014 struct sockaddr *osrc, *odst; 1015 struct sockaddr *nsrc, *ndst; 1016 in_port_t nsport = 0, ndport = 0; 1017 int error; 1018 1019 error = encap_lock_enter(); 1020 if (error) 1021 return error; 1022 1023 nsrc = sockaddr_dup(src, M_WAITOK); 1024 ndst = sockaddr_dup(dst, M_WAITOK); 1025 nvar = kmem_zalloc(sizeof(*nvar), KM_SLEEP); 1026 nullvar = kmem_zalloc(sizeof(*nullvar), KM_SLEEP); 1027 1028 mutex_enter(&sc->ipsec_lock); 1029 1030 ovar = sc->ipsec_var; 1031 1032 switch(nsrc->sa_family) { 1033 #ifdef INET 1034 case AF_INET: 1035 nsport = satosin(src)->sin_port; 1036 /* 1037 * avoid confuse SP when NAT-T disabled, 1038 * e.g. 1039 * expected: 10.0.1.2[any] 10.0.1.1[any] 4(ipv4) 1040 * confuse : 10.0.1.2[600] 10.0.1.1[600] 4(ipv4) 1041 */ 1042 satosin(nsrc)->sin_port = 0; 1043 ndport = satosin(dst)->sin_port; 1044 satosin(ndst)->sin_port = 0; 1045 break; 1046 #endif /* INET */ 1047 #ifdef INET6 1048 case AF_INET6: 1049 nsport = satosin6(src)->sin6_port; 1050 satosin6(nsrc)->sin6_port = 0; 1051 ndport = satosin6(dst)->sin6_port; 1052 satosin6(ndst)->sin6_port = 0; 1053 break; 1054 #endif /* INET6 */ 1055 default: 1056 log(LOG_DEBUG, 1057 "%s: Invalid address family: %d.\n", 1058 __func__, src->sa_family); 1059 error = EINVAL; 1060 goto out; 1061 } 1062 1063 /* 1064 * (1-1) Check the argument src and dst address pair will change 1065 * configuration from current src and dst address pair. 1066 */ 1067 if ((ovar->iv_pdst && sockaddr_cmp(ovar->iv_pdst, dst) == 0) && 1068 (ovar->iv_psrc && sockaddr_cmp(ovar->iv_psrc, src) == 0) && 1069 (ovar->iv_sport == nsport && ovar->iv_dport == ndport)) { 1070 /* address and port pair not changed. */ 1071 error = 0; 1072 goto out; 1073 } 1074 1075 /* 1076 * (1-2) Check any ipsec(4) I/F uses duplicated src and dst address pair 1077 * with argument src and dst address pair, except for NAT-T shared 1078 * tunnels. 1079 */ 1080 mutex_enter(&ipsec_softcs.lock); 1081 LIST_FOREACH(sc2, &ipsec_softcs.list, ipsec_list) { 1082 struct ipsec_variant *var2; 1083 struct psref psref; 1084 1085 if (sc2 == sc) 1086 continue; 1087 var2 = if_ipsec_getref_variant(sc2, &psref); 1088 if (if_ipsec_variant_is_unconfigured(var2)) { 1089 if_ipsec_putref_variant(var2, &psref); 1090 continue; 1091 } 1092 if (if_ipsec_nat_t(sc) || if_ipsec_nat_t(sc2)) { 1093 if_ipsec_putref_variant(var2, &psref); 1094 continue; /* NAT-T shared tunnel */ 1095 } 1096 if (sockaddr_cmp(var2->iv_pdst, dst) == 0 && 1097 sockaddr_cmp(var2->iv_psrc, src) == 0) { 1098 if_ipsec_putref_variant(var2, &psref); 1099 mutex_exit(&ipsec_softcs.lock); 1100 error = EADDRNOTAVAIL; 1101 goto out; 1102 } 1103 1104 if_ipsec_putref_variant(var2, &psref); 1105 /* XXX both end must be valid? (I mean, not 0.0.0.0) */ 1106 } 1107 mutex_exit(&ipsec_softcs.lock); 1108 1109 1110 osrc = ovar->iv_psrc; 1111 odst = ovar->iv_pdst; 1112 1113 /* 1114 * (2-1) Create ipsec_variant for new configuration. 1115 */ 1116 if_ipsec_copy_variant(nvar, ovar); 1117 nvar->iv_psrc = nsrc; 1118 nvar->iv_pdst = ndst; 1119 nvar->iv_sport = nsport; 1120 nvar->iv_dport = ndport; 1121 nvar->iv_encap_cookie4 = NULL; 1122 nvar->iv_encap_cookie6 = NULL; 1123 psref_target_init(&nvar->iv_psref, iv_psref_class); 1124 error = if_ipsec_encap_attach(nvar); 1125 if (error) 1126 goto out; 1127 1128 /* 1129 * (2-2) Create temporary "null" variant. 1130 */ 1131 if_ipsec_copy_variant(nullvar, ovar); 1132 if_ipsec_clear_config(nullvar); 1133 psref_target_init(&nullvar->iv_psref, iv_psref_class); 1134 /* 1135 * (2-3) Swap variant include its SPs. 1136 */ 1137 error = if_ipsec_update_variant(sc, nvar, nullvar); 1138 if (error) { 1139 if_ipsec_encap_detach(nvar); 1140 goto out; 1141 } 1142 1143 mutex_exit(&sc->ipsec_lock); 1144 1145 /* 1146 * (2-4) Cleanup last configurations. 1147 */ 1148 if (if_ipsec_variant_is_configured(ovar)) 1149 if_ipsec_encap_detach(ovar); 1150 encap_lock_exit(); 1151 1152 if (osrc != NULL) 1153 sockaddr_free(osrc); 1154 if (odst != NULL) 1155 sockaddr_free(odst); 1156 kmem_free(ovar, sizeof(*ovar)); 1157 kmem_free(nullvar, sizeof(*nullvar)); 1158 1159 return 0; 1160 1161 out: 1162 mutex_exit(&sc->ipsec_lock); 1163 encap_lock_exit(); 1164 1165 sockaddr_free(nsrc); 1166 sockaddr_free(ndst); 1167 kmem_free(nvar, sizeof(*nvar)); 1168 kmem_free(nullvar, sizeof(*nullvar)); 1169 1170 return error; 1171 } 1172 1173 /* 1174 * Validate and delete ipsec(4) I/F configurations. 1175 * (1) validate 1176 * (1-1) Check current src and dst address pair are null, 1177 * which means the ipsec(4) I/F is already done deletetunnel. 1178 * (2) delete 1179 * (2-1) Create variant for deleted status. 1180 * (2-2) Create temporary "null" variant used to avoid to access 1181 * dangling variant while SPs are deleted and added. 1182 * NOTE: 1183 * The contents of temporary "null" variant equal to the variant 1184 * of (2-1), however two psref_target_destroy() synchronization 1185 * points are necessary to avoid to access dangling variant 1186 * while SPs are deleted and added. To implement that simply, 1187 * we use the same manner as if_ipsec_set_tunnel(), that is, 1188 * create extra "null" variant and use it temporarily. 1189 * (2-3) Swap variant include its SPs. 1190 * (2-4) Cleanup last configurations. 1191 */ 1192 static void 1193 if_ipsec_delete_tunnel(struct ifnet *ifp) 1194 { 1195 struct ipsec_softc *sc = ifp->if_softc; 1196 struct ipsec_variant *ovar, *nvar, *nullvar; 1197 struct sockaddr *osrc, *odst; 1198 int error; 1199 1200 error = encap_lock_enter(); 1201 if (error) 1202 return; 1203 1204 nvar = kmem_zalloc(sizeof(*nvar), KM_SLEEP); 1205 nullvar = kmem_zalloc(sizeof(*nullvar), KM_SLEEP); 1206 1207 mutex_enter(&sc->ipsec_lock); 1208 1209 ovar = sc->ipsec_var; 1210 osrc = ovar->iv_psrc; 1211 odst = ovar->iv_pdst; 1212 /* 1213 * (1-1) Check current src and dst address pair are null, 1214 * which means the ipsec(4) I/F is already done deletetunnel. 1215 */ 1216 if (osrc == NULL || odst == NULL) { 1217 /* address pair not changed. */ 1218 mutex_exit(&sc->ipsec_lock); 1219 encap_lock_exit(); 1220 kmem_free(nvar, sizeof(*nvar)); 1221 kmem_free(nullvar, sizeof(*nullvar)); 1222 return; 1223 } 1224 1225 /* 1226 * (2-1) Create variant for deleted status. 1227 */ 1228 if_ipsec_copy_variant(nvar, ovar); 1229 if_ipsec_clear_config(nvar); 1230 psref_target_init(&nvar->iv_psref, iv_psref_class); 1231 1232 /* 1233 * (2-2) Create temporary "null" variant used to avoid to access 1234 * dangling variant while SPs are deleted and added. 1235 */ 1236 if_ipsec_copy_variant(nullvar, ovar); 1237 if_ipsec_clear_config(nullvar); 1238 psref_target_init(&nullvar->iv_psref, iv_psref_class); 1239 /* 1240 * (2-3) Swap variant include its SPs. 1241 */ 1242 /* if_ipsec_update_variant() does not fail when delete SP only. */ 1243 (void)if_ipsec_update_variant(sc, nvar, nullvar); 1244 1245 mutex_exit(&sc->ipsec_lock); 1246 1247 /* 1248 * (2-4) Cleanup last configurations. 1249 */ 1250 if (if_ipsec_variant_is_configured(ovar)) 1251 if_ipsec_encap_detach(ovar); 1252 encap_lock_exit(); 1253 1254 sockaddr_free(osrc); 1255 sockaddr_free(odst); 1256 kmem_free(ovar, sizeof(*ovar)); 1257 kmem_free(nullvar, sizeof(*nullvar)); 1258 } 1259 1260 /* 1261 * Check IFF_NAT_T and IFF_FWD_IPV6 flags, therefore update SPs if needed. 1262 * (1) check 1263 * (1-1) Check flags are changed. 1264 * (1-2) Check current src and dst address pair. If they are null, 1265 * that means the ipsec(4) I/F is deletetunnel'ed, so it is 1266 * not needed to update. 1267 * (2) update 1268 * (2-1) Create variant for new SPs. 1269 * (2-2) Create temporary "null" variant used to avoid to access 1270 * dangling variant while SPs are deleted and added. 1271 * NOTE: 1272 * There is the same problem as if_ipsec_delete_tunnel(). 1273 * (2-3) Swap variant include its SPs. 1274 * (2-4) Cleanup unused configurations. 1275 * NOTE: use the same encap_cookies. 1276 */ 1277 static int 1278 if_ipsec_ensure_flags(struct ifnet *ifp, u_short oflags) 1279 { 1280 struct ipsec_softc *sc = ifp->if_softc; 1281 struct ipsec_variant *ovar, *nvar, *nullvar; 1282 int error; 1283 1284 /* 1285 * (1) Check flags are changed. 1286 */ 1287 if ((oflags & (IFF_NAT_T|IFF_FWD_IPV6)) == 1288 (ifp->if_flags & (IFF_NAT_T|IFF_FWD_IPV6))) 1289 return 0; /* flags not changed. */ 1290 1291 error = encap_lock_enter(); 1292 if (error) 1293 return error; 1294 1295 nvar = kmem_zalloc(sizeof(*nvar), KM_SLEEP); 1296 nullvar = kmem_zalloc(sizeof(*nullvar), KM_SLEEP); 1297 1298 mutex_enter(&sc->ipsec_lock); 1299 1300 ovar = sc->ipsec_var; 1301 /* 1302 * (1-2) Check current src and dst address pair. 1303 */ 1304 if (if_ipsec_variant_is_unconfigured(ovar)) { 1305 /* nothing to do */ 1306 mutex_exit(&sc->ipsec_lock); 1307 encap_lock_exit(); 1308 kmem_free(nvar, sizeof(*nvar)); 1309 kmem_free(nullvar, sizeof(*nullvar)); 1310 return 0; 1311 } 1312 1313 /* 1314 * (2-1) Create variant for new SPs. 1315 */ 1316 if_ipsec_copy_variant(nvar, ovar); 1317 psref_target_init(&nvar->iv_psref, iv_psref_class); 1318 /* 1319 * (2-2) Create temporary "null" variant used to avoid to access 1320 * dangling variant while SPs are deleted and added. 1321 */ 1322 if_ipsec_copy_variant(nullvar, ovar); 1323 if_ipsec_clear_config(nullvar); 1324 psref_target_init(&nullvar->iv_psref, iv_psref_class); 1325 /* 1326 * (2-3) Swap variant include its SPs. 1327 */ 1328 error = if_ipsec_update_variant(sc, nvar, nullvar); 1329 1330 mutex_exit(&sc->ipsec_lock); 1331 encap_lock_exit(); 1332 1333 /* 1334 * (2-4) Cleanup unused configurations. 1335 */ 1336 if (!error) 1337 kmem_free(ovar, sizeof(*ovar)); 1338 else 1339 kmem_free(nvar, sizeof(*ovar)); 1340 kmem_free(nullvar, sizeof(*nullvar)); 1341 1342 return error; 1343 } 1344 1345 /* 1346 * SPD management 1347 */ 1348 1349 /* 1350 * Share SP set with other NAT-T ipsec(4) I/F(s). 1351 * Return 1, when "var" shares SP set. 1352 * Return 0, when "var" cannot share SP set. 1353 * 1354 * NOTE: 1355 * if_ipsec_share_sp() and if_ipsec_unshare_sp() would require global lock 1356 * to exclude other ipsec(4) I/Fs set_tunnel/delete_tunnel. E.g. when ipsec0 1357 * and ipsec1 can share SP set, running ipsec0's set_tunnel and ipsec1's 1358 * set_tunnel causes race. 1359 * Currently, (fortunately) encap_lock works as this global lock. 1360 */ 1361 static int 1362 if_ipsec_share_sp(struct ipsec_variant *var) 1363 { 1364 struct ipsec_softc *sc = var->iv_softc; 1365 struct ipsec_softc *sc2; 1366 struct ipsec_variant *var2; 1367 struct psref psref; 1368 1369 KASSERT(encap_lock_held()); 1370 KASSERT(var->iv_psrc != NULL && var->iv_pdst != NULL); 1371 1372 mutex_enter(&ipsec_softcs.lock); 1373 LIST_FOREACH(sc2, &ipsec_softcs.list, ipsec_list) { 1374 if (sc2 == sc) 1375 continue; 1376 var2 = if_ipsec_getref_variant(sc2, &psref); 1377 if (if_ipsec_variant_is_unconfigured(var2)) { 1378 if_ipsec_putref_variant(var2, &psref); 1379 continue; 1380 } 1381 if (sockaddr_cmp(var2->iv_pdst, var->iv_pdst) != 0 || 1382 sockaddr_cmp(var2->iv_psrc, var->iv_psrc) != 0) { 1383 if_ipsec_putref_variant(var2, &psref); 1384 continue; 1385 } 1386 1387 break; 1388 } 1389 mutex_exit(&ipsec_softcs.lock); 1390 if (sc2 == NULL) 1391 return 0; /* not shared */ 1392 1393 IV_SP_IN(var) = IV_SP_IN(var2); 1394 IV_SP_IN6(var) = IV_SP_IN6(var2); 1395 IV_SP_OUT(var) = IV_SP_OUT(var2); 1396 IV_SP_OUT6(var) = IV_SP_OUT6(var2); 1397 1398 if_ipsec_putref_variant(var2, &psref); 1399 return 1; /* shared */ 1400 } 1401 1402 /* 1403 * Unshare SP set with other NAT-T ipsec(4) I/F(s). 1404 * Return 1, when "var" shared SP set, and then unshare them. 1405 * Return 0, when "var" did not share SP set. 1406 * 1407 * NOTE: 1408 * See if_ipsec_share_sp()'s note. 1409 */ 1410 static int 1411 if_ipsec_unshare_sp(struct ipsec_variant *var) 1412 { 1413 struct ipsec_softc *sc = var->iv_softc; 1414 struct ipsec_softc *sc2; 1415 struct ipsec_variant *var2; 1416 struct psref psref; 1417 1418 KASSERT(encap_lock_held()); 1419 1420 if (!var->iv_pdst || !var->iv_psrc) 1421 return 0; 1422 1423 mutex_enter(&ipsec_softcs.lock); 1424 LIST_FOREACH(sc2, &ipsec_softcs.list, ipsec_list) { 1425 if (sc2 == sc) 1426 continue; 1427 var2 = if_ipsec_getref_variant(sc2, &psref); 1428 if (!var2->iv_pdst || !var2->iv_psrc) { 1429 if_ipsec_putref_variant(var2, &psref); 1430 continue; 1431 } 1432 if (sockaddr_cmp(var2->iv_pdst, var->iv_pdst) != 0 || 1433 sockaddr_cmp(var2->iv_psrc, var->iv_psrc) != 0) { 1434 if_ipsec_putref_variant(var2, &psref); 1435 continue; 1436 } 1437 1438 break; 1439 } 1440 mutex_exit(&ipsec_softcs.lock); 1441 if (sc2 == NULL) 1442 return 0; /* not shared */ 1443 1444 IV_SP_IN(var) = NULL; 1445 IV_SP_IN6(var) = NULL; 1446 IV_SP_OUT(var) = NULL; 1447 IV_SP_OUT6(var) = NULL; 1448 if_ipsec_putref_variant(var2, &psref); 1449 return 1; /* shared */ 1450 } 1451 1452 static inline void 1453 if_ipsec_add_mbuf_optalign(struct mbuf *m0, void *data, size_t len, bool align) 1454 { 1455 struct mbuf *m; 1456 1457 MGET(m, M_WAIT, MT_DATA); 1458 if (align) { 1459 m->m_len = PFKEY_ALIGN8(len); 1460 memset(mtod(m, void *), 0, m->m_len); 1461 } else 1462 m->m_len = len; 1463 m_copyback(m, 0, len, data); 1464 m_cat(m0, m); 1465 } 1466 1467 static inline void 1468 if_ipsec_add_mbuf(struct mbuf *m0, void *data, size_t len) 1469 { 1470 1471 if_ipsec_add_mbuf_optalign(m0, data, len, true); 1472 } 1473 1474 static inline void 1475 if_ipsec_add_mbuf_addr_port(struct mbuf *m0, struct sockaddr *addr, in_port_t port, bool align) 1476 { 1477 1478 if (port == 0) { 1479 if_ipsec_add_mbuf_optalign(m0, addr, addr->sa_len, align); 1480 } else { 1481 union sockaddr_union addrport_u; 1482 struct sockaddr *addrport = &addrport_u.sa; 1483 1484 if_ipsec_set_addr_port(addrport, addr, port); 1485 if_ipsec_add_mbuf_optalign(m0, addrport, addrport->sa_len, align); 1486 } 1487 } 1488 1489 static inline void 1490 if_ipsec_add_pad(struct mbuf *m0, size_t len) 1491 { 1492 struct mbuf *m; 1493 1494 if (len == 0) 1495 return; 1496 1497 MGET(m, M_WAIT, MT_DATA); 1498 m->m_len = len; 1499 memset(mtod(m, void *), 0, m->m_len); 1500 m_cat(m0, m); 1501 } 1502 1503 static inline size_t 1504 if_ipsec_set_sadb_addr(struct sadb_address *saaddr, struct sockaddr *addr, 1505 int proto, uint16_t exttype) 1506 { 1507 size_t size; 1508 1509 KASSERT(saaddr != NULL); 1510 KASSERT(addr != NULL); 1511 1512 size = sizeof(*saaddr) + PFKEY_ALIGN8(addr->sa_len); 1513 saaddr->sadb_address_len = PFKEY_UNIT64(size); 1514 saaddr->sadb_address_exttype = exttype; 1515 saaddr->sadb_address_proto = proto; 1516 switch (addr->sa_family) { 1517 #ifdef INET 1518 case AF_INET: 1519 saaddr->sadb_address_prefixlen = sizeof(struct in_addr) << 3; 1520 break; 1521 #endif /* INET */ 1522 #ifdef INET6 1523 case AF_INET6: 1524 saaddr->sadb_address_prefixlen = sizeof(struct in6_addr) << 3; 1525 break; 1526 #endif /* INET6 */ 1527 default: 1528 log(LOG_DEBUG, 1529 "%s: Invalid address family: %d.\n", 1530 __func__, addr->sa_family); 1531 break; 1532 } 1533 saaddr->sadb_address_reserved = 0; 1534 1535 return size; 1536 } 1537 1538 static inline size_t 1539 if_ipsec_set_sadb_src(struct sadb_address *sasrc, struct sockaddr *src, 1540 int proto) 1541 { 1542 1543 return if_ipsec_set_sadb_addr(sasrc, src, proto, 1544 SADB_EXT_ADDRESS_SRC); 1545 } 1546 1547 static inline size_t 1548 if_ipsec_set_sadb_dst(struct sadb_address *sadst, struct sockaddr *dst, 1549 int proto) 1550 { 1551 1552 return if_ipsec_set_sadb_addr(sadst, dst, proto, 1553 SADB_EXT_ADDRESS_DST); 1554 } 1555 1556 static inline size_t 1557 if_ipsec_set_sadb_x_policy(struct sadb_x_policy *xpl, 1558 struct sadb_x_ipsecrequest *xisr, uint16_t policy, uint8_t dir, uint32_t id, 1559 uint8_t level, struct sockaddr *src, struct sockaddr *dst) 1560 { 1561 size_t size; 1562 1563 KASSERT(policy != IPSEC_POLICY_IPSEC || xisr != NULL); 1564 1565 size = sizeof(*xpl); 1566 if (policy == IPSEC_POLICY_IPSEC) { 1567 size += PFKEY_ALIGN8(sizeof(*xisr)); 1568 if (src != NULL && dst != NULL) 1569 size += PFKEY_ALIGN8(src->sa_len + dst->sa_len); 1570 } 1571 xpl->sadb_x_policy_len = PFKEY_UNIT64(size); 1572 xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY; 1573 xpl->sadb_x_policy_type = policy; 1574 xpl->sadb_x_policy_dir = dir; 1575 xpl->sadb_x_policy_reserved = 0; 1576 xpl->sadb_x_policy_id = id; 1577 xpl->sadb_x_policy_reserved2 = 0; 1578 1579 if (policy == IPSEC_POLICY_IPSEC) { 1580 xisr->sadb_x_ipsecrequest_len = PFKEY_ALIGN8(sizeof(*xisr)); 1581 if (src != NULL && dst != NULL) 1582 xisr->sadb_x_ipsecrequest_len += 1583 PFKEY_ALIGN8(src->sa_len + dst->sa_len); 1584 xisr->sadb_x_ipsecrequest_proto = IPPROTO_ESP; 1585 xisr->sadb_x_ipsecrequest_mode = IPSEC_MODE_TRANSPORT; 1586 xisr->sadb_x_ipsecrequest_level = level; 1587 if (level == IPSEC_LEVEL_UNIQUE) 1588 xisr->sadb_x_ipsecrequest_reqid = key_newreqid(); 1589 else 1590 xisr->sadb_x_ipsecrequest_reqid = 0; 1591 } 1592 1593 return size; 1594 } 1595 1596 static inline void 1597 if_ipsec_set_sadb_msg(struct sadb_msg *msg, uint16_t extlen, uint8_t msgtype) 1598 { 1599 1600 KASSERT(msg != NULL); 1601 1602 msg->sadb_msg_version = PF_KEY_V2; 1603 msg->sadb_msg_type = msgtype; 1604 msg->sadb_msg_errno = 0; 1605 msg->sadb_msg_satype = SADB_SATYPE_UNSPEC; 1606 msg->sadb_msg_len = PFKEY_UNIT64(sizeof(*msg)) + extlen; 1607 msg->sadb_msg_reserved = 0; 1608 msg->sadb_msg_seq = 0; /* XXXX */ 1609 msg->sadb_msg_pid = 0; /* XXXX */ 1610 } 1611 1612 static inline void 1613 if_ipsec_set_sadb_msg_add(struct sadb_msg *msg, uint16_t extlen) 1614 { 1615 1616 if_ipsec_set_sadb_msg(msg, extlen, SADB_X_SPDADD); 1617 } 1618 1619 static inline void 1620 if_ipsec_set_sadb_msg_del(struct sadb_msg *msg, uint16_t extlen) 1621 { 1622 1623 if_ipsec_set_sadb_msg(msg, extlen, SADB_X_SPDDELETE2); 1624 } 1625 1626 static int 1627 if_ipsec_set_addr_port(struct sockaddr *addrport, struct sockaddr *addr, 1628 in_port_t port) 1629 { 1630 int error = 0; 1631 1632 sockaddr_copy(addrport, addr->sa_len, addr); 1633 1634 switch (addr->sa_family) { 1635 #ifdef INET 1636 case AF_INET: { 1637 struct sockaddr_in *sin = satosin(addrport); 1638 sin->sin_port = port; 1639 break; 1640 } 1641 #endif /* INET */ 1642 #ifdef INET6 1643 case AF_INET6: { 1644 struct sockaddr_in6 *sin6 = satosin6(addrport); 1645 sin6->sin6_port = port; 1646 break; 1647 } 1648 #endif /* INET6 */ 1649 default: 1650 log(LOG_DEBUG, 1651 "%s: Invalid address family: %d.\n", 1652 __func__, addr->sa_family); 1653 error = EINVAL; 1654 } 1655 1656 return error; 1657 } 1658 1659 static struct secpolicy * 1660 if_ipsec_add_sp0(struct sockaddr *src, in_port_t sport, 1661 struct sockaddr *dst, in_port_t dport, 1662 int dir, int proto, int level, u_int policy) 1663 { 1664 struct sadb_msg msg; 1665 struct sadb_address xsrc, xdst; 1666 struct sadb_x_policy xpl; 1667 struct sadb_x_ipsecrequest xisr; 1668 size_t size; 1669 size_t padlen; 1670 uint16_t ext_msg_len = 0; 1671 struct mbuf *m; 1672 1673 memset(&msg, 0, sizeof(msg)); 1674 memset(&xsrc, 0, sizeof(xsrc)); 1675 memset(&xdst, 0, sizeof(xdst)); 1676 memset(&xpl, 0, sizeof(xpl)); 1677 memset(&xisr, 0, sizeof(xisr)); 1678 1679 MGETHDR(m, M_WAIT, MT_DATA); 1680 1681 size = if_ipsec_set_sadb_src(&xsrc, src, proto); 1682 ext_msg_len += PFKEY_UNIT64(size); 1683 size = if_ipsec_set_sadb_dst(&xdst, dst, proto); 1684 ext_msg_len += PFKEY_UNIT64(size); 1685 size = if_ipsec_set_sadb_x_policy(&xpl, &xisr, policy, dir, 0, level, NULL, NULL); 1686 ext_msg_len += PFKEY_UNIT64(size); 1687 if_ipsec_set_sadb_msg_add(&msg, ext_msg_len); 1688 1689 /* build PF_KEY message */ 1690 1691 m->m_len = sizeof(msg); 1692 m_copyback(m, 0, sizeof(msg), &msg); 1693 1694 if_ipsec_add_mbuf(m, &xsrc, sizeof(xsrc)); 1695 /* 1696 * secpolicy.spidx.{src, dst} must not be set port number, 1697 * even if it is used for NAT-T. 1698 */ 1699 if_ipsec_add_mbuf_addr_port(m, src, 0, true); 1700 padlen = PFKEY_UNUNIT64(xsrc.sadb_address_len) 1701 - (sizeof(xsrc) + PFKEY_ALIGN8(src->sa_len)); 1702 if_ipsec_add_pad(m, padlen); 1703 1704 if_ipsec_add_mbuf(m, &xdst, sizeof(xdst)); 1705 /* ditto */ 1706 if_ipsec_add_mbuf_addr_port(m, dst, 0, true); 1707 padlen = PFKEY_UNUNIT64(xdst.sadb_address_len) 1708 - (sizeof(xdst) + PFKEY_ALIGN8(dst->sa_len)); 1709 if_ipsec_add_pad(m, padlen); 1710 1711 if_ipsec_add_mbuf(m, &xpl, sizeof(xpl)); 1712 padlen = PFKEY_UNUNIT64(xpl.sadb_x_policy_len) - sizeof(xpl); 1713 if (policy == IPSEC_POLICY_IPSEC) { 1714 if_ipsec_add_mbuf(m, &xisr, sizeof(xisr)); 1715 padlen -= PFKEY_ALIGN8(sizeof(xisr)); 1716 } 1717 if_ipsec_add_pad(m, padlen); 1718 1719 /* key_kpi_spdadd() has already done KEY_SP_REF(). */ 1720 return key_kpi_spdadd(m); 1721 } 1722 1723 static int 1724 if_ipsec_add_sp(struct ipsec_variant *var, 1725 struct sockaddr *src, in_port_t sport, 1726 struct sockaddr *dst, in_port_t dport) 1727 { 1728 struct ipsec_softc *sc = var->iv_softc; 1729 int level; 1730 u_int v6policy; 1731 1732 /* 1733 * must delete sp before add it. 1734 */ 1735 KASSERT(IV_SP_IN(var) == NULL); 1736 KASSERT(IV_SP_OUT(var) == NULL); 1737 KASSERT(IV_SP_IN6(var) == NULL); 1738 KASSERT(IV_SP_OUT6(var) == NULL); 1739 1740 /* 1741 * can be shared? 1742 */ 1743 if (if_ipsec_share_sp(var)) 1744 return 0; 1745 1746 if (if_ipsec_nat_t(sc)) 1747 level = IPSEC_LEVEL_REQUIRE; 1748 else 1749 level = IPSEC_LEVEL_UNIQUE; 1750 1751 if (if_ipsec_fwd_ipv6(sc)) 1752 v6policy = IPSEC_POLICY_IPSEC; 1753 else 1754 v6policy = IPSEC_POLICY_DISCARD; 1755 1756 IV_SP_IN(var) = if_ipsec_add_sp0(dst, dport, src, sport, 1757 IPSEC_DIR_INBOUND, IPPROTO_IPIP, level, IPSEC_POLICY_IPSEC); 1758 if (IV_SP_IN(var) == NULL) 1759 goto fail; 1760 IV_SP_OUT(var) = if_ipsec_add_sp0(src, sport, dst, dport, 1761 IPSEC_DIR_OUTBOUND, IPPROTO_IPIP, level, IPSEC_POLICY_IPSEC); 1762 if (IV_SP_OUT(var) == NULL) 1763 goto fail; 1764 IV_SP_IN6(var) = if_ipsec_add_sp0(dst, dport, src, sport, 1765 IPSEC_DIR_INBOUND, IPPROTO_IPV6, level, v6policy); 1766 if (IV_SP_IN6(var) == NULL) 1767 goto fail; 1768 IV_SP_OUT6(var) = if_ipsec_add_sp0(src, sport, dst, dport, 1769 IPSEC_DIR_OUTBOUND, IPPROTO_IPV6, level, v6policy); 1770 if (IV_SP_OUT6(var) == NULL) 1771 goto fail; 1772 1773 return 0; 1774 1775 fail: 1776 if (IV_SP_IN6(var) != NULL) { 1777 if_ipsec_del_sp0(IV_SP_IN6(var)); 1778 IV_SP_IN6(var) = NULL; 1779 } 1780 if (IV_SP_OUT(var) != NULL) { 1781 if_ipsec_del_sp0(IV_SP_OUT(var)); 1782 IV_SP_OUT(var) = NULL; 1783 } 1784 if (IV_SP_IN(var) != NULL) { 1785 if_ipsec_del_sp0(IV_SP_IN(var)); 1786 IV_SP_IN(var) = NULL; 1787 } 1788 1789 return EEXIST; 1790 } 1791 1792 static int 1793 if_ipsec_del_sp0(struct secpolicy *sp) 1794 { 1795 struct sadb_msg msg; 1796 struct sadb_x_policy xpl; 1797 size_t size; 1798 uint16_t ext_msg_len = 0; 1799 int error; 1800 struct mbuf *m; 1801 1802 if (sp == NULL) 1803 return 0; 1804 1805 memset(&msg, 0, sizeof(msg)); 1806 memset(&xpl, 0, sizeof(xpl)); 1807 1808 MGETHDR(m, M_WAIT, MT_DATA); 1809 1810 size = if_ipsec_set_sadb_x_policy(&xpl, NULL, 0, 0, sp->id, 0, NULL, NULL); 1811 ext_msg_len += PFKEY_UNIT64(size); 1812 1813 if_ipsec_set_sadb_msg_del(&msg, ext_msg_len); 1814 1815 m->m_len = sizeof(msg); 1816 m_copyback(m, 0, sizeof(msg), &msg); 1817 1818 if_ipsec_add_mbuf(m, &xpl, sizeof(xpl)); 1819 1820 /* unreference correspond to key_kpi_spdadd(). */ 1821 KEY_SP_UNREF(&sp); 1822 error = key_kpi_spddelete2(m); 1823 if (error != 0) { 1824 log(LOG_ERR, "%s: cannot delete SP(ID=%u) (error=%d).\n", 1825 __func__, sp->id, error); 1826 } 1827 return error; 1828 } 1829 1830 static void 1831 if_ipsec_del_sp(struct ipsec_variant *var) 1832 { 1833 1834 /* are the SPs shared? */ 1835 if (if_ipsec_unshare_sp(var)) 1836 return; 1837 1838 (void)if_ipsec_del_sp0(IV_SP_OUT(var)); 1839 (void)if_ipsec_del_sp0(IV_SP_IN(var)); 1840 (void)if_ipsec_del_sp0(IV_SP_OUT6(var)); 1841 (void)if_ipsec_del_sp0(IV_SP_IN6(var)); 1842 IV_SP_IN(var) = NULL; 1843 IV_SP_IN6(var) = NULL; 1844 IV_SP_OUT(var) = NULL; 1845 IV_SP_OUT6(var) = NULL; 1846 } 1847 1848 static int 1849 if_ipsec_replace_sp(struct ipsec_softc *sc, struct ipsec_variant *ovar, 1850 struct ipsec_variant *nvar) 1851 { 1852 in_port_t src_port = 0; 1853 in_port_t dst_port = 0; 1854 struct sockaddr *src; 1855 struct sockaddr *dst; 1856 int error = 0; 1857 1858 KASSERT(mutex_owned(&sc->ipsec_lock)); 1859 1860 if_ipsec_del_sp(ovar); 1861 1862 src = nvar->iv_psrc; 1863 dst = nvar->iv_pdst; 1864 if (if_ipsec_nat_t(sc)) { 1865 /* NAT-T enabled */ 1866 src_port = nvar->iv_sport; 1867 dst_port = nvar->iv_dport; 1868 } 1869 if (src && dst) 1870 error = if_ipsec_add_sp(nvar, src, src_port, dst, dst_port); 1871 1872 return error; 1873 } 1874 1875 /* 1876 * ipsec_variant and its SPs update API. 1877 * 1878 * Assumption: 1879 * reader side dereferences sc->ipsec_var in reader critical section only, 1880 * that is, all of reader sides do not reader the sc->ipsec_var after 1881 * pserialize_perform(). 1882 */ 1883 static int 1884 if_ipsec_update_variant(struct ipsec_softc *sc, struct ipsec_variant *nvar, 1885 struct ipsec_variant *nullvar) 1886 { 1887 struct ifnet *ifp = &sc->ipsec_if; 1888 struct ipsec_variant *ovar = sc->ipsec_var; 1889 int error; 1890 1891 KASSERT(mutex_owned(&sc->ipsec_lock)); 1892 1893 /* 1894 * To keep consistency between ipsec(4) I/F settings and SPs, 1895 * we stop packet processing while replacing SPs, that is, we set 1896 * "null" config variant to sc->ipsec_var. 1897 */ 1898 atomic_store_release(&sc->ipsec_var, nullvar); 1899 pserialize_perform(sc->ipsec_psz); 1900 psref_target_destroy(&ovar->iv_psref, iv_psref_class); 1901 1902 error = if_ipsec_replace_sp(sc, ovar, nvar); 1903 if (!error) 1904 atomic_store_release(&sc->ipsec_var, nvar); 1905 else { 1906 psref_target_init(&ovar->iv_psref, iv_psref_class); 1907 atomic_store_release(&sc->ipsec_var, ovar); /* rollback */ 1908 } 1909 1910 pserialize_perform(sc->ipsec_psz); 1911 psref_target_destroy(&nullvar->iv_psref, iv_psref_class); 1912 1913 if (if_ipsec_variant_is_configured(sc->ipsec_var)) 1914 ifp->if_flags |= IFF_RUNNING; 1915 else 1916 ifp->if_flags &= ~IFF_RUNNING; 1917 1918 return error; 1919 } 1920