if_gif.c revision 236955
1/* $FreeBSD: head/sys/net/if_gif.c 236955 2012-06-12 12:44:17Z rrs $ */ 2/* $KAME: if_gif.c,v 1.87 2001/10/19 08:50:27 itojun Exp $ */ 3 4/*- 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33#include "opt_inet.h" 34#include "opt_inet6.h" 35 36#include <sys/param.h> 37#include <sys/systm.h> 38#include <sys/jail.h> 39#include <sys/kernel.h> 40#include <sys/malloc.h> 41#include <sys/mbuf.h> 42#include <sys/module.h> 43#include <sys/socket.h> 44#include <sys/sockio.h> 45#include <sys/errno.h> 46#include <sys/time.h> 47#include <sys/sysctl.h> 48#include <sys/syslog.h> 49#include <sys/priv.h> 50#include <sys/proc.h> 51#include <sys/protosw.h> 52#include <sys/conf.h> 53#include <machine/cpu.h> 54 55#include <net/if.h> 56#include <net/if_clone.h> 57#include <net/if_types.h> 58#include <net/netisr.h> 59#include <net/route.h> 60#include <net/bpf.h> 61#include <net/vnet.h> 62 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/ip.h> 66#ifdef INET 67#include <netinet/in_var.h> 68#include <netinet/in_gif.h> 69#include <netinet/ip_var.h> 70#endif /* INET */ 71 72#ifdef INET6 73#ifndef INET 74#include <netinet/in.h> 75#endif 76#include <netinet6/in6_var.h> 77#include <netinet/ip6.h> 78#include <netinet6/ip6_var.h> 79#include <netinet6/scope6_var.h> 80#include <netinet6/in6_gif.h> 81#include <netinet6/ip6protosw.h> 82#endif /* INET6 */ 83 84#include <netinet/ip_encap.h> 85#include <net/ethernet.h> 86#include <net/if_bridgevar.h> 87#include <net/if_gif.h> 88 89#include <security/mac/mac_framework.h> 90 91#define GIFNAME "gif" 92 93/* 94 * gif_mtx protects the global gif_softc_list. 95 */ 96static struct mtx gif_mtx; 97static MALLOC_DEFINE(M_GIF, "gif", "Generic Tunnel Interface"); 98static VNET_DEFINE(LIST_HEAD(, gif_softc), gif_softc_list); 99#define V_gif_softc_list VNET(gif_softc_list) 100 101void (*ng_gif_input_p)(struct ifnet *ifp, struct mbuf **mp, int af); 102void (*ng_gif_input_orphan_p)(struct ifnet *ifp, struct mbuf *m, int af); 103void (*ng_gif_attach_p)(struct ifnet *ifp); 104void (*ng_gif_detach_p)(struct ifnet *ifp); 105 106static void gif_start(struct ifnet *); 107static int gif_clone_create(struct if_clone *, int, caddr_t); 108static void gif_clone_destroy(struct ifnet *); 109 110IFC_SIMPLE_DECLARE(gif, 0); 111 112static int gifmodevent(module_t, int, void *); 113 114SYSCTL_DECL(_net_link); 115static SYSCTL_NODE(_net_link, IFT_GIF, gif, CTLFLAG_RW, 0, 116 "Generic Tunnel Interface"); 117#ifndef MAX_GIF_NEST 118/* 119 * This macro controls the default upper limitation on nesting of gif tunnels. 120 * Since, setting a large value to this macro with a careless configuration 121 * may introduce system crash, we don't allow any nestings by default. 122 * If you need to configure nested gif tunnels, you can define this macro 123 * in your kernel configuration file. However, if you do so, please be 124 * careful to configure the tunnels so that it won't make a loop. 125 */ 126#define MAX_GIF_NEST 1 127#endif 128static VNET_DEFINE(int, max_gif_nesting) = MAX_GIF_NEST; 129#define V_max_gif_nesting VNET(max_gif_nesting) 130SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, max_nesting, CTLFLAG_RW, 131 &VNET_NAME(max_gif_nesting), 0, "Max nested tunnels"); 132 133/* 134 * By default, we disallow creation of multiple tunnels between the same 135 * pair of addresses. Some applications require this functionality so 136 * we allow control over this check here. 137 */ 138#ifdef XBONEHACK 139static VNET_DEFINE(int, parallel_tunnels) = 1; 140#else 141static VNET_DEFINE(int, parallel_tunnels) = 0; 142#endif 143#define V_parallel_tunnels VNET(parallel_tunnels) 144SYSCTL_VNET_INT(_net_link_gif, OID_AUTO, parallel_tunnels, CTLFLAG_RW, 145 &VNET_NAME(parallel_tunnels), 0, "Allow parallel tunnels?"); 146 147/* copy from src/sys/net/if_ethersubr.c */ 148static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 149 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 150#ifndef ETHER_IS_BROADCAST 151#define ETHER_IS_BROADCAST(addr) \ 152 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 153#endif 154 155static int 156gif_clone_create(ifc, unit, params) 157 struct if_clone *ifc; 158 int unit; 159 caddr_t params; 160{ 161 struct gif_softc *sc; 162 163 sc = malloc(sizeof(struct gif_softc), M_GIF, M_WAITOK | M_ZERO); 164 sc->gif_fibnum = curthread->td_proc->p_fibnum; 165 GIF2IFP(sc) = if_alloc(IFT_GIF); 166 if (GIF2IFP(sc) == NULL) { 167 free(sc, M_GIF); 168 return (ENOSPC); 169 } 170 171 GIF_LOCK_INIT(sc); 172 173 GIF2IFP(sc)->if_softc = sc; 174 if_initname(GIF2IFP(sc), ifc->ifc_name, unit); 175 176 sc->encap_cookie4 = sc->encap_cookie6 = NULL; 177 sc->gif_options = GIF_ACCEPT_REVETHIP; 178 179 GIF2IFP(sc)->if_addrlen = 0; 180 GIF2IFP(sc)->if_mtu = GIF_MTU; 181 GIF2IFP(sc)->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 182#if 0 183 /* turn off ingress filter */ 184 GIF2IFP(sc)->if_flags |= IFF_LINK2; 185#endif 186 GIF2IFP(sc)->if_ioctl = gif_ioctl; 187 GIF2IFP(sc)->if_start = gif_start; 188 GIF2IFP(sc)->if_output = gif_output; 189 GIF2IFP(sc)->if_snd.ifq_maxlen = ifqmaxlen; 190 if_attach(GIF2IFP(sc)); 191 bpfattach(GIF2IFP(sc), DLT_NULL, sizeof(u_int32_t)); 192 if (ng_gif_attach_p != NULL) 193 (*ng_gif_attach_p)(GIF2IFP(sc)); 194 195 mtx_lock(&gif_mtx); 196 LIST_INSERT_HEAD(&V_gif_softc_list, sc, gif_list); 197 mtx_unlock(&gif_mtx); 198 199 return (0); 200} 201 202static void 203gif_clone_destroy(ifp) 204 struct ifnet *ifp; 205{ 206#if defined(INET) || defined(INET6) 207 int err; 208#endif 209 struct gif_softc *sc = ifp->if_softc; 210 211 mtx_lock(&gif_mtx); 212 LIST_REMOVE(sc, gif_list); 213 mtx_unlock(&gif_mtx); 214 215 gif_delete_tunnel(ifp); 216#ifdef INET6 217 if (sc->encap_cookie6 != NULL) { 218 err = encap_detach(sc->encap_cookie6); 219 KASSERT(err == 0, ("Unexpected error detaching encap_cookie6")); 220 } 221#endif 222#ifdef INET 223 if (sc->encap_cookie4 != NULL) { 224 err = encap_detach(sc->encap_cookie4); 225 KASSERT(err == 0, ("Unexpected error detaching encap_cookie4")); 226 } 227#endif 228 229 if (ng_gif_detach_p != NULL) 230 (*ng_gif_detach_p)(ifp); 231 bpfdetach(ifp); 232 if_detach(ifp); 233 if_free(ifp); 234 235 GIF_LOCK_DESTROY(sc); 236 237 free(sc, M_GIF); 238} 239 240static void 241vnet_gif_init(const void *unused __unused) 242{ 243 244 LIST_INIT(&V_gif_softc_list); 245} 246VNET_SYSINIT(vnet_gif_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_gif_init, 247 NULL); 248 249static int 250gifmodevent(mod, type, data) 251 module_t mod; 252 int type; 253 void *data; 254{ 255 256 switch (type) { 257 case MOD_LOAD: 258 mtx_init(&gif_mtx, "gif_mtx", NULL, MTX_DEF); 259 if_clone_attach(&gif_cloner); 260 break; 261 262 case MOD_UNLOAD: 263 if_clone_detach(&gif_cloner); 264 mtx_destroy(&gif_mtx); 265 break; 266 default: 267 return EOPNOTSUPP; 268 } 269 return 0; 270} 271 272static moduledata_t gif_mod = { 273 "if_gif", 274 gifmodevent, 275 0 276}; 277 278DECLARE_MODULE(if_gif, gif_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 279MODULE_VERSION(if_gif, 1); 280 281int 282gif_encapcheck(m, off, proto, arg) 283 const struct mbuf *m; 284 int off; 285 int proto; 286 void *arg; 287{ 288 struct ip ip; 289 struct gif_softc *sc; 290 291 sc = (struct gif_softc *)arg; 292 if (sc == NULL) 293 return 0; 294 295 if ((GIF2IFP(sc)->if_flags & IFF_UP) == 0) 296 return 0; 297 298 /* no physical address */ 299 if (!sc->gif_psrc || !sc->gif_pdst) 300 return 0; 301 302 switch (proto) { 303#ifdef INET 304 case IPPROTO_IPV4: 305 break; 306#endif 307#ifdef INET6 308 case IPPROTO_IPV6: 309 break; 310#endif 311 case IPPROTO_ETHERIP: 312 break; 313 314 default: 315 return 0; 316 } 317 318 /* Bail on short packets */ 319 if (m->m_pkthdr.len < sizeof(ip)) 320 return 0; 321 322 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip); 323 324 switch (ip.ip_v) { 325#ifdef INET 326 case 4: 327 if (sc->gif_psrc->sa_family != AF_INET || 328 sc->gif_pdst->sa_family != AF_INET) 329 return 0; 330 return gif_encapcheck4(m, off, proto, arg); 331#endif 332#ifdef INET6 333 case 6: 334 if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) 335 return 0; 336 if (sc->gif_psrc->sa_family != AF_INET6 || 337 sc->gif_pdst->sa_family != AF_INET6) 338 return 0; 339 return gif_encapcheck6(m, off, proto, arg); 340#endif 341 default: 342 return 0; 343 } 344} 345#ifdef INET 346#define GIF_HDR_LEN (ETHER_HDR_LEN + sizeof (struct ip)) 347#endif 348#ifdef INET6 349#define GIF_HDR_LEN6 (ETHER_HDR_LEN + sizeof (struct ip6_hdr)) 350#endif 351 352static void 353gif_start(struct ifnet *ifp) 354{ 355 struct gif_softc *sc; 356 struct mbuf *m; 357 uint32_t af; 358 int error = 0; 359 360 sc = ifp->if_softc; 361 GIF_LOCK(sc); 362 if (ifp->if_drv_flags & IFF_DRV_OACTIVE) { 363 /* Already active */ 364 ifp->if_drv_flags |= IFF_GIF_WANTED; 365 GIF_UNLOCK(sc); 366 return; 367 } 368 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 369 GIF_UNLOCK(sc); 370keep_going: 371 while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) { 372 373 IFQ_DRV_DEQUEUE(&ifp->if_snd, m); 374 if (m == 0) 375 break; 376 377#ifdef ALTQ 378 /* Take out those altq bytes we add in gif_output */ 379#ifdef INET 380 if (sc->gif_psrc->sa_family == AF_INET) 381 m->m_pkthdr.len -= GIF_HDR_LEN; 382#endif 383#ifdef INET6 384 if (sc->gif_psrc->sa_family == AF_INET6) 385 m->m_pkthdr.len -= GIF_HDR_LEN6; 386#endif 387#endif 388 /* Now pull back the af in packet that 389 * was saved in the address location. 390 */ 391 af = m->m_pkthdr.csum_data; 392 393 if (ifp->if_bridge) 394 af = AF_LINK; 395 396 BPF_MTAP2(ifp, &af, sizeof(af), m); 397 ifp->if_opackets++; 398 399/* Done by IFQ_HANDOFF */ 400/* ifp->if_obytes += m->m_pkthdr.len;*/ 401 /* override to IPPROTO_ETHERIP for bridged traffic */ 402 403 M_SETFIB(m, sc->gif_fibnum); 404 /* inner AF-specific encapsulation */ 405 /* XXX should we check if our outer source is legal? */ 406 /* dispatch to output logic based on outer AF */ 407 switch (sc->gif_psrc->sa_family) { 408#ifdef INET 409 case AF_INET: 410 error = in_gif_output(ifp, af, m); 411 break; 412#endif 413#ifdef INET6 414 case AF_INET6: 415 error = in6_gif_output(ifp, af, m); 416 break; 417#endif 418 default: 419 m_freem(m); 420 error = ENETDOWN; 421 } 422 if (error) 423 ifp->if_oerrors++; 424 425 } 426 GIF_LOCK(sc); 427 if (ifp->if_drv_flags & IFF_GIF_WANTED) { 428 /* Someone did a start while 429 * we were unlocked and processing 430 * lets clear the flag and try again. 431 */ 432 ifp->if_drv_flags &= ~IFF_GIF_WANTED; 433 GIF_UNLOCK(sc); 434 goto keep_going; 435 } 436 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 437 GIF_UNLOCK(sc); 438 return; 439} 440 441int 442gif_output(ifp, m, dst, ro) 443 struct ifnet *ifp; 444 struct mbuf *m; 445 struct sockaddr *dst; 446 struct route *ro; 447{ 448 struct gif_softc *sc = ifp->if_softc; 449 struct m_tag *mtag; 450 int error = 0; 451 int gif_called; 452 uint32_t af; 453#ifdef MAC 454 error = mac_ifnet_check_transmit(ifp, m); 455 if (error) { 456 m_freem(m); 457 goto end; 458 } 459#endif 460 461 /* 462 * gif may cause infinite recursion calls when misconfigured. 463 * We'll prevent this by detecting loops. 464 * 465 * High nesting level may cause stack exhaustion. 466 * We'll prevent this by introducing upper limit. 467 */ 468 gif_called = 1; 469 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, NULL); 470 while (mtag != NULL) { 471 if (*(struct ifnet **)(mtag + 1) == ifp) { 472 log(LOG_NOTICE, 473 "gif_output: loop detected on %s\n", 474 (*(struct ifnet **)(mtag + 1))->if_xname); 475 m_freem(m); 476 error = EIO; /* is there better errno? */ 477 goto end; 478 } 479 mtag = m_tag_locate(m, MTAG_GIF, MTAG_GIF_CALLED, mtag); 480 gif_called++; 481 } 482 if (gif_called > V_max_gif_nesting) { 483 log(LOG_NOTICE, 484 "gif_output: recursively called too many times(%d)\n", 485 gif_called); 486 m_freem(m); 487 error = EIO; /* is there better errno? */ 488 goto end; 489 } 490 mtag = m_tag_alloc(MTAG_GIF, MTAG_GIF_CALLED, sizeof(struct ifnet *), 491 M_NOWAIT); 492 if (mtag == NULL) { 493 m_freem(m); 494 error = ENOMEM; 495 goto end; 496 } 497 *(struct ifnet **)(mtag + 1) = ifp; 498 m_tag_prepend(m, mtag); 499 500 m->m_flags &= ~(M_BCAST|M_MCAST); 501 /* BPF writes need to be handled specially. */ 502 if (dst->sa_family == AF_UNSPEC) { 503 bcopy(dst->sa_data, &af, sizeof(af)); 504 dst->sa_family = af; 505 } 506 af = dst->sa_family; 507 /* Now save the af in the inbound pkt csum 508 * data, this is a cheat since really 509 * gif tunnels don't do offload. 510 */ 511 m->m_pkthdr.csum_data = af; 512 if (!(ifp->if_flags & IFF_UP) || 513 sc->gif_psrc == NULL || sc->gif_pdst == NULL) { 514 m_freem(m); 515 error = ENETDOWN; 516 goto end; 517 } 518#ifdef ALTQ 519 /* Make altq aware of the bytes we will add 520 * when we actually send it. 521 */ 522#ifdef INET 523 if (sc->gif_psrc->sa_family == AF_INET) 524 m->m_pkthdr.len += GIF_HDR_LEN; 525#endif 526#ifdef INET6 527 if (sc->gif_psrc->sa_family == AF_INET6) 528 m->m_pkthdr.len += GIF_HDR_LEN6; 529#endif 530#endif 531 /* 532 * Queue message on interface, update output statistics if 533 * successful, and start output if interface not yet active. 534 */ 535 IFQ_HANDOFF(ifp, m, error); 536 end: 537 if (error) 538 ifp->if_oerrors++; 539 return (error); 540} 541 542void 543gif_input(m, af, ifp) 544 struct mbuf *m; 545 int af; 546 struct ifnet *ifp; 547{ 548 int isr, n; 549 struct gif_softc *sc; 550 struct etherip_header *eip; 551 struct ether_header *eh; 552 struct ifnet *oldifp; 553 554 if (ifp == NULL) { 555 /* just in case */ 556 m_freem(m); 557 return; 558 } 559 sc = ifp->if_softc; 560 m->m_pkthdr.rcvif = ifp; 561 562#ifdef MAC 563 mac_ifnet_create_mbuf(ifp, m); 564#endif 565 566 if (bpf_peers_present(ifp->if_bpf)) { 567 u_int32_t af1 = af; 568 bpf_mtap2(ifp->if_bpf, &af1, sizeof(af1), m); 569 } 570 571 if (ng_gif_input_p != NULL) { 572 (*ng_gif_input_p)(ifp, &m, af); 573 if (m == NULL) 574 return; 575 } 576 577 /* 578 * Put the packet to the network layer input queue according to the 579 * specified address family. 580 * Note: older versions of gif_input directly called network layer 581 * input functions, e.g. ip6_input, here. We changed the policy to 582 * prevent too many recursive calls of such input functions, which 583 * might cause kernel panic. But the change may introduce another 584 * problem; if the input queue is full, packets are discarded. 585 * The kernel stack overflow really happened, and we believed 586 * queue-full rarely occurs, so we changed the policy. 587 */ 588 switch (af) { 589#ifdef INET 590 case AF_INET: 591 isr = NETISR_IP; 592 break; 593#endif 594#ifdef INET6 595 case AF_INET6: 596 isr = NETISR_IPV6; 597 break; 598#endif 599 case AF_LINK: 600 n = sizeof(struct etherip_header) + sizeof(struct ether_header); 601 if (n > m->m_len) { 602 m = m_pullup(m, n); 603 if (m == NULL) { 604 ifp->if_ierrors++; 605 return; 606 } 607 } 608 609 eip = mtod(m, struct etherip_header *); 610 /* 611 * GIF_ACCEPT_REVETHIP (enabled by default) intentionally 612 * accepts an EtherIP packet with revered version field in 613 * the header. This is a knob for backward compatibility 614 * with FreeBSD 7.2R or prior. 615 */ 616 if (sc->gif_options & GIF_ACCEPT_REVETHIP) { 617 if (eip->eip_resvl != ETHERIP_VERSION 618 && eip->eip_ver != ETHERIP_VERSION) { 619 /* discard unknown versions */ 620 m_freem(m); 621 return; 622 } 623 } else { 624 if (eip->eip_ver != ETHERIP_VERSION) { 625 /* discard unknown versions */ 626 m_freem(m); 627 return; 628 } 629 } 630 m_adj(m, sizeof(struct etherip_header)); 631 632 m->m_flags &= ~(M_BCAST|M_MCAST); 633 m->m_pkthdr.rcvif = ifp; 634 635 if (ifp->if_bridge) { 636 oldifp = ifp; 637 eh = mtod(m, struct ether_header *); 638 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 639 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 640 m->m_flags |= M_BCAST; 641 else 642 m->m_flags |= M_MCAST; 643 ifp->if_imcasts++; 644 } 645 BRIDGE_INPUT(ifp, m); 646 647 if (m != NULL && ifp != oldifp) { 648 /* 649 * The bridge gave us back itself or one of the 650 * members for which the frame is addressed. 651 */ 652 ether_demux(ifp, m); 653 return; 654 } 655 } 656 if (m != NULL) 657 m_freem(m); 658 return; 659 660 default: 661 if (ng_gif_input_orphan_p != NULL) 662 (*ng_gif_input_orphan_p)(ifp, m, af); 663 else 664 m_freem(m); 665 return; 666 } 667 668 ifp->if_ipackets++; 669 ifp->if_ibytes += m->m_pkthdr.len; 670 M_SETFIB(m, ifp->if_fib); 671 netisr_dispatch(isr, m); 672} 673 674/* XXX how should we handle IPv6 scope on SIOC[GS]IFPHYADDR? */ 675int 676gif_ioctl(ifp, cmd, data) 677 struct ifnet *ifp; 678 u_long cmd; 679 caddr_t data; 680{ 681 struct gif_softc *sc = ifp->if_softc; 682 struct ifreq *ifr = (struct ifreq*)data; 683 int error = 0, size; 684 u_int options; 685 struct sockaddr *dst, *src; 686#ifdef SIOCSIFMTU /* xxx */ 687 u_long mtu; 688#endif 689 690 switch (cmd) { 691 case SIOCSIFADDR: 692 ifp->if_flags |= IFF_UP; 693 break; 694 695 case SIOCSIFDSTADDR: 696 break; 697 698 case SIOCADDMULTI: 699 case SIOCDELMULTI: 700 break; 701 702#ifdef SIOCSIFMTU /* xxx */ 703 case SIOCGIFMTU: 704 break; 705 706 case SIOCSIFMTU: 707 mtu = ifr->ifr_mtu; 708 if (mtu < GIF_MTU_MIN || mtu > GIF_MTU_MAX) 709 return (EINVAL); 710 ifp->if_mtu = mtu; 711 break; 712#endif /* SIOCSIFMTU */ 713 714#ifdef INET 715 case SIOCSIFPHYADDR: 716#endif 717#ifdef INET6 718 case SIOCSIFPHYADDR_IN6: 719#endif /* INET6 */ 720 case SIOCSLIFPHYADDR: 721 switch (cmd) { 722#ifdef INET 723 case SIOCSIFPHYADDR: 724 src = (struct sockaddr *) 725 &(((struct in_aliasreq *)data)->ifra_addr); 726 dst = (struct sockaddr *) 727 &(((struct in_aliasreq *)data)->ifra_dstaddr); 728 break; 729#endif 730#ifdef INET6 731 case SIOCSIFPHYADDR_IN6: 732 src = (struct sockaddr *) 733 &(((struct in6_aliasreq *)data)->ifra_addr); 734 dst = (struct sockaddr *) 735 &(((struct in6_aliasreq *)data)->ifra_dstaddr); 736 break; 737#endif 738 case SIOCSLIFPHYADDR: 739 src = (struct sockaddr *) 740 &(((struct if_laddrreq *)data)->addr); 741 dst = (struct sockaddr *) 742 &(((struct if_laddrreq *)data)->dstaddr); 743 break; 744 default: 745 return EINVAL; 746 } 747 748 /* sa_family must be equal */ 749 if (src->sa_family != dst->sa_family) 750 return EINVAL; 751 752 /* validate sa_len */ 753 switch (src->sa_family) { 754#ifdef INET 755 case AF_INET: 756 if (src->sa_len != sizeof(struct sockaddr_in)) 757 return EINVAL; 758 break; 759#endif 760#ifdef INET6 761 case AF_INET6: 762 if (src->sa_len != sizeof(struct sockaddr_in6)) 763 return EINVAL; 764 break; 765#endif 766 default: 767 return EAFNOSUPPORT; 768 } 769 switch (dst->sa_family) { 770#ifdef INET 771 case AF_INET: 772 if (dst->sa_len != sizeof(struct sockaddr_in)) 773 return EINVAL; 774 break; 775#endif 776#ifdef INET6 777 case AF_INET6: 778 if (dst->sa_len != sizeof(struct sockaddr_in6)) 779 return EINVAL; 780 break; 781#endif 782 default: 783 return EAFNOSUPPORT; 784 } 785 786 /* check sa_family looks sane for the cmd */ 787 switch (cmd) { 788 case SIOCSIFPHYADDR: 789 if (src->sa_family == AF_INET) 790 break; 791 return EAFNOSUPPORT; 792#ifdef INET6 793 case SIOCSIFPHYADDR_IN6: 794 if (src->sa_family == AF_INET6) 795 break; 796 return EAFNOSUPPORT; 797#endif /* INET6 */ 798 case SIOCSLIFPHYADDR: 799 /* checks done in the above */ 800 break; 801 } 802 803 error = gif_set_tunnel(GIF2IFP(sc), src, dst); 804 break; 805 806#ifdef SIOCDIFPHYADDR 807 case SIOCDIFPHYADDR: 808 gif_delete_tunnel(GIF2IFP(sc)); 809 break; 810#endif 811 812 case SIOCGIFPSRCADDR: 813#ifdef INET6 814 case SIOCGIFPSRCADDR_IN6: 815#endif /* INET6 */ 816 if (sc->gif_psrc == NULL) { 817 error = EADDRNOTAVAIL; 818 goto bad; 819 } 820 src = sc->gif_psrc; 821 switch (cmd) { 822#ifdef INET 823 case SIOCGIFPSRCADDR: 824 dst = &ifr->ifr_addr; 825 size = sizeof(ifr->ifr_addr); 826 break; 827#endif /* INET */ 828#ifdef INET6 829 case SIOCGIFPSRCADDR_IN6: 830 dst = (struct sockaddr *) 831 &(((struct in6_ifreq *)data)->ifr_addr); 832 size = sizeof(((struct in6_ifreq *)data)->ifr_addr); 833 break; 834#endif /* INET6 */ 835 default: 836 error = EADDRNOTAVAIL; 837 goto bad; 838 } 839 if (src->sa_len > size) 840 return EINVAL; 841 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len); 842#ifdef INET6 843 if (dst->sa_family == AF_INET6) { 844 error = sa6_recoverscope((struct sockaddr_in6 *)dst); 845 if (error != 0) 846 return (error); 847 } 848#endif 849 break; 850 851 case SIOCGIFPDSTADDR: 852#ifdef INET6 853 case SIOCGIFPDSTADDR_IN6: 854#endif /* INET6 */ 855 if (sc->gif_pdst == NULL) { 856 error = EADDRNOTAVAIL; 857 goto bad; 858 } 859 src = sc->gif_pdst; 860 switch (cmd) { 861#ifdef INET 862 case SIOCGIFPDSTADDR: 863 dst = &ifr->ifr_addr; 864 size = sizeof(ifr->ifr_addr); 865 break; 866#endif /* INET */ 867#ifdef INET6 868 case SIOCGIFPDSTADDR_IN6: 869 dst = (struct sockaddr *) 870 &(((struct in6_ifreq *)data)->ifr_addr); 871 size = sizeof(((struct in6_ifreq *)data)->ifr_addr); 872 break; 873#endif /* INET6 */ 874 default: 875 error = EADDRNOTAVAIL; 876 goto bad; 877 } 878 if (src->sa_len > size) 879 return EINVAL; 880 error = prison_if(curthread->td_ucred, src); 881 if (error != 0) 882 return (error); 883 error = prison_if(curthread->td_ucred, dst); 884 if (error != 0) 885 return (error); 886 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len); 887#ifdef INET6 888 if (dst->sa_family == AF_INET6) { 889 error = sa6_recoverscope((struct sockaddr_in6 *)dst); 890 if (error != 0) 891 return (error); 892 } 893#endif 894 break; 895 896 case SIOCGLIFPHYADDR: 897 if (sc->gif_psrc == NULL || sc->gif_pdst == NULL) { 898 error = EADDRNOTAVAIL; 899 goto bad; 900 } 901 902 /* copy src */ 903 src = sc->gif_psrc; 904 dst = (struct sockaddr *) 905 &(((struct if_laddrreq *)data)->addr); 906 size = sizeof(((struct if_laddrreq *)data)->addr); 907 if (src->sa_len > size) 908 return EINVAL; 909 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len); 910 911 /* copy dst */ 912 src = sc->gif_pdst; 913 dst = (struct sockaddr *) 914 &(((struct if_laddrreq *)data)->dstaddr); 915 size = sizeof(((struct if_laddrreq *)data)->dstaddr); 916 if (src->sa_len > size) 917 return EINVAL; 918 bcopy((caddr_t)src, (caddr_t)dst, src->sa_len); 919 break; 920 921 case SIOCSIFFLAGS: 922 /* if_ioctl() takes care of it */ 923 break; 924 925 case GIFGOPTS: 926 options = sc->gif_options; 927 error = copyout(&options, ifr->ifr_data, 928 sizeof(options)); 929 break; 930 931 case GIFSOPTS: 932 if ((error = priv_check(curthread, PRIV_NET_GIF)) != 0) 933 break; 934 error = copyin(ifr->ifr_data, &options, sizeof(options)); 935 if (error) 936 break; 937 if (options & ~GIF_OPTMASK) 938 error = EINVAL; 939 else 940 sc->gif_options = options; 941 break; 942 943 default: 944 error = EINVAL; 945 break; 946 } 947 bad: 948 return error; 949} 950 951/* 952 * XXXRW: There's a general event-ordering issue here: the code to check 953 * if a given tunnel is already present happens before we perform a 954 * potentially blocking setup of the tunnel. This code needs to be 955 * re-ordered so that the check and replacement can be atomic using 956 * a mutex. 957 */ 958int 959gif_set_tunnel(ifp, src, dst) 960 struct ifnet *ifp; 961 struct sockaddr *src; 962 struct sockaddr *dst; 963{ 964 struct gif_softc *sc = ifp->if_softc; 965 struct gif_softc *sc2; 966 struct sockaddr *osrc, *odst, *sa; 967 int error = 0; 968 969 mtx_lock(&gif_mtx); 970 LIST_FOREACH(sc2, &V_gif_softc_list, gif_list) { 971 if (sc2 == sc) 972 continue; 973 if (!sc2->gif_pdst || !sc2->gif_psrc) 974 continue; 975 if (sc2->gif_pdst->sa_family != dst->sa_family || 976 sc2->gif_pdst->sa_len != dst->sa_len || 977 sc2->gif_psrc->sa_family != src->sa_family || 978 sc2->gif_psrc->sa_len != src->sa_len) 979 continue; 980 981 /* 982 * Disallow parallel tunnels unless instructed 983 * otherwise. 984 */ 985 if (!V_parallel_tunnels && 986 bcmp(sc2->gif_pdst, dst, dst->sa_len) == 0 && 987 bcmp(sc2->gif_psrc, src, src->sa_len) == 0) { 988 error = EADDRNOTAVAIL; 989 mtx_unlock(&gif_mtx); 990 goto bad; 991 } 992 993 /* XXX both end must be valid? (I mean, not 0.0.0.0) */ 994 } 995 mtx_unlock(&gif_mtx); 996 997 /* XXX we can detach from both, but be polite just in case */ 998 if (sc->gif_psrc) 999 switch (sc->gif_psrc->sa_family) { 1000#ifdef INET 1001 case AF_INET: 1002 (void)in_gif_detach(sc); 1003 break; 1004#endif 1005#ifdef INET6 1006 case AF_INET6: 1007 (void)in6_gif_detach(sc); 1008 break; 1009#endif 1010 } 1011 1012 osrc = sc->gif_psrc; 1013 sa = (struct sockaddr *)malloc(src->sa_len, M_IFADDR, M_WAITOK); 1014 bcopy((caddr_t)src, (caddr_t)sa, src->sa_len); 1015 sc->gif_psrc = sa; 1016 1017 odst = sc->gif_pdst; 1018 sa = (struct sockaddr *)malloc(dst->sa_len, M_IFADDR, M_WAITOK); 1019 bcopy((caddr_t)dst, (caddr_t)sa, dst->sa_len); 1020 sc->gif_pdst = sa; 1021 1022 switch (sc->gif_psrc->sa_family) { 1023#ifdef INET 1024 case AF_INET: 1025 error = in_gif_attach(sc); 1026 break; 1027#endif 1028#ifdef INET6 1029 case AF_INET6: 1030 /* 1031 * Check validity of the scope zone ID of the addresses, and 1032 * convert it into the kernel internal form if necessary. 1033 */ 1034 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_psrc, 0); 1035 if (error != 0) 1036 break; 1037 error = sa6_embedscope((struct sockaddr_in6 *)sc->gif_pdst, 0); 1038 if (error != 0) 1039 break; 1040 error = in6_gif_attach(sc); 1041 break; 1042#endif 1043 } 1044 if (error) { 1045 /* rollback */ 1046 free((caddr_t)sc->gif_psrc, M_IFADDR); 1047 free((caddr_t)sc->gif_pdst, M_IFADDR); 1048 sc->gif_psrc = osrc; 1049 sc->gif_pdst = odst; 1050 goto bad; 1051 } 1052 1053 if (osrc) 1054 free((caddr_t)osrc, M_IFADDR); 1055 if (odst) 1056 free((caddr_t)odst, M_IFADDR); 1057 1058 bad: 1059 if (sc->gif_psrc && sc->gif_pdst) 1060 ifp->if_drv_flags |= IFF_DRV_RUNNING; 1061 else 1062 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 1063 1064 return error; 1065} 1066 1067void 1068gif_delete_tunnel(ifp) 1069 struct ifnet *ifp; 1070{ 1071 struct gif_softc *sc = ifp->if_softc; 1072 1073 if (sc->gif_psrc) { 1074 free((caddr_t)sc->gif_psrc, M_IFADDR); 1075 sc->gif_psrc = NULL; 1076 } 1077 if (sc->gif_pdst) { 1078 free((caddr_t)sc->gif_pdst, M_IFADDR); 1079 sc->gif_pdst = NULL; 1080 } 1081 /* it is safe to detach from both */ 1082#ifdef INET 1083 (void)in_gif_detach(sc); 1084#endif 1085#ifdef INET6 1086 (void)in6_gif_detach(sc); 1087#endif 1088 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 1089} 1090