if_gre.c revision 1.53
1 /* $NetBSD: if_gre.c,v 1.53 2004/12/04 18:31:43 peter Exp $ */ 2 3 /* 4 * Copyright (c) 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Heiko W.Rupp <hwr (at) pilhuhn.de> 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Encapsulate L3 protocols into IP 41 * See RFC 1701 and 1702 for more details. 42 * If_gre is compatible with Cisco GRE tunnels, so you can 43 * have a NetBSD box as the other end of a tunnel interface of a Cisco 44 * router. See gre(4) for more details. 45 * Also supported: IP in IP encaps (proto 55) as of RFC 2004 46 */ 47 48 #include <sys/cdefs.h> 49 __KERNEL_RCSID(0, "$NetBSD: if_gre.c,v 1.53 2004/12/04 18:31:43 peter Exp $"); 50 51 #include "opt_inet.h" 52 #include "opt_ns.h" 53 #include "bpfilter.h" 54 55 #include <sys/param.h> 56 #include <sys/malloc.h> 57 #include <sys/mbuf.h> 58 #include <sys/proc.h> 59 #include <sys/protosw.h> 60 #include <sys/socket.h> 61 #include <sys/ioctl.h> 62 #include <sys/queue.h> 63 #if __NetBSD__ 64 #include <sys/systm.h> 65 #endif 66 67 #include <machine/cpu.h> 68 69 #include <net/ethertypes.h> 70 #include <net/if.h> 71 #include <net/if_types.h> 72 #include <net/netisr.h> 73 #include <net/route.h> 74 75 #ifdef INET 76 #include <netinet/in.h> 77 #include <netinet/in_systm.h> 78 #include <netinet/in_var.h> 79 #include <netinet/ip.h> 80 #include <netinet/ip_var.h> 81 #else 82 #error "Huh? if_gre without inet?" 83 #endif 84 85 #ifdef NS 86 #include <netns/ns.h> 87 #include <netns/ns_if.h> 88 #endif 89 90 #ifdef NETATALK 91 #include <netatalk/at.h> 92 #include <netatalk/at_var.h> 93 #include <netatalk/at_extern.h> 94 #endif 95 96 #if NBPFILTER > 0 97 #include <sys/time.h> 98 #include <net/bpf.h> 99 #endif 100 101 #include <net/if_gre.h> 102 103 /* 104 * It is not easy to calculate the right value for a GRE MTU. 105 * We leave this task to the admin and use the same default that 106 * other vendors use. 107 */ 108 #define GREMTU 1476 109 110 struct gre_softc_head gre_softc_list; 111 int ip_gre_ttl = GRE_TTL; 112 113 int gre_clone_create __P((struct if_clone *, int)); 114 int gre_clone_destroy __P((struct ifnet *)); 115 116 struct if_clone gre_cloner = 117 IF_CLONE_INITIALIZER("gre", gre_clone_create, gre_clone_destroy); 118 119 int gre_compute_route(struct gre_softc *sc); 120 121 void greattach __P((int)); 122 123 /* ARGSUSED */ 124 void 125 greattach(count) 126 int count; 127 { 128 129 LIST_INIT(&gre_softc_list); 130 if_clone_attach(&gre_cloner); 131 } 132 133 int 134 gre_clone_create(ifc, unit) 135 struct if_clone *ifc; 136 int unit; 137 { 138 struct gre_softc *sc; 139 140 sc = malloc(sizeof(struct gre_softc), M_DEVBUF, M_WAITOK); 141 memset(sc, 0, sizeof(struct gre_softc)); 142 143 snprintf(sc->sc_if.if_xname, sizeof(sc->sc_if.if_xname), "%s%d", 144 ifc->ifc_name, unit); 145 sc->sc_if.if_softc = sc; 146 sc->sc_if.if_type = IFT_TUNNEL; 147 sc->sc_if.if_addrlen = 0; 148 sc->sc_if.if_hdrlen = 24; /* IP + GRE */ 149 sc->sc_if.if_dlt = DLT_NULL; 150 sc->sc_if.if_mtu = GREMTU; 151 sc->sc_if.if_flags = IFF_POINTOPOINT|IFF_MULTICAST; 152 sc->sc_if.if_output = gre_output; 153 sc->sc_if.if_ioctl = gre_ioctl; 154 sc->g_dst.s_addr = sc->g_src.s_addr = INADDR_ANY; 155 sc->g_proto = IPPROTO_GRE; 156 sc->sc_if.if_flags |= IFF_LINK0; 157 if_attach(&sc->sc_if); 158 if_alloc_sadl(&sc->sc_if); 159 #if NBPFILTER > 0 160 bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int32_t)); 161 #endif 162 LIST_INSERT_HEAD(&gre_softc_list, sc, sc_list); 163 return (0); 164 } 165 166 int 167 gre_clone_destroy(ifp) 168 struct ifnet *ifp; 169 { 170 struct gre_softc *sc = ifp->if_softc; 171 172 LIST_REMOVE(sc, sc_list); 173 #if NBPFILTER > 0 174 bpfdetach(ifp); 175 #endif 176 if_detach(ifp); 177 free(sc, M_DEVBUF); 178 179 return (0); 180 } 181 182 /* 183 * The output routine. Takes a packet and encapsulates it in the protocol 184 * given by sc->g_proto. See also RFC 1701 and RFC 2004 185 */ 186 int 187 gre_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 188 struct rtentry *rt) 189 { 190 int error = 0; 191 struct gre_softc *sc = ifp->if_softc; 192 struct greip *gh; 193 struct ip *ip; 194 u_int16_t etype = 0; 195 struct mobile_h mob_h; 196 197 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) == 0 || 198 sc->g_src.s_addr == INADDR_ANY || sc->g_dst.s_addr == INADDR_ANY) { 199 m_freem(m); 200 error = ENETDOWN; 201 goto end; 202 } 203 204 gh = NULL; 205 ip = NULL; 206 207 #if NBPFILTER >0 208 if (ifp->if_bpf) 209 bpf_mtap_af(ifp->if_bpf, dst->sa_family, m); 210 #endif 211 212 m->m_flags &= ~(M_BCAST|M_MCAST); 213 214 if (sc->g_proto == IPPROTO_MOBILE) { 215 if (dst->sa_family == AF_INET) { 216 struct mbuf *m0; 217 int msiz; 218 219 ip = mtod(m, struct ip *); 220 221 memset(&mob_h, 0, MOB_H_SIZ_L); 222 mob_h.proto = (ip->ip_p) << 8; 223 mob_h.odst = ip->ip_dst.s_addr; 224 ip->ip_dst.s_addr = sc->g_dst.s_addr; 225 226 /* 227 * If the packet comes from our host, we only change 228 * the destination address in the IP header. 229 * Else we also need to save and change the source 230 */ 231 if (in_hosteq(ip->ip_src, sc->g_src)) { 232 msiz = MOB_H_SIZ_S; 233 } else { 234 mob_h.proto |= MOB_H_SBIT; 235 mob_h.osrc = ip->ip_src.s_addr; 236 ip->ip_src.s_addr = sc->g_src.s_addr; 237 msiz = MOB_H_SIZ_L; 238 } 239 HTONS(mob_h.proto); 240 mob_h.hcrc = gre_in_cksum((u_int16_t *)&mob_h, msiz); 241 242 if ((m->m_data - msiz) < m->m_pktdat) { 243 /* need new mbuf */ 244 MGETHDR(m0, M_DONTWAIT, MT_HEADER); 245 if (m0 == NULL) { 246 IF_DROP(&ifp->if_snd); 247 m_freem(m); 248 error = ENOBUFS; 249 goto end; 250 } 251 m0->m_next = m; 252 m->m_data += sizeof(struct ip); 253 m->m_len -= sizeof(struct ip); 254 m0->m_pkthdr.len = m->m_pkthdr.len + msiz; 255 m0->m_len = msiz + sizeof(struct ip); 256 m0->m_data += max_linkhdr; 257 memcpy(mtod(m0, caddr_t), (caddr_t)ip, 258 sizeof(struct ip)); 259 m = m0; 260 } else { /* we have some space left in the old one */ 261 m->m_data -= msiz; 262 m->m_len += msiz; 263 m->m_pkthdr.len += msiz; 264 memmove(mtod(m, caddr_t), ip, 265 sizeof(struct ip)); 266 } 267 ip = mtod(m, struct ip *); 268 memcpy((caddr_t)(ip + 1), &mob_h, (unsigned)msiz); 269 ip->ip_len = htons(ntohs(ip->ip_len) + msiz); 270 } else { /* AF_INET */ 271 IF_DROP(&ifp->if_snd); 272 m_freem(m); 273 error = EINVAL; 274 goto end; 275 } 276 } else if (sc->g_proto == IPPROTO_GRE) { 277 switch (dst->sa_family) { 278 case AF_INET: 279 ip = mtod(m, struct ip *); 280 etype = ETHERTYPE_IP; 281 break; 282 #ifdef NETATALK 283 case AF_APPLETALK: 284 etype = ETHERTYPE_ATALK; 285 break; 286 #endif 287 #ifdef NS 288 case AF_NS: 289 etype = ETHERTYPE_NS; 290 break; 291 #endif 292 default: 293 IF_DROP(&ifp->if_snd); 294 m_freem(m); 295 error = EAFNOSUPPORT; 296 goto end; 297 } 298 M_PREPEND(m, sizeof(struct greip), M_DONTWAIT); 299 } else { 300 IF_DROP(&ifp->if_snd); 301 m_freem(m); 302 error = EINVAL; 303 goto end; 304 } 305 306 if (m == NULL) { /* impossible */ 307 IF_DROP(&ifp->if_snd); 308 error = ENOBUFS; 309 goto end; 310 } 311 312 gh = mtod(m, struct greip *); 313 if (sc->g_proto == IPPROTO_GRE) { 314 /* we don't have any GRE flags for now */ 315 316 memset((void *)&gh->gi_g, 0, sizeof(struct gre_h)); 317 gh->gi_ptype = htons(etype); 318 } 319 320 gh->gi_pr = sc->g_proto; 321 if (sc->g_proto != IPPROTO_MOBILE) { 322 gh->gi_src = sc->g_src; 323 gh->gi_dst = sc->g_dst; 324 ((struct ip*)gh)->ip_hl = (sizeof(struct ip)) >> 2; 325 ((struct ip*)gh)->ip_ttl = ip_gre_ttl; 326 ((struct ip*)gh)->ip_tos = ip->ip_tos; 327 gh->gi_len = htons(m->m_pkthdr.len); 328 } 329 330 ifp->if_opackets++; 331 ifp->if_obytes += m->m_pkthdr.len; 332 /* send it off */ 333 error = ip_output(m, NULL, &sc->route, 0, 334 (struct ip_moptions *)NULL, (struct socket *)NULL); 335 end: 336 if (error) 337 ifp->if_oerrors++; 338 return (error); 339 } 340 341 int 342 gre_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 343 { 344 struct proc *p = curproc; /* XXX */ 345 struct ifreq *ifr = (struct ifreq *)data; 346 struct if_laddrreq *lifr = (struct if_laddrreq *)data; 347 struct gre_softc *sc = ifp->if_softc; 348 int s; 349 struct sockaddr_in si; 350 struct sockaddr *sa = NULL; 351 int error; 352 353 error = 0; 354 355 s = splnet(); 356 switch (cmd) { 357 case SIOCSIFADDR: 358 ifp->if_flags |= IFF_UP; 359 break; 360 case SIOCSIFDSTADDR: 361 break; 362 case SIOCSIFFLAGS: 363 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 364 break; 365 if ((ifr->ifr_flags & IFF_LINK0) != 0) 366 sc->g_proto = IPPROTO_GRE; 367 else 368 sc->g_proto = IPPROTO_MOBILE; 369 break; 370 case SIOCSIFMTU: 371 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 372 break; 373 if (ifr->ifr_mtu < 576) { 374 error = EINVAL; 375 break; 376 } 377 ifp->if_mtu = ifr->ifr_mtu; 378 break; 379 case SIOCGIFMTU: 380 ifr->ifr_mtu = sc->sc_if.if_mtu; 381 break; 382 case SIOCADDMULTI: 383 case SIOCDELMULTI: 384 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 385 break; 386 if (ifr == 0) { 387 error = EAFNOSUPPORT; 388 break; 389 } 390 switch (ifr->ifr_addr.sa_family) { 391 #ifdef INET 392 case AF_INET: 393 break; 394 #endif 395 default: 396 error = EAFNOSUPPORT; 397 break; 398 } 399 break; 400 case GRESPROTO: 401 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 402 break; 403 sc->g_proto = ifr->ifr_flags; 404 switch (sc->g_proto) { 405 case IPPROTO_GRE: 406 ifp->if_flags |= IFF_LINK0; 407 break; 408 case IPPROTO_MOBILE: 409 ifp->if_flags &= ~IFF_LINK0; 410 break; 411 default: 412 error = EPROTONOSUPPORT; 413 break; 414 } 415 break; 416 case GREGPROTO: 417 ifr->ifr_flags = sc->g_proto; 418 break; 419 case GRESADDRS: 420 case GRESADDRD: 421 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 422 break; 423 /* 424 * set tunnel endpoints, compute a less specific route 425 * to the remote end and mark if as up 426 */ 427 sa = &ifr->ifr_addr; 428 if (cmd == GRESADDRS) 429 sc->g_src = (satosin(sa))->sin_addr; 430 if (cmd == GRESADDRD) 431 sc->g_dst = (satosin(sa))->sin_addr; 432 recompute: 433 if ((sc->g_src.s_addr != INADDR_ANY) && 434 (sc->g_dst.s_addr != INADDR_ANY)) { 435 if (sc->route.ro_rt != 0) /* free old route */ 436 RTFREE(sc->route.ro_rt); 437 if (gre_compute_route(sc) == 0) 438 ifp->if_flags |= IFF_RUNNING; 439 else 440 ifp->if_flags &= ~IFF_RUNNING; 441 } 442 break; 443 case GREGADDRS: 444 memset(&si, 0, sizeof(si)); 445 si.sin_family = AF_INET; 446 si.sin_len = sizeof(struct sockaddr_in); 447 si.sin_addr.s_addr = sc->g_src.s_addr; 448 sa = sintosa(&si); 449 ifr->ifr_addr = *sa; 450 break; 451 case GREGADDRD: 452 memset(&si, 0, sizeof(si)); 453 si.sin_family = AF_INET; 454 si.sin_len = sizeof(struct sockaddr_in); 455 si.sin_addr.s_addr = sc->g_dst.s_addr; 456 sa = sintosa(&si); 457 ifr->ifr_addr = *sa; 458 break; 459 case SIOCSLIFPHYADDR: 460 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 461 break; 462 if (lifr->addr.ss_family != AF_INET || 463 lifr->dstaddr.ss_family != AF_INET) { 464 error = EAFNOSUPPORT; 465 break; 466 } 467 if (lifr->addr.ss_len != sizeof(si) || 468 lifr->dstaddr.ss_len != sizeof(si)) { 469 error = EINVAL; 470 break; 471 } 472 sc->g_src = (satosin((struct sockadrr *)&lifr->addr))->sin_addr; 473 sc->g_dst = 474 (satosin((struct sockadrr *)&lifr->dstaddr))->sin_addr; 475 goto recompute; 476 case SIOCDIFPHYADDR: 477 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 478 break; 479 sc->g_src.s_addr = INADDR_ANY; 480 sc->g_dst.s_addr = INADDR_ANY; 481 break; 482 case SIOCGLIFPHYADDR: 483 if (sc->g_src.s_addr == INADDR_ANY || 484 sc->g_dst.s_addr == INADDR_ANY) { 485 error = EADDRNOTAVAIL; 486 break; 487 } 488 memset(&si, 0, sizeof(si)); 489 si.sin_family = AF_INET; 490 si.sin_len = sizeof(struct sockaddr_in); 491 si.sin_addr.s_addr = sc->g_src.s_addr; 492 memcpy(&lifr->addr, &si, sizeof(si)); 493 si.sin_addr.s_addr = sc->g_dst.s_addr; 494 memcpy(&lifr->dstaddr, &si, sizeof(si)); 495 break; 496 default: 497 error = EINVAL; 498 break; 499 } 500 501 splx(s); 502 return (error); 503 } 504 505 /* 506 * computes a route to our destination that is not the one 507 * which would be taken by ip_output(), as this one will loop back to 508 * us. If the interface is p2p as a--->b, then a routing entry exists 509 * If we now send a packet to b (e.g. ping b), this will come down here 510 * gets src=a, dst=b tacked on and would from ip_output() sent back to 511 * if_gre. 512 * Goal here is to compute a route to b that is less specific than 513 * a-->b. We know that this one exists as in normal operation we have 514 * at least a default route which matches. 515 */ 516 int 517 gre_compute_route(struct gre_softc *sc) 518 { 519 struct route *ro; 520 u_int32_t a, b, c; 521 522 ro = &sc->route; 523 524 memset(ro, 0, sizeof(struct route)); 525 ((struct sockaddr_in *)&ro->ro_dst)->sin_addr = sc->g_dst; 526 ro->ro_dst.sa_family = AF_INET; 527 ro->ro_dst.sa_len = sizeof(ro->ro_dst); 528 529 /* 530 * toggle last bit, so our interface is not found, but a less 531 * specific route. I'd rather like to specify a shorter mask, 532 * but this is not possible. Should work though. XXX 533 * there is a simpler way ... 534 */ 535 if ((sc->sc_if.if_flags & IFF_LINK1) == 0) { 536 a = ntohl(sc->g_dst.s_addr); 537 b = a & 0x01; 538 c = a & 0xfffffffe; 539 b = b ^ 0x01; 540 a = b | c; 541 ((struct sockaddr_in *)&ro->ro_dst)->sin_addr.s_addr 542 = htonl(a); 543 } 544 545 #ifdef DIAGNOSTIC 546 printf("%s: searching for a route to %s", sc->sc_if.if_xname, 547 inet_ntoa(((struct sockaddr_in *)&ro->ro_dst)->sin_addr)); 548 #endif 549 550 rtalloc(ro); 551 552 /* 553 * check if this returned a route at all and this route is no 554 * recursion to ourself 555 */ 556 if (ro->ro_rt == NULL || ro->ro_rt->rt_ifp->if_softc == sc) { 557 #ifdef DIAGNOSTIC 558 if (ro->ro_rt == NULL) 559 printf(" - no route found!\n"); 560 else 561 printf(" - route loops back to ourself!\n"); 562 #endif 563 return EADDRNOTAVAIL; 564 } 565 566 /* 567 * now change it back - else ip_output will just drop 568 * the route and search one to this interface ... 569 */ 570 if ((sc->sc_if.if_flags & IFF_LINK1) == 0) 571 ((struct sockaddr_in *)&ro->ro_dst)->sin_addr = sc->g_dst; 572 573 #ifdef DIAGNOSTIC 574 printf(", choosing %s with gateway %s", ro->ro_rt->rt_ifp->if_xname, 575 inet_ntoa(((struct sockaddr_in *)(ro->ro_rt->rt_gateway))->sin_addr)); 576 printf("\n"); 577 #endif 578 579 return 0; 580 } 581 582 /* 583 * do a checksum of a buffer - much like in_cksum, which operates on 584 * mbufs. 585 */ 586 u_int16_t 587 gre_in_cksum(u_int16_t *p, u_int len) 588 { 589 u_int32_t sum = 0; 590 int nwords = len >> 1; 591 592 while (nwords-- != 0) 593 sum += *p++; 594 595 if (len & 1) { 596 union { 597 u_short w; 598 u_char c[2]; 599 } u; 600 u.c[0] = *(u_char *)p; 601 u.c[1] = 0; 602 sum += u.w; 603 } 604 605 /* end-around-carry */ 606 sum = (sum >> 16) + (sum & 0xffff); 607 sum += (sum >> 16); 608 return (~sum); 609 } 610
Indexes created Fri Sep 26 08:10:20 GMT 2025