1 /* $OpenBSD: if_mpe.c,v 1.100 2021/03/26 19:00:21 kn Exp $ */ 2 3 /* 4 * Copyright (c) 2008 Pierre-Yves Ritschard <pyr@spootnik.org> 5 * 6 * Permission to use, copy, modify, and distribute this software for any 7 * purpose with or without fee is hereby granted, provided that the above 8 * copyright notice and this permission notice appear in all copies. 9 * 10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 17 */ 18 19 #include <sys/param.h> 20 #include <sys/systm.h> 21 #include <sys/mbuf.h> 22 #include <sys/socket.h> 23 #include <sys/sockio.h> 24 #include <sys/ioctl.h> 25 26 #include <net/if.h> 27 #include <net/if_dl.h> 28 #include <net/if_var.h> 29 #include <net/if_types.h> 30 #include <net/netisr.h> 31 #include <net/route.h> 32 33 #include <netinet/in.h> 34 #include <netinet/ip.h> 35 36 #ifdef INET6 37 #include <netinet/ip6.h> 38 #endif /* INET6 */ 39 40 #include "bpfilter.h" 41 #if NBPFILTER > 0 42 #include <net/bpf.h> 43 #endif 44 45 #include <netmpls/mpls.h> 46 47 48 49 #ifdef MPLS_DEBUG 50 #define DPRINTF(x) do { if (mpedebug) printf x ; } while (0) 51 #else 52 #define DPRINTF(x) 53 #endif 54 55 struct mpe_softc { 56 struct ifnet sc_if; /* the interface */ 57 int sc_txhprio; 58 int sc_rxhprio; 59 unsigned int sc_rdomain; 60 struct ifaddr sc_ifa; 61 struct sockaddr_mpls sc_smpls; 62 63 int sc_dead; 64 }; 65 66 #define MPE_HDRLEN sizeof(struct shim_hdr) 67 #define MPE_MTU 1500 68 #define MPE_MTU_MIN 256 69 #define MPE_MTU_MAX 8192 70 71 void mpeattach(int); 72 int mpe_output(struct ifnet *, struct mbuf *, struct sockaddr *, 73 struct rtentry *); 74 int mpe_ioctl(struct ifnet *, u_long, caddr_t); 75 void mpe_start(struct ifnet *); 76 int mpe_clone_create(struct if_clone *, int); 77 int mpe_clone_destroy(struct ifnet *); 78 void mpe_input(struct ifnet *, struct mbuf *); 79 80 struct if_clone mpe_cloner = 81 IF_CLONE_INITIALIZER("mpe", mpe_clone_create, mpe_clone_destroy); 82 83 extern int mpls_mapttl_ip; 84 #ifdef INET6 85 extern int mpls_mapttl_ip6; 86 #endif 87 88 void 89 mpeattach(int nmpe) 90 { 91 if_clone_attach(&mpe_cloner); 92 } 93 94 int 95 mpe_clone_create(struct if_clone *ifc, int unit) 96 { 97 struct mpe_softc *sc; 98 struct ifnet *ifp; 99 100 sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_CANFAIL|M_ZERO); 101 if (sc == NULL) 102 return (ENOMEM); 103 104 ifp = &sc->sc_if; 105 snprintf(ifp->if_xname, sizeof ifp->if_xname, "mpe%d", unit); 106 ifp->if_flags = IFF_POINTOPOINT; 107 ifp->if_xflags = IFXF_CLONED; 108 ifp->if_softc = sc; 109 ifp->if_mtu = MPE_MTU; 110 ifp->if_ioctl = mpe_ioctl; 111 ifp->if_bpf_mtap = p2p_bpf_mtap; 112 ifp->if_input = p2p_input; 113 ifp->if_output = mpe_output; 114 ifp->if_start = mpe_start; 115 ifp->if_type = IFT_MPLS; 116 ifp->if_hdrlen = MPE_HDRLEN; 117 118 sc->sc_dead = 0; 119 120 if_attach(ifp); 121 if_alloc_sadl(ifp); 122 if_counters_alloc(ifp); 123 124 #if NBPFILTER > 0 125 bpfattach(&ifp->if_bpf, ifp, DLT_LOOP, sizeof(u_int32_t)); 126 #endif 127 128 sc->sc_txhprio = 0; 129 sc->sc_rxhprio = IF_HDRPRIO_PACKET; 130 sc->sc_rdomain = 0; 131 sc->sc_ifa.ifa_ifp = ifp; 132 sc->sc_ifa.ifa_addr = sdltosa(ifp->if_sadl); 133 sc->sc_smpls.smpls_len = sizeof(sc->sc_smpls); 134 sc->sc_smpls.smpls_family = AF_MPLS; 135 136 return (0); 137 } 138 139 int 140 mpe_clone_destroy(struct ifnet *ifp) 141 { 142 struct mpe_softc *sc = ifp->if_softc; 143 144 NET_LOCK(); 145 CLR(ifp->if_flags, IFF_RUNNING); 146 sc->sc_dead = 1; 147 148 if (sc->sc_smpls.smpls_label) { 149 rt_ifa_del(&sc->sc_ifa, RTF_MPLS|RTF_LOCAL, 150 smplstosa(&sc->sc_smpls), sc->sc_rdomain); 151 } 152 NET_UNLOCK(); 153 154 ifq_barrier(&ifp->if_snd); 155 156 if_detach(ifp); 157 free(sc, M_DEVBUF, sizeof *sc); 158 return (0); 159 } 160 161 /* 162 * Start output on the mpe interface. 163 */ 164 void 165 mpe_start(struct ifnet *ifp) 166 { 167 struct mpe_softc *sc = ifp->if_softc; 168 struct mbuf *m; 169 struct sockaddr *sa; 170 struct sockaddr smpls = { .sa_family = AF_MPLS }; 171 struct rtentry *rt; 172 struct ifnet *ifp0; 173 174 while ((m = ifq_dequeue(&ifp->if_snd)) != NULL) { 175 sa = mtod(m, struct sockaddr *); 176 rt = rtalloc(sa, RT_RESOLVE, sc->sc_rdomain); 177 if (!rtisvalid(rt)) { 178 m_freem(m); 179 rtfree(rt); 180 continue; 181 } 182 183 ifp0 = if_get(rt->rt_ifidx); 184 if (ifp0 == NULL) { 185 m_freem(m); 186 rtfree(rt); 187 continue; 188 } 189 190 m_adj(m, sa->sa_len); 191 192 #if NBPFILTER > 0 193 if (ifp->if_bpf) { 194 /* remove MPLS label before passing packet to bpf */ 195 m->m_data += sizeof(struct shim_hdr); 196 m->m_len -= sizeof(struct shim_hdr); 197 m->m_pkthdr.len -= sizeof(struct shim_hdr); 198 bpf_mtap_af(ifp->if_bpf, m->m_pkthdr.ph_family, 199 m, BPF_DIRECTION_OUT); 200 m->m_data -= sizeof(struct shim_hdr); 201 m->m_len += sizeof(struct shim_hdr); 202 m->m_pkthdr.len += sizeof(struct shim_hdr); 203 } 204 #endif 205 206 m->m_pkthdr.ph_rtableid = sc->sc_rdomain; 207 CLR(m->m_flags, M_BCAST|M_MCAST); 208 209 mpls_output(ifp0, m, &smpls, rt); 210 if_put(ifp0); 211 rtfree(rt); 212 } 213 } 214 215 int 216 mpe_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 217 struct rtentry *rt) 218 { 219 struct mpe_softc *sc; 220 struct rt_mpls *rtmpls; 221 struct shim_hdr shim; 222 int error; 223 int txprio; 224 uint8_t ttl = mpls_defttl; 225 uint8_t tos, prio; 226 size_t ttloff; 227 socklen_t slen; 228 229 if (!rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_MPLS)) { 230 m_freem(m); 231 return (ENETUNREACH); 232 } 233 234 if (dst->sa_family == AF_LINK && ISSET(rt->rt_flags, RTF_LOCAL)) { 235 mpe_input(ifp, m); 236 return (0); 237 } 238 239 #ifdef DIAGNOSTIC 240 if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid)) { 241 printf("%s: trying to send packet on wrong domain. " 242 "if %d vs. mbuf %d\n", ifp->if_xname, 243 ifp->if_rdomain, rtable_l2(m->m_pkthdr.ph_rtableid)); 244 } 245 #endif 246 247 rtmpls = (struct rt_mpls *)rt->rt_llinfo; 248 if (rtmpls->mpls_operation != MPLS_OP_PUSH) { 249 m_freem(m); 250 return (ENETUNREACH); 251 } 252 253 error = 0; 254 switch (dst->sa_family) { 255 case AF_INET: { 256 struct ip *ip = mtod(m, struct ip *); 257 tos = ip->ip_tos; 258 ttloff = offsetof(struct ip, ip_ttl); 259 slen = sizeof(struct sockaddr_in); 260 break; 261 } 262 #ifdef INET6 263 case AF_INET6: { 264 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 265 uint32_t flow = bemtoh32(&ip6->ip6_flow); 266 tos = flow >> 20; 267 ttloff = offsetof(struct ip6_hdr, ip6_hlim); 268 slen = sizeof(struct sockaddr_in6); 269 break; 270 } 271 #endif 272 default: 273 m_freem(m); 274 return (EPFNOSUPPORT); 275 } 276 277 if (mpls_mapttl_ip) { 278 /* assumes the ip header is already contig */ 279 ttl = *(mtod(m, uint8_t *) + ttloff); 280 } 281 282 sc = ifp->if_softc; 283 txprio = sc->sc_txhprio; 284 285 switch (txprio) { 286 case IF_HDRPRIO_PACKET: 287 prio = m->m_pkthdr.pf.prio; 288 break; 289 case IF_HDRPRIO_PAYLOAD: 290 prio = IFQ_TOS2PRIO(tos); 291 break; 292 default: 293 prio = txprio; 294 break; 295 } 296 297 shim.shim_label = rtmpls->mpls_label | htonl(prio << MPLS_EXP_OFFSET) | 298 MPLS_BOS_MASK | htonl(ttl); 299 300 m = m_prepend(m, sizeof(shim), M_NOWAIT); 301 if (m == NULL) { 302 error = ENOMEM; 303 goto out; 304 } 305 *mtod(m, struct shim_hdr *) = shim; 306 307 m = m_prepend(m, slen, M_WAITOK); 308 if (m == NULL) { 309 error = ENOMEM; 310 goto out; 311 } 312 memcpy(mtod(m, struct sockaddr *), rt->rt_gateway, slen); 313 mtod(m, struct sockaddr *)->sa_len = slen; /* to be sure */ 314 315 m->m_pkthdr.ph_family = dst->sa_family; 316 317 error = if_enqueue(ifp, m); 318 out: 319 if (error) 320 ifp->if_oerrors++; 321 return (error); 322 } 323 324 int 325 mpe_set_label(struct mpe_softc *sc, uint32_t label, unsigned int rdomain) 326 { 327 int error; 328 329 if (sc->sc_dead) 330 return (ENXIO); 331 332 if (sc->sc_smpls.smpls_label) { 333 /* remove old MPLS route */ 334 rt_ifa_del(&sc->sc_ifa, RTF_MPLS|RTF_LOCAL, 335 smplstosa(&sc->sc_smpls), sc->sc_rdomain); 336 } 337 338 /* add new MPLS route */ 339 sc->sc_smpls.smpls_label = label; 340 sc->sc_rdomain = rdomain; 341 342 /* only install with a label or mpe_clone_destroy() will ignore it */ 343 if (sc->sc_smpls.smpls_label == MPLS_LABEL2SHIM(0)) 344 return 0; 345 346 error = rt_ifa_add(&sc->sc_ifa, RTF_MPLS|RTF_LOCAL, 347 smplstosa(&sc->sc_smpls), sc->sc_rdomain); 348 if (error) 349 sc->sc_smpls.smpls_label = 0; 350 351 return (error); 352 } 353 354 int 355 mpe_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 356 { 357 struct mpe_softc *sc = ifp->if_softc; 358 struct ifreq *ifr; 359 struct shim_hdr shim; 360 int error = 0; 361 362 ifr = (struct ifreq *)data; 363 switch (cmd) { 364 case SIOCSIFADDR: 365 break; 366 case SIOCSIFFLAGS: 367 if (ifp->if_flags & IFF_UP) 368 ifp->if_flags |= IFF_RUNNING; 369 else 370 ifp->if_flags &= ~IFF_RUNNING; 371 break; 372 case SIOCSIFMTU: 373 if (ifr->ifr_mtu < MPE_MTU_MIN || 374 ifr->ifr_mtu > MPE_MTU_MAX) 375 error = EINVAL; 376 else 377 ifp->if_mtu = ifr->ifr_mtu; 378 break; 379 case SIOCGETLABEL: 380 shim.shim_label = MPLS_SHIM2LABEL(sc->sc_smpls.smpls_label); 381 if (shim.shim_label == 0) { 382 error = EADDRNOTAVAIL; 383 break; 384 } 385 error = copyout(&shim, ifr->ifr_data, sizeof(shim)); 386 break; 387 case SIOCSETLABEL: 388 error = copyin(ifr->ifr_data, &shim, sizeof(shim)); 389 if (error != 0) 390 break; 391 if (shim.shim_label > MPLS_LABEL_MAX || 392 shim.shim_label <= MPLS_LABEL_RESERVED_MAX) { 393 error = EINVAL; 394 break; 395 } 396 shim.shim_label = MPLS_LABEL2SHIM(shim.shim_label); 397 if (sc->sc_smpls.smpls_label != shim.shim_label) { 398 error = mpe_set_label(sc, shim.shim_label, 399 sc->sc_rdomain); 400 } 401 break; 402 case SIOCDELLABEL: 403 if (sc->sc_smpls.smpls_label != MPLS_LABEL2SHIM(0)) { 404 rt_ifa_del(&sc->sc_ifa, RTF_MPLS|RTF_LOCAL, 405 smplstosa(&sc->sc_smpls), sc->sc_rdomain); 406 407 } 408 sc->sc_smpls.smpls_label = MPLS_LABEL2SHIM(0); 409 break; 410 411 case SIOCSLIFPHYRTABLE: 412 if (ifr->ifr_rdomainid < 0 || 413 ifr->ifr_rdomainid > RT_TABLEID_MAX || 414 !rtable_exists(ifr->ifr_rdomainid) || 415 ifr->ifr_rdomainid != rtable_l2(ifr->ifr_rdomainid)) { 416 error = EINVAL; 417 break; 418 } 419 if (sc->sc_rdomain != ifr->ifr_rdomainid) { 420 error = mpe_set_label(sc, sc->sc_smpls.smpls_label, 421 ifr->ifr_rdomainid); 422 } 423 break; 424 case SIOCGLIFPHYRTABLE: 425 ifr->ifr_rdomainid = sc->sc_rdomain; 426 break; 427 428 case SIOCSTXHPRIO: 429 error = if_txhprio_l3_check(ifr->ifr_hdrprio); 430 if (error != 0) 431 break; 432 433 sc->sc_txhprio = ifr->ifr_hdrprio; 434 break; 435 case SIOCGTXHPRIO: 436 ifr->ifr_hdrprio = sc->sc_txhprio; 437 break; 438 439 case SIOCSRXHPRIO: 440 error = if_rxhprio_l3_check(ifr->ifr_hdrprio); 441 if (error != 0) 442 break; 443 444 sc->sc_rxhprio = ifr->ifr_hdrprio; 445 break; 446 case SIOCGRXHPRIO: 447 ifr->ifr_hdrprio = sc->sc_rxhprio; 448 break; 449 450 default: 451 return (ENOTTY); 452 } 453 454 return (error); 455 } 456 457 void 458 mpe_input(struct ifnet *ifp, struct mbuf *m) 459 { 460 struct mpe_softc *sc = ifp->if_softc; 461 struct shim_hdr *shim; 462 struct mbuf *n; 463 uint8_t ttl, tos; 464 uint32_t exp; 465 int rxprio = sc->sc_rxhprio; 466 467 shim = mtod(m, struct shim_hdr *); 468 exp = ntohl(shim->shim_label & MPLS_EXP_MASK) >> MPLS_EXP_OFFSET; 469 if (!MPLS_BOS_ISSET(shim->shim_label)) 470 goto drop; 471 472 ttl = ntohl(shim->shim_label & MPLS_TTL_MASK); 473 m_adj(m, sizeof(*shim)); 474 475 n = m; 476 while (n->m_len == 0) { 477 n = n->m_next; 478 if (n == NULL) 479 goto drop; 480 } 481 482 switch (*mtod(n, uint8_t *) >> 4) { 483 case 4: { 484 struct ip *ip; 485 if (m->m_len < sizeof(*ip)) { 486 m = m_pullup(m, sizeof(*ip)); 487 if (m == NULL) 488 return; 489 } 490 ip = mtod(m, struct ip *); 491 tos = ip->ip_tos; 492 493 if (mpls_mapttl_ip) { 494 m = mpls_ip_adjttl(m, ttl); 495 if (m == NULL) 496 return; 497 } 498 499 m->m_pkthdr.ph_family = AF_INET; 500 break; 501 } 502 #ifdef INET6 503 case 6: { 504 struct ip6_hdr *ip6; 505 uint32_t flow; 506 if (m->m_len < sizeof(*ip6)) { 507 m = m_pullup(m, sizeof(*ip6)); 508 if (m == NULL) 509 return; 510 } 511 ip6 = mtod(m, struct ip6_hdr *); 512 flow = bemtoh32(&ip6->ip6_flow); 513 tos = flow >> 20; 514 515 if (mpls_mapttl_ip6) { 516 m = mpls_ip6_adjttl(m, ttl); 517 if (m == NULL) 518 return; 519 } 520 521 m->m_pkthdr.ph_family = AF_INET6; 522 break; 523 } 524 #endif /* INET6 */ 525 default: 526 goto drop; 527 } 528 529 switch (rxprio) { 530 case IF_HDRPRIO_PACKET: 531 /* nop */ 532 break; 533 case IF_HDRPRIO_OUTER: 534 m->m_pkthdr.pf.prio = exp; 535 break; 536 case IF_HDRPRIO_PAYLOAD: 537 m->m_pkthdr.pf.prio = IFQ_TOS2PRIO(tos); 538 break; 539 default: 540 m->m_pkthdr.pf.prio = rxprio; 541 break; 542 } 543 544 if_vinput(ifp, m); 545 return; 546 drop: 547 m_freem(m); 548 } 549