1 /* 2 * Copyright (c) 1982, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 34 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $ 35 * $DragonFly: src/sys/net/if_ethersubr.c,v 1.29 2005/03/12 02:42:28 joerg Exp $ 36 */ 37 38 #include "opt_atalk.h" 39 #include "opt_inet.h" 40 #include "opt_inet6.h" 41 #include "opt_ipx.h" 42 #include "opt_bdg.h" 43 #include "opt_netgraph.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/kernel.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/socket.h> 51 #include <sys/sockio.h> 52 #include <sys/sysctl.h> 53 54 #include <net/if.h> 55 #include <net/netisr.h> 56 #include <net/route.h> 57 #include <net/if_llc.h> 58 #include <net/if_dl.h> 59 #include <net/if_types.h> 60 #include <net/ifq_var.h> 61 #include <net/bpf.h> 62 #include <net/ethernet.h> 63 #include <net/bridge/bridge.h> 64 65 #if defined(INET) || defined(INET6) 66 #include <netinet/in.h> 67 #include <netinet/in_var.h> 68 #include <netinet/if_ether.h> 69 #include <net/ipfw/ip_fw.h> 70 #include <net/dummynet/ip_dummynet.h> 71 #endif 72 #ifdef INET6 73 #include <netinet6/nd6.h> 74 #endif 75 76 #ifdef IPX 77 #include <netproto/ipx/ipx.h> 78 #include <netproto/ipx/ipx_if.h> 79 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 80 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 81 short *tp, int *hlen); 82 #endif 83 84 #ifdef NS 85 #include <netns/ns.h> 86 #include <netns/ns_if.h> 87 ushort ns_nettype; 88 int ether_outputdebug = 0; 89 int ether_inputdebug = 0; 90 #endif 91 92 #ifdef NETATALK 93 #include <netproto/atalk/at.h> 94 #include <netproto/atalk/at_var.h> 95 #include <netproto/atalk/at_extern.h> 96 97 #define llc_snap_org_code llc_un.type_snap.org_code 98 #define llc_snap_ether_type llc_un.type_snap.ether_type 99 100 extern u_char at_org_code[3]; 101 extern u_char aarp_org_code[3]; 102 #endif /* NETATALK */ 103 104 /* netgraph node hooks for ng_ether(4) */ 105 void (*ng_ether_input_p)(struct ifnet *ifp, 106 struct mbuf **mp, struct ether_header *eh); 107 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 108 struct mbuf *m, struct ether_header *eh); 109 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 110 void (*ng_ether_attach_p)(struct ifnet *ifp); 111 void (*ng_ether_detach_p)(struct ifnet *ifp); 112 113 int (*vlan_input_p)(struct ether_header *eh, struct mbuf *m); 114 int (*vlan_input_tag_p)(struct mbuf *m, uint16_t t); 115 116 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 117 struct rtentry *); 118 119 /* 120 * bridge support 121 */ 122 int do_bridge; 123 bridge_in_t *bridge_in_ptr; 124 bdg_forward_t *bdg_forward_ptr; 125 bdgtakeifaces_t *bdgtakeifaces_ptr; 126 struct bdg_softc *ifp2sc; 127 128 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 129 struct sockaddr *); 130 131 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 132 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 133 }; 134 135 #define gotoerr(e) do { error = (e); goto bad; } while (0) 136 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 137 138 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 139 struct ip_fw **rule, struct ether_header *eh, 140 boolean_t shared); 141 142 static int ether_ipfw; 143 SYSCTL_DECL(_net_link); 144 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 145 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 146 ðer_ipfw, 0, "Pass ether pkts through firewall"); 147 148 /* 149 * Ethernet output routine. 150 * Encapsulate a packet of type family for the local net. 151 * Use trailer local net encapsulation if enough data in first 152 * packet leaves a multiple of 512 bytes of data in remainder. 153 * Assumes that ifp is actually pointer to arpcom structure. 154 */ 155 static int 156 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 157 struct rtentry *rt) 158 { 159 struct ether_header *eh, *deh; 160 u_char *edst; 161 int loop_copy = 0; 162 int hlen = ETHER_HDR_LEN; /* link layer header length */ 163 struct arpcom *ac = IFP2AC(ifp); 164 int error; 165 166 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 167 gotoerr(ENETDOWN); 168 169 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 170 if (m == NULL) 171 return (ENOBUFS); 172 eh = mtod(m, struct ether_header *); 173 edst = eh->ether_dhost; 174 175 /* 176 * Fill in the destination ethernet address and frame type. 177 */ 178 switch (dst->sa_family) { 179 #ifdef INET 180 case AF_INET: 181 if (!arpresolve(ifp, rt, m, dst, edst)) 182 return (0); /* if not yet resolved */ 183 eh->ether_type = htons(ETHERTYPE_IP); 184 break; 185 #endif 186 #ifdef INET6 187 case AF_INET6: 188 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 189 return (0); /* Something bad happenned. */ 190 eh->ether_type = htons(ETHERTYPE_IPV6); 191 break; 192 #endif 193 #ifdef IPX 194 case AF_IPX: 195 if (ef_outputp != NULL) { 196 error = ef_outputp(ifp, &m, dst, &eh->ether_type, 197 &hlen); 198 if (error) 199 goto bad; 200 } else { 201 eh->ether_type = htons(ETHERTYPE_IPX); 202 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 203 edst, ETHER_ADDR_LEN); 204 } 205 break; 206 #endif 207 #ifdef NETATALK 208 case AF_APPLETALK: { 209 struct at_ifaddr *aa; 210 211 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) { 212 error = 0; /* XXX */ 213 goto bad; 214 } 215 /* 216 * In the phase 2 case, need to prepend an mbuf for 217 * the llc header. Since we must preserve the value 218 * of m, which is passed to us by value, we m_copy() 219 * the first mbuf, and use it for our llc header. 220 */ 221 if (aa->aa_flags & AFA_PHASE2) { 222 struct llc llc; 223 224 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT); 225 eh = mtod(m, struct ether_header *); 226 edst = eh->ether_dhost; 227 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 228 llc.llc_control = LLC_UI; 229 bcopy(at_org_code, llc.llc_snap_org_code, 230 sizeof at_org_code); 231 llc.llc_snap_ether_type = htons(ETHERTYPE_AT); 232 bcopy(&llc, 233 mtod(m, caddr_t) + sizeof(struct ether_header), 234 sizeof(struct llc)); 235 eh->ether_type = htons(m->m_pkthdr.len); 236 hlen = sizeof(struct llc) + ETHER_HDR_LEN; 237 } else { 238 eh->ether_type = htons(ETHERTYPE_AT); 239 } 240 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) 241 return (0); 242 break; 243 } 244 #endif 245 #ifdef NS 246 case AF_NS: 247 switch(ns_nettype) { 248 default: 249 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 250 eh->ether_type = 0x8137; 251 break; 252 case 0x0: /* Novell 802.3 */ 253 eh->ether_type = htons(m->m_pkthdr.len); 254 break; 255 case 0xe0e0: /* Novell 802.2 and Token-Ring */ 256 M_PREPEND(m, 3, MB_DONTWAIT); 257 eh = mtod(m, struct ether_header *); 258 edst = eh->ether_dhost; 259 eh->ether_type = htons(m->m_pkthdr.len); 260 cp = mtod(m, u_char *) + sizeof(struct ether_header); 261 *cp++ = 0xE0; 262 *cp++ = 0xE0; 263 *cp++ = 0x03; 264 break; 265 } 266 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst, 267 ETHER_ADDR_LEN); 268 /* 269 * XXX if ns_thishost is the same as the node's ethernet 270 * address then just the default code will catch this anyhow. 271 * So I'm not sure if this next clause should be here at all? 272 * [JRE] 273 */ 274 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) { 275 m->m_pkthdr.rcvif = ifp; 276 netisr_dispatch(NETISR_NS, m); 277 return (error); 278 } 279 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0) 280 m->m_flags |= M_BCAST; 281 break; 282 #endif 283 case pseudo_AF_HDRCMPLT: 284 case AF_UNSPEC: 285 loop_copy = -1; /* if this is for us, don't do it */ 286 deh = (struct ether_header *)dst->sa_data; 287 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 288 eh->ether_type = deh->ether_type; 289 break; 290 291 default: 292 printf("%s: can't handle af%d\n", ifp->if_xname, 293 dst->sa_family); 294 gotoerr(EAFNOSUPPORT); 295 } 296 297 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 298 memcpy(eh->ether_shost, 299 ((struct ether_header *)dst->sa_data)->ether_shost, 300 ETHER_ADDR_LEN); 301 else 302 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 303 304 /* 305 * If a simplex interface, and the packet is being sent to our 306 * Ethernet address or a broadcast address, loopback a copy. 307 * XXX To make a simplex device behave exactly like a duplex 308 * device, we should copy in the case of sending to our own 309 * ethernet address (thus letting the original actually appear 310 * on the wire). However, we don't do that here for security 311 * reasons and compatibility with the original behavior. 312 */ 313 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 314 int csum_flags = 0; 315 316 if (m->m_pkthdr.csum_flags & CSUM_IP) 317 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 318 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 319 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 320 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 321 struct mbuf *n; 322 323 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 324 n->m_pkthdr.csum_flags |= csum_flags; 325 if (csum_flags & CSUM_DATA_VALID) 326 n->m_pkthdr.csum_data = 0xffff; 327 if_simloop(ifp, n, dst->sa_family, hlen); 328 } else 329 ifp->if_iqdrops++; 330 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 331 ETHER_ADDR_LEN) == 0) { 332 m->m_pkthdr.csum_flags |= csum_flags; 333 if (csum_flags & CSUM_DATA_VALID) 334 m->m_pkthdr.csum_data = 0xffff; 335 if_simloop(ifp, m, dst->sa_family, hlen); 336 return (0); /* XXX */ 337 } 338 } 339 340 /* Handle ng_ether(4) processing, if any */ 341 if (ng_ether_output_p != NULL) { 342 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) 343 goto bad; 344 if (m == NULL) 345 return (0); 346 } 347 348 /* Continue with link-layer output */ 349 return ether_output_frame(ifp, m); 350 351 bad: 352 m_freem(m); 353 return (error); 354 } 355 356 /* 357 * Ethernet link layer output routine to send a raw frame to the device. 358 * 359 * This assumes that the 14 byte Ethernet header is present and contiguous 360 * in the first mbuf (if BRIDGE'ing). 361 */ 362 int 363 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 364 { 365 struct ip_fw *rule = NULL; 366 int error = 0; 367 int s; 368 struct altq_pktattr pktattr; 369 370 /* Extract info from dummynet tag, ignore others */ 371 while (m->m_type == MT_TAG) { 372 if (m->m_flags == PACKET_TAG_DUMMYNET) { 373 rule = ((struct dn_pkt *)m)->rule; 374 break; 375 } 376 m = m->m_next; 377 } 378 if (rule != NULL) /* packet was already bridged */ 379 goto no_bridge; 380 381 if (BDG_ACTIVE(ifp)) { 382 struct ether_header *eh; /* a pointer suffices */ 383 384 m->m_pkthdr.rcvif = NULL; 385 eh = mtod(m, struct ether_header *); 386 m_adj(m, ETHER_HDR_LEN); 387 m = bdg_forward_ptr(m, eh, ifp); 388 m_freem(m); 389 return (0); 390 } 391 392 no_bridge: 393 if (ifq_is_enabled(&ifp->if_snd)) 394 altq_etherclassify(&ifp->if_snd, m, &pktattr); 395 s = splimp(); 396 if (IPFW_LOADED && ether_ipfw != 0) { 397 struct ether_header save_eh, *eh; 398 399 eh = mtod(m, struct ether_header *); 400 save_eh = *eh; 401 m_adj(m, ETHER_HDR_LEN); 402 if (!ether_ipfw_chk(&m, ifp, &rule, eh, FALSE)) { 403 if (m != NULL) { 404 m_freem(m); 405 return ENOBUFS; /* pkt dropped */ 406 } else 407 return 0; /* consumed e.g. in a pipe */ 408 } 409 eh = mtod(m, struct ether_header *); 410 /* packet was ok, restore the ethernet header */ 411 if ((void *)(eh + 1) == (void *)m->m_data) { 412 m->m_data -= ETHER_HDR_LEN ; 413 m->m_len += ETHER_HDR_LEN ; 414 m->m_pkthdr.len += ETHER_HDR_LEN ; 415 } else { 416 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 417 if (m == NULL) /* nope... */ 418 return ENOBUFS; 419 bcopy(&save_eh, mtod(m, struct ether_header *), 420 ETHER_HDR_LEN); 421 } 422 } 423 424 /* 425 * Queue message on interface, update output statistics if 426 * successful, and start output if interface not yet active. 427 */ 428 error = ifq_handoff(ifp, m, &pktattr); 429 splx(s); 430 return (error); 431 } 432 433 /* 434 * ipfw processing for ethernet packets (in and out). 435 * The second parameter is NULL from ether_demux(), and ifp from 436 * ether_output_frame(). This section of code could be used from 437 * bridge.c as well as long as we use some extra info 438 * to distinguish that case from ether_output_frame(). 439 */ 440 static boolean_t 441 ether_ipfw_chk( 442 struct mbuf **m0, 443 struct ifnet *dst, 444 struct ip_fw **rule, 445 struct ether_header *eh, 446 boolean_t shared) 447 { 448 struct ether_header save_eh = *eh; /* might be a ptr in m */ 449 struct ip_fw_args args; 450 int i; 451 452 if (*rule != NULL && fw_one_pass) 453 return TRUE; /* dummynet packet, already partially processed */ 454 455 /* 456 * I need some amount of data to be contiguous, and in case others 457 * need the packet (shared==TRUE), it also better be in the first mbuf. 458 */ 459 i = min((*m0)->m_pkthdr.len, max_protohdr); 460 if (shared || (*m0)->m_len < i) { 461 *m0 = m_pullup(*m0, i); 462 if (*m0 == NULL) 463 return FALSE; 464 } 465 466 args.m = *m0; /* the packet we are looking at */ 467 args.oif = dst; /* destination, if any */ 468 args.divert_rule = 0; /* we do not support divert yet */ 469 args.rule = *rule; /* matching rule to restart */ 470 args.next_hop = NULL; /* we do not support forward yet */ 471 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 472 i = ip_fw_chk_ptr(&args); 473 *m0 = args.m; 474 *rule = args.rule; 475 476 if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */ 477 return FALSE; 478 479 if (i == 0) /* a PASS rule. */ 480 return TRUE; 481 482 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) { 483 /* 484 * Pass the pkt to dummynet, which consumes it. 485 * If shared, make a copy and keep the original. 486 */ 487 struct mbuf *m ; 488 489 if (shared) { 490 m = m_copypacket(*m0, MB_DONTWAIT); 491 if (m == NULL) 492 return FALSE; 493 } else { 494 m = *m0 ; /* pass the original to dummynet */ 495 *m0 = NULL ; /* and nothing back to the caller */ 496 } 497 /* 498 * Prepend the header, optimize for the common case of 499 * eh pointing into the mbuf. 500 */ 501 if ((void *)(eh + 1) == (void *)m->m_data) { 502 m->m_data -= ETHER_HDR_LEN ; 503 m->m_len += ETHER_HDR_LEN ; 504 m->m_pkthdr.len += ETHER_HDR_LEN ; 505 } else { 506 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 507 if (m == NULL) 508 return FALSE; 509 bcopy(&save_eh, mtod(m, struct ether_header *), 510 ETHER_HDR_LEN); 511 } 512 ip_dn_io_ptr(m, (i & 0xffff), 513 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 514 return FALSE; 515 } 516 /* 517 * XXX at some point add support for divert/forward actions. 518 * If none of the above matches, we have to drop the pkt. 519 */ 520 return FALSE; 521 } 522 523 /* 524 * XXX merge this function with ether_input. 525 */ 526 static void 527 ether_input_internal(struct ifnet *ifp, struct mbuf *m) 528 { 529 ether_input(ifp, NULL, m); 530 } 531 532 /* 533 * Process a received Ethernet packet. We have two different interfaces: 534 * one (conventional) assumes the packet in the mbuf, with the ethernet 535 * header provided separately in *eh. The second one (new) has everything 536 * in the mbuf, and we can tell it because eh == NULL. 537 * The caller MUST MAKE SURE that there are at least 538 * sizeof(struct ether_header) bytes in the first mbuf. 539 * 540 * This allows us to concentrate in one place a bunch of code which 541 * is replicated in all device drivers. Also, many functions called 542 * from ether_input() try to put the eh back into the mbuf, so we 543 * can later propagate the 'contiguous packet' interface to them, 544 * and handle the old interface just here. 545 * 546 * NOTA BENE: for many drivers "eh" is a pointer into the first mbuf or 547 * cluster, right before m_data. So be very careful when working on m, 548 * as you could destroy *eh !! 549 * 550 * First we perform any link layer operations, then continue 551 * to the upper layers with ether_demux(). 552 */ 553 void 554 ether_input(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m) 555 { 556 struct ether_header save_eh; 557 558 if (eh == NULL) { 559 if (m->m_len < sizeof(struct ether_header)) { 560 /* XXX error in the caller. */ 561 m_freem(m); 562 return; 563 } 564 m->m_pkthdr.rcvif = ifp; 565 eh = mtod(m, struct ether_header *); 566 m_adj(m, sizeof(struct ether_header)); 567 /* XXX */ 568 /* m->m_pkthdr.len = m->m_len; */ 569 } 570 571 if (ifp->if_bpf) 572 bpf_ptap(ifp->if_bpf, m, eh, ETHER_HDR_LEN); 573 574 ifp->if_ibytes += m->m_pkthdr.len + (sizeof *eh); 575 576 /* Handle ng_ether(4) processing, if any */ 577 if (ng_ether_input_p != NULL) { 578 (*ng_ether_input_p)(ifp, &m, eh); 579 if (m == NULL) 580 return; 581 } 582 583 /* Check for bridging mode */ 584 if (BDG_ACTIVE(ifp)) { 585 struct ifnet *bif; 586 587 /* Check with bridging code */ 588 if ((bif = bridge_in_ptr(ifp, eh)) == BDG_DROP) { 589 m_freem(m); 590 return; 591 } 592 if (bif != BDG_LOCAL) { 593 save_eh = *eh ; /* because it might change */ 594 m = bdg_forward_ptr(m, eh, bif); /* needs forwarding */ 595 /* 596 * Do not continue if bdg_forward_ptr() processed our 597 * packet (and cleared the mbuf pointer m) or if 598 * it dropped (m_free'd) the packet itself. 599 */ 600 if (m == NULL) { 601 if (bif == BDG_BCAST || bif == BDG_MCAST) 602 printf("bdg_forward drop MULTICAST PKT\n"); 603 return; 604 } 605 eh = &save_eh ; 606 } 607 if (bif == BDG_LOCAL || bif == BDG_BCAST || bif == BDG_MCAST) 608 goto recvLocal; /* receive locally */ 609 610 /* If not local and not multicast, just drop it */ 611 m_freem(m); 612 return; 613 } 614 615 recvLocal: 616 /* Continue with upper layer processing */ 617 ether_demux(ifp, eh, m); 618 } 619 620 /* 621 * Upper layer processing for a received Ethernet packet. 622 */ 623 void 624 ether_demux(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m) 625 { 626 int isr; 627 u_short ether_type; 628 struct ip_fw *rule = NULL; 629 #ifdef NETATALK 630 struct llc *l; 631 #endif 632 633 /* Extract info from dummynet tag, ignore others */ 634 while (m->m_type == MT_TAG) { 635 if (m->m_flags == PACKET_TAG_DUMMYNET) { 636 rule = ((struct dn_pkt *)m)->rule; 637 ifp = m->m_next->m_pkthdr.rcvif; 638 break; 639 } 640 m = m->m_next; 641 } 642 if (rule) /* packet was already bridged */ 643 goto post_stats; 644 645 /* 646 * Discard packet if upper layers shouldn't see it because 647 * it was unicast to a different Ethernet address. If the 648 * driver is working properly, then this situation can only 649 * happen when the interface is in promiscuous mode. 650 */ 651 if (!BDG_ACTIVE(ifp) && 652 ((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 653 (eh->ether_dhost[0] & 1) == 0 && 654 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 655 m_freem(m); 656 return; 657 } 658 /* Discard packet if interface is not up */ 659 if (!(ifp->if_flags & IFF_UP)) { 660 m_freem(m); 661 return; 662 } 663 if (eh->ether_dhost[0] & 1) { 664 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 665 ifp->if_addrlen) == 0) 666 m->m_flags |= M_BCAST; 667 else 668 m->m_flags |= M_MCAST; 669 ifp->if_imcasts++; 670 } 671 672 post_stats: 673 if (IPFW_LOADED && ether_ipfw != 0) { 674 if (!ether_ipfw_chk(&m, NULL, &rule, eh, FALSE)) { 675 m_freem(m); 676 return; 677 } 678 eh = mtod(m, struct ether_header *); 679 } 680 681 ether_type = ntohs(eh->ether_type); 682 683 switch (ether_type) { 684 #ifdef INET 685 case ETHERTYPE_IP: 686 if (ipflow_fastforward(m)) 687 return; 688 isr = NETISR_IP; 689 break; 690 691 case ETHERTYPE_ARP: 692 if (ifp->if_flags & IFF_NOARP) { 693 /* Discard packet if ARP is disabled on interface */ 694 m_freem(m); 695 return; 696 } 697 isr = NETISR_ARP; 698 break; 699 #endif 700 701 #ifdef INET6 702 case ETHERTYPE_IPV6: 703 isr = NETISR_IPV6; 704 break; 705 #endif 706 707 #ifdef IPX 708 case ETHERTYPE_IPX: 709 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 710 return; 711 isr = NETISR_IPX; 712 break; 713 #endif 714 715 #ifdef NS 716 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 717 isr = NETISR_NS; 718 break; 719 720 #endif 721 722 #ifdef NETATALK 723 case ETHERTYPE_AT: 724 isr = NETISR_ATALK1; 725 break; 726 case ETHERTYPE_AARP: 727 isr = NETISR_AARP; 728 break; 729 #endif 730 731 case ETHERTYPE_VLAN: 732 if (vlan_input_p != NULL) 733 (*vlan_input_p)(eh, m); 734 else { 735 m->m_pkthdr.rcvif->if_noproto++; 736 m_freem(m); 737 } 738 return; 739 740 default: 741 #ifdef IPX 742 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 743 return; 744 #endif 745 #ifdef NS 746 checksum = mtod(m, ushort *); 747 /* Novell 802.3 */ 748 if ((ether_type <= ETHERMTU) && 749 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) { 750 if (*checksum == 0xE0E0) { 751 m->m_pkthdr.len -= 3; 752 m->m_len -= 3; 753 m->m_data += 3; 754 } 755 isr = NETISR_NS; 756 break; 757 } 758 #endif 759 #ifdef NETATALK 760 if (ether_type > ETHERMTU) 761 goto dropanyway; 762 l = mtod(m, struct llc *); 763 if (l->llc_dsap == LLC_SNAP_LSAP && 764 l->llc_ssap == LLC_SNAP_LSAP && 765 l->llc_control == LLC_UI) { 766 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 767 sizeof at_org_code) == 0 && 768 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 769 m_adj(m, sizeof(struct llc)); 770 isr = NETISR_ATALK2; 771 break; 772 } 773 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 774 sizeof aarp_org_code) == 0 && 775 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 776 m_adj(m, sizeof(struct llc)); 777 isr = NETISR_AARP; 778 break; 779 } 780 } 781 dropanyway: 782 #endif 783 if (ng_ether_input_orphan_p != NULL) 784 (*ng_ether_input_orphan_p)(ifp, m, eh); 785 else 786 m_freem(m); 787 return; 788 } 789 netisr_dispatch(isr, m); 790 } 791 792 /* 793 * Perform common duties while attaching to interface list 794 */ 795 796 void 797 ether_ifattach(struct ifnet *ifp, uint8_t *lla) 798 { 799 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header)); 800 } 801 802 void 803 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen) 804 { 805 struct ifaddr *ifa; 806 struct sockaddr_dl *sdl; 807 808 ifp->if_type = IFT_ETHER; 809 ifp->if_addrlen = ETHER_ADDR_LEN; 810 ifp->if_hdrlen = ETHER_HDR_LEN; 811 if_attach(ifp); 812 ifp->if_mtu = ETHERMTU; 813 if (ifp->if_baudrate == 0) 814 ifp->if_baudrate = 10000000; 815 ifp->if_output = ether_output; 816 ifp->if_input = ether_input_internal; 817 ifp->if_resolvemulti = ether_resolvemulti; 818 ifp->if_broadcastaddr = etherbroadcastaddr; 819 ifa = ifnet_addrs[ifp->if_index - 1]; 820 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 821 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 822 sdl->sdl_type = IFT_ETHER; 823 sdl->sdl_alen = ifp->if_addrlen; 824 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 825 /* 826 * XXX Keep the current drivers happy. 827 * XXX Remove once all drivers have been cleaned up 828 */ 829 if (lla != IFP2AC(ifp)->ac_enaddr) 830 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 831 bpfattach(ifp, dlt, hdrlen); 832 if (ng_ether_attach_p != NULL) 833 (*ng_ether_attach_p)(ifp); 834 if (BDG_LOADED) 835 bdgtakeifaces_ptr(); 836 837 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 838 } 839 840 /* 841 * Perform common duties while detaching an Ethernet interface 842 */ 843 void 844 ether_ifdetach(struct ifnet *ifp) 845 { 846 int s; 847 848 s = splnet(); 849 if_down(ifp); 850 splx(s); 851 852 if (ng_ether_detach_p != NULL) 853 (*ng_ether_detach_p)(ifp); 854 bpfdetach(ifp); 855 if_detach(ifp); 856 if (BDG_LOADED) 857 bdgtakeifaces_ptr(); 858 } 859 860 int 861 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 862 { 863 struct ifaddr *ifa = (struct ifaddr *) data; 864 struct ifreq *ifr = (struct ifreq *) data; 865 int error = 0; 866 867 switch (command) { 868 case SIOCSIFADDR: 869 ifp->if_flags |= IFF_UP; 870 871 switch (ifa->ifa_addr->sa_family) { 872 #ifdef INET 873 case AF_INET: 874 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 875 arp_ifinit(ifp, ifa); 876 break; 877 #endif 878 #ifdef IPX 879 /* 880 * XXX - This code is probably wrong 881 */ 882 case AF_IPX: 883 { 884 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 885 struct arpcom *ac = IFP2AC(ifp); 886 887 if (ipx_nullhost(*ina)) 888 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 889 else 890 bcopy(ina->x_host.c_host, ac->ac_enaddr, 891 sizeof ac->ac_enaddr); 892 893 ifp->if_init(ifp->if_softc); /* Set new address. */ 894 break; 895 } 896 #endif 897 #ifdef NS 898 /* 899 * XXX - This code is probably wrong 900 */ 901 case AF_NS: 902 { 903 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); 904 struct arpcom *ac = IFP2AC(ifp); 905 906 if (ns_nullhost(*ina)) 907 ina->x_host = *(union ns_host *)(ac->ac_enaddr); 908 else 909 bcopy(ina->x_host.c_host, ac->ac_enaddr, 910 sizeof ac->ac_enaddr); 911 912 /* 913 * Set new address 914 */ 915 ifp->if_init(ifp->if_softc); 916 break; 917 } 918 #endif 919 default: 920 ifp->if_init(ifp->if_softc); 921 break; 922 } 923 break; 924 925 case SIOCGIFADDR: 926 bcopy(IFP2AC(ifp)->ac_enaddr, 927 ((struct sockaddr *)ifr->ifr_data)->sa_data, 928 ETHER_ADDR_LEN); 929 break; 930 931 case SIOCSIFMTU: 932 /* 933 * Set the interface MTU. 934 */ 935 if (ifr->ifr_mtu > ETHERMTU) { 936 error = EINVAL; 937 } else { 938 ifp->if_mtu = ifr->ifr_mtu; 939 } 940 break; 941 default: 942 error = EINVAL; 943 break; 944 } 945 return (error); 946 } 947 948 int 949 ether_resolvemulti( 950 struct ifnet *ifp, 951 struct sockaddr **llsa, 952 struct sockaddr *sa) 953 { 954 struct sockaddr_dl *sdl; 955 struct sockaddr_in *sin; 956 #ifdef INET6 957 struct sockaddr_in6 *sin6; 958 #endif 959 u_char *e_addr; 960 961 switch(sa->sa_family) { 962 case AF_LINK: 963 /* 964 * No mapping needed. Just check that it's a valid MC address. 965 */ 966 sdl = (struct sockaddr_dl *)sa; 967 e_addr = LLADDR(sdl); 968 if ((e_addr[0] & 1) != 1) 969 return EADDRNOTAVAIL; 970 *llsa = 0; 971 return 0; 972 973 #ifdef INET 974 case AF_INET: 975 sin = (struct sockaddr_in *)sa; 976 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 977 return EADDRNOTAVAIL; 978 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 979 M_WAITOK | M_ZERO); 980 sdl->sdl_len = sizeof *sdl; 981 sdl->sdl_family = AF_LINK; 982 sdl->sdl_index = ifp->if_index; 983 sdl->sdl_type = IFT_ETHER; 984 sdl->sdl_alen = ETHER_ADDR_LEN; 985 e_addr = LLADDR(sdl); 986 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 987 *llsa = (struct sockaddr *)sdl; 988 return 0; 989 #endif 990 #ifdef INET6 991 case AF_INET6: 992 sin6 = (struct sockaddr_in6 *)sa; 993 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 994 /* 995 * An IP6 address of 0 means listen to all 996 * of the Ethernet multicast address used for IP6. 997 * (This is used for multicast routers.) 998 */ 999 ifp->if_flags |= IFF_ALLMULTI; 1000 *llsa = 0; 1001 return 0; 1002 } 1003 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1004 return EADDRNOTAVAIL; 1005 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1006 M_WAITOK | M_ZERO); 1007 sdl->sdl_len = sizeof *sdl; 1008 sdl->sdl_family = AF_LINK; 1009 sdl->sdl_index = ifp->if_index; 1010 sdl->sdl_type = IFT_ETHER; 1011 sdl->sdl_alen = ETHER_ADDR_LEN; 1012 e_addr = LLADDR(sdl); 1013 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1014 *llsa = (struct sockaddr *)sdl; 1015 return 0; 1016 #endif 1017 1018 default: 1019 /* 1020 * Well, the text isn't quite right, but it's the name 1021 * that counts... 1022 */ 1023 return EAFNOSUPPORT; 1024 } 1025 } 1026 1027 #if 0 1028 /* 1029 * This is for reference. We have a table-driven version 1030 * of the little-endian crc32 generator, which is faster 1031 * than the double-loop. 1032 */ 1033 uint32_t 1034 ether_crc32_le(const uint8_t *buf, size_t len) 1035 { 1036 uint32_t c, crc, carry; 1037 size_t i, j; 1038 1039 crc = 0xffffffffU; /* initial value */ 1040 1041 for (i = 0; i < len; i++) { 1042 c = buf[i]; 1043 for (j = 0; j < 8; j++) { 1044 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1045 crc >>= 1; 1046 c >>= 1; 1047 if (carry) 1048 crc = (crc ^ ETHER_CRC_POLY_LE); 1049 } 1050 } 1051 1052 return (crc); 1053 } 1054 #else 1055 uint32_t 1056 ether_crc32_le(const uint8_t *buf, size_t len) 1057 { 1058 static const uint32_t crctab[] = { 1059 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1060 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1061 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1062 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1063 }; 1064 uint32_t crc; 1065 size_t i; 1066 1067 crc = 0xffffffffU; /* initial value */ 1068 1069 for (i = 0; i < len; i++) { 1070 crc ^= buf[i]; 1071 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1072 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1073 } 1074 1075 return (crc); 1076 } 1077 #endif 1078 1079 uint32_t 1080 ether_crc32_be(const uint8_t *buf, size_t len) 1081 { 1082 uint32_t c, crc, carry; 1083 size_t i, j; 1084 1085 crc = 0xffffffffU; /* initial value */ 1086 1087 for (i = 0; i < len; i++) { 1088 c = buf[i]; 1089 for (j = 0; j < 8; j++) { 1090 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1091 crc <<= 1; 1092 c >>= 1; 1093 if (carry) 1094 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1095 } 1096 } 1097 1098 return (crc); 1099 } 1100 1101 /* 1102 * find the size of ethernet header, and call classifier 1103 */ 1104 void 1105 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 1106 struct altq_pktattr *pktattr) 1107 { 1108 struct ether_header *eh; 1109 uint16_t ether_type; 1110 int hlen, af, hdrsize; 1111 caddr_t hdr; 1112 1113 hlen = sizeof(struct ether_header); 1114 eh = mtod(m, struct ether_header *); 1115 1116 ether_type = ntohs(eh->ether_type); 1117 if (ether_type < ETHERMTU) { 1118 /* ick! LLC/SNAP */ 1119 struct llc *llc = (struct llc *)(eh + 1); 1120 hlen += 8; 1121 1122 if (m->m_len < hlen || 1123 llc->llc_dsap != LLC_SNAP_LSAP || 1124 llc->llc_ssap != LLC_SNAP_LSAP || 1125 llc->llc_control != LLC_UI) 1126 goto bad; /* not snap! */ 1127 1128 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 1129 } 1130 1131 if (ether_type == ETHERTYPE_IP) { 1132 af = AF_INET; 1133 hdrsize = 20; /* sizeof(struct ip) */ 1134 #ifdef INET6 1135 } else if (ether_type == ETHERTYPE_IPV6) { 1136 af = AF_INET6; 1137 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 1138 #endif 1139 } else 1140 goto bad; 1141 1142 while (m->m_len <= hlen) { 1143 hlen -= m->m_len; 1144 m = m->m_next; 1145 } 1146 hdr = m->m_data + hlen; 1147 if (m->m_len < hlen + hdrsize) { 1148 /* 1149 * ip header is not in a single mbuf. this should not 1150 * happen in the current code. 1151 * (todo: use m_pulldown in the future) 1152 */ 1153 goto bad; 1154 } 1155 m->m_data += hlen; 1156 m->m_len -= hlen; 1157 ifq_classify(ifq, m, af, pktattr); 1158 m->m_data -= hlen; 1159 m->m_len += hlen; 1160 1161 return; 1162 1163 bad: 1164 pktattr->pattr_class = NULL; 1165 pktattr->pattr_hdr = NULL; 1166 pktattr->pattr_af = AF_UNSPEC; 1167 } 1168