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 1.52 2007/11/27 11:06:31 sephe 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_netgraph.h" 43 #include "opt_carp.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 64 #if defined(INET) || defined(INET6) 65 #include <netinet/in.h> 66 #include <netinet/in_var.h> 67 #include <netinet/if_ether.h> 68 #include <net/ipfw/ip_fw.h> 69 #include <net/dummynet/ip_dummynet.h> 70 #endif 71 #ifdef INET6 72 #include <netinet6/nd6.h> 73 #endif 74 75 #ifdef CARP 76 #include <netinet/ip_carp.h> 77 #endif 78 79 #ifdef IPX 80 #include <netproto/ipx/ipx.h> 81 #include <netproto/ipx/ipx_if.h> 82 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 83 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 84 short *tp, int *hlen); 85 #endif 86 87 #ifdef NS 88 #include <netns/ns.h> 89 #include <netns/ns_if.h> 90 ushort ns_nettype; 91 int ether_outputdebug = 0; 92 int ether_inputdebug = 0; 93 #endif 94 95 #ifdef NETATALK 96 #include <netproto/atalk/at.h> 97 #include <netproto/atalk/at_var.h> 98 #include <netproto/atalk/at_extern.h> 99 100 #define llc_snap_org_code llc_un.type_snap.org_code 101 #define llc_snap_ether_type llc_un.type_snap.ether_type 102 103 extern u_char at_org_code[3]; 104 extern u_char aarp_org_code[3]; 105 #endif /* NETATALK */ 106 107 /* netgraph node hooks for ng_ether(4) */ 108 void (*ng_ether_input_p)(struct ifnet *ifp, 109 struct mbuf **mp, const struct ether_header *eh); 110 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 111 struct mbuf *m, const struct ether_header *eh); 112 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 113 void (*ng_ether_attach_p)(struct ifnet *ifp); 114 void (*ng_ether_detach_p)(struct ifnet *ifp); 115 116 int (*vlan_input_p)(const struct ether_header *eh, struct mbuf *m); 117 int (*vlan_input_tag_p)(struct mbuf *m, uint16_t t); 118 119 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 120 struct rtentry *); 121 122 /* 123 * if_bridge support 124 */ 125 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 126 int (*bridge_output_p)(struct ifnet *, struct mbuf *, 127 struct sockaddr *, struct rtentry *); 128 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 129 130 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 131 struct sockaddr *); 132 133 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 134 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 135 }; 136 137 #define gotoerr(e) do { error = (e); goto bad; } while (0) 138 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 139 140 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 141 struct ip_fw **rule, 142 const struct ether_header *eh, 143 boolean_t shared); 144 145 static int ether_ipfw; 146 SYSCTL_DECL(_net_link); 147 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 148 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 149 ðer_ipfw, 0, "Pass ether pkts through firewall"); 150 151 /* 152 * Ethernet output routine. 153 * Encapsulate a packet of type family for the local net. 154 * Use trailer local net encapsulation if enough data in first 155 * packet leaves a multiple of 512 bytes of data in remainder. 156 * Assumes that ifp is actually pointer to arpcom structure. 157 */ 158 static int 159 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 160 struct rtentry *rt) 161 { 162 struct ether_header *eh, *deh; 163 u_char *edst; 164 int loop_copy = 0; 165 int hlen = ETHER_HDR_LEN; /* link layer header length */ 166 struct arpcom *ac = IFP2AC(ifp); 167 int error; 168 169 ASSERT_SERIALIZED(ifp->if_serializer); 170 171 if (ifp->if_flags & IFF_MONITOR) 172 gotoerr(ENETDOWN); 173 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 174 gotoerr(ENETDOWN); 175 176 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 177 if (m == NULL) 178 return (ENOBUFS); 179 eh = mtod(m, struct ether_header *); 180 edst = eh->ether_dhost; 181 182 /* 183 * Fill in the destination ethernet address and frame type. 184 */ 185 switch (dst->sa_family) { 186 #ifdef INET 187 case AF_INET: 188 if (!arpresolve(ifp, rt, m, dst, edst)) 189 return (0); /* if not yet resolved */ 190 eh->ether_type = htons(ETHERTYPE_IP); 191 break; 192 #endif 193 #ifdef INET6 194 case AF_INET6: 195 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 196 return (0); /* Something bad happenned. */ 197 eh->ether_type = htons(ETHERTYPE_IPV6); 198 break; 199 #endif 200 #ifdef IPX 201 case AF_IPX: 202 if (ef_outputp != NULL) { 203 error = ef_outputp(ifp, &m, dst, &eh->ether_type, 204 &hlen); 205 if (error) 206 goto bad; 207 } else { 208 eh->ether_type = htons(ETHERTYPE_IPX); 209 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 210 edst, ETHER_ADDR_LEN); 211 } 212 break; 213 #endif 214 #ifdef NETATALK 215 case AF_APPLETALK: { 216 struct at_ifaddr *aa; 217 218 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) { 219 error = 0; /* XXX */ 220 goto bad; 221 } 222 /* 223 * In the phase 2 case, need to prepend an mbuf for 224 * the llc header. Since we must preserve the value 225 * of m, which is passed to us by value, we m_copy() 226 * the first mbuf, and use it for our llc header. 227 */ 228 if (aa->aa_flags & AFA_PHASE2) { 229 struct llc llc; 230 231 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT); 232 eh = mtod(m, struct ether_header *); 233 edst = eh->ether_dhost; 234 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 235 llc.llc_control = LLC_UI; 236 bcopy(at_org_code, llc.llc_snap_org_code, 237 sizeof at_org_code); 238 llc.llc_snap_ether_type = htons(ETHERTYPE_AT); 239 bcopy(&llc, 240 mtod(m, caddr_t) + sizeof(struct ether_header), 241 sizeof(struct llc)); 242 eh->ether_type = htons(m->m_pkthdr.len); 243 hlen = sizeof(struct llc) + ETHER_HDR_LEN; 244 } else { 245 eh->ether_type = htons(ETHERTYPE_AT); 246 } 247 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) 248 return (0); 249 break; 250 } 251 #endif 252 #ifdef NS 253 case AF_NS: 254 switch(ns_nettype) { 255 default: 256 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 257 eh->ether_type = 0x8137; 258 break; 259 case 0x0: /* Novell 802.3 */ 260 eh->ether_type = htons(m->m_pkthdr.len); 261 break; 262 case 0xe0e0: /* Novell 802.2 and Token-Ring */ 263 M_PREPEND(m, 3, MB_DONTWAIT); 264 eh = mtod(m, struct ether_header *); 265 edst = eh->ether_dhost; 266 eh->ether_type = htons(m->m_pkthdr.len); 267 cp = mtod(m, u_char *) + sizeof(struct ether_header); 268 *cp++ = 0xE0; 269 *cp++ = 0xE0; 270 *cp++ = 0x03; 271 break; 272 } 273 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst, 274 ETHER_ADDR_LEN); 275 /* 276 * XXX if ns_thishost is the same as the node's ethernet 277 * address then just the default code will catch this anyhow. 278 * So I'm not sure if this next clause should be here at all? 279 * [JRE] 280 */ 281 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) { 282 m->m_pkthdr.rcvif = ifp; 283 netisr_dispatch(NETISR_NS, m); 284 return (error); 285 } 286 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0) 287 m->m_flags |= M_BCAST; 288 break; 289 #endif 290 case pseudo_AF_HDRCMPLT: 291 case AF_UNSPEC: 292 loop_copy = -1; /* if this is for us, don't do it */ 293 deh = (struct ether_header *)dst->sa_data; 294 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 295 eh->ether_type = deh->ether_type; 296 break; 297 298 default: 299 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 300 gotoerr(EAFNOSUPPORT); 301 } 302 303 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 304 memcpy(eh->ether_shost, 305 ((struct ether_header *)dst->sa_data)->ether_shost, 306 ETHER_ADDR_LEN); 307 else 308 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 309 310 /* 311 * Bridges require special output handling. 312 */ 313 if (ifp->if_bridge) { 314 KASSERT(bridge_output_p != NULL, 315 ("%s: if_bridge not loaded!", __func__)); 316 return ((*bridge_output_p)(ifp, m, NULL, NULL)); 317 } 318 319 /* 320 * If a simplex interface, and the packet is being sent to our 321 * Ethernet address or a broadcast address, loopback a copy. 322 * XXX To make a simplex device behave exactly like a duplex 323 * device, we should copy in the case of sending to our own 324 * ethernet address (thus letting the original actually appear 325 * on the wire). However, we don't do that here for security 326 * reasons and compatibility with the original behavior. 327 */ 328 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 329 int csum_flags = 0; 330 331 if (m->m_pkthdr.csum_flags & CSUM_IP) 332 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 333 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 334 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 335 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 336 struct mbuf *n; 337 338 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 339 n->m_pkthdr.csum_flags |= csum_flags; 340 if (csum_flags & CSUM_DATA_VALID) 341 n->m_pkthdr.csum_data = 0xffff; 342 if_simloop(ifp, n, dst->sa_family, hlen); 343 } else 344 ifp->if_iqdrops++; 345 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 346 ETHER_ADDR_LEN) == 0) { 347 m->m_pkthdr.csum_flags |= csum_flags; 348 if (csum_flags & CSUM_DATA_VALID) 349 m->m_pkthdr.csum_data = 0xffff; 350 if_simloop(ifp, m, dst->sa_family, hlen); 351 return (0); /* XXX */ 352 } 353 } 354 355 #ifdef CARP 356 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) 357 goto bad; 358 #endif 359 360 361 /* Handle ng_ether(4) processing, if any */ 362 if (ng_ether_output_p != NULL) { 363 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) 364 goto bad; 365 if (m == NULL) 366 return (0); 367 } 368 369 /* Continue with link-layer output */ 370 return ether_output_frame(ifp, m); 371 372 bad: 373 m_freem(m); 374 return (error); 375 } 376 377 /* 378 * Ethernet link layer output routine to send a raw frame to the device. 379 * 380 * This assumes that the 14 byte Ethernet header is present and contiguous 381 * in the first mbuf. 382 */ 383 int 384 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 385 { 386 struct ip_fw *rule = NULL; 387 int error = 0; 388 struct altq_pktattr pktattr; 389 struct m_tag *mtag; 390 391 ASSERT_SERIALIZED(ifp->if_serializer); 392 393 /* Extract info from dummynet tag */ 394 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 395 if (mtag != NULL) { 396 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 397 398 m_tag_delete(m, mtag); 399 mtag = NULL; 400 } 401 402 if (ifq_is_enabled(&ifp->if_snd)) 403 altq_etherclassify(&ifp->if_snd, m, &pktattr); 404 crit_enter(); 405 if (IPFW_LOADED && ether_ipfw != 0) { 406 struct ether_header save_eh, *eh; 407 408 eh = mtod(m, struct ether_header *); 409 save_eh = *eh; 410 m_adj(m, ETHER_HDR_LEN); 411 if (!ether_ipfw_chk(&m, ifp, &rule, eh, FALSE)) { 412 crit_exit(); 413 if (m != NULL) { 414 m_freem(m); 415 return ENOBUFS; /* pkt dropped */ 416 } else 417 return 0; /* consumed e.g. in a pipe */ 418 } 419 eh = mtod(m, struct ether_header *); 420 /* packet was ok, restore the ethernet header */ 421 if ((void *)(eh + 1) == (void *)m->m_data) { 422 m->m_data -= ETHER_HDR_LEN ; 423 m->m_len += ETHER_HDR_LEN ; 424 m->m_pkthdr.len += ETHER_HDR_LEN ; 425 } else { 426 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 427 if (m == NULL) /* nope... */ { 428 crit_exit(); 429 return ENOBUFS; 430 } 431 bcopy(&save_eh, mtod(m, struct ether_header *), 432 ETHER_HDR_LEN); 433 } 434 } 435 crit_exit(); 436 437 /* 438 * Queue message on interface, update output statistics if 439 * successful, and start output if interface not yet active. 440 */ 441 error = ifq_handoff(ifp, m, &pktattr); 442 return (error); 443 } 444 445 /* 446 * ipfw processing for ethernet packets (in and out). 447 * The second parameter is NULL from ether_demux(), and ifp from 448 * ether_output_frame(). 449 */ 450 static boolean_t 451 ether_ipfw_chk( 452 struct mbuf **m0, 453 struct ifnet *dst, 454 struct ip_fw **rule, 455 const struct ether_header *eh, 456 boolean_t shared) 457 { 458 struct ether_header save_eh = *eh; /* might be a ptr in m */ 459 struct ip_fw_args args; 460 struct m_tag *mtag; 461 int i; 462 463 if (*rule != NULL && fw_one_pass) 464 return TRUE; /* dummynet packet, already partially processed */ 465 466 /* 467 * I need some amount of data to be contiguous, and in case others 468 * need the packet (shared==TRUE), it also better be in the first mbuf. 469 */ 470 i = min((*m0)->m_pkthdr.len, max_protohdr); 471 if (shared || (*m0)->m_len < i) { 472 *m0 = m_pullup(*m0, i); 473 if (*m0 == NULL) 474 return FALSE; 475 } 476 477 args.m = *m0; /* the packet we are looking at */ 478 args.oif = dst; /* destination, if any */ 479 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 480 m_tag_delete(*m0, mtag); 481 args.rule = *rule; /* matching rule to restart */ 482 args.next_hop = NULL; /* we do not support forward yet */ 483 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 484 i = ip_fw_chk_ptr(&args); 485 *m0 = args.m; 486 *rule = args.rule; 487 488 if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */ 489 return FALSE; 490 491 if (i == 0) /* a PASS rule. */ 492 return TRUE; 493 494 if (i & IP_FW_PORT_DYNT_FLAG) { 495 /* 496 * Pass the pkt to dummynet, which consumes it. 497 * If shared, make a copy and keep the original. 498 */ 499 struct mbuf *m ; 500 501 if (shared) { 502 m = m_copypacket(*m0, MB_DONTWAIT); 503 if (m == NULL) 504 return FALSE; 505 } else { 506 m = *m0 ; /* pass the original to dummynet */ 507 *m0 = NULL ; /* and nothing back to the caller */ 508 } 509 /* 510 * Prepend the header, optimize for the common case of 511 * eh pointing into the mbuf. 512 */ 513 if ((const void *)(eh + 1) == (void *)m->m_data) { 514 m->m_data -= ETHER_HDR_LEN ; 515 m->m_len += ETHER_HDR_LEN ; 516 m->m_pkthdr.len += ETHER_HDR_LEN ; 517 } else { 518 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 519 if (m == NULL) 520 return FALSE; 521 bcopy(&save_eh, mtod(m, struct ether_header *), 522 ETHER_HDR_LEN); 523 } 524 ip_fw_dn_io_ptr(m, (i & 0xffff), 525 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 526 return FALSE; 527 } 528 /* 529 * XXX at some point add support for divert/forward actions. 530 * If none of the above matches, we have to drop the pkt. 531 */ 532 return FALSE; 533 } 534 535 /* 536 * Process a received Ethernet packet. 537 * 538 * The ethernet header is assumed to be in the mbuf so the caller 539 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 540 * bytes in the first mbuf. 541 * 542 * This allows us to concentrate in one place a bunch of code which 543 * is replicated in all device drivers. Also, many functions called 544 * from ether_input() try to put the eh back into the mbuf, so we 545 * can later propagate the 'contiguous packet' interface to them. 546 * 547 * NOTA BENE: for all drivers "eh" is a pointer into the first mbuf or 548 * cluster, right before m_data. So be very careful when working on m, 549 * as you could destroy *eh !! 550 * 551 * First we perform any link layer operations, then continue to the 552 * upper layers with ether_demux(). 553 */ 554 void 555 ether_input(struct ifnet *ifp, struct mbuf *m) 556 { 557 struct ether_header *eh; 558 559 ASSERT_SERIALIZED(ifp->if_serializer); 560 M_ASSERTPKTHDR(m); 561 562 /* Discard packet if interface is not up */ 563 if (!(ifp->if_flags & IFF_UP)) { 564 m_freem(m); 565 return; 566 } 567 568 if (m->m_len < sizeof(struct ether_header)) { 569 /* XXX error in the caller. */ 570 m_freem(m); 571 return; 572 } 573 eh = mtod(m, struct ether_header *); 574 575 m->m_pkthdr.rcvif = ifp; 576 577 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 578 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 579 ifp->if_addrlen) == 0) 580 m->m_flags |= M_BCAST; 581 else 582 m->m_flags |= M_MCAST; 583 ifp->if_imcasts++; 584 } 585 586 BPF_MTAP(ifp, m); 587 588 ifp->if_ibytes += m->m_pkthdr.len; 589 590 if (ifp->if_flags & IFF_MONITOR) { 591 /* 592 * Interface marked for monitoring; discard packet. 593 */ 594 m_freem(m); 595 return; 596 } 597 598 /* 599 * Tap the packet off here for a bridge. bridge_input() 600 * will return NULL if it has consumed the packet, otherwise 601 * it gets processed as normal. Note that bridge_input() 602 * will always return the original packet if we need to 603 * process it locally. 604 */ 605 if (ifp->if_bridge) { 606 KASSERT(bridge_input_p != NULL, 607 ("%s: if_bridge not loaded!", __func__)); 608 609 if(m->m_flags & M_PROTO1) { 610 m->m_flags &= ~M_PROTO1; 611 } else { 612 /* clear M_PROMISC, in case the packets comes from a vlan */ 613 /* m->m_flags &= ~M_PROMISC; */ 614 lwkt_serialize_exit(ifp->if_serializer); 615 m = (*bridge_input_p)(ifp, m); 616 lwkt_serialize_enter(ifp->if_serializer); 617 if (m == NULL) 618 return; 619 620 KASSERT(ifp == m->m_pkthdr.rcvif, 621 ("bridge_input_p changed rcvif\n")); 622 623 /* 'm' may be changed by bridge_input_p() */ 624 eh = mtod(m, struct ether_header *); 625 } 626 } 627 628 /* XXX old crufty stuff, needs to be removed */ 629 m_adj(m, sizeof(struct ether_header)); 630 /* XXX */ 631 /* m->m_pkthdr.len = m->m_len; */ 632 633 /* Handle ng_ether(4) processing, if any */ 634 if (ng_ether_input_p != NULL) { 635 lwkt_serialize_exit(ifp->if_serializer); 636 (*ng_ether_input_p)(ifp, &m, eh); 637 lwkt_serialize_enter(ifp->if_serializer); 638 if (m == NULL) 639 return; 640 } 641 642 /* Continue with upper layer processing */ 643 ether_demux(ifp, eh, m); 644 } 645 646 /* 647 * Upper layer processing for a received Ethernet packet. 648 */ 649 void 650 ether_demux(struct ifnet *ifp, struct ether_header *eh0, struct mbuf *m) 651 { 652 struct ether_header eh; 653 int isr; 654 u_short ether_type; 655 struct ip_fw *rule = NULL; 656 struct m_tag *mtag; 657 #ifdef NETATALK 658 struct llc *l; 659 #endif 660 661 eh = *eh0; 662 663 /* Extract info from dummynet tag */ 664 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 665 if (mtag != NULL) { 666 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 667 KKASSERT(ifp == NULL); 668 ifp = m->m_pkthdr.rcvif; 669 670 m_tag_delete(m, mtag); 671 mtag = NULL; 672 } 673 if (rule) /* packet is passing the second time */ 674 goto post_stats; 675 676 #ifdef CARP 677 /* 678 * XXX: Okay, we need to call carp_forus() and - if it is for 679 * us jump over code that does the normal check 680 * "ac_enaddr == ether_dhost". The check sequence is a bit 681 * different from OpenBSD, so we jump over as few code as 682 * possible, to catch _all_ sanity checks. This needs 683 * evaluation, to see if the carp ether_dhost values break any 684 * of these checks! 685 */ 686 if (ifp->if_carp && carp_forus(ifp->if_carp, eh.ether_dhost)) 687 goto post_stats; 688 #endif 689 690 /* 691 * Discard packet if upper layers shouldn't see it because 692 * it was unicast to a different Ethernet address. If the 693 * driver is working properly, then this situation can only 694 * happen when the interface is in promiscuous mode. 695 */ 696 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 697 (eh.ether_dhost[0] & 1) == 0 && 698 bcmp(eh.ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 699 m_freem(m); 700 return; 701 } 702 703 post_stats: 704 if (IPFW_LOADED && ether_ipfw != 0) { 705 if (!ether_ipfw_chk(&m, NULL, &rule, &eh, FALSE)) { 706 m_freem(m); 707 return; 708 } 709 } 710 711 ether_type = ntohs(eh.ether_type); 712 713 switch (ether_type) { 714 #ifdef INET 715 case ETHERTYPE_IP: 716 if (ipflow_fastforward(m, ifp->if_serializer)) 717 return; 718 isr = NETISR_IP; 719 break; 720 721 case ETHERTYPE_ARP: 722 if (ifp->if_flags & IFF_NOARP) { 723 /* Discard packet if ARP is disabled on interface */ 724 m_freem(m); 725 return; 726 } 727 isr = NETISR_ARP; 728 break; 729 #endif 730 731 #ifdef INET6 732 case ETHERTYPE_IPV6: 733 isr = NETISR_IPV6; 734 break; 735 #endif 736 737 #ifdef IPX 738 case ETHERTYPE_IPX: 739 if (ef_inputp && ef_inputp(ifp, &eh, m) == 0) 740 return; 741 isr = NETISR_IPX; 742 break; 743 #endif 744 745 #ifdef NS 746 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 747 isr = NETISR_NS; 748 break; 749 750 #endif 751 752 #ifdef NETATALK 753 case ETHERTYPE_AT: 754 isr = NETISR_ATALK1; 755 break; 756 case ETHERTYPE_AARP: 757 isr = NETISR_AARP; 758 break; 759 #endif 760 761 case ETHERTYPE_VLAN: 762 if (vlan_input_p != NULL) 763 (*vlan_input_p)(&eh, m); 764 else { 765 m->m_pkthdr.rcvif->if_noproto++; 766 m_freem(m); 767 } 768 return; 769 770 default: 771 #ifdef IPX 772 if (ef_inputp && ef_inputp(ifp, &eh, m) == 0) 773 return; 774 #endif 775 #ifdef NS 776 checksum = mtod(m, ushort *); 777 /* Novell 802.3 */ 778 if ((ether_type <= ETHERMTU) && 779 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) { 780 if (*checksum == 0xE0E0) { 781 m->m_pkthdr.len -= 3; 782 m->m_len -= 3; 783 m->m_data += 3; 784 } 785 isr = NETISR_NS; 786 break; 787 } 788 #endif 789 #ifdef NETATALK 790 if (ether_type > ETHERMTU) 791 goto dropanyway; 792 l = mtod(m, struct llc *); 793 if (l->llc_dsap == LLC_SNAP_LSAP && 794 l->llc_ssap == LLC_SNAP_LSAP && 795 l->llc_control == LLC_UI) { 796 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 797 sizeof at_org_code) == 0 && 798 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 799 m_adj(m, sizeof(struct llc)); 800 isr = NETISR_ATALK2; 801 break; 802 } 803 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 804 sizeof aarp_org_code) == 0 && 805 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 806 m_adj(m, sizeof(struct llc)); 807 isr = NETISR_AARP; 808 break; 809 } 810 } 811 dropanyway: 812 #endif 813 if (ng_ether_input_orphan_p != NULL) 814 (*ng_ether_input_orphan_p)(ifp, m, &eh); 815 else 816 m_freem(m); 817 return; 818 } 819 netisr_dispatch(isr, m); 820 } 821 822 /* 823 * Perform common duties while attaching to interface list 824 */ 825 826 void 827 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 828 { 829 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 830 serializer); 831 } 832 833 void 834 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 835 lwkt_serialize_t serializer) 836 { 837 struct sockaddr_dl *sdl; 838 839 ifp->if_type = IFT_ETHER; 840 ifp->if_addrlen = ETHER_ADDR_LEN; 841 ifp->if_hdrlen = ETHER_HDR_LEN; 842 if_attach(ifp, serializer); 843 ifp->if_mtu = ETHERMTU; 844 if (ifp->if_baudrate == 0) 845 ifp->if_baudrate = 10000000; 846 ifp->if_output = ether_output; 847 ifp->if_input = ether_input; 848 ifp->if_resolvemulti = ether_resolvemulti; 849 ifp->if_broadcastaddr = etherbroadcastaddr; 850 sdl = IF_LLSOCKADDR(ifp); 851 sdl->sdl_type = IFT_ETHER; 852 sdl->sdl_alen = ifp->if_addrlen; 853 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 854 /* 855 * XXX Keep the current drivers happy. 856 * XXX Remove once all drivers have been cleaned up 857 */ 858 if (lla != IFP2AC(ifp)->ac_enaddr) 859 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 860 bpfattach(ifp, dlt, hdrlen); 861 if (ng_ether_attach_p != NULL) 862 (*ng_ether_attach_p)(ifp); 863 864 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 865 } 866 867 /* 868 * Perform common duties while detaching an Ethernet interface 869 */ 870 void 871 ether_ifdetach(struct ifnet *ifp) 872 { 873 if_down(ifp); 874 875 if (ng_ether_detach_p != NULL) 876 (*ng_ether_detach_p)(ifp); 877 bpfdetach(ifp); 878 if_detach(ifp); 879 } 880 881 int 882 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 883 { 884 struct ifaddr *ifa = (struct ifaddr *) data; 885 struct ifreq *ifr = (struct ifreq *) data; 886 int error = 0; 887 888 #define IF_INIT(ifp) \ 889 do { \ 890 if (((ifp)->if_flags & IFF_UP) == 0) { \ 891 (ifp)->if_flags |= IFF_UP; \ 892 (ifp)->if_init((ifp)->if_softc); \ 893 } \ 894 } while (0) 895 896 ASSERT_SERIALIZED(ifp->if_serializer); 897 898 switch (command) { 899 case SIOCSIFADDR: 900 switch (ifa->ifa_addr->sa_family) { 901 #ifdef INET 902 case AF_INET: 903 IF_INIT(ifp); /* before arpwhohas */ 904 arp_ifinit(ifp, ifa); 905 break; 906 #endif 907 #ifdef IPX 908 /* 909 * XXX - This code is probably wrong 910 */ 911 case AF_IPX: 912 { 913 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 914 struct arpcom *ac = IFP2AC(ifp); 915 916 if (ipx_nullhost(*ina)) 917 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 918 else 919 bcopy(ina->x_host.c_host, ac->ac_enaddr, 920 sizeof ac->ac_enaddr); 921 922 IF_INIT(ifp); /* Set new address. */ 923 break; 924 } 925 #endif 926 #ifdef NS 927 /* 928 * XXX - This code is probably wrong 929 */ 930 case AF_NS: 931 { 932 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); 933 struct arpcom *ac = IFP2AC(ifp); 934 935 if (ns_nullhost(*ina)) 936 ina->x_host = *(union ns_host *)(ac->ac_enaddr); 937 else 938 bcopy(ina->x_host.c_host, ac->ac_enaddr, 939 sizeof ac->ac_enaddr); 940 941 /* 942 * Set new address 943 */ 944 IF_INIT(ifp); 945 break; 946 } 947 #endif 948 default: 949 IF_INIT(ifp); 950 break; 951 } 952 break; 953 954 case SIOCGIFADDR: 955 bcopy(IFP2AC(ifp)->ac_enaddr, 956 ((struct sockaddr *)ifr->ifr_data)->sa_data, 957 ETHER_ADDR_LEN); 958 break; 959 960 case SIOCSIFMTU: 961 /* 962 * Set the interface MTU. 963 */ 964 if (ifr->ifr_mtu > ETHERMTU) { 965 error = EINVAL; 966 } else { 967 ifp->if_mtu = ifr->ifr_mtu; 968 } 969 break; 970 default: 971 error = EINVAL; 972 break; 973 } 974 return (error); 975 976 #undef IF_INIT 977 } 978 979 int 980 ether_resolvemulti( 981 struct ifnet *ifp, 982 struct sockaddr **llsa, 983 struct sockaddr *sa) 984 { 985 struct sockaddr_dl *sdl; 986 struct sockaddr_in *sin; 987 #ifdef INET6 988 struct sockaddr_in6 *sin6; 989 #endif 990 u_char *e_addr; 991 992 switch(sa->sa_family) { 993 case AF_LINK: 994 /* 995 * No mapping needed. Just check that it's a valid MC address. 996 */ 997 sdl = (struct sockaddr_dl *)sa; 998 e_addr = LLADDR(sdl); 999 if ((e_addr[0] & 1) != 1) 1000 return EADDRNOTAVAIL; 1001 *llsa = 0; 1002 return 0; 1003 1004 #ifdef INET 1005 case AF_INET: 1006 sin = (struct sockaddr_in *)sa; 1007 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1008 return EADDRNOTAVAIL; 1009 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1010 M_WAITOK | M_ZERO); 1011 sdl->sdl_len = sizeof *sdl; 1012 sdl->sdl_family = AF_LINK; 1013 sdl->sdl_index = ifp->if_index; 1014 sdl->sdl_type = IFT_ETHER; 1015 sdl->sdl_alen = ETHER_ADDR_LEN; 1016 e_addr = LLADDR(sdl); 1017 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1018 *llsa = (struct sockaddr *)sdl; 1019 return 0; 1020 #endif 1021 #ifdef INET6 1022 case AF_INET6: 1023 sin6 = (struct sockaddr_in6 *)sa; 1024 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1025 /* 1026 * An IP6 address of 0 means listen to all 1027 * of the Ethernet multicast address used for IP6. 1028 * (This is used for multicast routers.) 1029 */ 1030 ifp->if_flags |= IFF_ALLMULTI; 1031 *llsa = 0; 1032 return 0; 1033 } 1034 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1035 return EADDRNOTAVAIL; 1036 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1037 M_WAITOK | M_ZERO); 1038 sdl->sdl_len = sizeof *sdl; 1039 sdl->sdl_family = AF_LINK; 1040 sdl->sdl_index = ifp->if_index; 1041 sdl->sdl_type = IFT_ETHER; 1042 sdl->sdl_alen = ETHER_ADDR_LEN; 1043 e_addr = LLADDR(sdl); 1044 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1045 *llsa = (struct sockaddr *)sdl; 1046 return 0; 1047 #endif 1048 1049 default: 1050 /* 1051 * Well, the text isn't quite right, but it's the name 1052 * that counts... 1053 */ 1054 return EAFNOSUPPORT; 1055 } 1056 } 1057 1058 #if 0 1059 /* 1060 * This is for reference. We have a table-driven version 1061 * of the little-endian crc32 generator, which is faster 1062 * than the double-loop. 1063 */ 1064 uint32_t 1065 ether_crc32_le(const uint8_t *buf, size_t len) 1066 { 1067 uint32_t c, crc, carry; 1068 size_t i, j; 1069 1070 crc = 0xffffffffU; /* initial value */ 1071 1072 for (i = 0; i < len; i++) { 1073 c = buf[i]; 1074 for (j = 0; j < 8; j++) { 1075 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1076 crc >>= 1; 1077 c >>= 1; 1078 if (carry) 1079 crc = (crc ^ ETHER_CRC_POLY_LE); 1080 } 1081 } 1082 1083 return (crc); 1084 } 1085 #else 1086 uint32_t 1087 ether_crc32_le(const uint8_t *buf, size_t len) 1088 { 1089 static const uint32_t crctab[] = { 1090 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1091 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1092 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1093 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1094 }; 1095 uint32_t crc; 1096 size_t i; 1097 1098 crc = 0xffffffffU; /* initial value */ 1099 1100 for (i = 0; i < len; i++) { 1101 crc ^= buf[i]; 1102 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1103 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1104 } 1105 1106 return (crc); 1107 } 1108 #endif 1109 1110 uint32_t 1111 ether_crc32_be(const uint8_t *buf, size_t len) 1112 { 1113 uint32_t c, crc, carry; 1114 size_t i, j; 1115 1116 crc = 0xffffffffU; /* initial value */ 1117 1118 for (i = 0; i < len; i++) { 1119 c = buf[i]; 1120 for (j = 0; j < 8; j++) { 1121 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1122 crc <<= 1; 1123 c >>= 1; 1124 if (carry) 1125 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1126 } 1127 } 1128 1129 return (crc); 1130 } 1131 1132 /* 1133 * find the size of ethernet header, and call classifier 1134 */ 1135 void 1136 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 1137 struct altq_pktattr *pktattr) 1138 { 1139 struct ether_header *eh; 1140 uint16_t ether_type; 1141 int hlen, af, hdrsize; 1142 caddr_t hdr; 1143 1144 hlen = sizeof(struct ether_header); 1145 eh = mtod(m, struct ether_header *); 1146 1147 ether_type = ntohs(eh->ether_type); 1148 if (ether_type < ETHERMTU) { 1149 /* ick! LLC/SNAP */ 1150 struct llc *llc = (struct llc *)(eh + 1); 1151 hlen += 8; 1152 1153 if (m->m_len < hlen || 1154 llc->llc_dsap != LLC_SNAP_LSAP || 1155 llc->llc_ssap != LLC_SNAP_LSAP || 1156 llc->llc_control != LLC_UI) 1157 goto bad; /* not snap! */ 1158 1159 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 1160 } 1161 1162 if (ether_type == ETHERTYPE_IP) { 1163 af = AF_INET; 1164 hdrsize = 20; /* sizeof(struct ip) */ 1165 #ifdef INET6 1166 } else if (ether_type == ETHERTYPE_IPV6) { 1167 af = AF_INET6; 1168 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 1169 #endif 1170 } else 1171 goto bad; 1172 1173 while (m->m_len <= hlen) { 1174 hlen -= m->m_len; 1175 m = m->m_next; 1176 } 1177 hdr = m->m_data + hlen; 1178 if (m->m_len < hlen + hdrsize) { 1179 /* 1180 * ip header is not in a single mbuf. this should not 1181 * happen in the current code. 1182 * (todo: use m_pulldown in the future) 1183 */ 1184 goto bad; 1185 } 1186 m->m_data += hlen; 1187 m->m_len -= hlen; 1188 ifq_classify(ifq, m, af, pktattr); 1189 m->m_data -= hlen; 1190 m->m_len += hlen; 1191 1192 return; 1193 1194 bad: 1195 pktattr->pattr_class = NULL; 1196 pktattr->pattr_hdr = NULL; 1197 pktattr->pattr_af = AF_UNSPEC; 1198 } 1199