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 1.57 2008/03/20 14:08:45 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, struct mbuf **mp); 109 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 110 struct mbuf *m, const struct ether_header *eh); 111 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 112 void (*ng_ether_attach_p)(struct ifnet *ifp); 113 void (*ng_ether_detach_p)(struct ifnet *ifp); 114 115 int (*vlan_input_p)(const struct ether_header *eh, struct mbuf *m); 116 int (*vlan_input_tag_p)(struct mbuf *m, uint16_t t); 117 118 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 119 struct rtentry *); 120 static void ether_restore_header(struct mbuf **, const struct ether_header *, 121 const struct ether_header *); 122 123 /* 124 * if_bridge support 125 */ 126 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 127 int (*bridge_output_p)(struct ifnet *, struct mbuf *, 128 struct sockaddr *, struct rtentry *); 129 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 130 131 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 132 struct sockaddr *); 133 134 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 135 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 136 }; 137 138 #define gotoerr(e) do { error = (e); goto bad; } while (0) 139 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 140 141 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 142 struct ip_fw **rule, 143 const struct ether_header *eh); 144 145 static int ether_ipfw; 146 static u_int ether_restore_hdr; 147 static u_int ether_prepend_hdr; 148 149 SYSCTL_DECL(_net_link); 150 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 151 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 152 ðer_ipfw, 0, "Pass ether pkts through firewall"); 153 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 154 ðer_restore_hdr, 0, "# of ether header restoration"); 155 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 156 ðer_prepend_hdr, 0, 157 "# of ether header restoration which prepends mbuf"); 158 159 /* 160 * Ethernet output routine. 161 * Encapsulate a packet of type family for the local net. 162 * Use trailer local net encapsulation if enough data in first 163 * packet leaves a multiple of 512 bytes of data in remainder. 164 * Assumes that ifp is actually pointer to arpcom structure. 165 */ 166 static int 167 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 168 struct rtentry *rt) 169 { 170 struct ether_header *eh, *deh; 171 u_char *edst; 172 int loop_copy = 0; 173 int hlen = ETHER_HDR_LEN; /* link layer header length */ 174 struct arpcom *ac = IFP2AC(ifp); 175 int error; 176 177 ASSERT_SERIALIZED(ifp->if_serializer); 178 179 if (ifp->if_flags & IFF_MONITOR) 180 gotoerr(ENETDOWN); 181 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 182 gotoerr(ENETDOWN); 183 184 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 185 if (m == NULL) 186 return (ENOBUFS); 187 eh = mtod(m, struct ether_header *); 188 edst = eh->ether_dhost; 189 190 /* 191 * Fill in the destination ethernet address and frame type. 192 */ 193 switch (dst->sa_family) { 194 #ifdef INET 195 case AF_INET: 196 if (!arpresolve(ifp, rt, m, dst, edst)) 197 return (0); /* if not yet resolved */ 198 eh->ether_type = htons(ETHERTYPE_IP); 199 break; 200 #endif 201 #ifdef INET6 202 case AF_INET6: 203 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 204 return (0); /* Something bad happenned. */ 205 eh->ether_type = htons(ETHERTYPE_IPV6); 206 break; 207 #endif 208 #ifdef IPX 209 case AF_IPX: 210 if (ef_outputp != NULL) { 211 error = ef_outputp(ifp, &m, dst, &eh->ether_type, 212 &hlen); 213 if (error) 214 goto bad; 215 } else { 216 eh->ether_type = htons(ETHERTYPE_IPX); 217 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 218 edst, ETHER_ADDR_LEN); 219 } 220 break; 221 #endif 222 #ifdef NETATALK 223 case AF_APPLETALK: { 224 struct at_ifaddr *aa; 225 226 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) { 227 error = 0; /* XXX */ 228 goto bad; 229 } 230 /* 231 * In the phase 2 case, need to prepend an mbuf for 232 * the llc header. Since we must preserve the value 233 * of m, which is passed to us by value, we m_copy() 234 * the first mbuf, and use it for our llc header. 235 */ 236 if (aa->aa_flags & AFA_PHASE2) { 237 struct llc llc; 238 239 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT); 240 eh = mtod(m, struct ether_header *); 241 edst = eh->ether_dhost; 242 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 243 llc.llc_control = LLC_UI; 244 bcopy(at_org_code, llc.llc_snap_org_code, 245 sizeof at_org_code); 246 llc.llc_snap_ether_type = htons(ETHERTYPE_AT); 247 bcopy(&llc, 248 mtod(m, caddr_t) + sizeof(struct ether_header), 249 sizeof(struct llc)); 250 eh->ether_type = htons(m->m_pkthdr.len); 251 hlen = sizeof(struct llc) + ETHER_HDR_LEN; 252 } else { 253 eh->ether_type = htons(ETHERTYPE_AT); 254 } 255 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) 256 return (0); 257 break; 258 } 259 #endif 260 #ifdef NS 261 case AF_NS: 262 switch(ns_nettype) { 263 default: 264 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 265 eh->ether_type = 0x8137; 266 break; 267 case 0x0: /* Novell 802.3 */ 268 eh->ether_type = htons(m->m_pkthdr.len); 269 break; 270 case 0xe0e0: /* Novell 802.2 and Token-Ring */ 271 M_PREPEND(m, 3, MB_DONTWAIT); 272 eh = mtod(m, struct ether_header *); 273 edst = eh->ether_dhost; 274 eh->ether_type = htons(m->m_pkthdr.len); 275 cp = mtod(m, u_char *) + sizeof(struct ether_header); 276 *cp++ = 0xE0; 277 *cp++ = 0xE0; 278 *cp++ = 0x03; 279 break; 280 } 281 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst, 282 ETHER_ADDR_LEN); 283 /* 284 * XXX if ns_thishost is the same as the node's ethernet 285 * address then just the default code will catch this anyhow. 286 * So I'm not sure if this next clause should be here at all? 287 * [JRE] 288 */ 289 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) { 290 m->m_pkthdr.rcvif = ifp; 291 netisr_dispatch(NETISR_NS, m); 292 return (error); 293 } 294 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0) 295 m->m_flags |= M_BCAST; 296 break; 297 #endif 298 case pseudo_AF_HDRCMPLT: 299 case AF_UNSPEC: 300 loop_copy = -1; /* if this is for us, don't do it */ 301 deh = (struct ether_header *)dst->sa_data; 302 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 303 eh->ether_type = deh->ether_type; 304 break; 305 306 default: 307 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 308 gotoerr(EAFNOSUPPORT); 309 } 310 311 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 312 memcpy(eh->ether_shost, 313 ((struct ether_header *)dst->sa_data)->ether_shost, 314 ETHER_ADDR_LEN); 315 else 316 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 317 318 /* 319 * Bridges require special output handling. 320 */ 321 if (ifp->if_bridge) { 322 KASSERT(bridge_output_p != NULL, 323 ("%s: if_bridge not loaded!", __func__)); 324 return ((*bridge_output_p)(ifp, m, NULL, NULL)); 325 } 326 327 /* 328 * If a simplex interface, and the packet is being sent to our 329 * Ethernet address or a broadcast address, loopback a copy. 330 * XXX To make a simplex device behave exactly like a duplex 331 * device, we should copy in the case of sending to our own 332 * ethernet address (thus letting the original actually appear 333 * on the wire). However, we don't do that here for security 334 * reasons and compatibility with the original behavior. 335 */ 336 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 337 int csum_flags = 0; 338 339 if (m->m_pkthdr.csum_flags & CSUM_IP) 340 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 341 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 342 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 343 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 344 struct mbuf *n; 345 346 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 347 n->m_pkthdr.csum_flags |= csum_flags; 348 if (csum_flags & CSUM_DATA_VALID) 349 n->m_pkthdr.csum_data = 0xffff; 350 if_simloop(ifp, n, dst->sa_family, hlen); 351 } else 352 ifp->if_iqdrops++; 353 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 354 ETHER_ADDR_LEN) == 0) { 355 m->m_pkthdr.csum_flags |= csum_flags; 356 if (csum_flags & CSUM_DATA_VALID) 357 m->m_pkthdr.csum_data = 0xffff; 358 if_simloop(ifp, m, dst->sa_family, hlen); 359 return (0); /* XXX */ 360 } 361 } 362 363 #ifdef CARP 364 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) 365 goto bad; 366 #endif 367 368 369 /* Handle ng_ether(4) processing, if any */ 370 if (ng_ether_output_p != NULL) { 371 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) 372 goto bad; 373 if (m == NULL) 374 return (0); 375 } 376 377 /* Continue with link-layer output */ 378 return ether_output_frame(ifp, m); 379 380 bad: 381 m_freem(m); 382 return (error); 383 } 384 385 /* 386 * Ethernet link layer output routine to send a raw frame to the device. 387 * 388 * This assumes that the 14 byte Ethernet header is present and contiguous 389 * in the first mbuf. 390 */ 391 int 392 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 393 { 394 struct ip_fw *rule = NULL; 395 int error = 0; 396 struct altq_pktattr pktattr; 397 struct m_tag *mtag; 398 399 ASSERT_SERIALIZED(ifp->if_serializer); 400 401 /* Extract info from dummynet tag */ 402 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 403 if (mtag != NULL) { 404 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 405 406 m_tag_delete(m, mtag); 407 mtag = NULL; 408 } 409 410 if (ifq_is_enabled(&ifp->if_snd)) 411 altq_etherclassify(&ifp->if_snd, m, &pktattr); 412 crit_enter(); 413 if (IPFW_LOADED && ether_ipfw != 0) { 414 struct ether_header save_eh, *eh; 415 416 eh = mtod(m, struct ether_header *); 417 save_eh = *eh; 418 m_adj(m, ETHER_HDR_LEN); 419 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 420 crit_exit(); 421 if (m != NULL) { 422 m_freem(m); 423 return ENOBUFS; /* pkt dropped */ 424 } else 425 return 0; /* consumed e.g. in a pipe */ 426 } 427 428 /* packet was ok, restore the ethernet header */ 429 ether_restore_header(&m, eh, &save_eh); 430 if (m == NULL) { 431 crit_exit(); 432 return ENOBUFS; 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(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 452 const struct ether_header *eh) 453 { 454 struct ether_header save_eh = *eh; /* might be a ptr in m */ 455 struct ip_fw_args args; 456 struct m_tag *mtag; 457 int i; 458 459 if (*rule != NULL && fw_one_pass) 460 return TRUE; /* dummynet packet, already partially processed */ 461 462 /* 463 * I need some amount of data to be contiguous. 464 */ 465 i = min((*m0)->m_pkthdr.len, max_protohdr); 466 if ((*m0)->m_len < i) { 467 *m0 = m_pullup(*m0, i); 468 if (*m0 == NULL) 469 return FALSE; 470 } 471 472 args.m = *m0; /* the packet we are looking at */ 473 args.oif = dst; /* destination, if any */ 474 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 475 m_tag_delete(*m0, mtag); 476 args.rule = *rule; /* matching rule to restart */ 477 args.next_hop = NULL; /* we do not support forward yet */ 478 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 479 i = ip_fw_chk_ptr(&args); 480 *m0 = args.m; 481 *rule = args.rule; 482 483 if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */ 484 return FALSE; 485 486 if (i == 0) /* a PASS rule. */ 487 return TRUE; 488 489 if (i & IP_FW_PORT_DYNT_FLAG) { 490 /* 491 * Pass the pkt to dummynet, which consumes it. 492 */ 493 struct mbuf *m; 494 495 m = *m0; /* pass the original to dummynet */ 496 *m0 = NULL; /* and nothing back to the caller */ 497 498 ether_restore_header(&m, eh, &save_eh); 499 if (m == NULL) 500 return FALSE; 501 502 ip_fw_dn_io_ptr(m, (i & 0xffff), 503 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 504 return FALSE; 505 } 506 /* 507 * XXX at some point add support for divert/forward actions. 508 * If none of the above matches, we have to drop the pkt. 509 */ 510 return FALSE; 511 } 512 513 /* 514 * Process a received Ethernet packet. 515 * 516 * The ethernet header is assumed to be in the mbuf so the caller 517 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 518 * bytes in the first mbuf. 519 * 520 * This allows us to concentrate in one place a bunch of code which 521 * is replicated in all device drivers. Also, many functions called 522 * from ether_input() try to put the eh back into the mbuf, so we 523 * can later propagate the 'contiguous packet' interface to them. 524 * 525 * NOTA BENE: for all drivers "eh" is a pointer into the first mbuf or 526 * cluster, right before m_data. So be very careful when working on m, 527 * as you could destroy *eh !! 528 * 529 * First we perform any link layer operations, then continue to the 530 * upper layers with ether_demux(). 531 */ 532 void 533 ether_input(struct ifnet *ifp, struct mbuf *m) 534 { 535 struct ether_header *eh; 536 537 ASSERT_SERIALIZED(ifp->if_serializer); 538 M_ASSERTPKTHDR(m); 539 540 /* Discard packet if interface is not up */ 541 if (!(ifp->if_flags & IFF_UP)) { 542 m_freem(m); 543 return; 544 } 545 546 if (m->m_len < sizeof(struct ether_header)) { 547 /* XXX error in the caller. */ 548 m_freem(m); 549 return; 550 } 551 eh = mtod(m, struct ether_header *); 552 553 m->m_pkthdr.rcvif = ifp; 554 555 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 556 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 557 ifp->if_addrlen) == 0) 558 m->m_flags |= M_BCAST; 559 else 560 m->m_flags |= M_MCAST; 561 ifp->if_imcasts++; 562 } 563 564 BPF_MTAP(ifp, m); 565 566 ifp->if_ibytes += m->m_pkthdr.len; 567 568 if (ifp->if_flags & IFF_MONITOR) { 569 /* 570 * Interface marked for monitoring; discard packet. 571 */ 572 m_freem(m); 573 return; 574 } 575 576 /* 577 * Tap the packet off here for a bridge. bridge_input() 578 * will return NULL if it has consumed the packet, otherwise 579 * it gets processed as normal. Note that bridge_input() 580 * will always return the original packet if we need to 581 * process it locally. 582 */ 583 if (ifp->if_bridge) { 584 KASSERT(bridge_input_p != NULL, 585 ("%s: if_bridge not loaded!", __func__)); 586 587 if(m->m_flags & M_PROTO1) { 588 m->m_flags &= ~M_PROTO1; 589 } else { 590 /* clear M_PROMISC, in case the packets comes from a vlan */ 591 /* m->m_flags &= ~M_PROMISC; */ 592 lwkt_serialize_exit(ifp->if_serializer); 593 m = (*bridge_input_p)(ifp, m); 594 lwkt_serialize_enter(ifp->if_serializer); 595 if (m == NULL) 596 return; 597 598 KASSERT(ifp == m->m_pkthdr.rcvif, 599 ("bridge_input_p changed rcvif\n")); 600 601 /* 'm' may be changed by bridge_input_p() */ 602 eh = mtod(m, struct ether_header *); 603 } 604 } 605 606 /* Handle ng_ether(4) processing, if any */ 607 if (ng_ether_input_p != NULL) { 608 ng_ether_input_p(ifp, &m); 609 if (m == NULL) 610 return; 611 612 /* 'm' may be changed by ng_ether_input_p() */ 613 eh = mtod(m, struct ether_header *); 614 } 615 616 /* Continue with upper layer processing */ 617 ether_demux(ifp, m); 618 } 619 620 /* 621 * Upper layer processing for a received Ethernet packet. 622 */ 623 void 624 ether_demux(struct ifnet *ifp, struct mbuf *m) 625 { 626 struct ether_header save_eh, *eh; 627 int isr; 628 u_short ether_type; 629 struct ip_fw *rule = NULL; 630 struct m_tag *mtag; 631 #ifdef NETATALK 632 struct llc *l; 633 #endif 634 635 M_ASSERTPKTHDR(m); 636 KASSERT(m->m_len >= ETHER_HDR_LEN, 637 ("ether header is no contiguous!\n")); 638 639 eh = mtod(m, struct ether_header *); 640 save_eh = *eh; 641 642 /* XXX old crufty stuff, needs to be removed */ 643 m_adj(m, sizeof(struct ether_header)); 644 645 /* Extract info from dummynet tag */ 646 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 647 if (mtag != NULL) { 648 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 649 KKASSERT(ifp == NULL); 650 ifp = m->m_pkthdr.rcvif; 651 652 m_tag_delete(m, mtag); 653 mtag = NULL; 654 } 655 if (rule) /* packet is passing the second time */ 656 goto post_stats; 657 658 #ifdef CARP 659 /* 660 * XXX: Okay, we need to call carp_forus() and - if it is for 661 * us jump over code that does the normal check 662 * "ac_enaddr == ether_dhost". The check sequence is a bit 663 * different from OpenBSD, so we jump over as few code as 664 * possible, to catch _all_ sanity checks. This needs 665 * evaluation, to see if the carp ether_dhost values break any 666 * of these checks! 667 */ 668 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) 669 goto post_stats; 670 #endif 671 672 /* 673 * Discard packet if upper layers shouldn't see it because 674 * it was unicast to a different Ethernet address. If the 675 * driver is working properly, then this situation can only 676 * happen when the interface is in promiscuous mode. 677 */ 678 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 679 (eh->ether_dhost[0] & 1) == 0 && 680 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 681 m_freem(m); 682 return; 683 } 684 685 post_stats: 686 if (IPFW_LOADED && ether_ipfw != 0) { 687 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 688 m_freem(m); 689 return; 690 } 691 } 692 eh = NULL; /* catch any further usage */ 693 694 ether_type = ntohs(save_eh.ether_type); 695 696 switch (ether_type) { 697 #ifdef INET 698 case ETHERTYPE_IP: 699 if (ipflow_fastforward(m, ifp->if_serializer)) 700 return; 701 isr = NETISR_IP; 702 break; 703 704 case ETHERTYPE_ARP: 705 if (ifp->if_flags & IFF_NOARP) { 706 /* Discard packet if ARP is disabled on interface */ 707 m_freem(m); 708 return; 709 } 710 isr = NETISR_ARP; 711 break; 712 #endif 713 714 #ifdef INET6 715 case ETHERTYPE_IPV6: 716 isr = NETISR_IPV6; 717 break; 718 #endif 719 720 #ifdef IPX 721 case ETHERTYPE_IPX: 722 if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0) 723 return; 724 isr = NETISR_IPX; 725 break; 726 #endif 727 728 #ifdef NS 729 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 730 isr = NETISR_NS; 731 break; 732 733 #endif 734 735 #ifdef NETATALK 736 case ETHERTYPE_AT: 737 isr = NETISR_ATALK1; 738 break; 739 case ETHERTYPE_AARP: 740 isr = NETISR_AARP; 741 break; 742 #endif 743 744 case ETHERTYPE_VLAN: 745 if (vlan_input_p != NULL) { 746 (*vlan_input_p)(&save_eh, m); 747 } else { 748 m->m_pkthdr.rcvif->if_noproto++; 749 m_freem(m); 750 } 751 return; 752 753 default: 754 #ifdef IPX 755 if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0) 756 return; 757 #endif 758 #ifdef NS 759 checksum = mtod(m, ushort *); 760 /* Novell 802.3 */ 761 if ((ether_type <= ETHERMTU) && 762 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) { 763 if (*checksum == 0xE0E0) { 764 m->m_pkthdr.len -= 3; 765 m->m_len -= 3; 766 m->m_data += 3; 767 } 768 isr = NETISR_NS; 769 break; 770 } 771 #endif 772 #ifdef NETATALK 773 if (ether_type > ETHERMTU) 774 goto dropanyway; 775 l = mtod(m, struct llc *); 776 if (l->llc_dsap == LLC_SNAP_LSAP && 777 l->llc_ssap == LLC_SNAP_LSAP && 778 l->llc_control == LLC_UI) { 779 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 780 sizeof at_org_code) == 0 && 781 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 782 m_adj(m, sizeof(struct llc)); 783 isr = NETISR_ATALK2; 784 break; 785 } 786 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 787 sizeof aarp_org_code) == 0 && 788 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 789 m_adj(m, sizeof(struct llc)); 790 isr = NETISR_AARP; 791 break; 792 } 793 } 794 dropanyway: 795 #endif 796 if (ng_ether_input_orphan_p != NULL) 797 (*ng_ether_input_orphan_p)(ifp, m, &save_eh); 798 else 799 m_freem(m); 800 return; 801 } 802 netisr_dispatch(isr, m); 803 } 804 805 /* 806 * Perform common duties while attaching to interface list 807 */ 808 809 void 810 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 811 { 812 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 813 serializer); 814 } 815 816 void 817 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 818 lwkt_serialize_t serializer) 819 { 820 struct sockaddr_dl *sdl; 821 822 ifp->if_type = IFT_ETHER; 823 ifp->if_addrlen = ETHER_ADDR_LEN; 824 ifp->if_hdrlen = ETHER_HDR_LEN; 825 if_attach(ifp, serializer); 826 ifp->if_mtu = ETHERMTU; 827 if (ifp->if_baudrate == 0) 828 ifp->if_baudrate = 10000000; 829 ifp->if_output = ether_output; 830 ifp->if_input = ether_input; 831 ifp->if_resolvemulti = ether_resolvemulti; 832 ifp->if_broadcastaddr = etherbroadcastaddr; 833 sdl = IF_LLSOCKADDR(ifp); 834 sdl->sdl_type = IFT_ETHER; 835 sdl->sdl_alen = ifp->if_addrlen; 836 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 837 /* 838 * XXX Keep the current drivers happy. 839 * XXX Remove once all drivers have been cleaned up 840 */ 841 if (lla != IFP2AC(ifp)->ac_enaddr) 842 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 843 bpfattach(ifp, dlt, hdrlen); 844 if (ng_ether_attach_p != NULL) 845 (*ng_ether_attach_p)(ifp); 846 847 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 848 } 849 850 /* 851 * Perform common duties while detaching an Ethernet interface 852 */ 853 void 854 ether_ifdetach(struct ifnet *ifp) 855 { 856 if_down(ifp); 857 858 if (ng_ether_detach_p != NULL) 859 (*ng_ether_detach_p)(ifp); 860 bpfdetach(ifp); 861 if_detach(ifp); 862 } 863 864 int 865 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 866 { 867 struct ifaddr *ifa = (struct ifaddr *) data; 868 struct ifreq *ifr = (struct ifreq *) data; 869 int error = 0; 870 871 #define IF_INIT(ifp) \ 872 do { \ 873 if (((ifp)->if_flags & IFF_UP) == 0) { \ 874 (ifp)->if_flags |= IFF_UP; \ 875 (ifp)->if_init((ifp)->if_softc); \ 876 } \ 877 } while (0) 878 879 ASSERT_SERIALIZED(ifp->if_serializer); 880 881 switch (command) { 882 case SIOCSIFADDR: 883 switch (ifa->ifa_addr->sa_family) { 884 #ifdef INET 885 case AF_INET: 886 IF_INIT(ifp); /* before arpwhohas */ 887 arp_ifinit(ifp, ifa); 888 break; 889 #endif 890 #ifdef IPX 891 /* 892 * XXX - This code is probably wrong 893 */ 894 case AF_IPX: 895 { 896 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 897 struct arpcom *ac = IFP2AC(ifp); 898 899 if (ipx_nullhost(*ina)) 900 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 901 else 902 bcopy(ina->x_host.c_host, ac->ac_enaddr, 903 sizeof ac->ac_enaddr); 904 905 IF_INIT(ifp); /* Set new address. */ 906 break; 907 } 908 #endif 909 #ifdef NS 910 /* 911 * XXX - This code is probably wrong 912 */ 913 case AF_NS: 914 { 915 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); 916 struct arpcom *ac = IFP2AC(ifp); 917 918 if (ns_nullhost(*ina)) 919 ina->x_host = *(union ns_host *)(ac->ac_enaddr); 920 else 921 bcopy(ina->x_host.c_host, ac->ac_enaddr, 922 sizeof ac->ac_enaddr); 923 924 /* 925 * Set new address 926 */ 927 IF_INIT(ifp); 928 break; 929 } 930 #endif 931 default: 932 IF_INIT(ifp); 933 break; 934 } 935 break; 936 937 case SIOCGIFADDR: 938 bcopy(IFP2AC(ifp)->ac_enaddr, 939 ((struct sockaddr *)ifr->ifr_data)->sa_data, 940 ETHER_ADDR_LEN); 941 break; 942 943 case SIOCSIFMTU: 944 /* 945 * Set the interface MTU. 946 */ 947 if (ifr->ifr_mtu > ETHERMTU) { 948 error = EINVAL; 949 } else { 950 ifp->if_mtu = ifr->ifr_mtu; 951 } 952 break; 953 default: 954 error = EINVAL; 955 break; 956 } 957 return (error); 958 959 #undef IF_INIT 960 } 961 962 int 963 ether_resolvemulti( 964 struct ifnet *ifp, 965 struct sockaddr **llsa, 966 struct sockaddr *sa) 967 { 968 struct sockaddr_dl *sdl; 969 struct sockaddr_in *sin; 970 #ifdef INET6 971 struct sockaddr_in6 *sin6; 972 #endif 973 u_char *e_addr; 974 975 switch(sa->sa_family) { 976 case AF_LINK: 977 /* 978 * No mapping needed. Just check that it's a valid MC address. 979 */ 980 sdl = (struct sockaddr_dl *)sa; 981 e_addr = LLADDR(sdl); 982 if ((e_addr[0] & 1) != 1) 983 return EADDRNOTAVAIL; 984 *llsa = 0; 985 return 0; 986 987 #ifdef INET 988 case AF_INET: 989 sin = (struct sockaddr_in *)sa; 990 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 991 return EADDRNOTAVAIL; 992 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 993 M_WAITOK | M_ZERO); 994 sdl->sdl_len = sizeof *sdl; 995 sdl->sdl_family = AF_LINK; 996 sdl->sdl_index = ifp->if_index; 997 sdl->sdl_type = IFT_ETHER; 998 sdl->sdl_alen = ETHER_ADDR_LEN; 999 e_addr = LLADDR(sdl); 1000 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1001 *llsa = (struct sockaddr *)sdl; 1002 return 0; 1003 #endif 1004 #ifdef INET6 1005 case AF_INET6: 1006 sin6 = (struct sockaddr_in6 *)sa; 1007 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1008 /* 1009 * An IP6 address of 0 means listen to all 1010 * of the Ethernet multicast address used for IP6. 1011 * (This is used for multicast routers.) 1012 */ 1013 ifp->if_flags |= IFF_ALLMULTI; 1014 *llsa = 0; 1015 return 0; 1016 } 1017 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1018 return EADDRNOTAVAIL; 1019 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1020 M_WAITOK | M_ZERO); 1021 sdl->sdl_len = sizeof *sdl; 1022 sdl->sdl_family = AF_LINK; 1023 sdl->sdl_index = ifp->if_index; 1024 sdl->sdl_type = IFT_ETHER; 1025 sdl->sdl_alen = ETHER_ADDR_LEN; 1026 e_addr = LLADDR(sdl); 1027 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1028 *llsa = (struct sockaddr *)sdl; 1029 return 0; 1030 #endif 1031 1032 default: 1033 /* 1034 * Well, the text isn't quite right, but it's the name 1035 * that counts... 1036 */ 1037 return EAFNOSUPPORT; 1038 } 1039 } 1040 1041 #if 0 1042 /* 1043 * This is for reference. We have a table-driven version 1044 * of the little-endian crc32 generator, which is faster 1045 * than the double-loop. 1046 */ 1047 uint32_t 1048 ether_crc32_le(const uint8_t *buf, size_t len) 1049 { 1050 uint32_t c, crc, carry; 1051 size_t i, j; 1052 1053 crc = 0xffffffffU; /* initial value */ 1054 1055 for (i = 0; i < len; i++) { 1056 c = buf[i]; 1057 for (j = 0; j < 8; j++) { 1058 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1059 crc >>= 1; 1060 c >>= 1; 1061 if (carry) 1062 crc = (crc ^ ETHER_CRC_POLY_LE); 1063 } 1064 } 1065 1066 return (crc); 1067 } 1068 #else 1069 uint32_t 1070 ether_crc32_le(const uint8_t *buf, size_t len) 1071 { 1072 static const uint32_t crctab[] = { 1073 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1074 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1075 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1076 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1077 }; 1078 uint32_t crc; 1079 size_t i; 1080 1081 crc = 0xffffffffU; /* initial value */ 1082 1083 for (i = 0; i < len; i++) { 1084 crc ^= buf[i]; 1085 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1086 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1087 } 1088 1089 return (crc); 1090 } 1091 #endif 1092 1093 uint32_t 1094 ether_crc32_be(const uint8_t *buf, size_t len) 1095 { 1096 uint32_t c, crc, carry; 1097 size_t i, j; 1098 1099 crc = 0xffffffffU; /* initial value */ 1100 1101 for (i = 0; i < len; i++) { 1102 c = buf[i]; 1103 for (j = 0; j < 8; j++) { 1104 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1105 crc <<= 1; 1106 c >>= 1; 1107 if (carry) 1108 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1109 } 1110 } 1111 1112 return (crc); 1113 } 1114 1115 /* 1116 * find the size of ethernet header, and call classifier 1117 */ 1118 void 1119 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 1120 struct altq_pktattr *pktattr) 1121 { 1122 struct ether_header *eh; 1123 uint16_t ether_type; 1124 int hlen, af, hdrsize; 1125 caddr_t hdr; 1126 1127 hlen = sizeof(struct ether_header); 1128 eh = mtod(m, struct ether_header *); 1129 1130 ether_type = ntohs(eh->ether_type); 1131 if (ether_type < ETHERMTU) { 1132 /* ick! LLC/SNAP */ 1133 struct llc *llc = (struct llc *)(eh + 1); 1134 hlen += 8; 1135 1136 if (m->m_len < hlen || 1137 llc->llc_dsap != LLC_SNAP_LSAP || 1138 llc->llc_ssap != LLC_SNAP_LSAP || 1139 llc->llc_control != LLC_UI) 1140 goto bad; /* not snap! */ 1141 1142 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 1143 } 1144 1145 if (ether_type == ETHERTYPE_IP) { 1146 af = AF_INET; 1147 hdrsize = 20; /* sizeof(struct ip) */ 1148 #ifdef INET6 1149 } else if (ether_type == ETHERTYPE_IPV6) { 1150 af = AF_INET6; 1151 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 1152 #endif 1153 } else 1154 goto bad; 1155 1156 while (m->m_len <= hlen) { 1157 hlen -= m->m_len; 1158 m = m->m_next; 1159 } 1160 hdr = m->m_data + hlen; 1161 if (m->m_len < hlen + hdrsize) { 1162 /* 1163 * ip header is not in a single mbuf. this should not 1164 * happen in the current code. 1165 * (todo: use m_pulldown in the future) 1166 */ 1167 goto bad; 1168 } 1169 m->m_data += hlen; 1170 m->m_len -= hlen; 1171 ifq_classify(ifq, m, af, pktattr); 1172 m->m_data -= hlen; 1173 m->m_len += hlen; 1174 1175 return; 1176 1177 bad: 1178 pktattr->pattr_class = NULL; 1179 pktattr->pattr_hdr = NULL; 1180 pktattr->pattr_af = AF_UNSPEC; 1181 } 1182 1183 static void 1184 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 1185 const struct ether_header *save_eh) 1186 { 1187 struct mbuf *m = *m0; 1188 1189 ether_restore_hdr++; 1190 1191 /* 1192 * Prepend the header, optimize for the common case of 1193 * eh pointing into the mbuf. 1194 */ 1195 if ((const void *)(eh + 1) == (void *)m->m_data) { 1196 m->m_data -= ETHER_HDR_LEN; 1197 m->m_len += ETHER_HDR_LEN; 1198 m->m_pkthdr.len += ETHER_HDR_LEN; 1199 } else { 1200 ether_prepend_hdr++; 1201 1202 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1203 if (m != NULL) { 1204 bcopy(save_eh, mtod(m, struct ether_header *), 1205 ETHER_HDR_LEN); 1206 } 1207 } 1208 *m0 = m; 1209 } 1210