if_ethersubr.c revision 167716
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 30 * $FreeBSD: head/sys/net/if_ethersubr.c 167716 2007-03-19 18:39:36Z bms $ 31 */ 32 33#include "opt_atalk.h" 34#include "opt_inet.h" 35#include "opt_inet6.h" 36#include "opt_ipx.h" 37#include "opt_mac.h" 38#include "opt_netgraph.h" 39#include "opt_carp.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/kernel.h> 44#include <sys/malloc.h> 45#include <sys/module.h> 46#include <sys/mbuf.h> 47#include <sys/random.h> 48#include <sys/socket.h> 49#include <sys/sockio.h> 50#include <sys/sysctl.h> 51 52#include <net/if.h> 53#include <net/if_arp.h> 54#include <net/netisr.h> 55#include <net/route.h> 56#include <net/if_llc.h> 57#include <net/if_dl.h> 58#include <net/if_types.h> 59#include <net/bpf.h> 60#include <net/ethernet.h> 61#include <net/if_bridgevar.h> 62#include <net/if_vlan_var.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 <netinet/ip_fw.h> 69#include <netinet/ip_dummynet.h> 70#endif 71#ifdef INET6 72#include <netinet6/nd6.h> 73#endif 74 75#ifdef DEV_CARP 76#include <netinet/ip_carp.h> 77#endif 78 79#ifdef IPX 80#include <netipx/ipx.h> 81#include <netipx/ipx_if.h> 82#endif 83int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 84int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, 85 struct sockaddr *dst, short *tp, int *hlen); 86 87#ifdef NETATALK 88#include <netatalk/at.h> 89#include <netatalk/at_var.h> 90#include <netatalk/at_extern.h> 91 92#define llc_snap_org_code llc_un.type_snap.org_code 93#define llc_snap_ether_type llc_un.type_snap.ether_type 94 95extern u_char at_org_code[3]; 96extern u_char aarp_org_code[3]; 97#endif /* NETATALK */ 98 99#include <security/mac/mac_framework.h> 100 101/* netgraph node hooks for ng_ether(4) */ 102void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 103void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 104int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 105void (*ng_ether_attach_p)(struct ifnet *ifp); 106void (*ng_ether_detach_p)(struct ifnet *ifp); 107 108void (*vlan_input_p)(struct ifnet *, struct mbuf *); 109 110/* if_bridge(4) support */ 111struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 112int (*bridge_output_p)(struct ifnet *, struct mbuf *, 113 struct sockaddr *, struct rtentry *); 114void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 115 116static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 117 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 118 119static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 120 struct sockaddr *); 121 122/* XXX: should be in an arp support file, not here */ 123MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals"); 124 125#define ETHER_IS_BROADCAST(addr) \ 126 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 127 128#define senderr(e) do { error = (e); goto bad;} while (0) 129 130#if defined(INET) || defined(INET6) 131int 132ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 133 struct ip_fw **rule, int shared); 134static int ether_ipfw; 135#endif 136 137/* 138 * Ethernet output routine. 139 * Encapsulate a packet of type family for the local net. 140 * Use trailer local net encapsulation if enough data in first 141 * packet leaves a multiple of 512 bytes of data in remainder. 142 */ 143int 144ether_output(struct ifnet *ifp, struct mbuf *m, 145 struct sockaddr *dst, struct rtentry *rt0) 146{ 147 short type; 148 int error, hdrcmplt = 0; 149 u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 150 struct ether_header *eh; 151 int loop_copy = 1; 152 int hlen; /* link layer header length */ 153 154#ifdef MAC 155 error = mac_check_ifnet_transmit(ifp, m); 156 if (error) 157 senderr(error); 158#endif 159 160 if (ifp->if_flags & IFF_MONITOR) 161 senderr(ENETDOWN); 162 if (!((ifp->if_flags & IFF_UP) && 163 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 164 senderr(ENETDOWN); 165 166 hlen = ETHER_HDR_LEN; 167 switch (dst->sa_family) { 168#ifdef INET 169 case AF_INET: 170 error = arpresolve(ifp, rt0, m, dst, edst); 171 if (error) 172 return (error == EWOULDBLOCK ? 0 : error); 173 type = htons(ETHERTYPE_IP); 174 break; 175 case AF_ARP: 176 { 177 struct arphdr *ah; 178 ah = mtod(m, struct arphdr *); 179 ah->ar_hrd = htons(ARPHRD_ETHER); 180 181 loop_copy = 0; /* if this is for us, don't do it */ 182 183 switch(ntohs(ah->ar_op)) { 184 case ARPOP_REVREQUEST: 185 case ARPOP_REVREPLY: 186 type = htons(ETHERTYPE_REVARP); 187 break; 188 case ARPOP_REQUEST: 189 case ARPOP_REPLY: 190 default: 191 type = htons(ETHERTYPE_ARP); 192 break; 193 } 194 195 if (m->m_flags & M_BCAST) 196 bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); 197 else 198 bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); 199 200 } 201 break; 202#endif 203#ifdef INET6 204 case AF_INET6: 205 error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst); 206 if (error) 207 return error; 208 type = htons(ETHERTYPE_IPV6); 209 break; 210#endif 211#ifdef IPX 212 case AF_IPX: 213 if (ef_outputp) { 214 error = ef_outputp(ifp, &m, dst, &type, &hlen); 215 if (error) 216 goto bad; 217 } else 218 type = htons(ETHERTYPE_IPX); 219 bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 220 (caddr_t)edst, sizeof (edst)); 221 break; 222#endif 223#ifdef NETATALK 224 case AF_APPLETALK: 225 { 226 struct at_ifaddr *aa; 227 228 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) 229 senderr(EHOSTUNREACH); /* XXX */ 230 if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) 231 return (0); 232 /* 233 * In the phase 2 case, need to prepend an mbuf for the llc header. 234 */ 235 if ( aa->aa_flags & AFA_PHASE2 ) { 236 struct llc llc; 237 238 M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT); 239 if (m == NULL) 240 senderr(ENOBUFS); 241 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 242 llc.llc_control = LLC_UI; 243 bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code)); 244 llc.llc_snap_ether_type = htons( ETHERTYPE_AT ); 245 bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN); 246 type = htons(m->m_pkthdr.len); 247 hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN; 248 } else { 249 type = htons(ETHERTYPE_AT); 250 } 251 break; 252 } 253#endif /* NETATALK */ 254 255 case pseudo_AF_HDRCMPLT: 256 hdrcmplt = 1; 257 eh = (struct ether_header *)dst->sa_data; 258 (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); 259 /* FALLTHROUGH */ 260 261 case AF_UNSPEC: 262 loop_copy = 0; /* if this is for us, don't do it */ 263 eh = (struct ether_header *)dst->sa_data; 264 (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); 265 type = eh->ether_type; 266 break; 267 268 default: 269 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 270 senderr(EAFNOSUPPORT); 271 } 272 273 /* 274 * Add local net header. If no space in first mbuf, 275 * allocate another. 276 */ 277 M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT); 278 if (m == NULL) 279 senderr(ENOBUFS); 280 eh = mtod(m, struct ether_header *); 281 (void)memcpy(&eh->ether_type, &type, 282 sizeof(eh->ether_type)); 283 (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); 284 if (hdrcmplt) 285 (void)memcpy(eh->ether_shost, esrc, 286 sizeof(eh->ether_shost)); 287 else 288 (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), 289 sizeof(eh->ether_shost)); 290 291 /* 292 * If a simplex interface, and the packet is being sent to our 293 * Ethernet address or a broadcast address, loopback a copy. 294 * XXX To make a simplex device behave exactly like a duplex 295 * device, we should copy in the case of sending to our own 296 * ethernet address (thus letting the original actually appear 297 * on the wire). However, we don't do that here for security 298 * reasons and compatibility with the original behavior. 299 */ 300 if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && 301 m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) { 302 int csum_flags = 0; 303 304 if (m->m_pkthdr.csum_flags & CSUM_IP) 305 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 306 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 307 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 308 309 if (m->m_flags & M_BCAST) { 310 struct mbuf *n; 311 312 /* 313 * Because if_simloop() modifies the packet, we need a 314 * writable copy through m_dup() instead of a readonly 315 * one as m_copy[m] would give us. The alternative would 316 * be to modify if_simloop() to handle the readonly mbuf, 317 * but performancewise it is mostly equivalent (trading 318 * extra data copying vs. extra locking). 319 * 320 * XXX This is a local workaround. A number of less 321 * often used kernel parts suffer from the same bug. 322 * See PR kern/105943 for a proposed general solution. 323 */ 324 if ((n = m_dup(m, M_DONTWAIT)) != NULL) { 325 n->m_pkthdr.csum_flags |= csum_flags; 326 if (csum_flags & CSUM_DATA_VALID) 327 n->m_pkthdr.csum_data = 0xffff; 328 (void)if_simloop(ifp, n, dst->sa_family, hlen); 329 } else 330 ifp->if_iqdrops++; 331 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 332 ETHER_ADDR_LEN) == 0) { 333 m->m_pkthdr.csum_flags |= csum_flags; 334 if (csum_flags & CSUM_DATA_VALID) 335 m->m_pkthdr.csum_data = 0xffff; 336 (void) if_simloop(ifp, m, dst->sa_family, hlen); 337 return (0); /* XXX */ 338 } 339 } 340 341 /* 342 * Bridges require special output handling. 343 */ 344 if (ifp->if_bridge) { 345 BRIDGE_OUTPUT(ifp, m, error); 346 return (error); 347 } 348 349#ifdef DEV_CARP 350 if (ifp->if_carp && 351 (error = carp_output(ifp, m, dst, NULL))) 352 goto bad; 353#endif 354 355 /* Handle ng_ether(4) processing, if any */ 356 if (IFP2AC(ifp)->ac_netgraph != NULL) { 357 KASSERT(ng_ether_output_p != NULL, 358 ("ng_ether_output_p is NULL")); 359 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 360bad: if (m != NULL) 361 m_freem(m); 362 return (error); 363 } 364 if (m == NULL) 365 return (0); 366 } 367 368 /* Continue with link-layer output */ 369 return ether_output_frame(ifp, m); 370} 371 372/* 373 * Ethernet link layer output routine to send a raw frame to the device. 374 * 375 * This assumes that the 14 byte Ethernet header is present and contiguous 376 * in the first mbuf (if BRIDGE'ing). 377 */ 378int 379ether_output_frame(struct ifnet *ifp, struct mbuf *m) 380{ 381 int error; 382#if defined(INET) || defined(INET6) 383 struct ip_fw *rule = ip_dn_claim_rule(m); 384 385 if (IPFW_LOADED && ether_ipfw != 0) { 386 if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) { 387 if (m) { 388 m_freem(m); 389 return EACCES; /* pkt dropped */ 390 } else 391 return 0; /* consumed e.g. in a pipe */ 392 } 393 } 394#endif 395 396 /* 397 * Queue message on interface, update output statistics if 398 * successful, and start output if interface not yet active. 399 */ 400 IFQ_HANDOFF(ifp, m, error); 401 return (error); 402} 403 404#if defined(INET) || defined(INET6) 405/* 406 * ipfw processing for ethernet packets (in and out). 407 * The second parameter is NULL from ether_demux, and ifp from 408 * ether_output_frame. 409 */ 410int 411ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 412 struct ip_fw **rule, int shared) 413{ 414 struct ether_header *eh; 415 struct ether_header save_eh; 416 struct mbuf *m; 417 int i; 418 struct ip_fw_args args; 419 420 if (*rule != NULL && fw_one_pass) 421 return 1; /* dummynet packet, already partially processed */ 422 423 /* 424 * I need some amt of data to be contiguous, and in case others need 425 * the packet (shared==1) also better be in the first mbuf. 426 */ 427 m = *m0; 428 i = min( m->m_pkthdr.len, max_protohdr); 429 if ( shared || m->m_len < i) { 430 m = m_pullup(m, i); 431 if (m == NULL) { 432 *m0 = m; 433 return 0; 434 } 435 } 436 eh = mtod(m, struct ether_header *); 437 save_eh = *eh; /* save copy for restore below */ 438 m_adj(m, ETHER_HDR_LEN); /* strip ethernet header */ 439 440 args.m = m; /* the packet we are looking at */ 441 args.oif = dst; /* destination, if any */ 442 args.rule = *rule; /* matching rule to restart */ 443 args.next_hop = NULL; /* we do not support forward yet */ 444 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 445 args.inp = NULL; /* used by ipfw uid/gid/jail rules */ 446 i = ip_fw_chk_ptr(&args); 447 m = args.m; 448 if (m != NULL) { 449 /* 450 * Restore Ethernet header, as needed, in case the 451 * mbuf chain was replaced by ipfw. 452 */ 453 M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT); 454 if (m == NULL) { 455 *m0 = m; 456 return 0; 457 } 458 if (eh != mtod(m, struct ether_header *)) 459 bcopy(&save_eh, mtod(m, struct ether_header *), 460 ETHER_HDR_LEN); 461 } 462 *m0 = m; 463 *rule = args.rule; 464 465 if (i == IP_FW_DENY) /* drop */ 466 return 0; 467 468 KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL")); 469 470 if (i == IP_FW_PASS) /* a PASS rule. */ 471 return 1; 472 473 if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) { 474 /* 475 * Pass the pkt to dummynet, which consumes it. 476 * If shared, make a copy and keep the original. 477 */ 478 if (shared) { 479 m = m_copypacket(m, M_DONTWAIT); 480 if (m == NULL) 481 return 0; 482 } else { 483 /* 484 * Pass the original to dummynet and 485 * nothing back to the caller 486 */ 487 *m0 = NULL ; 488 } 489 ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 490 return 0; 491 } 492 /* 493 * XXX at some point add support for divert/forward actions. 494 * If none of the above matches, we have to drop the pkt. 495 */ 496 return 0; 497} 498#endif 499 500/* 501 * Process a received Ethernet packet; the packet is in the 502 * mbuf chain m with the ethernet header at the front. 503 */ 504static void 505ether_input(struct ifnet *ifp, struct mbuf *m) 506{ 507 struct ether_header *eh; 508 u_short etype; 509 510 if ((ifp->if_flags & IFF_UP) == 0) { 511 m_freem(m); 512 return; 513 } 514#ifdef DIAGNOSTIC 515 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 516 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 517 m_freem(m); 518 return; 519 } 520#endif 521 /* 522 * Do consistency checks to verify assumptions 523 * made by code past this point. 524 */ 525 if ((m->m_flags & M_PKTHDR) == 0) { 526 if_printf(ifp, "discard frame w/o packet header\n"); 527 ifp->if_ierrors++; 528 m_freem(m); 529 return; 530 } 531 if (m->m_len < ETHER_HDR_LEN) { 532 /* XXX maybe should pullup? */ 533 if_printf(ifp, "discard frame w/o leading ethernet " 534 "header (len %u pkt len %u)\n", 535 m->m_len, m->m_pkthdr.len); 536 ifp->if_ierrors++; 537 m_freem(m); 538 return; 539 } 540 eh = mtod(m, struct ether_header *); 541 etype = ntohs(eh->ether_type); 542 if (m->m_pkthdr.len > 543 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 544 if_printf(ifp, "discard oversize frame " 545 "(ether type %x flags %x len %u > max %lu)\n", 546 etype, m->m_flags, m->m_pkthdr.len, 547 ETHER_MAX_FRAME(ifp, etype, 548 m->m_flags & M_HASFCS)); 549 ifp->if_ierrors++; 550 m_freem(m); 551 return; 552 } 553 if (m->m_pkthdr.rcvif == NULL) { 554 if_printf(ifp, "discard frame w/o interface pointer\n"); 555 ifp->if_ierrors++; 556 m_freem(m); 557 return; 558 } 559#ifdef DIAGNOSTIC 560 if (m->m_pkthdr.rcvif != ifp) { 561 if_printf(ifp, "Warning, frame marked as received on %s\n", 562 m->m_pkthdr.rcvif->if_xname); 563 } 564#endif 565 566 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 567 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 568 m->m_flags |= M_BCAST; 569 else 570 m->m_flags |= M_MCAST; 571 ifp->if_imcasts++; 572 } 573 574#ifdef MAC 575 /* 576 * Tag the mbuf with an appropriate MAC label before any other 577 * consumers can get to it. 578 */ 579 mac_create_mbuf_from_ifnet(ifp, m); 580#endif 581 582 /* 583 * Give bpf a chance at the packet. 584 */ 585 ETHER_BPF_MTAP(ifp, m); 586 587 /* 588 * If the CRC is still on the packet, trim it off. We do this once 589 * and once only in case we are re-entered. Nothing else on the 590 * Ethernet receive path expects to see the FCS. 591 */ 592 if (m->m_flags & M_HASFCS) { 593 m_adj(m, -ETHER_CRC_LEN); 594 m->m_flags &= ~M_HASFCS; 595 } 596 597 ifp->if_ibytes += m->m_pkthdr.len; 598 599 /* Allow monitor mode to claim this frame, after stats are updated. */ 600 if (ifp->if_flags & IFF_MONITOR) { 601 m_freem(m); 602 return; 603 } 604 605 /* 606 * If the hardware did not process an 802.1Q tag, do this now, 607 * to allow 802.1P priority frames to be passed to the main input 608 * path correctly. 609 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 610 */ 611 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 612 struct ether_vlan_header *evl; 613 614 if (m->m_len < sizeof(*evl) && 615 (m = m_pullup(m, sizeof(*evl))) == NULL) { 616 if_printf(ifp, "cannot pullup VLAN header\n"); 617 ifp->if_ierrors++; 618 m_freem(m); 619 return; 620 } 621 622 evl = mtod(m, struct ether_vlan_header *); 623 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 624 m->m_flags |= M_VLANTAG; 625 626 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 627 ETHER_HDR_LEN - ETHER_TYPE_LEN); 628 m_adj(m, ETHER_VLAN_ENCAP_LEN); 629 } 630 631 /* Allow ng_ether(4) to claim this frame. */ 632 if (IFP2AC(ifp)->ac_netgraph != NULL) { 633 KASSERT(ng_ether_input_p != NULL, 634 ("%s: ng_ether_input_p is NULL", __func__)); 635 m->m_flags &= ~M_PROMISC; 636 (*ng_ether_input_p)(ifp, &m); 637 if (m == NULL) 638 return; 639 } 640 641 /* 642 * Allow if_bridge(4) to claim this frame. 643 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 644 * and the frame should be delivered locally. 645 */ 646 if (ifp->if_bridge != NULL) { 647 m->m_flags &= ~M_PROMISC; 648 BRIDGE_INPUT(ifp, m); 649 if (m == NULL) 650 return; 651 } 652 653#ifdef DEV_CARP 654 /* 655 * Clear M_PROMISC on frame so that carp(4) will see it when the 656 * mbuf flows up to Layer 3. 657 * FreeBSD's implementation of carp(4) uses the inprotosw 658 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 659 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 660 * is outside the scope of the M_PROMISC test below. 661 * TODO: Maintain a hash table of ethernet addresses other than 662 * ether_dhost which may be active on this ifp. 663 */ 664 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) { 665 m->m_flags &= ~M_PROMISC; 666 } else 667#endif 668 { 669 /* 670 * If the frame was received promiscuously, set the 671 * M_PROMISC flag on the mbuf chain. The frame may need to 672 * be seen by the rest of the Ethernet input path in case of 673 * re-entry (e.g. bridge, vlan, netgraph) but should not be 674 * seen by upper protocol layers. 675 */ 676 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 677 (ifp->if_flags & IFF_PROMISC) != 0 && 678 !bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN)) 679 m->m_flags |= M_PROMISC; 680 } 681 682 /* First chunk of an mbuf contains good entropy */ 683 if (harvest.ethernet) 684 random_harvest(m, 16, 3, 0, RANDOM_NET); 685 686 ether_demux(ifp, m); 687} 688 689/* 690 * Upper layer processing for a received Ethernet packet. 691 */ 692void 693ether_demux(struct ifnet *ifp, struct mbuf *m) 694{ 695 struct ether_header *eh; 696 int isr; 697 u_short ether_type; 698#if defined(NETATALK) 699 struct llc *l; 700#endif 701 702 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 703 704#if defined(INET) || defined(INET6) 705 /* 706 * Allow dummynet and/or ipfw to claim the frame. 707 * Do not do this for PROMISC frames in case we are re-entered. 708 */ 709 if (IPFW_LOADED && ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) { 710 struct ip_fw *rule = ip_dn_claim_rule(m); 711 712 if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) { 713 if (m) 714 m_freem(m); /* dropped; free mbuf chain */ 715 return; /* consumed */ 716 } 717 } 718#endif 719 eh = mtod(m, struct ether_header *); 720 ether_type = ntohs(eh->ether_type); 721 722 /* 723 * If this frame has a VLAN tag other than 0, call vlan_input() 724 * if its module is loaded. Otherwise, drop. 725 */ 726 if ((m->m_flags & M_VLANTAG) && 727 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 728 if (ifp->if_vlantrunk == NULL) { 729 ifp->if_noproto++; 730 m_freem(m); 731 return; 732 } 733 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 734 __func__)); 735 /* Clear before possibly re-entering ether_input(). */ 736 m->m_flags &= ~M_PROMISC; 737 (*vlan_input_p)(ifp, m); 738 return; 739 } 740 741 /* 742 * Pass promiscuously received frames to the upper layer if the user 743 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 744 */ 745 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 746 m_freem(m); 747 return; 748 } 749 750 /* 751 * Reset layer specific mbuf flags to avoid confusing upper layers. 752 * Strip off Ethernet header. 753 */ 754 m->m_flags &= ~M_VLANTAG; 755 m->m_flags &= ~(M_PROTOFLAGS); 756 m_adj(m, ETHER_HDR_LEN); 757 758 /* 759 * Dispatch frame to upper layer. 760 */ 761 switch (ether_type) { 762#ifdef INET 763 case ETHERTYPE_IP: 764 if ((m = ip_fastforward(m)) == NULL) 765 return; 766 isr = NETISR_IP; 767 break; 768 769 case ETHERTYPE_ARP: 770 if (ifp->if_flags & IFF_NOARP) { 771 /* Discard packet if ARP is disabled on interface */ 772 m_freem(m); 773 return; 774 } 775 isr = NETISR_ARP; 776 break; 777#endif 778#ifdef IPX 779 case ETHERTYPE_IPX: 780 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 781 return; 782 isr = NETISR_IPX; 783 break; 784#endif 785#ifdef INET6 786 case ETHERTYPE_IPV6: 787 isr = NETISR_IPV6; 788 break; 789#endif 790#ifdef NETATALK 791 case ETHERTYPE_AT: 792 isr = NETISR_ATALK1; 793 break; 794 case ETHERTYPE_AARP: 795 isr = NETISR_AARP; 796 break; 797#endif /* NETATALK */ 798 default: 799#ifdef IPX 800 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 801 return; 802#endif /* IPX */ 803#if defined(NETATALK) 804 if (ether_type > ETHERMTU) 805 goto discard; 806 l = mtod(m, struct llc *); 807 if (l->llc_dsap == LLC_SNAP_LSAP && 808 l->llc_ssap == LLC_SNAP_LSAP && 809 l->llc_control == LLC_UI) { 810 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 811 sizeof(at_org_code)) == 0 && 812 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 813 m_adj(m, LLC_SNAPFRAMELEN); 814 isr = NETISR_ATALK2; 815 break; 816 } 817 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 818 sizeof(aarp_org_code)) == 0 && 819 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 820 m_adj(m, LLC_SNAPFRAMELEN); 821 isr = NETISR_AARP; 822 break; 823 } 824 } 825#endif /* NETATALK */ 826 goto discard; 827 } 828 netisr_dispatch(isr, m); 829 return; 830 831discard: 832 /* 833 * Packet is to be discarded. If netgraph is present, 834 * hand the packet to it for last chance processing; 835 * otherwise dispose of it. 836 */ 837 if (IFP2AC(ifp)->ac_netgraph != NULL) { 838 KASSERT(ng_ether_input_orphan_p != NULL, 839 ("ng_ether_input_orphan_p is NULL")); 840 /* 841 * Put back the ethernet header so netgraph has a 842 * consistent view of inbound packets. 843 */ 844 M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT); 845 (*ng_ether_input_orphan_p)(ifp, m); 846 return; 847 } 848 m_freem(m); 849} 850 851/* 852 * Convert Ethernet address to printable (loggable) representation. 853 * This routine is for compatibility; it's better to just use 854 * 855 * printf("%6D", <pointer to address>, ":"); 856 * 857 * since there's no static buffer involved. 858 */ 859char * 860ether_sprintf(const u_char *ap) 861{ 862 static char etherbuf[18]; 863 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 864 return (etherbuf); 865} 866 867/* 868 * Perform common duties while attaching to interface list 869 */ 870void 871ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 872{ 873 int i; 874 struct ifaddr *ifa; 875 struct sockaddr_dl *sdl; 876 877 ifp->if_addrlen = ETHER_ADDR_LEN; 878 ifp->if_hdrlen = ETHER_HDR_LEN; 879 if_attach(ifp); 880 ifp->if_mtu = ETHERMTU; 881 ifp->if_output = ether_output; 882 ifp->if_input = ether_input; 883 ifp->if_resolvemulti = ether_resolvemulti; 884 if (ifp->if_baudrate == 0) 885 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 886 ifp->if_broadcastaddr = etherbroadcastaddr; 887 888 ifa = ifp->if_addr; 889 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 890 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 891 sdl->sdl_type = IFT_ETHER; 892 sdl->sdl_alen = ifp->if_addrlen; 893 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 894 895 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 896 if (ng_ether_attach_p != NULL) 897 (*ng_ether_attach_p)(ifp); 898 899 /* Announce Ethernet MAC address if non-zero. */ 900 for (i = 0; i < ifp->if_addrlen; i++) 901 if (lla[i] != 0) 902 break; 903 if (i != ifp->if_addrlen) 904 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 905 if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0) 906 if_printf(ifp, "if_start running deferred for Giant\n"); 907} 908 909/* 910 * Perform common duties while detaching an Ethernet interface 911 */ 912void 913ether_ifdetach(struct ifnet *ifp) 914{ 915 if (IFP2AC(ifp)->ac_netgraph != NULL) { 916 KASSERT(ng_ether_detach_p != NULL, 917 ("ng_ether_detach_p is NULL")); 918 (*ng_ether_detach_p)(ifp); 919 } 920 921 bpfdetach(ifp); 922 if_detach(ifp); 923} 924 925SYSCTL_DECL(_net_link); 926SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 927#if defined(INET) || defined(INET6) 928SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 929 ðer_ipfw,0,"Pass ether pkts through firewall"); 930#endif 931 932#if 0 933/* 934 * This is for reference. We have a table-driven version 935 * of the little-endian crc32 generator, which is faster 936 * than the double-loop. 937 */ 938uint32_t 939ether_crc32_le(const uint8_t *buf, size_t len) 940{ 941 size_t i; 942 uint32_t crc; 943 int bit; 944 uint8_t data; 945 946 crc = 0xffffffff; /* initial value */ 947 948 for (i = 0; i < len; i++) { 949 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) 950 carry = (crc ^ data) & 1; 951 crc >>= 1; 952 if (carry) 953 crc = (crc ^ ETHER_CRC_POLY_LE); 954 } 955 956 return (crc); 957} 958#else 959uint32_t 960ether_crc32_le(const uint8_t *buf, size_t len) 961{ 962 static const uint32_t crctab[] = { 963 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 964 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 965 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 966 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 967 }; 968 size_t i; 969 uint32_t crc; 970 971 crc = 0xffffffff; /* initial value */ 972 973 for (i = 0; i < len; i++) { 974 crc ^= buf[i]; 975 crc = (crc >> 4) ^ crctab[crc & 0xf]; 976 crc = (crc >> 4) ^ crctab[crc & 0xf]; 977 } 978 979 return (crc); 980} 981#endif 982 983uint32_t 984ether_crc32_be(const uint8_t *buf, size_t len) 985{ 986 size_t i; 987 uint32_t crc, carry; 988 int bit; 989 uint8_t data; 990 991 crc = 0xffffffff; /* initial value */ 992 993 for (i = 0; i < len; i++) { 994 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 995 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 996 crc <<= 1; 997 if (carry) 998 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 999 } 1000 } 1001 1002 return (crc); 1003} 1004 1005int 1006ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 1007{ 1008 struct ifaddr *ifa = (struct ifaddr *) data; 1009 struct ifreq *ifr = (struct ifreq *) data; 1010 int error = 0; 1011 1012 switch (command) { 1013 case SIOCSIFADDR: 1014 ifp->if_flags |= IFF_UP; 1015 1016 switch (ifa->ifa_addr->sa_family) { 1017#ifdef INET 1018 case AF_INET: 1019 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1020 arp_ifinit(ifp, ifa); 1021 break; 1022#endif 1023#ifdef IPX 1024 /* 1025 * XXX - This code is probably wrong 1026 */ 1027 case AF_IPX: 1028 { 1029 struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr); 1030 1031 if (ipx_nullhost(*ina)) 1032 ina->x_host = 1033 *(union ipx_host *) 1034 IF_LLADDR(ifp); 1035 else { 1036 bcopy((caddr_t) ina->x_host.c_host, 1037 (caddr_t) IF_LLADDR(ifp), 1038 ETHER_ADDR_LEN); 1039 } 1040 1041 /* 1042 * Set new address 1043 */ 1044 ifp->if_init(ifp->if_softc); 1045 break; 1046 } 1047#endif 1048 default: 1049 ifp->if_init(ifp->if_softc); 1050 break; 1051 } 1052 break; 1053 1054 case SIOCGIFADDR: 1055 { 1056 struct sockaddr *sa; 1057 1058 sa = (struct sockaddr *) & ifr->ifr_data; 1059 bcopy(IF_LLADDR(ifp), 1060 (caddr_t) sa->sa_data, ETHER_ADDR_LEN); 1061 } 1062 break; 1063 1064 case SIOCSIFMTU: 1065 /* 1066 * Set the interface MTU. 1067 */ 1068 if (ifr->ifr_mtu > ETHERMTU) { 1069 error = EINVAL; 1070 } else { 1071 ifp->if_mtu = ifr->ifr_mtu; 1072 } 1073 break; 1074 default: 1075 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1076 break; 1077 } 1078 return (error); 1079} 1080 1081static int 1082ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1083 struct sockaddr *sa) 1084{ 1085 struct sockaddr_dl *sdl; 1086#ifdef INET 1087 struct sockaddr_in *sin; 1088#endif 1089#ifdef INET6 1090 struct sockaddr_in6 *sin6; 1091#endif 1092 u_char *e_addr; 1093 1094 switch(sa->sa_family) { 1095 case AF_LINK: 1096 /* 1097 * No mapping needed. Just check that it's a valid MC address. 1098 */ 1099 sdl = (struct sockaddr_dl *)sa; 1100 e_addr = LLADDR(sdl); 1101 if (!ETHER_IS_MULTICAST(e_addr)) 1102 return EADDRNOTAVAIL; 1103 *llsa = 0; 1104 return 0; 1105 1106#ifdef INET 1107 case AF_INET: 1108 sin = (struct sockaddr_in *)sa; 1109 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1110 return EADDRNOTAVAIL; 1111 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1112 M_NOWAIT|M_ZERO); 1113 if (sdl == NULL) 1114 return ENOMEM; 1115 sdl->sdl_len = sizeof *sdl; 1116 sdl->sdl_family = AF_LINK; 1117 sdl->sdl_index = ifp->if_index; 1118 sdl->sdl_type = IFT_ETHER; 1119 sdl->sdl_alen = ETHER_ADDR_LEN; 1120 e_addr = LLADDR(sdl); 1121 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1122 *llsa = (struct sockaddr *)sdl; 1123 return 0; 1124#endif 1125#ifdef INET6 1126 case AF_INET6: 1127 sin6 = (struct sockaddr_in6 *)sa; 1128 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1129 /* 1130 * An IP6 address of 0 means listen to all 1131 * of the Ethernet multicast address used for IP6. 1132 * (This is used for multicast routers.) 1133 */ 1134 ifp->if_flags |= IFF_ALLMULTI; 1135 *llsa = 0; 1136 return 0; 1137 } 1138 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1139 return EADDRNOTAVAIL; 1140 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1141 M_NOWAIT|M_ZERO); 1142 if (sdl == NULL) 1143 return (ENOMEM); 1144 sdl->sdl_len = sizeof *sdl; 1145 sdl->sdl_family = AF_LINK; 1146 sdl->sdl_index = ifp->if_index; 1147 sdl->sdl_type = IFT_ETHER; 1148 sdl->sdl_alen = ETHER_ADDR_LEN; 1149 e_addr = LLADDR(sdl); 1150 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1151 *llsa = (struct sockaddr *)sdl; 1152 return 0; 1153#endif 1154 1155 default: 1156 /* 1157 * Well, the text isn't quite right, but it's the name 1158 * that counts... 1159 */ 1160 return EAFNOSUPPORT; 1161 } 1162} 1163 1164static void* 1165ether_alloc(u_char type, struct ifnet *ifp) 1166{ 1167 struct arpcom *ac; 1168 1169 ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO); 1170 ac->ac_ifp = ifp; 1171 1172 return (ac); 1173} 1174 1175static void 1176ether_free(void *com, u_char type) 1177{ 1178 1179 free(com, M_ARPCOM); 1180} 1181 1182static int 1183ether_modevent(module_t mod, int type, void *data) 1184{ 1185 1186 switch (type) { 1187 case MOD_LOAD: 1188 if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free); 1189 break; 1190 case MOD_UNLOAD: 1191 if_deregister_com_alloc(IFT_ETHER); 1192 break; 1193 default: 1194 return EOPNOTSUPP; 1195 } 1196 1197 return (0); 1198} 1199 1200static moduledata_t ether_mod = { 1201 "ether", 1202 ether_modevent, 1203 0 1204}; 1205 1206void 1207ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1208{ 1209 struct ether_vlan_header vlan; 1210 struct mbuf mv, mb; 1211 1212 KASSERT((m->m_flags & M_VLANTAG) != 0, 1213 ("%s: vlan information not present", __func__)); 1214 KASSERT(m->m_len >= sizeof(struct ether_header), 1215 ("%s: mbuf not large enough for header", __func__)); 1216 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1217 vlan.evl_proto = vlan.evl_encap_proto; 1218 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1219 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1220 m->m_len -= sizeof(struct ether_header); 1221 m->m_data += sizeof(struct ether_header); 1222 /* 1223 * If a data link has been supplied by the caller, then we will need to 1224 * re-create a stack allocated mbuf chain with the following structure: 1225 * 1226 * (1) mbuf #1 will contain the supplied data link 1227 * (2) mbuf #2 will contain the vlan header 1228 * (3) mbuf #3 will contain the original mbuf's packet data 1229 * 1230 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1231 */ 1232 if (data != NULL) { 1233 mv.m_next = m; 1234 mv.m_data = (caddr_t)&vlan; 1235 mv.m_len = sizeof(vlan); 1236 mb.m_next = &mv; 1237 mb.m_data = data; 1238 mb.m_len = dlen; 1239 bpf_mtap(bp, &mb); 1240 } else 1241 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1242 m->m_len += sizeof(struct ether_header); 1243 m->m_data -= sizeof(struct ether_header); 1244} 1245 1246DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1247MODULE_VERSION(ether, 1); 1248