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 1.61 2008/05/18 04:38:44 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 #include "opt_ethernet.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/globaldata.h> 49 #include <sys/kernel.h> 50 #include <sys/malloc.h> 51 #include <sys/mbuf.h> 52 #include <sys/msgport.h> 53 #include <sys/socket.h> 54 #include <sys/sockio.h> 55 #include <sys/sysctl.h> 56 #include <sys/thread.h> 57 #include <sys/thread2.h> 58 59 #include <net/if.h> 60 #include <net/netisr.h> 61 #include <net/route.h> 62 #include <net/if_llc.h> 63 #include <net/if_dl.h> 64 #include <net/if_types.h> 65 #include <net/ifq_var.h> 66 #include <net/bpf.h> 67 #include <net/ethernet.h> 68 #include <net/vlan/if_vlan_ether.h> 69 70 #if defined(INET) || defined(INET6) 71 #include <netinet/in.h> 72 #include <netinet/in_var.h> 73 #include <netinet/if_ether.h> 74 #include <net/ipfw/ip_fw.h> 75 #include <net/dummynet/ip_dummynet.h> 76 #endif 77 #ifdef INET6 78 #include <netinet6/nd6.h> 79 #endif 80 81 #ifdef CARP 82 #include <netinet/ip_carp.h> 83 #endif 84 85 #ifdef IPX 86 #include <netproto/ipx/ipx.h> 87 #include <netproto/ipx/ipx_if.h> 88 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 89 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 90 short *tp, int *hlen); 91 #endif 92 93 #ifdef NS 94 #include <netns/ns.h> 95 #include <netns/ns_if.h> 96 ushort ns_nettype; 97 int ether_outputdebug = 0; 98 int ether_inputdebug = 0; 99 #endif 100 101 #ifdef NETATALK 102 #include <netproto/atalk/at.h> 103 #include <netproto/atalk/at_var.h> 104 #include <netproto/atalk/at_extern.h> 105 106 #define llc_snap_org_code llc_un.type_snap.org_code 107 #define llc_snap_ether_type llc_un.type_snap.ether_type 108 109 extern u_char at_org_code[3]; 110 extern u_char aarp_org_code[3]; 111 #endif /* NETATALK */ 112 113 /* netgraph node hooks for ng_ether(4) */ 114 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 115 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 116 struct mbuf *m, const struct ether_header *eh); 117 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 118 void (*ng_ether_attach_p)(struct ifnet *ifp); 119 void (*ng_ether_detach_p)(struct ifnet *ifp); 120 121 int (*vlan_input_p)(struct mbuf *, struct mbuf_chain *); 122 123 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 124 struct rtentry *); 125 static void ether_restore_header(struct mbuf **, const struct ether_header *, 126 const struct ether_header *); 127 static void ether_demux_chain(struct ifnet *, struct mbuf *, 128 struct mbuf_chain *); 129 130 /* 131 * if_bridge support 132 */ 133 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 134 int (*bridge_output_p)(struct ifnet *, struct mbuf *, 135 struct sockaddr *, struct rtentry *); 136 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 137 138 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 139 struct sockaddr *); 140 141 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 142 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 143 }; 144 145 #define gotoerr(e) do { error = (e); goto bad; } while (0) 146 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 147 148 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 149 struct ip_fw **rule, 150 const struct ether_header *eh); 151 152 static int ether_ipfw; 153 static u_int ether_restore_hdr; 154 static u_int ether_prepend_hdr; 155 156 SYSCTL_DECL(_net_link); 157 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 158 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 159 ðer_ipfw, 0, "Pass ether pkts through firewall"); 160 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 161 ðer_restore_hdr, 0, "# of ether header restoration"); 162 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 163 ðer_prepend_hdr, 0, 164 "# of ether header restoration which prepends mbuf"); 165 166 /* 167 * Ethernet output routine. 168 * Encapsulate a packet of type family for the local net. 169 * Use trailer local net encapsulation if enough data in first 170 * packet leaves a multiple of 512 bytes of data in remainder. 171 * Assumes that ifp is actually pointer to arpcom structure. 172 */ 173 static int 174 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 175 struct rtentry *rt) 176 { 177 struct ether_header *eh, *deh; 178 u_char *edst; 179 int loop_copy = 0; 180 int hlen = ETHER_HDR_LEN; /* link layer header length */ 181 struct arpcom *ac = IFP2AC(ifp); 182 int error; 183 184 if (ifp->if_flags & IFF_MONITOR) 185 gotoerr(ENETDOWN); 186 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 187 gotoerr(ENETDOWN); 188 189 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 190 if (m == NULL) 191 return (ENOBUFS); 192 eh = mtod(m, struct ether_header *); 193 edst = eh->ether_dhost; 194 195 /* 196 * Fill in the destination ethernet address and frame type. 197 */ 198 switch (dst->sa_family) { 199 #ifdef INET 200 case AF_INET: 201 if (!arpresolve(ifp, rt, m, dst, edst)) 202 return (0); /* if not yet resolved */ 203 eh->ether_type = htons(ETHERTYPE_IP); 204 break; 205 #endif 206 #ifdef INET6 207 case AF_INET6: 208 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 209 return (0); /* Something bad happenned. */ 210 eh->ether_type = htons(ETHERTYPE_IPV6); 211 break; 212 #endif 213 #ifdef IPX 214 case AF_IPX: 215 if (ef_outputp != NULL) { 216 error = ef_outputp(ifp, &m, dst, &eh->ether_type, 217 &hlen); 218 if (error) 219 goto bad; 220 } else { 221 eh->ether_type = htons(ETHERTYPE_IPX); 222 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 223 edst, ETHER_ADDR_LEN); 224 } 225 break; 226 #endif 227 #ifdef NETATALK 228 case AF_APPLETALK: { 229 struct at_ifaddr *aa; 230 231 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) { 232 error = 0; /* XXX */ 233 goto bad; 234 } 235 /* 236 * In the phase 2 case, need to prepend an mbuf for 237 * the llc header. Since we must preserve the value 238 * of m, which is passed to us by value, we m_copy() 239 * the first mbuf, and use it for our llc header. 240 */ 241 if (aa->aa_flags & AFA_PHASE2) { 242 struct llc llc; 243 244 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT); 245 eh = mtod(m, struct ether_header *); 246 edst = eh->ether_dhost; 247 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 248 llc.llc_control = LLC_UI; 249 bcopy(at_org_code, llc.llc_snap_org_code, 250 sizeof at_org_code); 251 llc.llc_snap_ether_type = htons(ETHERTYPE_AT); 252 bcopy(&llc, 253 mtod(m, caddr_t) + sizeof(struct ether_header), 254 sizeof(struct llc)); 255 eh->ether_type = htons(m->m_pkthdr.len); 256 hlen = sizeof(struct llc) + ETHER_HDR_LEN; 257 } else { 258 eh->ether_type = htons(ETHERTYPE_AT); 259 } 260 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) 261 return (0); 262 break; 263 } 264 #endif 265 #ifdef NS 266 case AF_NS: 267 switch(ns_nettype) { 268 default: 269 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 270 eh->ether_type = 0x8137; 271 break; 272 case 0x0: /* Novell 802.3 */ 273 eh->ether_type = htons(m->m_pkthdr.len); 274 break; 275 case 0xe0e0: /* Novell 802.2 and Token-Ring */ 276 M_PREPEND(m, 3, MB_DONTWAIT); 277 eh = mtod(m, struct ether_header *); 278 edst = eh->ether_dhost; 279 eh->ether_type = htons(m->m_pkthdr.len); 280 cp = mtod(m, u_char *) + sizeof(struct ether_header); 281 *cp++ = 0xE0; 282 *cp++ = 0xE0; 283 *cp++ = 0x03; 284 break; 285 } 286 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst, 287 ETHER_ADDR_LEN); 288 /* 289 * XXX if ns_thishost is the same as the node's ethernet 290 * address then just the default code will catch this anyhow. 291 * So I'm not sure if this next clause should be here at all? 292 * [JRE] 293 */ 294 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) { 295 m->m_pkthdr.rcvif = ifp; 296 netisr_dispatch(NETISR_NS, m); 297 return (error); 298 } 299 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0) 300 m->m_flags |= M_BCAST; 301 break; 302 #endif 303 case pseudo_AF_HDRCMPLT: 304 case AF_UNSPEC: 305 loop_copy = -1; /* if this is for us, don't do it */ 306 deh = (struct ether_header *)dst->sa_data; 307 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 308 eh->ether_type = deh->ether_type; 309 break; 310 311 default: 312 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 313 gotoerr(EAFNOSUPPORT); 314 } 315 316 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 317 memcpy(eh->ether_shost, 318 ((struct ether_header *)dst->sa_data)->ether_shost, 319 ETHER_ADDR_LEN); 320 else 321 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 322 323 /* 324 * Bridges require special output handling. 325 */ 326 if (ifp->if_bridge) { 327 KASSERT(bridge_output_p != NULL, 328 ("%s: if_bridge not loaded!", __func__)); 329 lwkt_serialize_enter(ifp->if_serializer); 330 error = bridge_output_p(ifp, m, NULL, NULL); 331 lwkt_serialize_exit(ifp->if_serializer); 332 return error; 333 } 334 335 /* 336 * If a simplex interface, and the packet is being sent to our 337 * Ethernet address or a broadcast address, loopback a copy. 338 * XXX To make a simplex device behave exactly like a duplex 339 * device, we should copy in the case of sending to our own 340 * ethernet address (thus letting the original actually appear 341 * on the wire). However, we don't do that here for security 342 * reasons and compatibility with the original behavior. 343 */ 344 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 345 int csum_flags = 0; 346 347 if (m->m_pkthdr.csum_flags & CSUM_IP) 348 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 349 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 350 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 351 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 352 struct mbuf *n; 353 354 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 355 n->m_pkthdr.csum_flags |= csum_flags; 356 if (csum_flags & CSUM_DATA_VALID) 357 n->m_pkthdr.csum_data = 0xffff; 358 if_simloop(ifp, n, dst->sa_family, hlen); 359 } else 360 ifp->if_iqdrops++; 361 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 362 ETHER_ADDR_LEN) == 0) { 363 m->m_pkthdr.csum_flags |= csum_flags; 364 if (csum_flags & CSUM_DATA_VALID) 365 m->m_pkthdr.csum_data = 0xffff; 366 if_simloop(ifp, m, dst->sa_family, hlen); 367 return (0); /* XXX */ 368 } 369 } 370 371 #ifdef CARP 372 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) 373 goto bad; 374 #endif 375 376 377 /* Handle ng_ether(4) processing, if any */ 378 if (ng_ether_output_p != NULL) { 379 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) 380 goto bad; 381 if (m == NULL) 382 return (0); 383 } 384 385 /* Continue with link-layer output */ 386 return ether_output_frame(ifp, m); 387 388 bad: 389 m_freem(m); 390 return (error); 391 } 392 393 /* 394 * Ethernet link layer output routine to send a raw frame to the device. 395 * 396 * This assumes that the 14 byte Ethernet header is present and contiguous 397 * in the first mbuf. 398 */ 399 int 400 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 401 { 402 struct ip_fw *rule = NULL; 403 int error = 0; 404 struct altq_pktattr pktattr; 405 struct m_tag *mtag; 406 407 /* Extract info from dummynet tag */ 408 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 409 if (mtag != NULL) { 410 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 411 412 m_tag_delete(m, mtag); 413 mtag = NULL; 414 } 415 416 if (ifq_is_enabled(&ifp->if_snd)) 417 altq_etherclassify(&ifp->if_snd, m, &pktattr); 418 crit_enter(); 419 if (IPFW_LOADED && ether_ipfw != 0) { 420 struct ether_header save_eh, *eh; 421 422 eh = mtod(m, struct ether_header *); 423 save_eh = *eh; 424 m_adj(m, ETHER_HDR_LEN); 425 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 426 crit_exit(); 427 if (m != NULL) { 428 m_freem(m); 429 return ENOBUFS; /* pkt dropped */ 430 } else 431 return 0; /* consumed e.g. in a pipe */ 432 } 433 434 /* packet was ok, restore the ethernet header */ 435 ether_restore_header(&m, eh, &save_eh); 436 if (m == NULL) { 437 crit_exit(); 438 return ENOBUFS; 439 } 440 } 441 crit_exit(); 442 443 /* 444 * Queue message on interface, update output statistics if 445 * successful, and start output if interface not yet active. 446 */ 447 error = ifq_dispatch(ifp, m, &pktattr); 448 return (error); 449 } 450 451 /* 452 * ipfw processing for ethernet packets (in and out). 453 * The second parameter is NULL from ether_demux(), and ifp from 454 * ether_output_frame(). 455 */ 456 static boolean_t 457 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 458 const struct ether_header *eh) 459 { 460 struct ether_header save_eh = *eh; /* might be a ptr in m */ 461 struct ip_fw_args args; 462 struct m_tag *mtag; 463 int i; 464 465 if (*rule != NULL && fw_one_pass) 466 return TRUE; /* dummynet packet, already partially processed */ 467 468 /* 469 * I need some amount of data to be contiguous. 470 */ 471 i = min((*m0)->m_pkthdr.len, max_protohdr); 472 if ((*m0)->m_len < i) { 473 *m0 = m_pullup(*m0, i); 474 if (*m0 == NULL) 475 return FALSE; 476 } 477 478 args.m = *m0; /* the packet we are looking at */ 479 args.oif = dst; /* destination, if any */ 480 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 481 m_tag_delete(*m0, mtag); 482 args.rule = *rule; /* matching rule to restart */ 483 args.next_hop = NULL; /* we do not support forward yet */ 484 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 485 i = ip_fw_chk_ptr(&args); 486 *m0 = args.m; 487 *rule = args.rule; 488 489 if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */ 490 return FALSE; 491 492 if (i == 0) /* a PASS rule. */ 493 return TRUE; 494 495 if (i & IP_FW_PORT_DYNT_FLAG) { 496 /* 497 * Pass the pkt to dummynet, which consumes it. 498 */ 499 struct mbuf *m; 500 501 m = *m0; /* pass the original to dummynet */ 502 *m0 = NULL; /* and nothing back to the caller */ 503 504 ether_restore_header(&m, eh, &save_eh); 505 if (m == NULL) 506 return FALSE; 507 508 ip_fw_dn_io_ptr(m, (i & 0xffff), 509 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 510 return FALSE; 511 } 512 /* 513 * XXX at some point add support for divert/forward actions. 514 * If none of the above matches, we have to drop the pkt. 515 */ 516 return FALSE; 517 } 518 519 /* 520 * Process a received Ethernet packet. 521 * 522 * The ethernet header is assumed to be in the mbuf so the caller 523 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 524 * bytes in the first mbuf. 525 * 526 * This allows us to concentrate in one place a bunch of code which 527 * is replicated in all device drivers. Also, many functions called 528 * from ether_input() try to put the eh back into the mbuf, so we 529 * can later propagate the 'contiguous packet' interface to them. 530 * 531 * NOTA BENE: for all drivers "eh" is a pointer into the first mbuf or 532 * cluster, right before m_data. So be very careful when working on m, 533 * as you could destroy *eh !! 534 * 535 * First we perform any link layer operations, then continue to the 536 * upper layers with ether_demux(). 537 */ 538 void 539 ether_input_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain) 540 { 541 struct ether_header *eh; 542 543 ASSERT_SERIALIZED(ifp->if_serializer); 544 M_ASSERTPKTHDR(m); 545 546 /* Discard packet if interface is not up */ 547 if (!(ifp->if_flags & IFF_UP)) { 548 m_freem(m); 549 return; 550 } 551 552 if (m->m_len < sizeof(struct ether_header)) { 553 /* XXX error in the caller. */ 554 m_freem(m); 555 return; 556 } 557 eh = mtod(m, struct ether_header *); 558 559 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN && 560 (m->m_flags & M_VLANTAG) == 0) { 561 /* 562 * Extract vlan tag if hardware does not do it for us 563 */ 564 vlan_ether_decap(&m); 565 if (m == NULL) 566 return; 567 eh = mtod(m, struct ether_header *); 568 } 569 570 m->m_pkthdr.rcvif = ifp; 571 572 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 573 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 574 ifp->if_addrlen) == 0) 575 m->m_flags |= M_BCAST; 576 else 577 m->m_flags |= M_MCAST; 578 ifp->if_imcasts++; 579 } 580 581 ETHER_BPF_MTAP(ifp, m); 582 583 ifp->if_ibytes += m->m_pkthdr.len; 584 585 if (ifp->if_flags & IFF_MONITOR) { 586 /* 587 * Interface marked for monitoring; discard packet. 588 */ 589 m_freem(m); 590 return; 591 } 592 593 /* 594 * Tap the packet off here for a bridge. bridge_input() 595 * will return NULL if it has consumed the packet, otherwise 596 * it gets processed as normal. Note that bridge_input() 597 * will always return the original packet if we need to 598 * process it locally. 599 */ 600 if (ifp->if_bridge) { 601 KASSERT(bridge_input_p != NULL, 602 ("%s: if_bridge not loaded!", __func__)); 603 604 if(m->m_flags & M_PROTO1) { 605 m->m_flags &= ~M_PROTO1; 606 } else { 607 /* clear M_PROMISC, in case the packets comes from a vlan */ 608 /* m->m_flags &= ~M_PROMISC; */ 609 lwkt_serialize_exit(ifp->if_serializer); 610 m = (*bridge_input_p)(ifp, m); 611 lwkt_serialize_enter(ifp->if_serializer); 612 if (m == NULL) 613 return; 614 615 KASSERT(ifp == m->m_pkthdr.rcvif, 616 ("bridge_input_p changed rcvif\n")); 617 618 /* 'm' may be changed by bridge_input_p() */ 619 eh = mtod(m, struct ether_header *); 620 } 621 } 622 623 /* Handle ng_ether(4) processing, if any */ 624 if (ng_ether_input_p != NULL) { 625 ng_ether_input_p(ifp, &m); 626 if (m == NULL) 627 return; 628 629 /* 'm' may be changed by ng_ether_input_p() */ 630 eh = mtod(m, struct ether_header *); 631 } 632 633 /* Continue with upper layer processing */ 634 ether_demux_chain(ifp, m, chain); 635 } 636 637 void 638 ether_input(struct ifnet *ifp, struct mbuf *m) 639 { 640 ether_input_chain(ifp, m, NULL); 641 } 642 643 /* 644 * Upper layer processing for a received Ethernet packet. 645 */ 646 static void 647 ether_demux_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain) 648 { 649 struct ether_header save_eh, *eh; 650 int isr; 651 u_short ether_type; 652 struct ip_fw *rule = NULL; 653 struct m_tag *mtag; 654 #ifdef NETATALK 655 struct llc *l; 656 #endif 657 658 M_ASSERTPKTHDR(m); 659 KASSERT(m->m_len >= ETHER_HDR_LEN, 660 ("ether header is no contiguous!\n")); 661 662 eh = mtod(m, struct ether_header *); 663 save_eh = *eh; 664 665 /* XXX old crufty stuff, needs to be removed */ 666 m_adj(m, sizeof(struct ether_header)); 667 668 /* Extract info from dummynet tag */ 669 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 670 if (mtag != NULL) { 671 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 672 KKASSERT(ifp == NULL); 673 ifp = m->m_pkthdr.rcvif; 674 675 m_tag_delete(m, mtag); 676 mtag = NULL; 677 } 678 if (rule) /* packet is passing the second time */ 679 goto post_stats; 680 681 #ifdef CARP 682 /* 683 * XXX: Okay, we need to call carp_forus() and - if it is for 684 * us jump over code that does the normal check 685 * "ac_enaddr == ether_dhost". The check sequence is a bit 686 * different from OpenBSD, so we jump over as few code as 687 * possible, to catch _all_ sanity checks. This needs 688 * evaluation, to see if the carp ether_dhost values break any 689 * of these checks! 690 */ 691 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) 692 goto post_stats; 693 #endif 694 695 /* 696 * Discard packet if upper layers shouldn't see it because 697 * it was unicast to a different Ethernet address. If the 698 * driver is working properly, then this situation can only 699 * happen when the interface is in promiscuous mode. 700 */ 701 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 702 (eh->ether_dhost[0] & 1) == 0 && 703 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 704 m_freem(m); 705 return; 706 } 707 708 post_stats: 709 if (IPFW_LOADED && ether_ipfw != 0) { 710 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 711 m_freem(m); 712 return; 713 } 714 } 715 716 ether_type = ntohs(save_eh.ether_type); 717 718 if (m->m_flags & M_VLANTAG) { 719 if (ether_type == ETHERTYPE_VLAN) { 720 /* 721 * To prevent possible dangerous recursion, 722 * we don't do vlan-in-vlan 723 */ 724 m->m_pkthdr.rcvif->if_noproto++; 725 m_freem(m); 726 } 727 728 if (vlan_input_p != NULL) { 729 ether_restore_header(&m, eh, &save_eh); 730 if (m != NULL) 731 vlan_input_p(m, chain); 732 } else { 733 m->m_pkthdr.rcvif->if_noproto++; 734 m_freem(m); 735 } 736 return; 737 } 738 KKASSERT(ether_type != ETHERTYPE_VLAN); 739 740 switch (ether_type) { 741 #ifdef INET 742 case ETHERTYPE_IP: 743 if (ipflow_fastforward(m, ifp->if_serializer)) 744 return; 745 isr = NETISR_IP; 746 break; 747 748 case ETHERTYPE_ARP: 749 if (ifp->if_flags & IFF_NOARP) { 750 /* Discard packet if ARP is disabled on interface */ 751 m_freem(m); 752 return; 753 } 754 isr = NETISR_ARP; 755 break; 756 #endif 757 758 #ifdef INET6 759 case ETHERTYPE_IPV6: 760 isr = NETISR_IPV6; 761 break; 762 #endif 763 764 #ifdef IPX 765 case ETHERTYPE_IPX: 766 if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0) 767 return; 768 isr = NETISR_IPX; 769 break; 770 #endif 771 772 #ifdef NS 773 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 774 isr = NETISR_NS; 775 break; 776 777 #endif 778 779 #ifdef NETATALK 780 case ETHERTYPE_AT: 781 isr = NETISR_ATALK1; 782 break; 783 case ETHERTYPE_AARP: 784 isr = NETISR_AARP; 785 break; 786 #endif 787 788 default: 789 #ifdef IPX 790 if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0) 791 return; 792 #endif 793 #ifdef NS 794 checksum = mtod(m, ushort *); 795 /* Novell 802.3 */ 796 if ((ether_type <= ETHERMTU) && 797 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) { 798 if (*checksum == 0xE0E0) { 799 m->m_pkthdr.len -= 3; 800 m->m_len -= 3; 801 m->m_data += 3; 802 } 803 isr = NETISR_NS; 804 break; 805 } 806 #endif 807 #ifdef NETATALK 808 if (ether_type > ETHERMTU) 809 goto dropanyway; 810 l = mtod(m, struct llc *); 811 if (l->llc_dsap == LLC_SNAP_LSAP && 812 l->llc_ssap == LLC_SNAP_LSAP && 813 l->llc_control == LLC_UI) { 814 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 815 sizeof at_org_code) == 0 && 816 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 817 m_adj(m, sizeof(struct llc)); 818 isr = NETISR_ATALK2; 819 break; 820 } 821 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 822 sizeof aarp_org_code) == 0 && 823 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 824 m_adj(m, sizeof(struct llc)); 825 isr = NETISR_AARP; 826 break; 827 } 828 } 829 dropanyway: 830 #endif 831 if (ng_ether_input_orphan_p != NULL) 832 (*ng_ether_input_orphan_p)(ifp, m, &save_eh); 833 else 834 m_freem(m); 835 return; 836 } 837 838 #ifdef ETHER_INPUT_CHAIN 839 if (chain != NULL) { 840 struct mbuf_chain *c; 841 lwkt_port_t port; 842 int cpuid; 843 844 port = netisr_mport(isr, &m); 845 if (port == NULL) 846 return; 847 848 m->m_pkthdr.header = port; /* XXX */ 849 cpuid = port->mpu_td->td_gd->gd_cpuid; 850 851 c = &chain[cpuid]; 852 if (c->mc_head == NULL) { 853 c->mc_head = c->mc_tail = m; 854 } else { 855 c->mc_tail->m_nextpkt = m; 856 c->mc_tail = m; 857 } 858 m->m_nextpkt = NULL; 859 } else 860 #endif /* ETHER_INPUT_CHAIN */ 861 netisr_dispatch(isr, m); 862 } 863 864 void 865 ether_demux(struct ifnet *ifp, struct mbuf *m) 866 { 867 ether_demux_chain(ifp, m, NULL); 868 } 869 870 /* 871 * Perform common duties while attaching to interface list 872 */ 873 874 void 875 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 876 { 877 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 878 serializer); 879 } 880 881 void 882 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 883 lwkt_serialize_t serializer) 884 { 885 struct sockaddr_dl *sdl; 886 887 ifp->if_type = IFT_ETHER; 888 ifp->if_addrlen = ETHER_ADDR_LEN; 889 ifp->if_hdrlen = ETHER_HDR_LEN; 890 if_attach(ifp, serializer); 891 ifp->if_mtu = ETHERMTU; 892 if (ifp->if_baudrate == 0) 893 ifp->if_baudrate = 10000000; 894 ifp->if_output = ether_output; 895 ifp->if_input = ether_input; 896 ifp->if_resolvemulti = ether_resolvemulti; 897 ifp->if_broadcastaddr = etherbroadcastaddr; 898 sdl = IF_LLSOCKADDR(ifp); 899 sdl->sdl_type = IFT_ETHER; 900 sdl->sdl_alen = ifp->if_addrlen; 901 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 902 /* 903 * XXX Keep the current drivers happy. 904 * XXX Remove once all drivers have been cleaned up 905 */ 906 if (lla != IFP2AC(ifp)->ac_enaddr) 907 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 908 bpfattach(ifp, dlt, hdrlen); 909 if (ng_ether_attach_p != NULL) 910 (*ng_ether_attach_p)(ifp); 911 912 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 913 } 914 915 /* 916 * Perform common duties while detaching an Ethernet interface 917 */ 918 void 919 ether_ifdetach(struct ifnet *ifp) 920 { 921 if_down(ifp); 922 923 if (ng_ether_detach_p != NULL) 924 (*ng_ether_detach_p)(ifp); 925 bpfdetach(ifp); 926 if_detach(ifp); 927 } 928 929 int 930 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 931 { 932 struct ifaddr *ifa = (struct ifaddr *) data; 933 struct ifreq *ifr = (struct ifreq *) data; 934 int error = 0; 935 936 #define IF_INIT(ifp) \ 937 do { \ 938 if (((ifp)->if_flags & IFF_UP) == 0) { \ 939 (ifp)->if_flags |= IFF_UP; \ 940 (ifp)->if_init((ifp)->if_softc); \ 941 } \ 942 } while (0) 943 944 ASSERT_SERIALIZED(ifp->if_serializer); 945 946 switch (command) { 947 case SIOCSIFADDR: 948 switch (ifa->ifa_addr->sa_family) { 949 #ifdef INET 950 case AF_INET: 951 IF_INIT(ifp); /* before arpwhohas */ 952 arp_ifinit(ifp, ifa); 953 break; 954 #endif 955 #ifdef IPX 956 /* 957 * XXX - This code is probably wrong 958 */ 959 case AF_IPX: 960 { 961 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 962 struct arpcom *ac = IFP2AC(ifp); 963 964 if (ipx_nullhost(*ina)) 965 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 966 else 967 bcopy(ina->x_host.c_host, ac->ac_enaddr, 968 sizeof ac->ac_enaddr); 969 970 IF_INIT(ifp); /* Set new address. */ 971 break; 972 } 973 #endif 974 #ifdef NS 975 /* 976 * XXX - This code is probably wrong 977 */ 978 case AF_NS: 979 { 980 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); 981 struct arpcom *ac = IFP2AC(ifp); 982 983 if (ns_nullhost(*ina)) 984 ina->x_host = *(union ns_host *)(ac->ac_enaddr); 985 else 986 bcopy(ina->x_host.c_host, ac->ac_enaddr, 987 sizeof ac->ac_enaddr); 988 989 /* 990 * Set new address 991 */ 992 IF_INIT(ifp); 993 break; 994 } 995 #endif 996 default: 997 IF_INIT(ifp); 998 break; 999 } 1000 break; 1001 1002 case SIOCGIFADDR: 1003 bcopy(IFP2AC(ifp)->ac_enaddr, 1004 ((struct sockaddr *)ifr->ifr_data)->sa_data, 1005 ETHER_ADDR_LEN); 1006 break; 1007 1008 case SIOCSIFMTU: 1009 /* 1010 * Set the interface MTU. 1011 */ 1012 if (ifr->ifr_mtu > ETHERMTU) { 1013 error = EINVAL; 1014 } else { 1015 ifp->if_mtu = ifr->ifr_mtu; 1016 } 1017 break; 1018 default: 1019 error = EINVAL; 1020 break; 1021 } 1022 return (error); 1023 1024 #undef IF_INIT 1025 } 1026 1027 int 1028 ether_resolvemulti( 1029 struct ifnet *ifp, 1030 struct sockaddr **llsa, 1031 struct sockaddr *sa) 1032 { 1033 struct sockaddr_dl *sdl; 1034 struct sockaddr_in *sin; 1035 #ifdef INET6 1036 struct sockaddr_in6 *sin6; 1037 #endif 1038 u_char *e_addr; 1039 1040 switch(sa->sa_family) { 1041 case AF_LINK: 1042 /* 1043 * No mapping needed. Just check that it's a valid MC address. 1044 */ 1045 sdl = (struct sockaddr_dl *)sa; 1046 e_addr = LLADDR(sdl); 1047 if ((e_addr[0] & 1) != 1) 1048 return EADDRNOTAVAIL; 1049 *llsa = 0; 1050 return 0; 1051 1052 #ifdef INET 1053 case AF_INET: 1054 sin = (struct sockaddr_in *)sa; 1055 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1056 return EADDRNOTAVAIL; 1057 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1058 M_WAITOK | M_ZERO); 1059 sdl->sdl_len = sizeof *sdl; 1060 sdl->sdl_family = AF_LINK; 1061 sdl->sdl_index = ifp->if_index; 1062 sdl->sdl_type = IFT_ETHER; 1063 sdl->sdl_alen = ETHER_ADDR_LEN; 1064 e_addr = LLADDR(sdl); 1065 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1066 *llsa = (struct sockaddr *)sdl; 1067 return 0; 1068 #endif 1069 #ifdef INET6 1070 case AF_INET6: 1071 sin6 = (struct sockaddr_in6 *)sa; 1072 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1073 /* 1074 * An IP6 address of 0 means listen to all 1075 * of the Ethernet multicast address used for IP6. 1076 * (This is used for multicast routers.) 1077 */ 1078 ifp->if_flags |= IFF_ALLMULTI; 1079 *llsa = 0; 1080 return 0; 1081 } 1082 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1083 return EADDRNOTAVAIL; 1084 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 1085 M_WAITOK | M_ZERO); 1086 sdl->sdl_len = sizeof *sdl; 1087 sdl->sdl_family = AF_LINK; 1088 sdl->sdl_index = ifp->if_index; 1089 sdl->sdl_type = IFT_ETHER; 1090 sdl->sdl_alen = ETHER_ADDR_LEN; 1091 e_addr = LLADDR(sdl); 1092 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1093 *llsa = (struct sockaddr *)sdl; 1094 return 0; 1095 #endif 1096 1097 default: 1098 /* 1099 * Well, the text isn't quite right, but it's the name 1100 * that counts... 1101 */ 1102 return EAFNOSUPPORT; 1103 } 1104 } 1105 1106 #if 0 1107 /* 1108 * This is for reference. We have a table-driven version 1109 * of the little-endian crc32 generator, which is faster 1110 * than the double-loop. 1111 */ 1112 uint32_t 1113 ether_crc32_le(const uint8_t *buf, size_t len) 1114 { 1115 uint32_t c, crc, carry; 1116 size_t i, j; 1117 1118 crc = 0xffffffffU; /* initial value */ 1119 1120 for (i = 0; i < len; i++) { 1121 c = buf[i]; 1122 for (j = 0; j < 8; j++) { 1123 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1124 crc >>= 1; 1125 c >>= 1; 1126 if (carry) 1127 crc = (crc ^ ETHER_CRC_POLY_LE); 1128 } 1129 } 1130 1131 return (crc); 1132 } 1133 #else 1134 uint32_t 1135 ether_crc32_le(const uint8_t *buf, size_t len) 1136 { 1137 static const uint32_t crctab[] = { 1138 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1139 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1140 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1141 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1142 }; 1143 uint32_t crc; 1144 size_t i; 1145 1146 crc = 0xffffffffU; /* initial value */ 1147 1148 for (i = 0; i < len; i++) { 1149 crc ^= buf[i]; 1150 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1151 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1152 } 1153 1154 return (crc); 1155 } 1156 #endif 1157 1158 uint32_t 1159 ether_crc32_be(const uint8_t *buf, size_t len) 1160 { 1161 uint32_t c, crc, carry; 1162 size_t i, j; 1163 1164 crc = 0xffffffffU; /* initial value */ 1165 1166 for (i = 0; i < len; i++) { 1167 c = buf[i]; 1168 for (j = 0; j < 8; j++) { 1169 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1170 crc <<= 1; 1171 c >>= 1; 1172 if (carry) 1173 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1174 } 1175 } 1176 1177 return (crc); 1178 } 1179 1180 /* 1181 * find the size of ethernet header, and call classifier 1182 */ 1183 void 1184 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 1185 struct altq_pktattr *pktattr) 1186 { 1187 struct ether_header *eh; 1188 uint16_t ether_type; 1189 int hlen, af, hdrsize; 1190 caddr_t hdr; 1191 1192 hlen = sizeof(struct ether_header); 1193 eh = mtod(m, struct ether_header *); 1194 1195 ether_type = ntohs(eh->ether_type); 1196 if (ether_type < ETHERMTU) { 1197 /* ick! LLC/SNAP */ 1198 struct llc *llc = (struct llc *)(eh + 1); 1199 hlen += 8; 1200 1201 if (m->m_len < hlen || 1202 llc->llc_dsap != LLC_SNAP_LSAP || 1203 llc->llc_ssap != LLC_SNAP_LSAP || 1204 llc->llc_control != LLC_UI) 1205 goto bad; /* not snap! */ 1206 1207 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 1208 } 1209 1210 if (ether_type == ETHERTYPE_IP) { 1211 af = AF_INET; 1212 hdrsize = 20; /* sizeof(struct ip) */ 1213 #ifdef INET6 1214 } else if (ether_type == ETHERTYPE_IPV6) { 1215 af = AF_INET6; 1216 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 1217 #endif 1218 } else 1219 goto bad; 1220 1221 while (m->m_len <= hlen) { 1222 hlen -= m->m_len; 1223 m = m->m_next; 1224 } 1225 hdr = m->m_data + hlen; 1226 if (m->m_len < hlen + hdrsize) { 1227 /* 1228 * ip header is not in a single mbuf. this should not 1229 * happen in the current code. 1230 * (todo: use m_pulldown in the future) 1231 */ 1232 goto bad; 1233 } 1234 m->m_data += hlen; 1235 m->m_len -= hlen; 1236 ifq_classify(ifq, m, af, pktattr); 1237 m->m_data -= hlen; 1238 m->m_len += hlen; 1239 1240 return; 1241 1242 bad: 1243 pktattr->pattr_class = NULL; 1244 pktattr->pattr_hdr = NULL; 1245 pktattr->pattr_af = AF_UNSPEC; 1246 } 1247 1248 static void 1249 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 1250 const struct ether_header *save_eh) 1251 { 1252 struct mbuf *m = *m0; 1253 1254 ether_restore_hdr++; 1255 1256 /* 1257 * Prepend the header, optimize for the common case of 1258 * eh pointing into the mbuf. 1259 */ 1260 if ((const void *)(eh + 1) == (void *)m->m_data) { 1261 m->m_data -= ETHER_HDR_LEN; 1262 m->m_len += ETHER_HDR_LEN; 1263 m->m_pkthdr.len += ETHER_HDR_LEN; 1264 } else { 1265 ether_prepend_hdr++; 1266 1267 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1268 if (m != NULL) { 1269 bcopy(save_eh, mtod(m, struct ether_header *), 1270 ETHER_HDR_LEN); 1271 } 1272 } 1273 *m0 = m; 1274 } 1275 1276 #ifdef ETHER_INPUT_CHAIN 1277 1278 static void 1279 ether_input_ipifunc(void *arg) 1280 { 1281 struct mbuf *m, *next; 1282 lwkt_port_t port; 1283 1284 m = arg; 1285 do { 1286 next = m->m_nextpkt; 1287 m->m_nextpkt = NULL; 1288 1289 port = m->m_pkthdr.header; 1290 m->m_pkthdr.header = NULL; 1291 1292 lwkt_sendmsg(port, 1293 &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg); 1294 1295 m = next; 1296 } while (m != NULL); 1297 } 1298 1299 void 1300 ether_input_dispatch(struct mbuf_chain *chain) 1301 { 1302 #ifdef SMP 1303 int i; 1304 1305 for (i = 0; i < ncpus; ++i) { 1306 if (chain[i].mc_head != NULL) { 1307 lwkt_send_ipiq(globaldata_find(i), 1308 ether_input_ipifunc, chain[i].mc_head); 1309 } 1310 } 1311 #else 1312 ether_input_ipifunc(chain->mc_head); 1313 #endif 1314 } 1315 1316 #endif /* ETHER_INPUT_CHAIN */ 1317