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 * $FreeBSD$ 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_netgraph.h" 38 #include "opt_mbuf_profiling.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/kernel.h> 43 #include <sys/lock.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 #include <sys/uuid.h> 52 53 #include <net/if.h> 54 #include <net/if_var.h> 55 #include <net/if_arp.h> 56 #include <net/netisr.h> 57 #include <net/route.h> 58 #include <net/if_llc.h> 59 #include <net/if_dl.h> 60 #include <net/if_types.h> 61 #include <net/bpf.h> 62 #include <net/ethernet.h> 63 #include <net/if_bridgevar.h> 64 #include <net/if_vlan_var.h> 65 #include <net/if_llatbl.h> 66 #include <net/pfil.h> 67 #include <net/vnet.h> 68 69 #include <netpfil/pf/pf_mtag.h> 70 71 #if defined(INET) || defined(INET6) 72 #include <netinet/in.h> 73 #include <netinet/in_var.h> 74 #include <netinet/if_ether.h> 75 #include <netinet/ip_carp.h> 76 #include <netinet/ip_var.h> 77 #endif 78 #ifdef INET6 79 #include <netinet6/nd6.h> 80 #endif 81 82 #ifdef IPX 83 #include <netipx/ipx.h> 84 #include <netipx/ipx_if.h> 85 #endif 86 87 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 88 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, 89 const struct sockaddr *dst, short *tp, int *hlen); 90 91 #ifdef NETATALK 92 #include <netatalk/at.h> 93 #include <netatalk/at_var.h> 94 #include <netatalk/at_extern.h> 95 96 #define llc_snap_org_code llc_un.type_snap.org_code 97 #define llc_snap_ether_type llc_un.type_snap.ether_type 98 99 extern u_char at_org_code[3]; 100 extern u_char aarp_org_code[3]; 101 #endif /* NETATALK */ 102 103 #include <security/mac/mac_framework.h> 104 105 #ifdef CTASSERT 106 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 107 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 108 #endif 109 110 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 111 112 /* netgraph node hooks for ng_ether(4) */ 113 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 114 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 115 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 116 void (*ng_ether_attach_p)(struct ifnet *ifp); 117 void (*ng_ether_detach_p)(struct ifnet *ifp); 118 119 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 120 121 /* if_bridge(4) support */ 122 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 123 int (*bridge_output_p)(struct ifnet *, struct mbuf *, 124 struct sockaddr *, struct rtentry *); 125 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 126 127 /* if_lagg(4) support */ 128 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); 129 130 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 131 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 132 133 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 134 struct sockaddr *); 135 #ifdef VIMAGE 136 static void ether_reassign(struct ifnet *, struct vnet *, char *); 137 #endif 138 139 /* XXX: should be in an arp support file, not here */ 140 static MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals"); 141 142 #define ETHER_IS_BROADCAST(addr) \ 143 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 144 145 #define senderr(e) do { error = (e); goto bad;} while (0) 146 147 static void 148 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 149 { 150 int csum_flags = 0; 151 152 if (src->m_pkthdr.csum_flags & CSUM_IP) 153 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 154 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 155 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 156 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 157 csum_flags |= CSUM_SCTP_VALID; 158 dst->m_pkthdr.csum_flags |= csum_flags; 159 if (csum_flags & CSUM_DATA_VALID) 160 dst->m_pkthdr.csum_data = 0xffff; 161 } 162 163 /* 164 * Ethernet output routine. 165 * Encapsulate a packet of type family for the local net. 166 * Use trailer local net encapsulation if enough data in first 167 * packet leaves a multiple of 512 bytes of data in remainder. 168 */ 169 int 170 ether_output(struct ifnet *ifp, struct mbuf *m, 171 const struct sockaddr *dst, struct route *ro) 172 { 173 short type; 174 int error = 0, hdrcmplt = 0; 175 u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 176 struct llentry *lle = NULL; 177 struct rtentry *rt0 = NULL; 178 struct ether_header *eh; 179 struct pf_mtag *t; 180 int loop_copy = 1; 181 int hlen; /* link layer header length */ 182 183 if (ro != NULL) { 184 if (!(m->m_flags & (M_BCAST | M_MCAST))) 185 lle = ro->ro_lle; 186 rt0 = ro->ro_rt; 187 } 188 #ifdef MAC 189 error = mac_ifnet_check_transmit(ifp, m); 190 if (error) 191 senderr(error); 192 #endif 193 194 M_PROFILE(m); 195 if (ifp->if_flags & IFF_MONITOR) 196 senderr(ENETDOWN); 197 if (!((ifp->if_flags & IFF_UP) && 198 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 199 senderr(ENETDOWN); 200 201 hlen = ETHER_HDR_LEN; 202 switch (dst->sa_family) { 203 #ifdef INET 204 case AF_INET: 205 if (lle != NULL && (lle->la_flags & LLE_VALID)) 206 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 207 else 208 error = arpresolve(ifp, rt0, m, dst, edst, &lle); 209 if (error) 210 return (error == EWOULDBLOCK ? 0 : error); 211 type = htons(ETHERTYPE_IP); 212 break; 213 case AF_ARP: 214 { 215 struct arphdr *ah; 216 ah = mtod(m, struct arphdr *); 217 ah->ar_hrd = htons(ARPHRD_ETHER); 218 219 loop_copy = 0; /* if this is for us, don't do it */ 220 221 switch(ntohs(ah->ar_op)) { 222 case ARPOP_REVREQUEST: 223 case ARPOP_REVREPLY: 224 type = htons(ETHERTYPE_REVARP); 225 break; 226 case ARPOP_REQUEST: 227 case ARPOP_REPLY: 228 default: 229 type = htons(ETHERTYPE_ARP); 230 break; 231 } 232 233 if (m->m_flags & M_BCAST) 234 bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); 235 else 236 bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); 237 238 } 239 break; 240 #endif 241 #ifdef INET6 242 case AF_INET6: 243 if (lle != NULL && (lle->la_flags & LLE_VALID)) 244 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 245 else 246 error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle); 247 if (error) 248 return error; 249 type = htons(ETHERTYPE_IPV6); 250 break; 251 #endif 252 #ifdef IPX 253 case AF_IPX: 254 if (ef_outputp) { 255 error = ef_outputp(ifp, &m, dst, &type, &hlen); 256 if (error) 257 goto bad; 258 } else 259 type = htons(ETHERTYPE_IPX); 260 bcopy(&((const struct sockaddr_ipx *)dst)->sipx_addr.x_host, 261 edst, sizeof (edst)); 262 break; 263 #endif 264 #ifdef NETATALK 265 case AF_APPLETALK: 266 { 267 struct at_ifaddr *aa; 268 269 if ((aa = at_ifawithnet((const struct sockaddr_at *)dst)) == NULL) 270 senderr(EHOSTUNREACH); /* XXX */ 271 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 272 ifa_free(&aa->aa_ifa); 273 return (0); 274 } 275 /* 276 * In the phase 2 case, need to prepend an mbuf for the llc header. 277 */ 278 if ( aa->aa_flags & AFA_PHASE2 ) { 279 struct llc llc; 280 281 ifa_free(&aa->aa_ifa); 282 M_PREPEND(m, LLC_SNAPFRAMELEN, M_NOWAIT); 283 if (m == NULL) 284 senderr(ENOBUFS); 285 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 286 llc.llc_control = LLC_UI; 287 bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code)); 288 llc.llc_snap_ether_type = htons( ETHERTYPE_AT ); 289 bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN); 290 type = htons(m->m_pkthdr.len); 291 hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN; 292 } else { 293 ifa_free(&aa->aa_ifa); 294 type = htons(ETHERTYPE_AT); 295 } 296 break; 297 } 298 #endif /* NETATALK */ 299 300 case pseudo_AF_HDRCMPLT: 301 { 302 const struct ether_header *eh; 303 304 hdrcmplt = 1; 305 eh = (const struct ether_header *)dst->sa_data; 306 (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); 307 /* FALLTHROUGH */ 308 309 case AF_UNSPEC: 310 loop_copy = 0; /* if this is for us, don't do it */ 311 eh = (const struct ether_header *)dst->sa_data; 312 (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); 313 type = eh->ether_type; 314 break; 315 } 316 default: 317 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 318 senderr(EAFNOSUPPORT); 319 } 320 321 if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { 322 update_mbuf_csumflags(m, m); 323 return (if_simloop(ifp, m, dst->sa_family, 0)); 324 } 325 326 /* 327 * Add local net header. If no space in first mbuf, 328 * allocate another. 329 */ 330 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 331 if (m == NULL) 332 senderr(ENOBUFS); 333 eh = mtod(m, struct ether_header *); 334 (void)memcpy(&eh->ether_type, &type, 335 sizeof(eh->ether_type)); 336 (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); 337 if (hdrcmplt) 338 (void)memcpy(eh->ether_shost, esrc, 339 sizeof(eh->ether_shost)); 340 else 341 (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), 342 sizeof(eh->ether_shost)); 343 344 /* 345 * If a simplex interface, and the packet is being sent to our 346 * Ethernet address or a broadcast address, loopback a copy. 347 * XXX To make a simplex device behave exactly like a duplex 348 * device, we should copy in the case of sending to our own 349 * ethernet address (thus letting the original actually appear 350 * on the wire). However, we don't do that here for security 351 * reasons and compatibility with the original behavior. 352 */ 353 if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && 354 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 355 if (m->m_flags & M_BCAST) { 356 struct mbuf *n; 357 358 /* 359 * Because if_simloop() modifies the packet, we need a 360 * writable copy through m_dup() instead of a readonly 361 * one as m_copy[m] would give us. The alternative would 362 * be to modify if_simloop() to handle the readonly mbuf, 363 * but performancewise it is mostly equivalent (trading 364 * extra data copying vs. extra locking). 365 * 366 * XXX This is a local workaround. A number of less 367 * often used kernel parts suffer from the same bug. 368 * See PR kern/105943 for a proposed general solution. 369 */ 370 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 371 update_mbuf_csumflags(m, n); 372 (void)if_simloop(ifp, n, dst->sa_family, hlen); 373 } else 374 ifp->if_iqdrops++; 375 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 376 ETHER_ADDR_LEN) == 0) { 377 update_mbuf_csumflags(m, m); 378 (void) if_simloop(ifp, m, dst->sa_family, hlen); 379 return (0); /* XXX */ 380 } 381 } 382 383 /* 384 * Bridges require special output handling. 385 */ 386 if (ifp->if_bridge) { 387 BRIDGE_OUTPUT(ifp, m, error); 388 return (error); 389 } 390 391 #if defined(INET) || defined(INET6) 392 if (ifp->if_carp && 393 (error = (*carp_output_p)(ifp, m, dst))) 394 goto bad; 395 #endif 396 397 /* Handle ng_ether(4) processing, if any */ 398 if (IFP2AC(ifp)->ac_netgraph != NULL) { 399 KASSERT(ng_ether_output_p != NULL, 400 ("ng_ether_output_p is NULL")); 401 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 402 bad: if (m != NULL) 403 m_freem(m); 404 return (error); 405 } 406 if (m == NULL) 407 return (0); 408 } 409 410 /* Continue with link-layer output */ 411 return ether_output_frame(ifp, m); 412 } 413 414 /* 415 * Ethernet link layer output routine to send a raw frame to the device. 416 * 417 * This assumes that the 14 byte Ethernet header is present and contiguous 418 * in the first mbuf (if BRIDGE'ing). 419 */ 420 int 421 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 422 { 423 int i; 424 425 if (PFIL_HOOKED(&V_link_pfil_hook)) { 426 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); 427 428 if (i != 0) 429 return (EACCES); 430 431 if (m == NULL) 432 return (0); 433 } 434 435 /* 436 * Queue message on interface, update output statistics if 437 * successful, and start output if interface not yet active. 438 */ 439 return ((ifp->if_transmit)(ifp, m)); 440 } 441 442 #if defined(INET) || defined(INET6) 443 #endif 444 445 /* 446 * Process a received Ethernet packet; the packet is in the 447 * mbuf chain m with the ethernet header at the front. 448 */ 449 static void 450 ether_input_internal(struct ifnet *ifp, struct mbuf *m) 451 { 452 struct ether_header *eh; 453 u_short etype; 454 455 if ((ifp->if_flags & IFF_UP) == 0) { 456 m_freem(m); 457 return; 458 } 459 #ifdef DIAGNOSTIC 460 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 461 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 462 m_freem(m); 463 return; 464 } 465 #endif 466 /* 467 * Do consistency checks to verify assumptions 468 * made by code past this point. 469 */ 470 if ((m->m_flags & M_PKTHDR) == 0) { 471 if_printf(ifp, "discard frame w/o packet header\n"); 472 ifp->if_ierrors++; 473 m_freem(m); 474 return; 475 } 476 if (m->m_len < ETHER_HDR_LEN) { 477 /* XXX maybe should pullup? */ 478 if_printf(ifp, "discard frame w/o leading ethernet " 479 "header (len %u pkt len %u)\n", 480 m->m_len, m->m_pkthdr.len); 481 ifp->if_ierrors++; 482 m_freem(m); 483 return; 484 } 485 eh = mtod(m, struct ether_header *); 486 etype = ntohs(eh->ether_type); 487 if (m->m_pkthdr.rcvif == NULL) { 488 if_printf(ifp, "discard frame w/o interface pointer\n"); 489 ifp->if_ierrors++; 490 m_freem(m); 491 return; 492 } 493 #ifdef DIAGNOSTIC 494 if (m->m_pkthdr.rcvif != ifp) { 495 if_printf(ifp, "Warning, frame marked as received on %s\n", 496 m->m_pkthdr.rcvif->if_xname); 497 } 498 #endif 499 500 CURVNET_SET_QUIET(ifp->if_vnet); 501 502 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 503 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 504 m->m_flags |= M_BCAST; 505 else 506 m->m_flags |= M_MCAST; 507 ifp->if_imcasts++; 508 } 509 510 #ifdef MAC 511 /* 512 * Tag the mbuf with an appropriate MAC label before any other 513 * consumers can get to it. 514 */ 515 mac_ifnet_create_mbuf(ifp, m); 516 #endif 517 518 /* 519 * Give bpf a chance at the packet. 520 */ 521 ETHER_BPF_MTAP(ifp, m); 522 523 /* 524 * If the CRC is still on the packet, trim it off. We do this once 525 * and once only in case we are re-entered. Nothing else on the 526 * Ethernet receive path expects to see the FCS. 527 */ 528 if (m->m_flags & M_HASFCS) { 529 m_adj(m, -ETHER_CRC_LEN); 530 m->m_flags &= ~M_HASFCS; 531 } 532 533 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 534 ifp->if_ibytes += m->m_pkthdr.len; 535 536 /* Allow monitor mode to claim this frame, after stats are updated. */ 537 if (ifp->if_flags & IFF_MONITOR) { 538 m_freem(m); 539 CURVNET_RESTORE(); 540 return; 541 } 542 543 /* Handle input from a lagg(4) port */ 544 if (ifp->if_type == IFT_IEEE8023ADLAG) { 545 KASSERT(lagg_input_p != NULL, 546 ("%s: if_lagg not loaded!", __func__)); 547 m = (*lagg_input_p)(ifp, m); 548 if (m != NULL) 549 ifp = m->m_pkthdr.rcvif; 550 else { 551 CURVNET_RESTORE(); 552 return; 553 } 554 } 555 556 /* 557 * If the hardware did not process an 802.1Q tag, do this now, 558 * to allow 802.1P priority frames to be passed to the main input 559 * path correctly. 560 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 561 */ 562 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 563 struct ether_vlan_header *evl; 564 565 if (m->m_len < sizeof(*evl) && 566 (m = m_pullup(m, sizeof(*evl))) == NULL) { 567 #ifdef DIAGNOSTIC 568 if_printf(ifp, "cannot pullup VLAN header\n"); 569 #endif 570 ifp->if_ierrors++; 571 m_freem(m); 572 CURVNET_RESTORE(); 573 return; 574 } 575 576 evl = mtod(m, struct ether_vlan_header *); 577 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 578 m->m_flags |= M_VLANTAG; 579 580 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 581 ETHER_HDR_LEN - ETHER_TYPE_LEN); 582 m_adj(m, ETHER_VLAN_ENCAP_LEN); 583 eh = mtod(m, struct ether_header *); 584 } 585 586 M_SETFIB(m, ifp->if_fib); 587 588 /* Allow ng_ether(4) to claim this frame. */ 589 if (IFP2AC(ifp)->ac_netgraph != NULL) { 590 KASSERT(ng_ether_input_p != NULL, 591 ("%s: ng_ether_input_p is NULL", __func__)); 592 m->m_flags &= ~M_PROMISC; 593 (*ng_ether_input_p)(ifp, &m); 594 if (m == NULL) { 595 CURVNET_RESTORE(); 596 return; 597 } 598 eh = mtod(m, struct ether_header *); 599 } 600 601 /* 602 * Allow if_bridge(4) to claim this frame. 603 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 604 * and the frame should be delivered locally. 605 */ 606 if (ifp->if_bridge != NULL) { 607 m->m_flags &= ~M_PROMISC; 608 BRIDGE_INPUT(ifp, m); 609 if (m == NULL) { 610 CURVNET_RESTORE(); 611 return; 612 } 613 eh = mtod(m, struct ether_header *); 614 } 615 616 #if defined(INET) || defined(INET6) 617 /* 618 * Clear M_PROMISC on frame so that carp(4) will see it when the 619 * mbuf flows up to Layer 3. 620 * FreeBSD's implementation of carp(4) uses the inprotosw 621 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 622 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 623 * is outside the scope of the M_PROMISC test below. 624 * TODO: Maintain a hash table of ethernet addresses other than 625 * ether_dhost which may be active on this ifp. 626 */ 627 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 628 m->m_flags &= ~M_PROMISC; 629 } else 630 #endif 631 { 632 /* 633 * If the frame received was not for our MAC address, set the 634 * M_PROMISC flag on the mbuf chain. The frame may need to 635 * be seen by the rest of the Ethernet input path in case of 636 * re-entry (e.g. bridge, vlan, netgraph) but should not be 637 * seen by upper protocol layers. 638 */ 639 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 640 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 641 m->m_flags |= M_PROMISC; 642 } 643 644 if (harvest.ethernet) 645 random_harvest(&(m->m_data), 12, 2, RANDOM_NET_ETHER); 646 647 ether_demux(ifp, m); 648 CURVNET_RESTORE(); 649 } 650 651 /* 652 * Ethernet input dispatch; by default, direct dispatch here regardless of 653 * global configuration. 654 */ 655 static void 656 ether_nh_input(struct mbuf *m) 657 { 658 659 ether_input_internal(m->m_pkthdr.rcvif, m); 660 } 661 662 static struct netisr_handler ether_nh = { 663 .nh_name = "ether", 664 .nh_handler = ether_nh_input, 665 .nh_proto = NETISR_ETHER, 666 .nh_policy = NETISR_POLICY_SOURCE, 667 .nh_dispatch = NETISR_DISPATCH_DIRECT, 668 }; 669 670 static void 671 ether_init(__unused void *arg) 672 { 673 674 netisr_register(ðer_nh); 675 } 676 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 677 678 static void 679 vnet_ether_init(__unused void *arg) 680 { 681 int i; 682 683 /* Initialize packet filter hooks. */ 684 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 685 V_link_pfil_hook.ph_af = AF_LINK; 686 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 687 printf("%s: WARNING: unable to register pfil link hook, " 688 "error %d\n", __func__, i); 689 } 690 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 691 vnet_ether_init, NULL); 692 693 static void 694 vnet_ether_destroy(__unused void *arg) 695 { 696 int i; 697 698 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 699 printf("%s: WARNING: unable to unregister pfil link hook, " 700 "error %d\n", __func__, i); 701 } 702 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 703 vnet_ether_destroy, NULL); 704 705 706 707 static void 708 ether_input(struct ifnet *ifp, struct mbuf *m) 709 { 710 711 /* 712 * We will rely on rcvif being set properly in the deferred context, 713 * so assert it is correct here. 714 */ 715 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__)); 716 717 netisr_dispatch(NETISR_ETHER, m); 718 } 719 720 /* 721 * Upper layer processing for a received Ethernet packet. 722 */ 723 void 724 ether_demux(struct ifnet *ifp, struct mbuf *m) 725 { 726 struct ether_header *eh; 727 int i, isr; 728 u_short ether_type; 729 #if defined(NETATALK) 730 struct llc *l; 731 #endif 732 733 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 734 735 /* Do not grab PROMISC frames in case we are re-entered. */ 736 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 737 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); 738 739 if (i != 0 || m == NULL) 740 return; 741 } 742 743 eh = mtod(m, struct ether_header *); 744 ether_type = ntohs(eh->ether_type); 745 746 /* 747 * If this frame has a VLAN tag other than 0, call vlan_input() 748 * if its module is loaded. Otherwise, drop. 749 */ 750 if ((m->m_flags & M_VLANTAG) && 751 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 752 if (ifp->if_vlantrunk == NULL) { 753 ifp->if_noproto++; 754 m_freem(m); 755 return; 756 } 757 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 758 __func__)); 759 /* Clear before possibly re-entering ether_input(). */ 760 m->m_flags &= ~M_PROMISC; 761 (*vlan_input_p)(ifp, m); 762 return; 763 } 764 765 /* 766 * Pass promiscuously received frames to the upper layer if the user 767 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 768 */ 769 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 770 m_freem(m); 771 return; 772 } 773 774 /* 775 * Reset layer specific mbuf flags to avoid confusing upper layers. 776 * Strip off Ethernet header. 777 */ 778 m->m_flags &= ~M_VLANTAG; 779 m_clrprotoflags(m); 780 m_adj(m, ETHER_HDR_LEN); 781 782 /* 783 * Dispatch frame to upper layer. 784 */ 785 switch (ether_type) { 786 #ifdef INET 787 case ETHERTYPE_IP: 788 if ((m = ip_fastforward(m)) == NULL) 789 return; 790 isr = NETISR_IP; 791 break; 792 793 case ETHERTYPE_ARP: 794 if (ifp->if_flags & IFF_NOARP) { 795 /* Discard packet if ARP is disabled on interface */ 796 m_freem(m); 797 return; 798 } 799 isr = NETISR_ARP; 800 break; 801 #endif 802 #ifdef IPX 803 case ETHERTYPE_IPX: 804 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 805 return; 806 isr = NETISR_IPX; 807 break; 808 #endif 809 #ifdef INET6 810 case ETHERTYPE_IPV6: 811 isr = NETISR_IPV6; 812 break; 813 #endif 814 #ifdef NETATALK 815 case ETHERTYPE_AT: 816 isr = NETISR_ATALK1; 817 break; 818 case ETHERTYPE_AARP: 819 isr = NETISR_AARP; 820 break; 821 #endif /* NETATALK */ 822 default: 823 #ifdef IPX 824 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 825 return; 826 #endif /* IPX */ 827 #if defined(NETATALK) 828 if (ether_type > ETHERMTU) 829 goto discard; 830 l = mtod(m, struct llc *); 831 if (l->llc_dsap == LLC_SNAP_LSAP && 832 l->llc_ssap == LLC_SNAP_LSAP && 833 l->llc_control == LLC_UI) { 834 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 835 sizeof(at_org_code)) == 0 && 836 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 837 m_adj(m, LLC_SNAPFRAMELEN); 838 isr = NETISR_ATALK2; 839 break; 840 } 841 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 842 sizeof(aarp_org_code)) == 0 && 843 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 844 m_adj(m, LLC_SNAPFRAMELEN); 845 isr = NETISR_AARP; 846 break; 847 } 848 } 849 #endif /* NETATALK */ 850 goto discard; 851 } 852 netisr_dispatch(isr, m); 853 return; 854 855 discard: 856 /* 857 * Packet is to be discarded. If netgraph is present, 858 * hand the packet to it for last chance processing; 859 * otherwise dispose of it. 860 */ 861 if (IFP2AC(ifp)->ac_netgraph != NULL) { 862 KASSERT(ng_ether_input_orphan_p != NULL, 863 ("ng_ether_input_orphan_p is NULL")); 864 /* 865 * Put back the ethernet header so netgraph has a 866 * consistent view of inbound packets. 867 */ 868 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 869 (*ng_ether_input_orphan_p)(ifp, m); 870 return; 871 } 872 m_freem(m); 873 } 874 875 /* 876 * Convert Ethernet address to printable (loggable) representation. 877 * This routine is for compatibility; it's better to just use 878 * 879 * printf("%6D", <pointer to address>, ":"); 880 * 881 * since there's no static buffer involved. 882 */ 883 char * 884 ether_sprintf(const u_char *ap) 885 { 886 static char etherbuf[18]; 887 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 888 return (etherbuf); 889 } 890 891 /* 892 * Perform common duties while attaching to interface list 893 */ 894 void 895 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 896 { 897 int i; 898 struct ifaddr *ifa; 899 struct sockaddr_dl *sdl; 900 901 ifp->if_addrlen = ETHER_ADDR_LEN; 902 ifp->if_hdrlen = ETHER_HDR_LEN; 903 if_attach(ifp); 904 ifp->if_mtu = ETHERMTU; 905 ifp->if_output = ether_output; 906 ifp->if_input = ether_input; 907 ifp->if_resolvemulti = ether_resolvemulti; 908 #ifdef VIMAGE 909 ifp->if_reassign = ether_reassign; 910 #endif 911 if (ifp->if_baudrate == 0) 912 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 913 ifp->if_broadcastaddr = etherbroadcastaddr; 914 915 ifa = ifp->if_addr; 916 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 917 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 918 sdl->sdl_type = IFT_ETHER; 919 sdl->sdl_alen = ifp->if_addrlen; 920 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 921 922 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 923 if (ng_ether_attach_p != NULL) 924 (*ng_ether_attach_p)(ifp); 925 926 /* Announce Ethernet MAC address if non-zero. */ 927 for (i = 0; i < ifp->if_addrlen; i++) 928 if (lla[i] != 0) 929 break; 930 if (i != ifp->if_addrlen) 931 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 932 933 uuid_ether_add(LLADDR(sdl)); 934 } 935 936 /* 937 * Perform common duties while detaching an Ethernet interface 938 */ 939 void 940 ether_ifdetach(struct ifnet *ifp) 941 { 942 struct sockaddr_dl *sdl; 943 944 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 945 uuid_ether_del(LLADDR(sdl)); 946 947 if (IFP2AC(ifp)->ac_netgraph != NULL) { 948 KASSERT(ng_ether_detach_p != NULL, 949 ("ng_ether_detach_p is NULL")); 950 (*ng_ether_detach_p)(ifp); 951 } 952 953 bpfdetach(ifp); 954 if_detach(ifp); 955 } 956 957 #ifdef VIMAGE 958 void 959 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 960 { 961 962 if (IFP2AC(ifp)->ac_netgraph != NULL) { 963 KASSERT(ng_ether_detach_p != NULL, 964 ("ng_ether_detach_p is NULL")); 965 (*ng_ether_detach_p)(ifp); 966 } 967 968 if (ng_ether_attach_p != NULL) { 969 CURVNET_SET_QUIET(new_vnet); 970 (*ng_ether_attach_p)(ifp); 971 CURVNET_RESTORE(); 972 } 973 } 974 #endif 975 976 SYSCTL_DECL(_net_link); 977 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 978 979 #if 0 980 /* 981 * This is for reference. We have a table-driven version 982 * of the little-endian crc32 generator, which is faster 983 * than the double-loop. 984 */ 985 uint32_t 986 ether_crc32_le(const uint8_t *buf, size_t len) 987 { 988 size_t i; 989 uint32_t crc; 990 int bit; 991 uint8_t data; 992 993 crc = 0xffffffff; /* initial value */ 994 995 for (i = 0; i < len; i++) { 996 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 997 carry = (crc ^ data) & 1; 998 crc >>= 1; 999 if (carry) 1000 crc = (crc ^ ETHER_CRC_POLY_LE); 1001 } 1002 } 1003 1004 return (crc); 1005 } 1006 #else 1007 uint32_t 1008 ether_crc32_le(const uint8_t *buf, size_t len) 1009 { 1010 static const uint32_t crctab[] = { 1011 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1012 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1013 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1014 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1015 }; 1016 size_t i; 1017 uint32_t crc; 1018 1019 crc = 0xffffffff; /* initial value */ 1020 1021 for (i = 0; i < len; i++) { 1022 crc ^= buf[i]; 1023 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1024 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1025 } 1026 1027 return (crc); 1028 } 1029 #endif 1030 1031 uint32_t 1032 ether_crc32_be(const uint8_t *buf, size_t len) 1033 { 1034 size_t i; 1035 uint32_t crc, carry; 1036 int bit; 1037 uint8_t data; 1038 1039 crc = 0xffffffff; /* initial value */ 1040 1041 for (i = 0; i < len; i++) { 1042 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1043 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1044 crc <<= 1; 1045 if (carry) 1046 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1047 } 1048 } 1049 1050 return (crc); 1051 } 1052 1053 int 1054 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1055 { 1056 struct ifaddr *ifa = (struct ifaddr *) data; 1057 struct ifreq *ifr = (struct ifreq *) data; 1058 int error = 0; 1059 1060 switch (command) { 1061 case SIOCSIFADDR: 1062 ifp->if_flags |= IFF_UP; 1063 1064 switch (ifa->ifa_addr->sa_family) { 1065 #ifdef INET 1066 case AF_INET: 1067 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1068 arp_ifinit(ifp, ifa); 1069 break; 1070 #endif 1071 #ifdef IPX 1072 /* 1073 * XXX - This code is probably wrong 1074 */ 1075 case AF_IPX: 1076 { 1077 struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr); 1078 1079 if (ipx_nullhost(*ina)) 1080 ina->x_host = 1081 *(union ipx_host *) 1082 IF_LLADDR(ifp); 1083 else { 1084 bcopy((caddr_t) ina->x_host.c_host, 1085 (caddr_t) IF_LLADDR(ifp), 1086 ETHER_ADDR_LEN); 1087 } 1088 1089 /* 1090 * Set new address 1091 */ 1092 ifp->if_init(ifp->if_softc); 1093 break; 1094 } 1095 #endif 1096 default: 1097 ifp->if_init(ifp->if_softc); 1098 break; 1099 } 1100 break; 1101 1102 case SIOCGIFADDR: 1103 { 1104 struct sockaddr *sa; 1105 1106 sa = (struct sockaddr *) & ifr->ifr_data; 1107 bcopy(IF_LLADDR(ifp), 1108 (caddr_t) sa->sa_data, ETHER_ADDR_LEN); 1109 } 1110 break; 1111 1112 case SIOCSIFMTU: 1113 /* 1114 * Set the interface MTU. 1115 */ 1116 if (ifr->ifr_mtu > ETHERMTU) { 1117 error = EINVAL; 1118 } else { 1119 ifp->if_mtu = ifr->ifr_mtu; 1120 } 1121 break; 1122 default: 1123 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1124 break; 1125 } 1126 return (error); 1127 } 1128 1129 static int 1130 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1131 struct sockaddr *sa) 1132 { 1133 struct sockaddr_dl *sdl; 1134 #ifdef INET 1135 struct sockaddr_in *sin; 1136 #endif 1137 #ifdef INET6 1138 struct sockaddr_in6 *sin6; 1139 #endif 1140 u_char *e_addr; 1141 1142 switch(sa->sa_family) { 1143 case AF_LINK: 1144 /* 1145 * No mapping needed. Just check that it's a valid MC address. 1146 */ 1147 sdl = (struct sockaddr_dl *)sa; 1148 e_addr = LLADDR(sdl); 1149 if (!ETHER_IS_MULTICAST(e_addr)) 1150 return EADDRNOTAVAIL; 1151 *llsa = 0; 1152 return 0; 1153 1154 #ifdef INET 1155 case AF_INET: 1156 sin = (struct sockaddr_in *)sa; 1157 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1158 return EADDRNOTAVAIL; 1159 sdl = malloc(sizeof *sdl, M_IFMADDR, 1160 M_NOWAIT|M_ZERO); 1161 if (sdl == NULL) 1162 return ENOMEM; 1163 sdl->sdl_len = sizeof *sdl; 1164 sdl->sdl_family = AF_LINK; 1165 sdl->sdl_index = ifp->if_index; 1166 sdl->sdl_type = IFT_ETHER; 1167 sdl->sdl_alen = ETHER_ADDR_LEN; 1168 e_addr = LLADDR(sdl); 1169 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1170 *llsa = (struct sockaddr *)sdl; 1171 return 0; 1172 #endif 1173 #ifdef INET6 1174 case AF_INET6: 1175 sin6 = (struct sockaddr_in6 *)sa; 1176 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1177 /* 1178 * An IP6 address of 0 means listen to all 1179 * of the Ethernet multicast address used for IP6. 1180 * (This is used for multicast routers.) 1181 */ 1182 ifp->if_flags |= IFF_ALLMULTI; 1183 *llsa = 0; 1184 return 0; 1185 } 1186 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1187 return EADDRNOTAVAIL; 1188 sdl = malloc(sizeof *sdl, M_IFMADDR, 1189 M_NOWAIT|M_ZERO); 1190 if (sdl == NULL) 1191 return (ENOMEM); 1192 sdl->sdl_len = sizeof *sdl; 1193 sdl->sdl_family = AF_LINK; 1194 sdl->sdl_index = ifp->if_index; 1195 sdl->sdl_type = IFT_ETHER; 1196 sdl->sdl_alen = ETHER_ADDR_LEN; 1197 e_addr = LLADDR(sdl); 1198 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1199 *llsa = (struct sockaddr *)sdl; 1200 return 0; 1201 #endif 1202 1203 default: 1204 /* 1205 * Well, the text isn't quite right, but it's the name 1206 * that counts... 1207 */ 1208 return EAFNOSUPPORT; 1209 } 1210 } 1211 1212 static void* 1213 ether_alloc(u_char type, struct ifnet *ifp) 1214 { 1215 struct arpcom *ac; 1216 1217 ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO); 1218 ac->ac_ifp = ifp; 1219 1220 return (ac); 1221 } 1222 1223 static void 1224 ether_free(void *com, u_char type) 1225 { 1226 1227 free(com, M_ARPCOM); 1228 } 1229 1230 static int 1231 ether_modevent(module_t mod, int type, void *data) 1232 { 1233 1234 switch (type) { 1235 case MOD_LOAD: 1236 if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free); 1237 break; 1238 case MOD_UNLOAD: 1239 if_deregister_com_alloc(IFT_ETHER); 1240 break; 1241 default: 1242 return EOPNOTSUPP; 1243 } 1244 1245 return (0); 1246 } 1247 1248 static moduledata_t ether_mod = { 1249 "ether", 1250 ether_modevent, 1251 0 1252 }; 1253 1254 void 1255 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1256 { 1257 struct ether_vlan_header vlan; 1258 struct mbuf mv, mb; 1259 1260 KASSERT((m->m_flags & M_VLANTAG) != 0, 1261 ("%s: vlan information not present", __func__)); 1262 KASSERT(m->m_len >= sizeof(struct ether_header), 1263 ("%s: mbuf not large enough for header", __func__)); 1264 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1265 vlan.evl_proto = vlan.evl_encap_proto; 1266 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1267 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1268 m->m_len -= sizeof(struct ether_header); 1269 m->m_data += sizeof(struct ether_header); 1270 /* 1271 * If a data link has been supplied by the caller, then we will need to 1272 * re-create a stack allocated mbuf chain with the following structure: 1273 * 1274 * (1) mbuf #1 will contain the supplied data link 1275 * (2) mbuf #2 will contain the vlan header 1276 * (3) mbuf #3 will contain the original mbuf's packet data 1277 * 1278 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1279 */ 1280 if (data != NULL) { 1281 mv.m_next = m; 1282 mv.m_data = (caddr_t)&vlan; 1283 mv.m_len = sizeof(vlan); 1284 mb.m_next = &mv; 1285 mb.m_data = data; 1286 mb.m_len = dlen; 1287 bpf_mtap(bp, &mb); 1288 } else 1289 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1290 m->m_len += sizeof(struct ether_header); 1291 m->m_data -= sizeof(struct ether_header); 1292 } 1293 1294 struct mbuf * 1295 ether_vlanencap(struct mbuf *m, uint16_t tag) 1296 { 1297 struct ether_vlan_header *evl; 1298 1299 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1300 if (m == NULL) 1301 return (NULL); 1302 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1303 1304 if (m->m_len < sizeof(*evl)) { 1305 m = m_pullup(m, sizeof(*evl)); 1306 if (m == NULL) 1307 return (NULL); 1308 } 1309 1310 /* 1311 * Transform the Ethernet header into an Ethernet header 1312 * with 802.1Q encapsulation. 1313 */ 1314 evl = mtod(m, struct ether_vlan_header *); 1315 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1316 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1317 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1318 evl->evl_tag = htons(tag); 1319 return (m); 1320 } 1321 1322 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1323 MODULE_VERSION(ether, 1); 1324