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: stable/10/sys/net/if_ethersubr.c 332160 2018-04-07 00:04:28Z brooks $ 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/bus.h> 43#include <sys/eventhandler.h> 44#include <sys/kernel.h> 45#include <sys/lock.h> 46#include <sys/malloc.h> 47#include <sys/module.h> 48#include <sys/mbuf.h> 49#include <sys/random.h> 50#include <sys/socket.h> 51#include <sys/sockio.h> 52#include <sys/sysctl.h> 53#include <sys/uuid.h> 54 55#include <net/if.h> 56#include <net/if_arp.h> 57#include <net/netisr.h> 58#include <net/route.h> 59#include <net/if_llc.h> 60#include <net/if_dl.h> 61#include <net/if_types.h> 62#include <net/bpf.h> 63#include <net/ethernet.h> 64#include <net/if_bridgevar.h> 65#include <net/if_vlan_var.h> 66#include <net/if_llatbl.h> 67#include <net/pfil.h> 68#include <net/vnet.h> 69 70#include <netpfil/pf/pf_mtag.h> 71 72#if defined(INET) || defined(INET6) 73#include <netinet/in.h> 74#include <netinet/in_var.h> 75#include <netinet/if_ether.h> 76#include <netinet/ip_carp.h> 77#include <netinet/ip_var.h> 78#endif 79#ifdef INET6 80#include <netinet6/nd6.h> 81#endif 82 83#ifdef IPX 84#include <netipx/ipx.h> 85#include <netipx/ipx_if.h> 86#endif 87 88int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 89int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, 90 const struct sockaddr *dst, short *tp, int *hlen); 91 92#ifdef NETATALK 93#include <netatalk/at.h> 94#include <netatalk/at_var.h> 95#include <netatalk/at_extern.h> 96 97#define llc_snap_org_code llc_un.type_snap.org_code 98#define llc_snap_ether_type llc_un.type_snap.ether_type 99 100extern u_char at_org_code[3]; 101extern u_char aarp_org_code[3]; 102#endif /* NETATALK */ 103 104#include <security/mac/mac_framework.h> 105 106#ifdef CTASSERT 107CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 108CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 109#endif 110 111VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 112 113/* netgraph node hooks for ng_ether(4) */ 114void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 115void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 116int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 117void (*ng_ether_attach_p)(struct ifnet *ifp); 118void (*ng_ether_detach_p)(struct ifnet *ifp); 119 120void (*vlan_input_p)(struct ifnet *, struct mbuf *); 121 122/* if_bridge(4) support */ 123struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 124int (*bridge_output_p)(struct ifnet *, struct mbuf *, 125 struct sockaddr *, struct rtentry *); 126void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 127 128/* if_lagg(4) support */ 129struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); 130 131static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 132 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 133 134static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 135 struct sockaddr *); 136#ifdef VIMAGE 137static void ether_reassign(struct ifnet *, struct vnet *, char *); 138#endif 139 140/* XXX: should be in an arp support file, not here */ 141static MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals"); 142 143#define ETHER_IS_BROADCAST(addr) \ 144 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 145 146#define senderr(e) do { error = (e); goto bad;} while (0) 147 148static void 149update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 150{ 151 int csum_flags = 0; 152 153 if (src->m_pkthdr.csum_flags & CSUM_IP) 154 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 155 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 156 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 157 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 158 csum_flags |= CSUM_SCTP_VALID; 159 dst->m_pkthdr.csum_flags |= csum_flags; 160 if (csum_flags & CSUM_DATA_VALID) 161 dst->m_pkthdr.csum_data = 0xffff; 162} 163 164/* 165 * Ethernet output routine. 166 * Encapsulate a packet of type family for the local net. 167 * Use trailer local net encapsulation if enough data in first 168 * packet leaves a multiple of 512 bytes of data in remainder. 169 */ 170int 171ether_output(struct ifnet *ifp, struct mbuf *m, 172 const struct sockaddr *dst, struct route *ro) 173{ 174 short type; 175 int error = 0, hdrcmplt = 0; 176 u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 177 struct llentry *lle = NULL; 178 struct rtentry *rt0 = NULL; 179 struct ether_header *eh; 180 struct pf_mtag *t; 181 int loop_copy = 1; 182 int hlen; /* link layer header length */ 183 184 if (ro != NULL) { 185 if (!(m->m_flags & (M_BCAST | M_MCAST))) 186 lle = ro->ro_lle; 187 rt0 = ro->ro_rt; 188 } 189#ifdef MAC 190 error = mac_ifnet_check_transmit(ifp, m); 191 if (error) 192 senderr(error); 193#endif 194 195 M_PROFILE(m); 196 if (ifp->if_flags & IFF_MONITOR) 197 senderr(ENETDOWN); 198 if (!((ifp->if_flags & IFF_UP) && 199 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 200 senderr(ENETDOWN); 201 202 hlen = ETHER_HDR_LEN; 203 switch (dst->sa_family) { 204#ifdef INET 205 case AF_INET: 206 if (lle != NULL && (lle->la_flags & LLE_VALID)) 207 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 208 else 209 error = arpresolve(ifp, rt0, m, dst, edst, &lle); 210 if (error) 211 return (error == EWOULDBLOCK ? 0 : error); 212 type = htons(ETHERTYPE_IP); 213 break; 214 case AF_ARP: 215 { 216 struct arphdr *ah; 217 ah = mtod(m, struct arphdr *); 218 ah->ar_hrd = htons(ARPHRD_ETHER); 219 220 loop_copy = 0; /* if this is for us, don't do it */ 221 222 switch(ntohs(ah->ar_op)) { 223 case ARPOP_REVREQUEST: 224 case ARPOP_REVREPLY: 225 type = htons(ETHERTYPE_REVARP); 226 break; 227 case ARPOP_REQUEST: 228 case ARPOP_REPLY: 229 default: 230 type = htons(ETHERTYPE_ARP); 231 break; 232 } 233 234 if (m->m_flags & M_BCAST) 235 bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); 236 else 237 bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); 238 239 } 240 break; 241#endif 242#ifdef INET6 243 case AF_INET6: 244 if (lle != NULL && (lle->la_flags & LLE_VALID)) 245 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 246 else 247 error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle); 248 if (error) 249 return error; 250 type = htons(ETHERTYPE_IPV6); 251 break; 252#endif 253#ifdef IPX 254 case AF_IPX: 255 if (ef_outputp) { 256 error = ef_outputp(ifp, &m, dst, &type, &hlen); 257 if (error) 258 goto bad; 259 } else 260 type = htons(ETHERTYPE_IPX); 261 bcopy(&((const struct sockaddr_ipx *)dst)->sipx_addr.x_host, 262 edst, sizeof (edst)); 263 break; 264#endif 265#ifdef NETATALK 266 case AF_APPLETALK: 267 { 268 struct at_ifaddr *aa; 269 270 if ((aa = at_ifawithnet((const struct sockaddr_at *)dst)) == NULL) 271 senderr(EHOSTUNREACH); /* XXX */ 272 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 273 ifa_free(&aa->aa_ifa); 274 return (0); 275 } 276 /* 277 * In the phase 2 case, need to prepend an mbuf for the llc header. 278 */ 279 if ( aa->aa_flags & AFA_PHASE2 ) { 280 struct llc llc; 281 282 ifa_free(&aa->aa_ifa); 283 M_PREPEND(m, LLC_SNAPFRAMELEN, M_NOWAIT); 284 if (m == NULL) 285 senderr(ENOBUFS); 286 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 287 llc.llc_control = LLC_UI; 288 bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code)); 289 llc.llc_snap_ether_type = htons( ETHERTYPE_AT ); 290 bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN); 291 type = htons(m->m_pkthdr.len); 292 hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN; 293 } else { 294 ifa_free(&aa->aa_ifa); 295 type = htons(ETHERTYPE_AT); 296 } 297 break; 298 } 299#endif /* NETATALK */ 300 301 case pseudo_AF_HDRCMPLT: 302 { 303 const struct ether_header *eh; 304 305 hdrcmplt = 1; 306 eh = (const struct ether_header *)dst->sa_data; 307 (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); 308 /* FALLTHROUGH */ 309 310 case AF_UNSPEC: 311 loop_copy = 0; /* if this is for us, don't do it */ 312 eh = (const struct ether_header *)dst->sa_data; 313 (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); 314 type = eh->ether_type; 315 break; 316 } 317 default: 318 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 319 senderr(EAFNOSUPPORT); 320 } 321 322 if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { 323 update_mbuf_csumflags(m, m); 324 return (if_simloop(ifp, m, dst->sa_family, 0)); 325 } 326 327 /* 328 * Add local net header. If no space in first mbuf, 329 * allocate another. 330 */ 331 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 332 if (m == NULL) 333 senderr(ENOBUFS); 334 eh = mtod(m, struct ether_header *); 335 (void)memcpy(&eh->ether_type, &type, 336 sizeof(eh->ether_type)); 337 (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); 338 if (hdrcmplt) 339 (void)memcpy(eh->ether_shost, esrc, 340 sizeof(eh->ether_shost)); 341 else 342 (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), 343 sizeof(eh->ether_shost)); 344 345 /* 346 * If a simplex interface, and the packet is being sent to our 347 * Ethernet address or a broadcast address, loopback a copy. 348 * XXX To make a simplex device behave exactly like a duplex 349 * device, we should copy in the case of sending to our own 350 * ethernet address (thus letting the original actually appear 351 * on the wire). However, we don't do that here for security 352 * reasons and compatibility with the original behavior. 353 */ 354 if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && 355 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 356 if (m->m_flags & M_BCAST) { 357 struct mbuf *n; 358 359 /* 360 * Because if_simloop() modifies the packet, we need a 361 * writable copy through m_dup() instead of a readonly 362 * one as m_copy[m] would give us. The alternative would 363 * be to modify if_simloop() to handle the readonly mbuf, 364 * but performancewise it is mostly equivalent (trading 365 * extra data copying vs. extra locking). 366 * 367 * XXX This is a local workaround. A number of less 368 * often used kernel parts suffer from the same bug. 369 * See PR kern/105943 for a proposed general solution. 370 */ 371 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 372 update_mbuf_csumflags(m, n); 373 (void)if_simloop(ifp, n, dst->sa_family, hlen); 374 } else 375 ifp->if_iqdrops++; 376 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 377 ETHER_ADDR_LEN) == 0) { 378 update_mbuf_csumflags(m, m); 379 (void) if_simloop(ifp, m, dst->sa_family, hlen); 380 return (0); /* XXX */ 381 } 382 } 383 384 /* 385 * Bridges require special output handling. 386 */ 387 if (ifp->if_bridge) { 388 BRIDGE_OUTPUT(ifp, m, error); 389 return (error); 390 } 391 392#if defined(INET) || defined(INET6) 393 if (ifp->if_carp && 394 (error = (*carp_output_p)(ifp, m, dst))) 395 goto bad; 396#endif 397 398 /* Handle ng_ether(4) processing, if any */ 399 if (IFP2AC(ifp)->ac_netgraph != NULL) { 400 KASSERT(ng_ether_output_p != NULL, 401 ("ng_ether_output_p is NULL")); 402 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 403bad: if (m != NULL) 404 m_freem(m); 405 return (error); 406 } 407 if (m == NULL) 408 return (0); 409 } 410 411 /* Continue with link-layer output */ 412 return ether_output_frame(ifp, m); 413} 414 415/* 416 * Ethernet link layer output routine to send a raw frame to the device. 417 * 418 * This assumes that the 14 byte Ethernet header is present and contiguous 419 * in the first mbuf (if BRIDGE'ing). 420 */ 421int 422ether_output_frame(struct ifnet *ifp, struct mbuf *m) 423{ 424 int i; 425 426 if (PFIL_HOOKED(&V_link_pfil_hook)) { 427 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); 428 429 if (i != 0) 430 return (EACCES); 431 432 if (m == NULL) 433 return (0); 434 } 435 436 /* 437 * Queue message on interface, update output statistics if 438 * successful, and start output if interface not yet active. 439 */ 440 return ((ifp->if_transmit)(ifp, m)); 441} 442 443#if defined(INET) || defined(INET6) 444#endif 445 446/* 447 * Process a received Ethernet packet; the packet is in the 448 * mbuf chain m with the ethernet header at the front. 449 */ 450static void 451ether_input_internal(struct ifnet *ifp, struct mbuf *m) 452{ 453 struct ether_header *eh; 454 u_short etype; 455 456 if ((ifp->if_flags & IFF_UP) == 0) { 457 m_freem(m); 458 return; 459 } 460#ifdef DIAGNOSTIC 461 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 462 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 463 m_freem(m); 464 return; 465 } 466#endif 467 /* 468 * Do consistency checks to verify assumptions 469 * made by code past this point. 470 */ 471 if ((m->m_flags & M_PKTHDR) == 0) { 472 if_printf(ifp, "discard frame w/o packet header\n"); 473 ifp->if_ierrors++; 474 m_freem(m); 475 return; 476 } 477 if (m->m_len < ETHER_HDR_LEN) { 478 /* XXX maybe should pullup? */ 479 if_printf(ifp, "discard frame w/o leading ethernet " 480 "header (len %u pkt len %u)\n", 481 m->m_len, m->m_pkthdr.len); 482 ifp->if_ierrors++; 483 m_freem(m); 484 return; 485 } 486 eh = mtod(m, struct ether_header *); 487 etype = ntohs(eh->ether_type); 488 if (m->m_pkthdr.rcvif == NULL) { 489 if_printf(ifp, "discard frame w/o interface pointer\n"); 490 ifp->if_ierrors++; 491 m_freem(m); 492 return; 493 } 494#ifdef DIAGNOSTIC 495 if (m->m_pkthdr.rcvif != ifp) { 496 if_printf(ifp, "Warning, frame marked as received on %s\n", 497 m->m_pkthdr.rcvif->if_xname); 498 } 499#endif 500 501 CURVNET_SET_QUIET(ifp->if_vnet); 502 503 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 504 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 505 m->m_flags |= M_BCAST; 506 else 507 m->m_flags |= M_MCAST; 508 ifp->if_imcasts++; 509 } 510 511#ifdef MAC 512 /* 513 * Tag the mbuf with an appropriate MAC label before any other 514 * consumers can get to it. 515 */ 516 mac_ifnet_create_mbuf(ifp, m); 517#endif 518 519 /* 520 * Give bpf a chance at the packet. 521 */ 522 ETHER_BPF_MTAP(ifp, m); 523 524 /* 525 * If the CRC is still on the packet, trim it off. We do this once 526 * and once only in case we are re-entered. Nothing else on the 527 * Ethernet receive path expects to see the FCS. 528 */ 529 if (m->m_flags & M_HASFCS) { 530 m_adj(m, -ETHER_CRC_LEN); 531 m->m_flags &= ~M_HASFCS; 532 } 533 534 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 535 ifp->if_ibytes += m->m_pkthdr.len; 536 537 /* Allow monitor mode to claim this frame, after stats are updated. */ 538 if (ifp->if_flags & IFF_MONITOR) { 539 m_freem(m); 540 CURVNET_RESTORE(); 541 return; 542 } 543 544 /* Handle input from a lagg(4) port */ 545 if (ifp->if_type == IFT_IEEE8023ADLAG) { 546 KASSERT(lagg_input_p != NULL, 547 ("%s: if_lagg not loaded!", __func__)); 548 m = (*lagg_input_p)(ifp, m); 549 if (m != NULL) 550 ifp = m->m_pkthdr.rcvif; 551 else { 552 CURVNET_RESTORE(); 553 return; 554 } 555 } 556 557 /* 558 * If the hardware did not process an 802.1Q tag, do this now, 559 * to allow 802.1P priority frames to be passed to the main input 560 * path correctly. 561 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 562 */ 563 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 564 struct ether_vlan_header *evl; 565 566 if (m->m_len < sizeof(*evl) && 567 (m = m_pullup(m, sizeof(*evl))) == NULL) { 568#ifdef DIAGNOSTIC 569 if_printf(ifp, "cannot pullup VLAN header\n"); 570#endif 571 ifp->if_ierrors++; 572 m_freem(m); 573 CURVNET_RESTORE(); 574 return; 575 } 576 577 evl = mtod(m, struct ether_vlan_header *); 578 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 579 m->m_flags |= M_VLANTAG; 580 581 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 582 ETHER_HDR_LEN - ETHER_TYPE_LEN); 583 m_adj(m, ETHER_VLAN_ENCAP_LEN); 584 eh = mtod(m, struct ether_header *); 585 } 586 587 M_SETFIB(m, ifp->if_fib); 588 589 /* Allow ng_ether(4) to claim this frame. */ 590 if (IFP2AC(ifp)->ac_netgraph != NULL) { 591 KASSERT(ng_ether_input_p != NULL, 592 ("%s: ng_ether_input_p is NULL", __func__)); 593 m->m_flags &= ~M_PROMISC; 594 (*ng_ether_input_p)(ifp, &m); 595 if (m == NULL) { 596 CURVNET_RESTORE(); 597 return; 598 } 599 eh = mtod(m, struct ether_header *); 600 } 601 602 /* 603 * Allow if_bridge(4) to claim this frame. 604 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 605 * and the frame should be delivered locally. 606 */ 607 if (ifp->if_bridge != NULL) { 608 m->m_flags &= ~M_PROMISC; 609 BRIDGE_INPUT(ifp, m); 610 if (m == NULL) { 611 CURVNET_RESTORE(); 612 return; 613 } 614 eh = mtod(m, struct ether_header *); 615 } 616 617#if defined(INET) || defined(INET6) 618 /* 619 * Clear M_PROMISC on frame so that carp(4) will see it when the 620 * mbuf flows up to Layer 3. 621 * FreeBSD's implementation of carp(4) uses the inprotosw 622 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 623 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 624 * is outside the scope of the M_PROMISC test below. 625 * TODO: Maintain a hash table of ethernet addresses other than 626 * ether_dhost which may be active on this ifp. 627 */ 628 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 629 m->m_flags &= ~M_PROMISC; 630 } else 631#endif 632 { 633 /* 634 * If the frame received was not for our MAC address, set the 635 * M_PROMISC flag on the mbuf chain. The frame may need to 636 * be seen by the rest of the Ethernet input path in case of 637 * re-entry (e.g. bridge, vlan, netgraph) but should not be 638 * seen by upper protocol layers. 639 */ 640 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 641 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 642 m->m_flags |= M_PROMISC; 643 } 644 645 if (harvest.ethernet) 646 random_harvest(&(m->m_data), 12, 2, RANDOM_NET_ETHER); 647 648 ether_demux(ifp, m); 649 CURVNET_RESTORE(); 650} 651 652/* 653 * Ethernet input dispatch; by default, direct dispatch here regardless of 654 * global configuration. 655 */ 656static void 657ether_nh_input(struct mbuf *m) 658{ 659 660 ether_input_internal(m->m_pkthdr.rcvif, m); 661} 662 663static struct netisr_handler ether_nh = { 664 .nh_name = "ether", 665 .nh_handler = ether_nh_input, 666 .nh_proto = NETISR_ETHER, 667 .nh_policy = NETISR_POLICY_SOURCE, 668 .nh_dispatch = NETISR_DISPATCH_DIRECT, 669}; 670 671static void 672ether_init(__unused void *arg) 673{ 674 675 netisr_register(ðer_nh); 676} 677SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 678 679static void 680vnet_ether_init(__unused void *arg) 681{ 682 int i; 683 684 /* Initialize packet filter hooks. */ 685 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 686 V_link_pfil_hook.ph_af = AF_LINK; 687 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 688 printf("%s: WARNING: unable to register pfil link hook, " 689 "error %d\n", __func__, i); 690} 691VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 692 vnet_ether_init, NULL); 693 694static void 695vnet_ether_destroy(__unused void *arg) 696{ 697 int i; 698 699 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 700 printf("%s: WARNING: unable to unregister pfil link hook, " 701 "error %d\n", __func__, i); 702} 703VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 704 vnet_ether_destroy, NULL); 705 706 707 708static void 709ether_input(struct ifnet *ifp, struct mbuf *m) 710{ 711 712 /* 713 * We will rely on rcvif being set properly in the deferred context, 714 * so assert it is correct here. 715 */ 716 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__)); 717 718 netisr_dispatch(NETISR_ETHER, m); 719} 720 721/* 722 * Upper layer processing for a received Ethernet packet. 723 */ 724void 725ether_demux(struct ifnet *ifp, struct mbuf *m) 726{ 727 struct ether_header *eh; 728 int i, isr; 729 u_short ether_type; 730#if defined(NETATALK) 731 struct llc *l; 732#endif 733 734 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 735 736 /* Do not grab PROMISC frames in case we are re-entered. */ 737 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 738 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); 739 740 if (i != 0 || m == NULL) 741 return; 742 } 743 744 eh = mtod(m, struct ether_header *); 745 ether_type = ntohs(eh->ether_type); 746 747 /* 748 * If this frame has a VLAN tag other than 0, call vlan_input() 749 * if its module is loaded. Otherwise, drop. 750 */ 751 if ((m->m_flags & M_VLANTAG) && 752 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 753 if (ifp->if_vlantrunk == NULL) { 754 ifp->if_noproto++; 755 m_freem(m); 756 return; 757 } 758 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 759 __func__)); 760 /* Clear before possibly re-entering ether_input(). */ 761 m->m_flags &= ~M_PROMISC; 762 (*vlan_input_p)(ifp, m); 763 return; 764 } 765 766 /* 767 * Pass promiscuously received frames to the upper layer if the user 768 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 769 */ 770 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 771 m_freem(m); 772 return; 773 } 774 775 /* 776 * Reset layer specific mbuf flags to avoid confusing upper layers. 777 * Strip off Ethernet header. 778 */ 779 m->m_flags &= ~M_VLANTAG; 780 m_clrprotoflags(m); 781 m_adj(m, ETHER_HDR_LEN); 782 783 /* 784 * Dispatch frame to upper layer. 785 */ 786 switch (ether_type) { 787#ifdef INET 788 case ETHERTYPE_IP: 789 if ((m = ip_fastforward(m)) == NULL) 790 return; 791 isr = NETISR_IP; 792 break; 793 794 case ETHERTYPE_ARP: 795 if (ifp->if_flags & IFF_NOARP) { 796 /* Discard packet if ARP is disabled on interface */ 797 m_freem(m); 798 return; 799 } 800 isr = NETISR_ARP; 801 break; 802#endif 803#ifdef IPX 804 case ETHERTYPE_IPX: 805 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 806 return; 807 isr = NETISR_IPX; 808 break; 809#endif 810#ifdef INET6 811 case ETHERTYPE_IPV6: 812 isr = NETISR_IPV6; 813 break; 814#endif 815#ifdef NETATALK 816 case ETHERTYPE_AT: 817 isr = NETISR_ATALK1; 818 break; 819 case ETHERTYPE_AARP: 820 isr = NETISR_AARP; 821 break; 822#endif /* NETATALK */ 823 default: 824#ifdef IPX 825 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 826 return; 827#endif /* IPX */ 828#if defined(NETATALK) 829 if (ether_type > ETHERMTU) 830 goto discard; 831 l = mtod(m, struct llc *); 832 if (l->llc_dsap == LLC_SNAP_LSAP && 833 l->llc_ssap == LLC_SNAP_LSAP && 834 l->llc_control == LLC_UI) { 835 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 836 sizeof(at_org_code)) == 0 && 837 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 838 m_adj(m, LLC_SNAPFRAMELEN); 839 isr = NETISR_ATALK2; 840 break; 841 } 842 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 843 sizeof(aarp_org_code)) == 0 && 844 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 845 m_adj(m, LLC_SNAPFRAMELEN); 846 isr = NETISR_AARP; 847 break; 848 } 849 } 850#endif /* NETATALK */ 851 goto discard; 852 } 853 netisr_dispatch(isr, m); 854 return; 855 856discard: 857 /* 858 * Packet is to be discarded. If netgraph is present, 859 * hand the packet to it for last chance processing; 860 * otherwise dispose of it. 861 */ 862 if (IFP2AC(ifp)->ac_netgraph != NULL) { 863 KASSERT(ng_ether_input_orphan_p != NULL, 864 ("ng_ether_input_orphan_p is NULL")); 865 /* 866 * Put back the ethernet header so netgraph has a 867 * consistent view of inbound packets. 868 */ 869 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 870 (*ng_ether_input_orphan_p)(ifp, m); 871 return; 872 } 873 m_freem(m); 874} 875 876/* 877 * Convert Ethernet address to printable (loggable) representation. 878 * This routine is for compatibility; it's better to just use 879 * 880 * printf("%6D", <pointer to address>, ":"); 881 * 882 * since there's no static buffer involved. 883 */ 884char * 885ether_sprintf(const u_char *ap) 886{ 887 static char etherbuf[18]; 888 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 889 return (etherbuf); 890} 891 892/* 893 * Perform common duties while attaching to interface list 894 */ 895void 896ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 897{ 898 int i; 899 struct ifaddr *ifa; 900 struct sockaddr_dl *sdl; 901 902 ifp->if_addrlen = ETHER_ADDR_LEN; 903 ifp->if_hdrlen = ETHER_HDR_LEN; 904 if_attach(ifp); 905 ifp->if_mtu = ETHERMTU; 906 ifp->if_output = ether_output; 907 ifp->if_input = ether_input; 908 ifp->if_resolvemulti = ether_resolvemulti; 909#ifdef VIMAGE 910 ifp->if_reassign = ether_reassign; 911#endif 912 if (ifp->if_baudrate == 0) 913 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 914 ifp->if_broadcastaddr = etherbroadcastaddr; 915 916 ifa = ifp->if_addr; 917 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 918 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 919 sdl->sdl_type = IFT_ETHER; 920 sdl->sdl_alen = ifp->if_addrlen; 921 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 922 923 if (ifp->if_hw_addr != NULL) 924 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 925 926 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 927 if (ng_ether_attach_p != NULL) 928 (*ng_ether_attach_p)(ifp); 929 930 /* Announce Ethernet MAC address if non-zero. */ 931 for (i = 0; i < ifp->if_addrlen; i++) 932 if (lla[i] != 0) 933 break; 934 if (i != ifp->if_addrlen) 935 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 936 937 uuid_ether_add(LLADDR(sdl)); 938 939 /* Add necessary bits are setup; announce it now. */ 940 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 941 if (IS_DEFAULT_VNET(curvnet)) 942 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 943} 944 945/* 946 * Perform common duties while detaching an Ethernet interface 947 */ 948void 949ether_ifdetach(struct ifnet *ifp) 950{ 951 struct sockaddr_dl *sdl; 952 953 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 954 uuid_ether_del(LLADDR(sdl)); 955 956 if (IFP2AC(ifp)->ac_netgraph != NULL) { 957 KASSERT(ng_ether_detach_p != NULL, 958 ("ng_ether_detach_p is NULL")); 959 (*ng_ether_detach_p)(ifp); 960 } 961 962 bpfdetach(ifp); 963 if_detach(ifp); 964} 965 966#ifdef VIMAGE 967void 968ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 969{ 970 971 if (IFP2AC(ifp)->ac_netgraph != NULL) { 972 KASSERT(ng_ether_detach_p != NULL, 973 ("ng_ether_detach_p is NULL")); 974 (*ng_ether_detach_p)(ifp); 975 } 976 977 if (ng_ether_attach_p != NULL) { 978 CURVNET_SET_QUIET(new_vnet); 979 (*ng_ether_attach_p)(ifp); 980 CURVNET_RESTORE(); 981 } 982} 983#endif 984 985SYSCTL_DECL(_net_link); 986SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 987 988#if 0 989/* 990 * This is for reference. We have a table-driven version 991 * of the little-endian crc32 generator, which is faster 992 * than the double-loop. 993 */ 994uint32_t 995ether_crc32_le(const uint8_t *buf, size_t len) 996{ 997 size_t i; 998 uint32_t crc; 999 int bit; 1000 uint8_t data; 1001 1002 crc = 0xffffffff; /* initial value */ 1003 1004 for (i = 0; i < len; i++) { 1005 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1006 carry = (crc ^ data) & 1; 1007 crc >>= 1; 1008 if (carry) 1009 crc = (crc ^ ETHER_CRC_POLY_LE); 1010 } 1011 } 1012 1013 return (crc); 1014} 1015#else 1016uint32_t 1017ether_crc32_le(const uint8_t *buf, size_t len) 1018{ 1019 static const uint32_t crctab[] = { 1020 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1021 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1022 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1023 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1024 }; 1025 size_t i; 1026 uint32_t crc; 1027 1028 crc = 0xffffffff; /* initial value */ 1029 1030 for (i = 0; i < len; i++) { 1031 crc ^= buf[i]; 1032 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1033 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1034 } 1035 1036 return (crc); 1037} 1038#endif 1039 1040uint32_t 1041ether_crc32_be(const uint8_t *buf, size_t len) 1042{ 1043 size_t i; 1044 uint32_t crc, carry; 1045 int bit; 1046 uint8_t data; 1047 1048 crc = 0xffffffff; /* initial value */ 1049 1050 for (i = 0; i < len; i++) { 1051 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1052 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1053 crc <<= 1; 1054 if (carry) 1055 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1056 } 1057 } 1058 1059 return (crc); 1060} 1061 1062int 1063ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1064{ 1065 struct ifaddr *ifa = (struct ifaddr *) data; 1066 struct ifreq *ifr = (struct ifreq *) data; 1067 int error = 0; 1068 1069 switch (command) { 1070 case SIOCSIFADDR: 1071 ifp->if_flags |= IFF_UP; 1072 1073 switch (ifa->ifa_addr->sa_family) { 1074#ifdef INET 1075 case AF_INET: 1076 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1077 arp_ifinit(ifp, ifa); 1078 break; 1079#endif 1080#ifdef IPX 1081 /* 1082 * XXX - This code is probably wrong 1083 */ 1084 case AF_IPX: 1085 { 1086 struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr); 1087 1088 if (ipx_nullhost(*ina)) 1089 ina->x_host = 1090 *(union ipx_host *) 1091 IF_LLADDR(ifp); 1092 else { 1093 bcopy((caddr_t) ina->x_host.c_host, 1094 (caddr_t) IF_LLADDR(ifp), 1095 ETHER_ADDR_LEN); 1096 } 1097 1098 /* 1099 * Set new address 1100 */ 1101 ifp->if_init(ifp->if_softc); 1102 break; 1103 } 1104#endif 1105 default: 1106 ifp->if_init(ifp->if_softc); 1107 break; 1108 } 1109 break; 1110 1111 case SIOCGIFADDR: 1112 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1113 ETHER_ADDR_LEN); 1114 break; 1115 1116 case SIOCSIFMTU: 1117 /* 1118 * Set the interface MTU. 1119 */ 1120 if (ifr->ifr_mtu > ETHERMTU) { 1121 error = EINVAL; 1122 } else { 1123 ifp->if_mtu = ifr->ifr_mtu; 1124 } 1125 break; 1126 default: 1127 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1128 break; 1129 } 1130 return (error); 1131} 1132 1133static int 1134ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1135 struct sockaddr *sa) 1136{ 1137 struct sockaddr_dl *sdl; 1138#ifdef INET 1139 struct sockaddr_in *sin; 1140#endif 1141#ifdef INET6 1142 struct sockaddr_in6 *sin6; 1143#endif 1144 u_char *e_addr; 1145 1146 switch(sa->sa_family) { 1147 case AF_LINK: 1148 /* 1149 * No mapping needed. Just check that it's a valid MC address. 1150 */ 1151 sdl = (struct sockaddr_dl *)sa; 1152 e_addr = LLADDR(sdl); 1153 if (!ETHER_IS_MULTICAST(e_addr)) 1154 return EADDRNOTAVAIL; 1155 *llsa = 0; 1156 return 0; 1157 1158#ifdef INET 1159 case AF_INET: 1160 sin = (struct sockaddr_in *)sa; 1161 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1162 return EADDRNOTAVAIL; 1163 sdl = malloc(sizeof *sdl, M_IFMADDR, 1164 M_NOWAIT|M_ZERO); 1165 if (sdl == NULL) 1166 return ENOMEM; 1167 sdl->sdl_len = sizeof *sdl; 1168 sdl->sdl_family = AF_LINK; 1169 sdl->sdl_index = ifp->if_index; 1170 sdl->sdl_type = IFT_ETHER; 1171 sdl->sdl_alen = ETHER_ADDR_LEN; 1172 e_addr = LLADDR(sdl); 1173 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1174 *llsa = (struct sockaddr *)sdl; 1175 return 0; 1176#endif 1177#ifdef INET6 1178 case AF_INET6: 1179 sin6 = (struct sockaddr_in6 *)sa; 1180 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1181 /* 1182 * An IP6 address of 0 means listen to all 1183 * of the Ethernet multicast address used for IP6. 1184 * (This is used for multicast routers.) 1185 */ 1186 ifp->if_flags |= IFF_ALLMULTI; 1187 *llsa = 0; 1188 return 0; 1189 } 1190 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1191 return EADDRNOTAVAIL; 1192 sdl = malloc(sizeof *sdl, M_IFMADDR, 1193 M_NOWAIT|M_ZERO); 1194 if (sdl == NULL) 1195 return (ENOMEM); 1196 sdl->sdl_len = sizeof *sdl; 1197 sdl->sdl_family = AF_LINK; 1198 sdl->sdl_index = ifp->if_index; 1199 sdl->sdl_type = IFT_ETHER; 1200 sdl->sdl_alen = ETHER_ADDR_LEN; 1201 e_addr = LLADDR(sdl); 1202 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1203 *llsa = (struct sockaddr *)sdl; 1204 return 0; 1205#endif 1206 1207 default: 1208 /* 1209 * Well, the text isn't quite right, but it's the name 1210 * that counts... 1211 */ 1212 return EAFNOSUPPORT; 1213 } 1214} 1215 1216static void* 1217ether_alloc(u_char type, struct ifnet *ifp) 1218{ 1219 struct arpcom *ac; 1220 1221 ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO); 1222 ac->ac_ifp = ifp; 1223 1224 return (ac); 1225} 1226 1227static void 1228ether_free(void *com, u_char type) 1229{ 1230 1231 free(com, M_ARPCOM); 1232} 1233 1234static int 1235ether_modevent(module_t mod, int type, void *data) 1236{ 1237 1238 switch (type) { 1239 case MOD_LOAD: 1240 if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free); 1241 break; 1242 case MOD_UNLOAD: 1243 if_deregister_com_alloc(IFT_ETHER); 1244 break; 1245 default: 1246 return EOPNOTSUPP; 1247 } 1248 1249 return (0); 1250} 1251 1252static moduledata_t ether_mod = { 1253 "ether", 1254 ether_modevent, 1255 0 1256}; 1257 1258void 1259ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1260{ 1261 struct ether_vlan_header vlan; 1262 struct mbuf mv, mb; 1263 1264 KASSERT((m->m_flags & M_VLANTAG) != 0, 1265 ("%s: vlan information not present", __func__)); 1266 KASSERT(m->m_len >= sizeof(struct ether_header), 1267 ("%s: mbuf not large enough for header", __func__)); 1268 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1269 vlan.evl_proto = vlan.evl_encap_proto; 1270 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1271 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1272 m->m_len -= sizeof(struct ether_header); 1273 m->m_data += sizeof(struct ether_header); 1274 /* 1275 * If a data link has been supplied by the caller, then we will need to 1276 * re-create a stack allocated mbuf chain with the following structure: 1277 * 1278 * (1) mbuf #1 will contain the supplied data link 1279 * (2) mbuf #2 will contain the vlan header 1280 * (3) mbuf #3 will contain the original mbuf's packet data 1281 * 1282 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1283 */ 1284 if (data != NULL) { 1285 mv.m_next = m; 1286 mv.m_data = (caddr_t)&vlan; 1287 mv.m_len = sizeof(vlan); 1288 mb.m_next = &mv; 1289 mb.m_data = data; 1290 mb.m_len = dlen; 1291 bpf_mtap(bp, &mb); 1292 } else 1293 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1294 m->m_len += sizeof(struct ether_header); 1295 m->m_data -= sizeof(struct ether_header); 1296} 1297 1298struct mbuf * 1299ether_vlanencap(struct mbuf *m, uint16_t tag) 1300{ 1301 struct ether_vlan_header *evl; 1302 1303 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1304 if (m == NULL) 1305 return (NULL); 1306 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1307 1308 if (m->m_len < sizeof(*evl)) { 1309 m = m_pullup(m, sizeof(*evl)); 1310 if (m == NULL) 1311 return (NULL); 1312 } 1313 1314 /* 1315 * Transform the Ethernet header into an Ethernet header 1316 * with 802.1Q encapsulation. 1317 */ 1318 evl = mtod(m, struct ether_vlan_header *); 1319 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1320 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1321 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1322 evl->evl_tag = htons(tag); 1323 return (m); 1324} 1325 1326DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1327MODULE_VERSION(ether, 1); 1328