if_ethersubr.c revision 263086
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 263086 2014-03-12 10:45:58Z glebius $ 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_arp.h> 55#include <net/netisr.h> 56#include <net/route.h> 57#include <net/if_llc.h> 58#include <net/if_dl.h> 59#include <net/if_types.h> 60#include <net/bpf.h> 61#include <net/ethernet.h> 62#include <net/if_bridgevar.h> 63#include <net/if_vlan_var.h> 64#include <net/if_llatbl.h> 65#include <net/pfil.h> 66#include <net/vnet.h> 67 68#include <netpfil/pf/pf_mtag.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 <netinet/ip_carp.h> 75#include <netinet/ip_var.h> 76#endif 77#ifdef INET6 78#include <netinet6/nd6.h> 79#endif 80 81#ifdef IPX 82#include <netipx/ipx.h> 83#include <netipx/ipx_if.h> 84#endif 85 86int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 87int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, 88 const struct sockaddr *dst, short *tp, int *hlen); 89 90#ifdef NETATALK 91#include <netatalk/at.h> 92#include <netatalk/at_var.h> 93#include <netatalk/at_extern.h> 94 95#define llc_snap_org_code llc_un.type_snap.org_code 96#define llc_snap_ether_type llc_un.type_snap.ether_type 97 98extern u_char at_org_code[3]; 99extern u_char aarp_org_code[3]; 100#endif /* NETATALK */ 101 102#include <security/mac/mac_framework.h> 103 104#ifdef CTASSERT 105CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 106CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 107#endif 108 109VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 110 111/* netgraph node hooks for ng_ether(4) */ 112void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 113void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 114int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 115void (*ng_ether_attach_p)(struct ifnet *ifp); 116void (*ng_ether_detach_p)(struct ifnet *ifp); 117 118void (*vlan_input_p)(struct ifnet *, struct mbuf *); 119 120/* if_bridge(4) support */ 121struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 122int (*bridge_output_p)(struct ifnet *, struct mbuf *, 123 struct sockaddr *, struct rtentry *); 124void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 125 126/* if_lagg(4) support */ 127struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); 128 129static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 130 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 131 132static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 133 struct sockaddr *); 134#ifdef VIMAGE 135static void ether_reassign(struct ifnet *, struct vnet *, char *); 136#endif 137 138/* XXX: should be in an arp support file, not here */ 139static MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals"); 140 141#define ETHER_IS_BROADCAST(addr) \ 142 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 143 144#define senderr(e) do { error = (e); goto bad;} while (0) 145 146static void 147update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 148{ 149 int csum_flags = 0; 150 151 if (src->m_pkthdr.csum_flags & CSUM_IP) 152 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 153 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 154 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 155 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 156 csum_flags |= CSUM_SCTP_VALID; 157 dst->m_pkthdr.csum_flags |= csum_flags; 158 if (csum_flags & CSUM_DATA_VALID) 159 dst->m_pkthdr.csum_data = 0xffff; 160} 161 162/* 163 * Ethernet output routine. 164 * Encapsulate a packet of type family for the local net. 165 * Use trailer local net encapsulation if enough data in first 166 * packet leaves a multiple of 512 bytes of data in remainder. 167 */ 168int 169ether_output(struct ifnet *ifp, struct mbuf *m, 170 const struct sockaddr *dst, struct route *ro) 171{ 172 short type; 173 int error = 0, hdrcmplt = 0; 174 u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 175 struct llentry *lle = NULL; 176 struct rtentry *rt0 = NULL; 177 struct ether_header *eh; 178 struct pf_mtag *t; 179 int loop_copy = 1; 180 int hlen; /* link layer header length */ 181 182 if (ro != NULL) { 183 if (!(m->m_flags & (M_BCAST | M_MCAST))) 184 lle = ro->ro_lle; 185 rt0 = ro->ro_rt; 186 } 187#ifdef MAC 188 error = mac_ifnet_check_transmit(ifp, m); 189 if (error) 190 senderr(error); 191#endif 192 193 M_PROFILE(m); 194 if (ifp->if_flags & IFF_MONITOR) 195 senderr(ENETDOWN); 196 if (!((ifp->if_flags & IFF_UP) && 197 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 198 senderr(ENETDOWN); 199 200 hlen = ETHER_HDR_LEN; 201 switch (dst->sa_family) { 202#ifdef INET 203 case AF_INET: 204 if (lle != NULL && (lle->la_flags & LLE_VALID)) 205 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 206 else 207 error = arpresolve(ifp, rt0, m, dst, edst, &lle); 208 if (error) 209 return (error == EWOULDBLOCK ? 0 : error); 210 type = htons(ETHERTYPE_IP); 211 break; 212 case AF_ARP: 213 { 214 struct arphdr *ah; 215 ah = mtod(m, struct arphdr *); 216 ah->ar_hrd = htons(ARPHRD_ETHER); 217 218 loop_copy = 0; /* if this is for us, don't do it */ 219 220 switch(ntohs(ah->ar_op)) { 221 case ARPOP_REVREQUEST: 222 case ARPOP_REVREPLY: 223 type = htons(ETHERTYPE_REVARP); 224 break; 225 case ARPOP_REQUEST: 226 case ARPOP_REPLY: 227 default: 228 type = htons(ETHERTYPE_ARP); 229 break; 230 } 231 232 if (m->m_flags & M_BCAST) 233 bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); 234 else 235 bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); 236 237 } 238 break; 239#endif 240#ifdef INET6 241 case AF_INET6: 242 if (lle != NULL && (lle->la_flags & LLE_VALID)) 243 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 244 else 245 error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle); 246 if (error) 247 return error; 248 type = htons(ETHERTYPE_IPV6); 249 break; 250#endif 251#ifdef IPX 252 case AF_IPX: 253 if (ef_outputp) { 254 error = ef_outputp(ifp, &m, dst, &type, &hlen); 255 if (error) 256 goto bad; 257 } else 258 type = htons(ETHERTYPE_IPX); 259 bcopy(&((const struct sockaddr_ipx *)dst)->sipx_addr.x_host, 260 edst, sizeof (edst)); 261 break; 262#endif 263#ifdef NETATALK 264 case AF_APPLETALK: 265 { 266 struct at_ifaddr *aa; 267 268 if ((aa = at_ifawithnet((const struct sockaddr_at *)dst)) == NULL) 269 senderr(EHOSTUNREACH); /* XXX */ 270 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 271 ifa_free(&aa->aa_ifa); 272 return (0); 273 } 274 /* 275 * In the phase 2 case, need to prepend an mbuf for the llc header. 276 */ 277 if ( aa->aa_flags & AFA_PHASE2 ) { 278 struct llc llc; 279 280 ifa_free(&aa->aa_ifa); 281 M_PREPEND(m, LLC_SNAPFRAMELEN, M_NOWAIT); 282 if (m == NULL) 283 senderr(ENOBUFS); 284 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 285 llc.llc_control = LLC_UI; 286 bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code)); 287 llc.llc_snap_ether_type = htons( ETHERTYPE_AT ); 288 bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN); 289 type = htons(m->m_pkthdr.len); 290 hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN; 291 } else { 292 ifa_free(&aa->aa_ifa); 293 type = htons(ETHERTYPE_AT); 294 } 295 break; 296 } 297#endif /* NETATALK */ 298 299 case pseudo_AF_HDRCMPLT: 300 { 301 const struct ether_header *eh; 302 303 hdrcmplt = 1; 304 eh = (const struct ether_header *)dst->sa_data; 305 (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); 306 /* FALLTHROUGH */ 307 308 case AF_UNSPEC: 309 loop_copy = 0; /* if this is for us, don't do it */ 310 eh = (const struct ether_header *)dst->sa_data; 311 (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); 312 type = eh->ether_type; 313 break; 314 } 315 default: 316 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 317 senderr(EAFNOSUPPORT); 318 } 319 320 if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { 321 update_mbuf_csumflags(m, m); 322 return (if_simloop(ifp, m, dst->sa_family, 0)); 323 } 324 325 /* 326 * Add local net header. If no space in first mbuf, 327 * allocate another. 328 */ 329 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 330 if (m == NULL) 331 senderr(ENOBUFS); 332 eh = mtod(m, struct ether_header *); 333 (void)memcpy(&eh->ether_type, &type, 334 sizeof(eh->ether_type)); 335 (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); 336 if (hdrcmplt) 337 (void)memcpy(eh->ether_shost, esrc, 338 sizeof(eh->ether_shost)); 339 else 340 (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), 341 sizeof(eh->ether_shost)); 342 343 /* 344 * If a simplex interface, and the packet is being sent to our 345 * Ethernet address or a broadcast address, loopback a copy. 346 * XXX To make a simplex device behave exactly like a duplex 347 * device, we should copy in the case of sending to our own 348 * ethernet address (thus letting the original actually appear 349 * on the wire). However, we don't do that here for security 350 * reasons and compatibility with the original behavior. 351 */ 352 if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && 353 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 354 if (m->m_flags & M_BCAST) { 355 struct mbuf *n; 356 357 /* 358 * Because if_simloop() modifies the packet, we need a 359 * writable copy through m_dup() instead of a readonly 360 * one as m_copy[m] would give us. The alternative would 361 * be to modify if_simloop() to handle the readonly mbuf, 362 * but performancewise it is mostly equivalent (trading 363 * extra data copying vs. extra locking). 364 * 365 * XXX This is a local workaround. A number of less 366 * often used kernel parts suffer from the same bug. 367 * See PR kern/105943 for a proposed general solution. 368 */ 369 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 370 update_mbuf_csumflags(m, n); 371 (void)if_simloop(ifp, n, dst->sa_family, hlen); 372 } else 373 ifp->if_iqdrops++; 374 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 375 ETHER_ADDR_LEN) == 0) { 376 update_mbuf_csumflags(m, m); 377 (void) if_simloop(ifp, m, dst->sa_family, hlen); 378 return (0); /* XXX */ 379 } 380 } 381 382 /* 383 * Bridges require special output handling. 384 */ 385 if (ifp->if_bridge) { 386 BRIDGE_OUTPUT(ifp, m, error); 387 return (error); 388 } 389 390#if defined(INET) || defined(INET6) 391 if (ifp->if_carp && 392 (error = (*carp_output_p)(ifp, m, dst))) 393 goto bad; 394#endif 395 396 /* Handle ng_ether(4) processing, if any */ 397 if (IFP2AC(ifp)->ac_netgraph != NULL) { 398 KASSERT(ng_ether_output_p != NULL, 399 ("ng_ether_output_p is NULL")); 400 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 401bad: if (m != NULL) 402 m_freem(m); 403 return (error); 404 } 405 if (m == NULL) 406 return (0); 407 } 408 409 /* Continue with link-layer output */ 410 return ether_output_frame(ifp, m); 411} 412 413/* 414 * Ethernet link layer output routine to send a raw frame to the device. 415 * 416 * This assumes that the 14 byte Ethernet header is present and contiguous 417 * in the first mbuf (if BRIDGE'ing). 418 */ 419int 420ether_output_frame(struct ifnet *ifp, struct mbuf *m) 421{ 422 int i; 423 424 if (PFIL_HOOKED(&V_link_pfil_hook)) { 425 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); 426 427 if (i != 0) 428 return (EACCES); 429 430 if (m == NULL) 431 return (0); 432 } 433 434 /* 435 * Queue message on interface, update output statistics if 436 * successful, and start output if interface not yet active. 437 */ 438 return ((ifp->if_transmit)(ifp, m)); 439} 440 441#if defined(INET) || defined(INET6) 442#endif 443 444/* 445 * Process a received Ethernet packet; the packet is in the 446 * mbuf chain m with the ethernet header at the front. 447 */ 448static void 449ether_input_internal(struct ifnet *ifp, struct mbuf *m) 450{ 451 struct ether_header *eh; 452 u_short etype; 453 454 if ((ifp->if_flags & IFF_UP) == 0) { 455 m_freem(m); 456 return; 457 } 458#ifdef DIAGNOSTIC 459 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 460 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 461 m_freem(m); 462 return; 463 } 464#endif 465 /* 466 * Do consistency checks to verify assumptions 467 * made by code past this point. 468 */ 469 if ((m->m_flags & M_PKTHDR) == 0) { 470 if_printf(ifp, "discard frame w/o packet header\n"); 471 ifp->if_ierrors++; 472 m_freem(m); 473 return; 474 } 475 if (m->m_len < ETHER_HDR_LEN) { 476 /* XXX maybe should pullup? */ 477 if_printf(ifp, "discard frame w/o leading ethernet " 478 "header (len %u pkt len %u)\n", 479 m->m_len, m->m_pkthdr.len); 480 ifp->if_ierrors++; 481 m_freem(m); 482 return; 483 } 484 eh = mtod(m, struct ether_header *); 485 etype = ntohs(eh->ether_type); 486 if (m->m_pkthdr.rcvif == NULL) { 487 if_printf(ifp, "discard frame w/o interface pointer\n"); 488 ifp->if_ierrors++; 489 m_freem(m); 490 return; 491 } 492#ifdef DIAGNOSTIC 493 if (m->m_pkthdr.rcvif != ifp) { 494 if_printf(ifp, "Warning, frame marked as received on %s\n", 495 m->m_pkthdr.rcvif->if_xname); 496 } 497#endif 498 499 CURVNET_SET_QUIET(ifp->if_vnet); 500 501 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 502 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 503 m->m_flags |= M_BCAST; 504 else 505 m->m_flags |= M_MCAST; 506 ifp->if_imcasts++; 507 } 508 509#ifdef MAC 510 /* 511 * Tag the mbuf with an appropriate MAC label before any other 512 * consumers can get to it. 513 */ 514 mac_ifnet_create_mbuf(ifp, m); 515#endif 516 517 /* 518 * Give bpf a chance at the packet. 519 */ 520 ETHER_BPF_MTAP(ifp, m); 521 522 /* 523 * If the CRC is still on the packet, trim it off. We do this once 524 * and once only in case we are re-entered. Nothing else on the 525 * Ethernet receive path expects to see the FCS. 526 */ 527 if (m->m_flags & M_HASFCS) { 528 m_adj(m, -ETHER_CRC_LEN); 529 m->m_flags &= ~M_HASFCS; 530 } 531 532 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 533 ifp->if_ibytes += m->m_pkthdr.len; 534 535 /* Allow monitor mode to claim this frame, after stats are updated. */ 536 if (ifp->if_flags & IFF_MONITOR) { 537 m_freem(m); 538 CURVNET_RESTORE(); 539 return; 540 } 541 542 /* Handle input from a lagg(4) port */ 543 if (ifp->if_type == IFT_IEEE8023ADLAG) { 544 KASSERT(lagg_input_p != NULL, 545 ("%s: if_lagg not loaded!", __func__)); 546 m = (*lagg_input_p)(ifp, m); 547 if (m != NULL) 548 ifp = m->m_pkthdr.rcvif; 549 else { 550 CURVNET_RESTORE(); 551 return; 552 } 553 } 554 555 /* 556 * If the hardware did not process an 802.1Q tag, do this now, 557 * to allow 802.1P priority frames to be passed to the main input 558 * path correctly. 559 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 560 */ 561 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 562 struct ether_vlan_header *evl; 563 564 if (m->m_len < sizeof(*evl) && 565 (m = m_pullup(m, sizeof(*evl))) == NULL) { 566#ifdef DIAGNOSTIC 567 if_printf(ifp, "cannot pullup VLAN header\n"); 568#endif 569 ifp->if_ierrors++; 570 m_freem(m); 571 CURVNET_RESTORE(); 572 return; 573 } 574 575 evl = mtod(m, struct ether_vlan_header *); 576 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 577 m->m_flags |= M_VLANTAG; 578 579 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 580 ETHER_HDR_LEN - ETHER_TYPE_LEN); 581 m_adj(m, ETHER_VLAN_ENCAP_LEN); 582 eh = mtod(m, struct ether_header *); 583 } 584 585 M_SETFIB(m, ifp->if_fib); 586 587 /* Allow ng_ether(4) to claim this frame. */ 588 if (IFP2AC(ifp)->ac_netgraph != NULL) { 589 KASSERT(ng_ether_input_p != NULL, 590 ("%s: ng_ether_input_p is NULL", __func__)); 591 m->m_flags &= ~M_PROMISC; 592 (*ng_ether_input_p)(ifp, &m); 593 if (m == NULL) { 594 CURVNET_RESTORE(); 595 return; 596 } 597 eh = mtod(m, struct ether_header *); 598 } 599 600 /* 601 * Allow if_bridge(4) to claim this frame. 602 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 603 * and the frame should be delivered locally. 604 */ 605 if (ifp->if_bridge != NULL) { 606 m->m_flags &= ~M_PROMISC; 607 BRIDGE_INPUT(ifp, m); 608 if (m == NULL) { 609 CURVNET_RESTORE(); 610 return; 611 } 612 eh = mtod(m, struct ether_header *); 613 } 614 615#if defined(INET) || defined(INET6) 616 /* 617 * Clear M_PROMISC on frame so that carp(4) will see it when the 618 * mbuf flows up to Layer 3. 619 * FreeBSD's implementation of carp(4) uses the inprotosw 620 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 621 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 622 * is outside the scope of the M_PROMISC test below. 623 * TODO: Maintain a hash table of ethernet addresses other than 624 * ether_dhost which may be active on this ifp. 625 */ 626 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 627 m->m_flags &= ~M_PROMISC; 628 } else 629#endif 630 { 631 /* 632 * If the frame received was not for our MAC address, set the 633 * M_PROMISC flag on the mbuf chain. The frame may need to 634 * be seen by the rest of the Ethernet input path in case of 635 * re-entry (e.g. bridge, vlan, netgraph) but should not be 636 * seen by upper protocol layers. 637 */ 638 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 639 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 640 m->m_flags |= M_PROMISC; 641 } 642 643 if (harvest.ethernet) 644 random_harvest(&(m->m_data), 12, 2, RANDOM_NET_ETHER); 645 646 ether_demux(ifp, m); 647 CURVNET_RESTORE(); 648} 649 650/* 651 * Ethernet input dispatch; by default, direct dispatch here regardless of 652 * global configuration. 653 */ 654static void 655ether_nh_input(struct mbuf *m) 656{ 657 658 ether_input_internal(m->m_pkthdr.rcvif, m); 659} 660 661static struct netisr_handler ether_nh = { 662 .nh_name = "ether", 663 .nh_handler = ether_nh_input, 664 .nh_proto = NETISR_ETHER, 665 .nh_policy = NETISR_POLICY_SOURCE, 666 .nh_dispatch = NETISR_DISPATCH_DIRECT, 667}; 668 669static void 670ether_init(__unused void *arg) 671{ 672 673 netisr_register(ðer_nh); 674} 675SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 676 677static void 678vnet_ether_init(__unused void *arg) 679{ 680 int i; 681 682 /* Initialize packet filter hooks. */ 683 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 684 V_link_pfil_hook.ph_af = AF_LINK; 685 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 686 printf("%s: WARNING: unable to register pfil link hook, " 687 "error %d\n", __func__, i); 688} 689VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 690 vnet_ether_init, NULL); 691 692static void 693vnet_ether_destroy(__unused void *arg) 694{ 695 int i; 696 697 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 698 printf("%s: WARNING: unable to unregister pfil link hook, " 699 "error %d\n", __func__, i); 700} 701VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 702 vnet_ether_destroy, NULL); 703 704 705 706static void 707ether_input(struct ifnet *ifp, struct mbuf *m) 708{ 709 710 /* 711 * We will rely on rcvif being set properly in the deferred context, 712 * so assert it is correct here. 713 */ 714 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__)); 715 716 netisr_dispatch(NETISR_ETHER, m); 717} 718 719/* 720 * Upper layer processing for a received Ethernet packet. 721 */ 722void 723ether_demux(struct ifnet *ifp, struct mbuf *m) 724{ 725 struct ether_header *eh; 726 int i, isr; 727 u_short ether_type; 728#if defined(NETATALK) 729 struct llc *l; 730#endif 731 732 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 733 734 /* Do not grab PROMISC frames in case we are re-entered. */ 735 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 736 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); 737 738 if (i != 0 || m == NULL) 739 return; 740 } 741 742 eh = mtod(m, struct ether_header *); 743 ether_type = ntohs(eh->ether_type); 744 745 /* 746 * If this frame has a VLAN tag other than 0, call vlan_input() 747 * if its module is loaded. Otherwise, drop. 748 */ 749 if ((m->m_flags & M_VLANTAG) && 750 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 751 if (ifp->if_vlantrunk == NULL) { 752 ifp->if_noproto++; 753 m_freem(m); 754 return; 755 } 756 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 757 __func__)); 758 /* Clear before possibly re-entering ether_input(). */ 759 m->m_flags &= ~M_PROMISC; 760 (*vlan_input_p)(ifp, m); 761 return; 762 } 763 764 /* 765 * Pass promiscuously received frames to the upper layer if the user 766 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 767 */ 768 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 769 m_freem(m); 770 return; 771 } 772 773 /* 774 * Reset layer specific mbuf flags to avoid confusing upper layers. 775 * Strip off Ethernet header. 776 */ 777 m->m_flags &= ~M_VLANTAG; 778 m_clrprotoflags(m); 779 m_adj(m, ETHER_HDR_LEN); 780 781 /* 782 * Dispatch frame to upper layer. 783 */ 784 switch (ether_type) { 785#ifdef INET 786 case ETHERTYPE_IP: 787 if ((m = ip_fastforward(m)) == NULL) 788 return; 789 isr = NETISR_IP; 790 break; 791 792 case ETHERTYPE_ARP: 793 if (ifp->if_flags & IFF_NOARP) { 794 /* Discard packet if ARP is disabled on interface */ 795 m_freem(m); 796 return; 797 } 798 isr = NETISR_ARP; 799 break; 800#endif 801#ifdef IPX 802 case ETHERTYPE_IPX: 803 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 804 return; 805 isr = NETISR_IPX; 806 break; 807#endif 808#ifdef INET6 809 case ETHERTYPE_IPV6: 810 isr = NETISR_IPV6; 811 break; 812#endif 813#ifdef NETATALK 814 case ETHERTYPE_AT: 815 isr = NETISR_ATALK1; 816 break; 817 case ETHERTYPE_AARP: 818 isr = NETISR_AARP; 819 break; 820#endif /* NETATALK */ 821 default: 822#ifdef IPX 823 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) 824 return; 825#endif /* IPX */ 826#if defined(NETATALK) 827 if (ether_type > ETHERMTU) 828 goto discard; 829 l = mtod(m, struct llc *); 830 if (l->llc_dsap == LLC_SNAP_LSAP && 831 l->llc_ssap == LLC_SNAP_LSAP && 832 l->llc_control == LLC_UI) { 833 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 834 sizeof(at_org_code)) == 0 && 835 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 836 m_adj(m, LLC_SNAPFRAMELEN); 837 isr = NETISR_ATALK2; 838 break; 839 } 840 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 841 sizeof(aarp_org_code)) == 0 && 842 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 843 m_adj(m, LLC_SNAPFRAMELEN); 844 isr = NETISR_AARP; 845 break; 846 } 847 } 848#endif /* NETATALK */ 849 goto discard; 850 } 851 netisr_dispatch(isr, m); 852 return; 853 854discard: 855 /* 856 * Packet is to be discarded. If netgraph is present, 857 * hand the packet to it for last chance processing; 858 * otherwise dispose of it. 859 */ 860 if (IFP2AC(ifp)->ac_netgraph != NULL) { 861 KASSERT(ng_ether_input_orphan_p != NULL, 862 ("ng_ether_input_orphan_p is NULL")); 863 /* 864 * Put back the ethernet header so netgraph has a 865 * consistent view of inbound packets. 866 */ 867 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 868 (*ng_ether_input_orphan_p)(ifp, m); 869 return; 870 } 871 m_freem(m); 872} 873 874/* 875 * Convert Ethernet address to printable (loggable) representation. 876 * This routine is for compatibility; it's better to just use 877 * 878 * printf("%6D", <pointer to address>, ":"); 879 * 880 * since there's no static buffer involved. 881 */ 882char * 883ether_sprintf(const u_char *ap) 884{ 885 static char etherbuf[18]; 886 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 887 return (etherbuf); 888} 889 890/* 891 * Perform common duties while attaching to interface list 892 */ 893void 894ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 895{ 896 int i; 897 struct ifaddr *ifa; 898 struct sockaddr_dl *sdl; 899 900 ifp->if_addrlen = ETHER_ADDR_LEN; 901 ifp->if_hdrlen = ETHER_HDR_LEN; 902 if_attach(ifp); 903 ifp->if_mtu = ETHERMTU; 904 ifp->if_output = ether_output; 905 ifp->if_input = ether_input; 906 ifp->if_resolvemulti = ether_resolvemulti; 907#ifdef VIMAGE 908 ifp->if_reassign = ether_reassign; 909#endif 910 if (ifp->if_baudrate == 0) 911 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 912 ifp->if_broadcastaddr = etherbroadcastaddr; 913 914 ifa = ifp->if_addr; 915 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 916 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 917 sdl->sdl_type = IFT_ETHER; 918 sdl->sdl_alen = ifp->if_addrlen; 919 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 920 921 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 922 if (ng_ether_attach_p != NULL) 923 (*ng_ether_attach_p)(ifp); 924 925 /* Announce Ethernet MAC address if non-zero. */ 926 for (i = 0; i < ifp->if_addrlen; i++) 927 if (lla[i] != 0) 928 break; 929 if (i != ifp->if_addrlen) 930 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 931 932 uuid_ether_add(LLADDR(sdl)); 933} 934 935/* 936 * Perform common duties while detaching an Ethernet interface 937 */ 938void 939ether_ifdetach(struct ifnet *ifp) 940{ 941 struct sockaddr_dl *sdl; 942 943 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 944 uuid_ether_del(LLADDR(sdl)); 945 946 if (IFP2AC(ifp)->ac_netgraph != NULL) { 947 KASSERT(ng_ether_detach_p != NULL, 948 ("ng_ether_detach_p is NULL")); 949 (*ng_ether_detach_p)(ifp); 950 } 951 952 bpfdetach(ifp); 953 if_detach(ifp); 954} 955 956#ifdef VIMAGE 957void 958ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 959{ 960 961 if (IFP2AC(ifp)->ac_netgraph != NULL) { 962 KASSERT(ng_ether_detach_p != NULL, 963 ("ng_ether_detach_p is NULL")); 964 (*ng_ether_detach_p)(ifp); 965 } 966 967 if (ng_ether_attach_p != NULL) { 968 CURVNET_SET_QUIET(new_vnet); 969 (*ng_ether_attach_p)(ifp); 970 CURVNET_RESTORE(); 971 } 972} 973#endif 974 975SYSCTL_DECL(_net_link); 976SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 977 978#if 0 979/* 980 * This is for reference. We have a table-driven version 981 * of the little-endian crc32 generator, which is faster 982 * than the double-loop. 983 */ 984uint32_t 985ether_crc32_le(const uint8_t *buf, size_t len) 986{ 987 size_t i; 988 uint32_t crc; 989 int bit; 990 uint8_t data; 991 992 crc = 0xffffffff; /* initial value */ 993 994 for (i = 0; i < len; i++) { 995 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 996 carry = (crc ^ data) & 1; 997 crc >>= 1; 998 if (carry) 999 crc = (crc ^ ETHER_CRC_POLY_LE); 1000 } 1001 } 1002 1003 return (crc); 1004} 1005#else 1006uint32_t 1007ether_crc32_le(const uint8_t *buf, size_t len) 1008{ 1009 static const uint32_t crctab[] = { 1010 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1011 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1012 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1013 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1014 }; 1015 size_t i; 1016 uint32_t crc; 1017 1018 crc = 0xffffffff; /* initial value */ 1019 1020 for (i = 0; i < len; i++) { 1021 crc ^= buf[i]; 1022 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1023 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1024 } 1025 1026 return (crc); 1027} 1028#endif 1029 1030uint32_t 1031ether_crc32_be(const uint8_t *buf, size_t len) 1032{ 1033 size_t i; 1034 uint32_t crc, carry; 1035 int bit; 1036 uint8_t data; 1037 1038 crc = 0xffffffff; /* initial value */ 1039 1040 for (i = 0; i < len; i++) { 1041 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1042 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1043 crc <<= 1; 1044 if (carry) 1045 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1046 } 1047 } 1048 1049 return (crc); 1050} 1051 1052int 1053ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1054{ 1055 struct ifaddr *ifa = (struct ifaddr *) data; 1056 struct ifreq *ifr = (struct ifreq *) data; 1057 int error = 0; 1058 1059 switch (command) { 1060 case SIOCSIFADDR: 1061 ifp->if_flags |= IFF_UP; 1062 1063 switch (ifa->ifa_addr->sa_family) { 1064#ifdef INET 1065 case AF_INET: 1066 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1067 arp_ifinit(ifp, ifa); 1068 break; 1069#endif 1070#ifdef IPX 1071 /* 1072 * XXX - This code is probably wrong 1073 */ 1074 case AF_IPX: 1075 { 1076 struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr); 1077 1078 if (ipx_nullhost(*ina)) 1079 ina->x_host = 1080 *(union ipx_host *) 1081 IF_LLADDR(ifp); 1082 else { 1083 bcopy((caddr_t) ina->x_host.c_host, 1084 (caddr_t) IF_LLADDR(ifp), 1085 ETHER_ADDR_LEN); 1086 } 1087 1088 /* 1089 * Set new address 1090 */ 1091 ifp->if_init(ifp->if_softc); 1092 break; 1093 } 1094#endif 1095 default: 1096 ifp->if_init(ifp->if_softc); 1097 break; 1098 } 1099 break; 1100 1101 case SIOCGIFADDR: 1102 { 1103 struct sockaddr *sa; 1104 1105 sa = (struct sockaddr *) & ifr->ifr_data; 1106 bcopy(IF_LLADDR(ifp), 1107 (caddr_t) sa->sa_data, ETHER_ADDR_LEN); 1108 } 1109 break; 1110 1111 case SIOCSIFMTU: 1112 /* 1113 * Set the interface MTU. 1114 */ 1115 if (ifr->ifr_mtu > ETHERMTU) { 1116 error = EINVAL; 1117 } else { 1118 ifp->if_mtu = ifr->ifr_mtu; 1119 } 1120 break; 1121 default: 1122 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1123 break; 1124 } 1125 return (error); 1126} 1127 1128static int 1129ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1130 struct sockaddr *sa) 1131{ 1132 struct sockaddr_dl *sdl; 1133#ifdef INET 1134 struct sockaddr_in *sin; 1135#endif 1136#ifdef INET6 1137 struct sockaddr_in6 *sin6; 1138#endif 1139 u_char *e_addr; 1140 1141 switch(sa->sa_family) { 1142 case AF_LINK: 1143 /* 1144 * No mapping needed. Just check that it's a valid MC address. 1145 */ 1146 sdl = (struct sockaddr_dl *)sa; 1147 e_addr = LLADDR(sdl); 1148 if (!ETHER_IS_MULTICAST(e_addr)) 1149 return EADDRNOTAVAIL; 1150 *llsa = 0; 1151 return 0; 1152 1153#ifdef INET 1154 case AF_INET: 1155 sin = (struct sockaddr_in *)sa; 1156 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1157 return EADDRNOTAVAIL; 1158 sdl = malloc(sizeof *sdl, M_IFMADDR, 1159 M_NOWAIT|M_ZERO); 1160 if (sdl == NULL) 1161 return ENOMEM; 1162 sdl->sdl_len = sizeof *sdl; 1163 sdl->sdl_family = AF_LINK; 1164 sdl->sdl_index = ifp->if_index; 1165 sdl->sdl_type = IFT_ETHER; 1166 sdl->sdl_alen = ETHER_ADDR_LEN; 1167 e_addr = LLADDR(sdl); 1168 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1169 *llsa = (struct sockaddr *)sdl; 1170 return 0; 1171#endif 1172#ifdef INET6 1173 case AF_INET6: 1174 sin6 = (struct sockaddr_in6 *)sa; 1175 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1176 /* 1177 * An IP6 address of 0 means listen to all 1178 * of the Ethernet multicast address used for IP6. 1179 * (This is used for multicast routers.) 1180 */ 1181 ifp->if_flags |= IFF_ALLMULTI; 1182 *llsa = 0; 1183 return 0; 1184 } 1185 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1186 return EADDRNOTAVAIL; 1187 sdl = malloc(sizeof *sdl, M_IFMADDR, 1188 M_NOWAIT|M_ZERO); 1189 if (sdl == NULL) 1190 return (ENOMEM); 1191 sdl->sdl_len = sizeof *sdl; 1192 sdl->sdl_family = AF_LINK; 1193 sdl->sdl_index = ifp->if_index; 1194 sdl->sdl_type = IFT_ETHER; 1195 sdl->sdl_alen = ETHER_ADDR_LEN; 1196 e_addr = LLADDR(sdl); 1197 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1198 *llsa = (struct sockaddr *)sdl; 1199 return 0; 1200#endif 1201 1202 default: 1203 /* 1204 * Well, the text isn't quite right, but it's the name 1205 * that counts... 1206 */ 1207 return EAFNOSUPPORT; 1208 } 1209} 1210 1211static void* 1212ether_alloc(u_char type, struct ifnet *ifp) 1213{ 1214 struct arpcom *ac; 1215 1216 ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO); 1217 ac->ac_ifp = ifp; 1218 1219 return (ac); 1220} 1221 1222static void 1223ether_free(void *com, u_char type) 1224{ 1225 1226 free(com, M_ARPCOM); 1227} 1228 1229static int 1230ether_modevent(module_t mod, int type, void *data) 1231{ 1232 1233 switch (type) { 1234 case MOD_LOAD: 1235 if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free); 1236 break; 1237 case MOD_UNLOAD: 1238 if_deregister_com_alloc(IFT_ETHER); 1239 break; 1240 default: 1241 return EOPNOTSUPP; 1242 } 1243 1244 return (0); 1245} 1246 1247static moduledata_t ether_mod = { 1248 "ether", 1249 ether_modevent, 1250 0 1251}; 1252 1253void 1254ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1255{ 1256 struct ether_vlan_header vlan; 1257 struct mbuf mv, mb; 1258 1259 KASSERT((m->m_flags & M_VLANTAG) != 0, 1260 ("%s: vlan information not present", __func__)); 1261 KASSERT(m->m_len >= sizeof(struct ether_header), 1262 ("%s: mbuf not large enough for header", __func__)); 1263 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1264 vlan.evl_proto = vlan.evl_encap_proto; 1265 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1266 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1267 m->m_len -= sizeof(struct ether_header); 1268 m->m_data += sizeof(struct ether_header); 1269 /* 1270 * If a data link has been supplied by the caller, then we will need to 1271 * re-create a stack allocated mbuf chain with the following structure: 1272 * 1273 * (1) mbuf #1 will contain the supplied data link 1274 * (2) mbuf #2 will contain the vlan header 1275 * (3) mbuf #3 will contain the original mbuf's packet data 1276 * 1277 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1278 */ 1279 if (data != NULL) { 1280 mv.m_next = m; 1281 mv.m_data = (caddr_t)&vlan; 1282 mv.m_len = sizeof(vlan); 1283 mb.m_next = &mv; 1284 mb.m_data = data; 1285 mb.m_len = dlen; 1286 bpf_mtap(bp, &mb); 1287 } else 1288 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1289 m->m_len += sizeof(struct ether_header); 1290 m->m_data -= sizeof(struct ether_header); 1291} 1292 1293struct mbuf * 1294ether_vlanencap(struct mbuf *m, uint16_t tag) 1295{ 1296 struct ether_vlan_header *evl; 1297 1298 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1299 if (m == NULL) 1300 return (NULL); 1301 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1302 1303 if (m->m_len < sizeof(*evl)) { 1304 m = m_pullup(m, sizeof(*evl)); 1305 if (m == NULL) 1306 return (NULL); 1307 } 1308 1309 /* 1310 * Transform the Ethernet header into an Ethernet header 1311 * with 802.1Q encapsulation. 1312 */ 1313 evl = mtod(m, struct ether_vlan_header *); 1314 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1315 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1316 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1317 evl->evl_tag = htons(tag); 1318 return (m); 1319} 1320 1321DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1322MODULE_VERSION(ether, 1); 1323