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_inet.h" 34 #include "opt_inet6.h" 35 #include "opt_netgraph.h" 36 #include "opt_mbuf_profiling.h" 37 #include "opt_rss.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/malloc.h> 44 #include <sys/module.h> 45 #include <sys/mbuf.h> 46 #include <sys/random.h> 47 #include <sys/socket.h> 48 #include <sys/sockio.h> 49 #include <sys/sysctl.h> 50 #include <sys/uuid.h> 51 52 #include <net/if.h> 53 #include <net/if_var.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/in_rss.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 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); 83 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, 84 const struct sockaddr *dst, short *tp, int *hlen); 85 86 #include <security/mac/mac_framework.h> 87 88 #ifdef CTASSERT 89 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 90 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 91 #endif 92 93 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 94 95 /* netgraph node hooks for ng_ether(4) */ 96 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 97 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 98 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 99 void (*ng_ether_attach_p)(struct ifnet *ifp); 100 void (*ng_ether_detach_p)(struct ifnet *ifp); 101 102 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 103 104 /* if_bridge(4) support */ 105 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 106 int (*bridge_output_p)(struct ifnet *, struct mbuf *, 107 struct sockaddr *, struct rtentry *); 108 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 109 110 /* if_lagg(4) support */ 111 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); 112 113 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 114 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 115 116 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 117 struct sockaddr *); 118 #ifdef VIMAGE 119 static void ether_reassign(struct ifnet *, struct vnet *, char *); 120 #endif 121 122 /* XXX: should be in an arp support file, not here */ 123 static MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals"); 124 125 #define ETHER_IS_BROADCAST(addr) \ 126 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 127 128 #define senderr(e) do { error = (e); goto bad;} while (0) 129 130 static void 131 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 132 { 133 int csum_flags = 0; 134 135 if (src->m_pkthdr.csum_flags & CSUM_IP) 136 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 137 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 138 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 139 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 140 csum_flags |= CSUM_SCTP_VALID; 141 dst->m_pkthdr.csum_flags |= csum_flags; 142 if (csum_flags & CSUM_DATA_VALID) 143 dst->m_pkthdr.csum_data = 0xffff; 144 } 145 146 /* 147 * Ethernet output routine. 148 * Encapsulate a packet of type family for the local net. 149 * Use trailer local net encapsulation if enough data in first 150 * packet leaves a multiple of 512 bytes of data in remainder. 151 */ 152 int 153 ether_output(struct ifnet *ifp, struct mbuf *m, 154 const struct sockaddr *dst, struct route *ro) 155 { 156 short type; 157 int error = 0, hdrcmplt = 0; 158 u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 159 struct llentry *lle = NULL; 160 struct rtentry *rt0 = NULL; 161 struct ether_header *eh; 162 struct pf_mtag *t; 163 int loop_copy = 1; 164 int hlen; /* link layer header length */ 165 166 if (ro != NULL) { 167 if (!(m->m_flags & (M_BCAST | M_MCAST))) 168 lle = ro->ro_lle; 169 rt0 = ro->ro_rt; 170 } 171 #ifdef MAC 172 error = mac_ifnet_check_transmit(ifp, m); 173 if (error) 174 senderr(error); 175 #endif 176 177 M_PROFILE(m); 178 if (ifp->if_flags & IFF_MONITOR) 179 senderr(ENETDOWN); 180 if (!((ifp->if_flags & IFF_UP) && 181 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 182 senderr(ENETDOWN); 183 184 hlen = ETHER_HDR_LEN; 185 switch (dst->sa_family) { 186 #ifdef INET 187 case AF_INET: 188 if (lle != NULL && (lle->la_flags & LLE_VALID)) 189 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 190 else 191 error = arpresolve(ifp, rt0, m, dst, edst, &lle); 192 if (error) 193 return (error == EWOULDBLOCK ? 0 : error); 194 type = htons(ETHERTYPE_IP); 195 break; 196 case AF_ARP: 197 { 198 struct arphdr *ah; 199 ah = mtod(m, struct arphdr *); 200 ah->ar_hrd = htons(ARPHRD_ETHER); 201 202 loop_copy = 0; /* if this is for us, don't do it */ 203 204 switch(ntohs(ah->ar_op)) { 205 case ARPOP_REVREQUEST: 206 case ARPOP_REVREPLY: 207 type = htons(ETHERTYPE_REVARP); 208 break; 209 case ARPOP_REQUEST: 210 case ARPOP_REPLY: 211 default: 212 type = htons(ETHERTYPE_ARP); 213 break; 214 } 215 216 if (m->m_flags & M_BCAST) 217 bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); 218 else 219 bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); 220 221 } 222 break; 223 #endif 224 #ifdef INET6 225 case AF_INET6: 226 if (lle != NULL && (lle->la_flags & LLE_VALID)) 227 memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); 228 else 229 error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle); 230 if (error) 231 return error; 232 type = htons(ETHERTYPE_IPV6); 233 break; 234 #endif 235 case pseudo_AF_HDRCMPLT: 236 { 237 const struct ether_header *eh; 238 239 hdrcmplt = 1; 240 eh = (const struct ether_header *)dst->sa_data; 241 (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); 242 /* FALLTHROUGH */ 243 244 case AF_UNSPEC: 245 loop_copy = 0; /* if this is for us, don't do it */ 246 eh = (const struct ether_header *)dst->sa_data; 247 (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); 248 type = eh->ether_type; 249 break; 250 } 251 default: 252 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 253 senderr(EAFNOSUPPORT); 254 } 255 256 if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { 257 update_mbuf_csumflags(m, m); 258 return (if_simloop(ifp, m, dst->sa_family, 0)); 259 } 260 261 /* 262 * Add local net header. If no space in first mbuf, 263 * allocate another. 264 */ 265 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 266 if (m == NULL) 267 senderr(ENOBUFS); 268 eh = mtod(m, struct ether_header *); 269 (void)memcpy(&eh->ether_type, &type, 270 sizeof(eh->ether_type)); 271 (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); 272 if (hdrcmplt) 273 (void)memcpy(eh->ether_shost, esrc, 274 sizeof(eh->ether_shost)); 275 else 276 (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), 277 sizeof(eh->ether_shost)); 278 279 /* 280 * If a simplex interface, and the packet is being sent to our 281 * Ethernet address or a broadcast address, loopback a copy. 282 * XXX To make a simplex device behave exactly like a duplex 283 * device, we should copy in the case of sending to our own 284 * ethernet address (thus letting the original actually appear 285 * on the wire). However, we don't do that here for security 286 * reasons and compatibility with the original behavior. 287 */ 288 if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && 289 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 290 if (m->m_flags & M_BCAST) { 291 struct mbuf *n; 292 293 /* 294 * Because if_simloop() modifies the packet, we need a 295 * writable copy through m_dup() instead of a readonly 296 * one as m_copy[m] would give us. The alternative would 297 * be to modify if_simloop() to handle the readonly mbuf, 298 * but performancewise it is mostly equivalent (trading 299 * extra data copying vs. extra locking). 300 * 301 * XXX This is a local workaround. A number of less 302 * often used kernel parts suffer from the same bug. 303 * See PR kern/105943 for a proposed general solution. 304 */ 305 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 306 update_mbuf_csumflags(m, n); 307 (void)if_simloop(ifp, n, dst->sa_family, hlen); 308 } else 309 ifp->if_iqdrops++; 310 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 311 ETHER_ADDR_LEN) == 0) { 312 update_mbuf_csumflags(m, m); 313 (void) if_simloop(ifp, m, dst->sa_family, hlen); 314 return (0); /* XXX */ 315 } 316 } 317 318 /* 319 * Bridges require special output handling. 320 */ 321 if (ifp->if_bridge) { 322 BRIDGE_OUTPUT(ifp, m, error); 323 return (error); 324 } 325 326 #if defined(INET) || defined(INET6) 327 if (ifp->if_carp && 328 (error = (*carp_output_p)(ifp, m, dst))) 329 goto bad; 330 #endif 331 332 /* Handle ng_ether(4) processing, if any */ 333 if (IFP2AC(ifp)->ac_netgraph != NULL) { 334 KASSERT(ng_ether_output_p != NULL, 335 ("ng_ether_output_p is NULL")); 336 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 337 bad: if (m != NULL) 338 m_freem(m); 339 return (error); 340 } 341 if (m == NULL) 342 return (0); 343 } 344 345 /* Continue with link-layer output */ 346 return ether_output_frame(ifp, m); 347 } 348 349 /* 350 * Ethernet link layer output routine to send a raw frame to the device. 351 * 352 * This assumes that the 14 byte Ethernet header is present and contiguous 353 * in the first mbuf (if BRIDGE'ing). 354 */ 355 int 356 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 357 { 358 int i; 359 360 if (PFIL_HOOKED(&V_link_pfil_hook)) { 361 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); 362 363 if (i != 0) 364 return (EACCES); 365 366 if (m == NULL) 367 return (0); 368 } 369 370 /* 371 * Queue message on interface, update output statistics if 372 * successful, and start output if interface not yet active. 373 */ 374 return ((ifp->if_transmit)(ifp, m)); 375 } 376 377 #if defined(INET) || defined(INET6) 378 #endif 379 380 /* 381 * Process a received Ethernet packet; the packet is in the 382 * mbuf chain m with the ethernet header at the front. 383 */ 384 static void 385 ether_input_internal(struct ifnet *ifp, struct mbuf *m) 386 { 387 struct ether_header *eh; 388 u_short etype; 389 390 if ((ifp->if_flags & IFF_UP) == 0) { 391 m_freem(m); 392 return; 393 } 394 #ifdef DIAGNOSTIC 395 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 396 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 397 m_freem(m); 398 return; 399 } 400 #endif 401 /* 402 * Do consistency checks to verify assumptions 403 * made by code past this point. 404 */ 405 if ((m->m_flags & M_PKTHDR) == 0) { 406 if_printf(ifp, "discard frame w/o packet header\n"); 407 ifp->if_ierrors++; 408 m_freem(m); 409 return; 410 } 411 if (m->m_len < ETHER_HDR_LEN) { 412 /* XXX maybe should pullup? */ 413 if_printf(ifp, "discard frame w/o leading ethernet " 414 "header (len %u pkt len %u)\n", 415 m->m_len, m->m_pkthdr.len); 416 ifp->if_ierrors++; 417 m_freem(m); 418 return; 419 } 420 eh = mtod(m, struct ether_header *); 421 etype = ntohs(eh->ether_type); 422 if (m->m_pkthdr.rcvif == NULL) { 423 if_printf(ifp, "discard frame w/o interface pointer\n"); 424 ifp->if_ierrors++; 425 m_freem(m); 426 return; 427 } 428 #ifdef DIAGNOSTIC 429 if (m->m_pkthdr.rcvif != ifp) { 430 if_printf(ifp, "Warning, frame marked as received on %s\n", 431 m->m_pkthdr.rcvif->if_xname); 432 } 433 #endif 434 435 CURVNET_SET_QUIET(ifp->if_vnet); 436 437 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 438 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 439 m->m_flags |= M_BCAST; 440 else 441 m->m_flags |= M_MCAST; 442 ifp->if_imcasts++; 443 } 444 445 #ifdef MAC 446 /* 447 * Tag the mbuf with an appropriate MAC label before any other 448 * consumers can get to it. 449 */ 450 mac_ifnet_create_mbuf(ifp, m); 451 #endif 452 453 /* 454 * Give bpf a chance at the packet. 455 */ 456 ETHER_BPF_MTAP(ifp, m); 457 458 /* 459 * If the CRC is still on the packet, trim it off. We do this once 460 * and once only in case we are re-entered. Nothing else on the 461 * Ethernet receive path expects to see the FCS. 462 */ 463 if (m->m_flags & M_HASFCS) { 464 m_adj(m, -ETHER_CRC_LEN); 465 m->m_flags &= ~M_HASFCS; 466 } 467 468 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 469 ifp->if_ibytes += m->m_pkthdr.len; 470 471 /* Allow monitor mode to claim this frame, after stats are updated. */ 472 if (ifp->if_flags & IFF_MONITOR) { 473 m_freem(m); 474 CURVNET_RESTORE(); 475 return; 476 } 477 478 /* Handle input from a lagg(4) port */ 479 if (ifp->if_type == IFT_IEEE8023ADLAG) { 480 KASSERT(lagg_input_p != NULL, 481 ("%s: if_lagg not loaded!", __func__)); 482 m = (*lagg_input_p)(ifp, m); 483 if (m != NULL) 484 ifp = m->m_pkthdr.rcvif; 485 else { 486 CURVNET_RESTORE(); 487 return; 488 } 489 } 490 491 /* 492 * If the hardware did not process an 802.1Q tag, do this now, 493 * to allow 802.1P priority frames to be passed to the main input 494 * path correctly. 495 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 496 */ 497 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 498 struct ether_vlan_header *evl; 499 500 if (m->m_len < sizeof(*evl) && 501 (m = m_pullup(m, sizeof(*evl))) == NULL) { 502 #ifdef DIAGNOSTIC 503 if_printf(ifp, "cannot pullup VLAN header\n"); 504 #endif 505 ifp->if_ierrors++; 506 m_freem(m); 507 CURVNET_RESTORE(); 508 return; 509 } 510 511 evl = mtod(m, struct ether_vlan_header *); 512 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 513 m->m_flags |= M_VLANTAG; 514 515 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 516 ETHER_HDR_LEN - ETHER_TYPE_LEN); 517 m_adj(m, ETHER_VLAN_ENCAP_LEN); 518 eh = mtod(m, struct ether_header *); 519 } 520 521 M_SETFIB(m, ifp->if_fib); 522 523 /* Allow ng_ether(4) to claim this frame. */ 524 if (IFP2AC(ifp)->ac_netgraph != NULL) { 525 KASSERT(ng_ether_input_p != NULL, 526 ("%s: ng_ether_input_p is NULL", __func__)); 527 m->m_flags &= ~M_PROMISC; 528 (*ng_ether_input_p)(ifp, &m); 529 if (m == NULL) { 530 CURVNET_RESTORE(); 531 return; 532 } 533 eh = mtod(m, struct ether_header *); 534 } 535 536 /* 537 * Allow if_bridge(4) to claim this frame. 538 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 539 * and the frame should be delivered locally. 540 */ 541 if (ifp->if_bridge != NULL) { 542 m->m_flags &= ~M_PROMISC; 543 BRIDGE_INPUT(ifp, m); 544 if (m == NULL) { 545 CURVNET_RESTORE(); 546 return; 547 } 548 eh = mtod(m, struct ether_header *); 549 } 550 551 #if defined(INET) || defined(INET6) 552 /* 553 * Clear M_PROMISC on frame so that carp(4) will see it when the 554 * mbuf flows up to Layer 3. 555 * FreeBSD's implementation of carp(4) uses the inprotosw 556 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 557 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 558 * is outside the scope of the M_PROMISC test below. 559 * TODO: Maintain a hash table of ethernet addresses other than 560 * ether_dhost which may be active on this ifp. 561 */ 562 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 563 m->m_flags &= ~M_PROMISC; 564 } else 565 #endif 566 { 567 /* 568 * If the frame received was not for our MAC address, set the 569 * M_PROMISC flag on the mbuf chain. The frame may need to 570 * be seen by the rest of the Ethernet input path in case of 571 * re-entry (e.g. bridge, vlan, netgraph) but should not be 572 * seen by upper protocol layers. 573 */ 574 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 575 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 576 m->m_flags |= M_PROMISC; 577 } 578 579 if (harvest.ethernet) 580 random_harvest(&(m->m_data), 12, 2, RANDOM_NET_ETHER); 581 582 ether_demux(ifp, m); 583 CURVNET_RESTORE(); 584 } 585 586 /* 587 * Ethernet input dispatch; by default, direct dispatch here regardless of 588 * global configuration. However, if RSS is enabled, hook up RSS affinity 589 * so that when deferred or hybrid dispatch is enabled, we can redistribute 590 * load based on RSS. 591 * 592 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or 593 * not it had already done work distribution via multi-queue. Then we could 594 * direct dispatch in the event load balancing was already complete and 595 * handle the case of interfaces with different capabilities better. 596 * 597 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions 598 * at multiple layers? 599 * 600 * XXXRW: For now, enable all this only if RSS is compiled in, although it 601 * works fine without RSS. Need to characterise the performance overhead 602 * of the detour through the netisr code in the event the result is always 603 * direct dispatch. 604 */ 605 static void 606 ether_nh_input(struct mbuf *m) 607 { 608 609 ether_input_internal(m->m_pkthdr.rcvif, m); 610 } 611 612 static struct netisr_handler ether_nh = { 613 .nh_name = "ether", 614 .nh_handler = ether_nh_input, 615 .nh_proto = NETISR_ETHER, 616 #ifdef RSS 617 .nh_policy = NETISR_POLICY_CPU, 618 .nh_dispatch = NETISR_DISPATCH_DIRECT, 619 .nh_m2cpuid = rss_m2cpuid, 620 #else 621 .nh_policy = NETISR_POLICY_SOURCE, 622 .nh_dispatch = NETISR_DISPATCH_DIRECT, 623 #endif 624 }; 625 626 static void 627 ether_init(__unused void *arg) 628 { 629 630 netisr_register(ðer_nh); 631 } 632 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 633 634 static void 635 vnet_ether_init(__unused void *arg) 636 { 637 int i; 638 639 /* Initialize packet filter hooks. */ 640 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 641 V_link_pfil_hook.ph_af = AF_LINK; 642 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 643 printf("%s: WARNING: unable to register pfil link hook, " 644 "error %d\n", __func__, i); 645 } 646 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 647 vnet_ether_init, NULL); 648 649 static void 650 vnet_ether_destroy(__unused void *arg) 651 { 652 int i; 653 654 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 655 printf("%s: WARNING: unable to unregister pfil link hook, " 656 "error %d\n", __func__, i); 657 } 658 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 659 vnet_ether_destroy, NULL); 660 661 662 663 static void 664 ether_input(struct ifnet *ifp, struct mbuf *m) 665 { 666 667 struct mbuf *mn; 668 669 /* 670 * The drivers are allowed to pass in a chain of packets linked with 671 * m_nextpkt. We split them up into separate packets here and pass 672 * them up. This allows the drivers to amortize the receive lock. 673 */ 674 while (m) { 675 mn = m->m_nextpkt; 676 m->m_nextpkt = NULL; 677 678 /* 679 * We will rely on rcvif being set properly in the deferred context, 680 * so assert it is correct here. 681 */ 682 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__)); 683 netisr_dispatch(NETISR_ETHER, m); 684 m = mn; 685 } 686 } 687 688 /* 689 * Upper layer processing for a received Ethernet packet. 690 */ 691 void 692 ether_demux(struct ifnet *ifp, struct mbuf *m) 693 { 694 struct ether_header *eh; 695 int i, isr; 696 u_short ether_type; 697 698 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 699 700 /* Do not grab PROMISC frames in case we are re-entered. */ 701 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 702 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); 703 704 if (i != 0 || m == NULL) 705 return; 706 } 707 708 eh = mtod(m, struct ether_header *); 709 ether_type = ntohs(eh->ether_type); 710 711 /* 712 * If this frame has a VLAN tag other than 0, call vlan_input() 713 * if its module is loaded. Otherwise, drop. 714 */ 715 if ((m->m_flags & M_VLANTAG) && 716 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 717 if (ifp->if_vlantrunk == NULL) { 718 ifp->if_noproto++; 719 m_freem(m); 720 return; 721 } 722 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 723 __func__)); 724 /* Clear before possibly re-entering ether_input(). */ 725 m->m_flags &= ~M_PROMISC; 726 (*vlan_input_p)(ifp, m); 727 return; 728 } 729 730 /* 731 * Pass promiscuously received frames to the upper layer if the user 732 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 733 */ 734 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 735 m_freem(m); 736 return; 737 } 738 739 /* 740 * Reset layer specific mbuf flags to avoid confusing upper layers. 741 * Strip off Ethernet header. 742 */ 743 m->m_flags &= ~M_VLANTAG; 744 m_clrprotoflags(m); 745 m_adj(m, ETHER_HDR_LEN); 746 747 /* 748 * Dispatch frame to upper layer. 749 */ 750 switch (ether_type) { 751 #ifdef INET 752 case ETHERTYPE_IP: 753 if ((m = ip_fastforward(m)) == NULL) 754 return; 755 isr = NETISR_IP; 756 break; 757 758 case ETHERTYPE_ARP: 759 if (ifp->if_flags & IFF_NOARP) { 760 /* Discard packet if ARP is disabled on interface */ 761 m_freem(m); 762 return; 763 } 764 isr = NETISR_ARP; 765 break; 766 #endif 767 #ifdef INET6 768 case ETHERTYPE_IPV6: 769 isr = NETISR_IPV6; 770 break; 771 #endif 772 default: 773 goto discard; 774 } 775 netisr_dispatch(isr, m); 776 return; 777 778 discard: 779 /* 780 * Packet is to be discarded. If netgraph is present, 781 * hand the packet to it for last chance processing; 782 * otherwise dispose of it. 783 */ 784 if (IFP2AC(ifp)->ac_netgraph != NULL) { 785 KASSERT(ng_ether_input_orphan_p != NULL, 786 ("ng_ether_input_orphan_p is NULL")); 787 /* 788 * Put back the ethernet header so netgraph has a 789 * consistent view of inbound packets. 790 */ 791 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 792 (*ng_ether_input_orphan_p)(ifp, m); 793 return; 794 } 795 m_freem(m); 796 } 797 798 /* 799 * Convert Ethernet address to printable (loggable) representation. 800 * This routine is for compatibility; it's better to just use 801 * 802 * printf("%6D", <pointer to address>, ":"); 803 * 804 * since there's no static buffer involved. 805 */ 806 char * 807 ether_sprintf(const u_char *ap) 808 { 809 static char etherbuf[18]; 810 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 811 return (etherbuf); 812 } 813 814 /* 815 * Perform common duties while attaching to interface list 816 */ 817 void 818 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 819 { 820 int i; 821 struct ifaddr *ifa; 822 struct sockaddr_dl *sdl; 823 824 ifp->if_addrlen = ETHER_ADDR_LEN; 825 ifp->if_hdrlen = ETHER_HDR_LEN; 826 if_attach(ifp); 827 ifp->if_mtu = ETHERMTU; 828 ifp->if_output = ether_output; 829 ifp->if_input = ether_input; 830 ifp->if_resolvemulti = ether_resolvemulti; 831 #ifdef VIMAGE 832 ifp->if_reassign = ether_reassign; 833 #endif 834 if (ifp->if_baudrate == 0) 835 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 836 ifp->if_broadcastaddr = etherbroadcastaddr; 837 838 ifa = ifp->if_addr; 839 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 840 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 841 sdl->sdl_type = IFT_ETHER; 842 sdl->sdl_alen = ifp->if_addrlen; 843 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 844 845 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 846 if (ng_ether_attach_p != NULL) 847 (*ng_ether_attach_p)(ifp); 848 849 /* Announce Ethernet MAC address if non-zero. */ 850 for (i = 0; i < ifp->if_addrlen; i++) 851 if (lla[i] != 0) 852 break; 853 if (i != ifp->if_addrlen) 854 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 855 856 uuid_ether_add(LLADDR(sdl)); 857 } 858 859 /* 860 * Perform common duties while detaching an Ethernet interface 861 */ 862 void 863 ether_ifdetach(struct ifnet *ifp) 864 { 865 struct sockaddr_dl *sdl; 866 867 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 868 uuid_ether_del(LLADDR(sdl)); 869 870 if (IFP2AC(ifp)->ac_netgraph != NULL) { 871 KASSERT(ng_ether_detach_p != NULL, 872 ("ng_ether_detach_p is NULL")); 873 (*ng_ether_detach_p)(ifp); 874 } 875 876 bpfdetach(ifp); 877 if_detach(ifp); 878 } 879 880 #ifdef VIMAGE 881 void 882 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 883 { 884 885 if (IFP2AC(ifp)->ac_netgraph != NULL) { 886 KASSERT(ng_ether_detach_p != NULL, 887 ("ng_ether_detach_p is NULL")); 888 (*ng_ether_detach_p)(ifp); 889 } 890 891 if (ng_ether_attach_p != NULL) { 892 CURVNET_SET_QUIET(new_vnet); 893 (*ng_ether_attach_p)(ifp); 894 CURVNET_RESTORE(); 895 } 896 } 897 #endif 898 899 SYSCTL_DECL(_net_link); 900 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 901 902 #if 0 903 /* 904 * This is for reference. We have a table-driven version 905 * of the little-endian crc32 generator, which is faster 906 * than the double-loop. 907 */ 908 uint32_t 909 ether_crc32_le(const uint8_t *buf, size_t len) 910 { 911 size_t i; 912 uint32_t crc; 913 int bit; 914 uint8_t data; 915 916 crc = 0xffffffff; /* initial value */ 917 918 for (i = 0; i < len; i++) { 919 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 920 carry = (crc ^ data) & 1; 921 crc >>= 1; 922 if (carry) 923 crc = (crc ^ ETHER_CRC_POLY_LE); 924 } 925 } 926 927 return (crc); 928 } 929 #else 930 uint32_t 931 ether_crc32_le(const uint8_t *buf, size_t len) 932 { 933 static const uint32_t crctab[] = { 934 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 935 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 936 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 937 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 938 }; 939 size_t i; 940 uint32_t crc; 941 942 crc = 0xffffffff; /* initial value */ 943 944 for (i = 0; i < len; i++) { 945 crc ^= buf[i]; 946 crc = (crc >> 4) ^ crctab[crc & 0xf]; 947 crc = (crc >> 4) ^ crctab[crc & 0xf]; 948 } 949 950 return (crc); 951 } 952 #endif 953 954 uint32_t 955 ether_crc32_be(const uint8_t *buf, size_t len) 956 { 957 size_t i; 958 uint32_t crc, carry; 959 int bit; 960 uint8_t data; 961 962 crc = 0xffffffff; /* initial value */ 963 964 for (i = 0; i < len; i++) { 965 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 966 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 967 crc <<= 1; 968 if (carry) 969 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 970 } 971 } 972 973 return (crc); 974 } 975 976 int 977 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 978 { 979 struct ifaddr *ifa = (struct ifaddr *) data; 980 struct ifreq *ifr = (struct ifreq *) data; 981 int error = 0; 982 983 switch (command) { 984 case SIOCSIFADDR: 985 ifp->if_flags |= IFF_UP; 986 987 switch (ifa->ifa_addr->sa_family) { 988 #ifdef INET 989 case AF_INET: 990 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 991 arp_ifinit(ifp, ifa); 992 break; 993 #endif 994 default: 995 ifp->if_init(ifp->if_softc); 996 break; 997 } 998 break; 999 1000 case SIOCGIFADDR: 1001 { 1002 struct sockaddr *sa; 1003 1004 sa = (struct sockaddr *) & ifr->ifr_data; 1005 bcopy(IF_LLADDR(ifp), 1006 (caddr_t) sa->sa_data, ETHER_ADDR_LEN); 1007 } 1008 break; 1009 1010 case SIOCSIFMTU: 1011 /* 1012 * Set the interface MTU. 1013 */ 1014 if (ifr->ifr_mtu > ETHERMTU) { 1015 error = EINVAL; 1016 } else { 1017 ifp->if_mtu = ifr->ifr_mtu; 1018 } 1019 break; 1020 default: 1021 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1022 break; 1023 } 1024 return (error); 1025 } 1026 1027 static int 1028 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1029 struct sockaddr *sa) 1030 { 1031 struct sockaddr_dl *sdl; 1032 #ifdef INET 1033 struct sockaddr_in *sin; 1034 #endif 1035 #ifdef INET6 1036 struct sockaddr_in6 *sin6; 1037 #endif 1038 u_char *e_addr; 1039 1040 switch(sa->sa_family) { 1041 case AF_LINK: 1042 /* 1043 * No mapping needed. Just check that it's a valid MC address. 1044 */ 1045 sdl = (struct sockaddr_dl *)sa; 1046 e_addr = LLADDR(sdl); 1047 if (!ETHER_IS_MULTICAST(e_addr)) 1048 return EADDRNOTAVAIL; 1049 *llsa = 0; 1050 return 0; 1051 1052 #ifdef INET 1053 case AF_INET: 1054 sin = (struct sockaddr_in *)sa; 1055 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1056 return EADDRNOTAVAIL; 1057 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1058 sdl->sdl_alen = ETHER_ADDR_LEN; 1059 e_addr = LLADDR(sdl); 1060 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1061 *llsa = (struct sockaddr *)sdl; 1062 return 0; 1063 #endif 1064 #ifdef INET6 1065 case AF_INET6: 1066 sin6 = (struct sockaddr_in6 *)sa; 1067 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1068 /* 1069 * An IP6 address of 0 means listen to all 1070 * of the Ethernet multicast address used for IP6. 1071 * (This is used for multicast routers.) 1072 */ 1073 ifp->if_flags |= IFF_ALLMULTI; 1074 *llsa = 0; 1075 return 0; 1076 } 1077 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1078 return EADDRNOTAVAIL; 1079 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1080 sdl->sdl_alen = ETHER_ADDR_LEN; 1081 e_addr = LLADDR(sdl); 1082 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1083 *llsa = (struct sockaddr *)sdl; 1084 return 0; 1085 #endif 1086 1087 default: 1088 /* 1089 * Well, the text isn't quite right, but it's the name 1090 * that counts... 1091 */ 1092 return EAFNOSUPPORT; 1093 } 1094 } 1095 1096 static void* 1097 ether_alloc(u_char type, struct ifnet *ifp) 1098 { 1099 struct arpcom *ac; 1100 1101 ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO); 1102 ac->ac_ifp = ifp; 1103 1104 return (ac); 1105 } 1106 1107 static void 1108 ether_free(void *com, u_char type) 1109 { 1110 1111 free(com, M_ARPCOM); 1112 } 1113 1114 static int 1115 ether_modevent(module_t mod, int type, void *data) 1116 { 1117 1118 switch (type) { 1119 case MOD_LOAD: 1120 if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free); 1121 break; 1122 case MOD_UNLOAD: 1123 if_deregister_com_alloc(IFT_ETHER); 1124 break; 1125 default: 1126 return EOPNOTSUPP; 1127 } 1128 1129 return (0); 1130 } 1131 1132 static moduledata_t ether_mod = { 1133 "ether", 1134 ether_modevent, 1135 0 1136 }; 1137 1138 void 1139 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1140 { 1141 struct ether_vlan_header vlan; 1142 struct mbuf mv, mb; 1143 1144 KASSERT((m->m_flags & M_VLANTAG) != 0, 1145 ("%s: vlan information not present", __func__)); 1146 KASSERT(m->m_len >= sizeof(struct ether_header), 1147 ("%s: mbuf not large enough for header", __func__)); 1148 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1149 vlan.evl_proto = vlan.evl_encap_proto; 1150 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1151 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1152 m->m_len -= sizeof(struct ether_header); 1153 m->m_data += sizeof(struct ether_header); 1154 /* 1155 * If a data link has been supplied by the caller, then we will need to 1156 * re-create a stack allocated mbuf chain with the following structure: 1157 * 1158 * (1) mbuf #1 will contain the supplied data link 1159 * (2) mbuf #2 will contain the vlan header 1160 * (3) mbuf #3 will contain the original mbuf's packet data 1161 * 1162 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1163 */ 1164 if (data != NULL) { 1165 mv.m_next = m; 1166 mv.m_data = (caddr_t)&vlan; 1167 mv.m_len = sizeof(vlan); 1168 mb.m_next = &mv; 1169 mb.m_data = data; 1170 mb.m_len = dlen; 1171 bpf_mtap(bp, &mb); 1172 } else 1173 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1174 m->m_len += sizeof(struct ether_header); 1175 m->m_data -= sizeof(struct ether_header); 1176 } 1177 1178 struct mbuf * 1179 ether_vlanencap(struct mbuf *m, uint16_t tag) 1180 { 1181 struct ether_vlan_header *evl; 1182 1183 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1184 if (m == NULL) 1185 return (NULL); 1186 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1187 1188 if (m->m_len < sizeof(*evl)) { 1189 m = m_pullup(m, sizeof(*evl)); 1190 if (m == NULL) 1191 return (NULL); 1192 } 1193 1194 /* 1195 * Transform the Ethernet header into an Ethernet header 1196 * with 802.1Q encapsulation. 1197 */ 1198 evl = mtod(m, struct ether_vlan_header *); 1199 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1200 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1201 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1202 evl->evl_tag = htons(tag); 1203 return (m); 1204 } 1205 1206 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1207 MODULE_VERSION(ether, 1); 1208