1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 32 * $FreeBSD$ 33 */ 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 #include "opt_netgraph.h" 38 #include "opt_mbuf_profiling.h" 39 #include "opt_rss.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/devctl.h> 44 #include <sys/eventhandler.h> 45 #include <sys/jail.h> 46 #include <sys/kernel.h> 47 #include <sys/lock.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/module.h> 51 #include <sys/msan.h> 52 #include <sys/proc.h> 53 #include <sys/priv.h> 54 #include <sys/random.h> 55 #include <sys/socket.h> 56 #include <sys/sockio.h> 57 #include <sys/sysctl.h> 58 #include <sys/uuid.h> 59 #ifdef KDB 60 #include <sys/kdb.h> 61 #endif 62 63 #include <net/ieee_oui.h> 64 #include <net/if.h> 65 #include <net/if_var.h> 66 #include <net/if_private.h> 67 #include <net/if_arp.h> 68 #include <net/netisr.h> 69 #include <net/route.h> 70 #include <net/if_llc.h> 71 #include <net/if_dl.h> 72 #include <net/if_types.h> 73 #include <net/bpf.h> 74 #include <net/ethernet.h> 75 #include <net/if_bridgevar.h> 76 #include <net/if_vlan_var.h> 77 #include <net/if_llatbl.h> 78 #include <net/pfil.h> 79 #include <net/rss_config.h> 80 #include <net/vnet.h> 81 82 #include <netpfil/pf/pf_mtag.h> 83 84 #if defined(INET) || defined(INET6) 85 #include <netinet/in.h> 86 #include <netinet/in_var.h> 87 #include <netinet/if_ether.h> 88 #include <netinet/ip_carp.h> 89 #include <netinet/ip_var.h> 90 #endif 91 #ifdef INET6 92 #include <netinet6/nd6.h> 93 #endif 94 #include <security/mac/mac_framework.h> 95 96 #include <crypto/sha1.h> 97 98 #ifdef CTASSERT 99 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 100 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 101 #endif 102 103 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */ 104 105 /* netgraph node hooks for ng_ether(4) */ 106 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 107 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 108 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 109 void (*ng_ether_attach_p)(struct ifnet *ifp); 110 void (*ng_ether_detach_p)(struct ifnet *ifp); 111 112 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 113 114 /* if_bridge(4) support */ 115 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 116 117 /* if_lagg(4) support */ 118 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *); 119 120 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 121 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 122 123 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 124 struct sockaddr *); 125 static int ether_requestencap(struct ifnet *, struct if_encap_req *); 126 127 #define senderr(e) do { error = (e); goto bad;} while (0) 128 129 static void 130 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 131 { 132 int csum_flags = 0; 133 134 if (src->m_pkthdr.csum_flags & CSUM_IP) 135 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 136 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 137 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 138 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 139 csum_flags |= CSUM_SCTP_VALID; 140 dst->m_pkthdr.csum_flags |= csum_flags; 141 if (csum_flags & CSUM_DATA_VALID) 142 dst->m_pkthdr.csum_data = 0xffff; 143 } 144 145 /* 146 * Handle link-layer encapsulation requests. 147 */ 148 static int 149 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req) 150 { 151 struct ether_header *eh; 152 struct arphdr *ah; 153 uint16_t etype; 154 const u_char *lladdr; 155 156 if (req->rtype != IFENCAP_LL) 157 return (EOPNOTSUPP); 158 159 if (req->bufsize < ETHER_HDR_LEN) 160 return (ENOMEM); 161 162 eh = (struct ether_header *)req->buf; 163 lladdr = req->lladdr; 164 req->lladdr_off = 0; 165 166 switch (req->family) { 167 case AF_INET: 168 etype = htons(ETHERTYPE_IP); 169 break; 170 case AF_INET6: 171 etype = htons(ETHERTYPE_IPV6); 172 break; 173 case AF_ARP: 174 ah = (struct arphdr *)req->hdata; 175 ah->ar_hrd = htons(ARPHRD_ETHER); 176 177 switch(ntohs(ah->ar_op)) { 178 case ARPOP_REVREQUEST: 179 case ARPOP_REVREPLY: 180 etype = htons(ETHERTYPE_REVARP); 181 break; 182 case ARPOP_REQUEST: 183 case ARPOP_REPLY: 184 default: 185 etype = htons(ETHERTYPE_ARP); 186 break; 187 } 188 189 if (req->flags & IFENCAP_FLAG_BROADCAST) 190 lladdr = ifp->if_broadcastaddr; 191 break; 192 default: 193 return (EAFNOSUPPORT); 194 } 195 196 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 197 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN); 198 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 199 req->bufsize = sizeof(struct ether_header); 200 201 return (0); 202 } 203 204 static int 205 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m, 206 const struct sockaddr *dst, struct route *ro, u_char *phdr, 207 uint32_t *pflags, struct llentry **plle) 208 { 209 uint32_t lleflags = 0; 210 int error = 0; 211 #if defined(INET) || defined(INET6) 212 struct ether_header *eh = (struct ether_header *)phdr; 213 uint16_t etype; 214 #endif 215 216 if (plle) 217 *plle = NULL; 218 219 switch (dst->sa_family) { 220 #ifdef INET 221 case AF_INET: 222 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0) 223 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags, 224 plle); 225 else { 226 if (m->m_flags & M_BCAST) 227 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, 228 ETHER_ADDR_LEN); 229 else { 230 const struct in_addr *a; 231 a = &(((const struct sockaddr_in *)dst)->sin_addr); 232 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost); 233 } 234 etype = htons(ETHERTYPE_IP); 235 memcpy(&eh->ether_type, &etype, sizeof(etype)); 236 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 237 } 238 break; 239 #endif 240 #ifdef INET6 241 case AF_INET6: 242 if ((m->m_flags & M_MCAST) == 0) { 243 int af = RO_GET_FAMILY(ro, dst); 244 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr, 245 &lleflags, plle); 246 } else { 247 const struct in6_addr *a6; 248 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr); 249 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost); 250 etype = htons(ETHERTYPE_IPV6); 251 memcpy(&eh->ether_type, &etype, sizeof(etype)); 252 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 253 } 254 break; 255 #endif 256 default: 257 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 258 if (m != NULL) 259 m_freem(m); 260 return (EAFNOSUPPORT); 261 } 262 263 if (error == EHOSTDOWN) { 264 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0) 265 error = EHOSTUNREACH; 266 } 267 268 if (error != 0) 269 return (error); 270 271 *pflags = RT_MAY_LOOP; 272 if (lleflags & LLE_IFADDR) 273 *pflags |= RT_L2_ME; 274 275 return (0); 276 } 277 278 /* 279 * Ethernet output routine. 280 * Encapsulate a packet of type family for the local net. 281 * Use trailer local net encapsulation if enough data in first 282 * packet leaves a multiple of 512 bytes of data in remainder. 283 */ 284 int 285 ether_output(struct ifnet *ifp, struct mbuf *m, 286 const struct sockaddr *dst, struct route *ro) 287 { 288 int error = 0; 289 char linkhdr[ETHER_HDR_LEN], *phdr; 290 struct ether_header *eh; 291 struct pf_mtag *t; 292 bool loop_copy; 293 int hlen; /* link layer header length */ 294 uint32_t pflags; 295 struct llentry *lle = NULL; 296 int addref = 0; 297 298 phdr = NULL; 299 pflags = 0; 300 if (ro != NULL) { 301 /* XXX BPF uses ro_prepend */ 302 if (ro->ro_prepend != NULL) { 303 phdr = ro->ro_prepend; 304 hlen = ro->ro_plen; 305 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) { 306 if ((ro->ro_flags & RT_LLE_CACHE) != 0) { 307 lle = ro->ro_lle; 308 if (lle != NULL && 309 (lle->la_flags & LLE_VALID) == 0) { 310 LLE_FREE(lle); 311 lle = NULL; /* redundant */ 312 ro->ro_lle = NULL; 313 } 314 if (lle == NULL) { 315 /* if we lookup, keep cache */ 316 addref = 1; 317 } else 318 /* 319 * Notify LLE code that 320 * the entry was used 321 * by datapath. 322 */ 323 llentry_provide_feedback(lle); 324 } 325 if (lle != NULL) { 326 phdr = lle->r_linkdata; 327 hlen = lle->r_hdrlen; 328 pflags = lle->r_flags; 329 } 330 } 331 } 332 333 #ifdef MAC 334 error = mac_ifnet_check_transmit(ifp, m); 335 if (error) 336 senderr(error); 337 #endif 338 339 M_PROFILE(m); 340 if (ifp->if_flags & IFF_MONITOR) 341 senderr(ENETDOWN); 342 if (!((ifp->if_flags & IFF_UP) && 343 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 344 senderr(ENETDOWN); 345 346 if (phdr == NULL) { 347 /* No prepend data supplied. Try to calculate ourselves. */ 348 phdr = linkhdr; 349 hlen = ETHER_HDR_LEN; 350 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags, 351 addref ? &lle : NULL); 352 if (addref && lle != NULL) 353 ro->ro_lle = lle; 354 if (error != 0) 355 return (error == EWOULDBLOCK ? 0 : error); 356 } 357 358 if ((pflags & RT_L2_ME) != 0) { 359 update_mbuf_csumflags(m, m); 360 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0)); 361 } 362 loop_copy = (pflags & RT_MAY_LOOP) != 0; 363 364 /* 365 * Add local net header. If no space in first mbuf, 366 * allocate another. 367 * 368 * Note that we do prepend regardless of RT_HAS_HEADER flag. 369 * This is done because BPF code shifts m_data pointer 370 * to the end of ethernet header prior to calling if_output(). 371 */ 372 M_PREPEND(m, hlen, M_NOWAIT); 373 if (m == NULL) 374 senderr(ENOBUFS); 375 if ((pflags & RT_HAS_HEADER) == 0) { 376 eh = mtod(m, struct ether_header *); 377 memcpy(eh, phdr, hlen); 378 } 379 380 /* 381 * If a simplex interface, and the packet is being sent to our 382 * Ethernet address or a broadcast address, loopback a copy. 383 * XXX To make a simplex device behave exactly like a duplex 384 * device, we should copy in the case of sending to our own 385 * ethernet address (thus letting the original actually appear 386 * on the wire). However, we don't do that here for security 387 * reasons and compatibility with the original behavior. 388 */ 389 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) && 390 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 391 struct mbuf *n; 392 393 /* 394 * Because if_simloop() modifies the packet, we need a 395 * writable copy through m_dup() instead of a readonly 396 * one as m_copy[m] would give us. The alternative would 397 * be to modify if_simloop() to handle the readonly mbuf, 398 * but performancewise it is mostly equivalent (trading 399 * extra data copying vs. extra locking). 400 * 401 * XXX This is a local workaround. A number of less 402 * often used kernel parts suffer from the same bug. 403 * See PR kern/105943 for a proposed general solution. 404 */ 405 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 406 update_mbuf_csumflags(m, n); 407 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen); 408 } else 409 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); 410 } 411 412 /* 413 * Bridges require special output handling. 414 */ 415 if (ifp->if_bridge) { 416 BRIDGE_OUTPUT(ifp, m, error); 417 return (error); 418 } 419 420 #if defined(INET) || defined(INET6) 421 if (ifp->if_carp && 422 (error = (*carp_output_p)(ifp, m, dst))) 423 goto bad; 424 #endif 425 426 /* Handle ng_ether(4) processing, if any */ 427 if (ifp->if_l2com != NULL) { 428 KASSERT(ng_ether_output_p != NULL, 429 ("ng_ether_output_p is NULL")); 430 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 431 bad: if (m != NULL) 432 m_freem(m); 433 return (error); 434 } 435 if (m == NULL) 436 return (0); 437 } 438 439 /* Continue with link-layer output */ 440 return ether_output_frame(ifp, m); 441 } 442 443 static bool 444 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp) 445 { 446 struct ether_8021q_tag qtag; 447 struct ether_header *eh; 448 449 eh = mtod(*mp, struct ether_header *); 450 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN || 451 ntohs(eh->ether_type) == ETHERTYPE_QINQ) 452 return (true); 453 454 qtag.vid = 0; 455 qtag.pcp = pcp; 456 qtag.proto = ETHERTYPE_VLAN; 457 if (ether_8021q_frame(mp, ifp, ifp, &qtag)) 458 return (true); 459 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 460 return (false); 461 } 462 463 /* 464 * Ethernet link layer output routine to send a raw frame to the device. 465 * 466 * This assumes that the 14 byte Ethernet header is present and contiguous 467 * in the first mbuf (if BRIDGE'ing). 468 */ 469 int 470 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 471 { 472 uint8_t pcp; 473 474 pcp = ifp->if_pcp; 475 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN && 476 !ether_set_pcp(&m, ifp, pcp)) 477 return (0); 478 479 if (PFIL_HOOKED_OUT(V_link_pfil_head)) 480 switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) { 481 case PFIL_DROPPED: 482 return (EACCES); 483 case PFIL_CONSUMED: 484 return (0); 485 } 486 487 #ifdef EXPERIMENTAL 488 #if defined(INET6) && defined(INET) 489 /* draft-ietf-6man-ipv6only-flag */ 490 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */ 491 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) { 492 struct ether_header *eh; 493 494 eh = mtod(m, struct ether_header *); 495 switch (ntohs(eh->ether_type)) { 496 case ETHERTYPE_IP: 497 case ETHERTYPE_ARP: 498 case ETHERTYPE_REVARP: 499 m_freem(m); 500 return (EAFNOSUPPORT); 501 /* NOTREACHED */ 502 break; 503 }; 504 } 505 #endif 506 #endif 507 508 /* 509 * Queue message on interface, update output statistics if successful, 510 * and start output if interface not yet active. 511 * 512 * If KMSAN is enabled, use it to verify that the data does not contain 513 * any uninitialized bytes. 514 */ 515 kmsan_check_mbuf(m, "ether_output"); 516 return ((ifp->if_transmit)(ifp, m)); 517 } 518 519 /* 520 * Process a received Ethernet packet; the packet is in the 521 * mbuf chain m with the ethernet header at the front. 522 */ 523 static void 524 ether_input_internal(struct ifnet *ifp, struct mbuf *m) 525 { 526 struct ether_header *eh; 527 u_short etype; 528 529 if ((ifp->if_flags & IFF_UP) == 0) { 530 m_freem(m); 531 return; 532 } 533 #ifdef DIAGNOSTIC 534 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 535 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 536 m_freem(m); 537 return; 538 } 539 #endif 540 if (m->m_len < ETHER_HDR_LEN) { 541 /* XXX maybe should pullup? */ 542 if_printf(ifp, "discard frame w/o leading ethernet " 543 "header (len %u pkt len %u)\n", 544 m->m_len, m->m_pkthdr.len); 545 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 546 m_freem(m); 547 return; 548 } 549 eh = mtod(m, struct ether_header *); 550 etype = ntohs(eh->ether_type); 551 random_harvest_queue_ether(m, sizeof(*m)); 552 553 #ifdef EXPERIMENTAL 554 #if defined(INET6) && defined(INET) 555 /* draft-ietf-6man-ipv6only-flag */ 556 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */ 557 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) { 558 switch (etype) { 559 case ETHERTYPE_IP: 560 case ETHERTYPE_ARP: 561 case ETHERTYPE_REVARP: 562 m_freem(m); 563 return; 564 /* NOTREACHED */ 565 break; 566 }; 567 } 568 #endif 569 #endif 570 571 CURVNET_SET_QUIET(ifp->if_vnet); 572 573 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 574 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 575 m->m_flags |= M_BCAST; 576 else 577 m->m_flags |= M_MCAST; 578 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); 579 } 580 581 #ifdef MAC 582 /* 583 * Tag the mbuf with an appropriate MAC label before any other 584 * consumers can get to it. 585 */ 586 mac_ifnet_create_mbuf(ifp, m); 587 #endif 588 589 /* 590 * Give bpf a chance at the packet. 591 */ 592 ETHER_BPF_MTAP(ifp, m); 593 594 /* 595 * If the CRC is still on the packet, trim it off. We do this once 596 * and once only in case we are re-entered. Nothing else on the 597 * Ethernet receive path expects to see the FCS. 598 */ 599 if (m->m_flags & M_HASFCS) { 600 m_adj(m, -ETHER_CRC_LEN); 601 m->m_flags &= ~M_HASFCS; 602 } 603 604 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 605 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 606 607 /* Allow monitor mode to claim this frame, after stats are updated. */ 608 if (ifp->if_flags & IFF_MONITOR) { 609 m_freem(m); 610 CURVNET_RESTORE(); 611 return; 612 } 613 614 /* Handle input from a lagg(4) port */ 615 if (ifp->if_type == IFT_IEEE8023ADLAG) { 616 KASSERT(lagg_input_ethernet_p != NULL, 617 ("%s: if_lagg not loaded!", __func__)); 618 m = (*lagg_input_ethernet_p)(ifp, m); 619 if (m != NULL) 620 ifp = m->m_pkthdr.rcvif; 621 else { 622 CURVNET_RESTORE(); 623 return; 624 } 625 } 626 627 /* 628 * If the hardware did not process an 802.1Q tag, do this now, 629 * to allow 802.1P priority frames to be passed to the main input 630 * path correctly. 631 */ 632 if ((m->m_flags & M_VLANTAG) == 0 && 633 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) { 634 struct ether_vlan_header *evl; 635 636 if (m->m_len < sizeof(*evl) && 637 (m = m_pullup(m, sizeof(*evl))) == NULL) { 638 #ifdef DIAGNOSTIC 639 if_printf(ifp, "cannot pullup VLAN header\n"); 640 #endif 641 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 642 CURVNET_RESTORE(); 643 return; 644 } 645 646 evl = mtod(m, struct ether_vlan_header *); 647 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 648 m->m_flags |= M_VLANTAG; 649 650 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 651 ETHER_HDR_LEN - ETHER_TYPE_LEN); 652 m_adj(m, ETHER_VLAN_ENCAP_LEN); 653 eh = mtod(m, struct ether_header *); 654 } 655 656 M_SETFIB(m, ifp->if_fib); 657 658 /* Allow ng_ether(4) to claim this frame. */ 659 if (ifp->if_l2com != NULL) { 660 KASSERT(ng_ether_input_p != NULL, 661 ("%s: ng_ether_input_p is NULL", __func__)); 662 m->m_flags &= ~M_PROMISC; 663 (*ng_ether_input_p)(ifp, &m); 664 if (m == NULL) { 665 CURVNET_RESTORE(); 666 return; 667 } 668 eh = mtod(m, struct ether_header *); 669 } 670 671 /* 672 * Allow if_bridge(4) to claim this frame. 673 * 674 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 675 * and the frame should be delivered locally. 676 * 677 * If M_BRIDGE_INJECT is set, the packet was received directly by the 678 * bridge via netmap, so "ifp" is the bridge itself and the packet 679 * should be re-examined. 680 */ 681 if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) { 682 m->m_flags &= ~M_PROMISC; 683 BRIDGE_INPUT(ifp, m); 684 if (m == NULL) { 685 CURVNET_RESTORE(); 686 return; 687 } 688 eh = mtod(m, struct ether_header *); 689 } 690 691 #if defined(INET) || defined(INET6) 692 /* 693 * Clear M_PROMISC on frame so that carp(4) will see it when the 694 * mbuf flows up to Layer 3. 695 * FreeBSD's implementation of carp(4) uses the inprotosw 696 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 697 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 698 * is outside the scope of the M_PROMISC test below. 699 * TODO: Maintain a hash table of ethernet addresses other than 700 * ether_dhost which may be active on this ifp. 701 */ 702 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 703 m->m_flags &= ~M_PROMISC; 704 } else 705 #endif 706 { 707 /* 708 * If the frame received was not for our MAC address, set the 709 * M_PROMISC flag on the mbuf chain. The frame may need to 710 * be seen by the rest of the Ethernet input path in case of 711 * re-entry (e.g. bridge, vlan, netgraph) but should not be 712 * seen by upper protocol layers. 713 */ 714 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 715 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 716 m->m_flags |= M_PROMISC; 717 } 718 719 ether_demux(ifp, m); 720 CURVNET_RESTORE(); 721 } 722 723 /* 724 * Ethernet input dispatch; by default, direct dispatch here regardless of 725 * global configuration. However, if RSS is enabled, hook up RSS affinity 726 * so that when deferred or hybrid dispatch is enabled, we can redistribute 727 * load based on RSS. 728 * 729 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or 730 * not it had already done work distribution via multi-queue. Then we could 731 * direct dispatch in the event load balancing was already complete and 732 * handle the case of interfaces with different capabilities better. 733 * 734 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions 735 * at multiple layers? 736 * 737 * XXXRW: For now, enable all this only if RSS is compiled in, although it 738 * works fine without RSS. Need to characterise the performance overhead 739 * of the detour through the netisr code in the event the result is always 740 * direct dispatch. 741 */ 742 static void 743 ether_nh_input(struct mbuf *m) 744 { 745 746 M_ASSERTPKTHDR(m); 747 KASSERT(m->m_pkthdr.rcvif != NULL, 748 ("%s: NULL interface pointer", __func__)); 749 ether_input_internal(m->m_pkthdr.rcvif, m); 750 } 751 752 static struct netisr_handler ether_nh = { 753 .nh_name = "ether", 754 .nh_handler = ether_nh_input, 755 .nh_proto = NETISR_ETHER, 756 #ifdef RSS 757 .nh_policy = NETISR_POLICY_CPU, 758 .nh_dispatch = NETISR_DISPATCH_DIRECT, 759 .nh_m2cpuid = rss_m2cpuid, 760 #else 761 .nh_policy = NETISR_POLICY_SOURCE, 762 .nh_dispatch = NETISR_DISPATCH_DIRECT, 763 #endif 764 }; 765 766 static void 767 ether_init(__unused void *arg) 768 { 769 770 netisr_register(ðer_nh); 771 } 772 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 773 774 static void 775 vnet_ether_init(__unused void *arg) 776 { 777 struct pfil_head_args args; 778 779 args.pa_version = PFIL_VERSION; 780 args.pa_flags = PFIL_IN | PFIL_OUT; 781 args.pa_type = PFIL_TYPE_ETHERNET; 782 args.pa_headname = PFIL_ETHER_NAME; 783 V_link_pfil_head = pfil_head_register(&args); 784 785 #ifdef VIMAGE 786 netisr_register_vnet(ðer_nh); 787 #endif 788 } 789 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 790 vnet_ether_init, NULL); 791 792 #ifdef VIMAGE 793 static void 794 vnet_ether_pfil_destroy(__unused void *arg) 795 { 796 797 pfil_head_unregister(V_link_pfil_head); 798 } 799 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY, 800 vnet_ether_pfil_destroy, NULL); 801 802 static void 803 vnet_ether_destroy(__unused void *arg) 804 { 805 806 netisr_unregister_vnet(ðer_nh); 807 } 808 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 809 vnet_ether_destroy, NULL); 810 #endif 811 812 static void 813 ether_input(struct ifnet *ifp, struct mbuf *m) 814 { 815 struct epoch_tracker et; 816 struct mbuf *mn; 817 bool needs_epoch; 818 819 needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH); 820 #ifdef INVARIANTS 821 /* 822 * This temporary code is here to prevent epoch unaware and unmarked 823 * drivers to panic the system. Once all drivers are taken care of, 824 * the whole INVARIANTS block should go away. 825 */ 826 if (!needs_epoch && !in_epoch(net_epoch_preempt)) { 827 static bool printedonce; 828 829 needs_epoch = true; 830 if (!printedonce) { 831 printedonce = true; 832 if_printf(ifp, "called %s w/o net epoch! " 833 "PLEASE file a bug report.", __func__); 834 #ifdef KDB 835 kdb_backtrace(); 836 #endif 837 } 838 } 839 #endif 840 841 /* 842 * The drivers are allowed to pass in a chain of packets linked with 843 * m_nextpkt. We split them up into separate packets here and pass 844 * them up. This allows the drivers to amortize the receive lock. 845 */ 846 CURVNET_SET_QUIET(ifp->if_vnet); 847 if (__predict_false(needs_epoch)) 848 NET_EPOCH_ENTER(et); 849 while (m) { 850 mn = m->m_nextpkt; 851 m->m_nextpkt = NULL; 852 853 /* 854 * We will rely on rcvif being set properly in the deferred 855 * context, so assert it is correct here. 856 */ 857 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 858 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p " 859 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp)); 860 netisr_dispatch(NETISR_ETHER, m); 861 m = mn; 862 } 863 if (__predict_false(needs_epoch)) 864 NET_EPOCH_EXIT(et); 865 CURVNET_RESTORE(); 866 } 867 868 /* 869 * Upper layer processing for a received Ethernet packet. 870 */ 871 void 872 ether_demux(struct ifnet *ifp, struct mbuf *m) 873 { 874 struct ether_header *eh; 875 int i, isr; 876 u_short ether_type; 877 878 NET_EPOCH_ASSERT(); 879 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 880 881 /* Do not grab PROMISC frames in case we are re-entered. */ 882 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) { 883 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL); 884 if (i != 0 || m == NULL) 885 return; 886 } 887 888 eh = mtod(m, struct ether_header *); 889 ether_type = ntohs(eh->ether_type); 890 891 /* 892 * If this frame has a VLAN tag other than 0, call vlan_input() 893 * if its module is loaded. Otherwise, drop. 894 */ 895 if ((m->m_flags & M_VLANTAG) && 896 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 897 if (ifp->if_vlantrunk == NULL) { 898 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); 899 m_freem(m); 900 return; 901 } 902 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 903 __func__)); 904 /* Clear before possibly re-entering ether_input(). */ 905 m->m_flags &= ~M_PROMISC; 906 (*vlan_input_p)(ifp, m); 907 return; 908 } 909 910 /* 911 * Pass promiscuously received frames to the upper layer if the user 912 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 913 */ 914 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 915 m_freem(m); 916 return; 917 } 918 919 /* 920 * Reset layer specific mbuf flags to avoid confusing upper layers. 921 */ 922 m->m_flags &= ~M_VLANTAG; 923 m_clrprotoflags(m); 924 925 /* 926 * Dispatch frame to upper layer. 927 */ 928 switch (ether_type) { 929 #ifdef INET 930 case ETHERTYPE_IP: 931 isr = NETISR_IP; 932 break; 933 934 case ETHERTYPE_ARP: 935 if (ifp->if_flags & IFF_NOARP) { 936 /* Discard packet if ARP is disabled on interface */ 937 m_freem(m); 938 return; 939 } 940 isr = NETISR_ARP; 941 break; 942 #endif 943 #ifdef INET6 944 case ETHERTYPE_IPV6: 945 isr = NETISR_IPV6; 946 break; 947 #endif 948 default: 949 goto discard; 950 } 951 952 /* Strip off Ethernet header. */ 953 m_adj(m, ETHER_HDR_LEN); 954 955 netisr_dispatch(isr, m); 956 return; 957 958 discard: 959 /* 960 * Packet is to be discarded. If netgraph is present, 961 * hand the packet to it for last chance processing; 962 * otherwise dispose of it. 963 */ 964 if (ifp->if_l2com != NULL) { 965 KASSERT(ng_ether_input_orphan_p != NULL, 966 ("ng_ether_input_orphan_p is NULL")); 967 (*ng_ether_input_orphan_p)(ifp, m); 968 return; 969 } 970 m_freem(m); 971 } 972 973 /* 974 * Convert Ethernet address to printable (loggable) representation. 975 * This routine is for compatibility; it's better to just use 976 * 977 * printf("%6D", <pointer to address>, ":"); 978 * 979 * since there's no static buffer involved. 980 */ 981 char * 982 ether_sprintf(const u_char *ap) 983 { 984 static char etherbuf[18]; 985 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 986 return (etherbuf); 987 } 988 989 /* 990 * Perform common duties while attaching to interface list 991 */ 992 void 993 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 994 { 995 int i; 996 struct ifaddr *ifa; 997 struct sockaddr_dl *sdl; 998 999 ifp->if_addrlen = ETHER_ADDR_LEN; 1000 ifp->if_hdrlen = ETHER_HDR_LEN; 1001 ifp->if_mtu = ETHERMTU; 1002 if_attach(ifp); 1003 ifp->if_output = ether_output; 1004 ifp->if_input = ether_input; 1005 ifp->if_resolvemulti = ether_resolvemulti; 1006 ifp->if_requestencap = ether_requestencap; 1007 #ifdef VIMAGE 1008 ifp->if_reassign = ether_reassign; 1009 #endif 1010 if (ifp->if_baudrate == 0) 1011 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1012 ifp->if_broadcastaddr = etherbroadcastaddr; 1013 1014 ifa = ifp->if_addr; 1015 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 1016 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1017 sdl->sdl_type = IFT_ETHER; 1018 sdl->sdl_alen = ifp->if_addrlen; 1019 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 1020 1021 if (ifp->if_hw_addr != NULL) 1022 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 1023 1024 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 1025 if (ng_ether_attach_p != NULL) 1026 (*ng_ether_attach_p)(ifp); 1027 1028 /* Announce Ethernet MAC address if non-zero. */ 1029 for (i = 0; i < ifp->if_addrlen; i++) 1030 if (lla[i] != 0) 1031 break; 1032 if (i != ifp->if_addrlen) 1033 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 1034 1035 uuid_ether_add(LLADDR(sdl)); 1036 1037 /* Add necessary bits are setup; announce it now. */ 1038 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 1039 if (IS_DEFAULT_VNET(curvnet)) 1040 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 1041 } 1042 1043 /* 1044 * Perform common duties while detaching an Ethernet interface 1045 */ 1046 void 1047 ether_ifdetach(struct ifnet *ifp) 1048 { 1049 struct sockaddr_dl *sdl; 1050 1051 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 1052 uuid_ether_del(LLADDR(sdl)); 1053 1054 if (ifp->if_l2com != NULL) { 1055 KASSERT(ng_ether_detach_p != NULL, 1056 ("ng_ether_detach_p is NULL")); 1057 (*ng_ether_detach_p)(ifp); 1058 } 1059 1060 bpfdetach(ifp); 1061 if_detach(ifp); 1062 } 1063 1064 #ifdef VIMAGE 1065 void 1066 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 1067 { 1068 1069 if (ifp->if_l2com != NULL) { 1070 KASSERT(ng_ether_detach_p != NULL, 1071 ("ng_ether_detach_p is NULL")); 1072 (*ng_ether_detach_p)(ifp); 1073 } 1074 1075 if (ng_ether_attach_p != NULL) { 1076 CURVNET_SET_QUIET(new_vnet); 1077 (*ng_ether_attach_p)(ifp); 1078 CURVNET_RESTORE(); 1079 } 1080 } 1081 #endif 1082 1083 SYSCTL_DECL(_net_link); 1084 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1085 "Ethernet"); 1086 1087 #if 0 1088 /* 1089 * This is for reference. We have a table-driven version 1090 * of the little-endian crc32 generator, which is faster 1091 * than the double-loop. 1092 */ 1093 uint32_t 1094 ether_crc32_le(const uint8_t *buf, size_t len) 1095 { 1096 size_t i; 1097 uint32_t crc; 1098 int bit; 1099 uint8_t data; 1100 1101 crc = 0xffffffff; /* initial value */ 1102 1103 for (i = 0; i < len; i++) { 1104 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1105 carry = (crc ^ data) & 1; 1106 crc >>= 1; 1107 if (carry) 1108 crc = (crc ^ ETHER_CRC_POLY_LE); 1109 } 1110 } 1111 1112 return (crc); 1113 } 1114 #else 1115 uint32_t 1116 ether_crc32_le(const uint8_t *buf, size_t len) 1117 { 1118 static const uint32_t crctab[] = { 1119 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1120 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1121 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1122 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1123 }; 1124 size_t i; 1125 uint32_t crc; 1126 1127 crc = 0xffffffff; /* initial value */ 1128 1129 for (i = 0; i < len; i++) { 1130 crc ^= buf[i]; 1131 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1132 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1133 } 1134 1135 return (crc); 1136 } 1137 #endif 1138 1139 uint32_t 1140 ether_crc32_be(const uint8_t *buf, size_t len) 1141 { 1142 size_t i; 1143 uint32_t crc, carry; 1144 int bit; 1145 uint8_t data; 1146 1147 crc = 0xffffffff; /* initial value */ 1148 1149 for (i = 0; i < len; i++) { 1150 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1151 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1152 crc <<= 1; 1153 if (carry) 1154 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1155 } 1156 } 1157 1158 return (crc); 1159 } 1160 1161 int 1162 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1163 { 1164 struct ifaddr *ifa = (struct ifaddr *) data; 1165 struct ifreq *ifr = (struct ifreq *) data; 1166 int error = 0; 1167 1168 switch (command) { 1169 case SIOCSIFADDR: 1170 ifp->if_flags |= IFF_UP; 1171 1172 switch (ifa->ifa_addr->sa_family) { 1173 #ifdef INET 1174 case AF_INET: 1175 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1176 arp_ifinit(ifp, ifa); 1177 break; 1178 #endif 1179 default: 1180 ifp->if_init(ifp->if_softc); 1181 break; 1182 } 1183 break; 1184 1185 case SIOCGIFADDR: 1186 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1187 ETHER_ADDR_LEN); 1188 break; 1189 1190 case SIOCSIFMTU: 1191 /* 1192 * Set the interface MTU. 1193 */ 1194 if (ifr->ifr_mtu > ETHERMTU) { 1195 error = EINVAL; 1196 } else { 1197 ifp->if_mtu = ifr->ifr_mtu; 1198 } 1199 break; 1200 1201 case SIOCSLANPCP: 1202 error = priv_check(curthread, PRIV_NET_SETLANPCP); 1203 if (error != 0) 1204 break; 1205 if (ifr->ifr_lan_pcp > 7 && 1206 ifr->ifr_lan_pcp != IFNET_PCP_NONE) { 1207 error = EINVAL; 1208 } else { 1209 ifp->if_pcp = ifr->ifr_lan_pcp; 1210 /* broadcast event about PCP change */ 1211 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP); 1212 } 1213 break; 1214 1215 case SIOCGLANPCP: 1216 ifr->ifr_lan_pcp = ifp->if_pcp; 1217 break; 1218 1219 default: 1220 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1221 break; 1222 } 1223 return (error); 1224 } 1225 1226 static int 1227 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1228 struct sockaddr *sa) 1229 { 1230 struct sockaddr_dl *sdl; 1231 #ifdef INET 1232 struct sockaddr_in *sin; 1233 #endif 1234 #ifdef INET6 1235 struct sockaddr_in6 *sin6; 1236 #endif 1237 u_char *e_addr; 1238 1239 switch(sa->sa_family) { 1240 case AF_LINK: 1241 /* 1242 * No mapping needed. Just check that it's a valid MC address. 1243 */ 1244 sdl = (struct sockaddr_dl *)sa; 1245 e_addr = LLADDR(sdl); 1246 if (!ETHER_IS_MULTICAST(e_addr)) 1247 return EADDRNOTAVAIL; 1248 *llsa = NULL; 1249 return 0; 1250 1251 #ifdef INET 1252 case AF_INET: 1253 sin = (struct sockaddr_in *)sa; 1254 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1255 return EADDRNOTAVAIL; 1256 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1257 sdl->sdl_alen = ETHER_ADDR_LEN; 1258 e_addr = LLADDR(sdl); 1259 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1260 *llsa = (struct sockaddr *)sdl; 1261 return 0; 1262 #endif 1263 #ifdef INET6 1264 case AF_INET6: 1265 sin6 = (struct sockaddr_in6 *)sa; 1266 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1267 /* 1268 * An IP6 address of 0 means listen to all 1269 * of the Ethernet multicast address used for IP6. 1270 * (This is used for multicast routers.) 1271 */ 1272 ifp->if_flags |= IFF_ALLMULTI; 1273 *llsa = NULL; 1274 return 0; 1275 } 1276 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1277 return EADDRNOTAVAIL; 1278 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1279 sdl->sdl_alen = ETHER_ADDR_LEN; 1280 e_addr = LLADDR(sdl); 1281 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1282 *llsa = (struct sockaddr *)sdl; 1283 return 0; 1284 #endif 1285 1286 default: 1287 /* 1288 * Well, the text isn't quite right, but it's the name 1289 * that counts... 1290 */ 1291 return EAFNOSUPPORT; 1292 } 1293 } 1294 1295 static moduledata_t ether_mod = { 1296 .name = "ether", 1297 }; 1298 1299 void 1300 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1301 { 1302 struct ether_vlan_header vlan; 1303 struct mbuf mv, mb; 1304 1305 KASSERT((m->m_flags & M_VLANTAG) != 0, 1306 ("%s: vlan information not present", __func__)); 1307 KASSERT(m->m_len >= sizeof(struct ether_header), 1308 ("%s: mbuf not large enough for header", __func__)); 1309 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1310 vlan.evl_proto = vlan.evl_encap_proto; 1311 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1312 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1313 m->m_len -= sizeof(struct ether_header); 1314 m->m_data += sizeof(struct ether_header); 1315 /* 1316 * If a data link has been supplied by the caller, then we will need to 1317 * re-create a stack allocated mbuf chain with the following structure: 1318 * 1319 * (1) mbuf #1 will contain the supplied data link 1320 * (2) mbuf #2 will contain the vlan header 1321 * (3) mbuf #3 will contain the original mbuf's packet data 1322 * 1323 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1324 */ 1325 if (data != NULL) { 1326 mv.m_next = m; 1327 mv.m_data = (caddr_t)&vlan; 1328 mv.m_len = sizeof(vlan); 1329 mb.m_next = &mv; 1330 mb.m_data = data; 1331 mb.m_len = dlen; 1332 bpf_mtap(bp, &mb); 1333 } else 1334 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1335 m->m_len += sizeof(struct ether_header); 1336 m->m_data -= sizeof(struct ether_header); 1337 } 1338 1339 struct mbuf * 1340 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto) 1341 { 1342 struct ether_vlan_header *evl; 1343 1344 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1345 if (m == NULL) 1346 return (NULL); 1347 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1348 1349 if (m->m_len < sizeof(*evl)) { 1350 m = m_pullup(m, sizeof(*evl)); 1351 if (m == NULL) 1352 return (NULL); 1353 } 1354 1355 /* 1356 * Transform the Ethernet header into an Ethernet header 1357 * with 802.1Q encapsulation. 1358 */ 1359 evl = mtod(m, struct ether_vlan_header *); 1360 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1361 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1362 evl->evl_encap_proto = htons(proto); 1363 evl->evl_tag = htons(tag); 1364 return (m); 1365 } 1366 1367 void 1368 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m) 1369 { 1370 if (bpf_peers_present(ifp->if_bpf)) { 1371 M_ASSERTVALID(m); 1372 if ((m->m_flags & M_VLANTAG) != 0) 1373 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0); 1374 else 1375 bpf_mtap(ifp->if_bpf, m); 1376 } 1377 } 1378 1379 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1380 "IEEE 802.1Q VLAN"); 1381 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, 1382 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1383 "for consistency"); 1384 1385 VNET_DEFINE_STATIC(int, soft_pad); 1386 #define V_soft_pad VNET(soft_pad) 1387 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET, 1388 &VNET_NAME(soft_pad), 0, 1389 "pad short frames before tagging"); 1390 1391 /* 1392 * For now, make preserving PCP via an mbuf tag optional, as it increases 1393 * per-packet memory allocations and frees. In the future, it would be 1394 * preferable to reuse ether_vtag for this, or similar. 1395 */ 1396 VNET_DEFINE(int, vlan_mtag_pcp) = 0; 1397 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp) 1398 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET, 1399 &VNET_NAME(vlan_mtag_pcp), 0, 1400 "Retain VLAN PCP information as packets are passed up the stack"); 1401 1402 bool 1403 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p, 1404 struct ether_8021q_tag *qtag) 1405 { 1406 struct m_tag *mtag; 1407 int n; 1408 uint16_t tag; 1409 static const char pad[8]; /* just zeros */ 1410 1411 /* 1412 * Pad the frame to the minimum size allowed if told to. 1413 * This option is in accord with IEEE Std 802.1Q, 2003 Ed., 1414 * paragraph C.4.4.3.b. It can help to work around buggy 1415 * bridges that violate paragraph C.4.4.3.a from the same 1416 * document, i.e., fail to pad short frames after untagging. 1417 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but 1418 * untagging it will produce a 62-byte frame, which is a runt 1419 * and requires padding. There are VLAN-enabled network 1420 * devices that just discard such runts instead or mishandle 1421 * them somehow. 1422 */ 1423 if (V_soft_pad && p->if_type == IFT_ETHER) { 1424 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len; 1425 n > 0; n -= sizeof(pad)) { 1426 if (!m_append(*mp, min(n, sizeof(pad)), pad)) 1427 break; 1428 } 1429 if (n > 0) { 1430 m_freem(*mp); 1431 *mp = NULL; 1432 if_printf(ife, "cannot pad short frame"); 1433 return (false); 1434 } 1435 } 1436 1437 /* 1438 * If PCP is set in mbuf, use it 1439 */ 1440 if ((*mp)->m_flags & M_VLANTAG) { 1441 qtag->pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag); 1442 } 1443 1444 /* 1445 * If underlying interface can do VLAN tag insertion itself, 1446 * just pass the packet along. However, we need some way to 1447 * tell the interface where the packet came from so that it 1448 * knows how to find the VLAN tag to use, so we attach a 1449 * packet tag that holds it. 1450 */ 1451 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q, 1452 MTAG_8021Q_PCP_OUT, NULL)) != NULL) 1453 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0); 1454 else 1455 tag = EVL_MAKETAG(qtag->vid, qtag->pcp, 0); 1456 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) && 1457 (qtag->proto == ETHERTYPE_VLAN)) { 1458 (*mp)->m_pkthdr.ether_vtag = tag; 1459 (*mp)->m_flags |= M_VLANTAG; 1460 } else { 1461 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto); 1462 if (*mp == NULL) { 1463 if_printf(ife, "unable to prepend 802.1Q header"); 1464 return (false); 1465 } 1466 } 1467 return (true); 1468 } 1469 1470 /* 1471 * Allocate an address from the FreeBSD Foundation OUI. This uses a 1472 * cryptographic hash function on the containing jail's name, UUID and the 1473 * interface name to attempt to provide a unique but stable address. 1474 * Pseudo-interfaces which require a MAC address should use this function to 1475 * allocate non-locally-administered addresses. 1476 */ 1477 void 1478 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr) 1479 { 1480 SHA1_CTX ctx; 1481 char *buf; 1482 char uuid[HOSTUUIDLEN + 1]; 1483 uint64_t addr; 1484 int i, sz; 1485 char digest[SHA1_RESULTLEN]; 1486 char jailname[MAXHOSTNAMELEN]; 1487 1488 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid)); 1489 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) { 1490 /* Fall back to a random mac address. */ 1491 goto rando; 1492 } 1493 1494 /* If each (vnet) jail would also have a unique hostuuid this would not 1495 * be necessary. */ 1496 getjailname(curthread->td_ucred, jailname, sizeof(jailname)); 1497 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp), 1498 jailname); 1499 if (sz < 0) { 1500 /* Fall back to a random mac address. */ 1501 goto rando; 1502 } 1503 1504 SHA1Init(&ctx); 1505 SHA1Update(&ctx, buf, sz); 1506 SHA1Final(digest, &ctx); 1507 free(buf, M_TEMP); 1508 1509 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) & 1510 OUI_FREEBSD_GENERATED_MASK; 1511 addr = OUI_FREEBSD(addr); 1512 for (i = 0; i < ETHER_ADDR_LEN; ++i) { 1513 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) & 1514 0xFF; 1515 } 1516 1517 return; 1518 rando: 1519 arc4rand(hwaddr, sizeof(*hwaddr), 0); 1520 /* Unicast */ 1521 hwaddr->octet[0] &= 0xFE; 1522 /* Locally administered. */ 1523 hwaddr->octet[0] |= 0x02; 1524 } 1525 1526 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1527 MODULE_VERSION(ether, 1); 1528