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