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 */ 33 34 #include "opt_inet.h" 35 #include "opt_inet6.h" 36 #include "opt_netgraph.h" 37 #include "opt_mbuf_profiling.h" 38 #include "opt_rss.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/devctl.h> 43 #include <sys/eventhandler.h> 44 #include <sys/jail.h> 45 #include <sys/kernel.h> 46 #include <sys/lock.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/module.h> 50 #include <sys/msan.h> 51 #include <sys/proc.h> 52 #include <sys/priv.h> 53 #include <sys/random.h> 54 #include <sys/socket.h> 55 #include <sys/sockio.h> 56 #include <sys/sysctl.h> 57 #include <sys/uuid.h> 58 #ifdef KDB 59 #include <sys/kdb.h> 60 #endif 61 62 #include <net/ieee_oui.h> 63 #include <net/if.h> 64 #include <net/if_var.h> 65 #include <net/if_private.h> 66 #include <net/if_arp.h> 67 #include <net/netisr.h> 68 #include <net/route.h> 69 #include <net/if_llc.h> 70 #include <net/if_dl.h> 71 #include <net/if_types.h> 72 #include <net/bpf.h> 73 #include <net/ethernet.h> 74 #include <net/if_bridgevar.h> 75 #include <net/if_vlan_var.h> 76 #include <net/if_llatbl.h> 77 #include <net/pfil.h> 78 #include <net/rss_config.h> 79 #include <net/vnet.h> 80 81 #include <netpfil/pf/pf_mtag.h> 82 83 #if defined(INET) || defined(INET6) 84 #include <netinet/in.h> 85 #include <netinet/in_var.h> 86 #include <netinet/if_ether.h> 87 #include <netinet/ip_carp.h> 88 #include <netinet/ip_var.h> 89 #endif 90 #ifdef INET6 91 #include <netinet6/nd6.h> 92 #endif 93 #include <security/mac/mac_framework.h> 94 95 #include <crypto/sha1.h> 96 97 #ifdef CTASSERT 98 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 99 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 100 #endif 101 102 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */ 103 104 /* netgraph node hooks for ng_ether(4) */ 105 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 106 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 107 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 108 void (*ng_ether_attach_p)(struct ifnet *ifp); 109 void (*ng_ether_detach_p)(struct ifnet *ifp); 110 111 void (*vlan_input_p)(struct ifnet *, struct mbuf *); 112 113 /* if_bridge(4) support */ 114 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 115 116 /* if_lagg(4) support */ 117 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *); 118 119 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 120 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 121 122 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 123 struct sockaddr *); 124 static int ether_requestencap(struct ifnet *, struct if_encap_req *); 125 126 #define senderr(e) do { error = (e); goto bad;} while (0) 127 128 static void 129 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 130 { 131 int csum_flags = 0; 132 133 if (src->m_pkthdr.csum_flags & CSUM_IP) 134 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 135 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 136 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 137 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 138 csum_flags |= CSUM_SCTP_VALID; 139 dst->m_pkthdr.csum_flags |= csum_flags; 140 if (csum_flags & CSUM_DATA_VALID) 141 dst->m_pkthdr.csum_data = 0xffff; 142 } 143 144 /* 145 * Handle link-layer encapsulation requests. 146 */ 147 static int 148 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req) 149 { 150 struct ether_header *eh; 151 struct arphdr *ah; 152 uint16_t etype; 153 const u_char *lladdr; 154 155 if (req->rtype != IFENCAP_LL) 156 return (EOPNOTSUPP); 157 158 if (req->bufsize < ETHER_HDR_LEN) 159 return (ENOMEM); 160 161 eh = (struct ether_header *)req->buf; 162 lladdr = req->lladdr; 163 req->lladdr_off = 0; 164 165 switch (req->family) { 166 case AF_INET: 167 etype = htons(ETHERTYPE_IP); 168 break; 169 case AF_INET6: 170 etype = htons(ETHERTYPE_IPV6); 171 break; 172 case AF_ARP: 173 ah = (struct arphdr *)req->hdata; 174 ah->ar_hrd = htons(ARPHRD_ETHER); 175 176 switch(ntohs(ah->ar_op)) { 177 case ARPOP_REVREQUEST: 178 case ARPOP_REVREPLY: 179 etype = htons(ETHERTYPE_REVARP); 180 break; 181 case ARPOP_REQUEST: 182 case ARPOP_REPLY: 183 default: 184 etype = htons(ETHERTYPE_ARP); 185 break; 186 } 187 188 if (req->flags & IFENCAP_FLAG_BROADCAST) 189 lladdr = ifp->if_broadcastaddr; 190 break; 191 default: 192 return (EAFNOSUPPORT); 193 } 194 195 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 196 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN); 197 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 198 req->bufsize = sizeof(struct ether_header); 199 200 return (0); 201 } 202 203 static int 204 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m, 205 const struct sockaddr *dst, struct route *ro, u_char *phdr, 206 uint32_t *pflags, struct llentry **plle) 207 { 208 uint32_t lleflags = 0; 209 int error = 0; 210 #if defined(INET) || defined(INET6) 211 struct ether_header *eh = (struct ether_header *)phdr; 212 uint16_t etype; 213 #endif 214 215 if (plle) 216 *plle = NULL; 217 218 switch (dst->sa_family) { 219 #ifdef INET 220 case AF_INET: 221 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0) 222 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags, 223 plle); 224 else { 225 if (m->m_flags & M_BCAST) 226 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, 227 ETHER_ADDR_LEN); 228 else { 229 const struct in_addr *a; 230 a = &(((const struct sockaddr_in *)dst)->sin_addr); 231 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost); 232 } 233 etype = htons(ETHERTYPE_IP); 234 memcpy(&eh->ether_type, &etype, sizeof(etype)); 235 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 236 } 237 break; 238 #endif 239 #ifdef INET6 240 case AF_INET6: 241 if ((m->m_flags & M_MCAST) == 0) { 242 int af = RO_GET_FAMILY(ro, dst); 243 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr, 244 &lleflags, plle); 245 } else { 246 const struct in6_addr *a6; 247 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr); 248 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost); 249 etype = htons(ETHERTYPE_IPV6); 250 memcpy(&eh->ether_type, &etype, sizeof(etype)); 251 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 252 } 253 break; 254 #endif 255 default: 256 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 257 if (m != NULL) 258 m_freem(m); 259 return (EAFNOSUPPORT); 260 } 261 262 if (error == EHOSTDOWN) { 263 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0) 264 error = EHOSTUNREACH; 265 } 266 267 if (error != 0) 268 return (error); 269 270 *pflags = RT_MAY_LOOP; 271 if (lleflags & LLE_IFADDR) 272 *pflags |= RT_L2_ME; 273 274 return (0); 275 } 276 277 /* 278 * Ethernet output routine. 279 * Encapsulate a packet of type family for the local net. 280 * Use trailer local net encapsulation if enough data in first 281 * packet leaves a multiple of 512 bytes of data in remainder. 282 */ 283 int 284 ether_output(struct ifnet *ifp, struct mbuf *m, 285 const struct sockaddr *dst, struct route *ro) 286 { 287 int error = 0; 288 char linkhdr[ETHER_HDR_LEN], *phdr; 289 struct ether_header *eh; 290 struct pf_mtag *t; 291 bool loop_copy; 292 int hlen; /* link layer header length */ 293 uint32_t pflags; 294 struct llentry *lle = NULL; 295 int addref = 0; 296 297 phdr = NULL; 298 pflags = 0; 299 if (ro != NULL) { 300 /* XXX BPF uses ro_prepend */ 301 if (ro->ro_prepend != NULL) { 302 phdr = ro->ro_prepend; 303 hlen = ro->ro_plen; 304 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) { 305 if ((ro->ro_flags & RT_LLE_CACHE) != 0) { 306 lle = ro->ro_lle; 307 if (lle != NULL && 308 (lle->la_flags & LLE_VALID) == 0) { 309 LLE_FREE(lle); 310 lle = NULL; /* redundant */ 311 ro->ro_lle = NULL; 312 } 313 if (lle == NULL) { 314 /* if we lookup, keep cache */ 315 addref = 1; 316 } else 317 /* 318 * Notify LLE code that 319 * the entry was used 320 * by datapath. 321 */ 322 llentry_provide_feedback(lle); 323 } 324 if (lle != NULL) { 325 phdr = lle->r_linkdata; 326 hlen = lle->r_hdrlen; 327 pflags = lle->r_flags; 328 } 329 } 330 } 331 332 #ifdef MAC 333 error = mac_ifnet_check_transmit(ifp, m); 334 if (error) 335 senderr(error); 336 #endif 337 338 M_PROFILE(m); 339 if (ifp->if_flags & IFF_MONITOR) 340 senderr(ENETDOWN); 341 if (!((ifp->if_flags & IFF_UP) && 342 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 343 senderr(ENETDOWN); 344 345 if (phdr == NULL) { 346 /* No prepend data supplied. Try to calculate ourselves. */ 347 phdr = linkhdr; 348 hlen = ETHER_HDR_LEN; 349 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags, 350 addref ? &lle : NULL); 351 if (addref && lle != NULL) 352 ro->ro_lle = lle; 353 if (error != 0) 354 return (error == EWOULDBLOCK ? 0 : error); 355 } 356 357 if ((pflags & RT_L2_ME) != 0) { 358 update_mbuf_csumflags(m, m); 359 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0)); 360 } 361 loop_copy = (pflags & RT_MAY_LOOP) != 0; 362 363 /* 364 * Add local net header. If no space in first mbuf, 365 * allocate another. 366 * 367 * Note that we do prepend regardless of RT_HAS_HEADER flag. 368 * This is done because BPF code shifts m_data pointer 369 * to the end of ethernet header prior to calling if_output(). 370 */ 371 M_PREPEND(m, hlen, M_NOWAIT); 372 if (m == NULL) 373 senderr(ENOBUFS); 374 if ((pflags & RT_HAS_HEADER) == 0) { 375 eh = mtod(m, struct ether_header *); 376 memcpy(eh, phdr, hlen); 377 } 378 379 /* 380 * If a simplex interface, and the packet is being sent to our 381 * Ethernet address or a broadcast address, loopback a copy. 382 * XXX To make a simplex device behave exactly like a duplex 383 * device, we should copy in the case of sending to our own 384 * ethernet address (thus letting the original actually appear 385 * on the wire). However, we don't do that here for security 386 * reasons and compatibility with the original behavior. 387 */ 388 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) && 389 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 390 struct mbuf *n; 391 392 /* 393 * Because if_simloop() modifies the packet, we need a 394 * writable copy through m_dup() instead of a readonly 395 * one as m_copy[m] would give us. The alternative would 396 * be to modify if_simloop() to handle the readonly mbuf, 397 * but performancewise it is mostly equivalent (trading 398 * extra data copying vs. extra locking). 399 * 400 * XXX This is a local workaround. A number of less 401 * often used kernel parts suffer from the same bug. 402 * See PR kern/105943 for a proposed general solution. 403 */ 404 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 405 update_mbuf_csumflags(m, n); 406 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen); 407 } else 408 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); 409 } 410 411 /* 412 * Bridges require special output handling. 413 */ 414 if (ifp->if_bridge) { 415 BRIDGE_OUTPUT(ifp, m, error); 416 return (error); 417 } 418 419 #if defined(INET) || defined(INET6) 420 if (ifp->if_carp && 421 (error = (*carp_output_p)(ifp, m, dst))) 422 goto bad; 423 #endif 424 425 /* Handle ng_ether(4) processing, if any */ 426 if (ifp->if_l2com != NULL) { 427 KASSERT(ng_ether_output_p != NULL, 428 ("ng_ether_output_p is NULL")); 429 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 430 bad: if (m != NULL) 431 m_freem(m); 432 return (error); 433 } 434 if (m == NULL) 435 return (0); 436 } 437 438 /* Continue with link-layer output */ 439 return ether_output_frame(ifp, m); 440 } 441 442 static bool 443 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp) 444 { 445 struct ether_8021q_tag qtag; 446 struct ether_header *eh; 447 448 eh = mtod(*mp, struct ether_header *); 449 if (eh->ether_type == htons(ETHERTYPE_VLAN) || 450 eh->ether_type == htons(ETHERTYPE_QINQ)) { 451 (*mp)->m_flags &= ~M_VLANTAG; 452 return (true); 453 } 454 455 qtag.vid = 0; 456 qtag.pcp = pcp; 457 qtag.proto = ETHERTYPE_VLAN; 458 if (ether_8021q_frame(mp, ifp, ifp, &qtag)) 459 return (true); 460 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 461 return (false); 462 } 463 464 /* 465 * Ethernet link layer output routine to send a raw frame to the device. 466 * 467 * This assumes that the 14 byte Ethernet header is present and contiguous 468 * in the first mbuf (if BRIDGE'ing). 469 */ 470 int 471 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 472 { 473 uint8_t pcp; 474 475 pcp = ifp->if_pcp; 476 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN && 477 !ether_set_pcp(&m, ifp, pcp)) 478 return (0); 479 480 if (PFIL_HOOKED_OUT(V_link_pfil_head)) 481 switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) { 482 case PFIL_DROPPED: 483 return (EACCES); 484 case PFIL_CONSUMED: 485 return (0); 486 } 487 488 #ifdef EXPERIMENTAL 489 #if defined(INET6) && defined(INET) 490 /* draft-ietf-6man-ipv6only-flag */ 491 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */ 492 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) { 493 struct ether_header *eh; 494 495 eh = mtod(m, struct ether_header *); 496 switch (ntohs(eh->ether_type)) { 497 case ETHERTYPE_IP: 498 case ETHERTYPE_ARP: 499 case ETHERTYPE_REVARP: 500 m_freem(m); 501 return (EAFNOSUPPORT); 502 /* NOTREACHED */ 503 break; 504 }; 505 } 506 #endif 507 #endif 508 509 /* 510 * Queue message on interface, update output statistics if successful, 511 * and start output if interface not yet active. 512 * 513 * If KMSAN is enabled, use it to verify that the data does not contain 514 * any uninitialized bytes. 515 */ 516 kmsan_check_mbuf(m, "ether_output"); 517 return ((ifp->if_transmit)(ifp, m)); 518 } 519 520 /* 521 * Process a received Ethernet packet; the packet is in the 522 * mbuf chain m with the ethernet header at the front. 523 */ 524 static void 525 ether_input_internal(struct ifnet *ifp, struct mbuf *m) 526 { 527 struct ether_header *eh; 528 u_short etype; 529 530 if ((ifp->if_flags & IFF_UP) == 0) { 531 m_freem(m); 532 return; 533 } 534 #ifdef DIAGNOSTIC 535 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 536 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 537 m_freem(m); 538 return; 539 } 540 #endif 541 if (m->m_len < ETHER_HDR_LEN) { 542 /* XXX maybe should pullup? */ 543 if_printf(ifp, "discard frame w/o leading ethernet " 544 "header (len %u pkt len %u)\n", 545 m->m_len, m->m_pkthdr.len); 546 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 547 m_freem(m); 548 return; 549 } 550 eh = mtod(m, struct ether_header *); 551 etype = ntohs(eh->ether_type); 552 random_harvest_queue_ether(m, sizeof(*m)); 553 554 #ifdef EXPERIMENTAL 555 #if defined(INET6) && defined(INET) 556 /* draft-ietf-6man-ipv6only-flag */ 557 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */ 558 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) { 559 switch (etype) { 560 case ETHERTYPE_IP: 561 case ETHERTYPE_ARP: 562 case ETHERTYPE_REVARP: 563 m_freem(m); 564 return; 565 /* NOTREACHED */ 566 break; 567 }; 568 } 569 #endif 570 #endif 571 572 CURVNET_SET_QUIET(ifp->if_vnet); 573 574 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 575 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 576 m->m_flags |= M_BCAST; 577 else 578 m->m_flags |= M_MCAST; 579 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); 580 } 581 582 #ifdef MAC 583 /* 584 * Tag the mbuf with an appropriate MAC label before any other 585 * consumers can get to it. 586 */ 587 mac_ifnet_create_mbuf(ifp, m); 588 #endif 589 590 /* 591 * Give bpf a chance at the packet. 592 */ 593 ETHER_BPF_MTAP(ifp, m); 594 595 /* 596 * If the CRC is still on the packet, trim it off. We do this once 597 * and once only in case we are re-entered. Nothing else on the 598 * Ethernet receive path expects to see the FCS. 599 */ 600 if (m->m_flags & M_HASFCS) { 601 m_adj(m, -ETHER_CRC_LEN); 602 m->m_flags &= ~M_HASFCS; 603 } 604 605 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 606 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 607 608 /* Allow monitor mode to claim this frame, after stats are updated. */ 609 if (ifp->if_flags & IFF_MONITOR) { 610 m_freem(m); 611 CURVNET_RESTORE(); 612 return; 613 } 614 615 /* Handle input from a lagg(4) port */ 616 if (ifp->if_type == IFT_IEEE8023ADLAG) { 617 KASSERT(lagg_input_ethernet_p != NULL, 618 ("%s: if_lagg not loaded!", __func__)); 619 m = (*lagg_input_ethernet_p)(ifp, m); 620 if (m != NULL) 621 ifp = m->m_pkthdr.rcvif; 622 else { 623 CURVNET_RESTORE(); 624 return; 625 } 626 } 627 628 /* 629 * If the hardware did not process an 802.1Q tag, do this now, 630 * to allow 802.1P priority frames to be passed to the main input 631 * path correctly. 632 */ 633 if ((m->m_flags & M_VLANTAG) == 0 && 634 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) { 635 struct ether_vlan_header *evl; 636 637 if (m->m_len < sizeof(*evl) && 638 (m = m_pullup(m, sizeof(*evl))) == NULL) { 639 #ifdef DIAGNOSTIC 640 if_printf(ifp, "cannot pullup VLAN header\n"); 641 #endif 642 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 643 CURVNET_RESTORE(); 644 return; 645 } 646 647 evl = mtod(m, struct ether_vlan_header *); 648 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 649 m->m_flags |= M_VLANTAG; 650 651 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 652 ETHER_HDR_LEN - ETHER_TYPE_LEN); 653 m_adj(m, ETHER_VLAN_ENCAP_LEN); 654 eh = mtod(m, struct ether_header *); 655 } 656 657 M_SETFIB(m, ifp->if_fib); 658 659 /* Allow ng_ether(4) to claim this frame. */ 660 if (ifp->if_l2com != NULL) { 661 KASSERT(ng_ether_input_p != NULL, 662 ("%s: ng_ether_input_p is NULL", __func__)); 663 m->m_flags &= ~M_PROMISC; 664 (*ng_ether_input_p)(ifp, &m); 665 if (m == NULL) { 666 CURVNET_RESTORE(); 667 return; 668 } 669 eh = mtod(m, struct ether_header *); 670 } 671 672 /* 673 * Allow if_bridge(4) to claim this frame. 674 * 675 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 676 * and the frame should be delivered locally. 677 * 678 * If M_BRIDGE_INJECT is set, the packet was received directly by the 679 * bridge via netmap, so "ifp" is the bridge itself and the packet 680 * should be re-examined. 681 */ 682 if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) { 683 m->m_flags &= ~M_PROMISC; 684 BRIDGE_INPUT(ifp, m); 685 if (m == NULL) { 686 CURVNET_RESTORE(); 687 return; 688 } 689 eh = mtod(m, struct ether_header *); 690 } 691 692 #if defined(INET) || defined(INET6) 693 /* 694 * Clear M_PROMISC on frame so that carp(4) will see it when the 695 * mbuf flows up to Layer 3. 696 * FreeBSD's implementation of carp(4) uses the inprotosw 697 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 698 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 699 * is outside the scope of the M_PROMISC test below. 700 * TODO: Maintain a hash table of ethernet addresses other than 701 * ether_dhost which may be active on this ifp. 702 */ 703 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 704 m->m_flags &= ~M_PROMISC; 705 } else 706 #endif 707 { 708 /* 709 * If the frame received was not for our MAC address, set the 710 * M_PROMISC flag on the mbuf chain. The frame may need to 711 * be seen by the rest of the Ethernet input path in case of 712 * re-entry (e.g. bridge, vlan, netgraph) but should not be 713 * seen by upper protocol layers. 714 */ 715 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 716 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 717 m->m_flags |= M_PROMISC; 718 } 719 720 ether_demux(ifp, m); 721 CURVNET_RESTORE(); 722 } 723 724 /* 725 * Ethernet input dispatch; by default, direct dispatch here regardless of 726 * global configuration. However, if RSS is enabled, hook up RSS affinity 727 * so that when deferred or hybrid dispatch is enabled, we can redistribute 728 * load based on RSS. 729 * 730 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or 731 * not it had already done work distribution via multi-queue. Then we could 732 * direct dispatch in the event load balancing was already complete and 733 * handle the case of interfaces with different capabilities better. 734 * 735 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions 736 * at multiple layers? 737 * 738 * XXXRW: For now, enable all this only if RSS is compiled in, although it 739 * works fine without RSS. Need to characterise the performance overhead 740 * of the detour through the netisr code in the event the result is always 741 * direct dispatch. 742 */ 743 static void 744 ether_nh_input(struct mbuf *m) 745 { 746 747 M_ASSERTPKTHDR(m); 748 KASSERT(m->m_pkthdr.rcvif != NULL, 749 ("%s: NULL interface pointer", __func__)); 750 ether_input_internal(m->m_pkthdr.rcvif, m); 751 } 752 753 static struct netisr_handler ether_nh = { 754 .nh_name = "ether", 755 .nh_handler = ether_nh_input, 756 .nh_proto = NETISR_ETHER, 757 #ifdef RSS 758 .nh_policy = NETISR_POLICY_CPU, 759 .nh_dispatch = NETISR_DISPATCH_DIRECT, 760 .nh_m2cpuid = rss_m2cpuid, 761 #else 762 .nh_policy = NETISR_POLICY_SOURCE, 763 .nh_dispatch = NETISR_DISPATCH_DIRECT, 764 #endif 765 }; 766 767 static void 768 ether_init(__unused void *arg) 769 { 770 771 netisr_register(ðer_nh); 772 } 773 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 774 775 static void 776 vnet_ether_init(__unused void *arg) 777 { 778 struct pfil_head_args args; 779 780 args.pa_version = PFIL_VERSION; 781 args.pa_flags = PFIL_IN | PFIL_OUT; 782 args.pa_type = PFIL_TYPE_ETHERNET; 783 args.pa_headname = PFIL_ETHER_NAME; 784 V_link_pfil_head = pfil_head_register(&args); 785 786 #ifdef VIMAGE 787 netisr_register_vnet(ðer_nh); 788 #endif 789 } 790 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 791 vnet_ether_init, NULL); 792 793 #ifdef VIMAGE 794 static void 795 vnet_ether_pfil_destroy(__unused void *arg) 796 { 797 798 pfil_head_unregister(V_link_pfil_head); 799 } 800 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY, 801 vnet_ether_pfil_destroy, NULL); 802 803 static void 804 vnet_ether_destroy(__unused void *arg) 805 { 806 807 netisr_unregister_vnet(ðer_nh); 808 } 809 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 810 vnet_ether_destroy, NULL); 811 #endif 812 813 static void 814 ether_input(struct ifnet *ifp, struct mbuf *m) 815 { 816 struct epoch_tracker et; 817 struct mbuf *mn; 818 bool needs_epoch; 819 820 needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH); 821 #ifdef INVARIANTS 822 /* 823 * This temporary code is here to prevent epoch unaware and unmarked 824 * drivers to panic the system. Once all drivers are taken care of, 825 * the whole INVARIANTS block should go away. 826 */ 827 if (!needs_epoch && !in_epoch(net_epoch_preempt)) { 828 static bool printedonce; 829 830 needs_epoch = true; 831 if (!printedonce) { 832 printedonce = true; 833 if_printf(ifp, "called %s w/o net epoch! " 834 "PLEASE file a bug report.", __func__); 835 #ifdef KDB 836 kdb_backtrace(); 837 #endif 838 } 839 } 840 #endif 841 842 /* 843 * The drivers are allowed to pass in a chain of packets linked with 844 * m_nextpkt. We split them up into separate packets here and pass 845 * them up. This allows the drivers to amortize the receive lock. 846 */ 847 CURVNET_SET_QUIET(ifp->if_vnet); 848 if (__predict_false(needs_epoch)) 849 NET_EPOCH_ENTER(et); 850 while (m) { 851 mn = m->m_nextpkt; 852 m->m_nextpkt = NULL; 853 854 /* 855 * We will rely on rcvif being set properly in the deferred 856 * context, so assert it is correct here. 857 */ 858 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 859 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p " 860 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp)); 861 netisr_dispatch(NETISR_ETHER, m); 862 m = mn; 863 } 864 if (__predict_false(needs_epoch)) 865 NET_EPOCH_EXIT(et); 866 CURVNET_RESTORE(); 867 } 868 869 /* 870 * Upper layer processing for a received Ethernet packet. 871 */ 872 void 873 ether_demux(struct ifnet *ifp, struct mbuf *m) 874 { 875 struct ether_header *eh; 876 int i, isr; 877 u_short ether_type; 878 879 NET_EPOCH_ASSERT(); 880 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 881 882 /* Do not grab PROMISC frames in case we are re-entered. */ 883 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) { 884 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL); 885 if (i != 0 || m == NULL) 886 return; 887 } 888 889 eh = mtod(m, struct ether_header *); 890 ether_type = ntohs(eh->ether_type); 891 892 /* 893 * If this frame has a VLAN tag other than 0, call vlan_input() 894 * if its module is loaded. Otherwise, drop. 895 */ 896 if ((m->m_flags & M_VLANTAG) && 897 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 898 if (ifp->if_vlantrunk == NULL) { 899 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); 900 m_freem(m); 901 return; 902 } 903 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 904 __func__)); 905 /* Clear before possibly re-entering ether_input(). */ 906 m->m_flags &= ~M_PROMISC; 907 (*vlan_input_p)(ifp, m); 908 return; 909 } 910 911 /* 912 * Pass promiscuously received frames to the upper layer if the user 913 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 914 */ 915 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 916 m_freem(m); 917 return; 918 } 919 920 /* 921 * Reset layer specific mbuf flags to avoid confusing upper layers. 922 */ 923 m->m_flags &= ~M_VLANTAG; 924 m_clrprotoflags(m); 925 926 /* 927 * Dispatch frame to upper layer. 928 */ 929 switch (ether_type) { 930 #ifdef INET 931 case ETHERTYPE_IP: 932 isr = NETISR_IP; 933 break; 934 935 case ETHERTYPE_ARP: 936 if (ifp->if_flags & IFF_NOARP) { 937 /* Discard packet if ARP is disabled on interface */ 938 m_freem(m); 939 return; 940 } 941 isr = NETISR_ARP; 942 break; 943 #endif 944 #ifdef INET6 945 case ETHERTYPE_IPV6: 946 isr = NETISR_IPV6; 947 break; 948 #endif 949 default: 950 goto discard; 951 } 952 953 /* Strip off Ethernet header. */ 954 m_adj(m, ETHER_HDR_LEN); 955 956 netisr_dispatch(isr, m); 957 return; 958 959 discard: 960 /* 961 * Packet is to be discarded. If netgraph is present, 962 * hand the packet to it for last chance processing; 963 * otherwise dispose of it. 964 */ 965 if (ifp->if_l2com != NULL) { 966 KASSERT(ng_ether_input_orphan_p != NULL, 967 ("ng_ether_input_orphan_p is NULL")); 968 (*ng_ether_input_orphan_p)(ifp, m); 969 return; 970 } 971 m_freem(m); 972 } 973 974 /* 975 * Convert Ethernet address to printable (loggable) representation. 976 * This routine is for compatibility; it's better to just use 977 * 978 * printf("%6D", <pointer to address>, ":"); 979 * 980 * since there's no static buffer involved. 981 */ 982 char * 983 ether_sprintf(const u_char *ap) 984 { 985 static char etherbuf[18]; 986 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 987 return (etherbuf); 988 } 989 990 /* 991 * Perform common duties while attaching to interface list 992 */ 993 void 994 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 995 { 996 int i; 997 struct ifaddr *ifa; 998 struct sockaddr_dl *sdl; 999 1000 ifp->if_addrlen = ETHER_ADDR_LEN; 1001 ifp->if_hdrlen = ETHER_HDR_LEN; 1002 ifp->if_mtu = ETHERMTU; 1003 if_attach(ifp); 1004 ifp->if_output = ether_output; 1005 ifp->if_input = ether_input; 1006 ifp->if_resolvemulti = ether_resolvemulti; 1007 ifp->if_requestencap = ether_requestencap; 1008 #ifdef VIMAGE 1009 ifp->if_reassign = ether_reassign; 1010 #endif 1011 if (ifp->if_baudrate == 0) 1012 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1013 ifp->if_broadcastaddr = etherbroadcastaddr; 1014 1015 ifa = ifp->if_addr; 1016 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 1017 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1018 sdl->sdl_type = IFT_ETHER; 1019 sdl->sdl_alen = ifp->if_addrlen; 1020 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 1021 1022 if (ifp->if_hw_addr != NULL) 1023 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 1024 1025 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 1026 if (ng_ether_attach_p != NULL) 1027 (*ng_ether_attach_p)(ifp); 1028 1029 /* Announce Ethernet MAC address if non-zero. */ 1030 for (i = 0; i < ifp->if_addrlen; i++) 1031 if (lla[i] != 0) 1032 break; 1033 if (i != ifp->if_addrlen) 1034 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 1035 1036 uuid_ether_add(LLADDR(sdl)); 1037 1038 /* Add necessary bits are setup; announce it now. */ 1039 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 1040 if (IS_DEFAULT_VNET(curvnet)) 1041 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 1042 } 1043 1044 /* 1045 * Perform common duties while detaching an Ethernet interface 1046 */ 1047 void 1048 ether_ifdetach(struct ifnet *ifp) 1049 { 1050 struct sockaddr_dl *sdl; 1051 1052 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 1053 uuid_ether_del(LLADDR(sdl)); 1054 1055 if (ifp->if_l2com != NULL) { 1056 KASSERT(ng_ether_detach_p != NULL, 1057 ("ng_ether_detach_p is NULL")); 1058 (*ng_ether_detach_p)(ifp); 1059 } 1060 1061 bpfdetach(ifp); 1062 if_detach(ifp); 1063 } 1064 1065 #ifdef VIMAGE 1066 void 1067 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 1068 { 1069 1070 if (ifp->if_l2com != NULL) { 1071 KASSERT(ng_ether_detach_p != NULL, 1072 ("ng_ether_detach_p is NULL")); 1073 (*ng_ether_detach_p)(ifp); 1074 } 1075 1076 if (ng_ether_attach_p != NULL) { 1077 CURVNET_SET_QUIET(new_vnet); 1078 (*ng_ether_attach_p)(ifp); 1079 CURVNET_RESTORE(); 1080 } 1081 } 1082 #endif 1083 1084 SYSCTL_DECL(_net_link); 1085 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1086 "Ethernet"); 1087 1088 #if 0 1089 /* 1090 * This is for reference. We have a table-driven version 1091 * of the little-endian crc32 generator, which is faster 1092 * than the double-loop. 1093 */ 1094 uint32_t 1095 ether_crc32_le(const uint8_t *buf, size_t len) 1096 { 1097 size_t i; 1098 uint32_t crc; 1099 int bit; 1100 uint8_t data; 1101 1102 crc = 0xffffffff; /* initial value */ 1103 1104 for (i = 0; i < len; i++) { 1105 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1106 carry = (crc ^ data) & 1; 1107 crc >>= 1; 1108 if (carry) 1109 crc = (crc ^ ETHER_CRC_POLY_LE); 1110 } 1111 } 1112 1113 return (crc); 1114 } 1115 #else 1116 uint32_t 1117 ether_crc32_le(const uint8_t *buf, size_t len) 1118 { 1119 static const uint32_t crctab[] = { 1120 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1121 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1122 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1123 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1124 }; 1125 size_t i; 1126 uint32_t crc; 1127 1128 crc = 0xffffffff; /* initial value */ 1129 1130 for (i = 0; i < len; i++) { 1131 crc ^= buf[i]; 1132 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1133 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1134 } 1135 1136 return (crc); 1137 } 1138 #endif 1139 1140 uint32_t 1141 ether_crc32_be(const uint8_t *buf, size_t len) 1142 { 1143 size_t i; 1144 uint32_t crc, carry; 1145 int bit; 1146 uint8_t data; 1147 1148 crc = 0xffffffff; /* initial value */ 1149 1150 for (i = 0; i < len; i++) { 1151 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1152 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1153 crc <<= 1; 1154 if (carry) 1155 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1156 } 1157 } 1158 1159 return (crc); 1160 } 1161 1162 int 1163 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1164 { 1165 struct ifaddr *ifa = (struct ifaddr *) data; 1166 struct ifreq *ifr = (struct ifreq *) data; 1167 int error = 0; 1168 1169 switch (command) { 1170 case SIOCSIFADDR: 1171 ifp->if_flags |= IFF_UP; 1172 1173 switch (ifa->ifa_addr->sa_family) { 1174 #ifdef INET 1175 case AF_INET: 1176 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1177 arp_ifinit(ifp, ifa); 1178 break; 1179 #endif 1180 default: 1181 ifp->if_init(ifp->if_softc); 1182 break; 1183 } 1184 break; 1185 1186 case SIOCGIFADDR: 1187 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1188 ETHER_ADDR_LEN); 1189 break; 1190 1191 case SIOCSIFMTU: 1192 /* 1193 * Set the interface MTU. 1194 */ 1195 if (ifr->ifr_mtu > ETHERMTU) { 1196 error = EINVAL; 1197 } else { 1198 ifp->if_mtu = ifr->ifr_mtu; 1199 } 1200 break; 1201 1202 case SIOCSLANPCP: 1203 error = priv_check(curthread, PRIV_NET_SETLANPCP); 1204 if (error != 0) 1205 break; 1206 if (ifr->ifr_lan_pcp > 7 && 1207 ifr->ifr_lan_pcp != IFNET_PCP_NONE) { 1208 error = EINVAL; 1209 } else { 1210 ifp->if_pcp = ifr->ifr_lan_pcp; 1211 /* broadcast event about PCP change */ 1212 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP); 1213 } 1214 break; 1215 1216 case SIOCGLANPCP: 1217 ifr->ifr_lan_pcp = ifp->if_pcp; 1218 break; 1219 1220 default: 1221 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1222 break; 1223 } 1224 return (error); 1225 } 1226 1227 static int 1228 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1229 struct sockaddr *sa) 1230 { 1231 struct sockaddr_dl *sdl; 1232 #ifdef INET 1233 struct sockaddr_in *sin; 1234 #endif 1235 #ifdef INET6 1236 struct sockaddr_in6 *sin6; 1237 #endif 1238 u_char *e_addr; 1239 1240 switch(sa->sa_family) { 1241 case AF_LINK: 1242 /* 1243 * No mapping needed. Just check that it's a valid MC address. 1244 */ 1245 sdl = (struct sockaddr_dl *)sa; 1246 e_addr = LLADDR(sdl); 1247 if (!ETHER_IS_MULTICAST(e_addr)) 1248 return EADDRNOTAVAIL; 1249 *llsa = NULL; 1250 return 0; 1251 1252 #ifdef INET 1253 case AF_INET: 1254 sin = (struct sockaddr_in *)sa; 1255 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1256 return EADDRNOTAVAIL; 1257 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1258 sdl->sdl_alen = ETHER_ADDR_LEN; 1259 e_addr = LLADDR(sdl); 1260 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1261 *llsa = (struct sockaddr *)sdl; 1262 return 0; 1263 #endif 1264 #ifdef INET6 1265 case AF_INET6: 1266 sin6 = (struct sockaddr_in6 *)sa; 1267 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1268 /* 1269 * An IP6 address of 0 means listen to all 1270 * of the Ethernet multicast address used for IP6. 1271 * (This is used for multicast routers.) 1272 */ 1273 ifp->if_flags |= IFF_ALLMULTI; 1274 *llsa = NULL; 1275 return 0; 1276 } 1277 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1278 return EADDRNOTAVAIL; 1279 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1280 sdl->sdl_alen = ETHER_ADDR_LEN; 1281 e_addr = LLADDR(sdl); 1282 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1283 *llsa = (struct sockaddr *)sdl; 1284 return 0; 1285 #endif 1286 1287 default: 1288 /* 1289 * Well, the text isn't quite right, but it's the name 1290 * that counts... 1291 */ 1292 return EAFNOSUPPORT; 1293 } 1294 } 1295 1296 static moduledata_t ether_mod = { 1297 .name = "ether", 1298 }; 1299 1300 void 1301 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1302 { 1303 struct ether_vlan_header vlan; 1304 struct mbuf mv, mb; 1305 1306 KASSERT((m->m_flags & M_VLANTAG) != 0, 1307 ("%s: vlan information not present", __func__)); 1308 KASSERT(m->m_len >= sizeof(struct ether_header), 1309 ("%s: mbuf not large enough for header", __func__)); 1310 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1311 vlan.evl_proto = vlan.evl_encap_proto; 1312 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1313 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1314 m->m_len -= sizeof(struct ether_header); 1315 m->m_data += sizeof(struct ether_header); 1316 /* 1317 * If a data link has been supplied by the caller, then we will need to 1318 * re-create a stack allocated mbuf chain with the following structure: 1319 * 1320 * (1) mbuf #1 will contain the supplied data link 1321 * (2) mbuf #2 will contain the vlan header 1322 * (3) mbuf #3 will contain the original mbuf's packet data 1323 * 1324 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1325 */ 1326 if (data != NULL) { 1327 mv.m_next = m; 1328 mv.m_data = (caddr_t)&vlan; 1329 mv.m_len = sizeof(vlan); 1330 mb.m_next = &mv; 1331 mb.m_data = data; 1332 mb.m_len = dlen; 1333 bpf_mtap(bp, &mb); 1334 } else 1335 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1336 m->m_len += sizeof(struct ether_header); 1337 m->m_data -= sizeof(struct ether_header); 1338 } 1339 1340 struct mbuf * 1341 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto) 1342 { 1343 struct ether_vlan_header *evl; 1344 1345 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1346 if (m == NULL) 1347 return (NULL); 1348 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1349 1350 if (m->m_len < sizeof(*evl)) { 1351 m = m_pullup(m, sizeof(*evl)); 1352 if (m == NULL) 1353 return (NULL); 1354 } 1355 1356 /* 1357 * Transform the Ethernet header into an Ethernet header 1358 * with 802.1Q encapsulation. 1359 */ 1360 evl = mtod(m, struct ether_vlan_header *); 1361 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1362 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1363 evl->evl_encap_proto = htons(proto); 1364 evl->evl_tag = htons(tag); 1365 return (m); 1366 } 1367 1368 void 1369 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m) 1370 { 1371 if (bpf_peers_present(ifp->if_bpf)) { 1372 M_ASSERTVALID(m); 1373 if ((m->m_flags & M_VLANTAG) != 0) 1374 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0); 1375 else 1376 bpf_mtap(ifp->if_bpf, m); 1377 } 1378 } 1379 1380 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1381 "IEEE 802.1Q VLAN"); 1382 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, 1383 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1384 "for consistency"); 1385 1386 VNET_DEFINE_STATIC(int, soft_pad); 1387 #define V_soft_pad VNET(soft_pad) 1388 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET, 1389 &VNET_NAME(soft_pad), 0, 1390 "pad short frames before tagging"); 1391 1392 /* 1393 * For now, make preserving PCP via an mbuf tag optional, as it increases 1394 * per-packet memory allocations and frees. In the future, it would be 1395 * preferable to reuse ether_vtag for this, or similar. 1396 */ 1397 VNET_DEFINE(int, vlan_mtag_pcp) = 0; 1398 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp) 1399 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET, 1400 &VNET_NAME(vlan_mtag_pcp), 0, 1401 "Retain VLAN PCP information as packets are passed up the stack"); 1402 1403 bool 1404 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p, 1405 const struct ether_8021q_tag *qtag) 1406 { 1407 struct m_tag *mtag; 1408 int n; 1409 uint16_t tag; 1410 uint8_t pcp = qtag->pcp; 1411 static const char pad[8]; /* just zeros */ 1412 1413 /* 1414 * Pad the frame to the minimum size allowed if told to. 1415 * This option is in accord with IEEE Std 802.1Q, 2003 Ed., 1416 * paragraph C.4.4.3.b. It can help to work around buggy 1417 * bridges that violate paragraph C.4.4.3.a from the same 1418 * document, i.e., fail to pad short frames after untagging. 1419 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but 1420 * untagging it will produce a 62-byte frame, which is a runt 1421 * and requires padding. There are VLAN-enabled network 1422 * devices that just discard such runts instead or mishandle 1423 * them somehow. 1424 */ 1425 if (V_soft_pad && p->if_type == IFT_ETHER) { 1426 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len; 1427 n > 0; n -= sizeof(pad)) { 1428 if (!m_append(*mp, min(n, sizeof(pad)), pad)) 1429 break; 1430 } 1431 if (n > 0) { 1432 m_freem(*mp); 1433 *mp = NULL; 1434 if_printf(ife, "cannot pad short frame"); 1435 return (false); 1436 } 1437 } 1438 1439 /* 1440 * If PCP is set in mbuf, use it 1441 */ 1442 if ((*mp)->m_flags & M_VLANTAG) { 1443 pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag); 1444 } 1445 1446 /* 1447 * If underlying interface can do VLAN tag insertion itself, 1448 * just pass the packet along. However, we need some way to 1449 * tell the interface where the packet came from so that it 1450 * knows how to find the VLAN tag to use, so we attach a 1451 * packet tag that holds it. 1452 */ 1453 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q, 1454 MTAG_8021Q_PCP_OUT, NULL)) != NULL) 1455 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0); 1456 else 1457 tag = EVL_MAKETAG(qtag->vid, pcp, 0); 1458 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) && 1459 (qtag->proto == ETHERTYPE_VLAN)) { 1460 (*mp)->m_pkthdr.ether_vtag = tag; 1461 (*mp)->m_flags |= M_VLANTAG; 1462 } else { 1463 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto); 1464 if (*mp == NULL) { 1465 if_printf(ife, "unable to prepend 802.1Q header"); 1466 return (false); 1467 } 1468 (*mp)->m_flags &= ~M_VLANTAG; 1469 } 1470 return (true); 1471 } 1472 1473 /* 1474 * Allocate an address from the FreeBSD Foundation OUI. This uses a 1475 * cryptographic hash function on the containing jail's name, UUID and the 1476 * interface name to attempt to provide a unique but stable address. 1477 * Pseudo-interfaces which require a MAC address should use this function to 1478 * allocate non-locally-administered addresses. 1479 */ 1480 void 1481 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr) 1482 { 1483 SHA1_CTX ctx; 1484 char *buf; 1485 char uuid[HOSTUUIDLEN + 1]; 1486 uint64_t addr; 1487 int i, sz; 1488 char digest[SHA1_RESULTLEN]; 1489 char jailname[MAXHOSTNAMELEN]; 1490 1491 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid)); 1492 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) { 1493 /* Fall back to a random mac address. */ 1494 goto rando; 1495 } 1496 1497 /* If each (vnet) jail would also have a unique hostuuid this would not 1498 * be necessary. */ 1499 getjailname(curthread->td_ucred, jailname, sizeof(jailname)); 1500 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp), 1501 jailname); 1502 if (sz < 0) { 1503 /* Fall back to a random mac address. */ 1504 goto rando; 1505 } 1506 1507 SHA1Init(&ctx); 1508 SHA1Update(&ctx, buf, sz); 1509 SHA1Final(digest, &ctx); 1510 free(buf, M_TEMP); 1511 1512 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) & 1513 OUI_FREEBSD_GENERATED_MASK; 1514 addr = OUI_FREEBSD(addr); 1515 for (i = 0; i < ETHER_ADDR_LEN; ++i) { 1516 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) & 1517 0xFF; 1518 } 1519 1520 return; 1521 rando: 1522 arc4rand(hwaddr, sizeof(*hwaddr), 0); 1523 /* Unicast */ 1524 hwaddr->octet[0] &= 0xFE; 1525 /* Locally administered. */ 1526 hwaddr->octet[0] |= 0x02; 1527 } 1528 1529 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1530 MODULE_VERSION(ether, 1); 1531