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 32 #include "opt_inet.h" 33 #include "opt_inet6.h" 34 #include "opt_netgraph.h" 35 #include "opt_mbuf_profiling.h" 36 #include "opt_rss.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/devctl.h> 41 #include <sys/eventhandler.h> 42 #include <sys/jail.h> 43 #include <sys/kernel.h> 44 #include <sys/lock.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/module.h> 48 #include <sys/msan.h> 49 #include <sys/proc.h> 50 #include <sys/priv.h> 51 #include <sys/random.h> 52 #include <sys/socket.h> 53 #include <sys/sockio.h> 54 #include <sys/sysctl.h> 55 #include <sys/uuid.h> 56 #ifdef KDB 57 #include <sys/kdb.h> 58 #endif 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 static inline bool ether_do_pcp(struct ifnet *, struct mbuf *); 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 if (ether_do_pcp(ifp, m) && !ether_set_pcp(&m, ifp, ifp->if_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_NEEDSEPOCH); 817 #ifdef INVARIANTS 818 /* 819 * This temporary code is here to prevent epoch unaware and unmarked 820 * drivers to panic the system. Once all drivers are taken care of, 821 * the whole INVARIANTS block should go away. 822 */ 823 if (!needs_epoch && !in_epoch(net_epoch_preempt)) { 824 static bool printedonce; 825 826 needs_epoch = true; 827 if (!printedonce) { 828 printedonce = true; 829 if_printf(ifp, "called %s w/o net epoch! " 830 "PLEASE file a bug report.", __func__); 831 #ifdef KDB 832 kdb_backtrace(); 833 #endif 834 } 835 } 836 #endif 837 838 /* 839 * The drivers are allowed to pass in a chain of packets linked with 840 * m_nextpkt. We split them up into separate packets here and pass 841 * them up. This allows the drivers to amortize the receive lock. 842 */ 843 CURVNET_SET_QUIET(ifp->if_vnet); 844 if (__predict_false(needs_epoch)) 845 NET_EPOCH_ENTER(et); 846 while (m) { 847 mn = m->m_nextpkt; 848 m->m_nextpkt = NULL; 849 850 /* 851 * We will rely on rcvif being set properly in the deferred 852 * context, so assert it is correct here. 853 */ 854 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0); 855 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p " 856 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp)); 857 netisr_dispatch(NETISR_ETHER, m); 858 m = mn; 859 } 860 if (__predict_false(needs_epoch)) 861 NET_EPOCH_EXIT(et); 862 CURVNET_RESTORE(); 863 } 864 865 /* 866 * Upper layer processing for a received Ethernet packet. 867 */ 868 void 869 ether_demux(struct ifnet *ifp, struct mbuf *m) 870 { 871 struct ether_header *eh; 872 int i, isr; 873 u_short ether_type; 874 875 NET_EPOCH_ASSERT(); 876 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 877 878 /* Do not grab PROMISC frames in case we are re-entered. */ 879 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) { 880 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL); 881 if (i != PFIL_PASS) 882 return; 883 } 884 885 eh = mtod(m, struct ether_header *); 886 ether_type = ntohs(eh->ether_type); 887 888 /* 889 * If this frame has a VLAN tag other than 0, call vlan_input() 890 * if its module is loaded. Otherwise, drop. 891 */ 892 if ((m->m_flags & M_VLANTAG) && 893 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 894 if (ifp->if_vlantrunk == NULL) { 895 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); 896 m_freem(m); 897 return; 898 } 899 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 900 __func__)); 901 /* Clear before possibly re-entering ether_input(). */ 902 m->m_flags &= ~M_PROMISC; 903 (*vlan_input_p)(ifp, m); 904 return; 905 } 906 907 /* 908 * Pass promiscuously received frames to the upper layer if the user 909 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 910 */ 911 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 912 m_freem(m); 913 return; 914 } 915 916 /* 917 * Reset layer specific mbuf flags to avoid confusing upper layers. 918 */ 919 m->m_flags &= ~M_VLANTAG; 920 m_clrprotoflags(m); 921 922 /* 923 * Dispatch frame to upper layer. 924 */ 925 switch (ether_type) { 926 #ifdef INET 927 case ETHERTYPE_IP: 928 isr = NETISR_IP; 929 break; 930 931 case ETHERTYPE_ARP: 932 if (ifp->if_flags & IFF_NOARP) { 933 /* Discard packet if ARP is disabled on interface */ 934 m_freem(m); 935 return; 936 } 937 isr = NETISR_ARP; 938 break; 939 #endif 940 #ifdef INET6 941 case ETHERTYPE_IPV6: 942 isr = NETISR_IPV6; 943 break; 944 #endif 945 default: 946 goto discard; 947 } 948 949 /* Strip off Ethernet header. */ 950 m_adj(m, ETHER_HDR_LEN); 951 952 netisr_dispatch(isr, m); 953 return; 954 955 discard: 956 /* 957 * Packet is to be discarded. If netgraph is present, 958 * hand the packet to it for last chance processing; 959 * otherwise dispose of it. 960 */ 961 if (ifp->if_l2com != NULL) { 962 KASSERT(ng_ether_input_orphan_p != NULL, 963 ("ng_ether_input_orphan_p is NULL")); 964 (*ng_ether_input_orphan_p)(ifp, m); 965 return; 966 } 967 m_freem(m); 968 } 969 970 /* 971 * Convert Ethernet address to printable (loggable) representation. 972 * This routine is for compatibility; it's better to just use 973 * 974 * printf("%6D", <pointer to address>, ":"); 975 * 976 * since there's no static buffer involved. 977 */ 978 char * 979 ether_sprintf(const u_char *ap) 980 { 981 static char etherbuf[18]; 982 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 983 return (etherbuf); 984 } 985 986 /* 987 * Perform common duties while attaching to interface list 988 */ 989 void 990 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 991 { 992 int i; 993 struct ifaddr *ifa; 994 struct sockaddr_dl *sdl; 995 996 ifp->if_addrlen = ETHER_ADDR_LEN; 997 ifp->if_hdrlen = ETHER_HDR_LEN; 998 ifp->if_mtu = ETHERMTU; 999 if_attach(ifp); 1000 ifp->if_output = ether_output; 1001 ifp->if_input = ether_input; 1002 ifp->if_resolvemulti = ether_resolvemulti; 1003 ifp->if_requestencap = ether_requestencap; 1004 #ifdef VIMAGE 1005 ifp->if_reassign = ether_reassign; 1006 #endif 1007 if (ifp->if_baudrate == 0) 1008 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1009 ifp->if_broadcastaddr = etherbroadcastaddr; 1010 1011 ifa = ifp->if_addr; 1012 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 1013 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1014 sdl->sdl_type = IFT_ETHER; 1015 sdl->sdl_alen = ifp->if_addrlen; 1016 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 1017 1018 if (ifp->if_hw_addr != NULL) 1019 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 1020 1021 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 1022 if (ng_ether_attach_p != NULL) 1023 (*ng_ether_attach_p)(ifp); 1024 1025 /* Announce Ethernet MAC address if non-zero. */ 1026 for (i = 0; i < ifp->if_addrlen; i++) 1027 if (lla[i] != 0) 1028 break; 1029 if (i != ifp->if_addrlen) 1030 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 1031 1032 uuid_ether_add(LLADDR(sdl)); 1033 1034 /* Add necessary bits are setup; announce it now. */ 1035 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 1036 if (IS_DEFAULT_VNET(curvnet)) 1037 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 1038 } 1039 1040 /* 1041 * Perform common duties while detaching an Ethernet interface 1042 */ 1043 void 1044 ether_ifdetach(struct ifnet *ifp) 1045 { 1046 struct sockaddr_dl *sdl; 1047 1048 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 1049 uuid_ether_del(LLADDR(sdl)); 1050 1051 if (ifp->if_l2com != NULL) { 1052 KASSERT(ng_ether_detach_p != NULL, 1053 ("ng_ether_detach_p is NULL")); 1054 (*ng_ether_detach_p)(ifp); 1055 } 1056 1057 bpfdetach(ifp); 1058 if_detach(ifp); 1059 } 1060 1061 #ifdef VIMAGE 1062 void 1063 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 1064 { 1065 1066 if (ifp->if_l2com != NULL) { 1067 KASSERT(ng_ether_detach_p != NULL, 1068 ("ng_ether_detach_p is NULL")); 1069 (*ng_ether_detach_p)(ifp); 1070 } 1071 1072 if (ng_ether_attach_p != NULL) { 1073 CURVNET_SET_QUIET(new_vnet); 1074 (*ng_ether_attach_p)(ifp); 1075 CURVNET_RESTORE(); 1076 } 1077 } 1078 #endif 1079 1080 SYSCTL_DECL(_net_link); 1081 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1082 "Ethernet"); 1083 1084 #if 0 1085 /* 1086 * This is for reference. We have a table-driven version 1087 * of the little-endian crc32 generator, which is faster 1088 * than the double-loop. 1089 */ 1090 uint32_t 1091 ether_crc32_le(const uint8_t *buf, size_t len) 1092 { 1093 size_t i; 1094 uint32_t crc; 1095 int bit; 1096 uint8_t data; 1097 1098 crc = 0xffffffff; /* initial value */ 1099 1100 for (i = 0; i < len; i++) { 1101 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1102 carry = (crc ^ data) & 1; 1103 crc >>= 1; 1104 if (carry) 1105 crc = (crc ^ ETHER_CRC_POLY_LE); 1106 } 1107 } 1108 1109 return (crc); 1110 } 1111 #else 1112 uint32_t 1113 ether_crc32_le(const uint8_t *buf, size_t len) 1114 { 1115 static const uint32_t crctab[] = { 1116 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1117 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1118 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1119 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1120 }; 1121 size_t i; 1122 uint32_t crc; 1123 1124 crc = 0xffffffff; /* initial value */ 1125 1126 for (i = 0; i < len; i++) { 1127 crc ^= buf[i]; 1128 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1129 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1130 } 1131 1132 return (crc); 1133 } 1134 #endif 1135 1136 uint32_t 1137 ether_crc32_be(const uint8_t *buf, size_t len) 1138 { 1139 size_t i; 1140 uint32_t crc, carry; 1141 int bit; 1142 uint8_t data; 1143 1144 crc = 0xffffffff; /* initial value */ 1145 1146 for (i = 0; i < len; i++) { 1147 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1148 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1149 crc <<= 1; 1150 if (carry) 1151 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1152 } 1153 } 1154 1155 return (crc); 1156 } 1157 1158 int 1159 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1160 { 1161 struct ifaddr *ifa = (struct ifaddr *) data; 1162 struct ifreq *ifr = (struct ifreq *) data; 1163 int error = 0; 1164 1165 switch (command) { 1166 case SIOCSIFADDR: 1167 ifp->if_flags |= IFF_UP; 1168 1169 switch (ifa->ifa_addr->sa_family) { 1170 #ifdef INET 1171 case AF_INET: 1172 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1173 arp_ifinit(ifp, ifa); 1174 break; 1175 #endif 1176 default: 1177 ifp->if_init(ifp->if_softc); 1178 break; 1179 } 1180 break; 1181 1182 case SIOCGIFADDR: 1183 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1184 ETHER_ADDR_LEN); 1185 break; 1186 1187 case SIOCSIFMTU: 1188 /* 1189 * Set the interface MTU. 1190 */ 1191 if (ifr->ifr_mtu > ETHERMTU) { 1192 error = EINVAL; 1193 } else { 1194 ifp->if_mtu = ifr->ifr_mtu; 1195 } 1196 break; 1197 1198 case SIOCSLANPCP: 1199 error = priv_check(curthread, PRIV_NET_SETLANPCP); 1200 if (error != 0) 1201 break; 1202 if (ifr->ifr_lan_pcp > 7 && 1203 ifr->ifr_lan_pcp != IFNET_PCP_NONE) { 1204 error = EINVAL; 1205 } else { 1206 ifp->if_pcp = ifr->ifr_lan_pcp; 1207 /* broadcast event about PCP change */ 1208 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP); 1209 } 1210 break; 1211 1212 case SIOCGLANPCP: 1213 ifr->ifr_lan_pcp = ifp->if_pcp; 1214 break; 1215 1216 default: 1217 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1218 break; 1219 } 1220 return (error); 1221 } 1222 1223 static int 1224 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1225 struct sockaddr *sa) 1226 { 1227 struct sockaddr_dl *sdl; 1228 #ifdef INET 1229 struct sockaddr_in *sin; 1230 #endif 1231 #ifdef INET6 1232 struct sockaddr_in6 *sin6; 1233 #endif 1234 u_char *e_addr; 1235 1236 switch(sa->sa_family) { 1237 case AF_LINK: 1238 /* 1239 * No mapping needed. Just check that it's a valid MC address. 1240 */ 1241 sdl = (struct sockaddr_dl *)sa; 1242 e_addr = LLADDR(sdl); 1243 if (!ETHER_IS_MULTICAST(e_addr)) 1244 return EADDRNOTAVAIL; 1245 *llsa = NULL; 1246 return 0; 1247 1248 #ifdef INET 1249 case AF_INET: 1250 sin = (struct sockaddr_in *)sa; 1251 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1252 return EADDRNOTAVAIL; 1253 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1254 sdl->sdl_alen = ETHER_ADDR_LEN; 1255 e_addr = LLADDR(sdl); 1256 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1257 *llsa = (struct sockaddr *)sdl; 1258 return 0; 1259 #endif 1260 #ifdef INET6 1261 case AF_INET6: 1262 sin6 = (struct sockaddr_in6 *)sa; 1263 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1264 /* 1265 * An IP6 address of 0 means listen to all 1266 * of the Ethernet multicast address used for IP6. 1267 * (This is used for multicast routers.) 1268 */ 1269 ifp->if_flags |= IFF_ALLMULTI; 1270 *llsa = NULL; 1271 return 0; 1272 } 1273 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1274 return EADDRNOTAVAIL; 1275 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1276 sdl->sdl_alen = ETHER_ADDR_LEN; 1277 e_addr = LLADDR(sdl); 1278 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1279 *llsa = (struct sockaddr *)sdl; 1280 return 0; 1281 #endif 1282 1283 default: 1284 /* 1285 * Well, the text isn't quite right, but it's the name 1286 * that counts... 1287 */ 1288 return EAFNOSUPPORT; 1289 } 1290 } 1291 1292 static moduledata_t ether_mod = { 1293 .name = "ether", 1294 }; 1295 1296 void 1297 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1298 { 1299 struct ether_vlan_header vlan; 1300 struct mbuf mv, mb; 1301 1302 KASSERT((m->m_flags & M_VLANTAG) != 0, 1303 ("%s: vlan information not present", __func__)); 1304 KASSERT(m->m_len >= sizeof(struct ether_header), 1305 ("%s: mbuf not large enough for header", __func__)); 1306 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1307 vlan.evl_proto = vlan.evl_encap_proto; 1308 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1309 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1310 m->m_len -= sizeof(struct ether_header); 1311 m->m_data += sizeof(struct ether_header); 1312 /* 1313 * If a data link has been supplied by the caller, then we will need to 1314 * re-create a stack allocated mbuf chain with the following structure: 1315 * 1316 * (1) mbuf #1 will contain the supplied data link 1317 * (2) mbuf #2 will contain the vlan header 1318 * (3) mbuf #3 will contain the original mbuf's packet data 1319 * 1320 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1321 */ 1322 if (data != NULL) { 1323 mv.m_next = m; 1324 mv.m_data = (caddr_t)&vlan; 1325 mv.m_len = sizeof(vlan); 1326 mb.m_next = &mv; 1327 mb.m_data = data; 1328 mb.m_len = dlen; 1329 bpf_mtap(bp, &mb); 1330 } else 1331 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1332 m->m_len += sizeof(struct ether_header); 1333 m->m_data -= sizeof(struct ether_header); 1334 } 1335 1336 struct mbuf * 1337 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto) 1338 { 1339 struct ether_vlan_header *evl; 1340 1341 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1342 if (m == NULL) 1343 return (NULL); 1344 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1345 1346 if (m->m_len < sizeof(*evl)) { 1347 m = m_pullup(m, sizeof(*evl)); 1348 if (m == NULL) 1349 return (NULL); 1350 } 1351 1352 /* 1353 * Transform the Ethernet header into an Ethernet header 1354 * with 802.1Q encapsulation. 1355 */ 1356 evl = mtod(m, struct ether_vlan_header *); 1357 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1358 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1359 evl->evl_encap_proto = htons(proto); 1360 evl->evl_tag = htons(tag); 1361 return (m); 1362 } 1363 1364 void 1365 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m) 1366 { 1367 if (bpf_peers_present(ifp->if_bpf)) { 1368 M_ASSERTVALID(m); 1369 if ((m->m_flags & M_VLANTAG) != 0) 1370 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0); 1371 else 1372 bpf_mtap(ifp->if_bpf, m); 1373 } 1374 } 1375 1376 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1377 "IEEE 802.1Q VLAN"); 1378 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, 1379 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 1380 "for consistency"); 1381 1382 VNET_DEFINE_STATIC(int, soft_pad); 1383 #define V_soft_pad VNET(soft_pad) 1384 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET, 1385 &VNET_NAME(soft_pad), 0, 1386 "pad short frames before tagging"); 1387 1388 /* 1389 * For now, make preserving PCP via an mbuf tag optional, as it increases 1390 * per-packet memory allocations and frees. In the future, it would be 1391 * preferable to reuse ether_vtag for this, or similar. 1392 */ 1393 VNET_DEFINE(int, vlan_mtag_pcp) = 0; 1394 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp) 1395 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET, 1396 &VNET_NAME(vlan_mtag_pcp), 0, 1397 "Retain VLAN PCP information as packets are passed up the stack"); 1398 1399 static inline bool 1400 ether_do_pcp(struct ifnet *ifp, struct mbuf *m) 1401 { 1402 if (ifp->if_type == IFT_L2VLAN) 1403 return (false); 1404 if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0) 1405 return (true); 1406 if (V_vlan_mtag_pcp && 1407 m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL) 1408 return (true); 1409 return (false); 1410 } 1411 1412 bool 1413 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p, 1414 const struct ether_8021q_tag *qtag) 1415 { 1416 struct m_tag *mtag; 1417 int n; 1418 uint16_t tag; 1419 uint8_t pcp = qtag->pcp; 1420 static const char pad[8]; /* just zeros */ 1421 1422 /* 1423 * Pad the frame to the minimum size allowed if told to. 1424 * This option is in accord with IEEE Std 802.1Q, 2003 Ed., 1425 * paragraph C.4.4.3.b. It can help to work around buggy 1426 * bridges that violate paragraph C.4.4.3.a from the same 1427 * document, i.e., fail to pad short frames after untagging. 1428 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but 1429 * untagging it will produce a 62-byte frame, which is a runt 1430 * and requires padding. There are VLAN-enabled network 1431 * devices that just discard such runts instead or mishandle 1432 * them somehow. 1433 */ 1434 if (V_soft_pad && p->if_type == IFT_ETHER) { 1435 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len; 1436 n > 0; n -= sizeof(pad)) { 1437 if (!m_append(*mp, min(n, sizeof(pad)), pad)) 1438 break; 1439 } 1440 if (n > 0) { 1441 m_freem(*mp); 1442 *mp = NULL; 1443 if_printf(ife, "cannot pad short frame"); 1444 return (false); 1445 } 1446 } 1447 1448 /* 1449 * If PCP is set in mbuf, use it 1450 */ 1451 if ((*mp)->m_flags & M_VLANTAG) { 1452 pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag); 1453 } 1454 1455 /* 1456 * If underlying interface can do VLAN tag insertion itself, 1457 * just pass the packet along. However, we need some way to 1458 * tell the interface where the packet came from so that it 1459 * knows how to find the VLAN tag to use, so we attach a 1460 * packet tag that holds it. 1461 */ 1462 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q, 1463 MTAG_8021Q_PCP_OUT, NULL)) != NULL) 1464 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0); 1465 else 1466 tag = EVL_MAKETAG(qtag->vid, pcp, 0); 1467 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) && 1468 (qtag->proto == ETHERTYPE_VLAN)) { 1469 (*mp)->m_pkthdr.ether_vtag = tag; 1470 (*mp)->m_flags |= M_VLANTAG; 1471 } else { 1472 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto); 1473 if (*mp == NULL) { 1474 if_printf(ife, "unable to prepend 802.1Q header"); 1475 return (false); 1476 } 1477 (*mp)->m_flags &= ~M_VLANTAG; 1478 } 1479 return (true); 1480 } 1481 1482 /* 1483 * Allocate an address from the FreeBSD Foundation OUI. This uses a 1484 * cryptographic hash function on the containing jail's name, UUID and the 1485 * interface name to attempt to provide a unique but stable address. 1486 * Pseudo-interfaces which require a MAC address should use this function to 1487 * allocate non-locally-administered addresses. 1488 */ 1489 void 1490 ether_gen_addr_byname(const char *nameunit, struct ether_addr *hwaddr) 1491 { 1492 SHA1_CTX ctx; 1493 char *buf; 1494 char uuid[HOSTUUIDLEN + 1]; 1495 uint64_t addr; 1496 int i, sz; 1497 char digest[SHA1_RESULTLEN]; 1498 char jailname[MAXHOSTNAMELEN]; 1499 1500 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid)); 1501 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) { 1502 /* Fall back to a random mac address. */ 1503 goto rando; 1504 } 1505 1506 /* If each (vnet) jail would also have a unique hostuuid this would not 1507 * be necessary. */ 1508 getjailname(curthread->td_ucred, jailname, sizeof(jailname)); 1509 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, nameunit, 1510 jailname); 1511 if (sz < 0) { 1512 /* Fall back to a random mac address. */ 1513 goto rando; 1514 } 1515 1516 SHA1Init(&ctx); 1517 SHA1Update(&ctx, buf, sz); 1518 SHA1Final(digest, &ctx); 1519 free(buf, M_TEMP); 1520 1521 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) & 1522 OUI_FREEBSD_GENERATED_MASK; 1523 addr = OUI_FREEBSD(addr); 1524 for (i = 0; i < ETHER_ADDR_LEN; ++i) { 1525 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) & 1526 0xFF; 1527 } 1528 1529 return; 1530 rando: 1531 arc4rand(hwaddr, sizeof(*hwaddr), 0); 1532 /* Unicast */ 1533 hwaddr->octet[0] &= 0xFE; 1534 /* Locally administered. */ 1535 hwaddr->octet[0] |= 0x02; 1536 } 1537 1538 void 1539 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr) 1540 { 1541 ether_gen_addr_byname(if_name(ifp), hwaddr); 1542 } 1543 1544 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1545 MODULE_VERSION(ether, 1); 1546