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/bus.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/module.h> 50#include <sys/mbuf.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 59#include <net/ieee_oui.h> 60#include <net/if.h> 61#include <net/if_var.h> 62#include <net/if_arp.h> 63#include <net/netisr.h> 64#include <net/route.h> 65#include <net/if_llc.h> 66#include <net/if_dl.h> 67#include <net/if_types.h> 68#include <net/bpf.h> 69#include <net/ethernet.h> 70#include <net/if_bridgevar.h> 71#include <net/if_vlan_var.h> 72#include <net/if_llatbl.h> 73#include <net/pfil.h> 74#include <net/rss_config.h> 75#include <net/vnet.h> 76 77#include <netpfil/pf/pf_mtag.h> 78 79#if defined(INET) || defined(INET6) 80#include <netinet/in.h> 81#include <netinet/in_var.h> 82#include <netinet/if_ether.h> 83#include <netinet/ip_carp.h> 84#include <netinet/ip_var.h> 85#endif 86#ifdef INET6 87#include <netinet6/nd6.h> 88#endif 89#include <security/mac/mac_framework.h> 90 91#include <crypto/sha1.h> 92 93#ifdef CTASSERT 94CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 95CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 96#endif 97 98VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 99 100/* netgraph node hooks for ng_ether(4) */ 101void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 102void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 103int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 104void (*ng_ether_attach_p)(struct ifnet *ifp); 105void (*ng_ether_detach_p)(struct ifnet *ifp); 106 107void (*vlan_input_p)(struct ifnet *, struct mbuf *); 108 109/* if_bridge(4) support */ 110void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 111 112/* if_lagg(4) support */ 113struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *); 114 115static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 116 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 117 118static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 119 struct sockaddr *); 120static int ether_requestencap(struct ifnet *, struct if_encap_req *); 121 122 123#define senderr(e) do { error = (e); goto bad;} while (0) 124 125static void 126update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst) 127{ 128 int csum_flags = 0; 129 130 if (src->m_pkthdr.csum_flags & CSUM_IP) 131 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); 132 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 133 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); 134 if (src->m_pkthdr.csum_flags & CSUM_SCTP) 135 csum_flags |= CSUM_SCTP_VALID; 136 dst->m_pkthdr.csum_flags |= csum_flags; 137 if (csum_flags & CSUM_DATA_VALID) 138 dst->m_pkthdr.csum_data = 0xffff; 139} 140 141/* 142 * Handle link-layer encapsulation requests. 143 */ 144static int 145ether_requestencap(struct ifnet *ifp, struct if_encap_req *req) 146{ 147 struct ether_header *eh; 148 struct arphdr *ah; 149 uint16_t etype; 150 const u_char *lladdr; 151 152 if (req->rtype != IFENCAP_LL) 153 return (EOPNOTSUPP); 154 155 if (req->bufsize < ETHER_HDR_LEN) 156 return (ENOMEM); 157 158 eh = (struct ether_header *)req->buf; 159 lladdr = req->lladdr; 160 req->lladdr_off = 0; 161 162 switch (req->family) { 163 case AF_INET: 164 etype = htons(ETHERTYPE_IP); 165 break; 166 case AF_INET6: 167 etype = htons(ETHERTYPE_IPV6); 168 break; 169 case AF_ARP: 170 ah = (struct arphdr *)req->hdata; 171 ah->ar_hrd = htons(ARPHRD_ETHER); 172 173 switch(ntohs(ah->ar_op)) { 174 case ARPOP_REVREQUEST: 175 case ARPOP_REVREPLY: 176 etype = htons(ETHERTYPE_REVARP); 177 break; 178 case ARPOP_REQUEST: 179 case ARPOP_REPLY: 180 default: 181 etype = htons(ETHERTYPE_ARP); 182 break; 183 } 184 185 if (req->flags & IFENCAP_FLAG_BROADCAST) 186 lladdr = ifp->if_broadcastaddr; 187 break; 188 default: 189 return (EAFNOSUPPORT); 190 } 191 192 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 193 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN); 194 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 195 req->bufsize = sizeof(struct ether_header); 196 197 return (0); 198} 199 200 201static int 202ether_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 struct ether_header *eh; 207 uint32_t lleflags = 0; 208 int error = 0; 209#if defined(INET) || defined(INET6) 210 uint16_t etype; 211#endif 212 213 if (plle) 214 *plle = NULL; 215 eh = (struct ether_header *)phdr; 216 217 switch (dst->sa_family) { 218#ifdef INET 219 case AF_INET: 220 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0) 221 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags, 222 plle); 223 else { 224 if (m->m_flags & M_BCAST) 225 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, 226 ETHER_ADDR_LEN); 227 else { 228 const struct in_addr *a; 229 a = &(((const struct sockaddr_in *)dst)->sin_addr); 230 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost); 231 } 232 etype = htons(ETHERTYPE_IP); 233 memcpy(&eh->ether_type, &etype, sizeof(etype)); 234 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN); 235 } 236 break; 237#endif 238#ifdef INET6 239 case AF_INET6: 240 if ((m->m_flags & M_MCAST) == 0) 241 error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags, 242 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 */ 281int 282ether_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_mark_used(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, dst->sa_family, 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, dst->sa_family, 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) { 428bad: 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 440static bool 441ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp) 442{ 443 struct ether_header *eh; 444 445 eh = mtod(*mp, struct ether_header *); 446 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN || 447 ether_8021q_frame(mp, ifp, ifp, 0, pcp)) 448 return (true); 449 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 450 return (false); 451} 452 453/* 454 * Ethernet link layer output routine to send a raw frame to the device. 455 * 456 * This assumes that the 14 byte Ethernet header is present and contiguous 457 * in the first mbuf (if BRIDGE'ing). 458 */ 459int 460ether_output_frame(struct ifnet *ifp, struct mbuf *m) 461{ 462 int error; 463 uint8_t pcp; 464 465 pcp = ifp->if_pcp; 466 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN && 467 !ether_set_pcp(&m, ifp, pcp)) 468 return (0); 469 470 if (PFIL_HOOKED(&V_link_pfil_hook)) { 471 error = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, 472 PFIL_OUT, 0, NULL); 473 if (error != 0) 474 return (EACCES); 475 476 if (m == NULL) 477 return (0); 478 } 479 480 /* 481 * Queue message on interface, update output statistics if 482 * successful, and start output if interface not yet active. 483 */ 484 return ((ifp->if_transmit)(ifp, m)); 485} 486 487/* 488 * Process a received Ethernet packet; the packet is in the 489 * mbuf chain m with the ethernet header at the front. 490 */ 491static void 492ether_input_internal(struct ifnet *ifp, struct mbuf *m) 493{ 494 struct ether_header *eh; 495 u_short etype; 496 497 if ((ifp->if_flags & IFF_UP) == 0) { 498 m_freem(m); 499 return; 500 } 501#ifdef DIAGNOSTIC 502 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 503 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 504 m_freem(m); 505 return; 506 } 507#endif 508 if (m->m_len < ETHER_HDR_LEN) { 509 /* XXX maybe should pullup? */ 510 if_printf(ifp, "discard frame w/o leading ethernet " 511 "header (len %u pkt len %u)\n", 512 m->m_len, m->m_pkthdr.len); 513 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 514 m_freem(m); 515 return; 516 } 517 eh = mtod(m, struct ether_header *); 518 etype = ntohs(eh->ether_type); 519 random_harvest_queue_ether(m, sizeof(*m)); 520 521 CURVNET_SET_QUIET(ifp->if_vnet); 522 523 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 524 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 525 m->m_flags |= M_BCAST; 526 else 527 m->m_flags |= M_MCAST; 528 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); 529 } 530 531#ifdef MAC 532 /* 533 * Tag the mbuf with an appropriate MAC label before any other 534 * consumers can get to it. 535 */ 536 mac_ifnet_create_mbuf(ifp, m); 537#endif 538 539 /* 540 * Give bpf a chance at the packet. 541 */ 542 ETHER_BPF_MTAP(ifp, m); 543 544 /* 545 * If the CRC is still on the packet, trim it off. We do this once 546 * and once only in case we are re-entered. Nothing else on the 547 * Ethernet receive path expects to see the FCS. 548 */ 549 if (m->m_flags & M_HASFCS) { 550 m_adj(m, -ETHER_CRC_LEN); 551 m->m_flags &= ~M_HASFCS; 552 } 553 554 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 555 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 556 557 /* Allow monitor mode to claim this frame, after stats are updated. */ 558 if (ifp->if_flags & IFF_MONITOR) { 559 m_freem(m); 560 CURVNET_RESTORE(); 561 return; 562 } 563 564 /* Handle input from a lagg(4) port */ 565 if (ifp->if_type == IFT_IEEE8023ADLAG) { 566 KASSERT(lagg_input_ethernet_p != NULL, 567 ("%s: if_lagg not loaded!", __func__)); 568 m = (*lagg_input_ethernet_p)(ifp, m); 569 if (m != NULL) 570 ifp = m->m_pkthdr.rcvif; 571 else { 572 CURVNET_RESTORE(); 573 return; 574 } 575 } 576 577 /* 578 * If the hardware did not process an 802.1Q tag, do this now, 579 * to allow 802.1P priority frames to be passed to the main input 580 * path correctly. 581 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 582 */ 583 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 584 struct ether_vlan_header *evl; 585 586 if (m->m_len < sizeof(*evl) && 587 (m = m_pullup(m, sizeof(*evl))) == NULL) { 588#ifdef DIAGNOSTIC 589 if_printf(ifp, "cannot pullup VLAN header\n"); 590#endif 591 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 592 CURVNET_RESTORE(); 593 return; 594 } 595 596 evl = mtod(m, struct ether_vlan_header *); 597 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 598 m->m_flags |= M_VLANTAG; 599 600 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 601 ETHER_HDR_LEN - ETHER_TYPE_LEN); 602 m_adj(m, ETHER_VLAN_ENCAP_LEN); 603 eh = mtod(m, struct ether_header *); 604 } 605 606 M_SETFIB(m, ifp->if_fib); 607 608 /* Allow ng_ether(4) to claim this frame. */ 609 if (ifp->if_l2com != NULL) { 610 KASSERT(ng_ether_input_p != NULL, 611 ("%s: ng_ether_input_p is NULL", __func__)); 612 m->m_flags &= ~M_PROMISC; 613 (*ng_ether_input_p)(ifp, &m); 614 if (m == NULL) { 615 CURVNET_RESTORE(); 616 return; 617 } 618 eh = mtod(m, struct ether_header *); 619 } 620 621 /* 622 * Allow if_bridge(4) to claim this frame. 623 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 624 * and the frame should be delivered locally. 625 */ 626 if (ifp->if_bridge != NULL) { 627 m->m_flags &= ~M_PROMISC; 628 BRIDGE_INPUT(ifp, m); 629 if (m == NULL) { 630 CURVNET_RESTORE(); 631 return; 632 } 633 eh = mtod(m, struct ether_header *); 634 } 635 636#if defined(INET) || defined(INET6) 637 /* 638 * Clear M_PROMISC on frame so that carp(4) will see it when the 639 * mbuf flows up to Layer 3. 640 * FreeBSD's implementation of carp(4) uses the inprotosw 641 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 642 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 643 * is outside the scope of the M_PROMISC test below. 644 * TODO: Maintain a hash table of ethernet addresses other than 645 * ether_dhost which may be active on this ifp. 646 */ 647 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 648 m->m_flags &= ~M_PROMISC; 649 } else 650#endif 651 { 652 /* 653 * If the frame received was not for our MAC address, set the 654 * M_PROMISC flag on the mbuf chain. The frame may need to 655 * be seen by the rest of the Ethernet input path in case of 656 * re-entry (e.g. bridge, vlan, netgraph) but should not be 657 * seen by upper protocol layers. 658 */ 659 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 660 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 661 m->m_flags |= M_PROMISC; 662 } 663 664 ether_demux(ifp, m); 665 CURVNET_RESTORE(); 666} 667 668/* 669 * Ethernet input dispatch; by default, direct dispatch here regardless of 670 * global configuration. However, if RSS is enabled, hook up RSS affinity 671 * so that when deferred or hybrid dispatch is enabled, we can redistribute 672 * load based on RSS. 673 * 674 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or 675 * not it had already done work distribution via multi-queue. Then we could 676 * direct dispatch in the event load balancing was already complete and 677 * handle the case of interfaces with different capabilities better. 678 * 679 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions 680 * at multiple layers? 681 * 682 * XXXRW: For now, enable all this only if RSS is compiled in, although it 683 * works fine without RSS. Need to characterise the performance overhead 684 * of the detour through the netisr code in the event the result is always 685 * direct dispatch. 686 */ 687static void 688ether_nh_input(struct mbuf *m) 689{ 690 691 M_ASSERTPKTHDR(m); 692 KASSERT(m->m_pkthdr.rcvif != NULL, 693 ("%s: NULL interface pointer", __func__)); 694 ether_input_internal(m->m_pkthdr.rcvif, m); 695} 696 697static struct netisr_handler ether_nh = { 698 .nh_name = "ether", 699 .nh_handler = ether_nh_input, 700 .nh_proto = NETISR_ETHER, 701#ifdef RSS 702 .nh_policy = NETISR_POLICY_CPU, 703 .nh_dispatch = NETISR_DISPATCH_DIRECT, 704 .nh_m2cpuid = rss_m2cpuid, 705#else 706 .nh_policy = NETISR_POLICY_SOURCE, 707 .nh_dispatch = NETISR_DISPATCH_DIRECT, 708#endif 709}; 710 711static void 712ether_init(__unused void *arg) 713{ 714 715 netisr_register(ðer_nh); 716} 717SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 718 719static void 720vnet_ether_init(__unused void *arg) 721{ 722 int i; 723 724 /* Initialize packet filter hooks. */ 725 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 726 V_link_pfil_hook.ph_af = AF_LINK; 727 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 728 printf("%s: WARNING: unable to register pfil link hook, " 729 "error %d\n", __func__, i); 730#ifdef VIMAGE 731 netisr_register_vnet(ðer_nh); 732#endif 733} 734VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 735 vnet_ether_init, NULL); 736 737#ifdef VIMAGE 738static void 739vnet_ether_pfil_destroy(__unused void *arg) 740{ 741 int i; 742 743 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 744 printf("%s: WARNING: unable to unregister pfil link hook, " 745 "error %d\n", __func__, i); 746} 747VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY, 748 vnet_ether_pfil_destroy, NULL); 749 750static void 751vnet_ether_destroy(__unused void *arg) 752{ 753 754 netisr_unregister_vnet(ðer_nh); 755} 756VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 757 vnet_ether_destroy, NULL); 758#endif 759 760 761 762static void 763ether_input(struct ifnet *ifp, struct mbuf *m) 764{ 765 766 struct mbuf *mn; 767 768 /* 769 * The drivers are allowed to pass in a chain of packets linked with 770 * m_nextpkt. We split them up into separate packets here and pass 771 * them up. This allows the drivers to amortize the receive lock. 772 */ 773 while (m) { 774 mn = m->m_nextpkt; 775 m->m_nextpkt = NULL; 776 777 /* 778 * We will rely on rcvif being set properly in the deferred context, 779 * so assert it is correct here. 780 */ 781 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p " 782 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp)); 783 CURVNET_SET_QUIET(ifp->if_vnet); 784 netisr_dispatch(NETISR_ETHER, m); 785 CURVNET_RESTORE(); 786 m = mn; 787 } 788} 789 790/* 791 * Upper layer processing for a received Ethernet packet. 792 */ 793void 794ether_demux(struct ifnet *ifp, struct mbuf *m) 795{ 796 struct ether_header *eh; 797 int i, isr; 798 u_short ether_type; 799 800 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 801 802 /* Do not grab PROMISC frames in case we are re-entered. */ 803 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 804 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, 0, 805 NULL); 806 807 if (i != 0 || m == NULL) 808 return; 809 } 810 811 eh = mtod(m, struct ether_header *); 812 ether_type = ntohs(eh->ether_type); 813 814 /* 815 * If this frame has a VLAN tag other than 0, call vlan_input() 816 * if its module is loaded. Otherwise, drop. 817 */ 818 if ((m->m_flags & M_VLANTAG) && 819 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 820 if (ifp->if_vlantrunk == NULL) { 821 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); 822 m_freem(m); 823 return; 824 } 825 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 826 __func__)); 827 /* Clear before possibly re-entering ether_input(). */ 828 m->m_flags &= ~M_PROMISC; 829 (*vlan_input_p)(ifp, m); 830 return; 831 } 832 833 /* 834 * Pass promiscuously received frames to the upper layer if the user 835 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 836 */ 837 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 838 m_freem(m); 839 return; 840 } 841 842 /* 843 * Reset layer specific mbuf flags to avoid confusing upper layers. 844 * Strip off Ethernet header. 845 */ 846 m->m_flags &= ~M_VLANTAG; 847 m_clrprotoflags(m); 848 m_adj(m, ETHER_HDR_LEN); 849 850 /* 851 * Dispatch frame to upper layer. 852 */ 853 switch (ether_type) { 854#ifdef INET 855 case ETHERTYPE_IP: 856 isr = NETISR_IP; 857 break; 858 859 case ETHERTYPE_ARP: 860 if (ifp->if_flags & IFF_NOARP) { 861 /* Discard packet if ARP is disabled on interface */ 862 m_freem(m); 863 return; 864 } 865 isr = NETISR_ARP; 866 break; 867#endif 868#ifdef INET6 869 case ETHERTYPE_IPV6: 870 isr = NETISR_IPV6; 871 break; 872#endif 873 default: 874 goto discard; 875 } 876 netisr_dispatch(isr, m); 877 return; 878 879discard: 880 /* 881 * Packet is to be discarded. If netgraph is present, 882 * hand the packet to it for last chance processing; 883 * otherwise dispose of it. 884 */ 885 if (ifp->if_l2com != NULL) { 886 KASSERT(ng_ether_input_orphan_p != NULL, 887 ("ng_ether_input_orphan_p is NULL")); 888 /* 889 * Put back the ethernet header so netgraph has a 890 * consistent view of inbound packets. 891 */ 892 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 893 (*ng_ether_input_orphan_p)(ifp, m); 894 return; 895 } 896 m_freem(m); 897} 898 899/* 900 * Convert Ethernet address to printable (loggable) representation. 901 * This routine is for compatibility; it's better to just use 902 * 903 * printf("%6D", <pointer to address>, ":"); 904 * 905 * since there's no static buffer involved. 906 */ 907char * 908ether_sprintf(const u_char *ap) 909{ 910 static char etherbuf[18]; 911 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 912 return (etherbuf); 913} 914 915/* 916 * Perform common duties while attaching to interface list 917 */ 918void 919ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 920{ 921 int i; 922 struct ifaddr *ifa; 923 struct sockaddr_dl *sdl; 924 925 ifp->if_addrlen = ETHER_ADDR_LEN; 926 ifp->if_hdrlen = ETHER_HDR_LEN; 927 ifp->if_mtu = ETHERMTU; 928 if_attach(ifp); 929 ifp->if_output = ether_output; 930 ifp->if_input = ether_input; 931 ifp->if_resolvemulti = ether_resolvemulti; 932 ifp->if_requestencap = ether_requestencap; 933#ifdef VIMAGE 934 ifp->if_reassign = ether_reassign; 935#endif 936 if (ifp->if_baudrate == 0) 937 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 938 ifp->if_broadcastaddr = etherbroadcastaddr; 939 940 ifa = ifp->if_addr; 941 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 942 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 943 sdl->sdl_type = IFT_ETHER; 944 sdl->sdl_alen = ifp->if_addrlen; 945 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 946 947 if (ifp->if_hw_addr != NULL) 948 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 949 950 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 951 if (ng_ether_attach_p != NULL) 952 (*ng_ether_attach_p)(ifp); 953 954 /* Announce Ethernet MAC address if non-zero. */ 955 for (i = 0; i < ifp->if_addrlen; i++) 956 if (lla[i] != 0) 957 break; 958 if (i != ifp->if_addrlen) 959 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 960 961 uuid_ether_add(LLADDR(sdl)); 962 963 /* Add necessary bits are setup; announce it now. */ 964 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 965 if (IS_DEFAULT_VNET(curvnet)) 966 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 967} 968 969/* 970 * Perform common duties while detaching an Ethernet interface 971 */ 972void 973ether_ifdetach(struct ifnet *ifp) 974{ 975 struct sockaddr_dl *sdl; 976 977 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 978 uuid_ether_del(LLADDR(sdl)); 979 980 if (ifp->if_l2com != NULL) { 981 KASSERT(ng_ether_detach_p != NULL, 982 ("ng_ether_detach_p is NULL")); 983 (*ng_ether_detach_p)(ifp); 984 } 985 986 bpfdetach(ifp); 987 if_detach(ifp); 988} 989 990#ifdef VIMAGE 991void 992ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 993{ 994 995 if (ifp->if_l2com != NULL) { 996 KASSERT(ng_ether_detach_p != NULL, 997 ("ng_ether_detach_p is NULL")); 998 (*ng_ether_detach_p)(ifp); 999 } 1000 1001 if (ng_ether_attach_p != NULL) { 1002 CURVNET_SET_QUIET(new_vnet); 1003 (*ng_ether_attach_p)(ifp); 1004 CURVNET_RESTORE(); 1005 } 1006} 1007#endif 1008 1009SYSCTL_DECL(_net_link); 1010SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 1011 1012#if 0 1013/* 1014 * This is for reference. We have a table-driven version 1015 * of the little-endian crc32 generator, which is faster 1016 * than the double-loop. 1017 */ 1018uint32_t 1019ether_crc32_le(const uint8_t *buf, size_t len) 1020{ 1021 size_t i; 1022 uint32_t crc; 1023 int bit; 1024 uint8_t data; 1025 1026 crc = 0xffffffff; /* initial value */ 1027 1028 for (i = 0; i < len; i++) { 1029 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1030 carry = (crc ^ data) & 1; 1031 crc >>= 1; 1032 if (carry) 1033 crc = (crc ^ ETHER_CRC_POLY_LE); 1034 } 1035 } 1036 1037 return (crc); 1038} 1039#else 1040uint32_t 1041ether_crc32_le(const uint8_t *buf, size_t len) 1042{ 1043 static const uint32_t crctab[] = { 1044 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1045 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1046 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1047 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1048 }; 1049 size_t i; 1050 uint32_t crc; 1051 1052 crc = 0xffffffff; /* initial value */ 1053 1054 for (i = 0; i < len; i++) { 1055 crc ^= buf[i]; 1056 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1057 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1058 } 1059 1060 return (crc); 1061} 1062#endif 1063 1064uint32_t 1065ether_crc32_be(const uint8_t *buf, size_t len) 1066{ 1067 size_t i; 1068 uint32_t crc, carry; 1069 int bit; 1070 uint8_t data; 1071 1072 crc = 0xffffffff; /* initial value */ 1073 1074 for (i = 0; i < len; i++) { 1075 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1076 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1077 crc <<= 1; 1078 if (carry) 1079 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1080 } 1081 } 1082 1083 return (crc); 1084} 1085 1086int 1087ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1088{ 1089 struct ifaddr *ifa = (struct ifaddr *) data; 1090 struct ifreq *ifr = (struct ifreq *) data; 1091 int error = 0; 1092 1093 switch (command) { 1094 case SIOCSIFADDR: 1095 ifp->if_flags |= IFF_UP; 1096 1097 switch (ifa->ifa_addr->sa_family) { 1098#ifdef INET 1099 case AF_INET: 1100 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1101 arp_ifinit(ifp, ifa); 1102 break; 1103#endif 1104 default: 1105 ifp->if_init(ifp->if_softc); 1106 break; 1107 } 1108 break; 1109 1110 case SIOCGIFADDR: 1111 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1112 ETHER_ADDR_LEN); 1113 break; 1114 1115 case SIOCSIFMTU: 1116 /* 1117 * Set the interface MTU. 1118 */ 1119 if (ifr->ifr_mtu > ETHERMTU) { 1120 error = EINVAL; 1121 } else { 1122 ifp->if_mtu = ifr->ifr_mtu; 1123 } 1124 break; 1125 1126 case SIOCSLANPCP: 1127 error = priv_check(curthread, PRIV_NET_SETLANPCP); 1128 if (error != 0) 1129 break; 1130 if (ifr->ifr_lan_pcp > 7 && 1131 ifr->ifr_lan_pcp != IFNET_PCP_NONE) { 1132 error = EINVAL; 1133 } else { 1134 ifp->if_pcp = ifr->ifr_lan_pcp; 1135 /* broadcast event about PCP change */ 1136 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP); 1137 } 1138 break; 1139 1140 case SIOCGLANPCP: 1141 ifr->ifr_lan_pcp = ifp->if_pcp; 1142 break; 1143 1144 default: 1145 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1146 break; 1147 } 1148 return (error); 1149} 1150 1151static int 1152ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1153 struct sockaddr *sa) 1154{ 1155 struct sockaddr_dl *sdl; 1156#ifdef INET 1157 struct sockaddr_in *sin; 1158#endif 1159#ifdef INET6 1160 struct sockaddr_in6 *sin6; 1161#endif 1162 u_char *e_addr; 1163 1164 switch(sa->sa_family) { 1165 case AF_LINK: 1166 /* 1167 * No mapping needed. Just check that it's a valid MC address. 1168 */ 1169 sdl = (struct sockaddr_dl *)sa; 1170 e_addr = LLADDR(sdl); 1171 if (!ETHER_IS_MULTICAST(e_addr)) 1172 return EADDRNOTAVAIL; 1173 *llsa = NULL; 1174 return 0; 1175 1176#ifdef INET 1177 case AF_INET: 1178 sin = (struct sockaddr_in *)sa; 1179 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1180 return EADDRNOTAVAIL; 1181 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1182 sdl->sdl_alen = ETHER_ADDR_LEN; 1183 e_addr = LLADDR(sdl); 1184 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1185 *llsa = (struct sockaddr *)sdl; 1186 return 0; 1187#endif 1188#ifdef INET6 1189 case AF_INET6: 1190 sin6 = (struct sockaddr_in6 *)sa; 1191 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1192 /* 1193 * An IP6 address of 0 means listen to all 1194 * of the Ethernet multicast address used for IP6. 1195 * (This is used for multicast routers.) 1196 */ 1197 ifp->if_flags |= IFF_ALLMULTI; 1198 *llsa = NULL; 1199 return 0; 1200 } 1201 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1202 return EADDRNOTAVAIL; 1203 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1204 sdl->sdl_alen = ETHER_ADDR_LEN; 1205 e_addr = LLADDR(sdl); 1206 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1207 *llsa = (struct sockaddr *)sdl; 1208 return 0; 1209#endif 1210 1211 default: 1212 /* 1213 * Well, the text isn't quite right, but it's the name 1214 * that counts... 1215 */ 1216 return EAFNOSUPPORT; 1217 } 1218} 1219 1220static moduledata_t ether_mod = { 1221 .name = "ether", 1222}; 1223 1224void 1225ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1226{ 1227 struct ether_vlan_header vlan; 1228 struct mbuf mv, mb; 1229 1230 KASSERT((m->m_flags & M_VLANTAG) != 0, 1231 ("%s: vlan information not present", __func__)); 1232 KASSERT(m->m_len >= sizeof(struct ether_header), 1233 ("%s: mbuf not large enough for header", __func__)); 1234 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1235 vlan.evl_proto = vlan.evl_encap_proto; 1236 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1237 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1238 m->m_len -= sizeof(struct ether_header); 1239 m->m_data += sizeof(struct ether_header); 1240 /* 1241 * If a data link has been supplied by the caller, then we will need to 1242 * re-create a stack allocated mbuf chain with the following structure: 1243 * 1244 * (1) mbuf #1 will contain the supplied data link 1245 * (2) mbuf #2 will contain the vlan header 1246 * (3) mbuf #3 will contain the original mbuf's packet data 1247 * 1248 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1249 */ 1250 if (data != NULL) { 1251 mv.m_next = m; 1252 mv.m_data = (caddr_t)&vlan; 1253 mv.m_len = sizeof(vlan); 1254 mb.m_next = &mv; 1255 mb.m_data = data; 1256 mb.m_len = dlen; 1257 bpf_mtap(bp, &mb); 1258 } else 1259 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1260 m->m_len += sizeof(struct ether_header); 1261 m->m_data -= sizeof(struct ether_header); 1262} 1263 1264struct mbuf * 1265ether_vlanencap(struct mbuf *m, uint16_t tag) 1266{ 1267 struct ether_vlan_header *evl; 1268 1269 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1270 if (m == NULL) 1271 return (NULL); 1272 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1273 1274 if (m->m_len < sizeof(*evl)) { 1275 m = m_pullup(m, sizeof(*evl)); 1276 if (m == NULL) 1277 return (NULL); 1278 } 1279 1280 /* 1281 * Transform the Ethernet header into an Ethernet header 1282 * with 802.1Q encapsulation. 1283 */ 1284 evl = mtod(m, struct ether_vlan_header *); 1285 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1286 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1287 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1288 evl->evl_tag = htons(tag); 1289 return (m); 1290} 1291 1292static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0, 1293 "IEEE 802.1Q VLAN"); 1294static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0, 1295 "for consistency"); 1296 1297VNET_DEFINE_STATIC(int, soft_pad); 1298#define V_soft_pad VNET(soft_pad) 1299SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET, 1300 &VNET_NAME(soft_pad), 0, 1301 "pad short frames before tagging"); 1302 1303/* 1304 * For now, make preserving PCP via an mbuf tag optional, as it increases 1305 * per-packet memory allocations and frees. In the future, it would be 1306 * preferable to reuse ether_vtag for this, or similar. 1307 */ 1308int vlan_mtag_pcp = 0; 1309SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW, 1310 &vlan_mtag_pcp, 0, 1311 "Retain VLAN PCP information as packets are passed up the stack"); 1312 1313bool 1314ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p, 1315 uint16_t vid, uint8_t pcp) 1316{ 1317 struct m_tag *mtag; 1318 int n; 1319 uint16_t tag; 1320 static const char pad[8]; /* just zeros */ 1321 1322 /* 1323 * Pad the frame to the minimum size allowed if told to. 1324 * This option is in accord with IEEE Std 802.1Q, 2003 Ed., 1325 * paragraph C.4.4.3.b. It can help to work around buggy 1326 * bridges that violate paragraph C.4.4.3.a from the same 1327 * document, i.e., fail to pad short frames after untagging. 1328 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but 1329 * untagging it will produce a 62-byte frame, which is a runt 1330 * and requires padding. There are VLAN-enabled network 1331 * devices that just discard such runts instead or mishandle 1332 * them somehow. 1333 */ 1334 if (V_soft_pad && p->if_type == IFT_ETHER) { 1335 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len; 1336 n > 0; n -= sizeof(pad)) { 1337 if (!m_append(*mp, min(n, sizeof(pad)), pad)) 1338 break; 1339 } 1340 if (n > 0) { 1341 m_freem(*mp); 1342 *mp = NULL; 1343 if_printf(ife, "cannot pad short frame"); 1344 return (false); 1345 } 1346 } 1347 1348 /* 1349 * If PCP is set in mbuf, use it 1350 */ 1351 if ((*mp)->m_flags & M_VLANTAG) { 1352 pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag); 1353 } 1354 1355 /* 1356 * If underlying interface can do VLAN tag insertion itself, 1357 * just pass the packet along. However, we need some way to 1358 * tell the interface where the packet came from so that it 1359 * knows how to find the VLAN tag to use, so we attach a 1360 * packet tag that holds it. 1361 */ 1362 if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q, 1363 MTAG_8021Q_PCP_OUT, NULL)) != NULL) 1364 tag = EVL_MAKETAG(vid, *(uint8_t *)(mtag + 1), 0); 1365 else 1366 tag = EVL_MAKETAG(vid, pcp, 0); 1367 if (p->if_capenable & IFCAP_VLAN_HWTAGGING) { 1368 (*mp)->m_pkthdr.ether_vtag = tag; 1369 (*mp)->m_flags |= M_VLANTAG; 1370 } else { 1371 *mp = ether_vlanencap(*mp, tag); 1372 if (*mp == NULL) { 1373 if_printf(ife, "unable to prepend 802.1Q header"); 1374 return (false); 1375 } 1376 } 1377 return (true); 1378} 1379 1380/* 1381 * Allocate an address from the FreeBSD Foundation OUI. This uses a 1382 * cryptographic hash function on the containing jail's name, UUID and the 1383 * interface name to attempt to provide a unique but stable address. 1384 * Pseudo-interfaces which require a MAC address should use this function to 1385 * allocate non-locally-administered addresses. 1386 */ 1387void 1388ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr) 1389{ 1390 SHA1_CTX ctx; 1391 char *buf; 1392 char uuid[HOSTUUIDLEN + 1]; 1393 uint64_t addr; 1394 int i, sz; 1395 char digest[SHA1_RESULTLEN]; 1396 char jailname[MAXHOSTNAMELEN]; 1397 1398 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid)); 1399 /* If each (vnet) jail would also have a unique hostuuid this would not 1400 * be necessary. */ 1401 getjailname(curthread->td_ucred, jailname, sizeof(jailname)); 1402 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp), 1403 jailname); 1404 if (sz < 0) { 1405 /* Fall back to a random mac address. */ 1406 arc4rand(hwaddr, sizeof(*hwaddr), 0); 1407 hwaddr->octet[0] = 0x02; 1408 return; 1409 } 1410 1411 SHA1Init(&ctx); 1412 SHA1Update(&ctx, buf, sz); 1413 SHA1Final(digest, &ctx); 1414 free(buf, M_TEMP); 1415 1416 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) & 1417 OUI_FREEBSD_GENERATED_MASK; 1418 addr = OUI_FREEBSD(addr); 1419 for (i = 0; i < ETHER_ADDR_LEN; ++i) { 1420 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) & 1421 0xFF; 1422 } 1423} 1424 1425DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1426MODULE_VERSION(ether, 1); 1427