if_ethersubr.c revision 332159
1/*- 2 * Copyright (c) 1982, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 30 * $FreeBSD: stable/11/sys/net/if_ethersubr.c 332159 2018-04-06 23:31:47Z brooks $ 31 */ 32 33#include "opt_inet.h" 34#include "opt_inet6.h" 35#include "opt_netgraph.h" 36#include "opt_mbuf_profiling.h" 37#include "opt_rss.h" 38 39#include <sys/param.h> 40#include <sys/systm.h> 41#include <sys/bus.h> 42#include <sys/eventhandler.h> 43#include <sys/kernel.h> 44#include <sys/lock.h> 45#include <sys/malloc.h> 46#include <sys/module.h> 47#include <sys/mbuf.h> 48#include <sys/random.h> 49#include <sys/socket.h> 50#include <sys/sockio.h> 51#include <sys/sysctl.h> 52#include <sys/uuid.h> 53 54#include <net/if.h> 55#include <net/if_var.h> 56#include <net/if_arp.h> 57#include <net/netisr.h> 58#include <net/route.h> 59#include <net/if_llc.h> 60#include <net/if_dl.h> 61#include <net/if_types.h> 62#include <net/bpf.h> 63#include <net/ethernet.h> 64#include <net/if_bridgevar.h> 65#include <net/if_vlan_var.h> 66#include <net/if_llatbl.h> 67#include <net/pfil.h> 68#include <net/rss_config.h> 69#include <net/vnet.h> 70 71#include <netpfil/pf/pf_mtag.h> 72 73#if defined(INET) || defined(INET6) 74#include <netinet/in.h> 75#include <netinet/in_var.h> 76#include <netinet/if_ether.h> 77#include <netinet/ip_carp.h> 78#include <netinet/ip_var.h> 79#endif 80#ifdef INET6 81#include <netinet6/nd6.h> 82#endif 83#include <security/mac/mac_framework.h> 84 85#ifdef CTASSERT 86CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); 87CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); 88#endif 89 90VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */ 91 92/* netgraph node hooks for ng_ether(4) */ 93void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 94void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 95int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 96void (*ng_ether_attach_p)(struct ifnet *ifp); 97void (*ng_ether_detach_p)(struct ifnet *ifp); 98 99void (*vlan_input_p)(struct ifnet *, struct mbuf *); 100 101/* if_bridge(4) support */ 102struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 103int (*bridge_output_p)(struct ifnet *, struct mbuf *, 104 struct sockaddr *, struct rtentry *); 105void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 106 107/* if_lagg(4) support */ 108struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *); 109 110static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] = 111 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 112 113static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 114 struct sockaddr *); 115#ifdef VIMAGE 116static void ether_reassign(struct ifnet *, struct vnet *, char *); 117#endif 118static int ether_requestencap(struct ifnet *, struct if_encap_req *); 119 120#define ETHER_IS_BROADCAST(addr) \ 121 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0) 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 int loop_copy = 1; 290 int hlen; /* link layer header length */ 291 uint32_t pflags; 292 struct llentry *lle = NULL; 293 struct rtentry *rt0 = NULL; 294 int addref = 0; 295 296 phdr = NULL; 297 pflags = 0; 298 if (ro != NULL) { 299 /* XXX BPF uses ro_prepend */ 300 if (ro->ro_prepend != NULL) { 301 phdr = ro->ro_prepend; 302 hlen = ro->ro_plen; 303 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) { 304 if ((ro->ro_flags & RT_LLE_CACHE) != 0) { 305 lle = ro->ro_lle; 306 if (lle != NULL && 307 (lle->la_flags & LLE_VALID) == 0) { 308 LLE_FREE(lle); 309 lle = NULL; /* redundant */ 310 ro->ro_lle = NULL; 311 } 312 if (lle == NULL) { 313 /* if we lookup, keep cache */ 314 addref = 1; 315 } 316 } 317 if (lle != NULL) { 318 phdr = lle->r_linkdata; 319 hlen = lle->r_hdrlen; 320 pflags = lle->r_flags; 321 } 322 } 323 rt0 = ro->ro_rt; 324 } 325 326#ifdef MAC 327 error = mac_ifnet_check_transmit(ifp, m); 328 if (error) 329 senderr(error); 330#endif 331 332 M_PROFILE(m); 333 if (ifp->if_flags & IFF_MONITOR) 334 senderr(ENETDOWN); 335 if (!((ifp->if_flags & IFF_UP) && 336 (ifp->if_drv_flags & IFF_DRV_RUNNING))) 337 senderr(ENETDOWN); 338 339 if (phdr == NULL) { 340 /* No prepend data supplied. Try to calculate ourselves. */ 341 phdr = linkhdr; 342 hlen = ETHER_HDR_LEN; 343 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags, 344 addref ? &lle : NULL); 345 if (addref && lle != NULL) 346 ro->ro_lle = lle; 347 if (error != 0) 348 return (error == EWOULDBLOCK ? 0 : error); 349 } 350 351 if ((pflags & RT_L2_ME) != 0) { 352 update_mbuf_csumflags(m, m); 353 return (if_simloop(ifp, m, dst->sa_family, 0)); 354 } 355 loop_copy = pflags & RT_MAY_LOOP; 356 357 /* 358 * Add local net header. If no space in first mbuf, 359 * allocate another. 360 * 361 * Note that we do prepend regardless of RT_HAS_HEADER flag. 362 * This is done because BPF code shifts m_data pointer 363 * to the end of ethernet header prior to calling if_output(). 364 */ 365 M_PREPEND(m, hlen, M_NOWAIT); 366 if (m == NULL) 367 senderr(ENOBUFS); 368 if ((pflags & RT_HAS_HEADER) == 0) { 369 eh = mtod(m, struct ether_header *); 370 memcpy(eh, phdr, hlen); 371 } 372 373 /* 374 * If a simplex interface, and the packet is being sent to our 375 * Ethernet address or a broadcast address, loopback a copy. 376 * XXX To make a simplex device behave exactly like a duplex 377 * device, we should copy in the case of sending to our own 378 * ethernet address (thus letting the original actually appear 379 * on the wire). However, we don't do that here for security 380 * reasons and compatibility with the original behavior. 381 */ 382 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) && 383 ((t = pf_find_mtag(m)) == NULL || !t->routed)) { 384 struct mbuf *n; 385 386 /* 387 * Because if_simloop() modifies the packet, we need a 388 * writable copy through m_dup() instead of a readonly 389 * one as m_copy[m] would give us. The alternative would 390 * be to modify if_simloop() to handle the readonly mbuf, 391 * but performancewise it is mostly equivalent (trading 392 * extra data copying vs. extra locking). 393 * 394 * XXX This is a local workaround. A number of less 395 * often used kernel parts suffer from the same bug. 396 * See PR kern/105943 for a proposed general solution. 397 */ 398 if ((n = m_dup(m, M_NOWAIT)) != NULL) { 399 update_mbuf_csumflags(m, n); 400 (void)if_simloop(ifp, n, dst->sa_family, hlen); 401 } else 402 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); 403 } 404 405 /* 406 * Bridges require special output handling. 407 */ 408 if (ifp->if_bridge) { 409 BRIDGE_OUTPUT(ifp, m, error); 410 return (error); 411 } 412 413#if defined(INET) || defined(INET6) 414 if (ifp->if_carp && 415 (error = (*carp_output_p)(ifp, m, dst))) 416 goto bad; 417#endif 418 419 /* Handle ng_ether(4) processing, if any */ 420 if (ifp->if_l2com != NULL) { 421 KASSERT(ng_ether_output_p != NULL, 422 ("ng_ether_output_p is NULL")); 423 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { 424bad: if (m != NULL) 425 m_freem(m); 426 return (error); 427 } 428 if (m == NULL) 429 return (0); 430 } 431 432 /* Continue with link-layer output */ 433 return ether_output_frame(ifp, m); 434} 435 436/* 437 * Ethernet link layer output routine to send a raw frame to the device. 438 * 439 * This assumes that the 14 byte Ethernet header is present and contiguous 440 * in the first mbuf (if BRIDGE'ing). 441 */ 442int 443ether_output_frame(struct ifnet *ifp, struct mbuf *m) 444{ 445 int i; 446 447 if (PFIL_HOOKED(&V_link_pfil_hook)) { 448 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL); 449 450 if (i != 0) 451 return (EACCES); 452 453 if (m == NULL) 454 return (0); 455 } 456 457 /* 458 * Queue message on interface, update output statistics if 459 * successful, and start output if interface not yet active. 460 */ 461 return ((ifp->if_transmit)(ifp, m)); 462} 463 464/* 465 * Process a received Ethernet packet; the packet is in the 466 * mbuf chain m with the ethernet header at the front. 467 */ 468static void 469ether_input_internal(struct ifnet *ifp, struct mbuf *m) 470{ 471 struct ether_header *eh; 472 u_short etype; 473 474 if ((ifp->if_flags & IFF_UP) == 0) { 475 m_freem(m); 476 return; 477 } 478#ifdef DIAGNOSTIC 479 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 480 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n"); 481 m_freem(m); 482 return; 483 } 484#endif 485 if (m->m_len < ETHER_HDR_LEN) { 486 /* XXX maybe should pullup? */ 487 if_printf(ifp, "discard frame w/o leading ethernet " 488 "header (len %u pkt len %u)\n", 489 m->m_len, m->m_pkthdr.len); 490 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 491 m_freem(m); 492 return; 493 } 494 eh = mtod(m, struct ether_header *); 495 etype = ntohs(eh->ether_type); 496 random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_ETHER); 497 498 CURVNET_SET_QUIET(ifp->if_vnet); 499 500 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 501 if (ETHER_IS_BROADCAST(eh->ether_dhost)) 502 m->m_flags |= M_BCAST; 503 else 504 m->m_flags |= M_MCAST; 505 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1); 506 } 507 508#ifdef MAC 509 /* 510 * Tag the mbuf with an appropriate MAC label before any other 511 * consumers can get to it. 512 */ 513 mac_ifnet_create_mbuf(ifp, m); 514#endif 515 516 /* 517 * Give bpf a chance at the packet. 518 */ 519 ETHER_BPF_MTAP(ifp, m); 520 521 /* 522 * If the CRC is still on the packet, trim it off. We do this once 523 * and once only in case we are re-entered. Nothing else on the 524 * Ethernet receive path expects to see the FCS. 525 */ 526 if (m->m_flags & M_HASFCS) { 527 m_adj(m, -ETHER_CRC_LEN); 528 m->m_flags &= ~M_HASFCS; 529 } 530 531 if (!(ifp->if_capenable & IFCAP_HWSTATS)) 532 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 533 534 /* Allow monitor mode to claim this frame, after stats are updated. */ 535 if (ifp->if_flags & IFF_MONITOR) { 536 m_freem(m); 537 CURVNET_RESTORE(); 538 return; 539 } 540 541 /* Handle input from a lagg(4) port */ 542 if (ifp->if_type == IFT_IEEE8023ADLAG) { 543 KASSERT(lagg_input_p != NULL, 544 ("%s: if_lagg not loaded!", __func__)); 545 m = (*lagg_input_p)(ifp, m); 546 if (m != NULL) 547 ifp = m->m_pkthdr.rcvif; 548 else { 549 CURVNET_RESTORE(); 550 return; 551 } 552 } 553 554 /* 555 * If the hardware did not process an 802.1Q tag, do this now, 556 * to allow 802.1P priority frames to be passed to the main input 557 * path correctly. 558 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels. 559 */ 560 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) { 561 struct ether_vlan_header *evl; 562 563 if (m->m_len < sizeof(*evl) && 564 (m = m_pullup(m, sizeof(*evl))) == NULL) { 565#ifdef DIAGNOSTIC 566 if_printf(ifp, "cannot pullup VLAN header\n"); 567#endif 568 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 569 CURVNET_RESTORE(); 570 return; 571 } 572 573 evl = mtod(m, struct ether_vlan_header *); 574 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag); 575 m->m_flags |= M_VLANTAG; 576 577 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN, 578 ETHER_HDR_LEN - ETHER_TYPE_LEN); 579 m_adj(m, ETHER_VLAN_ENCAP_LEN); 580 eh = mtod(m, struct ether_header *); 581 } 582 583 M_SETFIB(m, ifp->if_fib); 584 585 /* Allow ng_ether(4) to claim this frame. */ 586 if (ifp->if_l2com != NULL) { 587 KASSERT(ng_ether_input_p != NULL, 588 ("%s: ng_ether_input_p is NULL", __func__)); 589 m->m_flags &= ~M_PROMISC; 590 (*ng_ether_input_p)(ifp, &m); 591 if (m == NULL) { 592 CURVNET_RESTORE(); 593 return; 594 } 595 eh = mtod(m, struct ether_header *); 596 } 597 598 /* 599 * Allow if_bridge(4) to claim this frame. 600 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it 601 * and the frame should be delivered locally. 602 */ 603 if (ifp->if_bridge != NULL) { 604 m->m_flags &= ~M_PROMISC; 605 BRIDGE_INPUT(ifp, m); 606 if (m == NULL) { 607 CURVNET_RESTORE(); 608 return; 609 } 610 eh = mtod(m, struct ether_header *); 611 } 612 613#if defined(INET) || defined(INET6) 614 /* 615 * Clear M_PROMISC on frame so that carp(4) will see it when the 616 * mbuf flows up to Layer 3. 617 * FreeBSD's implementation of carp(4) uses the inprotosw 618 * to dispatch IPPROTO_CARP. carp(4) also allocates its own 619 * Ethernet addresses of the form 00:00:5e:00:01:xx, which 620 * is outside the scope of the M_PROMISC test below. 621 * TODO: Maintain a hash table of ethernet addresses other than 622 * ether_dhost which may be active on this ifp. 623 */ 624 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) { 625 m->m_flags &= ~M_PROMISC; 626 } else 627#endif 628 { 629 /* 630 * If the frame received was not for our MAC address, set the 631 * M_PROMISC flag on the mbuf chain. The frame may need to 632 * be seen by the rest of the Ethernet input path in case of 633 * re-entry (e.g. bridge, vlan, netgraph) but should not be 634 * seen by upper protocol layers. 635 */ 636 if (!ETHER_IS_MULTICAST(eh->ether_dhost) && 637 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0) 638 m->m_flags |= M_PROMISC; 639 } 640 641 ether_demux(ifp, m); 642 CURVNET_RESTORE(); 643} 644 645/* 646 * Ethernet input dispatch; by default, direct dispatch here regardless of 647 * global configuration. However, if RSS is enabled, hook up RSS affinity 648 * so that when deferred or hybrid dispatch is enabled, we can redistribute 649 * load based on RSS. 650 * 651 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or 652 * not it had already done work distribution via multi-queue. Then we could 653 * direct dispatch in the event load balancing was already complete and 654 * handle the case of interfaces with different capabilities better. 655 * 656 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions 657 * at multiple layers? 658 * 659 * XXXRW: For now, enable all this only if RSS is compiled in, although it 660 * works fine without RSS. Need to characterise the performance overhead 661 * of the detour through the netisr code in the event the result is always 662 * direct dispatch. 663 */ 664static void 665ether_nh_input(struct mbuf *m) 666{ 667 668 M_ASSERTPKTHDR(m); 669 KASSERT(m->m_pkthdr.rcvif != NULL, 670 ("%s: NULL interface pointer", __func__)); 671 ether_input_internal(m->m_pkthdr.rcvif, m); 672} 673 674static struct netisr_handler ether_nh = { 675 .nh_name = "ether", 676 .nh_handler = ether_nh_input, 677 .nh_proto = NETISR_ETHER, 678#ifdef RSS 679 .nh_policy = NETISR_POLICY_CPU, 680 .nh_dispatch = NETISR_DISPATCH_DIRECT, 681 .nh_m2cpuid = rss_m2cpuid, 682#else 683 .nh_policy = NETISR_POLICY_SOURCE, 684 .nh_dispatch = NETISR_DISPATCH_DIRECT, 685#endif 686}; 687 688static void 689ether_init(__unused void *arg) 690{ 691 692 netisr_register(ðer_nh); 693} 694SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL); 695 696static void 697vnet_ether_init(__unused void *arg) 698{ 699 int i; 700 701 /* Initialize packet filter hooks. */ 702 V_link_pfil_hook.ph_type = PFIL_TYPE_AF; 703 V_link_pfil_hook.ph_af = AF_LINK; 704 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0) 705 printf("%s: WARNING: unable to register pfil link hook, " 706 "error %d\n", __func__, i); 707#ifdef VIMAGE 708 netisr_register_vnet(ðer_nh); 709#endif 710} 711VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, 712 vnet_ether_init, NULL); 713 714#ifdef VIMAGE 715static void 716vnet_ether_pfil_destroy(__unused void *arg) 717{ 718 int i; 719 720 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0) 721 printf("%s: WARNING: unable to unregister pfil link hook, " 722 "error %d\n", __func__, i); 723} 724VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY, 725 vnet_ether_pfil_destroy, NULL); 726 727static void 728vnet_ether_destroy(__unused void *arg) 729{ 730 731 netisr_unregister_vnet(ðer_nh); 732} 733VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, 734 vnet_ether_destroy, NULL); 735#endif 736 737 738 739static void 740ether_input(struct ifnet *ifp, struct mbuf *m) 741{ 742 743 struct mbuf *mn; 744 745 /* 746 * The drivers are allowed to pass in a chain of packets linked with 747 * m_nextpkt. We split them up into separate packets here and pass 748 * them up. This allows the drivers to amortize the receive lock. 749 */ 750 while (m) { 751 mn = m->m_nextpkt; 752 m->m_nextpkt = NULL; 753 754 /* 755 * We will rely on rcvif being set properly in the deferred context, 756 * so assert it is correct here. 757 */ 758 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p " 759 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp)); 760 CURVNET_SET_QUIET(ifp->if_vnet); 761 netisr_dispatch(NETISR_ETHER, m); 762 CURVNET_RESTORE(); 763 m = mn; 764 } 765} 766 767/* 768 * Upper layer processing for a received Ethernet packet. 769 */ 770void 771ether_demux(struct ifnet *ifp, struct mbuf *m) 772{ 773 struct ether_header *eh; 774 int i, isr; 775 u_short ether_type; 776 777 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__)); 778 779 /* Do not grab PROMISC frames in case we are re-entered. */ 780 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) { 781 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL); 782 783 if (i != 0 || m == NULL) 784 return; 785 } 786 787 eh = mtod(m, struct ether_header *); 788 ether_type = ntohs(eh->ether_type); 789 790 /* 791 * If this frame has a VLAN tag other than 0, call vlan_input() 792 * if its module is loaded. Otherwise, drop. 793 */ 794 if ((m->m_flags & M_VLANTAG) && 795 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) { 796 if (ifp->if_vlantrunk == NULL) { 797 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1); 798 m_freem(m); 799 return; 800 } 801 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!", 802 __func__)); 803 /* Clear before possibly re-entering ether_input(). */ 804 m->m_flags &= ~M_PROMISC; 805 (*vlan_input_p)(ifp, m); 806 return; 807 } 808 809 /* 810 * Pass promiscuously received frames to the upper layer if the user 811 * requested this by setting IFF_PPROMISC. Otherwise, drop them. 812 */ 813 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) { 814 m_freem(m); 815 return; 816 } 817 818 /* 819 * Reset layer specific mbuf flags to avoid confusing upper layers. 820 * Strip off Ethernet header. 821 */ 822 m->m_flags &= ~M_VLANTAG; 823 m_clrprotoflags(m); 824 m_adj(m, ETHER_HDR_LEN); 825 826 /* 827 * Dispatch frame to upper layer. 828 */ 829 switch (ether_type) { 830#ifdef INET 831 case ETHERTYPE_IP: 832 isr = NETISR_IP; 833 break; 834 835 case ETHERTYPE_ARP: 836 if (ifp->if_flags & IFF_NOARP) { 837 /* Discard packet if ARP is disabled on interface */ 838 m_freem(m); 839 return; 840 } 841 isr = NETISR_ARP; 842 break; 843#endif 844#ifdef INET6 845 case ETHERTYPE_IPV6: 846 isr = NETISR_IPV6; 847 break; 848#endif 849 default: 850 goto discard; 851 } 852 netisr_dispatch(isr, m); 853 return; 854 855discard: 856 /* 857 * Packet is to be discarded. If netgraph is present, 858 * hand the packet to it for last chance processing; 859 * otherwise dispose of it. 860 */ 861 if (ifp->if_l2com != NULL) { 862 KASSERT(ng_ether_input_orphan_p != NULL, 863 ("ng_ether_input_orphan_p is NULL")); 864 /* 865 * Put back the ethernet header so netgraph has a 866 * consistent view of inbound packets. 867 */ 868 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 869 (*ng_ether_input_orphan_p)(ifp, m); 870 return; 871 } 872 m_freem(m); 873} 874 875/* 876 * Convert Ethernet address to printable (loggable) representation. 877 * This routine is for compatibility; it's better to just use 878 * 879 * printf("%6D", <pointer to address>, ":"); 880 * 881 * since there's no static buffer involved. 882 */ 883char * 884ether_sprintf(const u_char *ap) 885{ 886 static char etherbuf[18]; 887 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":"); 888 return (etherbuf); 889} 890 891/* 892 * Perform common duties while attaching to interface list 893 */ 894void 895ether_ifattach(struct ifnet *ifp, const u_int8_t *lla) 896{ 897 int i; 898 struct ifaddr *ifa; 899 struct sockaddr_dl *sdl; 900 901 ifp->if_addrlen = ETHER_ADDR_LEN; 902 ifp->if_hdrlen = ETHER_HDR_LEN; 903 if_attach(ifp); 904 ifp->if_mtu = ETHERMTU; 905 ifp->if_output = ether_output; 906 ifp->if_input = ether_input; 907 ifp->if_resolvemulti = ether_resolvemulti; 908 ifp->if_requestencap = ether_requestencap; 909#ifdef VIMAGE 910 ifp->if_reassign = ether_reassign; 911#endif 912 if (ifp->if_baudrate == 0) 913 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 914 ifp->if_broadcastaddr = etherbroadcastaddr; 915 916 ifa = ifp->if_addr; 917 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__)); 918 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 919 sdl->sdl_type = IFT_ETHER; 920 sdl->sdl_alen = ifp->if_addrlen; 921 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 922 923 if (ifp->if_hw_addr != NULL) 924 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen); 925 926 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 927 if (ng_ether_attach_p != NULL) 928 (*ng_ether_attach_p)(ifp); 929 930 /* Announce Ethernet MAC address if non-zero. */ 931 for (i = 0; i < ifp->if_addrlen; i++) 932 if (lla[i] != 0) 933 break; 934 if (i != ifp->if_addrlen) 935 if_printf(ifp, "Ethernet address: %6D\n", lla, ":"); 936 937 uuid_ether_add(LLADDR(sdl)); 938 939 /* Add necessary bits are setup; announce it now. */ 940 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp); 941 if (IS_DEFAULT_VNET(curvnet)) 942 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL); 943} 944 945/* 946 * Perform common duties while detaching an Ethernet interface 947 */ 948void 949ether_ifdetach(struct ifnet *ifp) 950{ 951 struct sockaddr_dl *sdl; 952 953 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr); 954 uuid_ether_del(LLADDR(sdl)); 955 956 if (ifp->if_l2com != NULL) { 957 KASSERT(ng_ether_detach_p != NULL, 958 ("ng_ether_detach_p is NULL")); 959 (*ng_ether_detach_p)(ifp); 960 } 961 962 bpfdetach(ifp); 963 if_detach(ifp); 964} 965 966#ifdef VIMAGE 967void 968ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused) 969{ 970 971 if (ifp->if_l2com != NULL) { 972 KASSERT(ng_ether_detach_p != NULL, 973 ("ng_ether_detach_p is NULL")); 974 (*ng_ether_detach_p)(ifp); 975 } 976 977 if (ng_ether_attach_p != NULL) { 978 CURVNET_SET_QUIET(new_vnet); 979 (*ng_ether_attach_p)(ifp); 980 CURVNET_RESTORE(); 981 } 982} 983#endif 984 985SYSCTL_DECL(_net_link); 986SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 987 988#if 0 989/* 990 * This is for reference. We have a table-driven version 991 * of the little-endian crc32 generator, which is faster 992 * than the double-loop. 993 */ 994uint32_t 995ether_crc32_le(const uint8_t *buf, size_t len) 996{ 997 size_t i; 998 uint32_t crc; 999 int bit; 1000 uint8_t data; 1001 1002 crc = 0xffffffff; /* initial value */ 1003 1004 for (i = 0; i < len; i++) { 1005 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1006 carry = (crc ^ data) & 1; 1007 crc >>= 1; 1008 if (carry) 1009 crc = (crc ^ ETHER_CRC_POLY_LE); 1010 } 1011 } 1012 1013 return (crc); 1014} 1015#else 1016uint32_t 1017ether_crc32_le(const uint8_t *buf, size_t len) 1018{ 1019 static const uint32_t crctab[] = { 1020 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1021 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1022 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1023 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1024 }; 1025 size_t i; 1026 uint32_t crc; 1027 1028 crc = 0xffffffff; /* initial value */ 1029 1030 for (i = 0; i < len; i++) { 1031 crc ^= buf[i]; 1032 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1033 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1034 } 1035 1036 return (crc); 1037} 1038#endif 1039 1040uint32_t 1041ether_crc32_be(const uint8_t *buf, size_t len) 1042{ 1043 size_t i; 1044 uint32_t crc, carry; 1045 int bit; 1046 uint8_t data; 1047 1048 crc = 0xffffffff; /* initial value */ 1049 1050 for (i = 0; i < len; i++) { 1051 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) { 1052 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01); 1053 crc <<= 1; 1054 if (carry) 1055 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1056 } 1057 } 1058 1059 return (crc); 1060} 1061 1062int 1063ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 1064{ 1065 struct ifaddr *ifa = (struct ifaddr *) data; 1066 struct ifreq *ifr = (struct ifreq *) data; 1067 int error = 0; 1068 1069 switch (command) { 1070 case SIOCSIFADDR: 1071 ifp->if_flags |= IFF_UP; 1072 1073 switch (ifa->ifa_addr->sa_family) { 1074#ifdef INET 1075 case AF_INET: 1076 ifp->if_init(ifp->if_softc); /* before arpwhohas */ 1077 arp_ifinit(ifp, ifa); 1078 break; 1079#endif 1080 default: 1081 ifp->if_init(ifp->if_softc); 1082 break; 1083 } 1084 break; 1085 1086 case SIOCGIFADDR: 1087 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0], 1088 ETHER_ADDR_LEN); 1089 break; 1090 1091 case SIOCSIFMTU: 1092 /* 1093 * Set the interface MTU. 1094 */ 1095 if (ifr->ifr_mtu > ETHERMTU) { 1096 error = EINVAL; 1097 } else { 1098 ifp->if_mtu = ifr->ifr_mtu; 1099 } 1100 break; 1101 default: 1102 error = EINVAL; /* XXX netbsd has ENOTTY??? */ 1103 break; 1104 } 1105 return (error); 1106} 1107 1108static int 1109ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa, 1110 struct sockaddr *sa) 1111{ 1112 struct sockaddr_dl *sdl; 1113#ifdef INET 1114 struct sockaddr_in *sin; 1115#endif 1116#ifdef INET6 1117 struct sockaddr_in6 *sin6; 1118#endif 1119 u_char *e_addr; 1120 1121 switch(sa->sa_family) { 1122 case AF_LINK: 1123 /* 1124 * No mapping needed. Just check that it's a valid MC address. 1125 */ 1126 sdl = (struct sockaddr_dl *)sa; 1127 e_addr = LLADDR(sdl); 1128 if (!ETHER_IS_MULTICAST(e_addr)) 1129 return EADDRNOTAVAIL; 1130 *llsa = NULL; 1131 return 0; 1132 1133#ifdef INET 1134 case AF_INET: 1135 sin = (struct sockaddr_in *)sa; 1136 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 1137 return EADDRNOTAVAIL; 1138 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1139 sdl->sdl_alen = ETHER_ADDR_LEN; 1140 e_addr = LLADDR(sdl); 1141 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 1142 *llsa = (struct sockaddr *)sdl; 1143 return 0; 1144#endif 1145#ifdef INET6 1146 case AF_INET6: 1147 sin6 = (struct sockaddr_in6 *)sa; 1148 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1149 /* 1150 * An IP6 address of 0 means listen to all 1151 * of the Ethernet multicast address used for IP6. 1152 * (This is used for multicast routers.) 1153 */ 1154 ifp->if_flags |= IFF_ALLMULTI; 1155 *llsa = NULL; 1156 return 0; 1157 } 1158 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 1159 return EADDRNOTAVAIL; 1160 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER); 1161 sdl->sdl_alen = ETHER_ADDR_LEN; 1162 e_addr = LLADDR(sdl); 1163 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 1164 *llsa = (struct sockaddr *)sdl; 1165 return 0; 1166#endif 1167 1168 default: 1169 /* 1170 * Well, the text isn't quite right, but it's the name 1171 * that counts... 1172 */ 1173 return EAFNOSUPPORT; 1174 } 1175} 1176 1177static moduledata_t ether_mod = { 1178 .name = "ether", 1179}; 1180 1181void 1182ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen) 1183{ 1184 struct ether_vlan_header vlan; 1185 struct mbuf mv, mb; 1186 1187 KASSERT((m->m_flags & M_VLANTAG) != 0, 1188 ("%s: vlan information not present", __func__)); 1189 KASSERT(m->m_len >= sizeof(struct ether_header), 1190 ("%s: mbuf not large enough for header", __func__)); 1191 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header)); 1192 vlan.evl_proto = vlan.evl_encap_proto; 1193 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN); 1194 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag); 1195 m->m_len -= sizeof(struct ether_header); 1196 m->m_data += sizeof(struct ether_header); 1197 /* 1198 * If a data link has been supplied by the caller, then we will need to 1199 * re-create a stack allocated mbuf chain with the following structure: 1200 * 1201 * (1) mbuf #1 will contain the supplied data link 1202 * (2) mbuf #2 will contain the vlan header 1203 * (3) mbuf #3 will contain the original mbuf's packet data 1204 * 1205 * Otherwise, submit the packet and vlan header via bpf_mtap2(). 1206 */ 1207 if (data != NULL) { 1208 mv.m_next = m; 1209 mv.m_data = (caddr_t)&vlan; 1210 mv.m_len = sizeof(vlan); 1211 mb.m_next = &mv; 1212 mb.m_data = data; 1213 mb.m_len = dlen; 1214 bpf_mtap(bp, &mb); 1215 } else 1216 bpf_mtap2(bp, &vlan, sizeof(vlan), m); 1217 m->m_len += sizeof(struct ether_header); 1218 m->m_data -= sizeof(struct ether_header); 1219} 1220 1221struct mbuf * 1222ether_vlanencap(struct mbuf *m, uint16_t tag) 1223{ 1224 struct ether_vlan_header *evl; 1225 1226 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT); 1227 if (m == NULL) 1228 return (NULL); 1229 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */ 1230 1231 if (m->m_len < sizeof(*evl)) { 1232 m = m_pullup(m, sizeof(*evl)); 1233 if (m == NULL) 1234 return (NULL); 1235 } 1236 1237 /* 1238 * Transform the Ethernet header into an Ethernet header 1239 * with 802.1Q encapsulation. 1240 */ 1241 evl = mtod(m, struct ether_vlan_header *); 1242 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN, 1243 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN); 1244 evl->evl_encap_proto = htons(ETHERTYPE_VLAN); 1245 evl->evl_tag = htons(tag); 1246 return (m); 1247} 1248 1249DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY); 1250MODULE_VERSION(ether, 1); 1251