1 /* $NetBSD: if_ethersubr.c,v 1.320 2022/09/03 02:47:59 thorpej Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * 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 project 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 PROJECT 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 PROJECT 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 /* 33 * Copyright (c) 1982, 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)if_ethersubr.c 8.2 (Berkeley) 4/4/96 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: if_ethersubr.c,v 1.320 2022/09/03 02:47:59 thorpej Exp $"); 65 66 #ifdef _KERNEL_OPT 67 #include "opt_inet.h" 68 #include "opt_atalk.h" 69 #include "opt_mbuftrace.h" 70 #include "opt_mpls.h" 71 #include "opt_gateway.h" 72 #include "opt_pppoe.h" 73 #include "opt_net_mpsafe.h" 74 #endif 75 76 #include "vlan.h" 77 #include "pppoe.h" 78 #include "bridge.h" 79 #include "arp.h" 80 #include "agr.h" 81 82 #include <sys/sysctl.h> 83 #include <sys/mbuf.h> 84 #include <sys/mutex.h> 85 #include <sys/ioctl.h> 86 #include <sys/errno.h> 87 #include <sys/device.h> 88 #include <sys/entropy.h> 89 #include <sys/rndsource.h> 90 #include <sys/cpu.h> 91 #include <sys/kmem.h> 92 #include <sys/hook.h> 93 94 #include <net/if.h> 95 #include <net/route.h> 96 #include <net/if_llc.h> 97 #include <net/if_dl.h> 98 #include <net/if_types.h> 99 #include <net/pktqueue.h> 100 101 #include <net/if_media.h> 102 #include <dev/mii/mii.h> 103 #include <dev/mii/miivar.h> 104 105 #if NARP == 0 106 /* 107 * XXX there should really be a way to issue this warning from within config(8) 108 */ 109 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK. 110 #endif 111 112 #include <net/bpf.h> 113 114 #include <net/if_ether.h> 115 #include <net/if_vlanvar.h> 116 117 #if NPPPOE > 0 118 #include <net/if_pppoe.h> 119 #endif 120 121 #if NAGR > 0 122 #include <net/ether_slowprotocols.h> 123 #include <net/agr/ieee8023ad.h> 124 #include <net/agr/if_agrvar.h> 125 #endif 126 127 #if NBRIDGE > 0 128 #include <net/if_bridgevar.h> 129 #endif 130 131 #include <netinet/in.h> 132 #ifdef INET 133 #include <netinet/in_var.h> 134 #endif 135 #include <netinet/if_inarp.h> 136 137 #ifdef INET6 138 #ifndef INET 139 #include <netinet/in.h> 140 #endif 141 #include <netinet6/in6_var.h> 142 #include <netinet6/nd6.h> 143 #endif 144 145 #include "carp.h" 146 #if NCARP > 0 147 #include <netinet/ip_carp.h> 148 #endif 149 150 #ifdef NETATALK 151 #include <netatalk/at.h> 152 #include <netatalk/at_var.h> 153 #include <netatalk/at_extern.h> 154 155 #define llc_snap_org_code llc_un.type_snap.org_code 156 #define llc_snap_ether_type llc_un.type_snap.ether_type 157 158 extern u_char at_org_code[3]; 159 extern u_char aarp_org_code[3]; 160 #endif /* NETATALK */ 161 162 #ifdef MPLS 163 #include <netmpls/mpls.h> 164 #include <netmpls/mpls_var.h> 165 #endif 166 167 CTASSERT(sizeof(struct ether_addr) == 6); 168 CTASSERT(sizeof(struct ether_header) == 14); 169 170 #ifdef DIAGNOSTIC 171 static struct timeval bigpktppslim_last; 172 static int bigpktppslim = 2; /* XXX */ 173 static int bigpktpps_count; 174 static kmutex_t bigpktpps_lock __cacheline_aligned; 175 #endif 176 177 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = 178 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 179 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] = 180 { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 }; 181 #define senderr(e) { error = (e); goto bad;} 182 183 static pktq_rps_hash_func_t ether_pktq_rps_hash_p; 184 185 static int ether_output(struct ifnet *, struct mbuf *, 186 const struct sockaddr *, const struct rtentry *); 187 188 /* 189 * Ethernet output routine. 190 * Encapsulate a packet of type family for the local net. 191 * Assumes that ifp is actually pointer to ethercom structure. 192 */ 193 static int 194 ether_output(struct ifnet * const ifp0, struct mbuf * const m0, 195 const struct sockaddr * const dst, const struct rtentry *rt) 196 { 197 uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 198 uint16_t etype = 0; 199 int error = 0, hdrcmplt = 0; 200 struct mbuf *m = m0; 201 struct mbuf *mcopy = NULL; 202 struct ether_header *eh; 203 struct ifnet *ifp = ifp0; 204 #ifdef INET 205 struct arphdr *ah; 206 #endif 207 #ifdef NETATALK 208 struct at_ifaddr *aa; 209 #endif 210 211 #ifdef MBUFTRACE 212 m_claimm(m, ifp->if_mowner); 213 #endif 214 215 #if NCARP > 0 216 if (ifp->if_type == IFT_CARP) { 217 struct ifaddr *ifa; 218 int s = pserialize_read_enter(); 219 220 /* loop back if this is going to the carp interface */ 221 if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP && 222 (ifa = ifa_ifwithaddr(dst)) != NULL) { 223 if (ifa->ifa_ifp == ifp0) { 224 pserialize_read_exit(s); 225 return looutput(ifp0, m, dst, rt); 226 } 227 } 228 pserialize_read_exit(s); 229 230 ifp = ifp->if_carpdev; 231 /* ac = (struct arpcom *)ifp; */ 232 233 if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) != 234 (IFF_UP | IFF_RUNNING)) 235 senderr(ENETDOWN); 236 } 237 #endif 238 239 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 240 senderr(ENETDOWN); 241 242 switch (dst->sa_family) { 243 244 #ifdef INET 245 case AF_INET: 246 if (m->m_flags & M_BCAST) { 247 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 248 } else if (m->m_flags & M_MCAST) { 249 ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst); 250 } else { 251 error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst)); 252 if (error) 253 return (error == EWOULDBLOCK) ? 0 : error; 254 } 255 /* If broadcasting on a simplex interface, loopback a copy */ 256 if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX)) 257 mcopy = m_copypacket(m, M_DONTWAIT); 258 etype = htons(ETHERTYPE_IP); 259 break; 260 261 case AF_ARP: 262 ah = mtod(m, struct arphdr *); 263 if (m->m_flags & M_BCAST) { 264 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 265 } else { 266 void *tha = ar_tha(ah); 267 268 if (tha == NULL) { 269 /* fake with ARPHRD_IEEE1394 */ 270 m_freem(m); 271 return 0; 272 } 273 memcpy(edst, tha, sizeof(edst)); 274 } 275 276 ah->ar_hrd = htons(ARPHRD_ETHER); 277 278 switch (ntohs(ah->ar_op)) { 279 case ARPOP_REVREQUEST: 280 case ARPOP_REVREPLY: 281 etype = htons(ETHERTYPE_REVARP); 282 break; 283 284 case ARPOP_REQUEST: 285 case ARPOP_REPLY: 286 default: 287 etype = htons(ETHERTYPE_ARP); 288 } 289 break; 290 #endif 291 292 #ifdef INET6 293 case AF_INET6: 294 if (m->m_flags & M_BCAST) { 295 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 296 } else if (m->m_flags & M_MCAST) { 297 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 298 edst); 299 } else { 300 error = nd6_resolve(ifp0, rt, m, dst, edst, 301 sizeof(edst)); 302 if (error) 303 return (error == EWOULDBLOCK) ? 0 : error; 304 } 305 etype = htons(ETHERTYPE_IPV6); 306 break; 307 #endif 308 309 #ifdef NETATALK 310 case AF_APPLETALK: { 311 struct ifaddr *ifa; 312 int s; 313 314 KERNEL_LOCK(1, NULL); 315 316 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 317 KERNEL_UNLOCK_ONE(NULL); 318 return 0; 319 } 320 321 /* 322 * ifaddr is the first thing in at_ifaddr 323 */ 324 s = pserialize_read_enter(); 325 ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp); 326 if (ifa == NULL) { 327 pserialize_read_exit(s); 328 KERNEL_UNLOCK_ONE(NULL); 329 senderr(EADDRNOTAVAIL); 330 } 331 aa = (struct at_ifaddr *)ifa; 332 333 /* 334 * In the phase 2 case, we need to prepend an mbuf for the 335 * llc header. 336 */ 337 if (aa->aa_flags & AFA_PHASE2) { 338 struct llc llc; 339 340 M_PREPEND(m, sizeof(struct llc), M_DONTWAIT); 341 if (m == NULL) { 342 pserialize_read_exit(s); 343 KERNEL_UNLOCK_ONE(NULL); 344 senderr(ENOBUFS); 345 } 346 347 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 348 llc.llc_control = LLC_UI; 349 memcpy(llc.llc_snap_org_code, at_org_code, 350 sizeof(llc.llc_snap_org_code)); 351 llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK); 352 memcpy(mtod(m, void *), &llc, sizeof(struct llc)); 353 } else { 354 etype = htons(ETHERTYPE_ATALK); 355 } 356 pserialize_read_exit(s); 357 KERNEL_UNLOCK_ONE(NULL); 358 break; 359 } 360 #endif /* NETATALK */ 361 362 case pseudo_AF_HDRCMPLT: 363 hdrcmplt = 1; 364 memcpy(esrc, 365 ((const struct ether_header *)dst->sa_data)->ether_shost, 366 sizeof(esrc)); 367 /* FALLTHROUGH */ 368 369 case AF_UNSPEC: 370 memcpy(edst, 371 ((const struct ether_header *)dst->sa_data)->ether_dhost, 372 sizeof(edst)); 373 /* AF_UNSPEC doesn't swap the byte order of the ether_type. */ 374 etype = ((const struct ether_header *)dst->sa_data)->ether_type; 375 break; 376 377 default: 378 printf("%s: can't handle af%d\n", ifp->if_xname, 379 dst->sa_family); 380 senderr(EAFNOSUPPORT); 381 } 382 383 #ifdef MPLS 384 { 385 struct m_tag *mtag; 386 mtag = m_tag_find(m, PACKET_TAG_MPLS); 387 if (mtag != NULL) { 388 /* Having the tag itself indicates it's MPLS */ 389 etype = htons(ETHERTYPE_MPLS); 390 m_tag_delete(m, mtag); 391 } 392 } 393 #endif 394 395 if (mcopy) 396 (void)looutput(ifp, mcopy, dst, rt); 397 398 KASSERT((m->m_flags & M_PKTHDR) != 0); 399 400 /* 401 * If no ether type is set, this must be a 802.2 formatted packet. 402 */ 403 if (etype == 0) 404 etype = htons(m->m_pkthdr.len); 405 406 /* 407 * Add local net header. If no space in first mbuf, allocate another. 408 */ 409 M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT); 410 if (m == NULL) 411 senderr(ENOBUFS); 412 413 eh = mtod(m, struct ether_header *); 414 /* Note: etype is already in network byte order. */ 415 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 416 memcpy(eh->ether_dhost, edst, sizeof(edst)); 417 if (hdrcmplt) { 418 memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost)); 419 } else { 420 memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl), 421 sizeof(eh->ether_shost)); 422 } 423 424 #if NCARP > 0 425 if (ifp0 != ifp && ifp0->if_type == IFT_CARP) { 426 memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl), 427 sizeof(eh->ether_shost)); 428 } 429 #endif 430 431 if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0) 432 return error; 433 if (m == NULL) 434 return 0; 435 436 #if NBRIDGE > 0 437 /* 438 * Bridges require special output handling. 439 */ 440 if (ifp->if_bridge) 441 return bridge_output(ifp, m, NULL, NULL); 442 #endif 443 444 #if NCARP > 0 445 if (ifp != ifp0) 446 if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN); 447 #endif 448 449 #ifdef ALTQ 450 KERNEL_LOCK(1, NULL); 451 /* 452 * If ALTQ is enabled on the parent interface, do 453 * classification; the queueing discipline might not 454 * require classification, but might require the 455 * address family/header pointer in the pktattr. 456 */ 457 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 458 altq_etherclassify(&ifp->if_snd, m); 459 KERNEL_UNLOCK_ONE(NULL); 460 #endif 461 return ifq_enqueue(ifp, m); 462 463 bad: 464 if_statinc(ifp, if_oerrors); 465 if (m) 466 m_freem(m); 467 return error; 468 } 469 470 #ifdef ALTQ 471 /* 472 * This routine is a slight hack to allow a packet to be classified 473 * if the Ethernet headers are present. It will go away when ALTQ's 474 * classification engine understands link headers. 475 * 476 * XXX: We may need to do m_pullups here. First to ensure struct ether_header 477 * is indeed contiguous, then to read the LLC and so on. 478 */ 479 void 480 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m) 481 { 482 struct ether_header *eh; 483 struct mbuf *mtop = m; 484 uint16_t ether_type; 485 int hlen, af, hdrsize; 486 void *hdr; 487 488 KASSERT((mtop->m_flags & M_PKTHDR) != 0); 489 490 hlen = ETHER_HDR_LEN; 491 eh = mtod(m, struct ether_header *); 492 493 ether_type = htons(eh->ether_type); 494 495 if (ether_type < ETHERMTU) { 496 /* LLC/SNAP */ 497 struct llc *llc = (struct llc *)(eh + 1); 498 hlen += 8; 499 500 if (m->m_len < hlen || 501 llc->llc_dsap != LLC_SNAP_LSAP || 502 llc->llc_ssap != LLC_SNAP_LSAP || 503 llc->llc_control != LLC_UI) { 504 /* Not SNAP. */ 505 goto bad; 506 } 507 508 ether_type = htons(llc->llc_un.type_snap.ether_type); 509 } 510 511 switch (ether_type) { 512 case ETHERTYPE_IP: 513 af = AF_INET; 514 hdrsize = 20; /* sizeof(struct ip) */ 515 break; 516 517 case ETHERTYPE_IPV6: 518 af = AF_INET6; 519 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 520 break; 521 522 default: 523 af = AF_UNSPEC; 524 hdrsize = 0; 525 break; 526 } 527 528 while (m->m_len <= hlen) { 529 hlen -= m->m_len; 530 m = m->m_next; 531 if (m == NULL) 532 goto bad; 533 } 534 535 if (m->m_len < (hlen + hdrsize)) { 536 /* 537 * protocol header not in a single mbuf. 538 * We can't cope with this situation right 539 * now (but it shouldn't ever happen, really, anyhow). 540 */ 541 #ifdef DEBUG 542 printf("altq_etherclassify: headers span multiple mbufs: " 543 "%d < %d\n", m->m_len, (hlen + hdrsize)); 544 #endif 545 goto bad; 546 } 547 548 m->m_data += hlen; 549 m->m_len -= hlen; 550 551 hdr = mtod(m, void *); 552 553 if (ALTQ_NEEDS_CLASSIFY(ifq)) { 554 mtop->m_pkthdr.pattr_class = 555 (*ifq->altq_classify)(ifq->altq_clfier, m, af); 556 } 557 mtop->m_pkthdr.pattr_af = af; 558 mtop->m_pkthdr.pattr_hdr = hdr; 559 560 m->m_data -= hlen; 561 m->m_len += hlen; 562 563 return; 564 565 bad: 566 mtop->m_pkthdr.pattr_class = NULL; 567 mtop->m_pkthdr.pattr_hdr = NULL; 568 mtop->m_pkthdr.pattr_af = AF_UNSPEC; 569 } 570 #endif /* ALTQ */ 571 572 #if defined (LLC) || defined (NETATALK) 573 static void 574 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh) 575 { 576 pktqueue_t *pktq = NULL; 577 struct llc *l; 578 579 if (m->m_len < sizeof(*eh) + sizeof(struct llc)) 580 goto error; 581 582 l = (struct llc *)(eh+1); 583 switch (l->llc_dsap) { 584 #ifdef NETATALK 585 case LLC_SNAP_LSAP: 586 switch (l->llc_control) { 587 case LLC_UI: 588 if (l->llc_ssap != LLC_SNAP_LSAP) 589 goto error; 590 591 if (memcmp(&(l->llc_snap_org_code)[0], 592 at_org_code, sizeof(at_org_code)) == 0 && 593 ntohs(l->llc_snap_ether_type) == 594 ETHERTYPE_ATALK) { 595 pktq = at_pktq2; 596 m_adj(m, sizeof(struct ether_header) 597 + sizeof(struct llc)); 598 break; 599 } 600 601 if (memcmp(&(l->llc_snap_org_code)[0], 602 aarp_org_code, 603 sizeof(aarp_org_code)) == 0 && 604 ntohs(l->llc_snap_ether_type) == 605 ETHERTYPE_AARP) { 606 m_adj(m, sizeof(struct ether_header) 607 + sizeof(struct llc)); 608 aarpinput(ifp, m); /* XXX queue? */ 609 return; 610 } 611 612 default: 613 goto error; 614 } 615 break; 616 #endif 617 default: 618 goto noproto; 619 } 620 621 KASSERT(pktq != NULL); 622 if (__predict_false(!pktq_enqueue(pktq, m, 0))) { 623 m_freem(m); 624 } 625 return; 626 627 noproto: 628 m_freem(m); 629 if_statinc(ifp, if_noproto); 630 return; 631 error: 632 m_freem(m); 633 if_statinc(ifp, if_ierrors); 634 return; 635 } 636 #endif /* defined (LLC) || defined (NETATALK) */ 637 638 /* 639 * Process a received Ethernet packet; 640 * the packet is in the mbuf chain m with 641 * the ether header. 642 */ 643 void 644 ether_input(struct ifnet *ifp, struct mbuf *m) 645 { 646 #if NVLAN > 0 || defined(MBUFTRACE) 647 struct ethercom *ec = (struct ethercom *) ifp; 648 #endif 649 pktqueue_t *pktq = NULL; 650 uint16_t etype; 651 struct ether_header *eh; 652 size_t ehlen; 653 static int earlypkts; 654 655 /* No RPS for not-IP. */ 656 pktq_rps_hash_func_t rps_hash = NULL; 657 658 KASSERT(!cpu_intr_p()); 659 KASSERT((m->m_flags & M_PKTHDR) != 0); 660 661 if ((ifp->if_flags & IFF_UP) == 0) 662 goto drop; 663 664 #ifdef MBUFTRACE 665 m_claimm(m, &ec->ec_rx_mowner); 666 #endif 667 668 if (__predict_false(m->m_len < sizeof(*eh))) { 669 if ((m = m_pullup(m, sizeof(*eh))) == NULL) { 670 if_statinc(ifp, if_ierrors); 671 return; 672 } 673 } 674 675 eh = mtod(m, struct ether_header *); 676 etype = ntohs(eh->ether_type); 677 ehlen = sizeof(*eh); 678 679 if (__predict_false(earlypkts < 100 || 680 entropy_epoch() == (unsigned)-1)) { 681 rnd_add_data(NULL, eh, ehlen, 0); 682 earlypkts++; 683 } 684 685 /* 686 * Determine if the packet is within its size limits. For MPLS the 687 * header length is variable, so we skip the check. 688 */ 689 if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len > 690 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 691 #ifdef DIAGNOSTIC 692 mutex_enter(&bigpktpps_lock); 693 if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count, 694 bigpktppslim)) { 695 printf("%s: discarding oversize frame (len=%d)\n", 696 ifp->if_xname, m->m_pkthdr.len); 697 } 698 mutex_exit(&bigpktpps_lock); 699 #endif 700 goto error; 701 } 702 703 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 704 /* 705 * If this is not a simplex interface, drop the packet 706 * if it came from us. 707 */ 708 if ((ifp->if_flags & IFF_SIMPLEX) == 0 && 709 memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost, 710 ETHER_ADDR_LEN) == 0) { 711 goto drop; 712 } 713 714 if (memcmp(etherbroadcastaddr, 715 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 716 m->m_flags |= M_BCAST; 717 else 718 m->m_flags |= M_MCAST; 719 if_statinc(ifp, if_imcasts); 720 } 721 722 /* If the CRC is still on the packet, trim it off. */ 723 if (m->m_flags & M_HASFCS) { 724 m_adj(m, -ETHER_CRC_LEN); 725 m->m_flags &= ~M_HASFCS; 726 } 727 728 if_statadd(ifp, if_ibytes, m->m_pkthdr.len); 729 730 if (!vlan_has_tag(m) && etype == ETHERTYPE_VLAN) { 731 m = ether_strip_vlantag(m); 732 if (m == NULL) { 733 if_statinc(ifp, if_ierrors); 734 return; 735 } 736 737 eh = mtod(m, struct ether_header *); 738 etype = ntohs(eh->ether_type); 739 ehlen = sizeof(*eh); 740 } 741 742 if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 && 743 (ifp->if_flags & IFF_PROMISC) != 0 && 744 memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost, 745 ETHER_ADDR_LEN) != 0) { 746 m->m_flags |= M_PROMISC; 747 } 748 749 if ((m->m_flags & M_PROMISC) == 0) { 750 if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0) 751 return; 752 if (m == NULL) 753 return; 754 755 eh = mtod(m, struct ether_header *); 756 etype = ntohs(eh->ether_type); 757 } 758 759 /* 760 * Processing a logical interfaces that are able 761 * to configure vlan(4). 762 */ 763 #if NAGR > 0 764 if (ifp->if_lagg != NULL && 765 __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) { 766 m->m_flags &= ~M_PROMISC; 767 agr_input(ifp, m); 768 return; 769 } 770 #endif 771 772 /* 773 * VLAN processing. 774 * 775 * VLAN provides service delimiting so the frames are 776 * processed before other handlings. If a VLAN interface 777 * does not exist to take those frames, they're returned 778 * to ether_input(). 779 */ 780 781 if (vlan_has_tag(m)) { 782 if (EVL_VLANOFTAG(vlan_get_tag(m)) == 0) { 783 if (etype == ETHERTYPE_VLAN || 784 etype == ETHERTYPE_QINQ) 785 goto drop; 786 787 /* XXX we should actually use the prio value? */ 788 m->m_flags &= ~M_VLANTAG; 789 } else { 790 #if NVLAN > 0 791 if (ec->ec_nvlans > 0) { 792 m = vlan_input(ifp, m); 793 794 /* vlan_input() called ether_input() recursively */ 795 if (m == NULL) 796 return; 797 } 798 #endif 799 /* drop VLAN frames not for this port. */ 800 goto noproto; 801 } 802 } 803 804 #if NCARP > 0 805 if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) { 806 /* 807 * Clear M_PROMISC, in case the packet comes from a 808 * vlan. 809 */ 810 m->m_flags &= ~M_PROMISC; 811 if (carp_input(m, (uint8_t *)&eh->ether_shost, 812 (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0) 813 return; 814 } 815 #endif 816 817 /* 818 * Handle protocols that expect to have the Ethernet header 819 * (and possibly FCS) intact. 820 */ 821 switch (etype) { 822 #if NPPPOE > 0 823 case ETHERTYPE_PPPOEDISC: 824 pppoedisc_input(ifp, m); 825 return; 826 827 case ETHERTYPE_PPPOE: 828 pppoe_input(ifp, m); 829 return; 830 #endif 831 832 case ETHERTYPE_SLOWPROTOCOLS: { 833 uint8_t subtype; 834 835 if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) 836 goto error; 837 838 m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype); 839 switch (subtype) { 840 #if NAGR > 0 841 case SLOWPROTOCOLS_SUBTYPE_LACP: 842 if (ifp->if_lagg != NULL) { 843 ieee8023ad_lacp_input(ifp, m); 844 return; 845 } 846 break; 847 848 case SLOWPROTOCOLS_SUBTYPE_MARKER: 849 if (ifp->if_lagg != NULL) { 850 ieee8023ad_marker_input(ifp, m); 851 return; 852 } 853 break; 854 #endif 855 856 default: 857 if (subtype == 0 || subtype > 10) { 858 /* illegal value */ 859 goto error; 860 } 861 /* unknown subtype */ 862 break; 863 } 864 } 865 /* FALLTHROUGH */ 866 default: 867 if (m->m_flags & M_PROMISC) 868 goto drop; 869 } 870 871 /* If the CRC is still on the packet, trim it off. */ 872 if (m->m_flags & M_HASFCS) { 873 m_adj(m, -ETHER_CRC_LEN); 874 m->m_flags &= ~M_HASFCS; 875 } 876 877 /* etype represents the size of the payload in this case */ 878 if (etype <= ETHERMTU + sizeof(struct ether_header)) { 879 KASSERT(ehlen == sizeof(*eh)); 880 #if defined (LLC) || defined (NETATALK) 881 ether_input_llc(ifp, m, eh); 882 return; 883 #else 884 /* ethertype of 0-1500 is regarded as noproto */ 885 goto noproto; 886 #endif 887 } 888 889 /* Strip off the Ethernet header. */ 890 m_adj(m, ehlen); 891 892 switch (etype) { 893 #ifdef INET 894 case ETHERTYPE_IP: 895 #ifdef GATEWAY 896 if (ipflow_fastforward(m)) 897 return; 898 #endif 899 pktq = ip_pktq; 900 rps_hash = atomic_load_relaxed(ðer_pktq_rps_hash_p); 901 break; 902 903 case ETHERTYPE_ARP: 904 pktq = arp_pktq; 905 break; 906 907 case ETHERTYPE_REVARP: 908 revarpinput(m); /* XXX queue? */ 909 return; 910 #endif 911 912 #ifdef INET6 913 case ETHERTYPE_IPV6: 914 if (__predict_false(!in6_present)) 915 goto noproto; 916 #ifdef GATEWAY 917 if (ip6flow_fastforward(&m)) 918 return; 919 #endif 920 pktq = ip6_pktq; 921 rps_hash = atomic_load_relaxed(ðer_pktq_rps_hash_p); 922 break; 923 #endif 924 925 #ifdef NETATALK 926 case ETHERTYPE_ATALK: 927 pktq = at_pktq1; 928 break; 929 930 case ETHERTYPE_AARP: 931 aarpinput(ifp, m); /* XXX queue? */ 932 return; 933 #endif 934 935 #ifdef MPLS 936 case ETHERTYPE_MPLS: 937 pktq = mpls_pktq; 938 break; 939 #endif 940 941 default: 942 goto noproto; 943 } 944 945 KASSERT(pktq != NULL); 946 const uint32_t h = rps_hash ? pktq_rps_hash(&rps_hash, m) : 0; 947 if (__predict_false(!pktq_enqueue(pktq, m, h))) { 948 m_freem(m); 949 } 950 return; 951 952 drop: 953 m_freem(m); 954 if_statinc(ifp, if_iqdrops); 955 return; 956 noproto: 957 m_freem(m); 958 if_statinc(ifp, if_noproto); 959 return; 960 error: 961 m_freem(m); 962 if_statinc(ifp, if_ierrors); 963 return; 964 } 965 966 static void 967 ether_bpf_mtap(struct bpf_if *bp, struct mbuf *m, u_int direction) 968 { 969 struct ether_vlan_header evl; 970 struct m_hdr mh, md; 971 972 KASSERT(bp != NULL); 973 974 if (!vlan_has_tag(m)) { 975 bpf_mtap3(bp, m, direction); 976 return; 977 } 978 979 memcpy(&evl, mtod(m, char *), ETHER_HDR_LEN); 980 evl.evl_proto = evl.evl_encap_proto; 981 evl.evl_encap_proto = htons(ETHERTYPE_VLAN); 982 evl.evl_tag = htons(vlan_get_tag(m)); 983 984 md.mh_flags = 0; 985 md.mh_data = m->m_data + ETHER_HDR_LEN; 986 md.mh_len = m->m_len - ETHER_HDR_LEN; 987 md.mh_next = m->m_next; 988 989 mh.mh_flags = 0; 990 mh.mh_data = (char *)&evl; 991 mh.mh_len = sizeof(evl); 992 mh.mh_next = (struct mbuf *)&md; 993 994 bpf_mtap3(bp, (struct mbuf *)&mh, direction); 995 } 996 997 /* 998 * Convert Ethernet address to printable (loggable) representation. 999 */ 1000 char * 1001 ether_sprintf(const u_char *ap) 1002 { 1003 static char etherbuf[3 * ETHER_ADDR_LEN]; 1004 return ether_snprintf(etherbuf, sizeof(etherbuf), ap); 1005 } 1006 1007 char * 1008 ether_snprintf(char *buf, size_t len, const u_char *ap) 1009 { 1010 char *cp = buf; 1011 size_t i; 1012 1013 for (i = 0; i < len / 3; i++) { 1014 *cp++ = hexdigits[*ap >> 4]; 1015 *cp++ = hexdigits[*ap++ & 0xf]; 1016 *cp++ = ':'; 1017 } 1018 *--cp = '\0'; 1019 return buf; 1020 } 1021 1022 /* 1023 * Perform common duties while attaching to interface list 1024 */ 1025 void 1026 ether_ifattach(struct ifnet *ifp, const uint8_t *lla) 1027 { 1028 struct ethercom *ec = (struct ethercom *)ifp; 1029 char xnamebuf[HOOKNAMSIZ]; 1030 1031 ifp->if_type = IFT_ETHER; 1032 ifp->if_hdrlen = ETHER_HDR_LEN; 1033 ifp->if_dlt = DLT_EN10MB; 1034 ifp->if_mtu = ETHERMTU; 1035 ifp->if_output = ether_output; 1036 ifp->_if_input = ether_input; 1037 ifp->if_bpf_mtap = ether_bpf_mtap; 1038 if (ifp->if_baudrate == 0) 1039 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1040 1041 if (lla != NULL) 1042 if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla)); 1043 1044 LIST_INIT(&ec->ec_multiaddrs); 1045 SIMPLEQ_INIT(&ec->ec_vids); 1046 ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 1047 ec->ec_flags = 0; 1048 ifp->if_broadcastaddr = etherbroadcastaddr; 1049 bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header)); 1050 snprintf(xnamebuf, sizeof(xnamebuf), 1051 "%s-ether_ifdetachhooks", ifp->if_xname); 1052 ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf); 1053 #ifdef MBUFTRACE 1054 mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx"); 1055 mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx"); 1056 MOWNER_ATTACH(&ec->ec_tx_mowner); 1057 MOWNER_ATTACH(&ec->ec_rx_mowner); 1058 ifp->if_mowner = &ec->ec_tx_mowner; 1059 #endif 1060 } 1061 1062 void 1063 ether_ifdetach(struct ifnet *ifp) 1064 { 1065 struct ethercom *ec = (void *) ifp; 1066 struct ether_multi *enm; 1067 1068 IFNET_ASSERT_UNLOCKED(ifp); 1069 /* 1070 * Prevent further calls to ioctl (for example turning off 1071 * promiscuous mode from the bridge code), which eventually can 1072 * call if_init() which can cause panics because the interface 1073 * is in the process of being detached. Return device not configured 1074 * instead. 1075 */ 1076 ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *), 1077 enxio); 1078 1079 simplehook_dohooks(ec->ec_ifdetach_hooks); 1080 KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks)); 1081 simplehook_destroy(ec->ec_ifdetach_hooks); 1082 1083 bpf_detach(ifp); 1084 1085 ETHER_LOCK(ec); 1086 KASSERT(ec->ec_nvlans == 0); 1087 while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) { 1088 LIST_REMOVE(enm, enm_list); 1089 kmem_free(enm, sizeof(*enm)); 1090 ec->ec_multicnt--; 1091 } 1092 ETHER_UNLOCK(ec); 1093 1094 mutex_obj_free(ec->ec_lock); 1095 ec->ec_lock = NULL; 1096 1097 ifp->if_mowner = NULL; 1098 MOWNER_DETACH(&ec->ec_rx_mowner); 1099 MOWNER_DETACH(&ec->ec_tx_mowner); 1100 } 1101 1102 void * 1103 ether_ifdetachhook_establish(struct ifnet *ifp, 1104 void (*fn)(void *), void *arg) 1105 { 1106 struct ethercom *ec; 1107 khook_t *hk; 1108 1109 if (ifp->if_type != IFT_ETHER) 1110 return NULL; 1111 1112 ec = (struct ethercom *)ifp; 1113 hk = simplehook_establish(ec->ec_ifdetach_hooks, 1114 fn, arg); 1115 1116 return (void *)hk; 1117 } 1118 1119 void 1120 ether_ifdetachhook_disestablish(struct ifnet *ifp, 1121 void *vhook, kmutex_t *lock) 1122 { 1123 struct ethercom *ec; 1124 1125 if (vhook == NULL) 1126 return; 1127 1128 ec = (struct ethercom *)ifp; 1129 simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock); 1130 } 1131 1132 #if 0 1133 /* 1134 * This is for reference. We have a table-driven version 1135 * of the little-endian crc32 generator, which is faster 1136 * than the double-loop. 1137 */ 1138 uint32_t 1139 ether_crc32_le(const uint8_t *buf, size_t len) 1140 { 1141 uint32_t c, crc, carry; 1142 size_t i, j; 1143 1144 crc = 0xffffffffU; /* initial value */ 1145 1146 for (i = 0; i < len; i++) { 1147 c = buf[i]; 1148 for (j = 0; j < 8; j++) { 1149 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1150 crc >>= 1; 1151 c >>= 1; 1152 if (carry) 1153 crc = (crc ^ ETHER_CRC_POLY_LE); 1154 } 1155 } 1156 1157 return (crc); 1158 } 1159 #else 1160 uint32_t 1161 ether_crc32_le(const uint8_t *buf, size_t len) 1162 { 1163 static const uint32_t crctab[] = { 1164 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1165 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1166 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1167 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1168 }; 1169 uint32_t crc; 1170 size_t i; 1171 1172 crc = 0xffffffffU; /* initial value */ 1173 1174 for (i = 0; i < len; i++) { 1175 crc ^= buf[i]; 1176 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1177 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1178 } 1179 1180 return (crc); 1181 } 1182 #endif 1183 1184 uint32_t 1185 ether_crc32_be(const uint8_t *buf, size_t len) 1186 { 1187 uint32_t c, crc, carry; 1188 size_t i, j; 1189 1190 crc = 0xffffffffU; /* initial value */ 1191 1192 for (i = 0; i < len; i++) { 1193 c = buf[i]; 1194 for (j = 0; j < 8; j++) { 1195 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1196 crc <<= 1; 1197 c >>= 1; 1198 if (carry) 1199 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1200 } 1201 } 1202 1203 return (crc); 1204 } 1205 1206 #ifdef INET 1207 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] = 1208 { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }; 1209 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] = 1210 { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff }; 1211 #endif 1212 #ifdef INET6 1213 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] = 1214 { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 }; 1215 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] = 1216 { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff }; 1217 #endif 1218 1219 /* 1220 * ether_aton implementation, not using a static buffer. 1221 */ 1222 int 1223 ether_aton_r(u_char *dest, size_t len, const char *str) 1224 { 1225 const u_char *cp = (const void *)str; 1226 u_char *ep; 1227 1228 #define atox(c) (((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10)) 1229 1230 if (len < ETHER_ADDR_LEN) 1231 return ENOSPC; 1232 1233 ep = dest + ETHER_ADDR_LEN; 1234 1235 while (*cp) { 1236 if (!isxdigit(*cp)) 1237 return EINVAL; 1238 1239 *dest = atox(*cp); 1240 cp++; 1241 if (isxdigit(*cp)) { 1242 *dest = (*dest << 4) | atox(*cp); 1243 cp++; 1244 } 1245 dest++; 1246 1247 if (dest == ep) 1248 return (*cp == '\0') ? 0 : ENAMETOOLONG; 1249 1250 switch (*cp) { 1251 case ':': 1252 case '-': 1253 case '.': 1254 cp++; 1255 break; 1256 } 1257 } 1258 return ENOBUFS; 1259 } 1260 1261 /* 1262 * Convert a sockaddr into an Ethernet address or range of Ethernet 1263 * addresses. 1264 */ 1265 int 1266 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN], 1267 uint8_t addrhi[ETHER_ADDR_LEN]) 1268 { 1269 #ifdef INET 1270 const struct sockaddr_in *sin; 1271 #endif 1272 #ifdef INET6 1273 const struct sockaddr_in6 *sin6; 1274 #endif 1275 1276 switch (sa->sa_family) { 1277 1278 case AF_UNSPEC: 1279 memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN); 1280 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1281 break; 1282 1283 #ifdef INET 1284 case AF_INET: 1285 sin = satocsin(sa); 1286 if (sin->sin_addr.s_addr == INADDR_ANY) { 1287 /* 1288 * An IP address of INADDR_ANY means listen to 1289 * or stop listening to all of the Ethernet 1290 * multicast addresses used for IP. 1291 * (This is for the sake of IP multicast routers.) 1292 */ 1293 memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN); 1294 memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN); 1295 } else { 1296 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo); 1297 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1298 } 1299 break; 1300 #endif 1301 #ifdef INET6 1302 case AF_INET6: 1303 sin6 = satocsin6(sa); 1304 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1305 /* 1306 * An IP6 address of 0 means listen to or stop 1307 * listening to all of the Ethernet multicast 1308 * address used for IP6. 1309 * (This is used for multicast routers.) 1310 */ 1311 memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN); 1312 memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN); 1313 } else { 1314 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo); 1315 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1316 } 1317 break; 1318 #endif 1319 1320 default: 1321 return EAFNOSUPPORT; 1322 } 1323 return 0; 1324 } 1325 1326 /* 1327 * Add an Ethernet multicast address or range of addresses to the list for a 1328 * given interface. 1329 */ 1330 int 1331 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec) 1332 { 1333 struct ether_multi *enm, *_enm; 1334 u_char addrlo[ETHER_ADDR_LEN]; 1335 u_char addrhi[ETHER_ADDR_LEN]; 1336 int error = 0; 1337 1338 /* Allocate out of lock */ 1339 enm = kmem_alloc(sizeof(*enm), KM_SLEEP); 1340 1341 ETHER_LOCK(ec); 1342 error = ether_multiaddr(sa, addrlo, addrhi); 1343 if (error != 0) 1344 goto out; 1345 1346 /* 1347 * Verify that we have valid Ethernet multicast addresses. 1348 */ 1349 if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) { 1350 error = EINVAL; 1351 goto out; 1352 } 1353 1354 /* 1355 * See if the address range is already in the list. 1356 */ 1357 _enm = ether_lookup_multi(addrlo, addrhi, ec); 1358 if (_enm != NULL) { 1359 /* 1360 * Found it; just increment the reference count. 1361 */ 1362 ++_enm->enm_refcount; 1363 error = 0; 1364 goto out; 1365 } 1366 1367 /* 1368 * Link a new multicast record into the interface's multicast list. 1369 */ 1370 memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN); 1371 memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN); 1372 enm->enm_refcount = 1; 1373 LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list); 1374 ec->ec_multicnt++; 1375 1376 /* 1377 * Return ENETRESET to inform the driver that the list has changed 1378 * and its reception filter should be adjusted accordingly. 1379 */ 1380 error = ENETRESET; 1381 enm = NULL; 1382 1383 out: 1384 ETHER_UNLOCK(ec); 1385 if (enm != NULL) 1386 kmem_free(enm, sizeof(*enm)); 1387 return error; 1388 } 1389 1390 /* 1391 * Delete a multicast address record. 1392 */ 1393 int 1394 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec) 1395 { 1396 struct ether_multi *enm; 1397 u_char addrlo[ETHER_ADDR_LEN]; 1398 u_char addrhi[ETHER_ADDR_LEN]; 1399 int error; 1400 1401 ETHER_LOCK(ec); 1402 error = ether_multiaddr(sa, addrlo, addrhi); 1403 if (error != 0) 1404 goto error; 1405 1406 /* 1407 * Look up the address in our list. 1408 */ 1409 enm = ether_lookup_multi(addrlo, addrhi, ec); 1410 if (enm == NULL) { 1411 error = ENXIO; 1412 goto error; 1413 } 1414 if (--enm->enm_refcount != 0) { 1415 /* 1416 * Still some claims to this record. 1417 */ 1418 error = 0; 1419 goto error; 1420 } 1421 1422 /* 1423 * No remaining claims to this record; unlink and free it. 1424 */ 1425 LIST_REMOVE(enm, enm_list); 1426 ec->ec_multicnt--; 1427 ETHER_UNLOCK(ec); 1428 kmem_free(enm, sizeof(*enm)); 1429 1430 /* 1431 * Return ENETRESET to inform the driver that the list has changed 1432 * and its reception filter should be adjusted accordingly. 1433 */ 1434 return ENETRESET; 1435 1436 error: 1437 ETHER_UNLOCK(ec); 1438 return error; 1439 } 1440 1441 void 1442 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb) 1443 { 1444 ec->ec_ifflags_cb = cb; 1445 } 1446 1447 void 1448 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb) 1449 { 1450 1451 ec->ec_vlan_cb = cb; 1452 } 1453 1454 static int 1455 ether_ioctl_reinit(struct ethercom *ec) 1456 { 1457 struct ifnet *ifp = &ec->ec_if; 1458 int error; 1459 1460 KASSERTMSG(IFNET_LOCKED(ifp), "%s", ifp->if_xname); 1461 1462 switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) { 1463 case IFF_RUNNING: 1464 /* 1465 * If interface is marked down and it is running, 1466 * then stop and disable it. 1467 */ 1468 if_stop(ifp, 1); 1469 break; 1470 case IFF_UP: 1471 /* 1472 * If interface is marked up and it is stopped, then 1473 * start it. 1474 */ 1475 return if_init(ifp); 1476 case IFF_UP | IFF_RUNNING: 1477 error = 0; 1478 if (ec->ec_ifflags_cb != NULL) { 1479 error = (*ec->ec_ifflags_cb)(ec); 1480 if (error == ENETRESET) { 1481 /* 1482 * Reset the interface to pick up 1483 * changes in any other flags that 1484 * affect the hardware state. 1485 */ 1486 return if_init(ifp); 1487 } 1488 } else 1489 error = if_init(ifp); 1490 return error; 1491 case 0: 1492 break; 1493 } 1494 1495 return 0; 1496 } 1497 1498 /* 1499 * Common ioctls for Ethernet interfaces. Note, we must be 1500 * called at splnet(). 1501 */ 1502 int 1503 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1504 { 1505 struct ethercom *ec = (void *)ifp; 1506 struct eccapreq *eccr; 1507 struct ifreq *ifr = (struct ifreq *)data; 1508 struct if_laddrreq *iflr = data; 1509 const struct sockaddr_dl *sdl; 1510 static const uint8_t zero[ETHER_ADDR_LEN]; 1511 int error; 1512 1513 switch (cmd) { 1514 case SIOCINITIFADDR: 1515 { 1516 struct ifaddr *ifa = (struct ifaddr *)data; 1517 if (ifa->ifa_addr->sa_family != AF_LINK 1518 && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 1519 (IFF_UP | IFF_RUNNING)) { 1520 ifp->if_flags |= IFF_UP; 1521 if ((error = if_init(ifp)) != 0) 1522 return error; 1523 } 1524 #ifdef INET 1525 if (ifa->ifa_addr->sa_family == AF_INET) 1526 arp_ifinit(ifp, ifa); 1527 #endif 1528 return 0; 1529 } 1530 1531 case SIOCSIFMTU: 1532 { 1533 int maxmtu; 1534 1535 if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU) 1536 maxmtu = ETHERMTU_JUMBO; 1537 else 1538 maxmtu = ETHERMTU; 1539 1540 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu) 1541 return EINVAL; 1542 else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET) 1543 return error; 1544 else if (ifp->if_flags & IFF_UP) { 1545 /* Make sure the device notices the MTU change. */ 1546 return if_init(ifp); 1547 } else 1548 return 0; 1549 } 1550 1551 case SIOCSIFFLAGS: 1552 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1553 return error; 1554 return ether_ioctl_reinit(ec); 1555 case SIOCGIFFLAGS: 1556 error = ifioctl_common(ifp, cmd, data); 1557 if (error == 0) { 1558 /* Set IFF_ALLMULTI for backcompat */ 1559 ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ? 1560 IFF_ALLMULTI : 0; 1561 } 1562 return error; 1563 case SIOCGETHERCAP: 1564 eccr = (struct eccapreq *)data; 1565 eccr->eccr_capabilities = ec->ec_capabilities; 1566 eccr->eccr_capenable = ec->ec_capenable; 1567 return 0; 1568 case SIOCSETHERCAP: 1569 eccr = (struct eccapreq *)data; 1570 if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0) 1571 return EINVAL; 1572 if (eccr->eccr_capenable == ec->ec_capenable) 1573 return 0; 1574 #if 0 /* notyet */ 1575 ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE) 1576 | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE); 1577 #else 1578 ec->ec_capenable = eccr->eccr_capenable; 1579 #endif 1580 return ether_ioctl_reinit(ec); 1581 case SIOCADDMULTI: 1582 return ether_addmulti(ifreq_getaddr(cmd, ifr), ec); 1583 case SIOCDELMULTI: 1584 return ether_delmulti(ifreq_getaddr(cmd, ifr), ec); 1585 case SIOCSIFMEDIA: 1586 case SIOCGIFMEDIA: 1587 if (ec->ec_mii != NULL) 1588 return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, 1589 cmd); 1590 else if (ec->ec_ifmedia != NULL) 1591 return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd); 1592 else 1593 return ENOTTY; 1594 break; 1595 case SIOCALIFADDR: 1596 sdl = satocsdl(sstocsa(&iflr->addr)); 1597 if (sdl->sdl_family != AF_LINK) 1598 ; 1599 else if (ETHER_IS_MULTICAST(CLLADDR(sdl))) 1600 return EINVAL; 1601 else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0) 1602 return EINVAL; 1603 /*FALLTHROUGH*/ 1604 default: 1605 return ifioctl_common(ifp, cmd, data); 1606 } 1607 return 0; 1608 } 1609 1610 /* 1611 * Enable/disable passing VLAN packets if the parent interface supports it. 1612 * Return: 1613 * 0: Ok 1614 * -1: Parent interface does not support vlans 1615 * >0: Error 1616 */ 1617 int 1618 ether_enable_vlan_mtu(struct ifnet *ifp) 1619 { 1620 int error; 1621 struct ethercom *ec = (void *)ifp; 1622 1623 /* Parent does not support VLAN's */ 1624 if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0) 1625 return -1; 1626 1627 /* 1628 * Parent supports the VLAN_MTU capability, 1629 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames; 1630 * enable it. 1631 */ 1632 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1633 1634 /* Interface is down, defer for later */ 1635 if ((ifp->if_flags & IFF_UP) == 0) 1636 return 0; 1637 1638 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1639 return 0; 1640 1641 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1642 return error; 1643 } 1644 1645 int 1646 ether_disable_vlan_mtu(struct ifnet *ifp) 1647 { 1648 int error; 1649 struct ethercom *ec = (void *)ifp; 1650 1651 /* We still have VLAN's, defer for later */ 1652 if (ec->ec_nvlans != 0) 1653 return 0; 1654 1655 /* Parent does not support VLAB's, nothing to do. */ 1656 if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0) 1657 return -1; 1658 1659 /* 1660 * Disable Tx/Rx of VLAN-sized frames. 1661 */ 1662 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1663 1664 /* Interface is down, defer for later */ 1665 if ((ifp->if_flags & IFF_UP) == 0) 1666 return 0; 1667 1668 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1669 return 0; 1670 1671 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1672 return error; 1673 } 1674 1675 /* 1676 * Add and delete VLAN TAG 1677 */ 1678 int 1679 ether_add_vlantag(struct ifnet *ifp, uint16_t vtag, bool *vlanmtu_status) 1680 { 1681 struct ethercom *ec = (void *)ifp; 1682 struct vlanid_list *vidp; 1683 bool vlanmtu_enabled; 1684 uint16_t vid = EVL_VLANOFTAG(vtag); 1685 int error; 1686 1687 vlanmtu_enabled = false; 1688 1689 /* Add a vid to the list */ 1690 vidp = kmem_alloc(sizeof(*vidp), KM_SLEEP); 1691 vidp->vid = vid; 1692 1693 ETHER_LOCK(ec); 1694 ec->ec_nvlans++; 1695 SIMPLEQ_INSERT_TAIL(&ec->ec_vids, vidp, vid_list); 1696 ETHER_UNLOCK(ec); 1697 1698 if (ec->ec_nvlans == 1) { 1699 IFNET_LOCK(ifp); 1700 error = ether_enable_vlan_mtu(ifp); 1701 IFNET_UNLOCK(ifp); 1702 1703 if (error == 0) { 1704 vlanmtu_enabled = true; 1705 } else if (error != -1) { 1706 goto fail; 1707 } 1708 } 1709 1710 if (ec->ec_vlan_cb != NULL) { 1711 error = (*ec->ec_vlan_cb)(ec, vid, true); 1712 if (error != 0) 1713 goto fail; 1714 } 1715 1716 if (vlanmtu_status != NULL) 1717 *vlanmtu_status = vlanmtu_enabled; 1718 1719 return 0; 1720 fail: 1721 ETHER_LOCK(ec); 1722 ec->ec_nvlans--; 1723 SIMPLEQ_REMOVE(&ec->ec_vids, vidp, vlanid_list, vid_list); 1724 ETHER_UNLOCK(ec); 1725 1726 if (vlanmtu_enabled) { 1727 IFNET_LOCK(ifp); 1728 (void)ether_disable_vlan_mtu(ifp); 1729 IFNET_UNLOCK(ifp); 1730 } 1731 1732 kmem_free(vidp, sizeof(*vidp)); 1733 1734 return error; 1735 } 1736 1737 int 1738 ether_del_vlantag(struct ifnet *ifp, uint16_t vtag) 1739 { 1740 struct ethercom *ec = (void *)ifp; 1741 struct vlanid_list *vidp; 1742 uint16_t vid = EVL_VLANOFTAG(vtag); 1743 1744 ETHER_LOCK(ec); 1745 SIMPLEQ_FOREACH(vidp, &ec->ec_vids, vid_list) { 1746 if (vidp->vid == vid) { 1747 SIMPLEQ_REMOVE(&ec->ec_vids, vidp, 1748 vlanid_list, vid_list); 1749 ec->ec_nvlans--; 1750 break; 1751 } 1752 } 1753 ETHER_UNLOCK(ec); 1754 1755 if (vidp == NULL) 1756 return ENOENT; 1757 1758 if (ec->ec_vlan_cb != NULL) { 1759 (void)(*ec->ec_vlan_cb)(ec, vidp->vid, false); 1760 } 1761 1762 if (ec->ec_nvlans == 0) { 1763 IFNET_LOCK(ifp); 1764 (void)ether_disable_vlan_mtu(ifp); 1765 IFNET_UNLOCK(ifp); 1766 } 1767 1768 kmem_free(vidp, sizeof(*vidp)); 1769 1770 return 0; 1771 } 1772 1773 int 1774 ether_inject_vlantag(struct mbuf **mp, uint16_t etype, uint16_t tag) 1775 { 1776 static const size_t min_data_len = 1777 ETHER_MIN_LEN - ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN; 1778 /* Used to pad ethernet frames with < ETHER_MIN_LEN bytes */ 1779 static const char vlan_zero_pad_buff[ETHER_MIN_LEN] = { 0 }; 1780 1781 struct ether_vlan_header *evl; 1782 struct mbuf *m = *mp; 1783 int error; 1784 1785 error = 0; 1786 1787 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT); 1788 if (m == NULL) { 1789 error = ENOBUFS; 1790 goto out; 1791 } 1792 1793 if (m->m_len < sizeof(*evl)) { 1794 m = m_pullup(m, sizeof(*evl)); 1795 if (m == NULL) { 1796 error = ENOBUFS; 1797 goto out; 1798 } 1799 } 1800 1801 /* 1802 * Transform the Ethernet header into an 1803 * Ethernet header with 802.1Q encapsulation. 1804 */ 1805 memmove(mtod(m, void *), 1806 mtod(m, char *) + ETHER_VLAN_ENCAP_LEN, 1807 sizeof(struct ether_header)); 1808 evl = mtod(m, struct ether_vlan_header *); 1809 evl->evl_proto = evl->evl_encap_proto; 1810 evl->evl_encap_proto = htons(etype); 1811 evl->evl_tag = htons(tag); 1812 1813 /* 1814 * To cater for VLAN-aware layer 2 ethernet 1815 * switches which may need to strip the tag 1816 * before forwarding the packet, make sure 1817 * the packet+tag is at least 68 bytes long. 1818 * This is necessary because our parent will 1819 * only pad to 64 bytes (ETHER_MIN_LEN) and 1820 * some switches will not pad by themselves 1821 * after deleting a tag. 1822 */ 1823 if (m->m_pkthdr.len < min_data_len) { 1824 m_copyback(m, m->m_pkthdr.len, 1825 min_data_len - m->m_pkthdr.len, 1826 vlan_zero_pad_buff); 1827 } 1828 1829 m->m_flags &= ~M_VLANTAG; 1830 1831 out: 1832 *mp = m; 1833 return error; 1834 } 1835 1836 struct mbuf * 1837 ether_strip_vlantag(struct mbuf *m) 1838 { 1839 struct ether_vlan_header *evl; 1840 1841 if (m->m_len < sizeof(*evl) && 1842 (m = m_pullup(m, sizeof(*evl))) == NULL) { 1843 return NULL; 1844 } 1845 1846 if (m_makewritable(&m, 0, sizeof(*evl), M_DONTWAIT)) { 1847 m_freem(m); 1848 return NULL; 1849 } 1850 1851 evl = mtod(m, struct ether_vlan_header *); 1852 KASSERT(ntohs(evl->evl_encap_proto) == ETHERTYPE_VLAN); 1853 1854 vlan_set_tag(m, ntohs(evl->evl_tag)); 1855 1856 /* 1857 * Restore the original ethertype. We'll remove 1858 * the encapsulation after we've found the vlan 1859 * interface corresponding to the tag. 1860 */ 1861 evl->evl_encap_proto = evl->evl_proto; 1862 1863 /* 1864 * Remove the encapsulation header and append tag. 1865 * The original header has already been fixed up above. 1866 */ 1867 vlan_set_tag(m, ntohs(evl->evl_tag)); 1868 memmove((char *)evl + ETHER_VLAN_ENCAP_LEN, evl, 1869 offsetof(struct ether_vlan_header, evl_encap_proto)); 1870 m_adj(m, ETHER_VLAN_ENCAP_LEN); 1871 1872 return m; 1873 } 1874 1875 static int 1876 ether_multicast_sysctl(SYSCTLFN_ARGS) 1877 { 1878 struct ether_multi *enm; 1879 struct ifnet *ifp; 1880 struct ethercom *ec; 1881 int error = 0; 1882 size_t written; 1883 struct psref psref; 1884 int bound; 1885 unsigned int multicnt; 1886 struct ether_multi_sysctl *addrs; 1887 int i; 1888 1889 if (namelen != 1) 1890 return EINVAL; 1891 1892 bound = curlwp_bind(); 1893 ifp = if_get_byindex(name[0], &psref); 1894 if (ifp == NULL) { 1895 error = ENODEV; 1896 goto out; 1897 } 1898 if (ifp->if_type != IFT_ETHER) { 1899 if_put(ifp, &psref); 1900 *oldlenp = 0; 1901 goto out; 1902 } 1903 ec = (struct ethercom *)ifp; 1904 1905 if (oldp == NULL) { 1906 if_put(ifp, &psref); 1907 *oldlenp = ec->ec_multicnt * sizeof(*addrs); 1908 goto out; 1909 } 1910 1911 /* 1912 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which 1913 * is sleepable, while holding it. Copy data to a local buffer first 1914 * with the lock taken and then call sysctl_copyout without holding it. 1915 */ 1916 retry: 1917 multicnt = ec->ec_multicnt; 1918 1919 if (multicnt == 0) { 1920 if_put(ifp, &psref); 1921 *oldlenp = 0; 1922 goto out; 1923 } 1924 1925 addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP); 1926 1927 ETHER_LOCK(ec); 1928 if (multicnt != ec->ec_multicnt) { 1929 /* The number of multicast addresses has changed */ 1930 ETHER_UNLOCK(ec); 1931 kmem_free(addrs, sizeof(*addrs) * multicnt); 1932 goto retry; 1933 } 1934 1935 i = 0; 1936 LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) { 1937 struct ether_multi_sysctl *addr = &addrs[i]; 1938 addr->enm_refcount = enm->enm_refcount; 1939 memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN); 1940 memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN); 1941 i++; 1942 } 1943 ETHER_UNLOCK(ec); 1944 1945 error = 0; 1946 written = 0; 1947 for (i = 0; i < multicnt; i++) { 1948 struct ether_multi_sysctl *addr = &addrs[i]; 1949 1950 if (written + sizeof(*addr) > *oldlenp) 1951 break; 1952 error = sysctl_copyout(l, addr, oldp, sizeof(*addr)); 1953 if (error) 1954 break; 1955 written += sizeof(*addr); 1956 oldp = (char *)oldp + sizeof(*addr); 1957 } 1958 kmem_free(addrs, sizeof(*addrs) * multicnt); 1959 1960 if_put(ifp, &psref); 1961 1962 *oldlenp = written; 1963 out: 1964 curlwp_bindx(bound); 1965 return error; 1966 } 1967 1968 static void 1969 ether_sysctl_setup(struct sysctllog **clog) 1970 { 1971 const struct sysctlnode *rnode = NULL; 1972 1973 sysctl_createv(clog, 0, NULL, &rnode, 1974 CTLFLAG_PERMANENT, 1975 CTLTYPE_NODE, "ether", 1976 SYSCTL_DESCR("Ethernet-specific information"), 1977 NULL, 0, NULL, 0, 1978 CTL_NET, CTL_CREATE, CTL_EOL); 1979 1980 sysctl_createv(clog, 0, &rnode, NULL, 1981 CTLFLAG_PERMANENT, 1982 CTLTYPE_NODE, "multicast", 1983 SYSCTL_DESCR("multicast addresses"), 1984 ether_multicast_sysctl, 0, NULL, 0, 1985 CTL_CREATE, CTL_EOL); 1986 1987 sysctl_createv(clog, 0, &rnode, NULL, 1988 CTLFLAG_PERMANENT | CTLFLAG_READWRITE, 1989 CTLTYPE_STRING, "rps_hash", 1990 SYSCTL_DESCR("Interface rps hash function control"), 1991 sysctl_pktq_rps_hash_handler, 0, (void *)ðer_pktq_rps_hash_p, 1992 PKTQ_RPS_HASH_NAME_LEN, 1993 CTL_CREATE, CTL_EOL); 1994 } 1995 1996 void 1997 etherinit(void) 1998 { 1999 2000 #ifdef DIAGNOSTIC 2001 mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET); 2002 #endif 2003 ether_pktq_rps_hash_p = pktq_rps_hash_default; 2004 ether_sysctl_setup(NULL); 2005 } 2006