1 /* $NetBSD: if_ethersubr.c,v 1.283 2020/03/15 23:14:41 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 1.283 2020/03/15 23:14:41 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/rnd.h> 89 #include <sys/rndsource.h> 90 #include <sys/cpu.h> 91 #include <sys/kmem.h> 92 93 #include <net/if.h> 94 #include <net/netisr.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/agr/ieee8023_slowprotocols.h> /* XXX */ 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 #ifdef DIAGNOSTIC 168 static struct timeval bigpktppslim_last; 169 static int bigpktppslim = 2; /* XXX */ 170 static int bigpktpps_count; 171 static kmutex_t bigpktpps_lock __cacheline_aligned; 172 #endif 173 174 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = 175 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 176 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] = 177 { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 }; 178 #define senderr(e) { error = (e); goto bad;} 179 180 static int ether_output(struct ifnet *, struct mbuf *, 181 const struct sockaddr *, const struct rtentry *); 182 183 /* 184 * Ethernet output routine. 185 * Encapsulate a packet of type family for the local net. 186 * Assumes that ifp is actually pointer to ethercom structure. 187 */ 188 static int 189 ether_output(struct ifnet * const ifp0, struct mbuf * const m0, 190 const struct sockaddr * const dst, const struct rtentry *rt) 191 { 192 uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 193 uint16_t etype = 0; 194 int error = 0, hdrcmplt = 0; 195 struct mbuf *m = m0; 196 struct mbuf *mcopy = NULL; 197 struct ether_header *eh; 198 struct ifnet *ifp = ifp0; 199 #ifdef INET 200 struct arphdr *ah; 201 #endif 202 #ifdef NETATALK 203 struct at_ifaddr *aa; 204 #endif 205 206 #ifdef MBUFTRACE 207 m_claimm(m, ifp->if_mowner); 208 #endif 209 210 #if NCARP > 0 211 if (ifp->if_type == IFT_CARP) { 212 struct ifaddr *ifa; 213 int s = pserialize_read_enter(); 214 215 /* loop back if this is going to the carp interface */ 216 if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP && 217 (ifa = ifa_ifwithaddr(dst)) != NULL) { 218 if (ifa->ifa_ifp == ifp0) { 219 pserialize_read_exit(s); 220 return looutput(ifp0, m, dst, rt); 221 } 222 } 223 pserialize_read_exit(s); 224 225 ifp = ifp->if_carpdev; 226 /* ac = (struct arpcom *)ifp; */ 227 228 if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) != 229 (IFF_UP | IFF_RUNNING)) 230 senderr(ENETDOWN); 231 } 232 #endif 233 234 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 235 senderr(ENETDOWN); 236 237 switch (dst->sa_family) { 238 239 #ifdef INET 240 case AF_INET: 241 if (m->m_flags & M_BCAST) { 242 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 243 } else if (m->m_flags & M_MCAST) { 244 ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst); 245 } else { 246 error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst)); 247 if (error) 248 return (error == EWOULDBLOCK) ? 0 : error; 249 } 250 /* If broadcasting on a simplex interface, loopback a copy */ 251 if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX)) 252 mcopy = m_copypacket(m, M_DONTWAIT); 253 etype = htons(ETHERTYPE_IP); 254 break; 255 256 case AF_ARP: 257 ah = mtod(m, struct arphdr *); 258 if (m->m_flags & M_BCAST) { 259 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 260 } else { 261 void *tha = ar_tha(ah); 262 263 if (tha == NULL) { 264 /* fake with ARPHRD_IEEE1394 */ 265 m_freem(m); 266 return 0; 267 } 268 memcpy(edst, tha, sizeof(edst)); 269 } 270 271 ah->ar_hrd = htons(ARPHRD_ETHER); 272 273 switch (ntohs(ah->ar_op)) { 274 case ARPOP_REVREQUEST: 275 case ARPOP_REVREPLY: 276 etype = htons(ETHERTYPE_REVARP); 277 break; 278 279 case ARPOP_REQUEST: 280 case ARPOP_REPLY: 281 default: 282 etype = htons(ETHERTYPE_ARP); 283 } 284 break; 285 #endif 286 287 #ifdef INET6 288 case AF_INET6: 289 if (m->m_flags & M_BCAST) { 290 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 291 } else if (m->m_flags & M_MCAST) { 292 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 293 edst); 294 } else { 295 error = nd6_resolve(ifp0, rt, m, dst, edst, 296 sizeof(edst)); 297 if (error) 298 return (error == EWOULDBLOCK) ? 0 : error; 299 } 300 etype = htons(ETHERTYPE_IPV6); 301 break; 302 #endif 303 304 #ifdef NETATALK 305 case AF_APPLETALK: { 306 struct ifaddr *ifa; 307 int s; 308 309 KERNEL_LOCK(1, NULL); 310 311 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 312 KERNEL_UNLOCK_ONE(NULL); 313 return 0; 314 } 315 316 /* 317 * ifaddr is the first thing in at_ifaddr 318 */ 319 s = pserialize_read_enter(); 320 ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp); 321 if (ifa == NULL) { 322 pserialize_read_exit(s); 323 KERNEL_UNLOCK_ONE(NULL); 324 senderr(EADDRNOTAVAIL); 325 } 326 aa = (struct at_ifaddr *)ifa; 327 328 /* 329 * In the phase 2 case, we need to prepend an mbuf for the 330 * llc header. 331 */ 332 if (aa->aa_flags & AFA_PHASE2) { 333 struct llc llc; 334 335 M_PREPEND(m, sizeof(struct llc), M_DONTWAIT); 336 if (m == NULL) { 337 pserialize_read_exit(s); 338 KERNEL_UNLOCK_ONE(NULL); 339 senderr(ENOBUFS); 340 } 341 342 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 343 llc.llc_control = LLC_UI; 344 memcpy(llc.llc_snap_org_code, at_org_code, 345 sizeof(llc.llc_snap_org_code)); 346 llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK); 347 memcpy(mtod(m, void *), &llc, sizeof(struct llc)); 348 } else { 349 etype = htons(ETHERTYPE_ATALK); 350 } 351 pserialize_read_exit(s); 352 KERNEL_UNLOCK_ONE(NULL); 353 break; 354 } 355 #endif /* NETATALK */ 356 357 case pseudo_AF_HDRCMPLT: 358 hdrcmplt = 1; 359 memcpy(esrc, 360 ((const struct ether_header *)dst->sa_data)->ether_shost, 361 sizeof(esrc)); 362 /* FALLTHROUGH */ 363 364 case AF_UNSPEC: 365 memcpy(edst, 366 ((const struct ether_header *)dst->sa_data)->ether_dhost, 367 sizeof(edst)); 368 /* AF_UNSPEC doesn't swap the byte order of the ether_type. */ 369 etype = ((const struct ether_header *)dst->sa_data)->ether_type; 370 break; 371 372 default: 373 printf("%s: can't handle af%d\n", ifp->if_xname, 374 dst->sa_family); 375 senderr(EAFNOSUPPORT); 376 } 377 378 #ifdef MPLS 379 { 380 struct m_tag *mtag; 381 mtag = m_tag_find(m, PACKET_TAG_MPLS); 382 if (mtag != NULL) { 383 /* Having the tag itself indicates it's MPLS */ 384 etype = htons(ETHERTYPE_MPLS); 385 m_tag_delete(m, mtag); 386 } 387 } 388 #endif 389 390 if (mcopy) 391 (void)looutput(ifp, mcopy, dst, rt); 392 393 KASSERT((m->m_flags & M_PKTHDR) != 0); 394 395 /* 396 * If no ether type is set, this must be a 802.2 formatted packet. 397 */ 398 if (etype == 0) 399 etype = htons(m->m_pkthdr.len); 400 401 /* 402 * Add local net header. If no space in first mbuf, allocate another. 403 */ 404 M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT); 405 if (m == NULL) 406 senderr(ENOBUFS); 407 408 eh = mtod(m, struct ether_header *); 409 /* Note: etype is already in network byte order. */ 410 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 411 memcpy(eh->ether_dhost, edst, sizeof(edst)); 412 if (hdrcmplt) { 413 memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost)); 414 } else { 415 memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl), 416 sizeof(eh->ether_shost)); 417 } 418 419 #if NCARP > 0 420 if (ifp0 != ifp && ifp0->if_type == IFT_CARP) { 421 memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl), 422 sizeof(eh->ether_shost)); 423 } 424 #endif 425 426 if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0) 427 return error; 428 if (m == NULL) 429 return 0; 430 431 #if NBRIDGE > 0 432 /* 433 * Bridges require special output handling. 434 */ 435 if (ifp->if_bridge) 436 return bridge_output(ifp, m, NULL, NULL); 437 #endif 438 439 #if NCARP > 0 440 if (ifp != ifp0) 441 if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN); 442 #endif 443 444 #ifdef ALTQ 445 KERNEL_LOCK(1, NULL); 446 /* 447 * If ALTQ is enabled on the parent interface, do 448 * classification; the queueing discipline might not 449 * require classification, but might require the 450 * address family/header pointer in the pktattr. 451 */ 452 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 453 altq_etherclassify(&ifp->if_snd, m); 454 KERNEL_UNLOCK_ONE(NULL); 455 #endif 456 return ifq_enqueue(ifp, m); 457 458 bad: 459 if (m) 460 m_freem(m); 461 return error; 462 } 463 464 #ifdef ALTQ 465 /* 466 * This routine is a slight hack to allow a packet to be classified 467 * if the Ethernet headers are present. It will go away when ALTQ's 468 * classification engine understands link headers. 469 * 470 * XXX: We may need to do m_pullups here. First to ensure struct ether_header 471 * is indeed contiguous, then to read the LLC and so on. 472 */ 473 void 474 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m) 475 { 476 struct ether_header *eh; 477 struct mbuf *mtop = m; 478 uint16_t ether_type; 479 int hlen, af, hdrsize; 480 void *hdr; 481 482 KASSERT((mtop->m_flags & M_PKTHDR) != 0); 483 484 hlen = ETHER_HDR_LEN; 485 eh = mtod(m, struct ether_header *); 486 487 ether_type = htons(eh->ether_type); 488 489 if (ether_type < ETHERMTU) { 490 /* LLC/SNAP */ 491 struct llc *llc = (struct llc *)(eh + 1); 492 hlen += 8; 493 494 if (m->m_len < hlen || 495 llc->llc_dsap != LLC_SNAP_LSAP || 496 llc->llc_ssap != LLC_SNAP_LSAP || 497 llc->llc_control != LLC_UI) { 498 /* Not SNAP. */ 499 goto bad; 500 } 501 502 ether_type = htons(llc->llc_un.type_snap.ether_type); 503 } 504 505 switch (ether_type) { 506 case ETHERTYPE_IP: 507 af = AF_INET; 508 hdrsize = 20; /* sizeof(struct ip) */ 509 break; 510 511 case ETHERTYPE_IPV6: 512 af = AF_INET6; 513 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 514 break; 515 516 default: 517 af = AF_UNSPEC; 518 hdrsize = 0; 519 break; 520 } 521 522 while (m->m_len <= hlen) { 523 hlen -= m->m_len; 524 m = m->m_next; 525 if (m == NULL) 526 goto bad; 527 } 528 529 if (m->m_len < (hlen + hdrsize)) { 530 /* 531 * protocol header not in a single mbuf. 532 * We can't cope with this situation right 533 * now (but it shouldn't ever happen, really, anyhow). 534 */ 535 #ifdef DEBUG 536 printf("altq_etherclassify: headers span multiple mbufs: " 537 "%d < %d\n", m->m_len, (hlen + hdrsize)); 538 #endif 539 goto bad; 540 } 541 542 m->m_data += hlen; 543 m->m_len -= hlen; 544 545 hdr = mtod(m, void *); 546 547 if (ALTQ_NEEDS_CLASSIFY(ifq)) { 548 mtop->m_pkthdr.pattr_class = 549 (*ifq->altq_classify)(ifq->altq_clfier, m, af); 550 } 551 mtop->m_pkthdr.pattr_af = af; 552 mtop->m_pkthdr.pattr_hdr = hdr; 553 554 m->m_data -= hlen; 555 m->m_len += hlen; 556 557 return; 558 559 bad: 560 mtop->m_pkthdr.pattr_class = NULL; 561 mtop->m_pkthdr.pattr_hdr = NULL; 562 mtop->m_pkthdr.pattr_af = AF_UNSPEC; 563 } 564 #endif /* ALTQ */ 565 566 /* 567 * Process a received Ethernet packet; 568 * the packet is in the mbuf chain m with 569 * the ether header. 570 */ 571 void 572 ether_input(struct ifnet *ifp, struct mbuf *m) 573 { 574 struct ethercom *ec = (struct ethercom *) ifp; 575 pktqueue_t *pktq = NULL; 576 struct ifqueue *inq = NULL; 577 uint16_t etype; 578 struct ether_header *eh; 579 size_t ehlen; 580 static int earlypkts; 581 int isr = 0; 582 #if defined (LLC) || defined(NETATALK) 583 struct llc *l; 584 #endif 585 586 KASSERT(!cpu_intr_p()); 587 KASSERT((m->m_flags & M_PKTHDR) != 0); 588 589 if ((ifp->if_flags & IFF_UP) == 0) { 590 m_freem(m); 591 return; 592 } 593 if (m->m_len < sizeof(*eh)) { 594 m = m_pullup(m, sizeof(*eh)); 595 if (m == NULL) 596 return; 597 } 598 599 #ifdef MBUFTRACE 600 m_claimm(m, &ec->ec_rx_mowner); 601 #endif 602 eh = mtod(m, struct ether_header *); 603 etype = ntohs(eh->ether_type); 604 ehlen = sizeof(*eh); 605 606 if (__predict_false(earlypkts < 100 || !rnd_initial_entropy)) { 607 rnd_add_data(NULL, eh, ehlen, 0); 608 earlypkts++; 609 } 610 611 /* 612 * Determine if the packet is within its size limits. For MPLS the 613 * header length is variable, so we skip the check. 614 */ 615 if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len > 616 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 617 #ifdef DIAGNOSTIC 618 mutex_enter(&bigpktpps_lock); 619 if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count, 620 bigpktppslim)) { 621 printf("%s: discarding oversize frame (len=%d)\n", 622 ifp->if_xname, m->m_pkthdr.len); 623 } 624 mutex_exit(&bigpktpps_lock); 625 #endif 626 if_statinc(ifp, if_iqdrops); 627 m_freem(m); 628 return; 629 } 630 631 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 632 /* 633 * If this is not a simplex interface, drop the packet 634 * if it came from us. 635 */ 636 if ((ifp->if_flags & IFF_SIMPLEX) == 0 && 637 memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost, 638 ETHER_ADDR_LEN) == 0) { 639 m_freem(m); 640 return; 641 } 642 643 if (memcmp(etherbroadcastaddr, 644 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 645 m->m_flags |= M_BCAST; 646 else 647 m->m_flags |= M_MCAST; 648 if_statinc(ifp, if_imcasts); 649 } 650 651 /* If the CRC is still on the packet, trim it off. */ 652 if (m->m_flags & M_HASFCS) { 653 m_adj(m, -ETHER_CRC_LEN); 654 m->m_flags &= ~M_HASFCS; 655 } 656 657 if_statadd(ifp, if_ibytes, m->m_pkthdr.len); 658 659 #if NCARP > 0 660 if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) { 661 /* 662 * Clear M_PROMISC, in case the packet comes from a 663 * vlan. 664 */ 665 m->m_flags &= ~M_PROMISC; 666 if (carp_input(m, (uint8_t *)&eh->ether_shost, 667 (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0) 668 return; 669 } 670 #endif 671 672 if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 && 673 (ifp->if_flags & IFF_PROMISC) != 0 && 674 memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost, 675 ETHER_ADDR_LEN) != 0) { 676 m->m_flags |= M_PROMISC; 677 } 678 679 if ((m->m_flags & M_PROMISC) == 0) { 680 if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0) 681 return; 682 if (m == NULL) 683 return; 684 685 eh = mtod(m, struct ether_header *); 686 etype = ntohs(eh->ether_type); 687 } 688 689 #if NAGR > 0 690 if (ifp->if_agrprivate && 691 __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) { 692 m->m_flags &= ~M_PROMISC; 693 agr_input(ifp, m); 694 return; 695 } 696 #endif 697 698 /* 699 * If VLANs are configured on the interface, check to 700 * see if the device performed the decapsulation and 701 * provided us with the tag. 702 */ 703 if (ec->ec_nvlans && vlan_has_tag(m)) { 704 #if NVLAN > 0 705 /* 706 * vlan_input() will either recursively call ether_input() 707 * or drop the packet. 708 */ 709 vlan_input(ifp, m); 710 #else 711 m_freem(m); 712 #endif 713 return; 714 } 715 716 /* 717 * Handle protocols that expect to have the Ethernet header 718 * (and possibly FCS) intact. 719 */ 720 switch (etype) { 721 case ETHERTYPE_VLAN: { 722 struct ether_vlan_header *evl = (void *)eh; 723 724 /* 725 * If there is a tag of 0, then the VLAN header was probably 726 * just being used to store the priority. Extract the ether 727 * type, and if IP or IPV6, let them deal with it. 728 */ 729 if (m->m_len >= sizeof(*evl) && 730 EVL_VLANOFTAG(evl->evl_tag) == 0) { 731 etype = ntohs(evl->evl_proto); 732 ehlen = sizeof(*evl); 733 if ((m->m_flags & M_PROMISC) == 0 && 734 (etype == ETHERTYPE_IP || 735 etype == ETHERTYPE_IPV6)) 736 break; 737 } 738 739 #if NVLAN > 0 740 /* 741 * vlan_input() will either recursively call ether_input() 742 * or drop the packet. 743 */ 744 if (ec->ec_nvlans != 0) 745 vlan_input(ifp, m); 746 else 747 #endif 748 m_freem(m); 749 750 return; 751 } 752 753 #if NPPPOE > 0 754 case ETHERTYPE_PPPOEDISC: 755 pppoedisc_input(ifp, m); 756 return; 757 758 case ETHERTYPE_PPPOE: 759 pppoe_input(ifp, m); 760 return; 761 #endif 762 763 case ETHERTYPE_SLOWPROTOCOLS: { 764 uint8_t subtype; 765 766 if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) { 767 m_freem(m); 768 return; 769 } 770 771 m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype); 772 switch (subtype) { 773 #if NAGR > 0 774 case SLOWPROTOCOLS_SUBTYPE_LACP: 775 if (ifp->if_agrprivate) { 776 ieee8023ad_lacp_input(ifp, m); 777 return; 778 } 779 break; 780 781 case SLOWPROTOCOLS_SUBTYPE_MARKER: 782 if (ifp->if_agrprivate) { 783 ieee8023ad_marker_input(ifp, m); 784 return; 785 } 786 break; 787 #endif 788 789 default: 790 if (subtype == 0 || subtype > 10) { 791 /* illegal value */ 792 m_freem(m); 793 return; 794 } 795 /* unknown subtype */ 796 break; 797 } 798 } 799 /* FALLTHROUGH */ 800 default: 801 if (m->m_flags & M_PROMISC) { 802 m_freem(m); 803 return; 804 } 805 } 806 807 /* If the CRC is still on the packet, trim it off. */ 808 if (m->m_flags & M_HASFCS) { 809 m_adj(m, -ETHER_CRC_LEN); 810 m->m_flags &= ~M_HASFCS; 811 } 812 813 if (etype > ETHERMTU + sizeof(struct ether_header)) { 814 /* Strip off the Ethernet header. */ 815 m_adj(m, ehlen); 816 817 switch (etype) { 818 #ifdef INET 819 case ETHERTYPE_IP: 820 #ifdef GATEWAY 821 if (ipflow_fastforward(m)) 822 return; 823 #endif 824 pktq = ip_pktq; 825 break; 826 827 case ETHERTYPE_ARP: 828 isr = NETISR_ARP; 829 inq = &arpintrq; 830 break; 831 832 case ETHERTYPE_REVARP: 833 revarpinput(m); /* XXX queue? */ 834 return; 835 #endif 836 837 #ifdef INET6 838 case ETHERTYPE_IPV6: 839 if (__predict_false(!in6_present)) { 840 m_freem(m); 841 return; 842 } 843 #ifdef GATEWAY 844 if (ip6flow_fastforward(&m)) 845 return; 846 #endif 847 pktq = ip6_pktq; 848 break; 849 #endif 850 851 #ifdef NETATALK 852 case ETHERTYPE_ATALK: 853 isr = NETISR_ATALK; 854 inq = &atintrq1; 855 break; 856 857 case ETHERTYPE_AARP: 858 aarpinput(ifp, m); /* XXX queue? */ 859 return; 860 #endif 861 862 #ifdef MPLS 863 case ETHERTYPE_MPLS: 864 isr = NETISR_MPLS; 865 inq = &mplsintrq; 866 break; 867 #endif 868 869 default: 870 m_freem(m); 871 return; 872 } 873 } else { 874 KASSERT(ehlen == sizeof(*eh)); 875 #if defined (LLC) || defined (NETATALK) 876 if (m->m_len < sizeof(*eh) + sizeof(struct llc)) { 877 goto dropanyway; 878 } 879 l = (struct llc *)(eh+1); 880 881 switch (l->llc_dsap) { 882 #ifdef NETATALK 883 case LLC_SNAP_LSAP: 884 switch (l->llc_control) { 885 case LLC_UI: 886 if (l->llc_ssap != LLC_SNAP_LSAP) { 887 goto dropanyway; 888 } 889 890 if (memcmp(&(l->llc_snap_org_code)[0], 891 at_org_code, sizeof(at_org_code)) == 0 && 892 ntohs(l->llc_snap_ether_type) == 893 ETHERTYPE_ATALK) { 894 inq = &atintrq2; 895 m_adj(m, sizeof(struct ether_header) 896 + sizeof(struct llc)); 897 isr = NETISR_ATALK; 898 break; 899 } 900 901 if (memcmp(&(l->llc_snap_org_code)[0], 902 aarp_org_code, 903 sizeof(aarp_org_code)) == 0 && 904 ntohs(l->llc_snap_ether_type) == 905 ETHERTYPE_AARP) { 906 m_adj(m, sizeof(struct ether_header) 907 + sizeof(struct llc)); 908 aarpinput(ifp, m); /* XXX queue? */ 909 return; 910 } 911 912 default: 913 goto dropanyway; 914 } 915 break; 916 #endif 917 dropanyway: 918 default: 919 m_freem(m); 920 return; 921 } 922 #else /* LLC || NETATALK */ 923 m_freem(m); 924 return; 925 #endif /* LLC || NETATALK */ 926 } 927 928 if (__predict_true(pktq)) { 929 #ifdef NET_MPSAFE 930 const u_int h = curcpu()->ci_index; 931 #else 932 const uint32_t h = pktq_rps_hash(m); 933 #endif 934 if (__predict_false(!pktq_enqueue(pktq, m, h))) { 935 m_freem(m); 936 } 937 return; 938 } 939 940 if (__predict_false(!inq)) { 941 /* Should not happen. */ 942 m_freem(m); 943 return; 944 } 945 946 IFQ_LOCK(inq); 947 if (IF_QFULL(inq)) { 948 IF_DROP(inq); 949 IFQ_UNLOCK(inq); 950 m_freem(m); 951 } else { 952 IF_ENQUEUE(inq, m); 953 IFQ_UNLOCK(inq); 954 schednetisr(isr); 955 } 956 } 957 958 /* 959 * Convert Ethernet address to printable (loggable) representation. 960 */ 961 char * 962 ether_sprintf(const u_char *ap) 963 { 964 static char etherbuf[3 * ETHER_ADDR_LEN]; 965 return ether_snprintf(etherbuf, sizeof(etherbuf), ap); 966 } 967 968 char * 969 ether_snprintf(char *buf, size_t len, const u_char *ap) 970 { 971 char *cp = buf; 972 size_t i; 973 974 for (i = 0; i < len / 3; i++) { 975 *cp++ = hexdigits[*ap >> 4]; 976 *cp++ = hexdigits[*ap++ & 0xf]; 977 *cp++ = ':'; 978 } 979 *--cp = '\0'; 980 return buf; 981 } 982 983 /* 984 * Perform common duties while attaching to interface list 985 */ 986 void 987 ether_ifattach(struct ifnet *ifp, const uint8_t *lla) 988 { 989 struct ethercom *ec = (struct ethercom *)ifp; 990 991 ifp->if_type = IFT_ETHER; 992 ifp->if_hdrlen = ETHER_HDR_LEN; 993 ifp->if_dlt = DLT_EN10MB; 994 ifp->if_mtu = ETHERMTU; 995 ifp->if_output = ether_output; 996 ifp->_if_input = ether_input; 997 if (ifp->if_baudrate == 0) 998 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 999 1000 if (lla != NULL) 1001 if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla)); 1002 1003 LIST_INIT(&ec->ec_multiaddrs); 1004 SIMPLEQ_INIT(&ec->ec_vids); 1005 ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 1006 ec->ec_flags = 0; 1007 ifp->if_broadcastaddr = etherbroadcastaddr; 1008 bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header)); 1009 #ifdef MBUFTRACE 1010 mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx"); 1011 mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx"); 1012 MOWNER_ATTACH(&ec->ec_tx_mowner); 1013 MOWNER_ATTACH(&ec->ec_rx_mowner); 1014 ifp->if_mowner = &ec->ec_tx_mowner; 1015 #endif 1016 } 1017 1018 void 1019 ether_ifdetach(struct ifnet *ifp) 1020 { 1021 struct ethercom *ec = (void *) ifp; 1022 struct ether_multi *enm; 1023 1024 IFNET_ASSERT_UNLOCKED(ifp); 1025 /* 1026 * Prevent further calls to ioctl (for example turning off 1027 * promiscuous mode from the bridge code), which eventually can 1028 * call if_init() which can cause panics because the interface 1029 * is in the process of being detached. Return device not configured 1030 * instead. 1031 */ 1032 ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *), 1033 enxio); 1034 1035 #if NBRIDGE > 0 1036 if (ifp->if_bridge) 1037 bridge_ifdetach(ifp); 1038 #endif 1039 bpf_detach(ifp); 1040 #if NVLAN > 0 1041 if (ec->ec_nvlans) 1042 vlan_ifdetach(ifp); 1043 #endif 1044 1045 ETHER_LOCK(ec); 1046 KASSERT(ec->ec_nvlans == 0); 1047 while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) { 1048 LIST_REMOVE(enm, enm_list); 1049 kmem_free(enm, sizeof(*enm)); 1050 ec->ec_multicnt--; 1051 } 1052 ETHER_UNLOCK(ec); 1053 1054 mutex_obj_free(ec->ec_lock); 1055 ec->ec_lock = NULL; 1056 1057 ifp->if_mowner = NULL; 1058 MOWNER_DETACH(&ec->ec_rx_mowner); 1059 MOWNER_DETACH(&ec->ec_tx_mowner); 1060 } 1061 1062 #if 0 1063 /* 1064 * This is for reference. We have a table-driven version 1065 * of the little-endian crc32 generator, which is faster 1066 * than the double-loop. 1067 */ 1068 uint32_t 1069 ether_crc32_le(const uint8_t *buf, size_t len) 1070 { 1071 uint32_t c, crc, carry; 1072 size_t i, j; 1073 1074 crc = 0xffffffffU; /* initial value */ 1075 1076 for (i = 0; i < len; i++) { 1077 c = buf[i]; 1078 for (j = 0; j < 8; j++) { 1079 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1080 crc >>= 1; 1081 c >>= 1; 1082 if (carry) 1083 crc = (crc ^ ETHER_CRC_POLY_LE); 1084 } 1085 } 1086 1087 return (crc); 1088 } 1089 #else 1090 uint32_t 1091 ether_crc32_le(const uint8_t *buf, size_t len) 1092 { 1093 static const uint32_t crctab[] = { 1094 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1095 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1096 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1097 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1098 }; 1099 uint32_t crc; 1100 size_t i; 1101 1102 crc = 0xffffffffU; /* initial value */ 1103 1104 for (i = 0; i < len; i++) { 1105 crc ^= buf[i]; 1106 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1107 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1108 } 1109 1110 return (crc); 1111 } 1112 #endif 1113 1114 uint32_t 1115 ether_crc32_be(const uint8_t *buf, size_t len) 1116 { 1117 uint32_t c, crc, carry; 1118 size_t i, j; 1119 1120 crc = 0xffffffffU; /* initial value */ 1121 1122 for (i = 0; i < len; i++) { 1123 c = buf[i]; 1124 for (j = 0; j < 8; j++) { 1125 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1126 crc <<= 1; 1127 c >>= 1; 1128 if (carry) 1129 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1130 } 1131 } 1132 1133 return (crc); 1134 } 1135 1136 #ifdef INET 1137 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] = 1138 { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }; 1139 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] = 1140 { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff }; 1141 #endif 1142 #ifdef INET6 1143 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] = 1144 { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 }; 1145 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] = 1146 { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff }; 1147 #endif 1148 1149 /* 1150 * ether_aton implementation, not using a static buffer. 1151 */ 1152 int 1153 ether_aton_r(u_char *dest, size_t len, const char *str) 1154 { 1155 const u_char *cp = (const void *)str; 1156 u_char *ep; 1157 1158 #define atox(c) (((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10)) 1159 1160 if (len < ETHER_ADDR_LEN) 1161 return ENOSPC; 1162 1163 ep = dest + ETHER_ADDR_LEN; 1164 1165 while (*cp) { 1166 if (!isxdigit(*cp)) 1167 return EINVAL; 1168 1169 *dest = atox(*cp); 1170 cp++; 1171 if (isxdigit(*cp)) { 1172 *dest = (*dest << 4) | atox(*cp); 1173 cp++; 1174 } 1175 dest++; 1176 1177 if (dest == ep) 1178 return (*cp == '\0') ? 0 : ENAMETOOLONG; 1179 1180 switch (*cp) { 1181 case ':': 1182 case '-': 1183 case '.': 1184 cp++; 1185 break; 1186 } 1187 } 1188 return ENOBUFS; 1189 } 1190 1191 /* 1192 * Convert a sockaddr into an Ethernet address or range of Ethernet 1193 * addresses. 1194 */ 1195 int 1196 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN], 1197 uint8_t addrhi[ETHER_ADDR_LEN]) 1198 { 1199 #ifdef INET 1200 const struct sockaddr_in *sin; 1201 #endif 1202 #ifdef INET6 1203 const struct sockaddr_in6 *sin6; 1204 #endif 1205 1206 switch (sa->sa_family) { 1207 1208 case AF_UNSPEC: 1209 memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN); 1210 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1211 break; 1212 1213 #ifdef INET 1214 case AF_INET: 1215 sin = satocsin(sa); 1216 if (sin->sin_addr.s_addr == INADDR_ANY) { 1217 /* 1218 * An IP address of INADDR_ANY means listen to 1219 * or stop listening to all of the Ethernet 1220 * multicast addresses used for IP. 1221 * (This is for the sake of IP multicast routers.) 1222 */ 1223 memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN); 1224 memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN); 1225 } else { 1226 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo); 1227 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1228 } 1229 break; 1230 #endif 1231 #ifdef INET6 1232 case AF_INET6: 1233 sin6 = satocsin6(sa); 1234 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1235 /* 1236 * An IP6 address of 0 means listen to or stop 1237 * listening to all of the Ethernet multicast 1238 * address used for IP6. 1239 * (This is used for multicast routers.) 1240 */ 1241 memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN); 1242 memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN); 1243 } else { 1244 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo); 1245 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1246 } 1247 break; 1248 #endif 1249 1250 default: 1251 return EAFNOSUPPORT; 1252 } 1253 return 0; 1254 } 1255 1256 /* 1257 * Add an Ethernet multicast address or range of addresses to the list for a 1258 * given interface. 1259 */ 1260 int 1261 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec) 1262 { 1263 struct ether_multi *enm, *_enm; 1264 u_char addrlo[ETHER_ADDR_LEN]; 1265 u_char addrhi[ETHER_ADDR_LEN]; 1266 int error = 0; 1267 1268 /* Allocate out of lock */ 1269 enm = kmem_alloc(sizeof(*enm), KM_SLEEP); 1270 1271 ETHER_LOCK(ec); 1272 error = ether_multiaddr(sa, addrlo, addrhi); 1273 if (error != 0) 1274 goto out; 1275 1276 /* 1277 * Verify that we have valid Ethernet multicast addresses. 1278 */ 1279 if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) { 1280 error = EINVAL; 1281 goto out; 1282 } 1283 1284 /* 1285 * See if the address range is already in the list. 1286 */ 1287 _enm = ether_lookup_multi(addrlo, addrhi, ec); 1288 if (_enm != NULL) { 1289 /* 1290 * Found it; just increment the reference count. 1291 */ 1292 ++_enm->enm_refcount; 1293 error = 0; 1294 goto out; 1295 } 1296 1297 /* 1298 * Link a new multicast record into the interface's multicast list. 1299 */ 1300 memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN); 1301 memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN); 1302 enm->enm_refcount = 1; 1303 LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list); 1304 ec->ec_multicnt++; 1305 1306 /* 1307 * Return ENETRESET to inform the driver that the list has changed 1308 * and its reception filter should be adjusted accordingly. 1309 */ 1310 error = ENETRESET; 1311 enm = NULL; 1312 1313 out: 1314 ETHER_UNLOCK(ec); 1315 if (enm != NULL) 1316 kmem_free(enm, sizeof(*enm)); 1317 return error; 1318 } 1319 1320 /* 1321 * Delete a multicast address record. 1322 */ 1323 int 1324 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec) 1325 { 1326 struct ether_multi *enm; 1327 u_char addrlo[ETHER_ADDR_LEN]; 1328 u_char addrhi[ETHER_ADDR_LEN]; 1329 int error; 1330 1331 ETHER_LOCK(ec); 1332 error = ether_multiaddr(sa, addrlo, addrhi); 1333 if (error != 0) 1334 goto error; 1335 1336 /* 1337 * Look up the address in our list. 1338 */ 1339 enm = ether_lookup_multi(addrlo, addrhi, ec); 1340 if (enm == NULL) { 1341 error = ENXIO; 1342 goto error; 1343 } 1344 if (--enm->enm_refcount != 0) { 1345 /* 1346 * Still some claims to this record. 1347 */ 1348 error = 0; 1349 goto error; 1350 } 1351 1352 /* 1353 * No remaining claims to this record; unlink and free it. 1354 */ 1355 LIST_REMOVE(enm, enm_list); 1356 ec->ec_multicnt--; 1357 ETHER_UNLOCK(ec); 1358 kmem_free(enm, sizeof(*enm)); 1359 1360 /* 1361 * Return ENETRESET to inform the driver that the list has changed 1362 * and its reception filter should be adjusted accordingly. 1363 */ 1364 return ENETRESET; 1365 1366 error: 1367 ETHER_UNLOCK(ec); 1368 return error; 1369 } 1370 1371 void 1372 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb) 1373 { 1374 ec->ec_ifflags_cb = cb; 1375 } 1376 1377 void 1378 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb) 1379 { 1380 1381 ec->ec_vlan_cb = cb; 1382 } 1383 1384 static int 1385 ether_ioctl_reinit(struct ethercom *ec) 1386 { 1387 struct ifnet *ifp = &ec->ec_if; 1388 int error; 1389 1390 switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) { 1391 case IFF_RUNNING: 1392 /* 1393 * If interface is marked down and it is running, 1394 * then stop and disable it. 1395 */ 1396 (*ifp->if_stop)(ifp, 1); 1397 break; 1398 case IFF_UP: 1399 /* 1400 * If interface is marked up and it is stopped, then 1401 * start it. 1402 */ 1403 return (*ifp->if_init)(ifp); 1404 case IFF_UP | IFF_RUNNING: 1405 error = 0; 1406 if (ec->ec_ifflags_cb != NULL) { 1407 error = (*ec->ec_ifflags_cb)(ec); 1408 if (error == ENETRESET) { 1409 /* 1410 * Reset the interface to pick up 1411 * changes in any other flags that 1412 * affect the hardware state. 1413 */ 1414 return (*ifp->if_init)(ifp); 1415 } 1416 } else 1417 error = (*ifp->if_init)(ifp); 1418 return error; 1419 case 0: 1420 break; 1421 } 1422 1423 return 0; 1424 } 1425 1426 /* 1427 * Common ioctls for Ethernet interfaces. Note, we must be 1428 * called at splnet(). 1429 */ 1430 int 1431 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1432 { 1433 struct ethercom *ec = (void *)ifp; 1434 struct eccapreq *eccr; 1435 struct ifreq *ifr = (struct ifreq *)data; 1436 struct if_laddrreq *iflr = data; 1437 const struct sockaddr_dl *sdl; 1438 static const uint8_t zero[ETHER_ADDR_LEN]; 1439 int error; 1440 1441 switch (cmd) { 1442 case SIOCINITIFADDR: 1443 { 1444 struct ifaddr *ifa = (struct ifaddr *)data; 1445 if (ifa->ifa_addr->sa_family != AF_LINK 1446 && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 1447 (IFF_UP | IFF_RUNNING)) { 1448 ifp->if_flags |= IFF_UP; 1449 if ((error = (*ifp->if_init)(ifp)) != 0) 1450 return error; 1451 } 1452 #ifdef INET 1453 if (ifa->ifa_addr->sa_family == AF_INET) 1454 arp_ifinit(ifp, ifa); 1455 #endif 1456 return 0; 1457 } 1458 1459 case SIOCSIFMTU: 1460 { 1461 int maxmtu; 1462 1463 if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU) 1464 maxmtu = ETHERMTU_JUMBO; 1465 else 1466 maxmtu = ETHERMTU; 1467 1468 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu) 1469 return EINVAL; 1470 else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET) 1471 return error; 1472 else if (ifp->if_flags & IFF_UP) { 1473 /* Make sure the device notices the MTU change. */ 1474 return (*ifp->if_init)(ifp); 1475 } else 1476 return 0; 1477 } 1478 1479 case SIOCSIFFLAGS: 1480 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1481 return error; 1482 return ether_ioctl_reinit(ec); 1483 case SIOCGIFFLAGS: 1484 error = ifioctl_common(ifp, cmd, data); 1485 if (error == 0) { 1486 /* Set IFF_ALLMULTI for backcompat */ 1487 ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ? 1488 IFF_ALLMULTI : 0; 1489 } 1490 return error; 1491 case SIOCGETHERCAP: 1492 eccr = (struct eccapreq *)data; 1493 eccr->eccr_capabilities = ec->ec_capabilities; 1494 eccr->eccr_capenable = ec->ec_capenable; 1495 return 0; 1496 case SIOCSETHERCAP: 1497 eccr = (struct eccapreq *)data; 1498 if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0) 1499 return EINVAL; 1500 if (eccr->eccr_capenable == ec->ec_capenable) 1501 return 0; 1502 #if 0 /* notyet */ 1503 ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE) 1504 | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE); 1505 #else 1506 ec->ec_capenable = eccr->eccr_capenable; 1507 #endif 1508 return ether_ioctl_reinit(ec); 1509 case SIOCADDMULTI: 1510 return ether_addmulti(ifreq_getaddr(cmd, ifr), ec); 1511 case SIOCDELMULTI: 1512 return ether_delmulti(ifreq_getaddr(cmd, ifr), ec); 1513 case SIOCSIFMEDIA: 1514 case SIOCGIFMEDIA: 1515 if (ec->ec_mii != NULL) 1516 return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, 1517 cmd); 1518 else if (ec->ec_ifmedia != NULL) 1519 return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd); 1520 else 1521 return ENOTTY; 1522 break; 1523 case SIOCALIFADDR: 1524 sdl = satocsdl(sstocsa(&iflr->addr)); 1525 if (sdl->sdl_family != AF_LINK) 1526 ; 1527 else if (ETHER_IS_MULTICAST(CLLADDR(sdl))) 1528 return EINVAL; 1529 else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0) 1530 return EINVAL; 1531 /*FALLTHROUGH*/ 1532 default: 1533 return ifioctl_common(ifp, cmd, data); 1534 } 1535 return 0; 1536 } 1537 1538 /* 1539 * Enable/disable passing VLAN packets if the parent interface supports it. 1540 * Return: 1541 * 0: Ok 1542 * -1: Parent interface does not support vlans 1543 * >0: Error 1544 */ 1545 int 1546 ether_enable_vlan_mtu(struct ifnet *ifp) 1547 { 1548 int error; 1549 struct ethercom *ec = (void *)ifp; 1550 1551 /* Parent does not support VLAN's */ 1552 if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0) 1553 return -1; 1554 1555 /* 1556 * Parent supports the VLAN_MTU capability, 1557 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames; 1558 * enable it. 1559 */ 1560 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1561 1562 /* Interface is down, defer for later */ 1563 if ((ifp->if_flags & IFF_UP) == 0) 1564 return 0; 1565 1566 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1567 return 0; 1568 1569 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1570 return error; 1571 } 1572 1573 int 1574 ether_disable_vlan_mtu(struct ifnet *ifp) 1575 { 1576 int error; 1577 struct ethercom *ec = (void *)ifp; 1578 1579 /* We still have VLAN's, defer for later */ 1580 if (ec->ec_nvlans != 0) 1581 return 0; 1582 1583 /* Parent does not support VLAB's, nothing to do. */ 1584 if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0) 1585 return -1; 1586 1587 /* 1588 * Disable Tx/Rx of VLAN-sized frames. 1589 */ 1590 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1591 1592 /* Interface is down, defer for later */ 1593 if ((ifp->if_flags & IFF_UP) == 0) 1594 return 0; 1595 1596 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1597 return 0; 1598 1599 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1600 return error; 1601 } 1602 1603 static int 1604 ether_multicast_sysctl(SYSCTLFN_ARGS) 1605 { 1606 struct ether_multi *enm; 1607 struct ifnet *ifp; 1608 struct ethercom *ec; 1609 int error = 0; 1610 size_t written; 1611 struct psref psref; 1612 int bound; 1613 unsigned int multicnt; 1614 struct ether_multi_sysctl *addrs; 1615 int i; 1616 1617 if (namelen != 1) 1618 return EINVAL; 1619 1620 bound = curlwp_bind(); 1621 ifp = if_get_byindex(name[0], &psref); 1622 if (ifp == NULL) { 1623 error = ENODEV; 1624 goto out; 1625 } 1626 if (ifp->if_type != IFT_ETHER) { 1627 if_put(ifp, &psref); 1628 *oldlenp = 0; 1629 goto out; 1630 } 1631 ec = (struct ethercom *)ifp; 1632 1633 if (oldp == NULL) { 1634 if_put(ifp, &psref); 1635 *oldlenp = ec->ec_multicnt * sizeof(*addrs); 1636 goto out; 1637 } 1638 1639 /* 1640 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which 1641 * is sleepable, while holding it. Copy data to a local buffer first 1642 * with the lock taken and then call sysctl_copyout without holding it. 1643 */ 1644 retry: 1645 multicnt = ec->ec_multicnt; 1646 1647 if (multicnt == 0) { 1648 if_put(ifp, &psref); 1649 *oldlenp = 0; 1650 goto out; 1651 } 1652 1653 addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP); 1654 1655 ETHER_LOCK(ec); 1656 if (multicnt != ec->ec_multicnt) { 1657 /* The number of multicast addresses has changed */ 1658 ETHER_UNLOCK(ec); 1659 kmem_free(addrs, sizeof(*addrs) * multicnt); 1660 goto retry; 1661 } 1662 1663 i = 0; 1664 LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) { 1665 struct ether_multi_sysctl *addr = &addrs[i]; 1666 addr->enm_refcount = enm->enm_refcount; 1667 memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN); 1668 memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN); 1669 i++; 1670 } 1671 ETHER_UNLOCK(ec); 1672 1673 error = 0; 1674 written = 0; 1675 for (i = 0; i < multicnt; i++) { 1676 struct ether_multi_sysctl *addr = &addrs[i]; 1677 1678 if (written + sizeof(*addr) > *oldlenp) 1679 break; 1680 error = sysctl_copyout(l, addr, oldp, sizeof(*addr)); 1681 if (error) 1682 break; 1683 written += sizeof(*addr); 1684 oldp = (char *)oldp + sizeof(*addr); 1685 } 1686 kmem_free(addrs, sizeof(*addrs) * multicnt); 1687 1688 if_put(ifp, &psref); 1689 1690 *oldlenp = written; 1691 out: 1692 curlwp_bindx(bound); 1693 return error; 1694 } 1695 1696 static void 1697 ether_sysctl_setup(struct sysctllog **clog) 1698 { 1699 const struct sysctlnode *rnode = NULL; 1700 1701 sysctl_createv(clog, 0, NULL, &rnode, 1702 CTLFLAG_PERMANENT, 1703 CTLTYPE_NODE, "ether", 1704 SYSCTL_DESCR("Ethernet-specific information"), 1705 NULL, 0, NULL, 0, 1706 CTL_NET, CTL_CREATE, CTL_EOL); 1707 1708 sysctl_createv(clog, 0, &rnode, NULL, 1709 CTLFLAG_PERMANENT, 1710 CTLTYPE_NODE, "multicast", 1711 SYSCTL_DESCR("multicast addresses"), 1712 ether_multicast_sysctl, 0, NULL, 0, 1713 CTL_CREATE, CTL_EOL); 1714 } 1715 1716 void 1717 etherinit(void) 1718 { 1719 1720 #ifdef DIAGNOSTIC 1721 mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET); 1722 #endif 1723 ether_sysctl_setup(NULL); 1724 } 1725