if_ethersubr.c revision 1.290
1 /* $NetBSD: if_ethersubr.c,v 1.290 2021/02/13 07:28:04 roy 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 1.290 2021/02/13 07:28:04 roy 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 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_statinc(ifp, if_oerrors); 460 if (m) 461 m_freem(m); 462 return error; 463 } 464 465 #ifdef ALTQ 466 /* 467 * This routine is a slight hack to allow a packet to be classified 468 * if the Ethernet headers are present. It will go away when ALTQ's 469 * classification engine understands link headers. 470 * 471 * XXX: We may need to do m_pullups here. First to ensure struct ether_header 472 * is indeed contiguous, then to read the LLC and so on. 473 */ 474 void 475 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m) 476 { 477 struct ether_header *eh; 478 struct mbuf *mtop = m; 479 uint16_t ether_type; 480 int hlen, af, hdrsize; 481 void *hdr; 482 483 KASSERT((mtop->m_flags & M_PKTHDR) != 0); 484 485 hlen = ETHER_HDR_LEN; 486 eh = mtod(m, struct ether_header *); 487 488 ether_type = htons(eh->ether_type); 489 490 if (ether_type < ETHERMTU) { 491 /* LLC/SNAP */ 492 struct llc *llc = (struct llc *)(eh + 1); 493 hlen += 8; 494 495 if (m->m_len < hlen || 496 llc->llc_dsap != LLC_SNAP_LSAP || 497 llc->llc_ssap != LLC_SNAP_LSAP || 498 llc->llc_control != LLC_UI) { 499 /* Not SNAP. */ 500 goto bad; 501 } 502 503 ether_type = htons(llc->llc_un.type_snap.ether_type); 504 } 505 506 switch (ether_type) { 507 case ETHERTYPE_IP: 508 af = AF_INET; 509 hdrsize = 20; /* sizeof(struct ip) */ 510 break; 511 512 case ETHERTYPE_IPV6: 513 af = AF_INET6; 514 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 515 break; 516 517 default: 518 af = AF_UNSPEC; 519 hdrsize = 0; 520 break; 521 } 522 523 while (m->m_len <= hlen) { 524 hlen -= m->m_len; 525 m = m->m_next; 526 if (m == NULL) 527 goto bad; 528 } 529 530 if (m->m_len < (hlen + hdrsize)) { 531 /* 532 * protocol header not in a single mbuf. 533 * We can't cope with this situation right 534 * now (but it shouldn't ever happen, really, anyhow). 535 */ 536 #ifdef DEBUG 537 printf("altq_etherclassify: headers span multiple mbufs: " 538 "%d < %d\n", m->m_len, (hlen + hdrsize)); 539 #endif 540 goto bad; 541 } 542 543 m->m_data += hlen; 544 m->m_len -= hlen; 545 546 hdr = mtod(m, void *); 547 548 if (ALTQ_NEEDS_CLASSIFY(ifq)) { 549 mtop->m_pkthdr.pattr_class = 550 (*ifq->altq_classify)(ifq->altq_clfier, m, af); 551 } 552 mtop->m_pkthdr.pattr_af = af; 553 mtop->m_pkthdr.pattr_hdr = hdr; 554 555 m->m_data -= hlen; 556 m->m_len += hlen; 557 558 return; 559 560 bad: 561 mtop->m_pkthdr.pattr_class = NULL; 562 mtop->m_pkthdr.pattr_hdr = NULL; 563 mtop->m_pkthdr.pattr_af = AF_UNSPEC; 564 } 565 #endif /* ALTQ */ 566 567 #if defined (LLC) || defined (NETATALK) 568 static void 569 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh) 570 { 571 struct ifqueue *inq = NULL; 572 int isr = 0; 573 struct llc *l; 574 575 if (m->m_len < sizeof(*eh) + sizeof(struct llc)) 576 goto drop; 577 578 l = (struct llc *)(eh+1); 579 switch (l->llc_dsap) { 580 #ifdef NETATALK 581 case LLC_SNAP_LSAP: 582 switch (l->llc_control) { 583 case LLC_UI: 584 if (l->llc_ssap != LLC_SNAP_LSAP) 585 goto drop; 586 587 if (memcmp(&(l->llc_snap_org_code)[0], 588 at_org_code, sizeof(at_org_code)) == 0 && 589 ntohs(l->llc_snap_ether_type) == 590 ETHERTYPE_ATALK) { 591 inq = &atintrq2; 592 m_adj(m, sizeof(struct ether_header) 593 + sizeof(struct llc)); 594 isr = NETISR_ATALK; 595 break; 596 } 597 598 if (memcmp(&(l->llc_snap_org_code)[0], 599 aarp_org_code, 600 sizeof(aarp_org_code)) == 0 && 601 ntohs(l->llc_snap_ether_type) == 602 ETHERTYPE_AARP) { 603 m_adj(m, sizeof(struct ether_header) 604 + sizeof(struct llc)); 605 aarpinput(ifp, m); /* XXX queue? */ 606 return; 607 } 608 609 default: 610 goto drop; 611 } 612 break; 613 #endif 614 default: 615 goto drop; 616 } 617 618 KASSERT(inq != NULL); 619 IFQ_ENQUEUE_ISR(inq, m, isr); 620 return; 621 622 drop: 623 m_freem(m); 624 if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */ 625 return; 626 } 627 #endif /* defined (LLC) || defined (NETATALK) */ 628 629 /* 630 * Process a received Ethernet packet; 631 * the packet is in the mbuf chain m with 632 * the ether header. 633 */ 634 void 635 ether_input(struct ifnet *ifp, struct mbuf *m) 636 { 637 struct ethercom *ec = (struct ethercom *) ifp; 638 pktqueue_t *pktq = NULL; 639 struct ifqueue *inq = NULL; 640 uint16_t etype; 641 struct ether_header *eh; 642 size_t ehlen; 643 static int earlypkts; 644 int isr = 0; 645 646 KASSERT(!cpu_intr_p()); 647 KASSERT((m->m_flags & M_PKTHDR) != 0); 648 649 if ((ifp->if_flags & IFF_UP) == 0) 650 goto drop; 651 652 /* If the Ethernet header is not aligned, slurp it up into a new 653 * mbuf with space for link headers, in the event we forward 654 * it. Otherwise, if it is aligned, make sure the entire 655 * base Ethernet header is in the first mbuf of the chain. 656 */ 657 if (ETHER_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 658 if ((m = m_copyup(m, sizeof(*eh), 659 (max_linkhdr + 3) & ~3)) == NULL) 660 goto dropped; 661 } else if (__predict_false(m->m_len < sizeof(*eh))) { 662 if ((m = m_pullup(m, sizeof(*eh))) == NULL) 663 goto dropped; 664 } 665 666 #ifdef MBUFTRACE 667 m_claimm(m, &ec->ec_rx_mowner); 668 #endif 669 eh = mtod(m, struct ether_header *); 670 etype = ntohs(eh->ether_type); 671 ehlen = sizeof(*eh); 672 673 if (__predict_false(earlypkts < 100 || 674 entropy_epoch() == (unsigned)-1)) { 675 rnd_add_data(NULL, eh, ehlen, 0); 676 earlypkts++; 677 } 678 679 /* 680 * Determine if the packet is within its size limits. For MPLS the 681 * header length is variable, so we skip the check. 682 */ 683 if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len > 684 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 685 #ifdef DIAGNOSTIC 686 mutex_enter(&bigpktpps_lock); 687 if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count, 688 bigpktppslim)) { 689 printf("%s: discarding oversize frame (len=%d)\n", 690 ifp->if_xname, m->m_pkthdr.len); 691 } 692 mutex_exit(&bigpktpps_lock); 693 #endif 694 goto drop; 695 } 696 697 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 698 /* 699 * If this is not a simplex interface, drop the packet 700 * if it came from us. 701 */ 702 if ((ifp->if_flags & IFF_SIMPLEX) == 0 && 703 memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost, 704 ETHER_ADDR_LEN) == 0) { 705 goto drop; 706 } 707 708 if (memcmp(etherbroadcastaddr, 709 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 710 m->m_flags |= M_BCAST; 711 else 712 m->m_flags |= M_MCAST; 713 if_statinc(ifp, if_imcasts); 714 } 715 716 /* If the CRC is still on the packet, trim it off. */ 717 if (m->m_flags & M_HASFCS) { 718 m_adj(m, -ETHER_CRC_LEN); 719 m->m_flags &= ~M_HASFCS; 720 } 721 722 if_statadd(ifp, if_ibytes, m->m_pkthdr.len); 723 724 #if NCARP > 0 725 if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) { 726 /* 727 * Clear M_PROMISC, in case the packet comes from a 728 * vlan. 729 */ 730 m->m_flags &= ~M_PROMISC; 731 if (carp_input(m, (uint8_t *)&eh->ether_shost, 732 (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0) 733 return; 734 } 735 #endif 736 737 if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 && 738 (ifp->if_flags & IFF_PROMISC) != 0 && 739 memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost, 740 ETHER_ADDR_LEN) != 0) { 741 m->m_flags |= M_PROMISC; 742 } 743 744 if ((m->m_flags & M_PROMISC) == 0) { 745 if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0) 746 return; 747 if (m == NULL) 748 return; 749 750 eh = mtod(m, struct ether_header *); 751 etype = ntohs(eh->ether_type); 752 } 753 754 #if NAGR > 0 755 if (ifp->if_agrprivate && 756 __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) { 757 m->m_flags &= ~M_PROMISC; 758 agr_input(ifp, m); 759 return; 760 } 761 #endif 762 763 /* 764 * If VLANs are configured on the interface, check to 765 * see if the device performed the decapsulation and 766 * provided us with the tag. 767 */ 768 if (ec->ec_nvlans && vlan_has_tag(m)) { 769 #if NVLAN > 0 770 /* 771 * vlan_input() will either recursively call ether_input() 772 * or drop the packet. 773 */ 774 vlan_input(ifp, m); 775 return; 776 #else 777 goto drop; 778 #endif 779 } 780 781 /* 782 * Handle protocols that expect to have the Ethernet header 783 * (and possibly FCS) intact. 784 */ 785 switch (etype) { 786 case ETHERTYPE_VLAN: { 787 struct ether_vlan_header *evl = (void *)eh; 788 789 /* 790 * If there is a tag of 0, then the VLAN header was probably 791 * just being used to store the priority. Extract the ether 792 * type, and if IP or IPV6, let them deal with it. 793 */ 794 if (m->m_len >= sizeof(*evl) && 795 EVL_VLANOFTAG(evl->evl_tag) == 0) { 796 etype = ntohs(evl->evl_proto); 797 ehlen = sizeof(*evl); 798 if ((m->m_flags & M_PROMISC) == 0 && 799 (etype == ETHERTYPE_IP || 800 etype == ETHERTYPE_IPV6)) 801 break; 802 } 803 804 #if NVLAN > 0 805 /* 806 * vlan_input() will either recursively call ether_input() 807 * or drop the packet. 808 */ 809 if (ec->ec_nvlans != 0) { 810 vlan_input(ifp, m); 811 return; 812 } else 813 #endif 814 goto drop; 815 } 816 817 #if NPPPOE > 0 818 case ETHERTYPE_PPPOEDISC: 819 pppoedisc_input(ifp, m); 820 return; 821 822 case ETHERTYPE_PPPOE: 823 pppoe_input(ifp, m); 824 return; 825 #endif 826 827 case ETHERTYPE_SLOWPROTOCOLS: { 828 uint8_t subtype; 829 830 if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) 831 goto drop; 832 833 m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype); 834 switch (subtype) { 835 #if NAGR > 0 836 case SLOWPROTOCOLS_SUBTYPE_LACP: 837 if (ifp->if_agrprivate) { 838 ieee8023ad_lacp_input(ifp, m); 839 return; 840 } 841 break; 842 843 case SLOWPROTOCOLS_SUBTYPE_MARKER: 844 if (ifp->if_agrprivate) { 845 ieee8023ad_marker_input(ifp, m); 846 return; 847 } 848 break; 849 #endif 850 851 default: 852 if (subtype == 0 || subtype > 10) { 853 /* illegal value */ 854 goto drop; 855 } 856 /* unknown subtype */ 857 break; 858 } 859 } 860 /* FALLTHROUGH */ 861 default: 862 if (m->m_flags & M_PROMISC) 863 goto drop; 864 } 865 866 /* If the CRC is still on the packet, trim it off. */ 867 if (m->m_flags & M_HASFCS) { 868 m_adj(m, -ETHER_CRC_LEN); 869 m->m_flags &= ~M_HASFCS; 870 } 871 872 /* etype represents the size of the payload in this case */ 873 if (etype <= ETHERMTU + sizeof(struct ether_header)) { 874 KASSERT(ehlen == sizeof(*eh)); 875 #if defined (LLC) || defined (NETATALK) 876 ether_input_llc(ifp, m, eh); 877 return; 878 #else 879 goto drop; 880 #endif 881 } 882 883 /* Strip off the Ethernet header. */ 884 m_adj(m, ehlen); 885 886 switch (etype) { 887 #ifdef INET 888 case ETHERTYPE_IP: 889 #ifdef GATEWAY 890 if (ipflow_fastforward(m)) 891 return; 892 #endif 893 pktq = ip_pktq; 894 break; 895 896 case ETHERTYPE_ARP: 897 isr = NETISR_ARP; 898 inq = &arpintrq; 899 break; 900 901 case ETHERTYPE_REVARP: 902 revarpinput(m); /* XXX queue? */ 903 return; 904 #endif 905 906 #ifdef INET6 907 case ETHERTYPE_IPV6: 908 if (__predict_false(!in6_present)) 909 goto drop; 910 #ifdef GATEWAY 911 if (ip6flow_fastforward(&m)) 912 return; 913 #endif 914 pktq = ip6_pktq; 915 break; 916 #endif 917 918 #ifdef NETATALK 919 case ETHERTYPE_ATALK: 920 isr = NETISR_ATALK; 921 inq = &atintrq1; 922 break; 923 924 case ETHERTYPE_AARP: 925 aarpinput(ifp, m); /* XXX queue? */ 926 return; 927 #endif 928 929 #ifdef MPLS 930 case ETHERTYPE_MPLS: 931 isr = NETISR_MPLS; 932 inq = &mplsintrq; 933 break; 934 #endif 935 936 default: 937 goto drop; 938 } 939 940 if (__predict_true(pktq)) { 941 #ifdef NET_MPSAFE 942 const u_int h = curcpu()->ci_index; 943 #else 944 const uint32_t h = pktq_rps_hash(m); 945 #endif 946 if (__predict_false(!pktq_enqueue(pktq, m, h))) { 947 m_freem(m); 948 } 949 return; 950 } 951 952 if (__predict_false(!inq)) { 953 /* Should not happen. */ 954 goto drop; 955 } 956 957 IFQ_ENQUEUE_ISR(inq, m, isr); 958 return; 959 960 drop: 961 m_freem(m); 962 dropped: 963 if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */ 964 } 965 966 /* 967 * Convert Ethernet address to printable (loggable) representation. 968 */ 969 char * 970 ether_sprintf(const u_char *ap) 971 { 972 static char etherbuf[3 * ETHER_ADDR_LEN]; 973 return ether_snprintf(etherbuf, sizeof(etherbuf), ap); 974 } 975 976 char * 977 ether_snprintf(char *buf, size_t len, const u_char *ap) 978 { 979 char *cp = buf; 980 size_t i; 981 982 for (i = 0; i < len / 3; i++) { 983 *cp++ = hexdigits[*ap >> 4]; 984 *cp++ = hexdigits[*ap++ & 0xf]; 985 *cp++ = ':'; 986 } 987 *--cp = '\0'; 988 return buf; 989 } 990 991 static void 992 ether_link_state_changed(struct ifnet *ifp, int link_state) 993 { 994 #if NVLAN > 0 995 struct ethercom *ec = (void *)ifp; 996 997 if (ec->ec_nvlans) 998 vlan_link_state_changed(ifp, link_state); 999 #endif 1000 } 1001 1002 /* 1003 * Perform common duties while attaching to interface list 1004 */ 1005 void 1006 ether_ifattach(struct ifnet *ifp, const uint8_t *lla) 1007 { 1008 struct ethercom *ec = (struct ethercom *)ifp; 1009 1010 ifp->if_type = IFT_ETHER; 1011 ifp->if_hdrlen = ETHER_HDR_LEN; 1012 ifp->if_dlt = DLT_EN10MB; 1013 ifp->if_mtu = ETHERMTU; 1014 ifp->if_output = ether_output; 1015 ifp->_if_input = ether_input; 1016 ifp->if_link_state_changed = ether_link_state_changed; 1017 if (ifp->if_baudrate == 0) 1018 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1019 1020 if (lla != NULL) 1021 if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla)); 1022 1023 LIST_INIT(&ec->ec_multiaddrs); 1024 SIMPLEQ_INIT(&ec->ec_vids); 1025 ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 1026 ec->ec_flags = 0; 1027 ifp->if_broadcastaddr = etherbroadcastaddr; 1028 bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header)); 1029 #ifdef MBUFTRACE 1030 mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx"); 1031 mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx"); 1032 MOWNER_ATTACH(&ec->ec_tx_mowner); 1033 MOWNER_ATTACH(&ec->ec_rx_mowner); 1034 ifp->if_mowner = &ec->ec_tx_mowner; 1035 #endif 1036 } 1037 1038 void 1039 ether_ifdetach(struct ifnet *ifp) 1040 { 1041 struct ethercom *ec = (void *) ifp; 1042 struct ether_multi *enm; 1043 1044 IFNET_ASSERT_UNLOCKED(ifp); 1045 /* 1046 * Prevent further calls to ioctl (for example turning off 1047 * promiscuous mode from the bridge code), which eventually can 1048 * call if_init() which can cause panics because the interface 1049 * is in the process of being detached. Return device not configured 1050 * instead. 1051 */ 1052 ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *), 1053 enxio); 1054 1055 #if NBRIDGE > 0 1056 if (ifp->if_bridge) 1057 bridge_ifdetach(ifp); 1058 #endif 1059 bpf_detach(ifp); 1060 #if NVLAN > 0 1061 if (ec->ec_nvlans) 1062 vlan_ifdetach(ifp); 1063 #endif 1064 1065 ETHER_LOCK(ec); 1066 KASSERT(ec->ec_nvlans == 0); 1067 while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) { 1068 LIST_REMOVE(enm, enm_list); 1069 kmem_free(enm, sizeof(*enm)); 1070 ec->ec_multicnt--; 1071 } 1072 ETHER_UNLOCK(ec); 1073 1074 mutex_obj_free(ec->ec_lock); 1075 ec->ec_lock = NULL; 1076 1077 ifp->if_mowner = NULL; 1078 MOWNER_DETACH(&ec->ec_rx_mowner); 1079 MOWNER_DETACH(&ec->ec_tx_mowner); 1080 } 1081 1082 #if 0 1083 /* 1084 * This is for reference. We have a table-driven version 1085 * of the little-endian crc32 generator, which is faster 1086 * than the double-loop. 1087 */ 1088 uint32_t 1089 ether_crc32_le(const uint8_t *buf, size_t len) 1090 { 1091 uint32_t c, crc, carry; 1092 size_t i, j; 1093 1094 crc = 0xffffffffU; /* initial value */ 1095 1096 for (i = 0; i < len; i++) { 1097 c = buf[i]; 1098 for (j = 0; j < 8; j++) { 1099 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1100 crc >>= 1; 1101 c >>= 1; 1102 if (carry) 1103 crc = (crc ^ ETHER_CRC_POLY_LE); 1104 } 1105 } 1106 1107 return (crc); 1108 } 1109 #else 1110 uint32_t 1111 ether_crc32_le(const uint8_t *buf, size_t len) 1112 { 1113 static const uint32_t crctab[] = { 1114 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1115 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1116 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1117 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1118 }; 1119 uint32_t crc; 1120 size_t i; 1121 1122 crc = 0xffffffffU; /* initial value */ 1123 1124 for (i = 0; i < len; i++) { 1125 crc ^= buf[i]; 1126 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1127 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1128 } 1129 1130 return (crc); 1131 } 1132 #endif 1133 1134 uint32_t 1135 ether_crc32_be(const uint8_t *buf, size_t len) 1136 { 1137 uint32_t c, crc, carry; 1138 size_t i, j; 1139 1140 crc = 0xffffffffU; /* initial value */ 1141 1142 for (i = 0; i < len; i++) { 1143 c = buf[i]; 1144 for (j = 0; j < 8; j++) { 1145 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1146 crc <<= 1; 1147 c >>= 1; 1148 if (carry) 1149 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1150 } 1151 } 1152 1153 return (crc); 1154 } 1155 1156 #ifdef INET 1157 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] = 1158 { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }; 1159 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] = 1160 { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff }; 1161 #endif 1162 #ifdef INET6 1163 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] = 1164 { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 }; 1165 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] = 1166 { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff }; 1167 #endif 1168 1169 /* 1170 * ether_aton implementation, not using a static buffer. 1171 */ 1172 int 1173 ether_aton_r(u_char *dest, size_t len, const char *str) 1174 { 1175 const u_char *cp = (const void *)str; 1176 u_char *ep; 1177 1178 #define atox(c) (((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10)) 1179 1180 if (len < ETHER_ADDR_LEN) 1181 return ENOSPC; 1182 1183 ep = dest + ETHER_ADDR_LEN; 1184 1185 while (*cp) { 1186 if (!isxdigit(*cp)) 1187 return EINVAL; 1188 1189 *dest = atox(*cp); 1190 cp++; 1191 if (isxdigit(*cp)) { 1192 *dest = (*dest << 4) | atox(*cp); 1193 cp++; 1194 } 1195 dest++; 1196 1197 if (dest == ep) 1198 return (*cp == '\0') ? 0 : ENAMETOOLONG; 1199 1200 switch (*cp) { 1201 case ':': 1202 case '-': 1203 case '.': 1204 cp++; 1205 break; 1206 } 1207 } 1208 return ENOBUFS; 1209 } 1210 1211 /* 1212 * Convert a sockaddr into an Ethernet address or range of Ethernet 1213 * addresses. 1214 */ 1215 int 1216 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN], 1217 uint8_t addrhi[ETHER_ADDR_LEN]) 1218 { 1219 #ifdef INET 1220 const struct sockaddr_in *sin; 1221 #endif 1222 #ifdef INET6 1223 const struct sockaddr_in6 *sin6; 1224 #endif 1225 1226 switch (sa->sa_family) { 1227 1228 case AF_UNSPEC: 1229 memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN); 1230 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1231 break; 1232 1233 #ifdef INET 1234 case AF_INET: 1235 sin = satocsin(sa); 1236 if (sin->sin_addr.s_addr == INADDR_ANY) { 1237 /* 1238 * An IP address of INADDR_ANY means listen to 1239 * or stop listening to all of the Ethernet 1240 * multicast addresses used for IP. 1241 * (This is for the sake of IP multicast routers.) 1242 */ 1243 memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN); 1244 memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN); 1245 } else { 1246 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo); 1247 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1248 } 1249 break; 1250 #endif 1251 #ifdef INET6 1252 case AF_INET6: 1253 sin6 = satocsin6(sa); 1254 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1255 /* 1256 * An IP6 address of 0 means listen to or stop 1257 * listening to all of the Ethernet multicast 1258 * address used for IP6. 1259 * (This is used for multicast routers.) 1260 */ 1261 memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN); 1262 memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN); 1263 } else { 1264 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo); 1265 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1266 } 1267 break; 1268 #endif 1269 1270 default: 1271 return EAFNOSUPPORT; 1272 } 1273 return 0; 1274 } 1275 1276 /* 1277 * Add an Ethernet multicast address or range of addresses to the list for a 1278 * given interface. 1279 */ 1280 int 1281 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec) 1282 { 1283 struct ether_multi *enm, *_enm; 1284 u_char addrlo[ETHER_ADDR_LEN]; 1285 u_char addrhi[ETHER_ADDR_LEN]; 1286 int error = 0; 1287 1288 /* Allocate out of lock */ 1289 enm = kmem_alloc(sizeof(*enm), KM_SLEEP); 1290 1291 ETHER_LOCK(ec); 1292 error = ether_multiaddr(sa, addrlo, addrhi); 1293 if (error != 0) 1294 goto out; 1295 1296 /* 1297 * Verify that we have valid Ethernet multicast addresses. 1298 */ 1299 if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) { 1300 error = EINVAL; 1301 goto out; 1302 } 1303 1304 /* 1305 * See if the address range is already in the list. 1306 */ 1307 _enm = ether_lookup_multi(addrlo, addrhi, ec); 1308 if (_enm != NULL) { 1309 /* 1310 * Found it; just increment the reference count. 1311 */ 1312 ++_enm->enm_refcount; 1313 error = 0; 1314 goto out; 1315 } 1316 1317 /* 1318 * Link a new multicast record into the interface's multicast list. 1319 */ 1320 memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN); 1321 memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN); 1322 enm->enm_refcount = 1; 1323 LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list); 1324 ec->ec_multicnt++; 1325 1326 /* 1327 * Return ENETRESET to inform the driver that the list has changed 1328 * and its reception filter should be adjusted accordingly. 1329 */ 1330 error = ENETRESET; 1331 enm = NULL; 1332 1333 out: 1334 ETHER_UNLOCK(ec); 1335 if (enm != NULL) 1336 kmem_free(enm, sizeof(*enm)); 1337 return error; 1338 } 1339 1340 /* 1341 * Delete a multicast address record. 1342 */ 1343 int 1344 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec) 1345 { 1346 struct ether_multi *enm; 1347 u_char addrlo[ETHER_ADDR_LEN]; 1348 u_char addrhi[ETHER_ADDR_LEN]; 1349 int error; 1350 1351 ETHER_LOCK(ec); 1352 error = ether_multiaddr(sa, addrlo, addrhi); 1353 if (error != 0) 1354 goto error; 1355 1356 /* 1357 * Look up the address in our list. 1358 */ 1359 enm = ether_lookup_multi(addrlo, addrhi, ec); 1360 if (enm == NULL) { 1361 error = ENXIO; 1362 goto error; 1363 } 1364 if (--enm->enm_refcount != 0) { 1365 /* 1366 * Still some claims to this record. 1367 */ 1368 error = 0; 1369 goto error; 1370 } 1371 1372 /* 1373 * No remaining claims to this record; unlink and free it. 1374 */ 1375 LIST_REMOVE(enm, enm_list); 1376 ec->ec_multicnt--; 1377 ETHER_UNLOCK(ec); 1378 kmem_free(enm, sizeof(*enm)); 1379 1380 /* 1381 * Return ENETRESET to inform the driver that the list has changed 1382 * and its reception filter should be adjusted accordingly. 1383 */ 1384 return ENETRESET; 1385 1386 error: 1387 ETHER_UNLOCK(ec); 1388 return error; 1389 } 1390 1391 void 1392 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb) 1393 { 1394 ec->ec_ifflags_cb = cb; 1395 } 1396 1397 void 1398 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb) 1399 { 1400 1401 ec->ec_vlan_cb = cb; 1402 } 1403 1404 static int 1405 ether_ioctl_reinit(struct ethercom *ec) 1406 { 1407 struct ifnet *ifp = &ec->ec_if; 1408 int error; 1409 1410 switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) { 1411 case IFF_RUNNING: 1412 /* 1413 * If interface is marked down and it is running, 1414 * then stop and disable it. 1415 */ 1416 (*ifp->if_stop)(ifp, 1); 1417 break; 1418 case IFF_UP: 1419 /* 1420 * If interface is marked up and it is stopped, then 1421 * start it. 1422 */ 1423 return (*ifp->if_init)(ifp); 1424 case IFF_UP | IFF_RUNNING: 1425 error = 0; 1426 if (ec->ec_ifflags_cb != NULL) { 1427 error = (*ec->ec_ifflags_cb)(ec); 1428 if (error == ENETRESET) { 1429 /* 1430 * Reset the interface to pick up 1431 * changes in any other flags that 1432 * affect the hardware state. 1433 */ 1434 return (*ifp->if_init)(ifp); 1435 } 1436 } else 1437 error = (*ifp->if_init)(ifp); 1438 return error; 1439 case 0: 1440 break; 1441 } 1442 1443 return 0; 1444 } 1445 1446 /* 1447 * Common ioctls for Ethernet interfaces. Note, we must be 1448 * called at splnet(). 1449 */ 1450 int 1451 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1452 { 1453 struct ethercom *ec = (void *)ifp; 1454 struct eccapreq *eccr; 1455 struct ifreq *ifr = (struct ifreq *)data; 1456 struct if_laddrreq *iflr = data; 1457 const struct sockaddr_dl *sdl; 1458 static const uint8_t zero[ETHER_ADDR_LEN]; 1459 int error; 1460 1461 switch (cmd) { 1462 case SIOCINITIFADDR: 1463 { 1464 struct ifaddr *ifa = (struct ifaddr *)data; 1465 if (ifa->ifa_addr->sa_family != AF_LINK 1466 && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 1467 (IFF_UP | IFF_RUNNING)) { 1468 ifp->if_flags |= IFF_UP; 1469 if ((error = (*ifp->if_init)(ifp)) != 0) 1470 return error; 1471 } 1472 #ifdef INET 1473 if (ifa->ifa_addr->sa_family == AF_INET) 1474 arp_ifinit(ifp, ifa); 1475 #endif 1476 return 0; 1477 } 1478 1479 case SIOCSIFMTU: 1480 { 1481 int maxmtu; 1482 1483 if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU) 1484 maxmtu = ETHERMTU_JUMBO; 1485 else 1486 maxmtu = ETHERMTU; 1487 1488 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu) 1489 return EINVAL; 1490 else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET) 1491 return error; 1492 else if (ifp->if_flags & IFF_UP) { 1493 /* Make sure the device notices the MTU change. */ 1494 return (*ifp->if_init)(ifp); 1495 } else 1496 return 0; 1497 } 1498 1499 case SIOCSIFFLAGS: 1500 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1501 return error; 1502 return ether_ioctl_reinit(ec); 1503 case SIOCGIFFLAGS: 1504 error = ifioctl_common(ifp, cmd, data); 1505 if (error == 0) { 1506 /* Set IFF_ALLMULTI for backcompat */ 1507 ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ? 1508 IFF_ALLMULTI : 0; 1509 } 1510 return error; 1511 case SIOCGETHERCAP: 1512 eccr = (struct eccapreq *)data; 1513 eccr->eccr_capabilities = ec->ec_capabilities; 1514 eccr->eccr_capenable = ec->ec_capenable; 1515 return 0; 1516 case SIOCSETHERCAP: 1517 eccr = (struct eccapreq *)data; 1518 if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0) 1519 return EINVAL; 1520 if (eccr->eccr_capenable == ec->ec_capenable) 1521 return 0; 1522 #if 0 /* notyet */ 1523 ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE) 1524 | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE); 1525 #else 1526 ec->ec_capenable = eccr->eccr_capenable; 1527 #endif 1528 return ether_ioctl_reinit(ec); 1529 case SIOCADDMULTI: 1530 return ether_addmulti(ifreq_getaddr(cmd, ifr), ec); 1531 case SIOCDELMULTI: 1532 return ether_delmulti(ifreq_getaddr(cmd, ifr), ec); 1533 case SIOCSIFMEDIA: 1534 case SIOCGIFMEDIA: 1535 if (ec->ec_mii != NULL) 1536 return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, 1537 cmd); 1538 else if (ec->ec_ifmedia != NULL) 1539 return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd); 1540 else 1541 return ENOTTY; 1542 break; 1543 case SIOCALIFADDR: 1544 sdl = satocsdl(sstocsa(&iflr->addr)); 1545 if (sdl->sdl_family != AF_LINK) 1546 ; 1547 else if (ETHER_IS_MULTICAST(CLLADDR(sdl))) 1548 return EINVAL; 1549 else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0) 1550 return EINVAL; 1551 /*FALLTHROUGH*/ 1552 default: 1553 return ifioctl_common(ifp, cmd, data); 1554 } 1555 return 0; 1556 } 1557 1558 /* 1559 * Enable/disable passing VLAN packets if the parent interface supports it. 1560 * Return: 1561 * 0: Ok 1562 * -1: Parent interface does not support vlans 1563 * >0: Error 1564 */ 1565 int 1566 ether_enable_vlan_mtu(struct ifnet *ifp) 1567 { 1568 int error; 1569 struct ethercom *ec = (void *)ifp; 1570 1571 /* Parent does not support VLAN's */ 1572 if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0) 1573 return -1; 1574 1575 /* 1576 * Parent supports the VLAN_MTU capability, 1577 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames; 1578 * enable it. 1579 */ 1580 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1581 1582 /* Interface is down, defer for later */ 1583 if ((ifp->if_flags & IFF_UP) == 0) 1584 return 0; 1585 1586 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1587 return 0; 1588 1589 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1590 return error; 1591 } 1592 1593 int 1594 ether_disable_vlan_mtu(struct ifnet *ifp) 1595 { 1596 int error; 1597 struct ethercom *ec = (void *)ifp; 1598 1599 /* We still have VLAN's, defer for later */ 1600 if (ec->ec_nvlans != 0) 1601 return 0; 1602 1603 /* Parent does not support VLAB's, nothing to do. */ 1604 if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0) 1605 return -1; 1606 1607 /* 1608 * Disable Tx/Rx of VLAN-sized frames. 1609 */ 1610 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1611 1612 /* Interface is down, defer for later */ 1613 if ((ifp->if_flags & IFF_UP) == 0) 1614 return 0; 1615 1616 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1617 return 0; 1618 1619 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1620 return error; 1621 } 1622 1623 static int 1624 ether_multicast_sysctl(SYSCTLFN_ARGS) 1625 { 1626 struct ether_multi *enm; 1627 struct ifnet *ifp; 1628 struct ethercom *ec; 1629 int error = 0; 1630 size_t written; 1631 struct psref psref; 1632 int bound; 1633 unsigned int multicnt; 1634 struct ether_multi_sysctl *addrs; 1635 int i; 1636 1637 if (namelen != 1) 1638 return EINVAL; 1639 1640 bound = curlwp_bind(); 1641 ifp = if_get_byindex(name[0], &psref); 1642 if (ifp == NULL) { 1643 error = ENODEV; 1644 goto out; 1645 } 1646 if (ifp->if_type != IFT_ETHER) { 1647 if_put(ifp, &psref); 1648 *oldlenp = 0; 1649 goto out; 1650 } 1651 ec = (struct ethercom *)ifp; 1652 1653 if (oldp == NULL) { 1654 if_put(ifp, &psref); 1655 *oldlenp = ec->ec_multicnt * sizeof(*addrs); 1656 goto out; 1657 } 1658 1659 /* 1660 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which 1661 * is sleepable, while holding it. Copy data to a local buffer first 1662 * with the lock taken and then call sysctl_copyout without holding it. 1663 */ 1664 retry: 1665 multicnt = ec->ec_multicnt; 1666 1667 if (multicnt == 0) { 1668 if_put(ifp, &psref); 1669 *oldlenp = 0; 1670 goto out; 1671 } 1672 1673 addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP); 1674 1675 ETHER_LOCK(ec); 1676 if (multicnt != ec->ec_multicnt) { 1677 /* The number of multicast addresses has changed */ 1678 ETHER_UNLOCK(ec); 1679 kmem_free(addrs, sizeof(*addrs) * multicnt); 1680 goto retry; 1681 } 1682 1683 i = 0; 1684 LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) { 1685 struct ether_multi_sysctl *addr = &addrs[i]; 1686 addr->enm_refcount = enm->enm_refcount; 1687 memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN); 1688 memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN); 1689 i++; 1690 } 1691 ETHER_UNLOCK(ec); 1692 1693 error = 0; 1694 written = 0; 1695 for (i = 0; i < multicnt; i++) { 1696 struct ether_multi_sysctl *addr = &addrs[i]; 1697 1698 if (written + sizeof(*addr) > *oldlenp) 1699 break; 1700 error = sysctl_copyout(l, addr, oldp, sizeof(*addr)); 1701 if (error) 1702 break; 1703 written += sizeof(*addr); 1704 oldp = (char *)oldp + sizeof(*addr); 1705 } 1706 kmem_free(addrs, sizeof(*addrs) * multicnt); 1707 1708 if_put(ifp, &psref); 1709 1710 *oldlenp = written; 1711 out: 1712 curlwp_bindx(bound); 1713 return error; 1714 } 1715 1716 static void 1717 ether_sysctl_setup(struct sysctllog **clog) 1718 { 1719 const struct sysctlnode *rnode = NULL; 1720 1721 sysctl_createv(clog, 0, NULL, &rnode, 1722 CTLFLAG_PERMANENT, 1723 CTLTYPE_NODE, "ether", 1724 SYSCTL_DESCR("Ethernet-specific information"), 1725 NULL, 0, NULL, 0, 1726 CTL_NET, CTL_CREATE, CTL_EOL); 1727 1728 sysctl_createv(clog, 0, &rnode, NULL, 1729 CTLFLAG_PERMANENT, 1730 CTLTYPE_NODE, "multicast", 1731 SYSCTL_DESCR("multicast addresses"), 1732 ether_multicast_sysctl, 0, NULL, 0, 1733 CTL_CREATE, CTL_EOL); 1734 } 1735 1736 void 1737 etherinit(void) 1738 { 1739 1740 #ifdef DIAGNOSTIC 1741 mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET); 1742 #endif 1743 ether_sysctl_setup(NULL); 1744 } 1745
Indexes created Wed Sep 24 22:09:59 GMT 2025