if_ethersubr.c revision 1.299
1 /* $NetBSD: if_ethersubr.c,v 1.299 2021/09/30 04:13:42 yamaguchi 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 1.299 2021/09/30 04:13:42 yamaguchi 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 #include "lagg.h" 82 83 #include <sys/sysctl.h> 84 #include <sys/mbuf.h> 85 #include <sys/mutex.h> 86 #include <sys/ioctl.h> 87 #include <sys/errno.h> 88 #include <sys/device.h> 89 #include <sys/entropy.h> 90 #include <sys/rndsource.h> 91 #include <sys/cpu.h> 92 #include <sys/kmem.h> 93 #include <sys/hook.h> 94 95 #include <net/if.h> 96 #include <net/netisr.h> 97 #include <net/route.h> 98 #include <net/if_llc.h> 99 #include <net/if_dl.h> 100 #include <net/if_types.h> 101 #include <net/pktqueue.h> 102 103 #include <net/if_media.h> 104 #include <dev/mii/mii.h> 105 #include <dev/mii/miivar.h> 106 107 #if NARP == 0 108 /* 109 * XXX there should really be a way to issue this warning from within config(8) 110 */ 111 #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. 112 #endif 113 114 #include <net/bpf.h> 115 116 #include <net/if_ether.h> 117 #include <net/if_vlanvar.h> 118 119 #if NPPPOE > 0 120 #include <net/if_pppoe.h> 121 #endif 122 123 #if NAGR > 0 124 #include <net/agr/ieee8023_slowprotocols.h> /* XXX */ 125 #include <net/agr/ieee8023ad.h> 126 #include <net/agr/if_agrvar.h> 127 #endif 128 129 #include <net/lagg/if_laggvar.h> 130 131 #if NBRIDGE > 0 132 #include <net/if_bridgevar.h> 133 #endif 134 135 #include <netinet/in.h> 136 #ifdef INET 137 #include <netinet/in_var.h> 138 #endif 139 #include <netinet/if_inarp.h> 140 141 #ifdef INET6 142 #ifndef INET 143 #include <netinet/in.h> 144 #endif 145 #include <netinet6/in6_var.h> 146 #include <netinet6/nd6.h> 147 #endif 148 149 #include "carp.h" 150 #if NCARP > 0 151 #include <netinet/ip_carp.h> 152 #endif 153 154 #ifdef NETATALK 155 #include <netatalk/at.h> 156 #include <netatalk/at_var.h> 157 #include <netatalk/at_extern.h> 158 159 #define llc_snap_org_code llc_un.type_snap.org_code 160 #define llc_snap_ether_type llc_un.type_snap.ether_type 161 162 extern u_char at_org_code[3]; 163 extern u_char aarp_org_code[3]; 164 #endif /* NETATALK */ 165 166 #ifdef MPLS 167 #include <netmpls/mpls.h> 168 #include <netmpls/mpls_var.h> 169 #endif 170 171 CTASSERT(sizeof(struct ether_addr) == 6); 172 CTASSERT(sizeof(struct ether_header) == 14); 173 174 #ifdef DIAGNOSTIC 175 static struct timeval bigpktppslim_last; 176 static int bigpktppslim = 2; /* XXX */ 177 static int bigpktpps_count; 178 static kmutex_t bigpktpps_lock __cacheline_aligned; 179 #endif 180 181 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = 182 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; 183 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] = 184 { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 }; 185 #define senderr(e) { error = (e); goto bad;} 186 187 /* if_lagg(4) support */ 188 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *); 189 190 static int ether_output(struct ifnet *, struct mbuf *, 191 const struct sockaddr *, const struct rtentry *); 192 193 /* 194 * Ethernet output routine. 195 * Encapsulate a packet of type family for the local net. 196 * Assumes that ifp is actually pointer to ethercom structure. 197 */ 198 static int 199 ether_output(struct ifnet * const ifp0, struct mbuf * const m0, 200 const struct sockaddr * const dst, const struct rtentry *rt) 201 { 202 uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; 203 uint16_t etype = 0; 204 int error = 0, hdrcmplt = 0; 205 struct mbuf *m = m0; 206 struct mbuf *mcopy = NULL; 207 struct ether_header *eh; 208 struct ifnet *ifp = ifp0; 209 #ifdef INET 210 struct arphdr *ah; 211 #endif 212 #ifdef NETATALK 213 struct at_ifaddr *aa; 214 #endif 215 216 #ifdef MBUFTRACE 217 m_claimm(m, ifp->if_mowner); 218 #endif 219 220 #if NCARP > 0 221 if (ifp->if_type == IFT_CARP) { 222 struct ifaddr *ifa; 223 int s = pserialize_read_enter(); 224 225 /* loop back if this is going to the carp interface */ 226 if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP && 227 (ifa = ifa_ifwithaddr(dst)) != NULL) { 228 if (ifa->ifa_ifp == ifp0) { 229 pserialize_read_exit(s); 230 return looutput(ifp0, m, dst, rt); 231 } 232 } 233 pserialize_read_exit(s); 234 235 ifp = ifp->if_carpdev; 236 /* ac = (struct arpcom *)ifp; */ 237 238 if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) != 239 (IFF_UP | IFF_RUNNING)) 240 senderr(ENETDOWN); 241 } 242 #endif 243 244 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 245 senderr(ENETDOWN); 246 247 switch (dst->sa_family) { 248 249 #ifdef INET 250 case AF_INET: 251 if (m->m_flags & M_BCAST) { 252 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 253 } else if (m->m_flags & M_MCAST) { 254 ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst); 255 } else { 256 error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst)); 257 if (error) 258 return (error == EWOULDBLOCK) ? 0 : error; 259 } 260 /* If broadcasting on a simplex interface, loopback a copy */ 261 if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX)) 262 mcopy = m_copypacket(m, M_DONTWAIT); 263 etype = htons(ETHERTYPE_IP); 264 break; 265 266 case AF_ARP: 267 ah = mtod(m, struct arphdr *); 268 if (m->m_flags & M_BCAST) { 269 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 270 } else { 271 void *tha = ar_tha(ah); 272 273 if (tha == NULL) { 274 /* fake with ARPHRD_IEEE1394 */ 275 m_freem(m); 276 return 0; 277 } 278 memcpy(edst, tha, sizeof(edst)); 279 } 280 281 ah->ar_hrd = htons(ARPHRD_ETHER); 282 283 switch (ntohs(ah->ar_op)) { 284 case ARPOP_REVREQUEST: 285 case ARPOP_REVREPLY: 286 etype = htons(ETHERTYPE_REVARP); 287 break; 288 289 case ARPOP_REQUEST: 290 case ARPOP_REPLY: 291 default: 292 etype = htons(ETHERTYPE_ARP); 293 } 294 break; 295 #endif 296 297 #ifdef INET6 298 case AF_INET6: 299 if (m->m_flags & M_BCAST) { 300 memcpy(edst, etherbroadcastaddr, sizeof(edst)); 301 } else if (m->m_flags & M_MCAST) { 302 ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr, 303 edst); 304 } else { 305 error = nd6_resolve(ifp0, rt, m, dst, edst, 306 sizeof(edst)); 307 if (error) 308 return (error == EWOULDBLOCK) ? 0 : error; 309 } 310 etype = htons(ETHERTYPE_IPV6); 311 break; 312 #endif 313 314 #ifdef NETATALK 315 case AF_APPLETALK: { 316 struct ifaddr *ifa; 317 int s; 318 319 KERNEL_LOCK(1, NULL); 320 321 if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { 322 KERNEL_UNLOCK_ONE(NULL); 323 return 0; 324 } 325 326 /* 327 * ifaddr is the first thing in at_ifaddr 328 */ 329 s = pserialize_read_enter(); 330 ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp); 331 if (ifa == NULL) { 332 pserialize_read_exit(s); 333 KERNEL_UNLOCK_ONE(NULL); 334 senderr(EADDRNOTAVAIL); 335 } 336 aa = (struct at_ifaddr *)ifa; 337 338 /* 339 * In the phase 2 case, we need to prepend an mbuf for the 340 * llc header. 341 */ 342 if (aa->aa_flags & AFA_PHASE2) { 343 struct llc llc; 344 345 M_PREPEND(m, sizeof(struct llc), M_DONTWAIT); 346 if (m == NULL) { 347 pserialize_read_exit(s); 348 KERNEL_UNLOCK_ONE(NULL); 349 senderr(ENOBUFS); 350 } 351 352 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 353 llc.llc_control = LLC_UI; 354 memcpy(llc.llc_snap_org_code, at_org_code, 355 sizeof(llc.llc_snap_org_code)); 356 llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK); 357 memcpy(mtod(m, void *), &llc, sizeof(struct llc)); 358 } else { 359 etype = htons(ETHERTYPE_ATALK); 360 } 361 pserialize_read_exit(s); 362 KERNEL_UNLOCK_ONE(NULL); 363 break; 364 } 365 #endif /* NETATALK */ 366 367 case pseudo_AF_HDRCMPLT: 368 hdrcmplt = 1; 369 memcpy(esrc, 370 ((const struct ether_header *)dst->sa_data)->ether_shost, 371 sizeof(esrc)); 372 /* FALLTHROUGH */ 373 374 case AF_UNSPEC: 375 memcpy(edst, 376 ((const struct ether_header *)dst->sa_data)->ether_dhost, 377 sizeof(edst)); 378 /* AF_UNSPEC doesn't swap the byte order of the ether_type. */ 379 etype = ((const struct ether_header *)dst->sa_data)->ether_type; 380 break; 381 382 default: 383 printf("%s: can't handle af%d\n", ifp->if_xname, 384 dst->sa_family); 385 senderr(EAFNOSUPPORT); 386 } 387 388 #ifdef MPLS 389 { 390 struct m_tag *mtag; 391 mtag = m_tag_find(m, PACKET_TAG_MPLS); 392 if (mtag != NULL) { 393 /* Having the tag itself indicates it's MPLS */ 394 etype = htons(ETHERTYPE_MPLS); 395 m_tag_delete(m, mtag); 396 } 397 } 398 #endif 399 400 if (mcopy) 401 (void)looutput(ifp, mcopy, dst, rt); 402 403 KASSERT((m->m_flags & M_PKTHDR) != 0); 404 405 /* 406 * If no ether type is set, this must be a 802.2 formatted packet. 407 */ 408 if (etype == 0) 409 etype = htons(m->m_pkthdr.len); 410 411 /* 412 * Add local net header. If no space in first mbuf, allocate another. 413 */ 414 M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT); 415 if (m == NULL) 416 senderr(ENOBUFS); 417 418 eh = mtod(m, struct ether_header *); 419 /* Note: etype is already in network byte order. */ 420 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type)); 421 memcpy(eh->ether_dhost, edst, sizeof(edst)); 422 if (hdrcmplt) { 423 memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost)); 424 } else { 425 memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl), 426 sizeof(eh->ether_shost)); 427 } 428 429 #if NCARP > 0 430 if (ifp0 != ifp && ifp0->if_type == IFT_CARP) { 431 memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl), 432 sizeof(eh->ether_shost)); 433 } 434 #endif 435 436 if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0) 437 return error; 438 if (m == NULL) 439 return 0; 440 441 #if NBRIDGE > 0 442 /* 443 * Bridges require special output handling. 444 */ 445 if (ifp->if_bridge) 446 return bridge_output(ifp, m, NULL, NULL); 447 #endif 448 449 #if NCARP > 0 450 if (ifp != ifp0) 451 if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN); 452 #endif 453 454 #ifdef ALTQ 455 KERNEL_LOCK(1, NULL); 456 /* 457 * If ALTQ is enabled on the parent interface, do 458 * classification; the queueing discipline might not 459 * require classification, but might require the 460 * address family/header pointer in the pktattr. 461 */ 462 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 463 altq_etherclassify(&ifp->if_snd, m); 464 KERNEL_UNLOCK_ONE(NULL); 465 #endif 466 return ifq_enqueue(ifp, m); 467 468 bad: 469 if_statinc(ifp, if_oerrors); 470 if (m) 471 m_freem(m); 472 return error; 473 } 474 475 #ifdef ALTQ 476 /* 477 * This routine is a slight hack to allow a packet to be classified 478 * if the Ethernet headers are present. It will go away when ALTQ's 479 * classification engine understands link headers. 480 * 481 * XXX: We may need to do m_pullups here. First to ensure struct ether_header 482 * is indeed contiguous, then to read the LLC and so on. 483 */ 484 void 485 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m) 486 { 487 struct ether_header *eh; 488 struct mbuf *mtop = m; 489 uint16_t ether_type; 490 int hlen, af, hdrsize; 491 void *hdr; 492 493 KASSERT((mtop->m_flags & M_PKTHDR) != 0); 494 495 hlen = ETHER_HDR_LEN; 496 eh = mtod(m, struct ether_header *); 497 498 ether_type = htons(eh->ether_type); 499 500 if (ether_type < ETHERMTU) { 501 /* LLC/SNAP */ 502 struct llc *llc = (struct llc *)(eh + 1); 503 hlen += 8; 504 505 if (m->m_len < hlen || 506 llc->llc_dsap != LLC_SNAP_LSAP || 507 llc->llc_ssap != LLC_SNAP_LSAP || 508 llc->llc_control != LLC_UI) { 509 /* Not SNAP. */ 510 goto bad; 511 } 512 513 ether_type = htons(llc->llc_un.type_snap.ether_type); 514 } 515 516 switch (ether_type) { 517 case ETHERTYPE_IP: 518 af = AF_INET; 519 hdrsize = 20; /* sizeof(struct ip) */ 520 break; 521 522 case ETHERTYPE_IPV6: 523 af = AF_INET6; 524 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 525 break; 526 527 default: 528 af = AF_UNSPEC; 529 hdrsize = 0; 530 break; 531 } 532 533 while (m->m_len <= hlen) { 534 hlen -= m->m_len; 535 m = m->m_next; 536 if (m == NULL) 537 goto bad; 538 } 539 540 if (m->m_len < (hlen + hdrsize)) { 541 /* 542 * protocol header not in a single mbuf. 543 * We can't cope with this situation right 544 * now (but it shouldn't ever happen, really, anyhow). 545 */ 546 #ifdef DEBUG 547 printf("altq_etherclassify: headers span multiple mbufs: " 548 "%d < %d\n", m->m_len, (hlen + hdrsize)); 549 #endif 550 goto bad; 551 } 552 553 m->m_data += hlen; 554 m->m_len -= hlen; 555 556 hdr = mtod(m, void *); 557 558 if (ALTQ_NEEDS_CLASSIFY(ifq)) { 559 mtop->m_pkthdr.pattr_class = 560 (*ifq->altq_classify)(ifq->altq_clfier, m, af); 561 } 562 mtop->m_pkthdr.pattr_af = af; 563 mtop->m_pkthdr.pattr_hdr = hdr; 564 565 m->m_data -= hlen; 566 m->m_len += hlen; 567 568 return; 569 570 bad: 571 mtop->m_pkthdr.pattr_class = NULL; 572 mtop->m_pkthdr.pattr_hdr = NULL; 573 mtop->m_pkthdr.pattr_af = AF_UNSPEC; 574 } 575 #endif /* ALTQ */ 576 577 #if defined (LLC) || defined (NETATALK) 578 static void 579 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh) 580 { 581 struct ifqueue *inq = NULL; 582 int isr = 0; 583 struct llc *l; 584 585 if (m->m_len < sizeof(*eh) + sizeof(struct llc)) 586 goto drop; 587 588 l = (struct llc *)(eh+1); 589 switch (l->llc_dsap) { 590 #ifdef NETATALK 591 case LLC_SNAP_LSAP: 592 switch (l->llc_control) { 593 case LLC_UI: 594 if (l->llc_ssap != LLC_SNAP_LSAP) 595 goto drop; 596 597 if (memcmp(&(l->llc_snap_org_code)[0], 598 at_org_code, sizeof(at_org_code)) == 0 && 599 ntohs(l->llc_snap_ether_type) == 600 ETHERTYPE_ATALK) { 601 inq = &atintrq2; 602 m_adj(m, sizeof(struct ether_header) 603 + sizeof(struct llc)); 604 isr = NETISR_ATALK; 605 break; 606 } 607 608 if (memcmp(&(l->llc_snap_org_code)[0], 609 aarp_org_code, 610 sizeof(aarp_org_code)) == 0 && 611 ntohs(l->llc_snap_ether_type) == 612 ETHERTYPE_AARP) { 613 m_adj(m, sizeof(struct ether_header) 614 + sizeof(struct llc)); 615 aarpinput(ifp, m); /* XXX queue? */ 616 return; 617 } 618 619 default: 620 goto drop; 621 } 622 break; 623 #endif 624 default: 625 goto drop; 626 } 627 628 KASSERT(inq != NULL); 629 IFQ_ENQUEUE_ISR(inq, m, isr); 630 return; 631 632 drop: 633 m_freem(m); 634 if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */ 635 return; 636 } 637 #endif /* defined (LLC) || defined (NETATALK) */ 638 639 /* 640 * Process a received Ethernet packet; 641 * the packet is in the mbuf chain m with 642 * the ether header. 643 */ 644 void 645 ether_input(struct ifnet *ifp, struct mbuf *m) 646 { 647 struct ethercom *ec = (struct ethercom *) ifp; 648 pktqueue_t *pktq = NULL; 649 struct ifqueue *inq = NULL; 650 uint16_t etype; 651 struct ether_header *eh; 652 size_t ehlen; 653 static int earlypkts; 654 int isr = 0; 655 #if NAGR > 0 656 void *agrprivate; 657 #endif 658 659 KASSERT(!cpu_intr_p()); 660 KASSERT((m->m_flags & M_PKTHDR) != 0); 661 662 if ((ifp->if_flags & IFF_UP) == 0) 663 goto drop; 664 665 #ifdef MBUFTRACE 666 m_claimm(m, &ec->ec_rx_mowner); 667 #endif 668 669 if (__predict_false(m->m_len < sizeof(*eh))) { 670 if ((m = m_pullup(m, sizeof(*eh))) == NULL) 671 goto dropped; 672 } 673 674 eh = mtod(m, struct ether_header *); 675 etype = ntohs(eh->ether_type); 676 ehlen = sizeof(*eh); 677 678 if (__predict_false(earlypkts < 100 || 679 entropy_epoch() == (unsigned)-1)) { 680 rnd_add_data(NULL, eh, ehlen, 0); 681 earlypkts++; 682 } 683 684 /* 685 * Determine if the packet is within its size limits. For MPLS the 686 * header length is variable, so we skip the check. 687 */ 688 if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len > 689 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 690 #ifdef DIAGNOSTIC 691 mutex_enter(&bigpktpps_lock); 692 if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count, 693 bigpktppslim)) { 694 printf("%s: discarding oversize frame (len=%d)\n", 695 ifp->if_xname, m->m_pkthdr.len); 696 } 697 mutex_exit(&bigpktpps_lock); 698 #endif 699 goto drop; 700 } 701 702 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 703 /* 704 * If this is not a simplex interface, drop the packet 705 * if it came from us. 706 */ 707 if ((ifp->if_flags & IFF_SIMPLEX) == 0 && 708 memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost, 709 ETHER_ADDR_LEN) == 0) { 710 goto drop; 711 } 712 713 if (memcmp(etherbroadcastaddr, 714 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 715 m->m_flags |= M_BCAST; 716 else 717 m->m_flags |= M_MCAST; 718 if_statinc(ifp, if_imcasts); 719 } 720 721 /* If the CRC is still on the packet, trim it off. */ 722 if (m->m_flags & M_HASFCS) { 723 m_adj(m, -ETHER_CRC_LEN); 724 m->m_flags &= ~M_HASFCS; 725 } 726 727 if_statadd(ifp, if_ibytes, m->m_pkthdr.len); 728 729 #if NCARP > 0 730 if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) { 731 /* 732 * Clear M_PROMISC, in case the packet comes from a 733 * vlan. 734 */ 735 m->m_flags &= ~M_PROMISC; 736 if (carp_input(m, (uint8_t *)&eh->ether_shost, 737 (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0) 738 return; 739 } 740 #endif 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 #if NAGR > 0 760 if (ifp->if_type != IFT_IEEE8023ADLAG) { 761 agrprivate = ifp->if_lagg; 762 } else { 763 agrprivate = NULL; 764 } 765 if (agrprivate != NULL && 766 __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) { 767 m->m_flags &= ~M_PROMISC; 768 agr_input(ifp, m); 769 return; 770 } 771 #endif 772 773 /* Handle input from a lagg(4) port */ 774 if (ifp->if_type == IFT_IEEE8023ADLAG) { 775 KASSERT(lagg_input_ethernet_p != NULL); 776 m = (*lagg_input_ethernet_p)(ifp, m); 777 if (m == NULL) 778 return; 779 } 780 781 /* 782 * If VLANs are configured on the interface, check to 783 * see if the device performed the decapsulation and 784 * provided us with the tag. 785 */ 786 if (ec->ec_nvlans && vlan_has_tag(m)) { 787 #if NVLAN > 0 788 /* 789 * vlan_input() will either recursively call ether_input() 790 * or drop the packet. 791 */ 792 vlan_input(ifp, m); 793 return; 794 #else 795 goto drop; 796 #endif 797 } 798 799 /* 800 * Handle protocols that expect to have the Ethernet header 801 * (and possibly FCS) intact. 802 */ 803 switch (etype) { 804 case ETHERTYPE_VLAN: { 805 struct ether_vlan_header *evl = (void *)eh; 806 807 /* 808 * If there is a tag of 0, then the VLAN header was probably 809 * just being used to store the priority. Extract the ether 810 * type, and if IP or IPV6, let them deal with it. 811 */ 812 if (m->m_len >= sizeof(*evl) && 813 EVL_VLANOFTAG(evl->evl_tag) == 0) { 814 etype = ntohs(evl->evl_proto); 815 ehlen = sizeof(*evl); 816 if ((m->m_flags & M_PROMISC) == 0 && 817 (etype == ETHERTYPE_IP || 818 etype == ETHERTYPE_IPV6)) 819 break; 820 } 821 822 #if NVLAN > 0 823 /* 824 * vlan_input() will either recursively call ether_input() 825 * or drop the packet. 826 */ 827 if (ec->ec_nvlans != 0) { 828 vlan_input(ifp, m); 829 return; 830 } else 831 #endif 832 goto drop; 833 } 834 835 #if NPPPOE > 0 836 case ETHERTYPE_PPPOEDISC: 837 pppoedisc_input(ifp, m); 838 return; 839 840 case ETHERTYPE_PPPOE: 841 pppoe_input(ifp, m); 842 return; 843 #endif 844 845 case ETHERTYPE_SLOWPROTOCOLS: { 846 uint8_t subtype; 847 848 if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) 849 goto drop; 850 851 m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype); 852 switch (subtype) { 853 #if NAGR > 0 854 case SLOWPROTOCOLS_SUBTYPE_LACP: 855 if (agrprivate != NULL) { 856 ieee8023ad_lacp_input(ifp, m); 857 return; 858 } 859 break; 860 861 case SLOWPROTOCOLS_SUBTYPE_MARKER: 862 if (agrprivate != NULL) { 863 ieee8023ad_marker_input(ifp, m); 864 return; 865 } 866 break; 867 #endif 868 869 default: 870 if (subtype == 0 || subtype > 10) { 871 /* illegal value */ 872 goto drop; 873 } 874 /* unknown subtype */ 875 break; 876 } 877 } 878 /* FALLTHROUGH */ 879 default: 880 if (m->m_flags & M_PROMISC) 881 goto drop; 882 } 883 884 /* If the CRC is still on the packet, trim it off. */ 885 if (m->m_flags & M_HASFCS) { 886 m_adj(m, -ETHER_CRC_LEN); 887 m->m_flags &= ~M_HASFCS; 888 } 889 890 /* etype represents the size of the payload in this case */ 891 if (etype <= ETHERMTU + sizeof(struct ether_header)) { 892 KASSERT(ehlen == sizeof(*eh)); 893 #if defined (LLC) || defined (NETATALK) 894 ether_input_llc(ifp, m, eh); 895 return; 896 #else 897 goto drop; 898 #endif 899 } 900 901 /* Strip off the Ethernet header. */ 902 m_adj(m, ehlen); 903 904 switch (etype) { 905 #ifdef INET 906 case ETHERTYPE_IP: 907 #ifdef GATEWAY 908 if (ipflow_fastforward(m)) 909 return; 910 #endif 911 pktq = ip_pktq; 912 break; 913 914 case ETHERTYPE_ARP: 915 isr = NETISR_ARP; 916 inq = &arpintrq; 917 break; 918 919 case ETHERTYPE_REVARP: 920 revarpinput(m); /* XXX queue? */ 921 return; 922 #endif 923 924 #ifdef INET6 925 case ETHERTYPE_IPV6: 926 if (__predict_false(!in6_present)) 927 goto drop; 928 #ifdef GATEWAY 929 if (ip6flow_fastforward(&m)) 930 return; 931 #endif 932 pktq = ip6_pktq; 933 break; 934 #endif 935 936 #ifdef NETATALK 937 case ETHERTYPE_ATALK: 938 isr = NETISR_ATALK; 939 inq = &atintrq1; 940 break; 941 942 case ETHERTYPE_AARP: 943 aarpinput(ifp, m); /* XXX queue? */ 944 return; 945 #endif 946 947 #ifdef MPLS 948 case ETHERTYPE_MPLS: 949 isr = NETISR_MPLS; 950 inq = &mplsintrq; 951 break; 952 #endif 953 954 default: 955 goto drop; 956 } 957 958 if (__predict_true(pktq)) { 959 #ifdef NET_MPSAFE 960 const u_int h = curcpu()->ci_index; 961 #else 962 const uint32_t h = pktq_rps_hash(m); 963 #endif 964 if (__predict_false(!pktq_enqueue(pktq, m, h))) { 965 m_freem(m); 966 } 967 return; 968 } 969 970 if (__predict_false(!inq)) { 971 /* Should not happen. */ 972 goto drop; 973 } 974 975 IFQ_ENQUEUE_ISR(inq, m, isr); 976 return; 977 978 drop: 979 m_freem(m); 980 dropped: 981 if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */ 982 } 983 984 /* 985 * Convert Ethernet address to printable (loggable) representation. 986 */ 987 char * 988 ether_sprintf(const u_char *ap) 989 { 990 static char etherbuf[3 * ETHER_ADDR_LEN]; 991 return ether_snprintf(etherbuf, sizeof(etherbuf), ap); 992 } 993 994 char * 995 ether_snprintf(char *buf, size_t len, const u_char *ap) 996 { 997 char *cp = buf; 998 size_t i; 999 1000 for (i = 0; i < len / 3; i++) { 1001 *cp++ = hexdigits[*ap >> 4]; 1002 *cp++ = hexdigits[*ap++ & 0xf]; 1003 *cp++ = ':'; 1004 } 1005 *--cp = '\0'; 1006 return buf; 1007 } 1008 1009 /* 1010 * Perform common duties while attaching to interface list 1011 */ 1012 void 1013 ether_ifattach(struct ifnet *ifp, const uint8_t *lla) 1014 { 1015 struct ethercom *ec = (struct ethercom *)ifp; 1016 char xnamebuf[HOOKNAMSIZ]; 1017 1018 ifp->if_type = IFT_ETHER; 1019 ifp->if_hdrlen = ETHER_HDR_LEN; 1020 ifp->if_dlt = DLT_EN10MB; 1021 ifp->if_mtu = ETHERMTU; 1022 ifp->if_output = ether_output; 1023 ifp->_if_input = ether_input; 1024 if (ifp->if_baudrate == 0) 1025 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1026 1027 if (lla != NULL) 1028 if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla)); 1029 1030 LIST_INIT(&ec->ec_multiaddrs); 1031 SIMPLEQ_INIT(&ec->ec_vids); 1032 ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 1033 ec->ec_flags = 0; 1034 ifp->if_broadcastaddr = etherbroadcastaddr; 1035 bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header)); 1036 snprintf(xnamebuf, sizeof(xnamebuf), 1037 "%s-ether_ifdetachhooks", ifp->if_xname); 1038 ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf); 1039 #ifdef MBUFTRACE 1040 mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx"); 1041 mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx"); 1042 MOWNER_ATTACH(&ec->ec_tx_mowner); 1043 MOWNER_ATTACH(&ec->ec_rx_mowner); 1044 ifp->if_mowner = &ec->ec_tx_mowner; 1045 #endif 1046 } 1047 1048 void 1049 ether_ifdetach(struct ifnet *ifp) 1050 { 1051 struct ethercom *ec = (void *) ifp; 1052 struct ether_multi *enm; 1053 1054 IFNET_ASSERT_UNLOCKED(ifp); 1055 /* 1056 * Prevent further calls to ioctl (for example turning off 1057 * promiscuous mode from the bridge code), which eventually can 1058 * call if_init() which can cause panics because the interface 1059 * is in the process of being detached. Return device not configured 1060 * instead. 1061 */ 1062 ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *), 1063 enxio); 1064 1065 simplehook_dohooks(ec->ec_ifdetach_hooks); 1066 KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks)); 1067 simplehook_destroy(ec->ec_ifdetach_hooks); 1068 1069 bpf_detach(ifp); 1070 1071 #if NLAGG > 0 1072 if (ifp->if_lagg) 1073 lagg_ifdetach(ifp); 1074 #endif 1075 1076 ETHER_LOCK(ec); 1077 KASSERT(ec->ec_nvlans == 0); 1078 while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) { 1079 LIST_REMOVE(enm, enm_list); 1080 kmem_free(enm, sizeof(*enm)); 1081 ec->ec_multicnt--; 1082 } 1083 ETHER_UNLOCK(ec); 1084 1085 mutex_obj_free(ec->ec_lock); 1086 ec->ec_lock = NULL; 1087 1088 ifp->if_mowner = NULL; 1089 MOWNER_DETACH(&ec->ec_rx_mowner); 1090 MOWNER_DETACH(&ec->ec_tx_mowner); 1091 } 1092 1093 void * 1094 ether_ifdetachhook_establish(struct ifnet *ifp, 1095 void (*fn)(void *), void *arg) 1096 { 1097 struct ethercom *ec; 1098 khook_t *hk; 1099 1100 if (ifp->if_type != IFT_ETHER) 1101 return NULL; 1102 1103 ec = (struct ethercom *)ifp; 1104 hk = simplehook_establish(ec->ec_ifdetach_hooks, 1105 fn, arg); 1106 1107 return (void *)hk; 1108 } 1109 1110 void 1111 ether_ifdetachhook_disestablish(struct ifnet *ifp, 1112 void *vhook, kmutex_t *lock) 1113 { 1114 struct ethercom *ec; 1115 1116 if (vhook == NULL) 1117 return; 1118 1119 ec = (struct ethercom *)ifp; 1120 simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock); 1121 } 1122 1123 #if 0 1124 /* 1125 * This is for reference. We have a table-driven version 1126 * of the little-endian crc32 generator, which is faster 1127 * than the double-loop. 1128 */ 1129 uint32_t 1130 ether_crc32_le(const uint8_t *buf, size_t len) 1131 { 1132 uint32_t c, crc, carry; 1133 size_t i, j; 1134 1135 crc = 0xffffffffU; /* initial value */ 1136 1137 for (i = 0; i < len; i++) { 1138 c = buf[i]; 1139 for (j = 0; j < 8; j++) { 1140 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1141 crc >>= 1; 1142 c >>= 1; 1143 if (carry) 1144 crc = (crc ^ ETHER_CRC_POLY_LE); 1145 } 1146 } 1147 1148 return (crc); 1149 } 1150 #else 1151 uint32_t 1152 ether_crc32_le(const uint8_t *buf, size_t len) 1153 { 1154 static const uint32_t crctab[] = { 1155 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1156 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1157 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1158 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1159 }; 1160 uint32_t crc; 1161 size_t i; 1162 1163 crc = 0xffffffffU; /* initial value */ 1164 1165 for (i = 0; i < len; i++) { 1166 crc ^= buf[i]; 1167 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1168 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1169 } 1170 1171 return (crc); 1172 } 1173 #endif 1174 1175 uint32_t 1176 ether_crc32_be(const uint8_t *buf, size_t len) 1177 { 1178 uint32_t c, crc, carry; 1179 size_t i, j; 1180 1181 crc = 0xffffffffU; /* initial value */ 1182 1183 for (i = 0; i < len; i++) { 1184 c = buf[i]; 1185 for (j = 0; j < 8; j++) { 1186 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1187 crc <<= 1; 1188 c >>= 1; 1189 if (carry) 1190 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1191 } 1192 } 1193 1194 return (crc); 1195 } 1196 1197 #ifdef INET 1198 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] = 1199 { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }; 1200 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] = 1201 { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff }; 1202 #endif 1203 #ifdef INET6 1204 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] = 1205 { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 }; 1206 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] = 1207 { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff }; 1208 #endif 1209 1210 /* 1211 * ether_aton implementation, not using a static buffer. 1212 */ 1213 int 1214 ether_aton_r(u_char *dest, size_t len, const char *str) 1215 { 1216 const u_char *cp = (const void *)str; 1217 u_char *ep; 1218 1219 #define atox(c) (((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10)) 1220 1221 if (len < ETHER_ADDR_LEN) 1222 return ENOSPC; 1223 1224 ep = dest + ETHER_ADDR_LEN; 1225 1226 while (*cp) { 1227 if (!isxdigit(*cp)) 1228 return EINVAL; 1229 1230 *dest = atox(*cp); 1231 cp++; 1232 if (isxdigit(*cp)) { 1233 *dest = (*dest << 4) | atox(*cp); 1234 cp++; 1235 } 1236 dest++; 1237 1238 if (dest == ep) 1239 return (*cp == '\0') ? 0 : ENAMETOOLONG; 1240 1241 switch (*cp) { 1242 case ':': 1243 case '-': 1244 case '.': 1245 cp++; 1246 break; 1247 } 1248 } 1249 return ENOBUFS; 1250 } 1251 1252 /* 1253 * Convert a sockaddr into an Ethernet address or range of Ethernet 1254 * addresses. 1255 */ 1256 int 1257 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN], 1258 uint8_t addrhi[ETHER_ADDR_LEN]) 1259 { 1260 #ifdef INET 1261 const struct sockaddr_in *sin; 1262 #endif 1263 #ifdef INET6 1264 const struct sockaddr_in6 *sin6; 1265 #endif 1266 1267 switch (sa->sa_family) { 1268 1269 case AF_UNSPEC: 1270 memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN); 1271 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1272 break; 1273 1274 #ifdef INET 1275 case AF_INET: 1276 sin = satocsin(sa); 1277 if (sin->sin_addr.s_addr == INADDR_ANY) { 1278 /* 1279 * An IP address of INADDR_ANY means listen to 1280 * or stop listening to all of the Ethernet 1281 * multicast addresses used for IP. 1282 * (This is for the sake of IP multicast routers.) 1283 */ 1284 memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN); 1285 memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN); 1286 } else { 1287 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo); 1288 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1289 } 1290 break; 1291 #endif 1292 #ifdef INET6 1293 case AF_INET6: 1294 sin6 = satocsin6(sa); 1295 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1296 /* 1297 * An IP6 address of 0 means listen to or stop 1298 * listening to all of the Ethernet multicast 1299 * address used for IP6. 1300 * (This is used for multicast routers.) 1301 */ 1302 memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN); 1303 memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN); 1304 } else { 1305 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo); 1306 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1307 } 1308 break; 1309 #endif 1310 1311 default: 1312 return EAFNOSUPPORT; 1313 } 1314 return 0; 1315 } 1316 1317 /* 1318 * Add an Ethernet multicast address or range of addresses to the list for a 1319 * given interface. 1320 */ 1321 int 1322 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec) 1323 { 1324 struct ether_multi *enm, *_enm; 1325 u_char addrlo[ETHER_ADDR_LEN]; 1326 u_char addrhi[ETHER_ADDR_LEN]; 1327 int error = 0; 1328 1329 /* Allocate out of lock */ 1330 enm = kmem_alloc(sizeof(*enm), KM_SLEEP); 1331 1332 ETHER_LOCK(ec); 1333 error = ether_multiaddr(sa, addrlo, addrhi); 1334 if (error != 0) 1335 goto out; 1336 1337 /* 1338 * Verify that we have valid Ethernet multicast addresses. 1339 */ 1340 if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) { 1341 error = EINVAL; 1342 goto out; 1343 } 1344 1345 /* 1346 * See if the address range is already in the list. 1347 */ 1348 _enm = ether_lookup_multi(addrlo, addrhi, ec); 1349 if (_enm != NULL) { 1350 /* 1351 * Found it; just increment the reference count. 1352 */ 1353 ++_enm->enm_refcount; 1354 error = 0; 1355 goto out; 1356 } 1357 1358 /* 1359 * Link a new multicast record into the interface's multicast list. 1360 */ 1361 memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN); 1362 memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN); 1363 enm->enm_refcount = 1; 1364 LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list); 1365 ec->ec_multicnt++; 1366 1367 /* 1368 * Return ENETRESET to inform the driver that the list has changed 1369 * and its reception filter should be adjusted accordingly. 1370 */ 1371 error = ENETRESET; 1372 enm = NULL; 1373 1374 out: 1375 ETHER_UNLOCK(ec); 1376 if (enm != NULL) 1377 kmem_free(enm, sizeof(*enm)); 1378 return error; 1379 } 1380 1381 /* 1382 * Delete a multicast address record. 1383 */ 1384 int 1385 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec) 1386 { 1387 struct ether_multi *enm; 1388 u_char addrlo[ETHER_ADDR_LEN]; 1389 u_char addrhi[ETHER_ADDR_LEN]; 1390 int error; 1391 1392 ETHER_LOCK(ec); 1393 error = ether_multiaddr(sa, addrlo, addrhi); 1394 if (error != 0) 1395 goto error; 1396 1397 /* 1398 * Look up the address in our list. 1399 */ 1400 enm = ether_lookup_multi(addrlo, addrhi, ec); 1401 if (enm == NULL) { 1402 error = ENXIO; 1403 goto error; 1404 } 1405 if (--enm->enm_refcount != 0) { 1406 /* 1407 * Still some claims to this record. 1408 */ 1409 error = 0; 1410 goto error; 1411 } 1412 1413 /* 1414 * No remaining claims to this record; unlink and free it. 1415 */ 1416 LIST_REMOVE(enm, enm_list); 1417 ec->ec_multicnt--; 1418 ETHER_UNLOCK(ec); 1419 kmem_free(enm, sizeof(*enm)); 1420 1421 /* 1422 * Return ENETRESET to inform the driver that the list has changed 1423 * and its reception filter should be adjusted accordingly. 1424 */ 1425 return ENETRESET; 1426 1427 error: 1428 ETHER_UNLOCK(ec); 1429 return error; 1430 } 1431 1432 void 1433 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb) 1434 { 1435 ec->ec_ifflags_cb = cb; 1436 } 1437 1438 void 1439 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb) 1440 { 1441 1442 ec->ec_vlan_cb = cb; 1443 } 1444 1445 static int 1446 ether_ioctl_reinit(struct ethercom *ec) 1447 { 1448 struct ifnet *ifp = &ec->ec_if; 1449 int error; 1450 1451 switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) { 1452 case IFF_RUNNING: 1453 /* 1454 * If interface is marked down and it is running, 1455 * then stop and disable it. 1456 */ 1457 (*ifp->if_stop)(ifp, 1); 1458 break; 1459 case IFF_UP: 1460 /* 1461 * If interface is marked up and it is stopped, then 1462 * start it. 1463 */ 1464 return (*ifp->if_init)(ifp); 1465 case IFF_UP | IFF_RUNNING: 1466 error = 0; 1467 if (ec->ec_ifflags_cb != NULL) { 1468 error = (*ec->ec_ifflags_cb)(ec); 1469 if (error == ENETRESET) { 1470 /* 1471 * Reset the interface to pick up 1472 * changes in any other flags that 1473 * affect the hardware state. 1474 */ 1475 return (*ifp->if_init)(ifp); 1476 } 1477 } else 1478 error = (*ifp->if_init)(ifp); 1479 return error; 1480 case 0: 1481 break; 1482 } 1483 1484 return 0; 1485 } 1486 1487 /* 1488 * Common ioctls for Ethernet interfaces. Note, we must be 1489 * called at splnet(). 1490 */ 1491 int 1492 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1493 { 1494 struct ethercom *ec = (void *)ifp; 1495 struct eccapreq *eccr; 1496 struct ifreq *ifr = (struct ifreq *)data; 1497 struct if_laddrreq *iflr = data; 1498 const struct sockaddr_dl *sdl; 1499 static const uint8_t zero[ETHER_ADDR_LEN]; 1500 int error; 1501 1502 switch (cmd) { 1503 case SIOCINITIFADDR: 1504 { 1505 struct ifaddr *ifa = (struct ifaddr *)data; 1506 if (ifa->ifa_addr->sa_family != AF_LINK 1507 && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 1508 (IFF_UP | IFF_RUNNING)) { 1509 ifp->if_flags |= IFF_UP; 1510 if ((error = (*ifp->if_init)(ifp)) != 0) 1511 return error; 1512 } 1513 #ifdef INET 1514 if (ifa->ifa_addr->sa_family == AF_INET) 1515 arp_ifinit(ifp, ifa); 1516 #endif 1517 return 0; 1518 } 1519 1520 case SIOCSIFMTU: 1521 { 1522 int maxmtu; 1523 1524 if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU) 1525 maxmtu = ETHERMTU_JUMBO; 1526 else 1527 maxmtu = ETHERMTU; 1528 1529 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu) 1530 return EINVAL; 1531 else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET) 1532 return error; 1533 else if (ifp->if_flags & IFF_UP) { 1534 /* Make sure the device notices the MTU change. */ 1535 return (*ifp->if_init)(ifp); 1536 } else 1537 return 0; 1538 } 1539 1540 case SIOCSIFFLAGS: 1541 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1542 return error; 1543 return ether_ioctl_reinit(ec); 1544 case SIOCGIFFLAGS: 1545 error = ifioctl_common(ifp, cmd, data); 1546 if (error == 0) { 1547 /* Set IFF_ALLMULTI for backcompat */ 1548 ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ? 1549 IFF_ALLMULTI : 0; 1550 } 1551 return error; 1552 case SIOCGETHERCAP: 1553 eccr = (struct eccapreq *)data; 1554 eccr->eccr_capabilities = ec->ec_capabilities; 1555 eccr->eccr_capenable = ec->ec_capenable; 1556 return 0; 1557 case SIOCSETHERCAP: 1558 eccr = (struct eccapreq *)data; 1559 if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0) 1560 return EINVAL; 1561 if (eccr->eccr_capenable == ec->ec_capenable) 1562 return 0; 1563 #if 0 /* notyet */ 1564 ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE) 1565 | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE); 1566 #else 1567 ec->ec_capenable = eccr->eccr_capenable; 1568 #endif 1569 return ether_ioctl_reinit(ec); 1570 case SIOCADDMULTI: 1571 return ether_addmulti(ifreq_getaddr(cmd, ifr), ec); 1572 case SIOCDELMULTI: 1573 return ether_delmulti(ifreq_getaddr(cmd, ifr), ec); 1574 case SIOCSIFMEDIA: 1575 case SIOCGIFMEDIA: 1576 if (ec->ec_mii != NULL) 1577 return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, 1578 cmd); 1579 else if (ec->ec_ifmedia != NULL) 1580 return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd); 1581 else 1582 return ENOTTY; 1583 break; 1584 case SIOCALIFADDR: 1585 sdl = satocsdl(sstocsa(&iflr->addr)); 1586 if (sdl->sdl_family != AF_LINK) 1587 ; 1588 else if (ETHER_IS_MULTICAST(CLLADDR(sdl))) 1589 return EINVAL; 1590 else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0) 1591 return EINVAL; 1592 /*FALLTHROUGH*/ 1593 default: 1594 return ifioctl_common(ifp, cmd, data); 1595 } 1596 return 0; 1597 } 1598 1599 /* 1600 * Enable/disable passing VLAN packets if the parent interface supports it. 1601 * Return: 1602 * 0: Ok 1603 * -1: Parent interface does not support vlans 1604 * >0: Error 1605 */ 1606 int 1607 ether_enable_vlan_mtu(struct ifnet *ifp) 1608 { 1609 int error; 1610 struct ethercom *ec = (void *)ifp; 1611 1612 /* Parent does not support VLAN's */ 1613 if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0) 1614 return -1; 1615 1616 /* 1617 * Parent supports the VLAN_MTU capability, 1618 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames; 1619 * enable it. 1620 */ 1621 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1622 1623 /* Interface is down, defer for later */ 1624 if ((ifp->if_flags & IFF_UP) == 0) 1625 return 0; 1626 1627 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1628 return 0; 1629 1630 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1631 return error; 1632 } 1633 1634 int 1635 ether_disable_vlan_mtu(struct ifnet *ifp) 1636 { 1637 int error; 1638 struct ethercom *ec = (void *)ifp; 1639 1640 /* We still have VLAN's, defer for later */ 1641 if (ec->ec_nvlans != 0) 1642 return 0; 1643 1644 /* Parent does not support VLAB's, nothing to do. */ 1645 if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0) 1646 return -1; 1647 1648 /* 1649 * Disable Tx/Rx of VLAN-sized frames. 1650 */ 1651 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1652 1653 /* Interface is down, defer for later */ 1654 if ((ifp->if_flags & IFF_UP) == 0) 1655 return 0; 1656 1657 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1658 return 0; 1659 1660 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1661 return error; 1662 } 1663 1664 static int 1665 ether_multicast_sysctl(SYSCTLFN_ARGS) 1666 { 1667 struct ether_multi *enm; 1668 struct ifnet *ifp; 1669 struct ethercom *ec; 1670 int error = 0; 1671 size_t written; 1672 struct psref psref; 1673 int bound; 1674 unsigned int multicnt; 1675 struct ether_multi_sysctl *addrs; 1676 int i; 1677 1678 if (namelen != 1) 1679 return EINVAL; 1680 1681 bound = curlwp_bind(); 1682 ifp = if_get_byindex(name[0], &psref); 1683 if (ifp == NULL) { 1684 error = ENODEV; 1685 goto out; 1686 } 1687 if (ifp->if_type != IFT_ETHER) { 1688 if_put(ifp, &psref); 1689 *oldlenp = 0; 1690 goto out; 1691 } 1692 ec = (struct ethercom *)ifp; 1693 1694 if (oldp == NULL) { 1695 if_put(ifp, &psref); 1696 *oldlenp = ec->ec_multicnt * sizeof(*addrs); 1697 goto out; 1698 } 1699 1700 /* 1701 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which 1702 * is sleepable, while holding it. Copy data to a local buffer first 1703 * with the lock taken and then call sysctl_copyout without holding it. 1704 */ 1705 retry: 1706 multicnt = ec->ec_multicnt; 1707 1708 if (multicnt == 0) { 1709 if_put(ifp, &psref); 1710 *oldlenp = 0; 1711 goto out; 1712 } 1713 1714 addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP); 1715 1716 ETHER_LOCK(ec); 1717 if (multicnt != ec->ec_multicnt) { 1718 /* The number of multicast addresses has changed */ 1719 ETHER_UNLOCK(ec); 1720 kmem_free(addrs, sizeof(*addrs) * multicnt); 1721 goto retry; 1722 } 1723 1724 i = 0; 1725 LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) { 1726 struct ether_multi_sysctl *addr = &addrs[i]; 1727 addr->enm_refcount = enm->enm_refcount; 1728 memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN); 1729 memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN); 1730 i++; 1731 } 1732 ETHER_UNLOCK(ec); 1733 1734 error = 0; 1735 written = 0; 1736 for (i = 0; i < multicnt; i++) { 1737 struct ether_multi_sysctl *addr = &addrs[i]; 1738 1739 if (written + sizeof(*addr) > *oldlenp) 1740 break; 1741 error = sysctl_copyout(l, addr, oldp, sizeof(*addr)); 1742 if (error) 1743 break; 1744 written += sizeof(*addr); 1745 oldp = (char *)oldp + sizeof(*addr); 1746 } 1747 kmem_free(addrs, sizeof(*addrs) * multicnt); 1748 1749 if_put(ifp, &psref); 1750 1751 *oldlenp = written; 1752 out: 1753 curlwp_bindx(bound); 1754 return error; 1755 } 1756 1757 static void 1758 ether_sysctl_setup(struct sysctllog **clog) 1759 { 1760 const struct sysctlnode *rnode = NULL; 1761 1762 sysctl_createv(clog, 0, NULL, &rnode, 1763 CTLFLAG_PERMANENT, 1764 CTLTYPE_NODE, "ether", 1765 SYSCTL_DESCR("Ethernet-specific information"), 1766 NULL, 0, NULL, 0, 1767 CTL_NET, CTL_CREATE, CTL_EOL); 1768 1769 sysctl_createv(clog, 0, &rnode, NULL, 1770 CTLFLAG_PERMANENT, 1771 CTLTYPE_NODE, "multicast", 1772 SYSCTL_DESCR("multicast addresses"), 1773 ether_multicast_sysctl, 0, NULL, 0, 1774 CTL_CREATE, CTL_EOL); 1775 } 1776 1777 void 1778 etherinit(void) 1779 { 1780 1781 #ifdef DIAGNOSTIC 1782 mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET); 1783 #endif 1784 ether_sysctl_setup(NULL); 1785 } 1786
Indexes created Tue Sep 30 10:09:55 GMT 2025