if_ethersubr.c revision 1.291
1 /* $NetBSD: if_ethersubr.c,v 1.291 2021/02/13 13:00:16 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 1.291 2021/02/13 13:00:16 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 #ifdef MBUFTRACE 653 m_claimm(m, &ec->ec_rx_mowner); 654 #endif 655 656 /* Enforce alignement */ 657 if (ETHER_HDR_ALIGNED_P(mtod(m, void *)) == 0) { 658 if ((m = m_copyup(m, sizeof(*eh), 0)) == NULL) 659 goto dropped; 660 } else if (__predict_false(m->m_len < sizeof(*eh))) { 661 if ((m = m_pullup(m, sizeof(*eh))) == NULL) 662 goto dropped; 663 } 664 665 eh = mtod(m, struct ether_header *); 666 etype = ntohs(eh->ether_type); 667 ehlen = sizeof(*eh); 668 669 if (__predict_false(earlypkts < 100 || 670 entropy_epoch() == (unsigned)-1)) { 671 rnd_add_data(NULL, eh, ehlen, 0); 672 earlypkts++; 673 } 674 675 /* 676 * Determine if the packet is within its size limits. For MPLS the 677 * header length is variable, so we skip the check. 678 */ 679 if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len > 680 ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) { 681 #ifdef DIAGNOSTIC 682 mutex_enter(&bigpktpps_lock); 683 if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count, 684 bigpktppslim)) { 685 printf("%s: discarding oversize frame (len=%d)\n", 686 ifp->if_xname, m->m_pkthdr.len); 687 } 688 mutex_exit(&bigpktpps_lock); 689 #endif 690 goto drop; 691 } 692 693 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 694 /* 695 * If this is not a simplex interface, drop the packet 696 * if it came from us. 697 */ 698 if ((ifp->if_flags & IFF_SIMPLEX) == 0 && 699 memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost, 700 ETHER_ADDR_LEN) == 0) { 701 goto drop; 702 } 703 704 if (memcmp(etherbroadcastaddr, 705 eh->ether_dhost, ETHER_ADDR_LEN) == 0) 706 m->m_flags |= M_BCAST; 707 else 708 m->m_flags |= M_MCAST; 709 if_statinc(ifp, if_imcasts); 710 } 711 712 /* If the CRC is still on the packet, trim it off. */ 713 if (m->m_flags & M_HASFCS) { 714 m_adj(m, -ETHER_CRC_LEN); 715 m->m_flags &= ~M_HASFCS; 716 } 717 718 if_statadd(ifp, if_ibytes, m->m_pkthdr.len); 719 720 #if NCARP > 0 721 if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) { 722 /* 723 * Clear M_PROMISC, in case the packet comes from a 724 * vlan. 725 */ 726 m->m_flags &= ~M_PROMISC; 727 if (carp_input(m, (uint8_t *)&eh->ether_shost, 728 (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0) 729 return; 730 } 731 #endif 732 733 if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 && 734 (ifp->if_flags & IFF_PROMISC) != 0 && 735 memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost, 736 ETHER_ADDR_LEN) != 0) { 737 m->m_flags |= M_PROMISC; 738 } 739 740 if ((m->m_flags & M_PROMISC) == 0) { 741 if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0) 742 return; 743 if (m == NULL) 744 return; 745 746 eh = mtod(m, struct ether_header *); 747 etype = ntohs(eh->ether_type); 748 } 749 750 #if NAGR > 0 751 if (ifp->if_agrprivate && 752 __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) { 753 m->m_flags &= ~M_PROMISC; 754 agr_input(ifp, m); 755 return; 756 } 757 #endif 758 759 /* 760 * If VLANs are configured on the interface, check to 761 * see if the device performed the decapsulation and 762 * provided us with the tag. 763 */ 764 if (ec->ec_nvlans && vlan_has_tag(m)) { 765 #if NVLAN > 0 766 /* 767 * vlan_input() will either recursively call ether_input() 768 * or drop the packet. 769 */ 770 vlan_input(ifp, m); 771 return; 772 #else 773 goto drop; 774 #endif 775 } 776 777 /* 778 * Handle protocols that expect to have the Ethernet header 779 * (and possibly FCS) intact. 780 */ 781 switch (etype) { 782 case ETHERTYPE_VLAN: { 783 struct ether_vlan_header *evl = (void *)eh; 784 785 /* 786 * If there is a tag of 0, then the VLAN header was probably 787 * just being used to store the priority. Extract the ether 788 * type, and if IP or IPV6, let them deal with it. 789 */ 790 if (m->m_len >= sizeof(*evl) && 791 EVL_VLANOFTAG(evl->evl_tag) == 0) { 792 etype = ntohs(evl->evl_proto); 793 ehlen = sizeof(*evl); 794 if ((m->m_flags & M_PROMISC) == 0 && 795 (etype == ETHERTYPE_IP || 796 etype == ETHERTYPE_IPV6)) 797 break; 798 } 799 800 #if NVLAN > 0 801 /* 802 * vlan_input() will either recursively call ether_input() 803 * or drop the packet. 804 */ 805 if (ec->ec_nvlans != 0) { 806 vlan_input(ifp, m); 807 return; 808 } else 809 #endif 810 goto drop; 811 } 812 813 #if NPPPOE > 0 814 case ETHERTYPE_PPPOEDISC: 815 pppoedisc_input(ifp, m); 816 return; 817 818 case ETHERTYPE_PPPOE: 819 pppoe_input(ifp, m); 820 return; 821 #endif 822 823 case ETHERTYPE_SLOWPROTOCOLS: { 824 uint8_t subtype; 825 826 if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) 827 goto drop; 828 829 m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype); 830 switch (subtype) { 831 #if NAGR > 0 832 case SLOWPROTOCOLS_SUBTYPE_LACP: 833 if (ifp->if_agrprivate) { 834 ieee8023ad_lacp_input(ifp, m); 835 return; 836 } 837 break; 838 839 case SLOWPROTOCOLS_SUBTYPE_MARKER: 840 if (ifp->if_agrprivate) { 841 ieee8023ad_marker_input(ifp, m); 842 return; 843 } 844 break; 845 #endif 846 847 default: 848 if (subtype == 0 || subtype > 10) { 849 /* illegal value */ 850 goto drop; 851 } 852 /* unknown subtype */ 853 break; 854 } 855 } 856 /* FALLTHROUGH */ 857 default: 858 if (m->m_flags & M_PROMISC) 859 goto drop; 860 } 861 862 /* If the CRC is still on the packet, trim it off. */ 863 if (m->m_flags & M_HASFCS) { 864 m_adj(m, -ETHER_CRC_LEN); 865 m->m_flags &= ~M_HASFCS; 866 } 867 868 /* etype represents the size of the payload in this case */ 869 if (etype <= ETHERMTU + sizeof(struct ether_header)) { 870 KASSERT(ehlen == sizeof(*eh)); 871 #if defined (LLC) || defined (NETATALK) 872 ether_input_llc(ifp, m, eh); 873 return; 874 #else 875 goto drop; 876 #endif 877 } 878 879 /* Strip off the Ethernet header. */ 880 m_adj(m, ehlen); 881 882 switch (etype) { 883 #ifdef INET 884 case ETHERTYPE_IP: 885 #ifdef GATEWAY 886 if (ipflow_fastforward(m)) 887 return; 888 #endif 889 pktq = ip_pktq; 890 break; 891 892 case ETHERTYPE_ARP: 893 isr = NETISR_ARP; 894 inq = &arpintrq; 895 break; 896 897 case ETHERTYPE_REVARP: 898 revarpinput(m); /* XXX queue? */ 899 return; 900 #endif 901 902 #ifdef INET6 903 case ETHERTYPE_IPV6: 904 if (__predict_false(!in6_present)) 905 goto drop; 906 #ifdef GATEWAY 907 if (ip6flow_fastforward(&m)) 908 return; 909 #endif 910 pktq = ip6_pktq; 911 break; 912 #endif 913 914 #ifdef NETATALK 915 case ETHERTYPE_ATALK: 916 isr = NETISR_ATALK; 917 inq = &atintrq1; 918 break; 919 920 case ETHERTYPE_AARP: 921 aarpinput(ifp, m); /* XXX queue? */ 922 return; 923 #endif 924 925 #ifdef MPLS 926 case ETHERTYPE_MPLS: 927 isr = NETISR_MPLS; 928 inq = &mplsintrq; 929 break; 930 #endif 931 932 default: 933 goto drop; 934 } 935 936 if (__predict_true(pktq)) { 937 #ifdef NET_MPSAFE 938 const u_int h = curcpu()->ci_index; 939 #else 940 const uint32_t h = pktq_rps_hash(m); 941 #endif 942 if (__predict_false(!pktq_enqueue(pktq, m, h))) { 943 m_freem(m); 944 } 945 return; 946 } 947 948 if (__predict_false(!inq)) { 949 /* Should not happen. */ 950 goto drop; 951 } 952 953 IFQ_ENQUEUE_ISR(inq, m, isr); 954 return; 955 956 drop: 957 m_freem(m); 958 dropped: 959 if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */ 960 } 961 962 /* 963 * Convert Ethernet address to printable (loggable) representation. 964 */ 965 char * 966 ether_sprintf(const u_char *ap) 967 { 968 static char etherbuf[3 * ETHER_ADDR_LEN]; 969 return ether_snprintf(etherbuf, sizeof(etherbuf), ap); 970 } 971 972 char * 973 ether_snprintf(char *buf, size_t len, const u_char *ap) 974 { 975 char *cp = buf; 976 size_t i; 977 978 for (i = 0; i < len / 3; i++) { 979 *cp++ = hexdigits[*ap >> 4]; 980 *cp++ = hexdigits[*ap++ & 0xf]; 981 *cp++ = ':'; 982 } 983 *--cp = '\0'; 984 return buf; 985 } 986 987 static void 988 ether_link_state_changed(struct ifnet *ifp, int link_state) 989 { 990 #if NVLAN > 0 991 struct ethercom *ec = (void *)ifp; 992 993 if (ec->ec_nvlans) 994 vlan_link_state_changed(ifp, link_state); 995 #endif 996 } 997 998 /* 999 * Perform common duties while attaching to interface list 1000 */ 1001 void 1002 ether_ifattach(struct ifnet *ifp, const uint8_t *lla) 1003 { 1004 struct ethercom *ec = (struct ethercom *)ifp; 1005 1006 ifp->if_type = IFT_ETHER; 1007 ifp->if_hdrlen = ETHER_HDR_LEN; 1008 ifp->if_dlt = DLT_EN10MB; 1009 ifp->if_mtu = ETHERMTU; 1010 ifp->if_output = ether_output; 1011 ifp->_if_input = ether_input; 1012 ifp->if_link_state_changed = ether_link_state_changed; 1013 if (ifp->if_baudrate == 0) 1014 ifp->if_baudrate = IF_Mbps(10); /* just a default */ 1015 1016 if (lla != NULL) 1017 if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla)); 1018 1019 LIST_INIT(&ec->ec_multiaddrs); 1020 SIMPLEQ_INIT(&ec->ec_vids); 1021 ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET); 1022 ec->ec_flags = 0; 1023 ifp->if_broadcastaddr = etherbroadcastaddr; 1024 bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header)); 1025 #ifdef MBUFTRACE 1026 mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx"); 1027 mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx"); 1028 MOWNER_ATTACH(&ec->ec_tx_mowner); 1029 MOWNER_ATTACH(&ec->ec_rx_mowner); 1030 ifp->if_mowner = &ec->ec_tx_mowner; 1031 #endif 1032 } 1033 1034 void 1035 ether_ifdetach(struct ifnet *ifp) 1036 { 1037 struct ethercom *ec = (void *) ifp; 1038 struct ether_multi *enm; 1039 1040 IFNET_ASSERT_UNLOCKED(ifp); 1041 /* 1042 * Prevent further calls to ioctl (for example turning off 1043 * promiscuous mode from the bridge code), which eventually can 1044 * call if_init() which can cause panics because the interface 1045 * is in the process of being detached. Return device not configured 1046 * instead. 1047 */ 1048 ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *), 1049 enxio); 1050 1051 #if NBRIDGE > 0 1052 if (ifp->if_bridge) 1053 bridge_ifdetach(ifp); 1054 #endif 1055 bpf_detach(ifp); 1056 #if NVLAN > 0 1057 if (ec->ec_nvlans) 1058 vlan_ifdetach(ifp); 1059 #endif 1060 1061 ETHER_LOCK(ec); 1062 KASSERT(ec->ec_nvlans == 0); 1063 while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) { 1064 LIST_REMOVE(enm, enm_list); 1065 kmem_free(enm, sizeof(*enm)); 1066 ec->ec_multicnt--; 1067 } 1068 ETHER_UNLOCK(ec); 1069 1070 mutex_obj_free(ec->ec_lock); 1071 ec->ec_lock = NULL; 1072 1073 ifp->if_mowner = NULL; 1074 MOWNER_DETACH(&ec->ec_rx_mowner); 1075 MOWNER_DETACH(&ec->ec_tx_mowner); 1076 } 1077 1078 #if 0 1079 /* 1080 * This is for reference. We have a table-driven version 1081 * of the little-endian crc32 generator, which is faster 1082 * than the double-loop. 1083 */ 1084 uint32_t 1085 ether_crc32_le(const uint8_t *buf, size_t len) 1086 { 1087 uint32_t c, crc, carry; 1088 size_t i, j; 1089 1090 crc = 0xffffffffU; /* initial value */ 1091 1092 for (i = 0; i < len; i++) { 1093 c = buf[i]; 1094 for (j = 0; j < 8; j++) { 1095 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 1096 crc >>= 1; 1097 c >>= 1; 1098 if (carry) 1099 crc = (crc ^ ETHER_CRC_POLY_LE); 1100 } 1101 } 1102 1103 return (crc); 1104 } 1105 #else 1106 uint32_t 1107 ether_crc32_le(const uint8_t *buf, size_t len) 1108 { 1109 static const uint32_t crctab[] = { 1110 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 1111 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 1112 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 1113 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 1114 }; 1115 uint32_t crc; 1116 size_t i; 1117 1118 crc = 0xffffffffU; /* initial value */ 1119 1120 for (i = 0; i < len; i++) { 1121 crc ^= buf[i]; 1122 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1123 crc = (crc >> 4) ^ crctab[crc & 0xf]; 1124 } 1125 1126 return (crc); 1127 } 1128 #endif 1129 1130 uint32_t 1131 ether_crc32_be(const uint8_t *buf, size_t len) 1132 { 1133 uint32_t c, crc, carry; 1134 size_t i, j; 1135 1136 crc = 0xffffffffU; /* initial value */ 1137 1138 for (i = 0; i < len; i++) { 1139 c = buf[i]; 1140 for (j = 0; j < 8; j++) { 1141 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 1142 crc <<= 1; 1143 c >>= 1; 1144 if (carry) 1145 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 1146 } 1147 } 1148 1149 return (crc); 1150 } 1151 1152 #ifdef INET 1153 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] = 1154 { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 }; 1155 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] = 1156 { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff }; 1157 #endif 1158 #ifdef INET6 1159 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] = 1160 { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 }; 1161 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] = 1162 { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff }; 1163 #endif 1164 1165 /* 1166 * ether_aton implementation, not using a static buffer. 1167 */ 1168 int 1169 ether_aton_r(u_char *dest, size_t len, const char *str) 1170 { 1171 const u_char *cp = (const void *)str; 1172 u_char *ep; 1173 1174 #define atox(c) (((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10)) 1175 1176 if (len < ETHER_ADDR_LEN) 1177 return ENOSPC; 1178 1179 ep = dest + ETHER_ADDR_LEN; 1180 1181 while (*cp) { 1182 if (!isxdigit(*cp)) 1183 return EINVAL; 1184 1185 *dest = atox(*cp); 1186 cp++; 1187 if (isxdigit(*cp)) { 1188 *dest = (*dest << 4) | atox(*cp); 1189 cp++; 1190 } 1191 dest++; 1192 1193 if (dest == ep) 1194 return (*cp == '\0') ? 0 : ENAMETOOLONG; 1195 1196 switch (*cp) { 1197 case ':': 1198 case '-': 1199 case '.': 1200 cp++; 1201 break; 1202 } 1203 } 1204 return ENOBUFS; 1205 } 1206 1207 /* 1208 * Convert a sockaddr into an Ethernet address or range of Ethernet 1209 * addresses. 1210 */ 1211 int 1212 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN], 1213 uint8_t addrhi[ETHER_ADDR_LEN]) 1214 { 1215 #ifdef INET 1216 const struct sockaddr_in *sin; 1217 #endif 1218 #ifdef INET6 1219 const struct sockaddr_in6 *sin6; 1220 #endif 1221 1222 switch (sa->sa_family) { 1223 1224 case AF_UNSPEC: 1225 memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN); 1226 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1227 break; 1228 1229 #ifdef INET 1230 case AF_INET: 1231 sin = satocsin(sa); 1232 if (sin->sin_addr.s_addr == INADDR_ANY) { 1233 /* 1234 * An IP address of INADDR_ANY means listen to 1235 * or stop listening to all of the Ethernet 1236 * multicast addresses used for IP. 1237 * (This is for the sake of IP multicast routers.) 1238 */ 1239 memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN); 1240 memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN); 1241 } else { 1242 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo); 1243 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1244 } 1245 break; 1246 #endif 1247 #ifdef INET6 1248 case AF_INET6: 1249 sin6 = satocsin6(sa); 1250 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1251 /* 1252 * An IP6 address of 0 means listen to or stop 1253 * listening to all of the Ethernet multicast 1254 * address used for IP6. 1255 * (This is used for multicast routers.) 1256 */ 1257 memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN); 1258 memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN); 1259 } else { 1260 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo); 1261 memcpy(addrhi, addrlo, ETHER_ADDR_LEN); 1262 } 1263 break; 1264 #endif 1265 1266 default: 1267 return EAFNOSUPPORT; 1268 } 1269 return 0; 1270 } 1271 1272 /* 1273 * Add an Ethernet multicast address or range of addresses to the list for a 1274 * given interface. 1275 */ 1276 int 1277 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec) 1278 { 1279 struct ether_multi *enm, *_enm; 1280 u_char addrlo[ETHER_ADDR_LEN]; 1281 u_char addrhi[ETHER_ADDR_LEN]; 1282 int error = 0; 1283 1284 /* Allocate out of lock */ 1285 enm = kmem_alloc(sizeof(*enm), KM_SLEEP); 1286 1287 ETHER_LOCK(ec); 1288 error = ether_multiaddr(sa, addrlo, addrhi); 1289 if (error != 0) 1290 goto out; 1291 1292 /* 1293 * Verify that we have valid Ethernet multicast addresses. 1294 */ 1295 if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) { 1296 error = EINVAL; 1297 goto out; 1298 } 1299 1300 /* 1301 * See if the address range is already in the list. 1302 */ 1303 _enm = ether_lookup_multi(addrlo, addrhi, ec); 1304 if (_enm != NULL) { 1305 /* 1306 * Found it; just increment the reference count. 1307 */ 1308 ++_enm->enm_refcount; 1309 error = 0; 1310 goto out; 1311 } 1312 1313 /* 1314 * Link a new multicast record into the interface's multicast list. 1315 */ 1316 memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN); 1317 memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN); 1318 enm->enm_refcount = 1; 1319 LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list); 1320 ec->ec_multicnt++; 1321 1322 /* 1323 * Return ENETRESET to inform the driver that the list has changed 1324 * and its reception filter should be adjusted accordingly. 1325 */ 1326 error = ENETRESET; 1327 enm = NULL; 1328 1329 out: 1330 ETHER_UNLOCK(ec); 1331 if (enm != NULL) 1332 kmem_free(enm, sizeof(*enm)); 1333 return error; 1334 } 1335 1336 /* 1337 * Delete a multicast address record. 1338 */ 1339 int 1340 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec) 1341 { 1342 struct ether_multi *enm; 1343 u_char addrlo[ETHER_ADDR_LEN]; 1344 u_char addrhi[ETHER_ADDR_LEN]; 1345 int error; 1346 1347 ETHER_LOCK(ec); 1348 error = ether_multiaddr(sa, addrlo, addrhi); 1349 if (error != 0) 1350 goto error; 1351 1352 /* 1353 * Look up the address in our list. 1354 */ 1355 enm = ether_lookup_multi(addrlo, addrhi, ec); 1356 if (enm == NULL) { 1357 error = ENXIO; 1358 goto error; 1359 } 1360 if (--enm->enm_refcount != 0) { 1361 /* 1362 * Still some claims to this record. 1363 */ 1364 error = 0; 1365 goto error; 1366 } 1367 1368 /* 1369 * No remaining claims to this record; unlink and free it. 1370 */ 1371 LIST_REMOVE(enm, enm_list); 1372 ec->ec_multicnt--; 1373 ETHER_UNLOCK(ec); 1374 kmem_free(enm, sizeof(*enm)); 1375 1376 /* 1377 * Return ENETRESET to inform the driver that the list has changed 1378 * and its reception filter should be adjusted accordingly. 1379 */ 1380 return ENETRESET; 1381 1382 error: 1383 ETHER_UNLOCK(ec); 1384 return error; 1385 } 1386 1387 void 1388 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb) 1389 { 1390 ec->ec_ifflags_cb = cb; 1391 } 1392 1393 void 1394 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb) 1395 { 1396 1397 ec->ec_vlan_cb = cb; 1398 } 1399 1400 static int 1401 ether_ioctl_reinit(struct ethercom *ec) 1402 { 1403 struct ifnet *ifp = &ec->ec_if; 1404 int error; 1405 1406 switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) { 1407 case IFF_RUNNING: 1408 /* 1409 * If interface is marked down and it is running, 1410 * then stop and disable it. 1411 */ 1412 (*ifp->if_stop)(ifp, 1); 1413 break; 1414 case IFF_UP: 1415 /* 1416 * If interface is marked up and it is stopped, then 1417 * start it. 1418 */ 1419 return (*ifp->if_init)(ifp); 1420 case IFF_UP | IFF_RUNNING: 1421 error = 0; 1422 if (ec->ec_ifflags_cb != NULL) { 1423 error = (*ec->ec_ifflags_cb)(ec); 1424 if (error == ENETRESET) { 1425 /* 1426 * Reset the interface to pick up 1427 * changes in any other flags that 1428 * affect the hardware state. 1429 */ 1430 return (*ifp->if_init)(ifp); 1431 } 1432 } else 1433 error = (*ifp->if_init)(ifp); 1434 return error; 1435 case 0: 1436 break; 1437 } 1438 1439 return 0; 1440 } 1441 1442 /* 1443 * Common ioctls for Ethernet interfaces. Note, we must be 1444 * called at splnet(). 1445 */ 1446 int 1447 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1448 { 1449 struct ethercom *ec = (void *)ifp; 1450 struct eccapreq *eccr; 1451 struct ifreq *ifr = (struct ifreq *)data; 1452 struct if_laddrreq *iflr = data; 1453 const struct sockaddr_dl *sdl; 1454 static const uint8_t zero[ETHER_ADDR_LEN]; 1455 int error; 1456 1457 switch (cmd) { 1458 case SIOCINITIFADDR: 1459 { 1460 struct ifaddr *ifa = (struct ifaddr *)data; 1461 if (ifa->ifa_addr->sa_family != AF_LINK 1462 && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 1463 (IFF_UP | IFF_RUNNING)) { 1464 ifp->if_flags |= IFF_UP; 1465 if ((error = (*ifp->if_init)(ifp)) != 0) 1466 return error; 1467 } 1468 #ifdef INET 1469 if (ifa->ifa_addr->sa_family == AF_INET) 1470 arp_ifinit(ifp, ifa); 1471 #endif 1472 return 0; 1473 } 1474 1475 case SIOCSIFMTU: 1476 { 1477 int maxmtu; 1478 1479 if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU) 1480 maxmtu = ETHERMTU_JUMBO; 1481 else 1482 maxmtu = ETHERMTU; 1483 1484 if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu) 1485 return EINVAL; 1486 else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET) 1487 return error; 1488 else if (ifp->if_flags & IFF_UP) { 1489 /* Make sure the device notices the MTU change. */ 1490 return (*ifp->if_init)(ifp); 1491 } else 1492 return 0; 1493 } 1494 1495 case SIOCSIFFLAGS: 1496 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1497 return error; 1498 return ether_ioctl_reinit(ec); 1499 case SIOCGIFFLAGS: 1500 error = ifioctl_common(ifp, cmd, data); 1501 if (error == 0) { 1502 /* Set IFF_ALLMULTI for backcompat */ 1503 ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ? 1504 IFF_ALLMULTI : 0; 1505 } 1506 return error; 1507 case SIOCGETHERCAP: 1508 eccr = (struct eccapreq *)data; 1509 eccr->eccr_capabilities = ec->ec_capabilities; 1510 eccr->eccr_capenable = ec->ec_capenable; 1511 return 0; 1512 case SIOCSETHERCAP: 1513 eccr = (struct eccapreq *)data; 1514 if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0) 1515 return EINVAL; 1516 if (eccr->eccr_capenable == ec->ec_capenable) 1517 return 0; 1518 #if 0 /* notyet */ 1519 ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE) 1520 | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE); 1521 #else 1522 ec->ec_capenable = eccr->eccr_capenable; 1523 #endif 1524 return ether_ioctl_reinit(ec); 1525 case SIOCADDMULTI: 1526 return ether_addmulti(ifreq_getaddr(cmd, ifr), ec); 1527 case SIOCDELMULTI: 1528 return ether_delmulti(ifreq_getaddr(cmd, ifr), ec); 1529 case SIOCSIFMEDIA: 1530 case SIOCGIFMEDIA: 1531 if (ec->ec_mii != NULL) 1532 return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, 1533 cmd); 1534 else if (ec->ec_ifmedia != NULL) 1535 return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd); 1536 else 1537 return ENOTTY; 1538 break; 1539 case SIOCALIFADDR: 1540 sdl = satocsdl(sstocsa(&iflr->addr)); 1541 if (sdl->sdl_family != AF_LINK) 1542 ; 1543 else if (ETHER_IS_MULTICAST(CLLADDR(sdl))) 1544 return EINVAL; 1545 else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0) 1546 return EINVAL; 1547 /*FALLTHROUGH*/ 1548 default: 1549 return ifioctl_common(ifp, cmd, data); 1550 } 1551 return 0; 1552 } 1553 1554 /* 1555 * Enable/disable passing VLAN packets if the parent interface supports it. 1556 * Return: 1557 * 0: Ok 1558 * -1: Parent interface does not support vlans 1559 * >0: Error 1560 */ 1561 int 1562 ether_enable_vlan_mtu(struct ifnet *ifp) 1563 { 1564 int error; 1565 struct ethercom *ec = (void *)ifp; 1566 1567 /* Parent does not support VLAN's */ 1568 if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0) 1569 return -1; 1570 1571 /* 1572 * Parent supports the VLAN_MTU capability, 1573 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames; 1574 * enable it. 1575 */ 1576 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1577 1578 /* Interface is down, defer for later */ 1579 if ((ifp->if_flags & IFF_UP) == 0) 1580 return 0; 1581 1582 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1583 return 0; 1584 1585 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1586 return error; 1587 } 1588 1589 int 1590 ether_disable_vlan_mtu(struct ifnet *ifp) 1591 { 1592 int error; 1593 struct ethercom *ec = (void *)ifp; 1594 1595 /* We still have VLAN's, defer for later */ 1596 if (ec->ec_nvlans != 0) 1597 return 0; 1598 1599 /* Parent does not support VLAB's, nothing to do. */ 1600 if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0) 1601 return -1; 1602 1603 /* 1604 * Disable Tx/Rx of VLAN-sized frames. 1605 */ 1606 ec->ec_capenable &= ~ETHERCAP_VLAN_MTU; 1607 1608 /* Interface is down, defer for later */ 1609 if ((ifp->if_flags & IFF_UP) == 0) 1610 return 0; 1611 1612 if ((error = if_flags_set(ifp, ifp->if_flags)) == 0) 1613 return 0; 1614 1615 ec->ec_capenable |= ETHERCAP_VLAN_MTU; 1616 return error; 1617 } 1618 1619 static int 1620 ether_multicast_sysctl(SYSCTLFN_ARGS) 1621 { 1622 struct ether_multi *enm; 1623 struct ifnet *ifp; 1624 struct ethercom *ec; 1625 int error = 0; 1626 size_t written; 1627 struct psref psref; 1628 int bound; 1629 unsigned int multicnt; 1630 struct ether_multi_sysctl *addrs; 1631 int i; 1632 1633 if (namelen != 1) 1634 return EINVAL; 1635 1636 bound = curlwp_bind(); 1637 ifp = if_get_byindex(name[0], &psref); 1638 if (ifp == NULL) { 1639 error = ENODEV; 1640 goto out; 1641 } 1642 if (ifp->if_type != IFT_ETHER) { 1643 if_put(ifp, &psref); 1644 *oldlenp = 0; 1645 goto out; 1646 } 1647 ec = (struct ethercom *)ifp; 1648 1649 if (oldp == NULL) { 1650 if_put(ifp, &psref); 1651 *oldlenp = ec->ec_multicnt * sizeof(*addrs); 1652 goto out; 1653 } 1654 1655 /* 1656 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which 1657 * is sleepable, while holding it. Copy data to a local buffer first 1658 * with the lock taken and then call sysctl_copyout without holding it. 1659 */ 1660 retry: 1661 multicnt = ec->ec_multicnt; 1662 1663 if (multicnt == 0) { 1664 if_put(ifp, &psref); 1665 *oldlenp = 0; 1666 goto out; 1667 } 1668 1669 addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP); 1670 1671 ETHER_LOCK(ec); 1672 if (multicnt != ec->ec_multicnt) { 1673 /* The number of multicast addresses has changed */ 1674 ETHER_UNLOCK(ec); 1675 kmem_free(addrs, sizeof(*addrs) * multicnt); 1676 goto retry; 1677 } 1678 1679 i = 0; 1680 LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) { 1681 struct ether_multi_sysctl *addr = &addrs[i]; 1682 addr->enm_refcount = enm->enm_refcount; 1683 memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN); 1684 memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN); 1685 i++; 1686 } 1687 ETHER_UNLOCK(ec); 1688 1689 error = 0; 1690 written = 0; 1691 for (i = 0; i < multicnt; i++) { 1692 struct ether_multi_sysctl *addr = &addrs[i]; 1693 1694 if (written + sizeof(*addr) > *oldlenp) 1695 break; 1696 error = sysctl_copyout(l, addr, oldp, sizeof(*addr)); 1697 if (error) 1698 break; 1699 written += sizeof(*addr); 1700 oldp = (char *)oldp + sizeof(*addr); 1701 } 1702 kmem_free(addrs, sizeof(*addrs) * multicnt); 1703 1704 if_put(ifp, &psref); 1705 1706 *oldlenp = written; 1707 out: 1708 curlwp_bindx(bound); 1709 return error; 1710 } 1711 1712 static void 1713 ether_sysctl_setup(struct sysctllog **clog) 1714 { 1715 const struct sysctlnode *rnode = NULL; 1716 1717 sysctl_createv(clog, 0, NULL, &rnode, 1718 CTLFLAG_PERMANENT, 1719 CTLTYPE_NODE, "ether", 1720 SYSCTL_DESCR("Ethernet-specific information"), 1721 NULL, 0, NULL, 0, 1722 CTL_NET, CTL_CREATE, CTL_EOL); 1723 1724 sysctl_createv(clog, 0, &rnode, NULL, 1725 CTLFLAG_PERMANENT, 1726 CTLTYPE_NODE, "multicast", 1727 SYSCTL_DESCR("multicast addresses"), 1728 ether_multicast_sysctl, 0, NULL, 0, 1729 CTL_CREATE, CTL_EOL); 1730 } 1731 1732 void 1733 etherinit(void) 1734 { 1735 1736 #ifdef DIAGNOSTIC 1737 mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET); 1738 #endif 1739 ether_sysctl_setup(NULL); 1740 } 1741
Indexes created Tue Sep 23 01:09:51 GMT 2025