1 /* 2 * Copyright (c) 1982, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 34 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $ 35 */ 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 #include "opt_ipx.h" 40 #include "opt_mpls.h" 41 #include "opt_netgraph.h" 42 #include "opt_carp.h" 43 #include "opt_rss.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/globaldata.h> 48 #include <sys/kernel.h> 49 #include <sys/ktr.h> 50 #include <sys/lock.h> 51 #include <sys/malloc.h> 52 #include <sys/mbuf.h> 53 #include <sys/msgport.h> 54 #include <sys/socket.h> 55 #include <sys/sockio.h> 56 #include <sys/sysctl.h> 57 #include <sys/thread.h> 58 59 #include <sys/thread2.h> 60 #include <sys/mplock2.h> 61 62 #include <net/if.h> 63 #include <net/netisr.h> 64 #include <net/route.h> 65 #include <net/if_llc.h> 66 #include <net/if_dl.h> 67 #include <net/if_types.h> 68 #include <net/ifq_var.h> 69 #include <net/bpf.h> 70 #include <net/ethernet.h> 71 #include <net/vlan/if_vlan_ether.h> 72 #include <net/netmsg2.h> 73 74 #if defined(INET) || defined(INET6) 75 #include <netinet/in.h> 76 #include <netinet/ip_var.h> 77 #include <netinet/if_ether.h> 78 #include <netinet/ip_flow.h> 79 #include <net/ipfw/ip_fw.h> 80 #include <net/dummynet/ip_dummynet.h> 81 #endif 82 #ifdef INET6 83 #include <netinet6/nd6.h> 84 #endif 85 86 #ifdef CARP 87 #include <netinet/ip_carp.h> 88 #endif 89 90 #ifdef IPX 91 #include <netproto/ipx/ipx.h> 92 #include <netproto/ipx/ipx_if.h> 93 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 94 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 95 short *tp, int *hlen); 96 #endif 97 98 #ifdef MPLS 99 #include <netproto/mpls/mpls.h> 100 #endif 101 102 /* netgraph node hooks for ng_ether(4) */ 103 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 104 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 105 struct mbuf *m, const struct ether_header *eh); 106 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 107 void (*ng_ether_attach_p)(struct ifnet *ifp); 108 void (*ng_ether_detach_p)(struct ifnet *ifp); 109 110 void (*vlan_input_p)(struct mbuf *); 111 112 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 113 struct rtentry *); 114 static void ether_restore_header(struct mbuf **, const struct ether_header *, 115 const struct ether_header *); 116 static int ether_characterize(struct mbuf **); 117 118 /* 119 * if_bridge support 120 */ 121 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 122 int (*bridge_output_p)(struct ifnet *, struct mbuf *); 123 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 124 struct ifnet *(*bridge_interface_p)(void *if_bridge); 125 126 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 127 struct sockaddr *); 128 129 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 130 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 131 }; 132 133 #define gotoerr(e) do { error = (e); goto bad; } while (0) 134 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 135 136 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 137 struct ip_fw **rule, 138 const struct ether_header *eh); 139 140 static int ether_ipfw; 141 static u_long ether_restore_hdr; 142 static u_long ether_prepend_hdr; 143 static u_long ether_input_wronghash; 144 static int ether_debug; 145 146 #ifdef RSS_DEBUG 147 static u_long ether_pktinfo_try; 148 static u_long ether_pktinfo_hit; 149 static u_long ether_rss_nopi; 150 static u_long ether_rss_nohash; 151 static u_long ether_input_requeue; 152 #endif 153 154 SYSCTL_DECL(_net_link); 155 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 156 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 157 ðer_debug, 0, "Ether debug"); 158 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 159 ðer_ipfw, 0, "Pass ether pkts through firewall"); 160 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 161 ðer_restore_hdr, 0, "# of ether header restoration"); 162 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 163 ðer_prepend_hdr, 0, 164 "# of ether header restoration which prepends mbuf"); 165 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 166 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 167 #ifdef RSS_DEBUG 168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 169 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 171 ðer_rss_nohash, 0, "# of packets do not have hash"); 172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 173 ðer_pktinfo_try, 0, 174 "# of tries to find packets' msgport using pktinfo"); 175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 176 ðer_pktinfo_hit, 0, 177 "# of packets whose msgport are found using pktinfo"); 178 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 179 ðer_input_requeue, 0, "# of input packets gets requeued"); 180 #endif 181 182 #define ETHER_KTR_STR "ifp=%p" 183 #define ETHER_KTR_ARG_SIZE (sizeof(void *)) 184 #ifndef KTR_ETHERNET 185 #define KTR_ETHERNET KTR_ALL 186 #endif 187 KTR_INFO_MASTER(ether); 188 KTR_INFO(KTR_ETHERNET, ether, chain_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 189 KTR_INFO(KTR_ETHERNET, ether, chain_end, 1, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 190 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 191 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 192 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 193 194 /* 195 * Ethernet output routine. 196 * Encapsulate a packet of type family for the local net. 197 * Use trailer local net encapsulation if enough data in first 198 * packet leaves a multiple of 512 bytes of data in remainder. 199 * Assumes that ifp is actually pointer to arpcom structure. 200 */ 201 static int 202 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 203 struct rtentry *rt) 204 { 205 struct ether_header *eh, *deh; 206 u_char *edst; 207 int loop_copy = 0; 208 int hlen = ETHER_HDR_LEN; /* link layer header length */ 209 struct arpcom *ac = IFP2AC(ifp); 210 int error; 211 212 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 213 214 if (ifp->if_flags & IFF_MONITOR) 215 gotoerr(ENETDOWN); 216 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 217 gotoerr(ENETDOWN); 218 219 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 220 if (m == NULL) 221 return (ENOBUFS); 222 eh = mtod(m, struct ether_header *); 223 edst = eh->ether_dhost; 224 225 /* 226 * Fill in the destination ethernet address and frame type. 227 */ 228 switch (dst->sa_family) { 229 #ifdef INET 230 case AF_INET: 231 if (!arpresolve(ifp, rt, m, dst, edst)) 232 return (0); /* if not yet resolved */ 233 #ifdef MPLS 234 if (m->m_flags & M_MPLSLABELED) 235 eh->ether_type = htons(ETHERTYPE_MPLS); 236 else 237 #endif 238 eh->ether_type = htons(ETHERTYPE_IP); 239 break; 240 #endif 241 #ifdef INET6 242 case AF_INET6: 243 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 244 return (0); /* Something bad happenned. */ 245 eh->ether_type = htons(ETHERTYPE_IPV6); 246 break; 247 #endif 248 #ifdef IPX 249 case AF_IPX: 250 if (ef_outputp != NULL) { 251 /* 252 * Hold BGL and recheck ef_outputp 253 */ 254 get_mplock(); 255 if (ef_outputp != NULL) { 256 error = ef_outputp(ifp, &m, dst, 257 &eh->ether_type, &hlen); 258 rel_mplock(); 259 if (error) 260 goto bad; 261 else 262 break; 263 } 264 rel_mplock(); 265 } 266 eh->ether_type = htons(ETHERTYPE_IPX); 267 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 268 edst, ETHER_ADDR_LEN); 269 break; 270 #endif 271 case pseudo_AF_HDRCMPLT: 272 case AF_UNSPEC: 273 loop_copy = -1; /* if this is for us, don't do it */ 274 deh = (struct ether_header *)dst->sa_data; 275 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 276 eh->ether_type = deh->ether_type; 277 break; 278 279 default: 280 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 281 gotoerr(EAFNOSUPPORT); 282 } 283 284 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 285 memcpy(eh->ether_shost, 286 ((struct ether_header *)dst->sa_data)->ether_shost, 287 ETHER_ADDR_LEN); 288 else 289 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 290 291 /* 292 * Bridges require special output handling. 293 */ 294 if (ifp->if_bridge) { 295 KASSERT(bridge_output_p != NULL, 296 ("%s: if_bridge not loaded!", __func__)); 297 return bridge_output_p(ifp, m); 298 } 299 300 /* 301 * If a simplex interface, and the packet is being sent to our 302 * Ethernet address or a broadcast address, loopback a copy. 303 * XXX To make a simplex device behave exactly like a duplex 304 * device, we should copy in the case of sending to our own 305 * ethernet address (thus letting the original actually appear 306 * on the wire). However, we don't do that here for security 307 * reasons and compatibility with the original behavior. 308 */ 309 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 310 int csum_flags = 0; 311 312 if (m->m_pkthdr.csum_flags & CSUM_IP) 313 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 314 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 315 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 316 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 317 struct mbuf *n; 318 319 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 320 n->m_pkthdr.csum_flags |= csum_flags; 321 if (csum_flags & CSUM_DATA_VALID) 322 n->m_pkthdr.csum_data = 0xffff; 323 if_simloop(ifp, n, dst->sa_family, hlen); 324 } else 325 ifp->if_iqdrops++; 326 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 327 ETHER_ADDR_LEN) == 0) { 328 m->m_pkthdr.csum_flags |= csum_flags; 329 if (csum_flags & CSUM_DATA_VALID) 330 m->m_pkthdr.csum_data = 0xffff; 331 if_simloop(ifp, m, dst->sa_family, hlen); 332 return (0); /* XXX */ 333 } 334 } 335 336 #ifdef CARP 337 if (ifp->if_carp) { 338 /* 339 * Hold BGL and recheck ifp->if_carp 340 */ 341 get_mplock(); 342 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) { 343 rel_mplock(); 344 goto bad; 345 } 346 rel_mplock(); 347 } 348 #endif 349 350 351 /* Handle ng_ether(4) processing, if any */ 352 if (ng_ether_output_p != NULL) { 353 /* 354 * Hold BGL and recheck ng_ether_output_p 355 */ 356 get_mplock(); 357 if (ng_ether_output_p != NULL) { 358 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 359 rel_mplock(); 360 goto bad; 361 } 362 if (m == NULL) { 363 rel_mplock(); 364 return (0); 365 } 366 } 367 rel_mplock(); 368 } 369 370 /* Continue with link-layer output */ 371 return ether_output_frame(ifp, m); 372 373 bad: 374 m_freem(m); 375 return (error); 376 } 377 378 /* 379 * Returns the bridge interface an ifp is associated 380 * with. 381 * 382 * Only call if ifp->if_bridge != NULL. 383 */ 384 struct ifnet * 385 ether_bridge_interface(struct ifnet *ifp) 386 { 387 if (bridge_interface_p) 388 return(bridge_interface_p(ifp->if_bridge)); 389 return (ifp); 390 } 391 392 /* 393 * Ethernet link layer output routine to send a raw frame to the device. 394 * 395 * This assumes that the 14 byte Ethernet header is present and contiguous 396 * in the first mbuf. 397 */ 398 int 399 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 400 { 401 struct ip_fw *rule = NULL; 402 int error = 0; 403 struct altq_pktattr pktattr; 404 405 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 406 407 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 408 struct m_tag *mtag; 409 410 /* Extract info from dummynet tag */ 411 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 412 KKASSERT(mtag != NULL); 413 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 414 KKASSERT(rule != NULL); 415 416 m_tag_delete(m, mtag); 417 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 418 } 419 420 if (ifq_is_enabled(&ifp->if_snd)) 421 altq_etherclassify(&ifp->if_snd, m, &pktattr); 422 crit_enter(); 423 if (IPFW_LOADED && ether_ipfw != 0) { 424 struct ether_header save_eh, *eh; 425 426 eh = mtod(m, struct ether_header *); 427 save_eh = *eh; 428 m_adj(m, ETHER_HDR_LEN); 429 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 430 crit_exit(); 431 if (m != NULL) { 432 m_freem(m); 433 return ENOBUFS; /* pkt dropped */ 434 } else 435 return 0; /* consumed e.g. in a pipe */ 436 } 437 438 /* packet was ok, restore the ethernet header */ 439 ether_restore_header(&m, eh, &save_eh); 440 if (m == NULL) { 441 crit_exit(); 442 return ENOBUFS; 443 } 444 } 445 crit_exit(); 446 447 /* 448 * Queue message on interface, update output statistics if 449 * successful, and start output if interface not yet active. 450 */ 451 error = ifq_dispatch(ifp, m, &pktattr); 452 return (error); 453 } 454 455 /* 456 * ipfw processing for ethernet packets (in and out). 457 * The second parameter is NULL from ether_demux(), and ifp from 458 * ether_output_frame(). 459 */ 460 static boolean_t 461 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 462 const struct ether_header *eh) 463 { 464 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 465 struct ip_fw_args args; 466 struct m_tag *mtag; 467 struct mbuf *m; 468 int i; 469 470 if (*rule != NULL && fw_one_pass) 471 return TRUE; /* dummynet packet, already partially processed */ 472 473 /* 474 * I need some amount of data to be contiguous. 475 */ 476 i = min((*m0)->m_pkthdr.len, max_protohdr); 477 if ((*m0)->m_len < i) { 478 *m0 = m_pullup(*m0, i); 479 if (*m0 == NULL) 480 return FALSE; 481 } 482 483 /* 484 * Clean up tags 485 */ 486 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 487 m_tag_delete(*m0, mtag); 488 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 489 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 490 KKASSERT(mtag != NULL); 491 m_tag_delete(*m0, mtag); 492 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 493 } 494 495 args.m = *m0; /* the packet we are looking at */ 496 args.oif = dst; /* destination, if any */ 497 args.rule = *rule; /* matching rule to restart */ 498 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 499 i = ip_fw_chk_ptr(&args); 500 *m0 = args.m; 501 *rule = args.rule; 502 503 if (*m0 == NULL) 504 return FALSE; 505 506 switch (i) { 507 case IP_FW_PASS: 508 return TRUE; 509 510 case IP_FW_DIVERT: 511 case IP_FW_TEE: 512 case IP_FW_DENY: 513 /* 514 * XXX at some point add support for divert/forward actions. 515 * If none of the above matches, we have to drop the pkt. 516 */ 517 return FALSE; 518 519 case IP_FW_DUMMYNET: 520 /* 521 * Pass the pkt to dummynet, which consumes it. 522 */ 523 m = *m0; /* pass the original to dummynet */ 524 *m0 = NULL; /* and nothing back to the caller */ 525 526 ether_restore_header(&m, eh, &save_eh); 527 if (m == NULL) 528 return FALSE; 529 530 ip_fw_dn_io_ptr(m, args.cookie, 531 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 532 ip_dn_queue(m); 533 return FALSE; 534 535 default: 536 panic("unknown ipfw return value: %d\n", i); 537 } 538 } 539 540 static void 541 ether_input(struct ifnet *ifp, struct mbuf *m) 542 { 543 ether_input_chain(ifp, m, NULL, NULL); 544 } 545 546 /* 547 * Perform common duties while attaching to interface list 548 */ 549 void 550 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 551 { 552 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 553 serializer); 554 } 555 556 void 557 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 558 lwkt_serialize_t serializer) 559 { 560 struct sockaddr_dl *sdl; 561 562 ifp->if_type = IFT_ETHER; 563 ifp->if_addrlen = ETHER_ADDR_LEN; 564 ifp->if_hdrlen = ETHER_HDR_LEN; 565 if_attach(ifp, serializer); 566 ifp->if_mtu = ETHERMTU; 567 if (ifp->if_baudrate == 0) 568 ifp->if_baudrate = 10000000; 569 ifp->if_output = ether_output; 570 ifp->if_input = ether_input; 571 ifp->if_resolvemulti = ether_resolvemulti; 572 ifp->if_broadcastaddr = etherbroadcastaddr; 573 sdl = IF_LLSOCKADDR(ifp); 574 sdl->sdl_type = IFT_ETHER; 575 sdl->sdl_alen = ifp->if_addrlen; 576 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 577 /* 578 * XXX Keep the current drivers happy. 579 * XXX Remove once all drivers have been cleaned up 580 */ 581 if (lla != IFP2AC(ifp)->ac_enaddr) 582 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 583 bpfattach(ifp, dlt, hdrlen); 584 if (ng_ether_attach_p != NULL) 585 (*ng_ether_attach_p)(ifp); 586 587 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 588 } 589 590 /* 591 * Perform common duties while detaching an Ethernet interface 592 */ 593 void 594 ether_ifdetach(struct ifnet *ifp) 595 { 596 if_down(ifp); 597 598 if (ng_ether_detach_p != NULL) 599 (*ng_ether_detach_p)(ifp); 600 bpfdetach(ifp); 601 if_detach(ifp); 602 } 603 604 int 605 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 606 { 607 struct ifaddr *ifa = (struct ifaddr *) data; 608 struct ifreq *ifr = (struct ifreq *) data; 609 int error = 0; 610 611 #define IF_INIT(ifp) \ 612 do { \ 613 if (((ifp)->if_flags & IFF_UP) == 0) { \ 614 (ifp)->if_flags |= IFF_UP; \ 615 (ifp)->if_init((ifp)->if_softc); \ 616 } \ 617 } while (0) 618 619 ASSERT_IFNET_SERIALIZED_ALL(ifp); 620 621 switch (command) { 622 case SIOCSIFADDR: 623 switch (ifa->ifa_addr->sa_family) { 624 #ifdef INET 625 case AF_INET: 626 IF_INIT(ifp); /* before arpwhohas */ 627 arp_ifinit(ifp, ifa); 628 break; 629 #endif 630 #ifdef IPX 631 /* 632 * XXX - This code is probably wrong 633 */ 634 case AF_IPX: 635 { 636 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 637 struct arpcom *ac = IFP2AC(ifp); 638 639 if (ipx_nullhost(*ina)) 640 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 641 else 642 bcopy(ina->x_host.c_host, ac->ac_enaddr, 643 sizeof ac->ac_enaddr); 644 645 IF_INIT(ifp); /* Set new address. */ 646 break; 647 } 648 #endif 649 default: 650 IF_INIT(ifp); 651 break; 652 } 653 break; 654 655 case SIOCGIFADDR: 656 bcopy(IFP2AC(ifp)->ac_enaddr, 657 ((struct sockaddr *)ifr->ifr_data)->sa_data, 658 ETHER_ADDR_LEN); 659 break; 660 661 case SIOCSIFMTU: 662 /* 663 * Set the interface MTU. 664 */ 665 if (ifr->ifr_mtu > ETHERMTU) { 666 error = EINVAL; 667 } else { 668 ifp->if_mtu = ifr->ifr_mtu; 669 } 670 break; 671 default: 672 error = EINVAL; 673 break; 674 } 675 return (error); 676 677 #undef IF_INIT 678 } 679 680 int 681 ether_resolvemulti( 682 struct ifnet *ifp, 683 struct sockaddr **llsa, 684 struct sockaddr *sa) 685 { 686 struct sockaddr_dl *sdl; 687 struct sockaddr_in *sin; 688 #ifdef INET6 689 struct sockaddr_in6 *sin6; 690 #endif 691 u_char *e_addr; 692 693 switch(sa->sa_family) { 694 case AF_LINK: 695 /* 696 * No mapping needed. Just check that it's a valid MC address. 697 */ 698 sdl = (struct sockaddr_dl *)sa; 699 e_addr = LLADDR(sdl); 700 if ((e_addr[0] & 1) != 1) 701 return EADDRNOTAVAIL; 702 *llsa = 0; 703 return 0; 704 705 #ifdef INET 706 case AF_INET: 707 sin = (struct sockaddr_in *)sa; 708 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 709 return EADDRNOTAVAIL; 710 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 711 sdl->sdl_len = sizeof *sdl; 712 sdl->sdl_family = AF_LINK; 713 sdl->sdl_index = ifp->if_index; 714 sdl->sdl_type = IFT_ETHER; 715 sdl->sdl_alen = ETHER_ADDR_LEN; 716 e_addr = LLADDR(sdl); 717 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 718 *llsa = (struct sockaddr *)sdl; 719 return 0; 720 #endif 721 #ifdef INET6 722 case AF_INET6: 723 sin6 = (struct sockaddr_in6 *)sa; 724 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 725 /* 726 * An IP6 address of 0 means listen to all 727 * of the Ethernet multicast address used for IP6. 728 * (This is used for multicast routers.) 729 */ 730 ifp->if_flags |= IFF_ALLMULTI; 731 *llsa = 0; 732 return 0; 733 } 734 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 735 return EADDRNOTAVAIL; 736 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 737 sdl->sdl_len = sizeof *sdl; 738 sdl->sdl_family = AF_LINK; 739 sdl->sdl_index = ifp->if_index; 740 sdl->sdl_type = IFT_ETHER; 741 sdl->sdl_alen = ETHER_ADDR_LEN; 742 e_addr = LLADDR(sdl); 743 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 744 *llsa = (struct sockaddr *)sdl; 745 return 0; 746 #endif 747 748 default: 749 /* 750 * Well, the text isn't quite right, but it's the name 751 * that counts... 752 */ 753 return EAFNOSUPPORT; 754 } 755 } 756 757 #if 0 758 /* 759 * This is for reference. We have a table-driven version 760 * of the little-endian crc32 generator, which is faster 761 * than the double-loop. 762 */ 763 uint32_t 764 ether_crc32_le(const uint8_t *buf, size_t len) 765 { 766 uint32_t c, crc, carry; 767 size_t i, j; 768 769 crc = 0xffffffffU; /* initial value */ 770 771 for (i = 0; i < len; i++) { 772 c = buf[i]; 773 for (j = 0; j < 8; j++) { 774 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 775 crc >>= 1; 776 c >>= 1; 777 if (carry) 778 crc = (crc ^ ETHER_CRC_POLY_LE); 779 } 780 } 781 782 return (crc); 783 } 784 #else 785 uint32_t 786 ether_crc32_le(const uint8_t *buf, size_t len) 787 { 788 static const uint32_t crctab[] = { 789 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 790 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 791 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 792 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 793 }; 794 uint32_t crc; 795 size_t i; 796 797 crc = 0xffffffffU; /* initial value */ 798 799 for (i = 0; i < len; i++) { 800 crc ^= buf[i]; 801 crc = (crc >> 4) ^ crctab[crc & 0xf]; 802 crc = (crc >> 4) ^ crctab[crc & 0xf]; 803 } 804 805 return (crc); 806 } 807 #endif 808 809 uint32_t 810 ether_crc32_be(const uint8_t *buf, size_t len) 811 { 812 uint32_t c, crc, carry; 813 size_t i, j; 814 815 crc = 0xffffffffU; /* initial value */ 816 817 for (i = 0; i < len; i++) { 818 c = buf[i]; 819 for (j = 0; j < 8; j++) { 820 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 821 crc <<= 1; 822 c >>= 1; 823 if (carry) 824 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 825 } 826 } 827 828 return (crc); 829 } 830 831 /* 832 * find the size of ethernet header, and call classifier 833 */ 834 void 835 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 836 struct altq_pktattr *pktattr) 837 { 838 struct ether_header *eh; 839 uint16_t ether_type; 840 int hlen, af, hdrsize; 841 caddr_t hdr; 842 843 hlen = sizeof(struct ether_header); 844 eh = mtod(m, struct ether_header *); 845 846 ether_type = ntohs(eh->ether_type); 847 if (ether_type < ETHERMTU) { 848 /* ick! LLC/SNAP */ 849 struct llc *llc = (struct llc *)(eh + 1); 850 hlen += 8; 851 852 if (m->m_len < hlen || 853 llc->llc_dsap != LLC_SNAP_LSAP || 854 llc->llc_ssap != LLC_SNAP_LSAP || 855 llc->llc_control != LLC_UI) 856 goto bad; /* not snap! */ 857 858 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 859 } 860 861 if (ether_type == ETHERTYPE_IP) { 862 af = AF_INET; 863 hdrsize = 20; /* sizeof(struct ip) */ 864 #ifdef INET6 865 } else if (ether_type == ETHERTYPE_IPV6) { 866 af = AF_INET6; 867 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 868 #endif 869 } else 870 goto bad; 871 872 while (m->m_len <= hlen) { 873 hlen -= m->m_len; 874 m = m->m_next; 875 } 876 hdr = m->m_data + hlen; 877 if (m->m_len < hlen + hdrsize) { 878 /* 879 * ip header is not in a single mbuf. this should not 880 * happen in the current code. 881 * (todo: use m_pulldown in the future) 882 */ 883 goto bad; 884 } 885 m->m_data += hlen; 886 m->m_len -= hlen; 887 ifq_classify(ifq, m, af, pktattr); 888 m->m_data -= hlen; 889 m->m_len += hlen; 890 891 return; 892 893 bad: 894 pktattr->pattr_class = NULL; 895 pktattr->pattr_hdr = NULL; 896 pktattr->pattr_af = AF_UNSPEC; 897 } 898 899 static void 900 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 901 const struct ether_header *save_eh) 902 { 903 struct mbuf *m = *m0; 904 905 ether_restore_hdr++; 906 907 /* 908 * Prepend the header, optimize for the common case of 909 * eh pointing into the mbuf. 910 */ 911 if ((const void *)(eh + 1) == (void *)m->m_data) { 912 m->m_data -= ETHER_HDR_LEN; 913 m->m_len += ETHER_HDR_LEN; 914 m->m_pkthdr.len += ETHER_HDR_LEN; 915 } else { 916 ether_prepend_hdr++; 917 918 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 919 if (m != NULL) { 920 bcopy(save_eh, mtod(m, struct ether_header *), 921 ETHER_HDR_LEN); 922 } 923 } 924 *m0 = m; 925 } 926 927 static void 928 ether_input_ipifunc(void *arg) 929 { 930 struct mbuf *m, *next; 931 lwkt_port_t port = cpu_portfn(mycpu->gd_cpuid); 932 933 m = arg; 934 do { 935 next = m->m_nextpkt; 936 m->m_nextpkt = NULL; 937 lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.base.lmsg); 938 m = next; 939 } while (m != NULL); 940 } 941 942 void 943 ether_input_dispatch(struct mbuf_chain *chain) 944 { 945 #ifdef SMP 946 int i; 947 948 logether(disp_beg, NULL); 949 for (i = 0; i < ncpus; ++i) { 950 if (chain[i].mc_head != NULL) { 951 lwkt_send_ipiq(globaldata_find(i), 952 ether_input_ipifunc, chain[i].mc_head); 953 } 954 } 955 #else 956 logether(disp_beg, NULL); 957 if (chain->mc_head != NULL) 958 ether_input_ipifunc(chain->mc_head); 959 #endif 960 logether(disp_end, NULL); 961 } 962 963 void 964 ether_input_chain_init(struct mbuf_chain *chain) 965 { 966 #ifdef SMP 967 int i; 968 969 for (i = 0; i < ncpus; ++i) 970 chain[i].mc_head = chain[i].mc_tail = NULL; 971 #else 972 chain->mc_head = chain->mc_tail = NULL; 973 #endif 974 } 975 976 /* 977 * Upper layer processing for a received Ethernet packet. 978 */ 979 void 980 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 981 { 982 struct ether_header *eh; 983 int isr, discard = 0; 984 u_short ether_type; 985 struct ip_fw *rule = NULL; 986 987 M_ASSERTPKTHDR(m); 988 KASSERT(m->m_len >= ETHER_HDR_LEN, 989 ("ether header is no contiguous!\n")); 990 991 eh = mtod(m, struct ether_header *); 992 993 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 994 struct m_tag *mtag; 995 996 /* Extract info from dummynet tag */ 997 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 998 KKASSERT(mtag != NULL); 999 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 1000 KKASSERT(rule != NULL); 1001 1002 m_tag_delete(m, mtag); 1003 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 1004 1005 /* packet is passing the second time */ 1006 goto post_stats; 1007 } 1008 1009 #ifdef CARP 1010 /* 1011 * XXX: Okay, we need to call carp_forus() and - if it is for 1012 * us jump over code that does the normal check 1013 * "ac_enaddr == ether_dhost". The check sequence is a bit 1014 * different from OpenBSD, so we jump over as few code as 1015 * possible, to catch _all_ sanity checks. This needs 1016 * evaluation, to see if the carp ether_dhost values break any 1017 * of these checks! 1018 */ 1019 if (ifp->if_carp) { 1020 /* 1021 * Hold BGL and recheck ifp->if_carp 1022 */ 1023 get_mplock(); 1024 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) { 1025 rel_mplock(); 1026 goto post_stats; 1027 } 1028 rel_mplock(); 1029 } 1030 #endif 1031 1032 /* 1033 * We got a packet which was unicast to a different Ethernet 1034 * address. If the driver is working properly, then this 1035 * situation can only happen when the interface is in 1036 * promiscuous mode. We defer the packet discarding until the 1037 * vlan processing is done, so that vlan/bridge or vlan/netgraph 1038 * could work. 1039 */ 1040 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 1041 !ETHER_IS_MULTICAST(eh->ether_dhost) && 1042 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 1043 if (ether_debug & 1) { 1044 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 1045 "%02x:%02x:%02x:%02x:%02x:%02x " 1046 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 1047 eh->ether_dhost[0], 1048 eh->ether_dhost[1], 1049 eh->ether_dhost[2], 1050 eh->ether_dhost[3], 1051 eh->ether_dhost[4], 1052 eh->ether_dhost[5], 1053 eh->ether_shost[0], 1054 eh->ether_shost[1], 1055 eh->ether_shost[2], 1056 eh->ether_shost[3], 1057 eh->ether_shost[4], 1058 eh->ether_shost[5], 1059 eh->ether_type, 1060 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1061 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1062 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1063 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1064 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1065 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1066 ); 1067 } 1068 if ((ether_debug & 2) == 0) 1069 discard = 1; 1070 } 1071 1072 post_stats: 1073 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1074 struct ether_header save_eh = *eh; 1075 1076 /* XXX old crufty stuff, needs to be removed */ 1077 m_adj(m, sizeof(struct ether_header)); 1078 1079 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1080 m_freem(m); 1081 return; 1082 } 1083 1084 ether_restore_header(&m, eh, &save_eh); 1085 if (m == NULL) 1086 return; 1087 eh = mtod(m, struct ether_header *); 1088 } 1089 1090 ether_type = ntohs(eh->ether_type); 1091 KKASSERT(ether_type != ETHERTYPE_VLAN); 1092 1093 if (m->m_flags & M_VLANTAG) { 1094 void (*vlan_input_func)(struct mbuf *); 1095 1096 vlan_input_func = vlan_input_p; 1097 if (vlan_input_func != NULL) { 1098 vlan_input_func(m); 1099 } else { 1100 m->m_pkthdr.rcvif->if_noproto++; 1101 m_freem(m); 1102 } 1103 return; 1104 } 1105 1106 /* 1107 * If we have been asked to discard this packet 1108 * (e.g. not for us), drop it before entering 1109 * the upper layer. 1110 */ 1111 if (discard) { 1112 m_freem(m); 1113 return; 1114 } 1115 1116 /* 1117 * Clear protocol specific flags, 1118 * before entering the upper layer. 1119 */ 1120 m->m_flags &= ~M_ETHER_FLAGS; 1121 1122 /* Strip ethernet header. */ 1123 m_adj(m, sizeof(struct ether_header)); 1124 1125 switch (ether_type) { 1126 #ifdef INET 1127 case ETHERTYPE_IP: 1128 if ((m->m_flags & M_LENCHECKED) == 0) { 1129 if (!ip_lengthcheck(&m, 0)) 1130 return; 1131 } 1132 if (ipflow_fastforward(m)) 1133 return; 1134 isr = NETISR_IP; 1135 break; 1136 1137 case ETHERTYPE_ARP: 1138 if (ifp->if_flags & IFF_NOARP) { 1139 /* Discard packet if ARP is disabled on interface */ 1140 m_freem(m); 1141 return; 1142 } 1143 isr = NETISR_ARP; 1144 break; 1145 #endif 1146 1147 #ifdef INET6 1148 case ETHERTYPE_IPV6: 1149 isr = NETISR_IPV6; 1150 break; 1151 #endif 1152 1153 #ifdef IPX 1154 case ETHERTYPE_IPX: 1155 if (ef_inputp) { 1156 /* 1157 * Hold BGL and recheck ef_inputp 1158 */ 1159 get_mplock(); 1160 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1161 rel_mplock(); 1162 return; 1163 } 1164 rel_mplock(); 1165 } 1166 isr = NETISR_IPX; 1167 break; 1168 #endif 1169 1170 #ifdef MPLS 1171 case ETHERTYPE_MPLS: 1172 case ETHERTYPE_MPLS_MCAST: 1173 /* Should have been set by ether_input_chain(). */ 1174 KKASSERT(m->m_flags & M_MPLSLABELED); 1175 isr = NETISR_MPLS; 1176 break; 1177 #endif 1178 1179 default: 1180 /* 1181 * The accurate msgport is not determined before 1182 * we reach here, so recharacterize packet. 1183 */ 1184 m->m_flags &= ~M_HASH; 1185 #ifdef IPX 1186 if (ef_inputp) { 1187 /* 1188 * Hold BGL and recheck ef_inputp 1189 */ 1190 get_mplock(); 1191 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1192 rel_mplock(); 1193 return; 1194 } 1195 rel_mplock(); 1196 } 1197 #endif 1198 if (ng_ether_input_orphan_p != NULL) { 1199 /* 1200 * Hold BGL and recheck ng_ether_input_orphan_p 1201 */ 1202 get_mplock(); 1203 if (ng_ether_input_orphan_p != NULL) { 1204 ng_ether_input_orphan_p(ifp, m, eh); 1205 rel_mplock(); 1206 return; 1207 } 1208 rel_mplock(); 1209 } 1210 m_freem(m); 1211 return; 1212 } 1213 1214 if (m->m_flags & M_HASH) { 1215 if (&curthread->td_msgport == cpu_portfn(m->m_pkthdr.hash)) { 1216 netisr_handle(isr, m); 1217 return; 1218 } else { 1219 /* 1220 * XXX Something is wrong, 1221 * we probably should panic here! 1222 */ 1223 m->m_flags &= ~M_HASH; 1224 ether_input_wronghash++; 1225 } 1226 } 1227 #ifdef RSS_DEBUG 1228 ether_input_requeue++; 1229 #endif 1230 netisr_queue(isr, m); 1231 } 1232 1233 /* 1234 * First we perform any link layer operations, then continue to the 1235 * upper layers with ether_demux_oncpu(). 1236 */ 1237 static void 1238 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1239 { 1240 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1241 /* 1242 * Receiving interface's flags are changed, when this 1243 * packet is waiting for processing; discard it. 1244 */ 1245 m_freem(m); 1246 return; 1247 } 1248 1249 /* 1250 * Tap the packet off here for a bridge. bridge_input() 1251 * will return NULL if it has consumed the packet, otherwise 1252 * it gets processed as normal. Note that bridge_input() 1253 * will always return the original packet if we need to 1254 * process it locally. 1255 */ 1256 if (ifp->if_bridge) { 1257 KASSERT(bridge_input_p != NULL, 1258 ("%s: if_bridge not loaded!", __func__)); 1259 1260 if(m->m_flags & M_ETHER_BRIDGED) { 1261 m->m_flags &= ~M_ETHER_BRIDGED; 1262 } else { 1263 m = bridge_input_p(ifp, m); 1264 if (m == NULL) 1265 return; 1266 1267 KASSERT(ifp == m->m_pkthdr.rcvif, 1268 ("bridge_input_p changed rcvif\n")); 1269 } 1270 } 1271 1272 /* Handle ng_ether(4) processing, if any */ 1273 if (ng_ether_input_p != NULL) { 1274 /* 1275 * Hold BGL and recheck ng_ether_input_p 1276 */ 1277 get_mplock(); 1278 if (ng_ether_input_p != NULL) 1279 ng_ether_input_p(ifp, &m); 1280 rel_mplock(); 1281 1282 if (m == NULL) 1283 return; 1284 } 1285 1286 /* Continue with upper layer processing */ 1287 ether_demux_oncpu(ifp, m); 1288 } 1289 1290 /* 1291 * Perform certain functions of ether_input_chain(): 1292 * - Test IFF_UP 1293 * - Update statistics 1294 * - Run bpf(4) tap if requested 1295 * Then pass the packet to ether_input_oncpu(). 1296 * 1297 * This function should be used by pseudo interface (e.g. vlan(4)), 1298 * when it tries to claim that the packet is received by it. 1299 * 1300 * REINPUT_KEEPRCVIF 1301 * REINPUT_RUNBPF 1302 */ 1303 void 1304 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1305 { 1306 /* Discard packet if interface is not up */ 1307 if (!(ifp->if_flags & IFF_UP)) { 1308 m_freem(m); 1309 return; 1310 } 1311 1312 /* 1313 * Change receiving interface. The bridge will often pass a flag to 1314 * ask that this not be done so ARPs get applied to the correct 1315 * side. 1316 */ 1317 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1318 m->m_pkthdr.rcvif == NULL) { 1319 m->m_pkthdr.rcvif = ifp; 1320 } 1321 1322 /* Update statistics */ 1323 ifp->if_ipackets++; 1324 ifp->if_ibytes += m->m_pkthdr.len; 1325 if (m->m_flags & (M_MCAST | M_BCAST)) 1326 ifp->if_imcasts++; 1327 1328 if (reinput_flags & REINPUT_RUNBPF) 1329 BPF_MTAP(ifp, m); 1330 1331 ether_input_oncpu(ifp, m); 1332 } 1333 1334 static __inline boolean_t 1335 ether_vlancheck(struct mbuf **m0) 1336 { 1337 struct mbuf *m = *m0; 1338 struct ether_header *eh; 1339 uint16_t ether_type; 1340 1341 eh = mtod(m, struct ether_header *); 1342 ether_type = ntohs(eh->ether_type); 1343 1344 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1345 /* 1346 * Extract vlan tag if hardware does not do it for us 1347 */ 1348 vlan_ether_decap(&m); 1349 if (m == NULL) 1350 goto failed; 1351 1352 eh = mtod(m, struct ether_header *); 1353 ether_type = ntohs(eh->ether_type); 1354 } 1355 1356 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1357 /* 1358 * To prevent possible dangerous recursion, 1359 * we don't do vlan-in-vlan 1360 */ 1361 m->m_pkthdr.rcvif->if_noproto++; 1362 goto failed; 1363 } 1364 KKASSERT(ether_type != ETHERTYPE_VLAN); 1365 1366 m->m_flags |= M_ETHER_VLANCHECKED; 1367 *m0 = m; 1368 return TRUE; 1369 failed: 1370 if (m != NULL) 1371 m_freem(m); 1372 *m0 = NULL; 1373 return FALSE; 1374 } 1375 1376 static void 1377 ether_input_handler(netmsg_t nmsg) 1378 { 1379 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1380 struct ether_header *eh; 1381 struct ifnet *ifp; 1382 struct mbuf *m; 1383 1384 m = nmp->nm_packet; 1385 M_ASSERTPKTHDR(m); 1386 ifp = m->m_pkthdr.rcvif; 1387 1388 eh = mtod(m, struct ether_header *); 1389 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1390 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1391 ifp->if_addrlen) == 0) 1392 m->m_flags |= M_BCAST; 1393 else 1394 m->m_flags |= M_MCAST; 1395 ifp->if_imcasts++; 1396 } 1397 1398 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1399 if (!ether_vlancheck(&m)) { 1400 KKASSERT(m == NULL); 1401 return; 1402 } 1403 } 1404 1405 ether_input_oncpu(ifp, m); 1406 } 1407 1408 /* 1409 * Send the packet to the target msgport or queue it into 'chain'. 1410 * 1411 * At this point the packet had better be characterized (M_HASH set), 1412 * so we know which cpu to send it to. 1413 */ 1414 static void 1415 ether_dispatch(int isr, struct mbuf *m, struct mbuf_chain *chain) 1416 { 1417 struct netmsg_packet *pmsg; 1418 1419 KKASSERT(m->m_flags & M_HASH); 1420 pmsg = &m->m_hdr.mh_netmsg; 1421 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1422 0, ether_input_handler); 1423 pmsg->nm_packet = m; 1424 pmsg->base.lmsg.u.ms_result = isr; 1425 1426 if (chain != NULL) { 1427 int cpuid = m->m_pkthdr.hash; 1428 struct mbuf_chain *c; 1429 1430 c = &chain[cpuid]; 1431 if (c->mc_head == NULL) { 1432 c->mc_head = c->mc_tail = m; 1433 } else { 1434 c->mc_tail->m_nextpkt = m; 1435 c->mc_tail = m; 1436 } 1437 m->m_nextpkt = NULL; 1438 } else { 1439 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1440 } 1441 } 1442 1443 /* 1444 * Process a received Ethernet packet. 1445 * 1446 * The ethernet header is assumed to be in the mbuf so the caller 1447 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1448 * bytes in the first mbuf. 1449 * 1450 * - If 'chain' is NULL, this ether frame is sent to the target msgport 1451 * immediately. This situation happens when ether_input_chain is 1452 * accessed through ifnet.if_input. 1453 * 1454 * - If 'chain' is not NULL, this ether frame is queued to the 'chain' 1455 * bucket indexed by the target msgport's cpuid and the target msgport 1456 * is saved in mbuf's m_pkthdr.m_head. Caller of ether_input_chain 1457 * must initialize 'chain' by calling ether_input_chain_init(). 1458 * ether_input_dispatch must be called later to send ether frames 1459 * queued on 'chain' to their target msgport. 1460 */ 1461 void 1462 ether_input_chain(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1463 struct mbuf_chain *chain) 1464 { 1465 int isr; 1466 1467 M_ASSERTPKTHDR(m); 1468 1469 /* Discard packet if interface is not up */ 1470 if (!(ifp->if_flags & IFF_UP)) { 1471 m_freem(m); 1472 return; 1473 } 1474 1475 if (m->m_len < sizeof(struct ether_header)) { 1476 /* XXX error in the caller. */ 1477 m_freem(m); 1478 return; 1479 } 1480 1481 m->m_pkthdr.rcvif = ifp; 1482 1483 logether(chain_beg, ifp); 1484 1485 ETHER_BPF_MTAP(ifp, m); 1486 1487 ifp->if_ibytes += m->m_pkthdr.len; 1488 1489 if (ifp->if_flags & IFF_MONITOR) { 1490 struct ether_header *eh; 1491 1492 eh = mtod(m, struct ether_header *); 1493 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1494 ifp->if_imcasts++; 1495 1496 /* 1497 * Interface marked for monitoring; discard packet. 1498 */ 1499 m_freem(m); 1500 1501 logether(chain_end, ifp); 1502 return; 1503 } 1504 1505 /* 1506 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1507 * we can dispatch it immediately without further inspection. 1508 */ 1509 if (pi != NULL && (m->m_flags & M_HASH)) { 1510 #ifdef RSS_DEBUG 1511 ether_pktinfo_try++; 1512 #endif 1513 netisr_hashcheck(pi->pi_netisr, m, pi); 1514 if (m->m_flags & M_HASH) { 1515 ether_dispatch(pi->pi_netisr, m, chain); 1516 #ifdef RSS_DEBUG 1517 ether_pktinfo_hit++; 1518 #endif 1519 logether(chain_end, ifp); 1520 return; 1521 } 1522 } 1523 #ifdef RSS_DEBUG 1524 else if (ifp->if_capenable & IFCAP_RSS) { 1525 if (pi == NULL) 1526 ether_rss_nopi++; 1527 else 1528 ether_rss_nohash++; 1529 } 1530 #endif 1531 1532 /* 1533 * Packet hash will be recalculated by software, 1534 * so clear the M_HASH flag set by the driver; 1535 * the hash value calculated by the hardware may 1536 * not be exactly what we want. 1537 */ 1538 m->m_flags &= ~M_HASH; 1539 1540 if (!ether_vlancheck(&m)) { 1541 KKASSERT(m == NULL); 1542 logether(chain_end, ifp); 1543 return; 1544 } 1545 1546 isr = ether_characterize(&m); 1547 if (m == NULL) { 1548 logether(chain_end, ifp); 1549 return; 1550 } 1551 1552 /* 1553 * Finally dispatch it 1554 */ 1555 ether_dispatch(isr, m, chain); 1556 1557 logether(chain_end, ifp); 1558 } 1559 1560 static int 1561 ether_characterize(struct mbuf **m0) 1562 { 1563 struct mbuf *m = *m0; 1564 struct ether_header *eh; 1565 uint16_t ether_type; 1566 int isr; 1567 1568 eh = mtod(m, struct ether_header *); 1569 ether_type = ntohs(eh->ether_type); 1570 1571 /* 1572 * Map ether type to netisr id. 1573 */ 1574 switch (ether_type) { 1575 #ifdef INET 1576 case ETHERTYPE_IP: 1577 isr = NETISR_IP; 1578 break; 1579 1580 case ETHERTYPE_ARP: 1581 isr = NETISR_ARP; 1582 break; 1583 #endif 1584 1585 #ifdef INET6 1586 case ETHERTYPE_IPV6: 1587 isr = NETISR_IPV6; 1588 break; 1589 #endif 1590 1591 #ifdef IPX 1592 case ETHERTYPE_IPX: 1593 isr = NETISR_IPX; 1594 break; 1595 #endif 1596 1597 #ifdef MPLS 1598 case ETHERTYPE_MPLS: 1599 case ETHERTYPE_MPLS_MCAST: 1600 m->m_flags |= M_MPLSLABELED; 1601 isr = NETISR_MPLS; 1602 break; 1603 #endif 1604 1605 default: 1606 /* 1607 * NETISR_MAX is an invalid value; it is chosen to let 1608 * netisr_characterize() know that we have no clear 1609 * idea where this packet should go. 1610 */ 1611 isr = NETISR_MAX; 1612 break; 1613 } 1614 1615 /* 1616 * Ask the isr to characterize the packet since we couldn't. 1617 * This is an attempt to optimally get us onto the correct protocol 1618 * thread. 1619 */ 1620 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1621 1622 *m0 = m; 1623 return isr; 1624 } 1625 1626 static void 1627 ether_demux_handler(netmsg_t nmsg) 1628 { 1629 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1630 struct ifnet *ifp; 1631 struct mbuf *m; 1632 1633 m = nmp->nm_packet; 1634 M_ASSERTPKTHDR(m); 1635 ifp = m->m_pkthdr.rcvif; 1636 1637 ether_demux_oncpu(ifp, m); 1638 } 1639 1640 void 1641 ether_demux(struct mbuf *m) 1642 { 1643 struct netmsg_packet *pmsg; 1644 int isr; 1645 1646 isr = ether_characterize(&m); 1647 if (m == NULL) 1648 return; 1649 1650 KKASSERT(m->m_flags & M_HASH); 1651 pmsg = &m->m_hdr.mh_netmsg; 1652 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1653 0, ether_demux_handler); 1654 pmsg->nm_packet = m; 1655 pmsg->base.lmsg.u.ms_result = isr; 1656 1657 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1658 } 1659 1660 MODULE_VERSION(ether, 1); 1661