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/vlan/if_vlan_var.h> 73 #include <net/netmsg2.h> 74 75 #if defined(INET) || defined(INET6) 76 #include <netinet/in.h> 77 #include <netinet/ip_var.h> 78 #include <netinet/tcp_var.h> 79 #include <netinet/if_ether.h> 80 #include <netinet/ip_flow.h> 81 #include <net/ipfw/ip_fw.h> 82 #include <net/dummynet/ip_dummynet.h> 83 #endif 84 #ifdef INET6 85 #include <netinet6/nd6.h> 86 #endif 87 88 #ifdef CARP 89 #include <netinet/ip_carp.h> 90 #endif 91 92 #ifdef IPX 93 #include <netproto/ipx/ipx.h> 94 #include <netproto/ipx/ipx_if.h> 95 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 96 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 97 short *tp, int *hlen); 98 #endif 99 100 #ifdef MPLS 101 #include <netproto/mpls/mpls.h> 102 #endif 103 104 /* netgraph node hooks for ng_ether(4) */ 105 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 106 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 107 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 108 void (*ng_ether_attach_p)(struct ifnet *ifp); 109 void (*ng_ether_detach_p)(struct ifnet *ifp); 110 111 void (*vlan_input_p)(struct mbuf *); 112 113 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 114 struct rtentry *); 115 static void ether_restore_header(struct mbuf **, const struct ether_header *, 116 const struct ether_header *); 117 static int ether_characterize(struct mbuf **); 118 static void ether_dispatch(int, struct mbuf *); 119 120 /* 121 * if_bridge support 122 */ 123 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 124 int (*bridge_output_p)(struct ifnet *, struct mbuf *); 125 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 126 struct ifnet *(*bridge_interface_p)(void *if_bridge); 127 128 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 129 struct sockaddr *); 130 131 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 132 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 133 }; 134 135 #define gotoerr(e) do { error = (e); goto bad; } while (0) 136 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 137 138 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 139 struct ip_fw **rule, 140 const struct ether_header *eh); 141 142 static int ether_ipfw; 143 static u_long ether_restore_hdr; 144 static u_long ether_prepend_hdr; 145 static u_long ether_input_wronghash; 146 static int ether_debug; 147 148 #ifdef RSS_DEBUG 149 static u_long ether_pktinfo_try; 150 static u_long ether_pktinfo_hit; 151 static u_long ether_rss_nopi; 152 static u_long ether_rss_nohash; 153 static u_long ether_input_requeue; 154 static u_long ether_input_wronghwhash; 155 static int ether_input_ckhash; 156 #endif 157 158 SYSCTL_DECL(_net_link); 159 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 160 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 161 ðer_debug, 0, "Ether debug"); 162 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 163 ðer_ipfw, 0, "Pass ether pkts through firewall"); 164 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 165 ðer_restore_hdr, 0, "# of ether header restoration"); 166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 167 ðer_prepend_hdr, 0, 168 "# of ether header restoration which prepends mbuf"); 169 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 170 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 171 #ifdef RSS_DEBUG 172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 173 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 174 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 175 ðer_rss_nohash, 0, "# of packets do not have hash"); 176 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 177 ðer_pktinfo_try, 0, 178 "# of tries to find packets' msgport using pktinfo"); 179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 180 ðer_pktinfo_hit, 0, 181 "# of packets whose msgport are found using pktinfo"); 182 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 183 ðer_input_requeue, 0, "# of input packets gets requeued"); 184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 185 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 186 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 187 ðer_input_ckhash, 0, "always check hash"); 188 #endif 189 190 #define ETHER_KTR_STR "ifp=%p" 191 #define ETHER_KTR_ARGS struct ifnet *ifp 192 #ifndef KTR_ETHERNET 193 #define KTR_ETHERNET KTR_ALL 194 #endif 195 KTR_INFO_MASTER(ether); 196 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 197 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 198 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 199 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 200 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 201 202 /* 203 * Ethernet output routine. 204 * Encapsulate a packet of type family for the local net. 205 * Use trailer local net encapsulation if enough data in first 206 * packet leaves a multiple of 512 bytes of data in remainder. 207 * Assumes that ifp is actually pointer to arpcom structure. 208 */ 209 static int 210 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 211 struct rtentry *rt) 212 { 213 struct ether_header *eh, *deh; 214 u_char *edst; 215 int loop_copy = 0; 216 int hlen = ETHER_HDR_LEN; /* link layer header length */ 217 struct arpcom *ac = IFP2AC(ifp); 218 int error; 219 220 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 221 222 if (ifp->if_flags & IFF_MONITOR) 223 gotoerr(ENETDOWN); 224 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 225 gotoerr(ENETDOWN); 226 227 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 228 if (m == NULL) 229 return (ENOBUFS); 230 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 231 eh = mtod(m, struct ether_header *); 232 edst = eh->ether_dhost; 233 234 /* 235 * Fill in the destination ethernet address and frame type. 236 */ 237 switch (dst->sa_family) { 238 #ifdef INET 239 case AF_INET: 240 if (!arpresolve(ifp, rt, m, dst, edst)) 241 return (0); /* if not yet resolved */ 242 #ifdef MPLS 243 if (m->m_flags & M_MPLSLABELED) 244 eh->ether_type = htons(ETHERTYPE_MPLS); 245 else 246 #endif 247 eh->ether_type = htons(ETHERTYPE_IP); 248 break; 249 #endif 250 #ifdef INET6 251 case AF_INET6: 252 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 253 return (0); /* Something bad happenned. */ 254 eh->ether_type = htons(ETHERTYPE_IPV6); 255 break; 256 #endif 257 #ifdef IPX 258 case AF_IPX: 259 if (ef_outputp != NULL) { 260 /* 261 * Hold BGL and recheck ef_outputp 262 */ 263 get_mplock(); 264 if (ef_outputp != NULL) { 265 error = ef_outputp(ifp, &m, dst, 266 &eh->ether_type, &hlen); 267 rel_mplock(); 268 if (error) 269 goto bad; 270 else 271 break; 272 } 273 rel_mplock(); 274 } 275 eh->ether_type = htons(ETHERTYPE_IPX); 276 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 277 edst, ETHER_ADDR_LEN); 278 break; 279 #endif 280 case pseudo_AF_HDRCMPLT: 281 case AF_UNSPEC: 282 loop_copy = -1; /* if this is for us, don't do it */ 283 deh = (struct ether_header *)dst->sa_data; 284 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 285 eh->ether_type = deh->ether_type; 286 break; 287 288 default: 289 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 290 gotoerr(EAFNOSUPPORT); 291 } 292 293 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 294 memcpy(eh->ether_shost, 295 ((struct ether_header *)dst->sa_data)->ether_shost, 296 ETHER_ADDR_LEN); 297 else 298 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 299 300 /* 301 * Bridges require special output handling. 302 */ 303 if (ifp->if_bridge) { 304 KASSERT(bridge_output_p != NULL, 305 ("%s: if_bridge not loaded!", __func__)); 306 return bridge_output_p(ifp, m); 307 } 308 309 /* 310 * If a simplex interface, and the packet is being sent to our 311 * Ethernet address or a broadcast address, loopback a copy. 312 * XXX To make a simplex device behave exactly like a duplex 313 * device, we should copy in the case of sending to our own 314 * ethernet address (thus letting the original actually appear 315 * on the wire). However, we don't do that here for security 316 * reasons and compatibility with the original behavior. 317 */ 318 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 319 int csum_flags = 0; 320 321 if (m->m_pkthdr.csum_flags & CSUM_IP) 322 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 323 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 324 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 325 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 326 struct mbuf *n; 327 328 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 329 n->m_pkthdr.csum_flags |= csum_flags; 330 if (csum_flags & CSUM_DATA_VALID) 331 n->m_pkthdr.csum_data = 0xffff; 332 if_simloop(ifp, n, dst->sa_family, hlen); 333 } else 334 ifp->if_iqdrops++; 335 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 336 ETHER_ADDR_LEN) == 0) { 337 m->m_pkthdr.csum_flags |= csum_flags; 338 if (csum_flags & CSUM_DATA_VALID) 339 m->m_pkthdr.csum_data = 0xffff; 340 if_simloop(ifp, m, dst->sa_family, hlen); 341 return (0); /* XXX */ 342 } 343 } 344 345 #ifdef CARP 346 if (ifp->if_type == IFT_CARP) { 347 ifp = carp_parent(ifp); 348 if (ifp == NULL) 349 gotoerr(ENETUNREACH); 350 351 ac = IFP2AC(ifp); 352 353 /* 354 * Check precondition again 355 */ 356 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 357 358 if (ifp->if_flags & IFF_MONITOR) 359 gotoerr(ENETDOWN); 360 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 361 (IFF_UP | IFF_RUNNING)) 362 gotoerr(ENETDOWN); 363 } 364 #endif 365 366 /* Handle ng_ether(4) processing, if any */ 367 if (ng_ether_output_p != NULL) { 368 /* 369 * Hold BGL and recheck ng_ether_output_p 370 */ 371 get_mplock(); 372 if (ng_ether_output_p != NULL) { 373 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 374 rel_mplock(); 375 goto bad; 376 } 377 if (m == NULL) { 378 rel_mplock(); 379 return (0); 380 } 381 } 382 rel_mplock(); 383 } 384 385 /* Continue with link-layer output */ 386 return ether_output_frame(ifp, m); 387 388 bad: 389 m_freem(m); 390 return (error); 391 } 392 393 /* 394 * Returns the bridge interface an ifp is associated 395 * with. 396 * 397 * Only call if ifp->if_bridge != NULL. 398 */ 399 struct ifnet * 400 ether_bridge_interface(struct ifnet *ifp) 401 { 402 if (bridge_interface_p) 403 return(bridge_interface_p(ifp->if_bridge)); 404 return (ifp); 405 } 406 407 /* 408 * Ethernet link layer output routine to send a raw frame to the device. 409 * 410 * This assumes that the 14 byte Ethernet header is present and contiguous 411 * in the first mbuf. 412 */ 413 int 414 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 415 { 416 struct ip_fw *rule = NULL; 417 int error = 0; 418 struct altq_pktattr pktattr; 419 420 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 421 422 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 423 struct m_tag *mtag; 424 425 /* Extract info from dummynet tag */ 426 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 427 KKASSERT(mtag != NULL); 428 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 429 KKASSERT(rule != NULL); 430 431 m_tag_delete(m, mtag); 432 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 433 } 434 435 if (ifq_is_enabled(&ifp->if_snd)) 436 altq_etherclassify(&ifp->if_snd, m, &pktattr); 437 crit_enter(); 438 if (IPFW_LOADED && ether_ipfw != 0) { 439 struct ether_header save_eh, *eh; 440 441 eh = mtod(m, struct ether_header *); 442 save_eh = *eh; 443 m_adj(m, ETHER_HDR_LEN); 444 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 445 crit_exit(); 446 if (m != NULL) { 447 m_freem(m); 448 return ENOBUFS; /* pkt dropped */ 449 } else 450 return 0; /* consumed e.g. in a pipe */ 451 } 452 453 /* packet was ok, restore the ethernet header */ 454 ether_restore_header(&m, eh, &save_eh); 455 if (m == NULL) { 456 crit_exit(); 457 return ENOBUFS; 458 } 459 } 460 crit_exit(); 461 462 /* 463 * Queue message on interface, update output statistics if 464 * successful, and start output if interface not yet active. 465 */ 466 error = ifq_dispatch(ifp, m, &pktattr); 467 return (error); 468 } 469 470 /* 471 * ipfw processing for ethernet packets (in and out). 472 * The second parameter is NULL from ether_demux(), and ifp from 473 * ether_output_frame(). 474 */ 475 static boolean_t 476 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 477 const struct ether_header *eh) 478 { 479 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 480 struct ip_fw_args args; 481 struct m_tag *mtag; 482 struct mbuf *m; 483 int i; 484 485 if (*rule != NULL && fw_one_pass) 486 return TRUE; /* dummynet packet, already partially processed */ 487 488 /* 489 * I need some amount of data to be contiguous. 490 */ 491 i = min((*m0)->m_pkthdr.len, max_protohdr); 492 if ((*m0)->m_len < i) { 493 *m0 = m_pullup(*m0, i); 494 if (*m0 == NULL) 495 return FALSE; 496 } 497 498 /* 499 * Clean up tags 500 */ 501 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 502 m_tag_delete(*m0, mtag); 503 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 504 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 505 KKASSERT(mtag != NULL); 506 m_tag_delete(*m0, mtag); 507 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 508 } 509 510 args.m = *m0; /* the packet we are looking at */ 511 args.oif = dst; /* destination, if any */ 512 args.rule = *rule; /* matching rule to restart */ 513 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 514 i = ip_fw_chk_ptr(&args); 515 *m0 = args.m; 516 *rule = args.rule; 517 518 if (*m0 == NULL) 519 return FALSE; 520 521 switch (i) { 522 case IP_FW_PASS: 523 return TRUE; 524 525 case IP_FW_DIVERT: 526 case IP_FW_TEE: 527 case IP_FW_DENY: 528 /* 529 * XXX at some point add support for divert/forward actions. 530 * If none of the above matches, we have to drop the pkt. 531 */ 532 return FALSE; 533 534 case IP_FW_DUMMYNET: 535 /* 536 * Pass the pkt to dummynet, which consumes it. 537 */ 538 m = *m0; /* pass the original to dummynet */ 539 *m0 = NULL; /* and nothing back to the caller */ 540 541 ether_restore_header(&m, eh, &save_eh); 542 if (m == NULL) 543 return FALSE; 544 545 ip_fw_dn_io_ptr(m, args.cookie, 546 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 547 ip_dn_queue(m); 548 return FALSE; 549 550 default: 551 panic("unknown ipfw return value: %d", i); 552 } 553 } 554 555 static void 556 ether_input(struct ifnet *ifp, struct mbuf *m) 557 { 558 ether_input_pkt(ifp, m, NULL); 559 } 560 561 /* 562 * Perform common duties while attaching to interface list 563 */ 564 void 565 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 566 { 567 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 568 serializer); 569 } 570 571 void 572 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 573 lwkt_serialize_t serializer) 574 { 575 struct sockaddr_dl *sdl; 576 char ethstr[ETHER_ADDRSTRLEN + 1]; 577 578 ifp->if_type = IFT_ETHER; 579 ifp->if_addrlen = ETHER_ADDR_LEN; 580 ifp->if_hdrlen = ETHER_HDR_LEN; 581 if_attach(ifp, serializer); 582 ifp->if_mtu = ETHERMTU; 583 if (ifp->if_baudrate == 0) 584 ifp->if_baudrate = 10000000; 585 ifp->if_output = ether_output; 586 ifp->if_input = ether_input; 587 ifp->if_resolvemulti = ether_resolvemulti; 588 ifp->if_broadcastaddr = etherbroadcastaddr; 589 sdl = IF_LLSOCKADDR(ifp); 590 sdl->sdl_type = IFT_ETHER; 591 sdl->sdl_alen = ifp->if_addrlen; 592 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 593 /* 594 * XXX Keep the current drivers happy. 595 * XXX Remove once all drivers have been cleaned up 596 */ 597 if (lla != IFP2AC(ifp)->ac_enaddr) 598 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 599 bpfattach(ifp, dlt, hdrlen); 600 if (ng_ether_attach_p != NULL) 601 (*ng_ether_attach_p)(ifp); 602 603 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 604 } 605 606 /* 607 * Perform common duties while detaching an Ethernet interface 608 */ 609 void 610 ether_ifdetach(struct ifnet *ifp) 611 { 612 if_down(ifp); 613 614 if (ng_ether_detach_p != NULL) 615 (*ng_ether_detach_p)(ifp); 616 bpfdetach(ifp); 617 if_detach(ifp); 618 } 619 620 int 621 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 622 { 623 struct ifaddr *ifa = (struct ifaddr *) data; 624 struct ifreq *ifr = (struct ifreq *) data; 625 int error = 0; 626 627 #define IF_INIT(ifp) \ 628 do { \ 629 if (((ifp)->if_flags & IFF_UP) == 0) { \ 630 (ifp)->if_flags |= IFF_UP; \ 631 (ifp)->if_init((ifp)->if_softc); \ 632 } \ 633 } while (0) 634 635 ASSERT_IFNET_SERIALIZED_ALL(ifp); 636 637 switch (command) { 638 case SIOCSIFADDR: 639 switch (ifa->ifa_addr->sa_family) { 640 #ifdef INET 641 case AF_INET: 642 IF_INIT(ifp); /* before arpwhohas */ 643 arp_ifinit(ifp, ifa); 644 break; 645 #endif 646 #ifdef IPX 647 /* 648 * XXX - This code is probably wrong 649 */ 650 case AF_IPX: 651 { 652 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 653 struct arpcom *ac = IFP2AC(ifp); 654 655 if (ipx_nullhost(*ina)) 656 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 657 else 658 bcopy(ina->x_host.c_host, ac->ac_enaddr, 659 sizeof ac->ac_enaddr); 660 661 IF_INIT(ifp); /* Set new address. */ 662 break; 663 } 664 #endif 665 default: 666 IF_INIT(ifp); 667 break; 668 } 669 break; 670 671 case SIOCGIFADDR: 672 bcopy(IFP2AC(ifp)->ac_enaddr, 673 ((struct sockaddr *)ifr->ifr_data)->sa_data, 674 ETHER_ADDR_LEN); 675 break; 676 677 case SIOCSIFMTU: 678 /* 679 * Set the interface MTU. 680 */ 681 if (ifr->ifr_mtu > ETHERMTU) { 682 error = EINVAL; 683 } else { 684 ifp->if_mtu = ifr->ifr_mtu; 685 } 686 break; 687 default: 688 error = EINVAL; 689 break; 690 } 691 return (error); 692 693 #undef IF_INIT 694 } 695 696 int 697 ether_resolvemulti( 698 struct ifnet *ifp, 699 struct sockaddr **llsa, 700 struct sockaddr *sa) 701 { 702 struct sockaddr_dl *sdl; 703 #ifdef INET 704 struct sockaddr_in *sin; 705 #endif 706 #ifdef INET6 707 struct sockaddr_in6 *sin6; 708 #endif 709 u_char *e_addr; 710 711 switch(sa->sa_family) { 712 case AF_LINK: 713 /* 714 * No mapping needed. Just check that it's a valid MC address. 715 */ 716 sdl = (struct sockaddr_dl *)sa; 717 e_addr = LLADDR(sdl); 718 if ((e_addr[0] & 1) != 1) 719 return EADDRNOTAVAIL; 720 *llsa = NULL; 721 return 0; 722 723 #ifdef INET 724 case AF_INET: 725 sin = (struct sockaddr_in *)sa; 726 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 727 return EADDRNOTAVAIL; 728 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 729 sdl->sdl_len = sizeof *sdl; 730 sdl->sdl_family = AF_LINK; 731 sdl->sdl_index = ifp->if_index; 732 sdl->sdl_type = IFT_ETHER; 733 sdl->sdl_alen = ETHER_ADDR_LEN; 734 e_addr = LLADDR(sdl); 735 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 736 *llsa = (struct sockaddr *)sdl; 737 return 0; 738 #endif 739 #ifdef INET6 740 case AF_INET6: 741 sin6 = (struct sockaddr_in6 *)sa; 742 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 743 /* 744 * An IP6 address of 0 means listen to all 745 * of the Ethernet multicast address used for IP6. 746 * (This is used for multicast routers.) 747 */ 748 ifp->if_flags |= IFF_ALLMULTI; 749 *llsa = NULL; 750 return 0; 751 } 752 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 753 return EADDRNOTAVAIL; 754 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 755 sdl->sdl_len = sizeof *sdl; 756 sdl->sdl_family = AF_LINK; 757 sdl->sdl_index = ifp->if_index; 758 sdl->sdl_type = IFT_ETHER; 759 sdl->sdl_alen = ETHER_ADDR_LEN; 760 e_addr = LLADDR(sdl); 761 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 762 *llsa = (struct sockaddr *)sdl; 763 return 0; 764 #endif 765 766 default: 767 /* 768 * Well, the text isn't quite right, but it's the name 769 * that counts... 770 */ 771 return EAFNOSUPPORT; 772 } 773 } 774 775 #if 0 776 /* 777 * This is for reference. We have a table-driven version 778 * of the little-endian crc32 generator, which is faster 779 * than the double-loop. 780 */ 781 uint32_t 782 ether_crc32_le(const uint8_t *buf, size_t len) 783 { 784 uint32_t c, crc, carry; 785 size_t i, j; 786 787 crc = 0xffffffffU; /* initial value */ 788 789 for (i = 0; i < len; i++) { 790 c = buf[i]; 791 for (j = 0; j < 8; j++) { 792 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 793 crc >>= 1; 794 c >>= 1; 795 if (carry) 796 crc = (crc ^ ETHER_CRC_POLY_LE); 797 } 798 } 799 800 return (crc); 801 } 802 #else 803 uint32_t 804 ether_crc32_le(const uint8_t *buf, size_t len) 805 { 806 static const uint32_t crctab[] = { 807 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 808 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 809 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 810 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 811 }; 812 uint32_t crc; 813 size_t i; 814 815 crc = 0xffffffffU; /* initial value */ 816 817 for (i = 0; i < len; i++) { 818 crc ^= buf[i]; 819 crc = (crc >> 4) ^ crctab[crc & 0xf]; 820 crc = (crc >> 4) ^ crctab[crc & 0xf]; 821 } 822 823 return (crc); 824 } 825 #endif 826 827 uint32_t 828 ether_crc32_be(const uint8_t *buf, size_t len) 829 { 830 uint32_t c, crc, carry; 831 size_t i, j; 832 833 crc = 0xffffffffU; /* initial value */ 834 835 for (i = 0; i < len; i++) { 836 c = buf[i]; 837 for (j = 0; j < 8; j++) { 838 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 839 crc <<= 1; 840 c >>= 1; 841 if (carry) 842 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 843 } 844 } 845 846 return (crc); 847 } 848 849 /* 850 * find the size of ethernet header, and call classifier 851 */ 852 void 853 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 854 struct altq_pktattr *pktattr) 855 { 856 struct ether_header *eh; 857 uint16_t ether_type; 858 int hlen, af, hdrsize; 859 860 hlen = sizeof(struct ether_header); 861 eh = mtod(m, struct ether_header *); 862 863 ether_type = ntohs(eh->ether_type); 864 if (ether_type < ETHERMTU) { 865 /* ick! LLC/SNAP */ 866 struct llc *llc = (struct llc *)(eh + 1); 867 hlen += 8; 868 869 if (m->m_len < hlen || 870 llc->llc_dsap != LLC_SNAP_LSAP || 871 llc->llc_ssap != LLC_SNAP_LSAP || 872 llc->llc_control != LLC_UI) 873 goto bad; /* not snap! */ 874 875 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 876 } 877 878 if (ether_type == ETHERTYPE_IP) { 879 af = AF_INET; 880 hdrsize = 20; /* sizeof(struct ip) */ 881 #ifdef INET6 882 } else if (ether_type == ETHERTYPE_IPV6) { 883 af = AF_INET6; 884 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 885 #endif 886 } else 887 goto bad; 888 889 while (m->m_len <= hlen) { 890 hlen -= m->m_len; 891 m = m->m_next; 892 } 893 if (m->m_len < hlen + hdrsize) { 894 /* 895 * ip header is not in a single mbuf. this should not 896 * happen in the current code. 897 * (todo: use m_pulldown in the future) 898 */ 899 goto bad; 900 } 901 m->m_data += hlen; 902 m->m_len -= hlen; 903 ifq_classify(ifq, m, af, pktattr); 904 m->m_data -= hlen; 905 m->m_len += hlen; 906 907 return; 908 909 bad: 910 pktattr->pattr_class = NULL; 911 pktattr->pattr_hdr = NULL; 912 pktattr->pattr_af = AF_UNSPEC; 913 } 914 915 static void 916 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 917 const struct ether_header *save_eh) 918 { 919 struct mbuf *m = *m0; 920 921 ether_restore_hdr++; 922 923 /* 924 * Prepend the header, optimize for the common case of 925 * eh pointing into the mbuf. 926 */ 927 if ((const void *)(eh + 1) == (void *)m->m_data) { 928 m->m_data -= ETHER_HDR_LEN; 929 m->m_len += ETHER_HDR_LEN; 930 m->m_pkthdr.len += ETHER_HDR_LEN; 931 } else { 932 ether_prepend_hdr++; 933 934 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 935 if (m != NULL) { 936 bcopy(save_eh, mtod(m, struct ether_header *), 937 ETHER_HDR_LEN); 938 } 939 } 940 *m0 = m; 941 } 942 943 /* 944 * Upper layer processing for a received Ethernet packet. 945 */ 946 void 947 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 948 { 949 struct ether_header *eh; 950 int isr, discard = 0; 951 u_short ether_type; 952 struct ip_fw *rule = NULL; 953 954 M_ASSERTPKTHDR(m); 955 KASSERT(m->m_len >= ETHER_HDR_LEN, 956 ("ether header is not contiguous!")); 957 958 eh = mtod(m, struct ether_header *); 959 960 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 961 struct m_tag *mtag; 962 963 /* Extract info from dummynet tag */ 964 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 965 KKASSERT(mtag != NULL); 966 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 967 KKASSERT(rule != NULL); 968 969 m_tag_delete(m, mtag); 970 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 971 972 /* packet is passing the second time */ 973 goto post_stats; 974 } 975 976 /* 977 * We got a packet which was unicast to a different Ethernet 978 * address. If the driver is working properly, then this 979 * situation can only happen when the interface is in 980 * promiscuous mode. We defer the packet discarding until the 981 * vlan processing is done, so that vlan/bridge or vlan/netgraph 982 * could work. 983 */ 984 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 985 !ETHER_IS_MULTICAST(eh->ether_dhost) && 986 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 987 if (ether_debug & 1) { 988 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 989 "%02x:%02x:%02x:%02x:%02x:%02x " 990 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 991 eh->ether_dhost[0], 992 eh->ether_dhost[1], 993 eh->ether_dhost[2], 994 eh->ether_dhost[3], 995 eh->ether_dhost[4], 996 eh->ether_dhost[5], 997 eh->ether_shost[0], 998 eh->ether_shost[1], 999 eh->ether_shost[2], 1000 eh->ether_shost[3], 1001 eh->ether_shost[4], 1002 eh->ether_shost[5], 1003 eh->ether_type, 1004 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1005 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1006 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1007 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1008 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1009 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1010 ); 1011 } 1012 if ((ether_debug & 2) == 0) 1013 discard = 1; 1014 } 1015 1016 post_stats: 1017 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1018 struct ether_header save_eh = *eh; 1019 1020 /* XXX old crufty stuff, needs to be removed */ 1021 m_adj(m, sizeof(struct ether_header)); 1022 1023 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1024 m_freem(m); 1025 return; 1026 } 1027 1028 ether_restore_header(&m, eh, &save_eh); 1029 if (m == NULL) 1030 return; 1031 eh = mtod(m, struct ether_header *); 1032 } 1033 1034 ether_type = ntohs(eh->ether_type); 1035 KKASSERT(ether_type != ETHERTYPE_VLAN); 1036 1037 if (m->m_flags & M_VLANTAG) { 1038 void (*vlan_input_func)(struct mbuf *); 1039 1040 vlan_input_func = vlan_input_p; 1041 if (vlan_input_func != NULL) { 1042 vlan_input_func(m); 1043 } else { 1044 m->m_pkthdr.rcvif->if_noproto++; 1045 m_freem(m); 1046 } 1047 return; 1048 } 1049 1050 /* 1051 * If we have been asked to discard this packet 1052 * (e.g. not for us), drop it before entering 1053 * the upper layer. 1054 */ 1055 if (discard) { 1056 m_freem(m); 1057 return; 1058 } 1059 1060 /* 1061 * Clear protocol specific flags, 1062 * before entering the upper layer. 1063 */ 1064 m->m_flags &= ~M_ETHER_FLAGS; 1065 1066 /* Strip ethernet header. */ 1067 m_adj(m, sizeof(struct ether_header)); 1068 1069 switch (ether_type) { 1070 #ifdef INET 1071 case ETHERTYPE_IP: 1072 if ((m->m_flags & M_LENCHECKED) == 0) { 1073 if (!ip_lengthcheck(&m, 0)) 1074 return; 1075 } 1076 if (ipflow_fastforward(m)) 1077 return; 1078 isr = NETISR_IP; 1079 break; 1080 1081 case ETHERTYPE_ARP: 1082 if (ifp->if_flags & IFF_NOARP) { 1083 /* Discard packet if ARP is disabled on interface */ 1084 m_freem(m); 1085 return; 1086 } 1087 isr = NETISR_ARP; 1088 break; 1089 #endif 1090 1091 #ifdef INET6 1092 case ETHERTYPE_IPV6: 1093 isr = NETISR_IPV6; 1094 break; 1095 #endif 1096 1097 #ifdef IPX 1098 case ETHERTYPE_IPX: 1099 if (ef_inputp) { 1100 /* 1101 * Hold BGL and recheck ef_inputp 1102 */ 1103 get_mplock(); 1104 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1105 rel_mplock(); 1106 return; 1107 } 1108 rel_mplock(); 1109 } 1110 isr = NETISR_IPX; 1111 break; 1112 #endif 1113 1114 #ifdef MPLS 1115 case ETHERTYPE_MPLS: 1116 case ETHERTYPE_MPLS_MCAST: 1117 /* Should have been set by ether_input_pkt(). */ 1118 KKASSERT(m->m_flags & M_MPLSLABELED); 1119 isr = NETISR_MPLS; 1120 break; 1121 #endif 1122 1123 default: 1124 /* 1125 * The accurate msgport is not determined before 1126 * we reach here, so recharacterize packet. 1127 */ 1128 m->m_flags &= ~M_HASH; 1129 #ifdef IPX 1130 if (ef_inputp) { 1131 /* 1132 * Hold BGL and recheck ef_inputp 1133 */ 1134 get_mplock(); 1135 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1136 rel_mplock(); 1137 return; 1138 } 1139 rel_mplock(); 1140 } 1141 #endif 1142 if (ng_ether_input_orphan_p != NULL) { 1143 /* 1144 * Put back the ethernet header so netgraph has a 1145 * consistent view of inbound packets. 1146 */ 1147 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1148 if (m == NULL) { 1149 /* 1150 * M_PREPEND frees the mbuf in case of failure. 1151 */ 1152 return; 1153 } 1154 /* 1155 * Hold BGL and recheck ng_ether_input_orphan_p 1156 */ 1157 get_mplock(); 1158 if (ng_ether_input_orphan_p != NULL) { 1159 ng_ether_input_orphan_p(ifp, m); 1160 rel_mplock(); 1161 return; 1162 } 1163 rel_mplock(); 1164 } 1165 m_freem(m); 1166 return; 1167 } 1168 1169 if (m->m_flags & M_HASH) { 1170 if (&curthread->td_msgport == netisr_portfn(m->m_pkthdr.hash)) { 1171 netisr_handle(isr, m); 1172 return; 1173 } else { 1174 /* 1175 * XXX Something is wrong, 1176 * we probably should panic here! 1177 */ 1178 m->m_flags &= ~M_HASH; 1179 atomic_add_long(ðer_input_wronghash, 1); 1180 } 1181 } 1182 #ifdef RSS_DEBUG 1183 atomic_add_long(ðer_input_requeue, 1); 1184 #endif 1185 netisr_queue(isr, m); 1186 } 1187 1188 /* 1189 * First we perform any link layer operations, then continue to the 1190 * upper layers with ether_demux_oncpu(). 1191 */ 1192 static void 1193 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1194 { 1195 #ifdef CARP 1196 void *carp; 1197 #endif 1198 1199 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1200 /* 1201 * Receiving interface's flags are changed, when this 1202 * packet is waiting for processing; discard it. 1203 */ 1204 m_freem(m); 1205 return; 1206 } 1207 1208 /* 1209 * Tap the packet off here for a bridge. bridge_input() 1210 * will return NULL if it has consumed the packet, otherwise 1211 * it gets processed as normal. Note that bridge_input() 1212 * will always return the original packet if we need to 1213 * process it locally. 1214 */ 1215 if (ifp->if_bridge) { 1216 KASSERT(bridge_input_p != NULL, 1217 ("%s: if_bridge not loaded!", __func__)); 1218 1219 if(m->m_flags & M_ETHER_BRIDGED) { 1220 m->m_flags &= ~M_ETHER_BRIDGED; 1221 } else { 1222 m = bridge_input_p(ifp, m); 1223 if (m == NULL) 1224 return; 1225 1226 KASSERT(ifp == m->m_pkthdr.rcvif, 1227 ("bridge_input_p changed rcvif")); 1228 } 1229 } 1230 1231 #ifdef CARP 1232 carp = ifp->if_carp; 1233 if (carp) { 1234 m = carp_input(carp, m); 1235 if (m == NULL) 1236 return; 1237 KASSERT(ifp == m->m_pkthdr.rcvif, 1238 ("carp_input changed rcvif")); 1239 } 1240 #endif 1241 1242 /* Handle ng_ether(4) processing, if any */ 1243 if (ng_ether_input_p != NULL) { 1244 /* 1245 * Hold BGL and recheck ng_ether_input_p 1246 */ 1247 get_mplock(); 1248 if (ng_ether_input_p != NULL) 1249 ng_ether_input_p(ifp, &m); 1250 rel_mplock(); 1251 1252 if (m == NULL) 1253 return; 1254 } 1255 1256 /* Continue with upper layer processing */ 1257 ether_demux_oncpu(ifp, m); 1258 } 1259 1260 /* 1261 * Perform certain functions of ether_input_pkt(): 1262 * - Test IFF_UP 1263 * - Update statistics 1264 * - Run bpf(4) tap if requested 1265 * Then pass the packet to ether_input_oncpu(). 1266 * 1267 * This function should be used by pseudo interface (e.g. vlan(4)), 1268 * when it tries to claim that the packet is received by it. 1269 * 1270 * REINPUT_KEEPRCVIF 1271 * REINPUT_RUNBPF 1272 */ 1273 void 1274 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1275 { 1276 /* Discard packet if interface is not up */ 1277 if (!(ifp->if_flags & IFF_UP)) { 1278 m_freem(m); 1279 return; 1280 } 1281 1282 /* 1283 * Change receiving interface. The bridge will often pass a flag to 1284 * ask that this not be done so ARPs get applied to the correct 1285 * side. 1286 */ 1287 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1288 m->m_pkthdr.rcvif == NULL) { 1289 m->m_pkthdr.rcvif = ifp; 1290 } 1291 1292 /* Update statistics */ 1293 ifp->if_ipackets++; 1294 ifp->if_ibytes += m->m_pkthdr.len; 1295 if (m->m_flags & (M_MCAST | M_BCAST)) 1296 ifp->if_imcasts++; 1297 1298 if (reinput_flags & REINPUT_RUNBPF) 1299 BPF_MTAP(ifp, m); 1300 1301 ether_input_oncpu(ifp, m); 1302 } 1303 1304 static __inline boolean_t 1305 ether_vlancheck(struct mbuf **m0) 1306 { 1307 struct mbuf *m = *m0; 1308 struct ether_header *eh; 1309 uint16_t ether_type; 1310 1311 eh = mtod(m, struct ether_header *); 1312 ether_type = ntohs(eh->ether_type); 1313 1314 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1315 /* 1316 * Extract vlan tag if hardware does not do it for us 1317 */ 1318 vlan_ether_decap(&m); 1319 if (m == NULL) 1320 goto failed; 1321 1322 eh = mtod(m, struct ether_header *); 1323 ether_type = ntohs(eh->ether_type); 1324 } 1325 1326 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1327 /* 1328 * To prevent possible dangerous recursion, 1329 * we don't do vlan-in-vlan 1330 */ 1331 m->m_pkthdr.rcvif->if_noproto++; 1332 goto failed; 1333 } 1334 KKASSERT(ether_type != ETHERTYPE_VLAN); 1335 1336 m->m_flags |= M_ETHER_VLANCHECKED; 1337 *m0 = m; 1338 return TRUE; 1339 failed: 1340 if (m != NULL) 1341 m_freem(m); 1342 *m0 = NULL; 1343 return FALSE; 1344 } 1345 1346 static void 1347 ether_input_handler(netmsg_t nmsg) 1348 { 1349 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1350 struct ether_header *eh; 1351 struct ifnet *ifp; 1352 struct mbuf *m; 1353 1354 m = nmp->nm_packet; 1355 M_ASSERTPKTHDR(m); 1356 1357 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1358 if (!ether_vlancheck(&m)) { 1359 KKASSERT(m == NULL); 1360 return; 1361 } 1362 } 1363 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) 1364 #ifdef RSS_DEBUG 1365 || ether_input_ckhash 1366 #endif 1367 ) { 1368 int isr; 1369 1370 /* 1371 * Need to verify the hash supplied by the hardware 1372 * which could be wrong. 1373 */ 1374 m->m_flags &= ~(M_HASH | M_CKHASH); 1375 isr = ether_characterize(&m); 1376 if (m == NULL) 1377 return; 1378 KKASSERT(m->m_flags & M_HASH); 1379 1380 if (m->m_pkthdr.hash != mycpuid) { 1381 /* 1382 * Wrong hardware supplied hash; redispatch 1383 */ 1384 ether_dispatch(isr, m); 1385 #ifdef RSS_DEBUG 1386 atomic_add_long(ðer_input_wronghwhash, 1); 1387 #endif 1388 return; 1389 } 1390 } 1391 ifp = m->m_pkthdr.rcvif; 1392 1393 eh = mtod(m, struct ether_header *); 1394 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1395 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1396 ifp->if_addrlen) == 0) 1397 m->m_flags |= M_BCAST; 1398 else 1399 m->m_flags |= M_MCAST; 1400 ifp->if_imcasts++; 1401 } 1402 1403 ether_input_oncpu(ifp, m); 1404 } 1405 1406 /* 1407 * Send the packet to the target msgport 1408 * 1409 * At this point the packet had better be characterized (M_HASH set), 1410 * so we know which cpu to send it to. 1411 */ 1412 static void 1413 ether_dispatch(int isr, struct mbuf *m) 1414 { 1415 struct netmsg_packet *pmsg; 1416 1417 KKASSERT(m->m_flags & M_HASH); 1418 pmsg = &m->m_hdr.mh_netmsg; 1419 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1420 0, ether_input_handler); 1421 pmsg->nm_packet = m; 1422 pmsg->base.lmsg.u.ms_result = isr; 1423 1424 logether(disp_beg, NULL); 1425 lwkt_sendmsg(netisr_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1426 logether(disp_end, NULL); 1427 } 1428 1429 /* 1430 * Process a received Ethernet packet. 1431 * 1432 * The ethernet header is assumed to be in the mbuf so the caller 1433 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1434 * bytes in the first mbuf. 1435 */ 1436 void 1437 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi) 1438 { 1439 int isr; 1440 1441 M_ASSERTPKTHDR(m); 1442 1443 /* Discard packet if interface is not up */ 1444 if (!(ifp->if_flags & IFF_UP)) { 1445 m_freem(m); 1446 return; 1447 } 1448 1449 if (m->m_len < sizeof(struct ether_header)) { 1450 /* XXX error in the caller. */ 1451 m_freem(m); 1452 return; 1453 } 1454 1455 m->m_pkthdr.rcvif = ifp; 1456 1457 logether(pkt_beg, ifp); 1458 1459 ETHER_BPF_MTAP(ifp, m); 1460 1461 ifp->if_ibytes += m->m_pkthdr.len; 1462 1463 if (ifp->if_flags & IFF_MONITOR) { 1464 struct ether_header *eh; 1465 1466 eh = mtod(m, struct ether_header *); 1467 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1468 ifp->if_imcasts++; 1469 1470 /* 1471 * Interface marked for monitoring; discard packet. 1472 */ 1473 m_freem(m); 1474 1475 logether(pkt_end, ifp); 1476 return; 1477 } 1478 1479 /* 1480 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1481 * we can dispatch it immediately without further inspection. 1482 */ 1483 if (pi != NULL && (m->m_flags & M_HASH)) { 1484 #ifdef RSS_DEBUG 1485 atomic_add_long(ðer_pktinfo_try, 1); 1486 #endif 1487 netisr_hashcheck(pi->pi_netisr, m, pi); 1488 if (m->m_flags & M_HASH) { 1489 ether_dispatch(pi->pi_netisr, m); 1490 #ifdef RSS_DEBUG 1491 atomic_add_long(ðer_pktinfo_hit, 1); 1492 #endif 1493 logether(pkt_end, ifp); 1494 return; 1495 } 1496 } 1497 #ifdef RSS_DEBUG 1498 else if (ifp->if_capenable & IFCAP_RSS) { 1499 if (pi == NULL) 1500 atomic_add_long(ðer_rss_nopi, 1); 1501 else 1502 atomic_add_long(ðer_rss_nohash, 1); 1503 } 1504 #endif 1505 1506 /* 1507 * Packet hash will be recalculated by software, so clear 1508 * the M_HASH and M_CKHASH flag set by the driver; the hash 1509 * value calculated by the hardware may not be exactly what 1510 * we want. 1511 */ 1512 m->m_flags &= ~(M_HASH | M_CKHASH); 1513 1514 if (!ether_vlancheck(&m)) { 1515 KKASSERT(m == NULL); 1516 logether(pkt_end, ifp); 1517 return; 1518 } 1519 1520 isr = ether_characterize(&m); 1521 if (m == NULL) { 1522 logether(pkt_end, ifp); 1523 return; 1524 } 1525 1526 /* 1527 * Finally dispatch it 1528 */ 1529 ether_dispatch(isr, m); 1530 1531 logether(pkt_end, ifp); 1532 } 1533 1534 static int 1535 ether_characterize(struct mbuf **m0) 1536 { 1537 struct mbuf *m = *m0; 1538 struct ether_header *eh; 1539 uint16_t ether_type; 1540 int isr; 1541 1542 eh = mtod(m, struct ether_header *); 1543 ether_type = ntohs(eh->ether_type); 1544 1545 /* 1546 * Map ether type to netisr id. 1547 */ 1548 switch (ether_type) { 1549 #ifdef INET 1550 case ETHERTYPE_IP: 1551 isr = NETISR_IP; 1552 break; 1553 1554 case ETHERTYPE_ARP: 1555 isr = NETISR_ARP; 1556 break; 1557 #endif 1558 1559 #ifdef INET6 1560 case ETHERTYPE_IPV6: 1561 isr = NETISR_IPV6; 1562 break; 1563 #endif 1564 1565 #ifdef IPX 1566 case ETHERTYPE_IPX: 1567 isr = NETISR_IPX; 1568 break; 1569 #endif 1570 1571 #ifdef MPLS 1572 case ETHERTYPE_MPLS: 1573 case ETHERTYPE_MPLS_MCAST: 1574 m->m_flags |= M_MPLSLABELED; 1575 isr = NETISR_MPLS; 1576 break; 1577 #endif 1578 1579 default: 1580 /* 1581 * NETISR_MAX is an invalid value; it is chosen to let 1582 * netisr_characterize() know that we have no clear 1583 * idea where this packet should go. 1584 */ 1585 isr = NETISR_MAX; 1586 break; 1587 } 1588 1589 /* 1590 * Ask the isr to characterize the packet since we couldn't. 1591 * This is an attempt to optimally get us onto the correct protocol 1592 * thread. 1593 */ 1594 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1595 1596 *m0 = m; 1597 return isr; 1598 } 1599 1600 static void 1601 ether_demux_handler(netmsg_t nmsg) 1602 { 1603 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1604 struct ifnet *ifp; 1605 struct mbuf *m; 1606 1607 m = nmp->nm_packet; 1608 M_ASSERTPKTHDR(m); 1609 ifp = m->m_pkthdr.rcvif; 1610 1611 ether_demux_oncpu(ifp, m); 1612 } 1613 1614 void 1615 ether_demux(struct mbuf *m) 1616 { 1617 struct netmsg_packet *pmsg; 1618 int isr; 1619 1620 isr = ether_characterize(&m); 1621 if (m == NULL) 1622 return; 1623 1624 KKASSERT(m->m_flags & M_HASH); 1625 pmsg = &m->m_hdr.mh_netmsg; 1626 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1627 0, ether_demux_handler); 1628 pmsg->nm_packet = m; 1629 pmsg->base.lmsg.u.ms_result = isr; 1630 1631 lwkt_sendmsg(netisr_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1632 } 1633 1634 boolean_t 1635 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen, 1636 struct tcphdr **th, int *thoff) 1637 { 1638 struct mbuf *m = *mp; 1639 struct ether_header *eh; 1640 uint16_t type; 1641 int hoff; 1642 1643 KASSERT(M_WRITABLE(m), ("not writable")); 1644 1645 hoff = ETHER_HDR_LEN; 1646 if (m->m_len < hoff) { 1647 m = m_pullup(m, hoff); 1648 if (m == NULL) 1649 goto failed; 1650 } 1651 eh = mtod(m, struct ether_header *); 1652 type = eh->ether_type; 1653 1654 if (type == htons(ETHERTYPE_VLAN)) { 1655 struct ether_vlan_header *evh; 1656 1657 hoff += EVL_ENCAPLEN; 1658 if (m->m_len < hoff) { 1659 m = m_pullup(m, hoff); 1660 if (m == NULL) 1661 goto failed; 1662 } 1663 evh = mtod(m, struct ether_vlan_header *); 1664 type = evh->evl_proto; 1665 } 1666 KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type))); 1667 1668 *mp = m; 1669 *hoff0 = hoff; 1670 return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff); 1671 1672 failed: 1673 if (m != NULL) 1674 m_freem(m); 1675 *mp = NULL; 1676 return FALSE; 1677 } 1678 1679 u_char * 1680 kether_aton(const char *macstr, u_char *addr) 1681 { 1682 unsigned int o0, o1, o2, o3, o4, o5; 1683 int n; 1684 1685 if (macstr == NULL || addr == NULL) 1686 return NULL; 1687 1688 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1689 &o3, &o4, &o5); 1690 if (n != 6) 1691 return NULL; 1692 1693 addr[0] = o0; 1694 addr[1] = o1; 1695 addr[2] = o2; 1696 addr[3] = o3; 1697 addr[4] = o4; 1698 addr[5] = o5; 1699 1700 return addr; 1701 } 1702 1703 char * 1704 kether_ntoa(const u_char *addr, char *buf) 1705 { 1706 int len = ETHER_ADDRSTRLEN + 1; 1707 int n; 1708 1709 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1710 addr[1], addr[2], addr[3], addr[4], addr[5]); 1711 1712 if (n < 17) 1713 return NULL; 1714 1715 return buf; 1716 } 1717 1718 MODULE_VERSION(ether, 1); 1719