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_ARGS struct ifnet *ifp 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_ARGS); 189 KTR_INFO(KTR_ETHERNET, ether, chain_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 190 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 191 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 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, u_long 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 #ifdef INET 688 struct sockaddr_in *sin; 689 #endif 690 #ifdef INET6 691 struct sockaddr_in6 *sin6; 692 #endif 693 u_char *e_addr; 694 695 switch(sa->sa_family) { 696 case AF_LINK: 697 /* 698 * No mapping needed. Just check that it's a valid MC address. 699 */ 700 sdl = (struct sockaddr_dl *)sa; 701 e_addr = LLADDR(sdl); 702 if ((e_addr[0] & 1) != 1) 703 return EADDRNOTAVAIL; 704 *llsa = NULL; 705 return 0; 706 707 #ifdef INET 708 case AF_INET: 709 sin = (struct sockaddr_in *)sa; 710 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 711 return EADDRNOTAVAIL; 712 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 713 sdl->sdl_len = sizeof *sdl; 714 sdl->sdl_family = AF_LINK; 715 sdl->sdl_index = ifp->if_index; 716 sdl->sdl_type = IFT_ETHER; 717 sdl->sdl_alen = ETHER_ADDR_LEN; 718 e_addr = LLADDR(sdl); 719 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 720 *llsa = (struct sockaddr *)sdl; 721 return 0; 722 #endif 723 #ifdef INET6 724 case AF_INET6: 725 sin6 = (struct sockaddr_in6 *)sa; 726 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 727 /* 728 * An IP6 address of 0 means listen to all 729 * of the Ethernet multicast address used for IP6. 730 * (This is used for multicast routers.) 731 */ 732 ifp->if_flags |= IFF_ALLMULTI; 733 *llsa = NULL; 734 return 0; 735 } 736 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 737 return EADDRNOTAVAIL; 738 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 739 sdl->sdl_len = sizeof *sdl; 740 sdl->sdl_family = AF_LINK; 741 sdl->sdl_index = ifp->if_index; 742 sdl->sdl_type = IFT_ETHER; 743 sdl->sdl_alen = ETHER_ADDR_LEN; 744 e_addr = LLADDR(sdl); 745 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 746 *llsa = (struct sockaddr *)sdl; 747 return 0; 748 #endif 749 750 default: 751 /* 752 * Well, the text isn't quite right, but it's the name 753 * that counts... 754 */ 755 return EAFNOSUPPORT; 756 } 757 } 758 759 #if 0 760 /* 761 * This is for reference. We have a table-driven version 762 * of the little-endian crc32 generator, which is faster 763 * than the double-loop. 764 */ 765 uint32_t 766 ether_crc32_le(const uint8_t *buf, size_t len) 767 { 768 uint32_t c, crc, carry; 769 size_t i, j; 770 771 crc = 0xffffffffU; /* initial value */ 772 773 for (i = 0; i < len; i++) { 774 c = buf[i]; 775 for (j = 0; j < 8; j++) { 776 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 777 crc >>= 1; 778 c >>= 1; 779 if (carry) 780 crc = (crc ^ ETHER_CRC_POLY_LE); 781 } 782 } 783 784 return (crc); 785 } 786 #else 787 uint32_t 788 ether_crc32_le(const uint8_t *buf, size_t len) 789 { 790 static const uint32_t crctab[] = { 791 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 792 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 793 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 794 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 795 }; 796 uint32_t crc; 797 size_t i; 798 799 crc = 0xffffffffU; /* initial value */ 800 801 for (i = 0; i < len; i++) { 802 crc ^= buf[i]; 803 crc = (crc >> 4) ^ crctab[crc & 0xf]; 804 crc = (crc >> 4) ^ crctab[crc & 0xf]; 805 } 806 807 return (crc); 808 } 809 #endif 810 811 uint32_t 812 ether_crc32_be(const uint8_t *buf, size_t len) 813 { 814 uint32_t c, crc, carry; 815 size_t i, j; 816 817 crc = 0xffffffffU; /* initial value */ 818 819 for (i = 0; i < len; i++) { 820 c = buf[i]; 821 for (j = 0; j < 8; j++) { 822 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 823 crc <<= 1; 824 c >>= 1; 825 if (carry) 826 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 827 } 828 } 829 830 return (crc); 831 } 832 833 /* 834 * find the size of ethernet header, and call classifier 835 */ 836 void 837 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 838 struct altq_pktattr *pktattr) 839 { 840 struct ether_header *eh; 841 uint16_t ether_type; 842 int hlen, af, hdrsize; 843 caddr_t hdr; 844 845 hlen = sizeof(struct ether_header); 846 eh = mtod(m, struct ether_header *); 847 848 ether_type = ntohs(eh->ether_type); 849 if (ether_type < ETHERMTU) { 850 /* ick! LLC/SNAP */ 851 struct llc *llc = (struct llc *)(eh + 1); 852 hlen += 8; 853 854 if (m->m_len < hlen || 855 llc->llc_dsap != LLC_SNAP_LSAP || 856 llc->llc_ssap != LLC_SNAP_LSAP || 857 llc->llc_control != LLC_UI) 858 goto bad; /* not snap! */ 859 860 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 861 } 862 863 if (ether_type == ETHERTYPE_IP) { 864 af = AF_INET; 865 hdrsize = 20; /* sizeof(struct ip) */ 866 #ifdef INET6 867 } else if (ether_type == ETHERTYPE_IPV6) { 868 af = AF_INET6; 869 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 870 #endif 871 } else 872 goto bad; 873 874 while (m->m_len <= hlen) { 875 hlen -= m->m_len; 876 m = m->m_next; 877 } 878 hdr = m->m_data + hlen; 879 if (m->m_len < hlen + hdrsize) { 880 /* 881 * ip header is not in a single mbuf. this should not 882 * happen in the current code. 883 * (todo: use m_pulldown in the future) 884 */ 885 goto bad; 886 } 887 m->m_data += hlen; 888 m->m_len -= hlen; 889 ifq_classify(ifq, m, af, pktattr); 890 m->m_data -= hlen; 891 m->m_len += hlen; 892 893 return; 894 895 bad: 896 pktattr->pattr_class = NULL; 897 pktattr->pattr_hdr = NULL; 898 pktattr->pattr_af = AF_UNSPEC; 899 } 900 901 static void 902 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 903 const struct ether_header *save_eh) 904 { 905 struct mbuf *m = *m0; 906 907 ether_restore_hdr++; 908 909 /* 910 * Prepend the header, optimize for the common case of 911 * eh pointing into the mbuf. 912 */ 913 if ((const void *)(eh + 1) == (void *)m->m_data) { 914 m->m_data -= ETHER_HDR_LEN; 915 m->m_len += ETHER_HDR_LEN; 916 m->m_pkthdr.len += ETHER_HDR_LEN; 917 } else { 918 ether_prepend_hdr++; 919 920 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 921 if (m != NULL) { 922 bcopy(save_eh, mtod(m, struct ether_header *), 923 ETHER_HDR_LEN); 924 } 925 } 926 *m0 = m; 927 } 928 929 static void 930 ether_input_ipifunc(void *arg) 931 { 932 struct mbuf *m, *next; 933 lwkt_port_t port = cpu_portfn(mycpu->gd_cpuid); 934 935 m = arg; 936 do { 937 next = m->m_nextpkt; 938 m->m_nextpkt = NULL; 939 lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.base.lmsg); 940 m = next; 941 } while (m != NULL); 942 } 943 944 void 945 ether_input_dispatch(struct mbuf_chain *chain) 946 { 947 #ifdef SMP 948 int i; 949 950 logether(disp_beg, NULL); 951 for (i = 0; i < ncpus; ++i) { 952 if (chain[i].mc_head != NULL) { 953 lwkt_send_ipiq(globaldata_find(i), 954 ether_input_ipifunc, chain[i].mc_head); 955 } 956 } 957 #else 958 logether(disp_beg, NULL); 959 if (chain->mc_head != NULL) 960 ether_input_ipifunc(chain->mc_head); 961 #endif 962 logether(disp_end, NULL); 963 } 964 965 void 966 ether_input_chain_init(struct mbuf_chain *chain) 967 { 968 #ifdef SMP 969 int i; 970 971 for (i = 0; i < ncpus; ++i) 972 chain[i].mc_head = chain[i].mc_tail = NULL; 973 #else 974 chain->mc_head = chain->mc_tail = NULL; 975 #endif 976 } 977 978 /* 979 * Upper layer processing for a received Ethernet packet. 980 */ 981 void 982 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 983 { 984 struct ether_header *eh; 985 int isr, discard = 0; 986 u_short ether_type; 987 struct ip_fw *rule = NULL; 988 989 M_ASSERTPKTHDR(m); 990 KASSERT(m->m_len >= ETHER_HDR_LEN, 991 ("ether header is no contiguous!\n")); 992 993 eh = mtod(m, struct ether_header *); 994 995 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 996 struct m_tag *mtag; 997 998 /* Extract info from dummynet tag */ 999 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 1000 KKASSERT(mtag != NULL); 1001 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 1002 KKASSERT(rule != NULL); 1003 1004 m_tag_delete(m, mtag); 1005 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 1006 1007 /* packet is passing the second time */ 1008 goto post_stats; 1009 } 1010 1011 #ifdef CARP 1012 /* 1013 * XXX: Okay, we need to call carp_forus() and - if it is for 1014 * us jump over code that does the normal check 1015 * "ac_enaddr == ether_dhost". The check sequence is a bit 1016 * different from OpenBSD, so we jump over as few code as 1017 * possible, to catch _all_ sanity checks. This needs 1018 * evaluation, to see if the carp ether_dhost values break any 1019 * of these checks! 1020 */ 1021 if (ifp->if_carp) { 1022 /* 1023 * Hold BGL and recheck ifp->if_carp 1024 */ 1025 get_mplock(); 1026 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) { 1027 rel_mplock(); 1028 goto post_stats; 1029 } 1030 rel_mplock(); 1031 } 1032 #endif 1033 1034 /* 1035 * We got a packet which was unicast to a different Ethernet 1036 * address. If the driver is working properly, then this 1037 * situation can only happen when the interface is in 1038 * promiscuous mode. We defer the packet discarding until the 1039 * vlan processing is done, so that vlan/bridge or vlan/netgraph 1040 * could work. 1041 */ 1042 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 1043 !ETHER_IS_MULTICAST(eh->ether_dhost) && 1044 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 1045 if (ether_debug & 1) { 1046 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 1047 "%02x:%02x:%02x:%02x:%02x:%02x " 1048 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 1049 eh->ether_dhost[0], 1050 eh->ether_dhost[1], 1051 eh->ether_dhost[2], 1052 eh->ether_dhost[3], 1053 eh->ether_dhost[4], 1054 eh->ether_dhost[5], 1055 eh->ether_shost[0], 1056 eh->ether_shost[1], 1057 eh->ether_shost[2], 1058 eh->ether_shost[3], 1059 eh->ether_shost[4], 1060 eh->ether_shost[5], 1061 eh->ether_type, 1062 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1063 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1064 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1065 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1066 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1067 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1068 ); 1069 } 1070 if ((ether_debug & 2) == 0) 1071 discard = 1; 1072 } 1073 1074 post_stats: 1075 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1076 struct ether_header save_eh = *eh; 1077 1078 /* XXX old crufty stuff, needs to be removed */ 1079 m_adj(m, sizeof(struct ether_header)); 1080 1081 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1082 m_freem(m); 1083 return; 1084 } 1085 1086 ether_restore_header(&m, eh, &save_eh); 1087 if (m == NULL) 1088 return; 1089 eh = mtod(m, struct ether_header *); 1090 } 1091 1092 ether_type = ntohs(eh->ether_type); 1093 KKASSERT(ether_type != ETHERTYPE_VLAN); 1094 1095 if (m->m_flags & M_VLANTAG) { 1096 void (*vlan_input_func)(struct mbuf *); 1097 1098 vlan_input_func = vlan_input_p; 1099 if (vlan_input_func != NULL) { 1100 vlan_input_func(m); 1101 } else { 1102 m->m_pkthdr.rcvif->if_noproto++; 1103 m_freem(m); 1104 } 1105 return; 1106 } 1107 1108 /* 1109 * If we have been asked to discard this packet 1110 * (e.g. not for us), drop it before entering 1111 * the upper layer. 1112 */ 1113 if (discard) { 1114 m_freem(m); 1115 return; 1116 } 1117 1118 /* 1119 * Clear protocol specific flags, 1120 * before entering the upper layer. 1121 */ 1122 m->m_flags &= ~M_ETHER_FLAGS; 1123 1124 /* Strip ethernet header. */ 1125 m_adj(m, sizeof(struct ether_header)); 1126 1127 switch (ether_type) { 1128 #ifdef INET 1129 case ETHERTYPE_IP: 1130 if ((m->m_flags & M_LENCHECKED) == 0) { 1131 if (!ip_lengthcheck(&m, 0)) 1132 return; 1133 } 1134 if (ipflow_fastforward(m)) 1135 return; 1136 isr = NETISR_IP; 1137 break; 1138 1139 case ETHERTYPE_ARP: 1140 if (ifp->if_flags & IFF_NOARP) { 1141 /* Discard packet if ARP is disabled on interface */ 1142 m_freem(m); 1143 return; 1144 } 1145 isr = NETISR_ARP; 1146 break; 1147 #endif 1148 1149 #ifdef INET6 1150 case ETHERTYPE_IPV6: 1151 isr = NETISR_IPV6; 1152 break; 1153 #endif 1154 1155 #ifdef IPX 1156 case ETHERTYPE_IPX: 1157 if (ef_inputp) { 1158 /* 1159 * Hold BGL and recheck ef_inputp 1160 */ 1161 get_mplock(); 1162 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1163 rel_mplock(); 1164 return; 1165 } 1166 rel_mplock(); 1167 } 1168 isr = NETISR_IPX; 1169 break; 1170 #endif 1171 1172 #ifdef MPLS 1173 case ETHERTYPE_MPLS: 1174 case ETHERTYPE_MPLS_MCAST: 1175 /* Should have been set by ether_input_chain(). */ 1176 KKASSERT(m->m_flags & M_MPLSLABELED); 1177 isr = NETISR_MPLS; 1178 break; 1179 #endif 1180 1181 default: 1182 /* 1183 * The accurate msgport is not determined before 1184 * we reach here, so recharacterize packet. 1185 */ 1186 m->m_flags &= ~M_HASH; 1187 #ifdef IPX 1188 if (ef_inputp) { 1189 /* 1190 * Hold BGL and recheck ef_inputp 1191 */ 1192 get_mplock(); 1193 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1194 rel_mplock(); 1195 return; 1196 } 1197 rel_mplock(); 1198 } 1199 #endif 1200 if (ng_ether_input_orphan_p != NULL) { 1201 /* 1202 * Hold BGL and recheck ng_ether_input_orphan_p 1203 */ 1204 get_mplock(); 1205 if (ng_ether_input_orphan_p != NULL) { 1206 ng_ether_input_orphan_p(ifp, m, eh); 1207 rel_mplock(); 1208 return; 1209 } 1210 rel_mplock(); 1211 } 1212 m_freem(m); 1213 return; 1214 } 1215 1216 if (m->m_flags & M_HASH) { 1217 if (&curthread->td_msgport == cpu_portfn(m->m_pkthdr.hash)) { 1218 netisr_handle(isr, m); 1219 return; 1220 } else { 1221 /* 1222 * XXX Something is wrong, 1223 * we probably should panic here! 1224 */ 1225 m->m_flags &= ~M_HASH; 1226 ether_input_wronghash++; 1227 } 1228 } 1229 #ifdef RSS_DEBUG 1230 ether_input_requeue++; 1231 #endif 1232 netisr_queue(isr, m); 1233 } 1234 1235 /* 1236 * First we perform any link layer operations, then continue to the 1237 * upper layers with ether_demux_oncpu(). 1238 */ 1239 static void 1240 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1241 { 1242 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1243 /* 1244 * Receiving interface's flags are changed, when this 1245 * packet is waiting for processing; discard it. 1246 */ 1247 m_freem(m); 1248 return; 1249 } 1250 1251 /* 1252 * Tap the packet off here for a bridge. bridge_input() 1253 * will return NULL if it has consumed the packet, otherwise 1254 * it gets processed as normal. Note that bridge_input() 1255 * will always return the original packet if we need to 1256 * process it locally. 1257 */ 1258 if (ifp->if_bridge) { 1259 KASSERT(bridge_input_p != NULL, 1260 ("%s: if_bridge not loaded!", __func__)); 1261 1262 if(m->m_flags & M_ETHER_BRIDGED) { 1263 m->m_flags &= ~M_ETHER_BRIDGED; 1264 } else { 1265 m = bridge_input_p(ifp, m); 1266 if (m == NULL) 1267 return; 1268 1269 KASSERT(ifp == m->m_pkthdr.rcvif, 1270 ("bridge_input_p changed rcvif\n")); 1271 } 1272 } 1273 1274 /* Handle ng_ether(4) processing, if any */ 1275 if (ng_ether_input_p != NULL) { 1276 /* 1277 * Hold BGL and recheck ng_ether_input_p 1278 */ 1279 get_mplock(); 1280 if (ng_ether_input_p != NULL) 1281 ng_ether_input_p(ifp, &m); 1282 rel_mplock(); 1283 1284 if (m == NULL) 1285 return; 1286 } 1287 1288 /* Continue with upper layer processing */ 1289 ether_demux_oncpu(ifp, m); 1290 } 1291 1292 /* 1293 * Perform certain functions of ether_input_chain(): 1294 * - Test IFF_UP 1295 * - Update statistics 1296 * - Run bpf(4) tap if requested 1297 * Then pass the packet to ether_input_oncpu(). 1298 * 1299 * This function should be used by pseudo interface (e.g. vlan(4)), 1300 * when it tries to claim that the packet is received by it. 1301 * 1302 * REINPUT_KEEPRCVIF 1303 * REINPUT_RUNBPF 1304 */ 1305 void 1306 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1307 { 1308 /* Discard packet if interface is not up */ 1309 if (!(ifp->if_flags & IFF_UP)) { 1310 m_freem(m); 1311 return; 1312 } 1313 1314 /* 1315 * Change receiving interface. The bridge will often pass a flag to 1316 * ask that this not be done so ARPs get applied to the correct 1317 * side. 1318 */ 1319 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1320 m->m_pkthdr.rcvif == NULL) { 1321 m->m_pkthdr.rcvif = ifp; 1322 } 1323 1324 /* Update statistics */ 1325 ifp->if_ipackets++; 1326 ifp->if_ibytes += m->m_pkthdr.len; 1327 if (m->m_flags & (M_MCAST | M_BCAST)) 1328 ifp->if_imcasts++; 1329 1330 if (reinput_flags & REINPUT_RUNBPF) 1331 BPF_MTAP(ifp, m); 1332 1333 ether_input_oncpu(ifp, m); 1334 } 1335 1336 static __inline boolean_t 1337 ether_vlancheck(struct mbuf **m0) 1338 { 1339 struct mbuf *m = *m0; 1340 struct ether_header *eh; 1341 uint16_t ether_type; 1342 1343 eh = mtod(m, struct ether_header *); 1344 ether_type = ntohs(eh->ether_type); 1345 1346 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1347 /* 1348 * Extract vlan tag if hardware does not do it for us 1349 */ 1350 vlan_ether_decap(&m); 1351 if (m == NULL) 1352 goto failed; 1353 1354 eh = mtod(m, struct ether_header *); 1355 ether_type = ntohs(eh->ether_type); 1356 } 1357 1358 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1359 /* 1360 * To prevent possible dangerous recursion, 1361 * we don't do vlan-in-vlan 1362 */ 1363 m->m_pkthdr.rcvif->if_noproto++; 1364 goto failed; 1365 } 1366 KKASSERT(ether_type != ETHERTYPE_VLAN); 1367 1368 m->m_flags |= M_ETHER_VLANCHECKED; 1369 *m0 = m; 1370 return TRUE; 1371 failed: 1372 if (m != NULL) 1373 m_freem(m); 1374 *m0 = NULL; 1375 return FALSE; 1376 } 1377 1378 static void 1379 ether_input_handler(netmsg_t nmsg) 1380 { 1381 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1382 struct ether_header *eh; 1383 struct ifnet *ifp; 1384 struct mbuf *m; 1385 1386 m = nmp->nm_packet; 1387 M_ASSERTPKTHDR(m); 1388 ifp = m->m_pkthdr.rcvif; 1389 1390 eh = mtod(m, struct ether_header *); 1391 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1392 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1393 ifp->if_addrlen) == 0) 1394 m->m_flags |= M_BCAST; 1395 else 1396 m->m_flags |= M_MCAST; 1397 ifp->if_imcasts++; 1398 } 1399 1400 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1401 if (!ether_vlancheck(&m)) { 1402 KKASSERT(m == NULL); 1403 return; 1404 } 1405 } 1406 1407 ether_input_oncpu(ifp, m); 1408 } 1409 1410 /* 1411 * Send the packet to the target msgport or queue it into 'chain'. 1412 * 1413 * At this point the packet had better be characterized (M_HASH set), 1414 * so we know which cpu to send it to. 1415 */ 1416 static void 1417 ether_dispatch(int isr, struct mbuf *m, struct mbuf_chain *chain) 1418 { 1419 struct netmsg_packet *pmsg; 1420 1421 KKASSERT(m->m_flags & M_HASH); 1422 pmsg = &m->m_hdr.mh_netmsg; 1423 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1424 0, ether_input_handler); 1425 pmsg->nm_packet = m; 1426 pmsg->base.lmsg.u.ms_result = isr; 1427 1428 if (chain != NULL) { 1429 int cpuid = m->m_pkthdr.hash; 1430 struct mbuf_chain *c; 1431 1432 c = &chain[cpuid]; 1433 if (c->mc_head == NULL) { 1434 c->mc_head = c->mc_tail = m; 1435 } else { 1436 c->mc_tail->m_nextpkt = m; 1437 c->mc_tail = m; 1438 } 1439 m->m_nextpkt = NULL; 1440 } else { 1441 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1442 } 1443 } 1444 1445 /* 1446 * Process a received Ethernet packet. 1447 * 1448 * The ethernet header is assumed to be in the mbuf so the caller 1449 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1450 * bytes in the first mbuf. 1451 * 1452 * - If 'chain' is NULL, this ether frame is sent to the target msgport 1453 * immediately. This situation happens when ether_input_chain is 1454 * accessed through ifnet.if_input. 1455 * 1456 * - If 'chain' is not NULL, this ether frame is queued to the 'chain' 1457 * bucket indexed by the target msgport's cpuid and the target msgport 1458 * is saved in mbuf's m_pkthdr.m_head. Caller of ether_input_chain 1459 * must initialize 'chain' by calling ether_input_chain_init(). 1460 * ether_input_dispatch must be called later to send ether frames 1461 * queued on 'chain' to their target msgport. 1462 */ 1463 void 1464 ether_input_chain(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1465 struct mbuf_chain *chain) 1466 { 1467 int isr; 1468 1469 M_ASSERTPKTHDR(m); 1470 1471 /* Discard packet if interface is not up */ 1472 if (!(ifp->if_flags & IFF_UP)) { 1473 m_freem(m); 1474 return; 1475 } 1476 1477 if (m->m_len < sizeof(struct ether_header)) { 1478 /* XXX error in the caller. */ 1479 m_freem(m); 1480 return; 1481 } 1482 1483 m->m_pkthdr.rcvif = ifp; 1484 1485 logether(chain_beg, ifp); 1486 1487 ETHER_BPF_MTAP(ifp, m); 1488 1489 ifp->if_ibytes += m->m_pkthdr.len; 1490 1491 if (ifp->if_flags & IFF_MONITOR) { 1492 struct ether_header *eh; 1493 1494 eh = mtod(m, struct ether_header *); 1495 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1496 ifp->if_imcasts++; 1497 1498 /* 1499 * Interface marked for monitoring; discard packet. 1500 */ 1501 m_freem(m); 1502 1503 logether(chain_end, ifp); 1504 return; 1505 } 1506 1507 /* 1508 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1509 * we can dispatch it immediately without further inspection. 1510 */ 1511 if (pi != NULL && (m->m_flags & M_HASH)) { 1512 #ifdef RSS_DEBUG 1513 ether_pktinfo_try++; 1514 #endif 1515 netisr_hashcheck(pi->pi_netisr, m, pi); 1516 if (m->m_flags & M_HASH) { 1517 ether_dispatch(pi->pi_netisr, m, chain); 1518 #ifdef RSS_DEBUG 1519 ether_pktinfo_hit++; 1520 #endif 1521 logether(chain_end, ifp); 1522 return; 1523 } 1524 } 1525 #ifdef RSS_DEBUG 1526 else if (ifp->if_capenable & IFCAP_RSS) { 1527 if (pi == NULL) 1528 ether_rss_nopi++; 1529 else 1530 ether_rss_nohash++; 1531 } 1532 #endif 1533 1534 /* 1535 * Packet hash will be recalculated by software, 1536 * so clear the M_HASH flag set by the driver; 1537 * the hash value calculated by the hardware may 1538 * not be exactly what we want. 1539 */ 1540 m->m_flags &= ~M_HASH; 1541 1542 if (!ether_vlancheck(&m)) { 1543 KKASSERT(m == NULL); 1544 logether(chain_end, ifp); 1545 return; 1546 } 1547 1548 isr = ether_characterize(&m); 1549 if (m == NULL) { 1550 logether(chain_end, ifp); 1551 return; 1552 } 1553 1554 /* 1555 * Finally dispatch it 1556 */ 1557 ether_dispatch(isr, m, chain); 1558 1559 logether(chain_end, ifp); 1560 } 1561 1562 static int 1563 ether_characterize(struct mbuf **m0) 1564 { 1565 struct mbuf *m = *m0; 1566 struct ether_header *eh; 1567 uint16_t ether_type; 1568 int isr; 1569 1570 eh = mtod(m, struct ether_header *); 1571 ether_type = ntohs(eh->ether_type); 1572 1573 /* 1574 * Map ether type to netisr id. 1575 */ 1576 switch (ether_type) { 1577 #ifdef INET 1578 case ETHERTYPE_IP: 1579 isr = NETISR_IP; 1580 break; 1581 1582 case ETHERTYPE_ARP: 1583 isr = NETISR_ARP; 1584 break; 1585 #endif 1586 1587 #ifdef INET6 1588 case ETHERTYPE_IPV6: 1589 isr = NETISR_IPV6; 1590 break; 1591 #endif 1592 1593 #ifdef IPX 1594 case ETHERTYPE_IPX: 1595 isr = NETISR_IPX; 1596 break; 1597 #endif 1598 1599 #ifdef MPLS 1600 case ETHERTYPE_MPLS: 1601 case ETHERTYPE_MPLS_MCAST: 1602 m->m_flags |= M_MPLSLABELED; 1603 isr = NETISR_MPLS; 1604 break; 1605 #endif 1606 1607 default: 1608 /* 1609 * NETISR_MAX is an invalid value; it is chosen to let 1610 * netisr_characterize() know that we have no clear 1611 * idea where this packet should go. 1612 */ 1613 isr = NETISR_MAX; 1614 break; 1615 } 1616 1617 /* 1618 * Ask the isr to characterize the packet since we couldn't. 1619 * This is an attempt to optimally get us onto the correct protocol 1620 * thread. 1621 */ 1622 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1623 1624 *m0 = m; 1625 return isr; 1626 } 1627 1628 static void 1629 ether_demux_handler(netmsg_t nmsg) 1630 { 1631 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1632 struct ifnet *ifp; 1633 struct mbuf *m; 1634 1635 m = nmp->nm_packet; 1636 M_ASSERTPKTHDR(m); 1637 ifp = m->m_pkthdr.rcvif; 1638 1639 ether_demux_oncpu(ifp, m); 1640 } 1641 1642 void 1643 ether_demux(struct mbuf *m) 1644 { 1645 struct netmsg_packet *pmsg; 1646 int isr; 1647 1648 isr = ether_characterize(&m); 1649 if (m == NULL) 1650 return; 1651 1652 KKASSERT(m->m_flags & M_HASH); 1653 pmsg = &m->m_hdr.mh_netmsg; 1654 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1655 0, ether_demux_handler); 1656 pmsg->nm_packet = m; 1657 pmsg->base.lmsg.u.ms_result = isr; 1658 1659 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1660 } 1661 1662 MODULE_VERSION(ether, 1); 1663