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