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 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU) 159 160 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 161 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default); 162 163 SYSCTL_DECL(_net_link); 164 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 165 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 166 ðer_debug, 0, "Ether debug"); 167 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 168 ðer_ipfw, 0, "Pass ether pkts through firewall"); 169 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 170 ðer_restore_hdr, 0, "# of ether header restoration"); 171 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 172 ðer_prepend_hdr, 0, 173 "# of ether header restoration which prepends mbuf"); 174 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 175 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 176 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW, 177 ðer_tsolen_default, 0, "Default max TSO length"); 178 179 #ifdef RSS_DEBUG 180 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 181 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 182 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 183 ðer_rss_nohash, 0, "# of packets do not have hash"); 184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 185 ðer_pktinfo_try, 0, 186 "# of tries to find packets' msgport using pktinfo"); 187 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 188 ðer_pktinfo_hit, 0, 189 "# of packets whose msgport are found using pktinfo"); 190 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 191 ðer_input_requeue, 0, "# of input packets gets requeued"); 192 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 193 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 194 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 195 ðer_input_ckhash, 0, "always check hash"); 196 #endif 197 198 #define ETHER_KTR_STR "ifp=%p" 199 #define ETHER_KTR_ARGS struct ifnet *ifp 200 #ifndef KTR_ETHERNET 201 #define KTR_ETHERNET KTR_ALL 202 #endif 203 KTR_INFO_MASTER(ether); 204 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 205 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 206 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 207 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 208 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 209 210 /* 211 * Ethernet output routine. 212 * Encapsulate a packet of type family for the local net. 213 * Use trailer local net encapsulation if enough data in first 214 * packet leaves a multiple of 512 bytes of data in remainder. 215 * Assumes that ifp is actually pointer to arpcom structure. 216 */ 217 static int 218 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 219 struct rtentry *rt) 220 { 221 struct ether_header *eh, *deh; 222 u_char *edst; 223 int loop_copy = 0; 224 int hlen = ETHER_HDR_LEN; /* link layer header length */ 225 struct arpcom *ac = IFP2AC(ifp); 226 int error; 227 228 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 229 230 if (ifp->if_flags & IFF_MONITOR) 231 gotoerr(ENETDOWN); 232 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 233 gotoerr(ENETDOWN); 234 235 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 236 if (m == NULL) 237 return (ENOBUFS); 238 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 239 eh = mtod(m, struct ether_header *); 240 edst = eh->ether_dhost; 241 242 /* 243 * Fill in the destination ethernet address and frame type. 244 */ 245 switch (dst->sa_family) { 246 #ifdef INET 247 case AF_INET: 248 if (!arpresolve(ifp, rt, m, dst, edst)) 249 return (0); /* if not yet resolved */ 250 #ifdef MPLS 251 if (m->m_flags & M_MPLSLABELED) 252 eh->ether_type = htons(ETHERTYPE_MPLS); 253 else 254 #endif 255 eh->ether_type = htons(ETHERTYPE_IP); 256 break; 257 #endif 258 #ifdef INET6 259 case AF_INET6: 260 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 261 return (0); /* Something bad happenned. */ 262 eh->ether_type = htons(ETHERTYPE_IPV6); 263 break; 264 #endif 265 #ifdef IPX 266 case AF_IPX: 267 if (ef_outputp != NULL) { 268 /* 269 * Hold BGL and recheck ef_outputp 270 */ 271 get_mplock(); 272 if (ef_outputp != NULL) { 273 error = ef_outputp(ifp, &m, dst, 274 &eh->ether_type, &hlen); 275 rel_mplock(); 276 if (error) 277 goto bad; 278 else 279 break; 280 } 281 rel_mplock(); 282 } 283 eh->ether_type = htons(ETHERTYPE_IPX); 284 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 285 edst, ETHER_ADDR_LEN); 286 break; 287 #endif 288 case pseudo_AF_HDRCMPLT: 289 case AF_UNSPEC: 290 loop_copy = -1; /* if this is for us, don't do it */ 291 deh = (struct ether_header *)dst->sa_data; 292 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 293 eh->ether_type = deh->ether_type; 294 break; 295 296 default: 297 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 298 gotoerr(EAFNOSUPPORT); 299 } 300 301 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 302 memcpy(eh->ether_shost, 303 ((struct ether_header *)dst->sa_data)->ether_shost, 304 ETHER_ADDR_LEN); 305 else 306 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 307 308 /* 309 * Bridges require special output handling. 310 */ 311 if (ifp->if_bridge) { 312 KASSERT(bridge_output_p != NULL, 313 ("%s: if_bridge not loaded!", __func__)); 314 return bridge_output_p(ifp, m); 315 } 316 317 /* 318 * If a simplex interface, and the packet is being sent to our 319 * Ethernet address or a broadcast address, loopback a copy. 320 * XXX To make a simplex device behave exactly like a duplex 321 * device, we should copy in the case of sending to our own 322 * ethernet address (thus letting the original actually appear 323 * on the wire). However, we don't do that here for security 324 * reasons and compatibility with the original behavior. 325 */ 326 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 327 int csum_flags = 0; 328 329 if (m->m_pkthdr.csum_flags & CSUM_IP) 330 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 331 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 332 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 333 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 334 struct mbuf *n; 335 336 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 337 n->m_pkthdr.csum_flags |= csum_flags; 338 if (csum_flags & CSUM_DATA_VALID) 339 n->m_pkthdr.csum_data = 0xffff; 340 if_simloop(ifp, n, dst->sa_family, hlen); 341 } else 342 IFNET_STAT_INC(ifp, iqdrops, 1); 343 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 344 ETHER_ADDR_LEN) == 0) { 345 m->m_pkthdr.csum_flags |= csum_flags; 346 if (csum_flags & CSUM_DATA_VALID) 347 m->m_pkthdr.csum_data = 0xffff; 348 if_simloop(ifp, m, dst->sa_family, hlen); 349 return (0); /* XXX */ 350 } 351 } 352 353 #ifdef CARP 354 if (ifp->if_type == IFT_CARP) { 355 ifp = carp_parent(ifp); 356 if (ifp == NULL) 357 gotoerr(ENETUNREACH); 358 359 ac = IFP2AC(ifp); 360 361 /* 362 * Check precondition again 363 */ 364 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 365 366 if (ifp->if_flags & IFF_MONITOR) 367 gotoerr(ENETDOWN); 368 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 369 (IFF_UP | IFF_RUNNING)) 370 gotoerr(ENETDOWN); 371 } 372 #endif 373 374 /* Handle ng_ether(4) processing, if any */ 375 if (ng_ether_output_p != NULL) { 376 /* 377 * Hold BGL and recheck ng_ether_output_p 378 */ 379 get_mplock(); 380 if (ng_ether_output_p != NULL) { 381 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 382 rel_mplock(); 383 goto bad; 384 } 385 if (m == NULL) { 386 rel_mplock(); 387 return (0); 388 } 389 } 390 rel_mplock(); 391 } 392 393 /* Continue with link-layer output */ 394 return ether_output_frame(ifp, m); 395 396 bad: 397 m_freem(m); 398 return (error); 399 } 400 401 /* 402 * Returns the bridge interface an ifp is associated 403 * with. 404 * 405 * Only call if ifp->if_bridge != NULL. 406 */ 407 struct ifnet * 408 ether_bridge_interface(struct ifnet *ifp) 409 { 410 if (bridge_interface_p) 411 return(bridge_interface_p(ifp->if_bridge)); 412 return (ifp); 413 } 414 415 /* 416 * Ethernet link layer output routine to send a raw frame to the device. 417 * 418 * This assumes that the 14 byte Ethernet header is present and contiguous 419 * in the first mbuf. 420 */ 421 int 422 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 423 { 424 struct ip_fw *rule = NULL; 425 int error = 0; 426 struct altq_pktattr pktattr; 427 428 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 429 430 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 431 struct m_tag *mtag; 432 433 /* Extract info from dummynet tag */ 434 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 435 KKASSERT(mtag != NULL); 436 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 437 KKASSERT(rule != NULL); 438 439 m_tag_delete(m, mtag); 440 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 441 } 442 443 if (ifq_is_enabled(&ifp->if_snd)) 444 altq_etherclassify(&ifp->if_snd, m, &pktattr); 445 crit_enter(); 446 if (IPFW_LOADED && ether_ipfw != 0) { 447 struct ether_header save_eh, *eh; 448 449 eh = mtod(m, struct ether_header *); 450 save_eh = *eh; 451 m_adj(m, ETHER_HDR_LEN); 452 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 453 crit_exit(); 454 if (m != NULL) { 455 m_freem(m); 456 return ENOBUFS; /* pkt dropped */ 457 } else 458 return 0; /* consumed e.g. in a pipe */ 459 } 460 461 /* packet was ok, restore the ethernet header */ 462 ether_restore_header(&m, eh, &save_eh); 463 if (m == NULL) { 464 crit_exit(); 465 return ENOBUFS; 466 } 467 } 468 crit_exit(); 469 470 /* 471 * Queue message on interface, update output statistics if 472 * successful, and start output if interface not yet active. 473 */ 474 error = ifq_dispatch(ifp, m, &pktattr); 475 return (error); 476 } 477 478 /* 479 * ipfw processing for ethernet packets (in and out). 480 * The second parameter is NULL from ether_demux(), and ifp from 481 * ether_output_frame(). 482 */ 483 static boolean_t 484 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 485 const struct ether_header *eh) 486 { 487 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 488 struct ip_fw_args args; 489 struct m_tag *mtag; 490 struct mbuf *m; 491 int i; 492 493 if (*rule != NULL && fw_one_pass) 494 return TRUE; /* dummynet packet, already partially processed */ 495 496 /* 497 * I need some amount of data to be contiguous. 498 */ 499 i = min((*m0)->m_pkthdr.len, max_protohdr); 500 if ((*m0)->m_len < i) { 501 *m0 = m_pullup(*m0, i); 502 if (*m0 == NULL) 503 return FALSE; 504 } 505 506 /* 507 * Clean up tags 508 */ 509 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 510 m_tag_delete(*m0, mtag); 511 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 512 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 513 KKASSERT(mtag != NULL); 514 m_tag_delete(*m0, mtag); 515 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 516 } 517 518 args.m = *m0; /* the packet we are looking at */ 519 args.oif = dst; /* destination, if any */ 520 args.rule = *rule; /* matching rule to restart */ 521 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 522 i = ip_fw_chk_ptr(&args); 523 *m0 = args.m; 524 *rule = args.rule; 525 526 if (*m0 == NULL) 527 return FALSE; 528 529 switch (i) { 530 case IP_FW_PASS: 531 return TRUE; 532 533 case IP_FW_DIVERT: 534 case IP_FW_TEE: 535 case IP_FW_DENY: 536 /* 537 * XXX at some point add support for divert/forward actions. 538 * If none of the above matches, we have to drop the pkt. 539 */ 540 return FALSE; 541 542 case IP_FW_DUMMYNET: 543 /* 544 * Pass the pkt to dummynet, which consumes it. 545 */ 546 m = *m0; /* pass the original to dummynet */ 547 *m0 = NULL; /* and nothing back to the caller */ 548 549 ether_restore_header(&m, eh, &save_eh); 550 if (m == NULL) 551 return FALSE; 552 553 ip_fw_dn_io_ptr(m, args.cookie, 554 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 555 ip_dn_queue(m); 556 return FALSE; 557 558 default: 559 panic("unknown ipfw return value: %d", i); 560 } 561 } 562 563 static void 564 ether_input(struct ifnet *ifp, struct mbuf *m) 565 { 566 ether_input_pkt(ifp, m, NULL); 567 } 568 569 /* 570 * Perform common duties while attaching to interface list 571 */ 572 void 573 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 574 { 575 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 576 serializer); 577 } 578 579 void 580 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 581 lwkt_serialize_t serializer) 582 { 583 struct sockaddr_dl *sdl; 584 char ethstr[ETHER_ADDRSTRLEN + 1]; 585 586 ifp->if_type = IFT_ETHER; 587 ifp->if_addrlen = ETHER_ADDR_LEN; 588 ifp->if_hdrlen = ETHER_HDR_LEN; 589 if_attach(ifp, serializer); 590 ifp->if_mtu = ETHERMTU; 591 if (ifp->if_tsolen <= 0) { 592 if ((ether_tsolen_default / ETHERMTU) < 2) { 593 kprintf("ether TSO maxlen %d -> %d\n", 594 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 595 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 596 } 597 ifp->if_tsolen = ether_tsolen_default; 598 } 599 if (ifp->if_baudrate == 0) 600 ifp->if_baudrate = 10000000; 601 ifp->if_output = ether_output; 602 ifp->if_input = ether_input; 603 ifp->if_resolvemulti = ether_resolvemulti; 604 ifp->if_broadcastaddr = etherbroadcastaddr; 605 sdl = IF_LLSOCKADDR(ifp); 606 sdl->sdl_type = IFT_ETHER; 607 sdl->sdl_alen = ifp->if_addrlen; 608 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 609 /* 610 * XXX Keep the current drivers happy. 611 * XXX Remove once all drivers have been cleaned up 612 */ 613 if (lla != IFP2AC(ifp)->ac_enaddr) 614 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 615 bpfattach(ifp, dlt, hdrlen); 616 if (ng_ether_attach_p != NULL) 617 (*ng_ether_attach_p)(ifp); 618 619 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 620 } 621 622 /* 623 * Perform common duties while detaching an Ethernet interface 624 */ 625 void 626 ether_ifdetach(struct ifnet *ifp) 627 { 628 if_down(ifp); 629 630 if (ng_ether_detach_p != NULL) 631 (*ng_ether_detach_p)(ifp); 632 bpfdetach(ifp); 633 if_detach(ifp); 634 } 635 636 int 637 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 638 { 639 struct ifaddr *ifa = (struct ifaddr *) data; 640 struct ifreq *ifr = (struct ifreq *) data; 641 int error = 0; 642 643 #define IF_INIT(ifp) \ 644 do { \ 645 if (((ifp)->if_flags & IFF_UP) == 0) { \ 646 (ifp)->if_flags |= IFF_UP; \ 647 (ifp)->if_init((ifp)->if_softc); \ 648 } \ 649 } while (0) 650 651 ASSERT_IFNET_SERIALIZED_ALL(ifp); 652 653 switch (command) { 654 case SIOCSIFADDR: 655 switch (ifa->ifa_addr->sa_family) { 656 #ifdef INET 657 case AF_INET: 658 IF_INIT(ifp); /* before arpwhohas */ 659 arp_ifinit(ifp, ifa); 660 break; 661 #endif 662 #ifdef IPX 663 /* 664 * XXX - This code is probably wrong 665 */ 666 case AF_IPX: 667 { 668 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 669 struct arpcom *ac = IFP2AC(ifp); 670 671 if (ipx_nullhost(*ina)) 672 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 673 else 674 bcopy(ina->x_host.c_host, ac->ac_enaddr, 675 sizeof ac->ac_enaddr); 676 677 IF_INIT(ifp); /* Set new address. */ 678 break; 679 } 680 #endif 681 default: 682 IF_INIT(ifp); 683 break; 684 } 685 break; 686 687 case SIOCGIFADDR: 688 bcopy(IFP2AC(ifp)->ac_enaddr, 689 ((struct sockaddr *)ifr->ifr_data)->sa_data, 690 ETHER_ADDR_LEN); 691 break; 692 693 case SIOCSIFMTU: 694 /* 695 * Set the interface MTU. 696 */ 697 if (ifr->ifr_mtu > ETHERMTU) { 698 error = EINVAL; 699 } else { 700 ifp->if_mtu = ifr->ifr_mtu; 701 } 702 break; 703 default: 704 error = EINVAL; 705 break; 706 } 707 return (error); 708 709 #undef IF_INIT 710 } 711 712 int 713 ether_resolvemulti( 714 struct ifnet *ifp, 715 struct sockaddr **llsa, 716 struct sockaddr *sa) 717 { 718 struct sockaddr_dl *sdl; 719 #ifdef INET 720 struct sockaddr_in *sin; 721 #endif 722 #ifdef INET6 723 struct sockaddr_in6 *sin6; 724 #endif 725 u_char *e_addr; 726 727 switch(sa->sa_family) { 728 case AF_LINK: 729 /* 730 * No mapping needed. Just check that it's a valid MC address. 731 */ 732 sdl = (struct sockaddr_dl *)sa; 733 e_addr = LLADDR(sdl); 734 if ((e_addr[0] & 1) != 1) 735 return EADDRNOTAVAIL; 736 *llsa = NULL; 737 return 0; 738 739 #ifdef INET 740 case AF_INET: 741 sin = (struct sockaddr_in *)sa; 742 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 743 return EADDRNOTAVAIL; 744 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 745 sdl->sdl_len = sizeof *sdl; 746 sdl->sdl_family = AF_LINK; 747 sdl->sdl_index = ifp->if_index; 748 sdl->sdl_type = IFT_ETHER; 749 sdl->sdl_alen = ETHER_ADDR_LEN; 750 e_addr = LLADDR(sdl); 751 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 752 *llsa = (struct sockaddr *)sdl; 753 return 0; 754 #endif 755 #ifdef INET6 756 case AF_INET6: 757 sin6 = (struct sockaddr_in6 *)sa; 758 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 759 /* 760 * An IP6 address of 0 means listen to all 761 * of the Ethernet multicast address used for IP6. 762 * (This is used for multicast routers.) 763 */ 764 ifp->if_flags |= IFF_ALLMULTI; 765 *llsa = NULL; 766 return 0; 767 } 768 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 769 return EADDRNOTAVAIL; 770 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 771 sdl->sdl_len = sizeof *sdl; 772 sdl->sdl_family = AF_LINK; 773 sdl->sdl_index = ifp->if_index; 774 sdl->sdl_type = IFT_ETHER; 775 sdl->sdl_alen = ETHER_ADDR_LEN; 776 e_addr = LLADDR(sdl); 777 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 778 *llsa = (struct sockaddr *)sdl; 779 return 0; 780 #endif 781 782 default: 783 /* 784 * Well, the text isn't quite right, but it's the name 785 * that counts... 786 */ 787 return EAFNOSUPPORT; 788 } 789 } 790 791 #if 0 792 /* 793 * This is for reference. We have a table-driven version 794 * of the little-endian crc32 generator, which is faster 795 * than the double-loop. 796 */ 797 uint32_t 798 ether_crc32_le(const uint8_t *buf, size_t len) 799 { 800 uint32_t c, crc, carry; 801 size_t i, j; 802 803 crc = 0xffffffffU; /* initial value */ 804 805 for (i = 0; i < len; i++) { 806 c = buf[i]; 807 for (j = 0; j < 8; j++) { 808 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 809 crc >>= 1; 810 c >>= 1; 811 if (carry) 812 crc = (crc ^ ETHER_CRC_POLY_LE); 813 } 814 } 815 816 return (crc); 817 } 818 #else 819 uint32_t 820 ether_crc32_le(const uint8_t *buf, size_t len) 821 { 822 static const uint32_t crctab[] = { 823 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 824 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 825 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 826 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 827 }; 828 uint32_t crc; 829 size_t i; 830 831 crc = 0xffffffffU; /* initial value */ 832 833 for (i = 0; i < len; i++) { 834 crc ^= buf[i]; 835 crc = (crc >> 4) ^ crctab[crc & 0xf]; 836 crc = (crc >> 4) ^ crctab[crc & 0xf]; 837 } 838 839 return (crc); 840 } 841 #endif 842 843 uint32_t 844 ether_crc32_be(const uint8_t *buf, size_t len) 845 { 846 uint32_t c, crc, carry; 847 size_t i, j; 848 849 crc = 0xffffffffU; /* initial value */ 850 851 for (i = 0; i < len; i++) { 852 c = buf[i]; 853 for (j = 0; j < 8; j++) { 854 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 855 crc <<= 1; 856 c >>= 1; 857 if (carry) 858 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 859 } 860 } 861 862 return (crc); 863 } 864 865 /* 866 * find the size of ethernet header, and call classifier 867 */ 868 void 869 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 870 struct altq_pktattr *pktattr) 871 { 872 struct ether_header *eh; 873 uint16_t ether_type; 874 int hlen, af, hdrsize; 875 876 hlen = sizeof(struct ether_header); 877 eh = mtod(m, struct ether_header *); 878 879 ether_type = ntohs(eh->ether_type); 880 if (ether_type < ETHERMTU) { 881 /* ick! LLC/SNAP */ 882 struct llc *llc = (struct llc *)(eh + 1); 883 hlen += 8; 884 885 if (m->m_len < hlen || 886 llc->llc_dsap != LLC_SNAP_LSAP || 887 llc->llc_ssap != LLC_SNAP_LSAP || 888 llc->llc_control != LLC_UI) 889 goto bad; /* not snap! */ 890 891 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 892 } 893 894 if (ether_type == ETHERTYPE_IP) { 895 af = AF_INET; 896 hdrsize = 20; /* sizeof(struct ip) */ 897 #ifdef INET6 898 } else if (ether_type == ETHERTYPE_IPV6) { 899 af = AF_INET6; 900 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 901 #endif 902 } else 903 goto bad; 904 905 while (m->m_len <= hlen) { 906 hlen -= m->m_len; 907 m = m->m_next; 908 } 909 if (m->m_len < hlen + hdrsize) { 910 /* 911 * ip header is not in a single mbuf. this should not 912 * happen in the current code. 913 * (todo: use m_pulldown in the future) 914 */ 915 goto bad; 916 } 917 m->m_data += hlen; 918 m->m_len -= hlen; 919 ifq_classify(ifq, m, af, pktattr); 920 m->m_data -= hlen; 921 m->m_len += hlen; 922 923 return; 924 925 bad: 926 pktattr->pattr_class = NULL; 927 pktattr->pattr_hdr = NULL; 928 pktattr->pattr_af = AF_UNSPEC; 929 } 930 931 static void 932 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 933 const struct ether_header *save_eh) 934 { 935 struct mbuf *m = *m0; 936 937 ether_restore_hdr++; 938 939 /* 940 * Prepend the header, optimize for the common case of 941 * eh pointing into the mbuf. 942 */ 943 if ((const void *)(eh + 1) == (void *)m->m_data) { 944 m->m_data -= ETHER_HDR_LEN; 945 m->m_len += ETHER_HDR_LEN; 946 m->m_pkthdr.len += ETHER_HDR_LEN; 947 } else { 948 ether_prepend_hdr++; 949 950 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 951 if (m != NULL) { 952 bcopy(save_eh, mtod(m, struct ether_header *), 953 ETHER_HDR_LEN); 954 } 955 } 956 *m0 = m; 957 } 958 959 /* 960 * Upper layer processing for a received Ethernet packet. 961 */ 962 void 963 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 964 { 965 struct ether_header *eh; 966 int isr, discard = 0; 967 u_short ether_type; 968 struct ip_fw *rule = NULL; 969 970 M_ASSERTPKTHDR(m); 971 KASSERT(m->m_len >= ETHER_HDR_LEN, 972 ("ether header is not contiguous!")); 973 974 eh = mtod(m, struct ether_header *); 975 976 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 977 struct m_tag *mtag; 978 979 /* Extract info from dummynet tag */ 980 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 981 KKASSERT(mtag != NULL); 982 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 983 KKASSERT(rule != NULL); 984 985 m_tag_delete(m, mtag); 986 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 987 988 /* packet is passing the second time */ 989 goto post_stats; 990 } 991 992 /* 993 * We got a packet which was unicast to a different Ethernet 994 * address. If the driver is working properly, then this 995 * situation can only happen when the interface is in 996 * promiscuous mode. We defer the packet discarding until the 997 * vlan processing is done, so that vlan/bridge or vlan/netgraph 998 * could work. 999 */ 1000 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 1001 !ETHER_IS_MULTICAST(eh->ether_dhost) && 1002 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 1003 if (ether_debug & 1) { 1004 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 1005 "%02x:%02x:%02x:%02x:%02x:%02x " 1006 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 1007 eh->ether_dhost[0], 1008 eh->ether_dhost[1], 1009 eh->ether_dhost[2], 1010 eh->ether_dhost[3], 1011 eh->ether_dhost[4], 1012 eh->ether_dhost[5], 1013 eh->ether_shost[0], 1014 eh->ether_shost[1], 1015 eh->ether_shost[2], 1016 eh->ether_shost[3], 1017 eh->ether_shost[4], 1018 eh->ether_shost[5], 1019 eh->ether_type, 1020 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1021 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1022 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1023 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1024 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1025 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1026 ); 1027 } 1028 if ((ether_debug & 2) == 0) 1029 discard = 1; 1030 } 1031 1032 post_stats: 1033 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1034 struct ether_header save_eh = *eh; 1035 1036 /* XXX old crufty stuff, needs to be removed */ 1037 m_adj(m, sizeof(struct ether_header)); 1038 1039 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1040 m_freem(m); 1041 return; 1042 } 1043 1044 ether_restore_header(&m, eh, &save_eh); 1045 if (m == NULL) 1046 return; 1047 eh = mtod(m, struct ether_header *); 1048 } 1049 1050 ether_type = ntohs(eh->ether_type); 1051 KKASSERT(ether_type != ETHERTYPE_VLAN); 1052 1053 if (m->m_flags & M_VLANTAG) { 1054 void (*vlan_input_func)(struct mbuf *); 1055 1056 vlan_input_func = vlan_input_p; 1057 if (vlan_input_func != NULL) { 1058 vlan_input_func(m); 1059 } else { 1060 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1061 m_freem(m); 1062 } 1063 return; 1064 } 1065 1066 /* 1067 * If we have been asked to discard this packet 1068 * (e.g. not for us), drop it before entering 1069 * the upper layer. 1070 */ 1071 if (discard) { 1072 m_freem(m); 1073 return; 1074 } 1075 1076 /* 1077 * Clear protocol specific flags, 1078 * before entering the upper layer. 1079 */ 1080 m->m_flags &= ~M_ETHER_FLAGS; 1081 1082 /* Strip ethernet header. */ 1083 m_adj(m, sizeof(struct ether_header)); 1084 1085 switch (ether_type) { 1086 #ifdef INET 1087 case ETHERTYPE_IP: 1088 if ((m->m_flags & M_LENCHECKED) == 0) { 1089 if (!ip_lengthcheck(&m, 0)) 1090 return; 1091 } 1092 if (ipflow_fastforward(m)) 1093 return; 1094 isr = NETISR_IP; 1095 break; 1096 1097 case ETHERTYPE_ARP: 1098 if (ifp->if_flags & IFF_NOARP) { 1099 /* Discard packet if ARP is disabled on interface */ 1100 m_freem(m); 1101 return; 1102 } 1103 isr = NETISR_ARP; 1104 break; 1105 #endif 1106 1107 #ifdef INET6 1108 case ETHERTYPE_IPV6: 1109 isr = NETISR_IPV6; 1110 break; 1111 #endif 1112 1113 #ifdef IPX 1114 case ETHERTYPE_IPX: 1115 if (ef_inputp) { 1116 /* 1117 * Hold BGL and recheck ef_inputp 1118 */ 1119 get_mplock(); 1120 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1121 rel_mplock(); 1122 return; 1123 } 1124 rel_mplock(); 1125 } 1126 isr = NETISR_IPX; 1127 break; 1128 #endif 1129 1130 #ifdef MPLS 1131 case ETHERTYPE_MPLS: 1132 case ETHERTYPE_MPLS_MCAST: 1133 /* Should have been set by ether_input_pkt(). */ 1134 KKASSERT(m->m_flags & M_MPLSLABELED); 1135 isr = NETISR_MPLS; 1136 break; 1137 #endif 1138 1139 default: 1140 /* 1141 * The accurate msgport is not determined before 1142 * we reach here, so recharacterize packet. 1143 */ 1144 m->m_flags &= ~M_HASH; 1145 #ifdef IPX 1146 if (ef_inputp) { 1147 /* 1148 * Hold BGL and recheck ef_inputp 1149 */ 1150 get_mplock(); 1151 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1152 rel_mplock(); 1153 return; 1154 } 1155 rel_mplock(); 1156 } 1157 #endif 1158 if (ng_ether_input_orphan_p != NULL) { 1159 /* 1160 * Put back the ethernet header so netgraph has a 1161 * consistent view of inbound packets. 1162 */ 1163 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1164 if (m == NULL) { 1165 /* 1166 * M_PREPEND frees the mbuf in case of failure. 1167 */ 1168 return; 1169 } 1170 /* 1171 * Hold BGL and recheck ng_ether_input_orphan_p 1172 */ 1173 get_mplock(); 1174 if (ng_ether_input_orphan_p != NULL) { 1175 ng_ether_input_orphan_p(ifp, m); 1176 rel_mplock(); 1177 return; 1178 } 1179 rel_mplock(); 1180 } 1181 m_freem(m); 1182 return; 1183 } 1184 1185 if (m->m_flags & M_HASH) { 1186 if (&curthread->td_msgport == netisr_portfn(m->m_pkthdr.hash)) { 1187 netisr_handle(isr, m); 1188 return; 1189 } else { 1190 /* 1191 * XXX Something is wrong, 1192 * we probably should panic here! 1193 */ 1194 m->m_flags &= ~M_HASH; 1195 atomic_add_long(ðer_input_wronghash, 1); 1196 } 1197 } 1198 #ifdef RSS_DEBUG 1199 atomic_add_long(ðer_input_requeue, 1); 1200 #endif 1201 netisr_queue(isr, m); 1202 } 1203 1204 /* 1205 * First we perform any link layer operations, then continue to the 1206 * upper layers with ether_demux_oncpu(). 1207 */ 1208 static void 1209 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1210 { 1211 #ifdef CARP 1212 void *carp; 1213 #endif 1214 1215 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1216 /* 1217 * Receiving interface's flags are changed, when this 1218 * packet is waiting for processing; discard it. 1219 */ 1220 m_freem(m); 1221 return; 1222 } 1223 1224 /* 1225 * Tap the packet off here for a bridge. bridge_input() 1226 * will return NULL if it has consumed the packet, otherwise 1227 * it gets processed as normal. Note that bridge_input() 1228 * will always return the original packet if we need to 1229 * process it locally. 1230 */ 1231 if (ifp->if_bridge) { 1232 KASSERT(bridge_input_p != NULL, 1233 ("%s: if_bridge not loaded!", __func__)); 1234 1235 if(m->m_flags & M_ETHER_BRIDGED) { 1236 m->m_flags &= ~M_ETHER_BRIDGED; 1237 } else { 1238 m = bridge_input_p(ifp, m); 1239 if (m == NULL) 1240 return; 1241 1242 KASSERT(ifp == m->m_pkthdr.rcvif, 1243 ("bridge_input_p changed rcvif")); 1244 } 1245 } 1246 1247 #ifdef CARP 1248 carp = ifp->if_carp; 1249 if (carp) { 1250 m = carp_input(carp, m); 1251 if (m == NULL) 1252 return; 1253 KASSERT(ifp == m->m_pkthdr.rcvif, 1254 ("carp_input changed rcvif")); 1255 } 1256 #endif 1257 1258 /* Handle ng_ether(4) processing, if any */ 1259 if (ng_ether_input_p != NULL) { 1260 /* 1261 * Hold BGL and recheck ng_ether_input_p 1262 */ 1263 get_mplock(); 1264 if (ng_ether_input_p != NULL) 1265 ng_ether_input_p(ifp, &m); 1266 rel_mplock(); 1267 1268 if (m == NULL) 1269 return; 1270 } 1271 1272 /* Continue with upper layer processing */ 1273 ether_demux_oncpu(ifp, m); 1274 } 1275 1276 /* 1277 * Perform certain functions of ether_input_pkt(): 1278 * - Test IFF_UP 1279 * - Update statistics 1280 * - Run bpf(4) tap if requested 1281 * Then pass the packet to ether_input_oncpu(). 1282 * 1283 * This function should be used by pseudo interface (e.g. vlan(4)), 1284 * when it tries to claim that the packet is received by it. 1285 * 1286 * REINPUT_KEEPRCVIF 1287 * REINPUT_RUNBPF 1288 */ 1289 void 1290 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1291 { 1292 /* Discard packet if interface is not up */ 1293 if (!(ifp->if_flags & IFF_UP)) { 1294 m_freem(m); 1295 return; 1296 } 1297 1298 /* 1299 * Change receiving interface. The bridge will often pass a flag to 1300 * ask that this not be done so ARPs get applied to the correct 1301 * side. 1302 */ 1303 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1304 m->m_pkthdr.rcvif == NULL) { 1305 m->m_pkthdr.rcvif = ifp; 1306 } 1307 1308 /* Update statistics */ 1309 IFNET_STAT_INC(ifp, ipackets, 1); 1310 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1311 if (m->m_flags & (M_MCAST | M_BCAST)) 1312 IFNET_STAT_INC(ifp, imcasts, 1); 1313 1314 if (reinput_flags & REINPUT_RUNBPF) 1315 BPF_MTAP(ifp, m); 1316 1317 ether_input_oncpu(ifp, m); 1318 } 1319 1320 static __inline boolean_t 1321 ether_vlancheck(struct mbuf **m0) 1322 { 1323 struct mbuf *m = *m0; 1324 struct ether_header *eh; 1325 uint16_t ether_type; 1326 1327 eh = mtod(m, struct ether_header *); 1328 ether_type = ntohs(eh->ether_type); 1329 1330 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1331 /* 1332 * Extract vlan tag if hardware does not do it for us 1333 */ 1334 vlan_ether_decap(&m); 1335 if (m == NULL) 1336 goto failed; 1337 1338 eh = mtod(m, struct ether_header *); 1339 ether_type = ntohs(eh->ether_type); 1340 } 1341 1342 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1343 /* 1344 * To prevent possible dangerous recursion, 1345 * we don't do vlan-in-vlan 1346 */ 1347 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1348 goto failed; 1349 } 1350 KKASSERT(ether_type != ETHERTYPE_VLAN); 1351 1352 m->m_flags |= M_ETHER_VLANCHECKED; 1353 *m0 = m; 1354 return TRUE; 1355 failed: 1356 if (m != NULL) 1357 m_freem(m); 1358 *m0 = NULL; 1359 return FALSE; 1360 } 1361 1362 static void 1363 ether_input_handler(netmsg_t nmsg) 1364 { 1365 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1366 struct ether_header *eh; 1367 struct ifnet *ifp; 1368 struct mbuf *m; 1369 1370 m = nmp->nm_packet; 1371 M_ASSERTPKTHDR(m); 1372 1373 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1374 if (!ether_vlancheck(&m)) { 1375 KKASSERT(m == NULL); 1376 return; 1377 } 1378 } 1379 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) 1380 #ifdef RSS_DEBUG 1381 || ether_input_ckhash 1382 #endif 1383 ) { 1384 int isr; 1385 1386 /* 1387 * Need to verify the hash supplied by the hardware 1388 * which could be wrong. 1389 */ 1390 m->m_flags &= ~(M_HASH | M_CKHASH); 1391 isr = ether_characterize(&m); 1392 if (m == NULL) 1393 return; 1394 KKASSERT(m->m_flags & M_HASH); 1395 1396 if (m->m_pkthdr.hash != mycpuid) { 1397 /* 1398 * Wrong hardware supplied hash; redispatch 1399 */ 1400 ether_dispatch(isr, m); 1401 #ifdef RSS_DEBUG 1402 atomic_add_long(ðer_input_wronghwhash, 1); 1403 #endif 1404 return; 1405 } 1406 } 1407 ifp = m->m_pkthdr.rcvif; 1408 1409 eh = mtod(m, struct ether_header *); 1410 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1411 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1412 ifp->if_addrlen) == 0) 1413 m->m_flags |= M_BCAST; 1414 else 1415 m->m_flags |= M_MCAST; 1416 IFNET_STAT_INC(ifp, imcasts, 1); 1417 } 1418 1419 ether_input_oncpu(ifp, m); 1420 } 1421 1422 /* 1423 * Send the packet to the target msgport 1424 * 1425 * At this point the packet had better be characterized (M_HASH set), 1426 * so we know which cpu to send it to. 1427 */ 1428 static void 1429 ether_dispatch(int isr, struct mbuf *m) 1430 { 1431 struct netmsg_packet *pmsg; 1432 1433 KKASSERT(m->m_flags & M_HASH); 1434 pmsg = &m->m_hdr.mh_netmsg; 1435 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1436 0, ether_input_handler); 1437 pmsg->nm_packet = m; 1438 pmsg->base.lmsg.u.ms_result = isr; 1439 1440 logether(disp_beg, NULL); 1441 lwkt_sendmsg(netisr_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1442 logether(disp_end, NULL); 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 void 1453 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi) 1454 { 1455 int isr; 1456 1457 M_ASSERTPKTHDR(m); 1458 1459 /* Discard packet if interface is not up */ 1460 if (!(ifp->if_flags & IFF_UP)) { 1461 m_freem(m); 1462 return; 1463 } 1464 1465 if (m->m_len < sizeof(struct ether_header)) { 1466 /* XXX error in the caller. */ 1467 m_freem(m); 1468 return; 1469 } 1470 1471 m->m_pkthdr.rcvif = ifp; 1472 1473 logether(pkt_beg, ifp); 1474 1475 ETHER_BPF_MTAP(ifp, m); 1476 1477 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1478 1479 if (ifp->if_flags & IFF_MONITOR) { 1480 struct ether_header *eh; 1481 1482 eh = mtod(m, struct ether_header *); 1483 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1484 IFNET_STAT_INC(ifp, imcasts, 1); 1485 1486 /* 1487 * Interface marked for monitoring; discard packet. 1488 */ 1489 m_freem(m); 1490 1491 logether(pkt_end, ifp); 1492 return; 1493 } 1494 1495 /* 1496 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1497 * we can dispatch it immediately without further inspection. 1498 */ 1499 if (pi != NULL && (m->m_flags & M_HASH)) { 1500 #ifdef RSS_DEBUG 1501 atomic_add_long(ðer_pktinfo_try, 1); 1502 #endif 1503 netisr_hashcheck(pi->pi_netisr, m, pi); 1504 if (m->m_flags & M_HASH) { 1505 ether_dispatch(pi->pi_netisr, m); 1506 #ifdef RSS_DEBUG 1507 atomic_add_long(ðer_pktinfo_hit, 1); 1508 #endif 1509 logether(pkt_end, ifp); 1510 return; 1511 } 1512 } 1513 #ifdef RSS_DEBUG 1514 else if (ifp->if_capenable & IFCAP_RSS) { 1515 if (pi == NULL) 1516 atomic_add_long(ðer_rss_nopi, 1); 1517 else 1518 atomic_add_long(ðer_rss_nohash, 1); 1519 } 1520 #endif 1521 1522 /* 1523 * Packet hash will be recalculated by software, so clear 1524 * the M_HASH and M_CKHASH flag set by the driver; the hash 1525 * value calculated by the hardware may not be exactly what 1526 * we want. 1527 */ 1528 m->m_flags &= ~(M_HASH | M_CKHASH); 1529 1530 if (!ether_vlancheck(&m)) { 1531 KKASSERT(m == NULL); 1532 logether(pkt_end, ifp); 1533 return; 1534 } 1535 1536 isr = ether_characterize(&m); 1537 if (m == NULL) { 1538 logether(pkt_end, ifp); 1539 return; 1540 } 1541 1542 /* 1543 * Finally dispatch it 1544 */ 1545 ether_dispatch(isr, m); 1546 1547 logether(pkt_end, ifp); 1548 } 1549 1550 static int 1551 ether_characterize(struct mbuf **m0) 1552 { 1553 struct mbuf *m = *m0; 1554 struct ether_header *eh; 1555 uint16_t ether_type; 1556 int isr; 1557 1558 eh = mtod(m, struct ether_header *); 1559 ether_type = ntohs(eh->ether_type); 1560 1561 /* 1562 * Map ether type to netisr id. 1563 */ 1564 switch (ether_type) { 1565 #ifdef INET 1566 case ETHERTYPE_IP: 1567 isr = NETISR_IP; 1568 break; 1569 1570 case ETHERTYPE_ARP: 1571 isr = NETISR_ARP; 1572 break; 1573 #endif 1574 1575 #ifdef INET6 1576 case ETHERTYPE_IPV6: 1577 isr = NETISR_IPV6; 1578 break; 1579 #endif 1580 1581 #ifdef IPX 1582 case ETHERTYPE_IPX: 1583 isr = NETISR_IPX; 1584 break; 1585 #endif 1586 1587 #ifdef MPLS 1588 case ETHERTYPE_MPLS: 1589 case ETHERTYPE_MPLS_MCAST: 1590 m->m_flags |= M_MPLSLABELED; 1591 isr = NETISR_MPLS; 1592 break; 1593 #endif 1594 1595 default: 1596 /* 1597 * NETISR_MAX is an invalid value; it is chosen to let 1598 * netisr_characterize() know that we have no clear 1599 * idea where this packet should go. 1600 */ 1601 isr = NETISR_MAX; 1602 break; 1603 } 1604 1605 /* 1606 * Ask the isr to characterize the packet since we couldn't. 1607 * This is an attempt to optimally get us onto the correct protocol 1608 * thread. 1609 */ 1610 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1611 1612 *m0 = m; 1613 return isr; 1614 } 1615 1616 static void 1617 ether_demux_handler(netmsg_t nmsg) 1618 { 1619 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1620 struct ifnet *ifp; 1621 struct mbuf *m; 1622 1623 m = nmp->nm_packet; 1624 M_ASSERTPKTHDR(m); 1625 ifp = m->m_pkthdr.rcvif; 1626 1627 ether_demux_oncpu(ifp, m); 1628 } 1629 1630 void 1631 ether_demux(struct mbuf *m) 1632 { 1633 struct netmsg_packet *pmsg; 1634 int isr; 1635 1636 isr = ether_characterize(&m); 1637 if (m == NULL) 1638 return; 1639 1640 KKASSERT(m->m_flags & M_HASH); 1641 pmsg = &m->m_hdr.mh_netmsg; 1642 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1643 0, ether_demux_handler); 1644 pmsg->nm_packet = m; 1645 pmsg->base.lmsg.u.ms_result = isr; 1646 1647 lwkt_sendmsg(netisr_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1648 } 1649 1650 boolean_t 1651 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen, 1652 struct tcphdr **th, int *thoff) 1653 { 1654 struct mbuf *m = *mp; 1655 struct ether_header *eh; 1656 uint16_t type; 1657 int hoff; 1658 1659 KASSERT(M_WRITABLE(m), ("not writable")); 1660 1661 hoff = ETHER_HDR_LEN; 1662 if (m->m_len < hoff) { 1663 m = m_pullup(m, hoff); 1664 if (m == NULL) 1665 goto failed; 1666 } 1667 eh = mtod(m, struct ether_header *); 1668 type = eh->ether_type; 1669 1670 if (type == htons(ETHERTYPE_VLAN)) { 1671 struct ether_vlan_header *evh; 1672 1673 hoff += EVL_ENCAPLEN; 1674 if (m->m_len < hoff) { 1675 m = m_pullup(m, hoff); 1676 if (m == NULL) 1677 goto failed; 1678 } 1679 evh = mtod(m, struct ether_vlan_header *); 1680 type = evh->evl_proto; 1681 } 1682 KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type))); 1683 1684 *mp = m; 1685 *hoff0 = hoff; 1686 return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff); 1687 1688 failed: 1689 if (m != NULL) 1690 m_freem(m); 1691 *mp = NULL; 1692 return FALSE; 1693 } 1694 1695 u_char * 1696 kether_aton(const char *macstr, u_char *addr) 1697 { 1698 unsigned int o0, o1, o2, o3, o4, o5; 1699 int n; 1700 1701 if (macstr == NULL || addr == NULL) 1702 return NULL; 1703 1704 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1705 &o3, &o4, &o5); 1706 if (n != 6) 1707 return NULL; 1708 1709 addr[0] = o0; 1710 addr[1] = o1; 1711 addr[2] = o2; 1712 addr[3] = o3; 1713 addr[4] = o4; 1714 addr[5] = o5; 1715 1716 return addr; 1717 } 1718 1719 char * 1720 kether_ntoa(const u_char *addr, char *buf) 1721 { 1722 int len = ETHER_ADDRSTRLEN + 1; 1723 int n; 1724 1725 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1726 addr[1], addr[2], addr[3], addr[4], addr[5]); 1727 1728 if (n < 17) 1729 return NULL; 1730 1731 return buf; 1732 } 1733 1734 MODULE_VERSION(ether, 1); 1735