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. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 30 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $ 31 */ 32 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 #include "opt_mpls.h" 36 #include "opt_netgraph.h" 37 #include "opt_carp.h" 38 #include "opt_rss.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/globaldata.h> 43 #include <sys/kernel.h> 44 #include <sys/ktr.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/msgport.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/sysctl.h> 52 #include <sys/thread.h> 53 54 #include <sys/thread2.h> 55 #include <sys/mplock2.h> 56 57 #include <net/if.h> 58 #include <net/netisr.h> 59 #include <net/route.h> 60 #include <net/if_llc.h> 61 #include <net/if_dl.h> 62 #include <net/if_types.h> 63 #include <net/ifq_var.h> 64 #include <net/bpf.h> 65 #include <net/ethernet.h> 66 #include <net/vlan/if_vlan_ether.h> 67 #include <net/vlan/if_vlan_var.h> 68 #include <net/netmsg2.h> 69 #include <net/netisr2.h> 70 71 #if defined(INET) || defined(INET6) 72 #include <netinet/in.h> 73 #include <netinet/ip_var.h> 74 #include <netinet/tcp_var.h> 75 #include <netinet/if_ether.h> 76 #include <netinet/ip_flow.h> 77 #include <net/ipfw/ip_fw.h> 78 #include <net/dummynet/ip_dummynet.h> 79 #endif 80 #ifdef INET6 81 #include <netinet6/nd6.h> 82 #endif 83 84 #ifdef CARP 85 #include <netinet/ip_carp.h> 86 #endif 87 88 #ifdef MPLS 89 #include <netproto/mpls/mpls.h> 90 #endif 91 92 /* netgraph node hooks for ng_ether(4) */ 93 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 94 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 95 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 96 void (*ng_ether_attach_p)(struct ifnet *ifp); 97 void (*ng_ether_detach_p)(struct ifnet *ifp); 98 99 void (*vlan_input_p)(struct mbuf *); 100 101 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 102 struct rtentry *); 103 static void ether_restore_header(struct mbuf **, const struct ether_header *, 104 const struct ether_header *); 105 static int ether_characterize(struct mbuf **); 106 static void ether_dispatch(int, struct mbuf *, int); 107 108 /* 109 * if_bridge support 110 */ 111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 112 int (*bridge_output_p)(struct ifnet *, struct mbuf *); 113 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 114 struct ifnet *(*bridge_interface_p)(void *if_bridge); 115 116 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 117 struct sockaddr *); 118 119 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 120 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 121 }; 122 123 #define gotoerr(e) do { error = (e); goto bad; } while (0) 124 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 125 126 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 127 struct ip_fw **rule, 128 const struct ether_header *eh); 129 130 static int ether_ipfw; 131 static u_long ether_restore_hdr; 132 static u_long ether_prepend_hdr; 133 static u_long ether_input_wronghash; 134 static int ether_debug; 135 136 #ifdef RSS_DEBUG 137 static u_long ether_pktinfo_try; 138 static u_long ether_pktinfo_hit; 139 static u_long ether_rss_nopi; 140 static u_long ether_rss_nohash; 141 static u_long ether_input_requeue; 142 #endif 143 static u_long ether_input_wronghwhash; 144 static int ether_input_ckhash; 145 146 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU) 147 148 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 149 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default); 150 151 SYSCTL_DECL(_net_link); 152 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 153 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 154 ðer_debug, 0, "Ether debug"); 155 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 156 ðer_ipfw, 0, "Pass ether pkts through firewall"); 157 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 158 ðer_restore_hdr, 0, "# of ether header restoration"); 159 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 160 ðer_prepend_hdr, 0, 161 "# of ether header restoration which prepends mbuf"); 162 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 163 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 164 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW, 165 ðer_tsolen_default, 0, "Default max TSO length"); 166 167 #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 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 182 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 183 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 184 ðer_input_ckhash, 0, "always check hash"); 185 186 #define ETHER_KTR_STR "ifp=%p" 187 #define ETHER_KTR_ARGS struct ifnet *ifp 188 #ifndef KTR_ETHERNET 189 #define KTR_ETHERNET KTR_ALL 190 #endif 191 KTR_INFO_MASTER(ether); 192 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 193 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 194 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 195 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 196 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 197 198 /* 199 * Ethernet output routine. 200 * Encapsulate a packet of type family for the local net. 201 * Use trailer local net encapsulation if enough data in first 202 * packet leaves a multiple of 512 bytes of data in remainder. 203 * Assumes that ifp is actually pointer to arpcom structure. 204 */ 205 static int 206 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 207 struct rtentry *rt) 208 { 209 struct ether_header *eh, *deh; 210 u_char *edst; 211 int loop_copy = 0; 212 int hlen = ETHER_HDR_LEN; /* link layer header length */ 213 struct arpcom *ac = IFP2AC(ifp); 214 int error; 215 216 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 217 218 if (ifp->if_flags & IFF_MONITOR) 219 gotoerr(ENETDOWN); 220 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 221 gotoerr(ENETDOWN); 222 223 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 224 if (m == NULL) 225 return (ENOBUFS); 226 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 227 eh = mtod(m, struct ether_header *); 228 edst = eh->ether_dhost; 229 230 /* 231 * Fill in the destination ethernet address and frame type. 232 */ 233 switch (dst->sa_family) { 234 #ifdef INET 235 case AF_INET: 236 if (!arpresolve(ifp, rt, m, dst, edst)) 237 return (0); /* if not yet resolved */ 238 #ifdef MPLS 239 if (m->m_flags & M_MPLSLABELED) 240 eh->ether_type = htons(ETHERTYPE_MPLS); 241 else 242 #endif 243 eh->ether_type = htons(ETHERTYPE_IP); 244 break; 245 #endif 246 #ifdef INET6 247 case AF_INET6: 248 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 249 return (0); /* Something bad happenned. */ 250 eh->ether_type = htons(ETHERTYPE_IPV6); 251 break; 252 #endif 253 case pseudo_AF_HDRCMPLT: 254 case AF_UNSPEC: 255 loop_copy = -1; /* if this is for us, don't do it */ 256 deh = (struct ether_header *)dst->sa_data; 257 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 258 eh->ether_type = deh->ether_type; 259 break; 260 261 default: 262 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 263 gotoerr(EAFNOSUPPORT); 264 } 265 266 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 267 memcpy(eh->ether_shost, 268 ((struct ether_header *)dst->sa_data)->ether_shost, 269 ETHER_ADDR_LEN); 270 else 271 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 272 273 /* 274 * Bridges require special output handling. 275 */ 276 if (ifp->if_bridge) { 277 KASSERT(bridge_output_p != NULL, 278 ("%s: if_bridge not loaded!", __func__)); 279 return bridge_output_p(ifp, m); 280 } 281 282 /* 283 * If a simplex interface, and the packet is being sent to our 284 * Ethernet address or a broadcast address, loopback a copy. 285 * XXX To make a simplex device behave exactly like a duplex 286 * device, we should copy in the case of sending to our own 287 * ethernet address (thus letting the original actually appear 288 * on the wire). However, we don't do that here for security 289 * reasons and compatibility with the original behavior. 290 */ 291 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 292 int csum_flags = 0; 293 294 if (m->m_pkthdr.csum_flags & CSUM_IP) 295 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 296 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 297 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 298 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 299 struct mbuf *n; 300 301 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 302 n->m_pkthdr.csum_flags |= csum_flags; 303 if (csum_flags & CSUM_DATA_VALID) 304 n->m_pkthdr.csum_data = 0xffff; 305 if_simloop(ifp, n, dst->sa_family, hlen); 306 } else 307 IFNET_STAT_INC(ifp, iqdrops, 1); 308 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 309 ETHER_ADDR_LEN) == 0) { 310 m->m_pkthdr.csum_flags |= csum_flags; 311 if (csum_flags & CSUM_DATA_VALID) 312 m->m_pkthdr.csum_data = 0xffff; 313 if_simloop(ifp, m, dst->sa_family, hlen); 314 return (0); /* XXX */ 315 } 316 } 317 318 #ifdef CARP 319 if (ifp->if_type == IFT_CARP) { 320 ifp = carp_parent(ifp); 321 if (ifp == NULL) 322 gotoerr(ENETUNREACH); 323 324 ac = IFP2AC(ifp); 325 326 /* 327 * Check precondition again 328 */ 329 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 330 331 if (ifp->if_flags & IFF_MONITOR) 332 gotoerr(ENETDOWN); 333 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 334 (IFF_UP | IFF_RUNNING)) 335 gotoerr(ENETDOWN); 336 } 337 #endif 338 339 /* Handle ng_ether(4) processing, if any */ 340 if (ng_ether_output_p != NULL) { 341 /* 342 * Hold BGL and recheck ng_ether_output_p 343 */ 344 get_mplock(); 345 if (ng_ether_output_p != NULL) { 346 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 347 rel_mplock(); 348 goto bad; 349 } 350 if (m == NULL) { 351 rel_mplock(); 352 return (0); 353 } 354 } 355 rel_mplock(); 356 } 357 358 /* Continue with link-layer output */ 359 return ether_output_frame(ifp, m); 360 361 bad: 362 m_freem(m); 363 return (error); 364 } 365 366 /* 367 * Returns the bridge interface an ifp is associated 368 * with. 369 * 370 * Only call if ifp->if_bridge != NULL. 371 */ 372 struct ifnet * 373 ether_bridge_interface(struct ifnet *ifp) 374 { 375 if (bridge_interface_p) 376 return(bridge_interface_p(ifp->if_bridge)); 377 return (ifp); 378 } 379 380 /* 381 * Ethernet link layer output routine to send a raw frame to the device. 382 * 383 * This assumes that the 14 byte Ethernet header is present and contiguous 384 * in the first mbuf. 385 */ 386 int 387 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 388 { 389 struct ip_fw *rule = NULL; 390 int error = 0; 391 struct altq_pktattr pktattr; 392 393 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 394 395 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 396 struct m_tag *mtag; 397 398 /* Extract info from dummynet tag */ 399 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 400 KKASSERT(mtag != NULL); 401 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 402 KKASSERT(rule != NULL); 403 404 m_tag_delete(m, mtag); 405 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 406 } 407 408 if (ifq_is_enabled(&ifp->if_snd)) 409 altq_etherclassify(&ifp->if_snd, m, &pktattr); 410 crit_enter(); 411 if (IPFW_LOADED && ether_ipfw != 0) { 412 struct ether_header save_eh, *eh; 413 414 eh = mtod(m, struct ether_header *); 415 save_eh = *eh; 416 m_adj(m, ETHER_HDR_LEN); 417 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 418 crit_exit(); 419 if (m != NULL) { 420 m_freem(m); 421 return ENOBUFS; /* pkt dropped */ 422 } else 423 return 0; /* consumed e.g. in a pipe */ 424 } 425 426 /* packet was ok, restore the ethernet header */ 427 ether_restore_header(&m, eh, &save_eh); 428 if (m == NULL) { 429 crit_exit(); 430 return ENOBUFS; 431 } 432 } 433 crit_exit(); 434 435 /* 436 * Queue message on interface, update output statistics if 437 * successful, and start output if interface not yet active. 438 */ 439 error = ifq_dispatch(ifp, m, &pktattr); 440 return (error); 441 } 442 443 /* 444 * ipfw processing for ethernet packets (in and out). 445 * The second parameter is NULL from ether_demux(), and ifp from 446 * ether_output_frame(). 447 */ 448 static boolean_t 449 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 450 const struct ether_header *eh) 451 { 452 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 453 struct ip_fw_args args; 454 struct m_tag *mtag; 455 struct mbuf *m; 456 int i; 457 458 if (*rule != NULL && fw_one_pass) 459 return TRUE; /* dummynet packet, already partially processed */ 460 461 /* 462 * I need some amount of data to be contiguous. 463 */ 464 i = min((*m0)->m_pkthdr.len, max_protohdr); 465 if ((*m0)->m_len < i) { 466 *m0 = m_pullup(*m0, i); 467 if (*m0 == NULL) 468 return FALSE; 469 } 470 471 /* 472 * Clean up tags 473 */ 474 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 475 m_tag_delete(*m0, mtag); 476 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 477 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 478 KKASSERT(mtag != NULL); 479 m_tag_delete(*m0, mtag); 480 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 481 } 482 483 args.m = *m0; /* the packet we are looking at */ 484 args.oif = dst; /* destination, if any */ 485 args.rule = *rule; /* matching rule to restart */ 486 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 487 i = ip_fw_chk_ptr(&args); 488 *m0 = args.m; 489 *rule = args.rule; 490 491 if (*m0 == NULL) 492 return FALSE; 493 494 switch (i) { 495 case IP_FW_PASS: 496 return TRUE; 497 498 case IP_FW_DIVERT: 499 case IP_FW_TEE: 500 case IP_FW_DENY: 501 /* 502 * XXX at some point add support for divert/forward actions. 503 * If none of the above matches, we have to drop the pkt. 504 */ 505 return FALSE; 506 507 case IP_FW_DUMMYNET: 508 /* 509 * Pass the pkt to dummynet, which consumes it. 510 */ 511 m = *m0; /* pass the original to dummynet */ 512 *m0 = NULL; /* and nothing back to the caller */ 513 514 ether_restore_header(&m, eh, &save_eh); 515 if (m == NULL) 516 return FALSE; 517 518 ip_fw_dn_io_ptr(m, args.cookie, 519 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 520 ip_dn_queue(m); 521 return FALSE; 522 523 default: 524 panic("unknown ipfw return value: %d", i); 525 } 526 } 527 528 static void 529 ether_input(struct ifnet *ifp, struct mbuf *m) 530 { 531 ether_input_pkt(ifp, m, NULL, -1); 532 } 533 534 /* 535 * Perform common duties while attaching to interface list 536 */ 537 void 538 ether_ifattach(struct ifnet *ifp, const uint8_t *lla, 539 lwkt_serialize_t serializer) 540 { 541 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 542 serializer); 543 } 544 545 void 546 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla, 547 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer) 548 { 549 struct sockaddr_dl *sdl; 550 char ethstr[ETHER_ADDRSTRLEN + 1]; 551 struct ifaltq *ifq; 552 int i; 553 554 ifp->if_type = IFT_ETHER; 555 ifp->if_addrlen = ETHER_ADDR_LEN; 556 ifp->if_hdrlen = ETHER_HDR_LEN; 557 if_attach(ifp, serializer); 558 ifq = &ifp->if_snd; 559 for (i = 0; i < ifq->altq_subq_cnt; ++i) { 560 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i); 561 562 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen * 563 (ETHER_MAX_LEN - ETHER_CRC_LEN); 564 } 565 ifp->if_mtu = ETHERMTU; 566 if (ifp->if_tsolen <= 0) { 567 if ((ether_tsolen_default / ETHERMTU) < 2) { 568 kprintf("ether TSO maxlen %d -> %d\n", 569 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 570 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 571 } 572 ifp->if_tsolen = ether_tsolen_default; 573 } 574 if (ifp->if_baudrate == 0) 575 ifp->if_baudrate = 10000000; 576 ifp->if_output = ether_output; 577 ifp->if_input = ether_input; 578 ifp->if_resolvemulti = ether_resolvemulti; 579 ifp->if_broadcastaddr = etherbroadcastaddr; 580 sdl = IF_LLSOCKADDR(ifp); 581 sdl->sdl_type = IFT_ETHER; 582 sdl->sdl_alen = ifp->if_addrlen; 583 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 584 /* 585 * XXX Keep the current drivers happy. 586 * XXX Remove once all drivers have been cleaned up 587 */ 588 if (lla != IFP2AC(ifp)->ac_enaddr) 589 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 590 bpfattach(ifp, dlt, hdrlen); 591 if (ng_ether_attach_p != NULL) 592 (*ng_ether_attach_p)(ifp); 593 594 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 595 } 596 597 /* 598 * Perform common duties while detaching an Ethernet interface 599 */ 600 void 601 ether_ifdetach(struct ifnet *ifp) 602 { 603 if_down(ifp); 604 605 if (ng_ether_detach_p != NULL) 606 (*ng_ether_detach_p)(ifp); 607 bpfdetach(ifp); 608 if_detach(ifp); 609 } 610 611 int 612 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 613 { 614 struct ifaddr *ifa = (struct ifaddr *) data; 615 struct ifreq *ifr = (struct ifreq *) data; 616 int error = 0; 617 618 #define IF_INIT(ifp) \ 619 do { \ 620 if (((ifp)->if_flags & IFF_UP) == 0) { \ 621 (ifp)->if_flags |= IFF_UP; \ 622 (ifp)->if_init((ifp)->if_softc); \ 623 } \ 624 } while (0) 625 626 ASSERT_IFNET_SERIALIZED_ALL(ifp); 627 628 switch (command) { 629 case SIOCSIFADDR: 630 switch (ifa->ifa_addr->sa_family) { 631 #ifdef INET 632 case AF_INET: 633 IF_INIT(ifp); /* before arpwhohas */ 634 arp_ifinit(ifp, ifa); 635 break; 636 #endif 637 default: 638 IF_INIT(ifp); 639 break; 640 } 641 break; 642 643 case SIOCGIFADDR: 644 bcopy(IFP2AC(ifp)->ac_enaddr, 645 ((struct sockaddr *)ifr->ifr_data)->sa_data, 646 ETHER_ADDR_LEN); 647 break; 648 649 case SIOCSIFMTU: 650 /* 651 * Set the interface MTU. 652 */ 653 if (ifr->ifr_mtu > ETHERMTU) { 654 error = EINVAL; 655 } else { 656 ifp->if_mtu = ifr->ifr_mtu; 657 } 658 break; 659 default: 660 error = EINVAL; 661 break; 662 } 663 return (error); 664 665 #undef IF_INIT 666 } 667 668 int 669 ether_resolvemulti( 670 struct ifnet *ifp, 671 struct sockaddr **llsa, 672 struct sockaddr *sa) 673 { 674 struct sockaddr_dl *sdl; 675 #ifdef INET 676 struct sockaddr_in *sin; 677 #endif 678 #ifdef INET6 679 struct sockaddr_in6 *sin6; 680 #endif 681 u_char *e_addr; 682 683 switch(sa->sa_family) { 684 case AF_LINK: 685 /* 686 * No mapping needed. Just check that it's a valid MC address. 687 */ 688 sdl = (struct sockaddr_dl *)sa; 689 e_addr = LLADDR(sdl); 690 if ((e_addr[0] & 1) != 1) 691 return EADDRNOTAVAIL; 692 *llsa = NULL; 693 return 0; 694 695 #ifdef INET 696 case AF_INET: 697 sin = (struct sockaddr_in *)sa; 698 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 699 return EADDRNOTAVAIL; 700 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 701 sdl->sdl_len = sizeof *sdl; 702 sdl->sdl_family = AF_LINK; 703 sdl->sdl_index = ifp->if_index; 704 sdl->sdl_type = IFT_ETHER; 705 sdl->sdl_alen = ETHER_ADDR_LEN; 706 e_addr = LLADDR(sdl); 707 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 708 *llsa = (struct sockaddr *)sdl; 709 return 0; 710 #endif 711 #ifdef INET6 712 case AF_INET6: 713 sin6 = (struct sockaddr_in6 *)sa; 714 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 715 /* 716 * An IP6 address of 0 means listen to all 717 * of the Ethernet multicast address used for IP6. 718 * (This is used for multicast routers.) 719 */ 720 ifp->if_flags |= IFF_ALLMULTI; 721 *llsa = NULL; 722 return 0; 723 } 724 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 725 return EADDRNOTAVAIL; 726 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 727 sdl->sdl_len = sizeof *sdl; 728 sdl->sdl_family = AF_LINK; 729 sdl->sdl_index = ifp->if_index; 730 sdl->sdl_type = IFT_ETHER; 731 sdl->sdl_alen = ETHER_ADDR_LEN; 732 e_addr = LLADDR(sdl); 733 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 734 *llsa = (struct sockaddr *)sdl; 735 return 0; 736 #endif 737 738 default: 739 /* 740 * Well, the text isn't quite right, but it's the name 741 * that counts... 742 */ 743 return EAFNOSUPPORT; 744 } 745 } 746 747 #if 0 748 /* 749 * This is for reference. We have a table-driven version 750 * of the little-endian crc32 generator, which is faster 751 * than the double-loop. 752 */ 753 uint32_t 754 ether_crc32_le(const uint8_t *buf, size_t len) 755 { 756 uint32_t c, crc, carry; 757 size_t i, j; 758 759 crc = 0xffffffffU; /* initial value */ 760 761 for (i = 0; i < len; i++) { 762 c = buf[i]; 763 for (j = 0; j < 8; j++) { 764 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 765 crc >>= 1; 766 c >>= 1; 767 if (carry) 768 crc = (crc ^ ETHER_CRC_POLY_LE); 769 } 770 } 771 772 return (crc); 773 } 774 #else 775 uint32_t 776 ether_crc32_le(const uint8_t *buf, size_t len) 777 { 778 static const uint32_t crctab[] = { 779 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 780 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 781 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 782 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 783 }; 784 uint32_t crc; 785 size_t i; 786 787 crc = 0xffffffffU; /* initial value */ 788 789 for (i = 0; i < len; i++) { 790 crc ^= buf[i]; 791 crc = (crc >> 4) ^ crctab[crc & 0xf]; 792 crc = (crc >> 4) ^ crctab[crc & 0xf]; 793 } 794 795 return (crc); 796 } 797 #endif 798 799 uint32_t 800 ether_crc32_be(const uint8_t *buf, size_t len) 801 { 802 uint32_t c, crc, carry; 803 size_t i, j; 804 805 crc = 0xffffffffU; /* initial value */ 806 807 for (i = 0; i < len; i++) { 808 c = buf[i]; 809 for (j = 0; j < 8; j++) { 810 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 811 crc <<= 1; 812 c >>= 1; 813 if (carry) 814 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 815 } 816 } 817 818 return (crc); 819 } 820 821 /* 822 * find the size of ethernet header, and call classifier 823 */ 824 void 825 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 826 struct altq_pktattr *pktattr) 827 { 828 struct ether_header *eh; 829 uint16_t ether_type; 830 int hlen, af, hdrsize; 831 832 hlen = sizeof(struct ether_header); 833 eh = mtod(m, struct ether_header *); 834 835 ether_type = ntohs(eh->ether_type); 836 if (ether_type < ETHERMTU) { 837 /* ick! LLC/SNAP */ 838 struct llc *llc = (struct llc *)(eh + 1); 839 hlen += 8; 840 841 if (m->m_len < hlen || 842 llc->llc_dsap != LLC_SNAP_LSAP || 843 llc->llc_ssap != LLC_SNAP_LSAP || 844 llc->llc_control != LLC_UI) 845 goto bad; /* not snap! */ 846 847 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 848 } 849 850 if (ether_type == ETHERTYPE_IP) { 851 af = AF_INET; 852 hdrsize = 20; /* sizeof(struct ip) */ 853 #ifdef INET6 854 } else if (ether_type == ETHERTYPE_IPV6) { 855 af = AF_INET6; 856 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 857 #endif 858 } else 859 goto bad; 860 861 while (m->m_len <= hlen) { 862 hlen -= m->m_len; 863 m = m->m_next; 864 } 865 if (m->m_len < hlen + hdrsize) { 866 /* 867 * ip header is not in a single mbuf. this should not 868 * happen in the current code. 869 * (todo: use m_pulldown in the future) 870 */ 871 goto bad; 872 } 873 m->m_data += hlen; 874 m->m_len -= hlen; 875 ifq_classify(ifq, m, af, pktattr); 876 m->m_data -= hlen; 877 m->m_len += hlen; 878 879 return; 880 881 bad: 882 pktattr->pattr_class = NULL; 883 pktattr->pattr_hdr = NULL; 884 pktattr->pattr_af = AF_UNSPEC; 885 } 886 887 static void 888 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 889 const struct ether_header *save_eh) 890 { 891 struct mbuf *m = *m0; 892 893 ether_restore_hdr++; 894 895 /* 896 * Prepend the header, optimize for the common case of 897 * eh pointing into the mbuf. 898 */ 899 if ((const void *)(eh + 1) == (void *)m->m_data) { 900 m->m_data -= ETHER_HDR_LEN; 901 m->m_len += ETHER_HDR_LEN; 902 m->m_pkthdr.len += ETHER_HDR_LEN; 903 } else { 904 ether_prepend_hdr++; 905 906 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 907 if (m != NULL) { 908 bcopy(save_eh, mtod(m, struct ether_header *), 909 ETHER_HDR_LEN); 910 } 911 } 912 *m0 = m; 913 } 914 915 /* 916 * Upper layer processing for a received Ethernet packet. 917 */ 918 void 919 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 920 { 921 struct ether_header *eh; 922 int isr, discard = 0; 923 u_short ether_type; 924 struct ip_fw *rule = NULL; 925 926 M_ASSERTPKTHDR(m); 927 KASSERT(m->m_len >= ETHER_HDR_LEN, 928 ("ether header is not contiguous!")); 929 930 eh = mtod(m, struct ether_header *); 931 932 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 933 struct m_tag *mtag; 934 935 /* Extract info from dummynet tag */ 936 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 937 KKASSERT(mtag != NULL); 938 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 939 KKASSERT(rule != NULL); 940 941 m_tag_delete(m, mtag); 942 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 943 944 /* packet is passing the second time */ 945 goto post_stats; 946 } 947 948 /* 949 * We got a packet which was unicast to a different Ethernet 950 * address. If the driver is working properly, then this 951 * situation can only happen when the interface is in 952 * promiscuous mode. We defer the packet discarding until the 953 * vlan processing is done, so that vlan/bridge or vlan/netgraph 954 * could work. 955 */ 956 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 957 !ETHER_IS_MULTICAST(eh->ether_dhost) && 958 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 959 if (ether_debug & 1) { 960 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 961 "%02x:%02x:%02x:%02x:%02x:%02x " 962 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 963 eh->ether_dhost[0], 964 eh->ether_dhost[1], 965 eh->ether_dhost[2], 966 eh->ether_dhost[3], 967 eh->ether_dhost[4], 968 eh->ether_dhost[5], 969 eh->ether_shost[0], 970 eh->ether_shost[1], 971 eh->ether_shost[2], 972 eh->ether_shost[3], 973 eh->ether_shost[4], 974 eh->ether_shost[5], 975 eh->ether_type, 976 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 977 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 978 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 979 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 980 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 981 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 982 ); 983 } 984 if ((ether_debug & 2) == 0) 985 discard = 1; 986 } 987 988 post_stats: 989 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 990 struct ether_header save_eh = *eh; 991 992 /* XXX old crufty stuff, needs to be removed */ 993 m_adj(m, sizeof(struct ether_header)); 994 995 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 996 m_freem(m); 997 return; 998 } 999 1000 ether_restore_header(&m, eh, &save_eh); 1001 if (m == NULL) 1002 return; 1003 eh = mtod(m, struct ether_header *); 1004 } 1005 1006 ether_type = ntohs(eh->ether_type); 1007 KKASSERT(ether_type != ETHERTYPE_VLAN); 1008 1009 if (m->m_flags & M_VLANTAG) { 1010 void (*vlan_input_func)(struct mbuf *); 1011 1012 vlan_input_func = vlan_input_p; 1013 if (vlan_input_func != NULL) { 1014 vlan_input_func(m); 1015 } else { 1016 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1017 m_freem(m); 1018 } 1019 return; 1020 } 1021 1022 /* 1023 * If we have been asked to discard this packet 1024 * (e.g. not for us), drop it before entering 1025 * the upper layer. 1026 */ 1027 if (discard) { 1028 m_freem(m); 1029 return; 1030 } 1031 1032 /* 1033 * Clear protocol specific flags, 1034 * before entering the upper layer. 1035 */ 1036 m->m_flags &= ~M_ETHER_FLAGS; 1037 1038 /* Strip ethernet header. */ 1039 m_adj(m, sizeof(struct ether_header)); 1040 1041 switch (ether_type) { 1042 #ifdef INET 1043 case ETHERTYPE_IP: 1044 if ((m->m_flags & M_LENCHECKED) == 0) { 1045 if (!ip_lengthcheck(&m, 0)) 1046 return; 1047 } 1048 if (ipflow_fastforward(m)) 1049 return; 1050 isr = NETISR_IP; 1051 break; 1052 1053 case ETHERTYPE_ARP: 1054 if (ifp->if_flags & IFF_NOARP) { 1055 /* Discard packet if ARP is disabled on interface */ 1056 m_freem(m); 1057 return; 1058 } 1059 isr = NETISR_ARP; 1060 break; 1061 #endif 1062 1063 #ifdef INET6 1064 case ETHERTYPE_IPV6: 1065 isr = NETISR_IPV6; 1066 break; 1067 #endif 1068 1069 #ifdef MPLS 1070 case ETHERTYPE_MPLS: 1071 case ETHERTYPE_MPLS_MCAST: 1072 /* Should have been set by ether_input_pkt(). */ 1073 KKASSERT(m->m_flags & M_MPLSLABELED); 1074 isr = NETISR_MPLS; 1075 break; 1076 #endif 1077 1078 default: 1079 /* 1080 * The accurate msgport is not determined before 1081 * we reach here, so recharacterize packet. 1082 */ 1083 m->m_flags &= ~M_HASH; 1084 if (ng_ether_input_orphan_p != NULL) { 1085 /* 1086 * Put back the ethernet header so netgraph has a 1087 * consistent view of inbound packets. 1088 */ 1089 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1090 if (m == NULL) { 1091 /* 1092 * M_PREPEND frees the mbuf in case of failure. 1093 */ 1094 return; 1095 } 1096 /* 1097 * Hold BGL and recheck ng_ether_input_orphan_p 1098 */ 1099 get_mplock(); 1100 if (ng_ether_input_orphan_p != NULL) { 1101 ng_ether_input_orphan_p(ifp, m); 1102 rel_mplock(); 1103 return; 1104 } 1105 rel_mplock(); 1106 } 1107 m_freem(m); 1108 return; 1109 } 1110 1111 if (m->m_flags & M_HASH) { 1112 if (&curthread->td_msgport == 1113 netisr_hashport(m->m_pkthdr.hash)) { 1114 netisr_handle(isr, m); 1115 return; 1116 } else { 1117 /* 1118 * XXX Something is wrong, 1119 * we probably should panic here! 1120 */ 1121 m->m_flags &= ~M_HASH; 1122 atomic_add_long(ðer_input_wronghash, 1); 1123 } 1124 } 1125 #ifdef RSS_DEBUG 1126 atomic_add_long(ðer_input_requeue, 1); 1127 #endif 1128 netisr_queue(isr, m); 1129 } 1130 1131 /* 1132 * First we perform any link layer operations, then continue to the 1133 * upper layers with ether_demux_oncpu(). 1134 */ 1135 static void 1136 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1137 { 1138 #ifdef CARP 1139 void *carp; 1140 #endif 1141 1142 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1143 /* 1144 * Receiving interface's flags are changed, when this 1145 * packet is waiting for processing; discard it. 1146 */ 1147 m_freem(m); 1148 return; 1149 } 1150 1151 /* 1152 * Tap the packet off here for a bridge. bridge_input() 1153 * will return NULL if it has consumed the packet, otherwise 1154 * it gets processed as normal. Note that bridge_input() 1155 * will always return the original packet if we need to 1156 * process it locally. 1157 */ 1158 if (ifp->if_bridge) { 1159 KASSERT(bridge_input_p != NULL, 1160 ("%s: if_bridge not loaded!", __func__)); 1161 1162 if(m->m_flags & M_ETHER_BRIDGED) { 1163 m->m_flags &= ~M_ETHER_BRIDGED; 1164 } else { 1165 m = bridge_input_p(ifp, m); 1166 if (m == NULL) 1167 return; 1168 1169 KASSERT(ifp == m->m_pkthdr.rcvif, 1170 ("bridge_input_p changed rcvif")); 1171 } 1172 } 1173 1174 #ifdef CARP 1175 carp = ifp->if_carp; 1176 if (carp) { 1177 m = carp_input(carp, m); 1178 if (m == NULL) 1179 return; 1180 KASSERT(ifp == m->m_pkthdr.rcvif, 1181 ("carp_input changed rcvif")); 1182 } 1183 #endif 1184 1185 /* Handle ng_ether(4) processing, if any */ 1186 if (ng_ether_input_p != NULL) { 1187 /* 1188 * Hold BGL and recheck ng_ether_input_p 1189 */ 1190 get_mplock(); 1191 if (ng_ether_input_p != NULL) 1192 ng_ether_input_p(ifp, &m); 1193 rel_mplock(); 1194 1195 if (m == NULL) 1196 return; 1197 } 1198 1199 /* Continue with upper layer processing */ 1200 ether_demux_oncpu(ifp, m); 1201 } 1202 1203 /* 1204 * Perform certain functions of ether_input_pkt(): 1205 * - Test IFF_UP 1206 * - Update statistics 1207 * - Run bpf(4) tap if requested 1208 * Then pass the packet to ether_input_oncpu(). 1209 * 1210 * This function should be used by pseudo interface (e.g. vlan(4)), 1211 * when it tries to claim that the packet is received by it. 1212 * 1213 * REINPUT_KEEPRCVIF 1214 * REINPUT_RUNBPF 1215 */ 1216 void 1217 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1218 { 1219 /* Discard packet if interface is not up */ 1220 if (!(ifp->if_flags & IFF_UP)) { 1221 m_freem(m); 1222 return; 1223 } 1224 1225 /* 1226 * Change receiving interface. The bridge will often pass a flag to 1227 * ask that this not be done so ARPs get applied to the correct 1228 * side. 1229 */ 1230 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1231 m->m_pkthdr.rcvif == NULL) { 1232 m->m_pkthdr.rcvif = ifp; 1233 } 1234 1235 /* Update statistics */ 1236 IFNET_STAT_INC(ifp, ipackets, 1); 1237 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1238 if (m->m_flags & (M_MCAST | M_BCAST)) 1239 IFNET_STAT_INC(ifp, imcasts, 1); 1240 1241 if (reinput_flags & REINPUT_RUNBPF) 1242 BPF_MTAP(ifp, m); 1243 1244 ether_input_oncpu(ifp, m); 1245 } 1246 1247 static __inline boolean_t 1248 ether_vlancheck(struct mbuf **m0) 1249 { 1250 struct mbuf *m = *m0; 1251 struct ether_header *eh; 1252 uint16_t ether_type; 1253 1254 eh = mtod(m, struct ether_header *); 1255 ether_type = ntohs(eh->ether_type); 1256 1257 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1258 /* 1259 * Extract vlan tag if hardware does not do it for us 1260 */ 1261 vlan_ether_decap(&m); 1262 if (m == NULL) 1263 goto failed; 1264 1265 eh = mtod(m, struct ether_header *); 1266 ether_type = ntohs(eh->ether_type); 1267 } 1268 1269 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1270 /* 1271 * To prevent possible dangerous recursion, 1272 * we don't do vlan-in-vlan 1273 */ 1274 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1275 goto failed; 1276 } 1277 KKASSERT(ether_type != ETHERTYPE_VLAN); 1278 1279 m->m_flags |= M_ETHER_VLANCHECKED; 1280 *m0 = m; 1281 return TRUE; 1282 failed: 1283 if (m != NULL) 1284 m_freem(m); 1285 *m0 = NULL; 1286 return FALSE; 1287 } 1288 1289 static void 1290 ether_input_handler(netmsg_t nmsg) 1291 { 1292 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1293 struct ether_header *eh; 1294 struct ifnet *ifp; 1295 struct mbuf *m; 1296 1297 m = nmp->nm_packet; 1298 M_ASSERTPKTHDR(m); 1299 1300 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1301 if (!ether_vlancheck(&m)) { 1302 KKASSERT(m == NULL); 1303 return; 1304 } 1305 } 1306 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) || 1307 __predict_false(ether_input_ckhash)) { 1308 int isr; 1309 1310 /* 1311 * Need to verify the hash supplied by the hardware 1312 * which could be wrong. 1313 */ 1314 m->m_flags &= ~(M_HASH | M_CKHASH); 1315 isr = ether_characterize(&m); 1316 if (m == NULL) 1317 return; 1318 KKASSERT(m->m_flags & M_HASH); 1319 1320 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) { 1321 /* 1322 * Wrong hardware supplied hash; redispatch 1323 */ 1324 ether_dispatch(isr, m, -1); 1325 if (__predict_false(ether_input_ckhash)) 1326 atomic_add_long(ðer_input_wronghwhash, 1); 1327 return; 1328 } 1329 } 1330 ifp = m->m_pkthdr.rcvif; 1331 1332 eh = mtod(m, struct ether_header *); 1333 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1334 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1335 ifp->if_addrlen) == 0) 1336 m->m_flags |= M_BCAST; 1337 else 1338 m->m_flags |= M_MCAST; 1339 IFNET_STAT_INC(ifp, imcasts, 1); 1340 } 1341 1342 ether_input_oncpu(ifp, m); 1343 } 1344 1345 /* 1346 * Send the packet to the target netisr msgport 1347 * 1348 * At this point the packet must be characterized (M_HASH set), 1349 * so we know which netisr to send it to. 1350 */ 1351 static void 1352 ether_dispatch(int isr, struct mbuf *m, int cpuid) 1353 { 1354 struct netmsg_packet *pmsg; 1355 int target_cpuid; 1356 1357 KKASSERT(m->m_flags & M_HASH); 1358 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash); 1359 1360 pmsg = &m->m_hdr.mh_netmsg; 1361 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1362 0, ether_input_handler); 1363 pmsg->nm_packet = m; 1364 pmsg->base.lmsg.u.ms_result = isr; 1365 1366 logether(disp_beg, NULL); 1367 if (target_cpuid == cpuid) { 1368 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid), 1369 &pmsg->base.lmsg); 1370 } else { 1371 lwkt_sendmsg(netisr_cpuport(target_cpuid), 1372 &pmsg->base.lmsg); 1373 } 1374 logether(disp_end, NULL); 1375 } 1376 1377 /* 1378 * Process a received Ethernet packet. 1379 * 1380 * The ethernet header is assumed to be in the mbuf so the caller 1381 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1382 * bytes in the first mbuf. 1383 * 1384 * If the caller knows that the current thread is stick to the current 1385 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid 1386 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should 1387 * be passed as 'cpuid' argument. 1388 */ 1389 void 1390 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1391 int cpuid) 1392 { 1393 int isr; 1394 1395 M_ASSERTPKTHDR(m); 1396 1397 /* Discard packet if interface is not up */ 1398 if (!(ifp->if_flags & IFF_UP)) { 1399 m_freem(m); 1400 return; 1401 } 1402 1403 if (m->m_len < sizeof(struct ether_header)) { 1404 /* XXX error in the caller. */ 1405 m_freem(m); 1406 return; 1407 } 1408 1409 m->m_pkthdr.rcvif = ifp; 1410 1411 logether(pkt_beg, ifp); 1412 1413 ETHER_BPF_MTAP(ifp, m); 1414 1415 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1416 1417 if (ifp->if_flags & IFF_MONITOR) { 1418 struct ether_header *eh; 1419 1420 eh = mtod(m, struct ether_header *); 1421 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1422 IFNET_STAT_INC(ifp, imcasts, 1); 1423 1424 /* 1425 * Interface marked for monitoring; discard packet. 1426 */ 1427 m_freem(m); 1428 1429 logether(pkt_end, ifp); 1430 return; 1431 } 1432 1433 /* 1434 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1435 * we can dispatch it immediately with trivial checks. 1436 */ 1437 if (pi != NULL && (m->m_flags & M_HASH)) { 1438 #ifdef RSS_DEBUG 1439 atomic_add_long(ðer_pktinfo_try, 1); 1440 #endif 1441 netisr_hashcheck(pi->pi_netisr, m, pi); 1442 if (m->m_flags & M_HASH) { 1443 ether_dispatch(pi->pi_netisr, m, cpuid); 1444 #ifdef RSS_DEBUG 1445 atomic_add_long(ðer_pktinfo_hit, 1); 1446 #endif 1447 logether(pkt_end, ifp); 1448 return; 1449 } 1450 } 1451 #ifdef RSS_DEBUG 1452 else if (ifp->if_capenable & IFCAP_RSS) { 1453 if (pi == NULL) 1454 atomic_add_long(ðer_rss_nopi, 1); 1455 else 1456 atomic_add_long(ðer_rss_nohash, 1); 1457 } 1458 #endif 1459 1460 /* 1461 * Packet hash will be recalculated by software, so clear 1462 * the M_HASH and M_CKHASH flag set by the driver; the hash 1463 * value calculated by the hardware may not be exactly what 1464 * we want. 1465 */ 1466 m->m_flags &= ~(M_HASH | M_CKHASH); 1467 1468 if (!ether_vlancheck(&m)) { 1469 KKASSERT(m == NULL); 1470 logether(pkt_end, ifp); 1471 return; 1472 } 1473 1474 isr = ether_characterize(&m); 1475 if (m == NULL) { 1476 logether(pkt_end, ifp); 1477 return; 1478 } 1479 1480 /* 1481 * Finally dispatch it 1482 */ 1483 ether_dispatch(isr, m, cpuid); 1484 1485 logether(pkt_end, ifp); 1486 } 1487 1488 static int 1489 ether_characterize(struct mbuf **m0) 1490 { 1491 struct mbuf *m = *m0; 1492 struct ether_header *eh; 1493 uint16_t ether_type; 1494 int isr; 1495 1496 eh = mtod(m, struct ether_header *); 1497 ether_type = ntohs(eh->ether_type); 1498 1499 /* 1500 * Map ether type to netisr id. 1501 */ 1502 switch (ether_type) { 1503 #ifdef INET 1504 case ETHERTYPE_IP: 1505 isr = NETISR_IP; 1506 break; 1507 1508 case ETHERTYPE_ARP: 1509 isr = NETISR_ARP; 1510 break; 1511 #endif 1512 1513 #ifdef INET6 1514 case ETHERTYPE_IPV6: 1515 isr = NETISR_IPV6; 1516 break; 1517 #endif 1518 1519 #ifdef MPLS 1520 case ETHERTYPE_MPLS: 1521 case ETHERTYPE_MPLS_MCAST: 1522 m->m_flags |= M_MPLSLABELED; 1523 isr = NETISR_MPLS; 1524 break; 1525 #endif 1526 1527 default: 1528 /* 1529 * NETISR_MAX is an invalid value; it is chosen to let 1530 * netisr_characterize() know that we have no clear 1531 * idea where this packet should go. 1532 */ 1533 isr = NETISR_MAX; 1534 break; 1535 } 1536 1537 /* 1538 * Ask the isr to characterize the packet since we couldn't. 1539 * This is an attempt to optimally get us onto the correct protocol 1540 * thread. 1541 */ 1542 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1543 1544 *m0 = m; 1545 return isr; 1546 } 1547 1548 static void 1549 ether_demux_handler(netmsg_t nmsg) 1550 { 1551 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1552 struct ifnet *ifp; 1553 struct mbuf *m; 1554 1555 m = nmp->nm_packet; 1556 M_ASSERTPKTHDR(m); 1557 ifp = m->m_pkthdr.rcvif; 1558 1559 ether_demux_oncpu(ifp, m); 1560 } 1561 1562 void 1563 ether_demux(struct mbuf *m) 1564 { 1565 struct netmsg_packet *pmsg; 1566 int isr; 1567 1568 isr = ether_characterize(&m); 1569 if (m == NULL) 1570 return; 1571 1572 KKASSERT(m->m_flags & M_HASH); 1573 pmsg = &m->m_hdr.mh_netmsg; 1574 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1575 0, ether_demux_handler); 1576 pmsg->nm_packet = m; 1577 pmsg->base.lmsg.u.ms_result = isr; 1578 1579 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1580 } 1581 1582 u_char * 1583 kether_aton(const char *macstr, u_char *addr) 1584 { 1585 unsigned int o0, o1, o2, o3, o4, o5; 1586 int n; 1587 1588 if (macstr == NULL || addr == NULL) 1589 return NULL; 1590 1591 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1592 &o3, &o4, &o5); 1593 if (n != 6) 1594 return NULL; 1595 1596 addr[0] = o0; 1597 addr[1] = o1; 1598 addr[2] = o2; 1599 addr[3] = o3; 1600 addr[4] = o4; 1601 addr[5] = o5; 1602 1603 return addr; 1604 } 1605 1606 char * 1607 kether_ntoa(const u_char *addr, char *buf) 1608 { 1609 int len = ETHER_ADDRSTRLEN + 1; 1610 int n; 1611 1612 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1613 addr[1], addr[2], addr[3], addr[4], addr[5]); 1614 1615 if (n < 17) 1616 return NULL; 1617 1618 return buf; 1619 } 1620 1621 MODULE_VERSION(ether, 1); 1622