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_ipx.h" 36 #include "opt_mpls.h" 37 #include "opt_netgraph.h" 38 #include "opt_carp.h" 39 #include "opt_rss.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/globaldata.h> 44 #include <sys/kernel.h> 45 #include <sys/ktr.h> 46 #include <sys/lock.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/msgport.h> 50 #include <sys/socket.h> 51 #include <sys/sockio.h> 52 #include <sys/sysctl.h> 53 #include <sys/thread.h> 54 55 #include <sys/thread2.h> 56 #include <sys/mplock2.h> 57 58 #include <net/if.h> 59 #include <net/netisr.h> 60 #include <net/route.h> 61 #include <net/if_llc.h> 62 #include <net/if_dl.h> 63 #include <net/if_types.h> 64 #include <net/ifq_var.h> 65 #include <net/bpf.h> 66 #include <net/ethernet.h> 67 #include <net/vlan/if_vlan_ether.h> 68 #include <net/vlan/if_vlan_var.h> 69 #include <net/netmsg2.h> 70 #include <net/netisr2.h> 71 72 #if defined(INET) || defined(INET6) 73 #include <netinet/in.h> 74 #include <netinet/ip_var.h> 75 #include <netinet/tcp_var.h> 76 #include <netinet/if_ether.h> 77 #include <netinet/ip_flow.h> 78 #include <net/ipfw/ip_fw.h> 79 #include <net/dummynet/ip_dummynet.h> 80 #endif 81 #ifdef INET6 82 #include <netinet6/nd6.h> 83 #endif 84 85 #ifdef CARP 86 #include <netinet/ip_carp.h> 87 #endif 88 89 #ifdef IPX 90 #include <netproto/ipx/ipx.h> 91 #include <netproto/ipx/ipx_if.h> 92 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 93 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 94 short *tp, int *hlen); 95 #endif 96 97 #ifdef MPLS 98 #include <netproto/mpls/mpls.h> 99 #endif 100 101 /* netgraph node hooks for ng_ether(4) */ 102 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 103 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 104 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 105 void (*ng_ether_attach_p)(struct ifnet *ifp); 106 void (*ng_ether_detach_p)(struct ifnet *ifp); 107 108 void (*vlan_input_p)(struct mbuf *); 109 110 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 111 struct rtentry *); 112 static void ether_restore_header(struct mbuf **, const struct ether_header *, 113 const struct ether_header *); 114 static int ether_characterize(struct mbuf **); 115 static void ether_dispatch(int, 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 static u_long ether_input_wronghwhash; 153 static int ether_input_ckhash; 154 155 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU) 156 157 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 158 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default); 159 160 SYSCTL_DECL(_net_link); 161 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 162 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 163 ðer_debug, 0, "Ether debug"); 164 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 165 ðer_ipfw, 0, "Pass ether pkts through firewall"); 166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 167 ðer_restore_hdr, 0, "# of ether header restoration"); 168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 169 ðer_prepend_hdr, 0, 170 "# of ether header restoration which prepends mbuf"); 171 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 172 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 173 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW, 174 ðer_tsolen_default, 0, "Default max TSO length"); 175 176 #ifdef RSS_DEBUG 177 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 178 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 180 ðer_rss_nohash, 0, "# of packets do not have hash"); 181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 182 ðer_pktinfo_try, 0, 183 "# of tries to find packets' msgport using pktinfo"); 184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 185 ðer_pktinfo_hit, 0, 186 "# of packets whose msgport are found using pktinfo"); 187 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 188 ðer_input_requeue, 0, "# of input packets gets requeued"); 189 #endif 190 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 191 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 192 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 193 ðer_input_ckhash, 0, "always check hash"); 194 195 #define ETHER_KTR_STR "ifp=%p" 196 #define ETHER_KTR_ARGS struct ifnet *ifp 197 #ifndef KTR_ETHERNET 198 #define KTR_ETHERNET KTR_ALL 199 #endif 200 KTR_INFO_MASTER(ether); 201 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 202 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 203 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 204 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 205 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 206 207 /* 208 * Ethernet output routine. 209 * Encapsulate a packet of type family for the local net. 210 * Use trailer local net encapsulation if enough data in first 211 * packet leaves a multiple of 512 bytes of data in remainder. 212 * Assumes that ifp is actually pointer to arpcom structure. 213 */ 214 static int 215 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 216 struct rtentry *rt) 217 { 218 struct ether_header *eh, *deh; 219 u_char *edst; 220 int loop_copy = 0; 221 int hlen = ETHER_HDR_LEN; /* link layer header length */ 222 struct arpcom *ac = IFP2AC(ifp); 223 int error; 224 225 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 226 227 if (ifp->if_flags & IFF_MONITOR) 228 gotoerr(ENETDOWN); 229 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 230 gotoerr(ENETDOWN); 231 232 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT); 233 if (m == NULL) 234 return (ENOBUFS); 235 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 236 eh = mtod(m, struct ether_header *); 237 edst = eh->ether_dhost; 238 239 /* 240 * Fill in the destination ethernet address and frame type. 241 */ 242 switch (dst->sa_family) { 243 #ifdef INET 244 case AF_INET: 245 if (!arpresolve(ifp, rt, m, dst, edst)) 246 return (0); /* if not yet resolved */ 247 #ifdef MPLS 248 if (m->m_flags & M_MPLSLABELED) 249 eh->ether_type = htons(ETHERTYPE_MPLS); 250 else 251 #endif 252 eh->ether_type = htons(ETHERTYPE_IP); 253 break; 254 #endif 255 #ifdef INET6 256 case AF_INET6: 257 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 258 return (0); /* Something bad happenned. */ 259 eh->ether_type = htons(ETHERTYPE_IPV6); 260 break; 261 #endif 262 #ifdef IPX 263 case AF_IPX: 264 if (ef_outputp != NULL) { 265 /* 266 * Hold BGL and recheck ef_outputp 267 */ 268 get_mplock(); 269 if (ef_outputp != NULL) { 270 error = ef_outputp(ifp, &m, dst, 271 &eh->ether_type, &hlen); 272 rel_mplock(); 273 if (error) 274 goto bad; 275 else 276 break; 277 } 278 rel_mplock(); 279 } 280 eh->ether_type = htons(ETHERTYPE_IPX); 281 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 282 edst, ETHER_ADDR_LEN); 283 break; 284 #endif 285 case pseudo_AF_HDRCMPLT: 286 case AF_UNSPEC: 287 loop_copy = -1; /* if this is for us, don't do it */ 288 deh = (struct ether_header *)dst->sa_data; 289 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 290 eh->ether_type = deh->ether_type; 291 break; 292 293 default: 294 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 295 gotoerr(EAFNOSUPPORT); 296 } 297 298 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 299 memcpy(eh->ether_shost, 300 ((struct ether_header *)dst->sa_data)->ether_shost, 301 ETHER_ADDR_LEN); 302 else 303 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 304 305 /* 306 * Bridges require special output handling. 307 */ 308 if (ifp->if_bridge) { 309 KASSERT(bridge_output_p != NULL, 310 ("%s: if_bridge not loaded!", __func__)); 311 return bridge_output_p(ifp, m); 312 } 313 314 /* 315 * If a simplex interface, and the packet is being sent to our 316 * Ethernet address or a broadcast address, loopback a copy. 317 * XXX To make a simplex device behave exactly like a duplex 318 * device, we should copy in the case of sending to our own 319 * ethernet address (thus letting the original actually appear 320 * on the wire). However, we don't do that here for security 321 * reasons and compatibility with the original behavior. 322 */ 323 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 324 int csum_flags = 0; 325 326 if (m->m_pkthdr.csum_flags & CSUM_IP) 327 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 328 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 329 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 330 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 331 struct mbuf *n; 332 333 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 334 n->m_pkthdr.csum_flags |= csum_flags; 335 if (csum_flags & CSUM_DATA_VALID) 336 n->m_pkthdr.csum_data = 0xffff; 337 if_simloop(ifp, n, dst->sa_family, hlen); 338 } else 339 IFNET_STAT_INC(ifp, iqdrops, 1); 340 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 341 ETHER_ADDR_LEN) == 0) { 342 m->m_pkthdr.csum_flags |= csum_flags; 343 if (csum_flags & CSUM_DATA_VALID) 344 m->m_pkthdr.csum_data = 0xffff; 345 if_simloop(ifp, m, dst->sa_family, hlen); 346 return (0); /* XXX */ 347 } 348 } 349 350 #ifdef CARP 351 if (ifp->if_type == IFT_CARP) { 352 ifp = carp_parent(ifp); 353 if (ifp == NULL) 354 gotoerr(ENETUNREACH); 355 356 ac = IFP2AC(ifp); 357 358 /* 359 * Check precondition again 360 */ 361 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 362 363 if (ifp->if_flags & IFF_MONITOR) 364 gotoerr(ENETDOWN); 365 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 366 (IFF_UP | IFF_RUNNING)) 367 gotoerr(ENETDOWN); 368 } 369 #endif 370 371 /* Handle ng_ether(4) processing, if any */ 372 if (ng_ether_output_p != NULL) { 373 /* 374 * Hold BGL and recheck ng_ether_output_p 375 */ 376 get_mplock(); 377 if (ng_ether_output_p != NULL) { 378 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 379 rel_mplock(); 380 goto bad; 381 } 382 if (m == NULL) { 383 rel_mplock(); 384 return (0); 385 } 386 } 387 rel_mplock(); 388 } 389 390 /* Continue with link-layer output */ 391 return ether_output_frame(ifp, m); 392 393 bad: 394 m_freem(m); 395 return (error); 396 } 397 398 /* 399 * Returns the bridge interface an ifp is associated 400 * with. 401 * 402 * Only call if ifp->if_bridge != NULL. 403 */ 404 struct ifnet * 405 ether_bridge_interface(struct ifnet *ifp) 406 { 407 if (bridge_interface_p) 408 return(bridge_interface_p(ifp->if_bridge)); 409 return (ifp); 410 } 411 412 /* 413 * Ethernet link layer output routine to send a raw frame to the device. 414 * 415 * This assumes that the 14 byte Ethernet header is present and contiguous 416 * in the first mbuf. 417 */ 418 int 419 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 420 { 421 struct ip_fw *rule = NULL; 422 int error = 0; 423 struct altq_pktattr pktattr; 424 425 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 426 427 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 428 struct m_tag *mtag; 429 430 /* Extract info from dummynet tag */ 431 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 432 KKASSERT(mtag != NULL); 433 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 434 KKASSERT(rule != NULL); 435 436 m_tag_delete(m, mtag); 437 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 438 } 439 440 if (ifq_is_enabled(&ifp->if_snd)) 441 altq_etherclassify(&ifp->if_snd, m, &pktattr); 442 crit_enter(); 443 if (IPFW_LOADED && ether_ipfw != 0) { 444 struct ether_header save_eh, *eh; 445 446 eh = mtod(m, struct ether_header *); 447 save_eh = *eh; 448 m_adj(m, ETHER_HDR_LEN); 449 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 450 crit_exit(); 451 if (m != NULL) { 452 m_freem(m); 453 return ENOBUFS; /* pkt dropped */ 454 } else 455 return 0; /* consumed e.g. in a pipe */ 456 } 457 458 /* packet was ok, restore the ethernet header */ 459 ether_restore_header(&m, eh, &save_eh); 460 if (m == NULL) { 461 crit_exit(); 462 return ENOBUFS; 463 } 464 } 465 crit_exit(); 466 467 /* 468 * Queue message on interface, update output statistics if 469 * successful, and start output if interface not yet active. 470 */ 471 error = ifq_dispatch(ifp, m, &pktattr); 472 return (error); 473 } 474 475 /* 476 * ipfw processing for ethernet packets (in and out). 477 * The second parameter is NULL from ether_demux(), and ifp from 478 * ether_output_frame(). 479 */ 480 static boolean_t 481 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 482 const struct ether_header *eh) 483 { 484 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 485 struct ip_fw_args args; 486 struct m_tag *mtag; 487 struct mbuf *m; 488 int i; 489 490 if (*rule != NULL && fw_one_pass) 491 return TRUE; /* dummynet packet, already partially processed */ 492 493 /* 494 * I need some amount of data to be contiguous. 495 */ 496 i = min((*m0)->m_pkthdr.len, max_protohdr); 497 if ((*m0)->m_len < i) { 498 *m0 = m_pullup(*m0, i); 499 if (*m0 == NULL) 500 return FALSE; 501 } 502 503 /* 504 * Clean up tags 505 */ 506 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 507 m_tag_delete(*m0, mtag); 508 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 509 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 510 KKASSERT(mtag != NULL); 511 m_tag_delete(*m0, mtag); 512 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 513 } 514 515 args.m = *m0; /* the packet we are looking at */ 516 args.oif = dst; /* destination, if any */ 517 args.rule = *rule; /* matching rule to restart */ 518 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 519 i = ip_fw_chk_ptr(&args); 520 *m0 = args.m; 521 *rule = args.rule; 522 523 if (*m0 == NULL) 524 return FALSE; 525 526 switch (i) { 527 case IP_FW_PASS: 528 return TRUE; 529 530 case IP_FW_DIVERT: 531 case IP_FW_TEE: 532 case IP_FW_DENY: 533 /* 534 * XXX at some point add support for divert/forward actions. 535 * If none of the above matches, we have to drop the pkt. 536 */ 537 return FALSE; 538 539 case IP_FW_DUMMYNET: 540 /* 541 * Pass the pkt to dummynet, which consumes it. 542 */ 543 m = *m0; /* pass the original to dummynet */ 544 *m0 = NULL; /* and nothing back to the caller */ 545 546 ether_restore_header(&m, eh, &save_eh); 547 if (m == NULL) 548 return FALSE; 549 550 ip_fw_dn_io_ptr(m, args.cookie, 551 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 552 ip_dn_queue(m); 553 return FALSE; 554 555 default: 556 panic("unknown ipfw return value: %d", i); 557 } 558 } 559 560 static void 561 ether_input(struct ifnet *ifp, struct mbuf *m) 562 { 563 ether_input_pkt(ifp, m, NULL); 564 } 565 566 /* 567 * Perform common duties while attaching to interface list 568 */ 569 void 570 ether_ifattach(struct ifnet *ifp, const uint8_t *lla, 571 lwkt_serialize_t serializer) 572 { 573 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 574 serializer); 575 } 576 577 void 578 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla, 579 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer) 580 { 581 struct sockaddr_dl *sdl; 582 char ethstr[ETHER_ADDRSTRLEN + 1]; 583 struct ifaltq *ifq; 584 int i; 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 ifq = &ifp->if_snd; 591 for (i = 0; i < ifq->altq_subq_cnt; ++i) { 592 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i); 593 594 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen * 595 (ETHER_MAX_LEN - ETHER_CRC_LEN); 596 } 597 ifp->if_mtu = ETHERMTU; 598 if (ifp->if_tsolen <= 0) { 599 if ((ether_tsolen_default / ETHERMTU) < 2) { 600 kprintf("ether TSO maxlen %d -> %d\n", 601 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 602 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 603 } 604 ifp->if_tsolen = ether_tsolen_default; 605 } 606 if (ifp->if_baudrate == 0) 607 ifp->if_baudrate = 10000000; 608 ifp->if_output = ether_output; 609 ifp->if_input = ether_input; 610 ifp->if_resolvemulti = ether_resolvemulti; 611 ifp->if_broadcastaddr = etherbroadcastaddr; 612 sdl = IF_LLSOCKADDR(ifp); 613 sdl->sdl_type = IFT_ETHER; 614 sdl->sdl_alen = ifp->if_addrlen; 615 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 616 /* 617 * XXX Keep the current drivers happy. 618 * XXX Remove once all drivers have been cleaned up 619 */ 620 if (lla != IFP2AC(ifp)->ac_enaddr) 621 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 622 bpfattach(ifp, dlt, hdrlen); 623 if (ng_ether_attach_p != NULL) 624 (*ng_ether_attach_p)(ifp); 625 626 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 627 } 628 629 /* 630 * Perform common duties while detaching an Ethernet interface 631 */ 632 void 633 ether_ifdetach(struct ifnet *ifp) 634 { 635 if_down(ifp); 636 637 if (ng_ether_detach_p != NULL) 638 (*ng_ether_detach_p)(ifp); 639 bpfdetach(ifp); 640 if_detach(ifp); 641 } 642 643 int 644 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 645 { 646 struct ifaddr *ifa = (struct ifaddr *) data; 647 struct ifreq *ifr = (struct ifreq *) data; 648 int error = 0; 649 650 #define IF_INIT(ifp) \ 651 do { \ 652 if (((ifp)->if_flags & IFF_UP) == 0) { \ 653 (ifp)->if_flags |= IFF_UP; \ 654 (ifp)->if_init((ifp)->if_softc); \ 655 } \ 656 } while (0) 657 658 ASSERT_IFNET_SERIALIZED_ALL(ifp); 659 660 switch (command) { 661 case SIOCSIFADDR: 662 switch (ifa->ifa_addr->sa_family) { 663 #ifdef INET 664 case AF_INET: 665 IF_INIT(ifp); /* before arpwhohas */ 666 arp_ifinit(ifp, ifa); 667 break; 668 #endif 669 #ifdef IPX 670 /* 671 * XXX - This code is probably wrong 672 */ 673 case AF_IPX: 674 { 675 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 676 struct arpcom *ac = IFP2AC(ifp); 677 678 if (ipx_nullhost(*ina)) 679 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 680 else 681 bcopy(ina->x_host.c_host, ac->ac_enaddr, 682 sizeof ac->ac_enaddr); 683 684 IF_INIT(ifp); /* Set new address. */ 685 break; 686 } 687 #endif 688 default: 689 IF_INIT(ifp); 690 break; 691 } 692 break; 693 694 case SIOCGIFADDR: 695 bcopy(IFP2AC(ifp)->ac_enaddr, 696 ((struct sockaddr *)ifr->ifr_data)->sa_data, 697 ETHER_ADDR_LEN); 698 break; 699 700 case SIOCSIFMTU: 701 /* 702 * Set the interface MTU. 703 */ 704 if (ifr->ifr_mtu > ETHERMTU) { 705 error = EINVAL; 706 } else { 707 ifp->if_mtu = ifr->ifr_mtu; 708 } 709 break; 710 default: 711 error = EINVAL; 712 break; 713 } 714 return (error); 715 716 #undef IF_INIT 717 } 718 719 int 720 ether_resolvemulti( 721 struct ifnet *ifp, 722 struct sockaddr **llsa, 723 struct sockaddr *sa) 724 { 725 struct sockaddr_dl *sdl; 726 #ifdef INET 727 struct sockaddr_in *sin; 728 #endif 729 #ifdef INET6 730 struct sockaddr_in6 *sin6; 731 #endif 732 u_char *e_addr; 733 734 switch(sa->sa_family) { 735 case AF_LINK: 736 /* 737 * No mapping needed. Just check that it's a valid MC address. 738 */ 739 sdl = (struct sockaddr_dl *)sa; 740 e_addr = LLADDR(sdl); 741 if ((e_addr[0] & 1) != 1) 742 return EADDRNOTAVAIL; 743 *llsa = NULL; 744 return 0; 745 746 #ifdef INET 747 case AF_INET: 748 sin = (struct sockaddr_in *)sa; 749 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 750 return EADDRNOTAVAIL; 751 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 752 sdl->sdl_len = sizeof *sdl; 753 sdl->sdl_family = AF_LINK; 754 sdl->sdl_index = ifp->if_index; 755 sdl->sdl_type = IFT_ETHER; 756 sdl->sdl_alen = ETHER_ADDR_LEN; 757 e_addr = LLADDR(sdl); 758 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 759 *llsa = (struct sockaddr *)sdl; 760 return 0; 761 #endif 762 #ifdef INET6 763 case AF_INET6: 764 sin6 = (struct sockaddr_in6 *)sa; 765 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 766 /* 767 * An IP6 address of 0 means listen to all 768 * of the Ethernet multicast address used for IP6. 769 * (This is used for multicast routers.) 770 */ 771 ifp->if_flags |= IFF_ALLMULTI; 772 *llsa = NULL; 773 return 0; 774 } 775 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 776 return EADDRNOTAVAIL; 777 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 778 sdl->sdl_len = sizeof *sdl; 779 sdl->sdl_family = AF_LINK; 780 sdl->sdl_index = ifp->if_index; 781 sdl->sdl_type = IFT_ETHER; 782 sdl->sdl_alen = ETHER_ADDR_LEN; 783 e_addr = LLADDR(sdl); 784 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 785 *llsa = (struct sockaddr *)sdl; 786 return 0; 787 #endif 788 789 default: 790 /* 791 * Well, the text isn't quite right, but it's the name 792 * that counts... 793 */ 794 return EAFNOSUPPORT; 795 } 796 } 797 798 #if 0 799 /* 800 * This is for reference. We have a table-driven version 801 * of the little-endian crc32 generator, which is faster 802 * than the double-loop. 803 */ 804 uint32_t 805 ether_crc32_le(const uint8_t *buf, size_t len) 806 { 807 uint32_t c, crc, carry; 808 size_t i, j; 809 810 crc = 0xffffffffU; /* initial value */ 811 812 for (i = 0; i < len; i++) { 813 c = buf[i]; 814 for (j = 0; j < 8; j++) { 815 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 816 crc >>= 1; 817 c >>= 1; 818 if (carry) 819 crc = (crc ^ ETHER_CRC_POLY_LE); 820 } 821 } 822 823 return (crc); 824 } 825 #else 826 uint32_t 827 ether_crc32_le(const uint8_t *buf, size_t len) 828 { 829 static const uint32_t crctab[] = { 830 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 831 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 832 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 833 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 834 }; 835 uint32_t crc; 836 size_t i; 837 838 crc = 0xffffffffU; /* initial value */ 839 840 for (i = 0; i < len; i++) { 841 crc ^= buf[i]; 842 crc = (crc >> 4) ^ crctab[crc & 0xf]; 843 crc = (crc >> 4) ^ crctab[crc & 0xf]; 844 } 845 846 return (crc); 847 } 848 #endif 849 850 uint32_t 851 ether_crc32_be(const uint8_t *buf, size_t len) 852 { 853 uint32_t c, crc, carry; 854 size_t i, j; 855 856 crc = 0xffffffffU; /* initial value */ 857 858 for (i = 0; i < len; i++) { 859 c = buf[i]; 860 for (j = 0; j < 8; j++) { 861 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 862 crc <<= 1; 863 c >>= 1; 864 if (carry) 865 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 866 } 867 } 868 869 return (crc); 870 } 871 872 /* 873 * find the size of ethernet header, and call classifier 874 */ 875 void 876 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 877 struct altq_pktattr *pktattr) 878 { 879 struct ether_header *eh; 880 uint16_t ether_type; 881 int hlen, af, hdrsize; 882 883 hlen = sizeof(struct ether_header); 884 eh = mtod(m, struct ether_header *); 885 886 ether_type = ntohs(eh->ether_type); 887 if (ether_type < ETHERMTU) { 888 /* ick! LLC/SNAP */ 889 struct llc *llc = (struct llc *)(eh + 1); 890 hlen += 8; 891 892 if (m->m_len < hlen || 893 llc->llc_dsap != LLC_SNAP_LSAP || 894 llc->llc_ssap != LLC_SNAP_LSAP || 895 llc->llc_control != LLC_UI) 896 goto bad; /* not snap! */ 897 898 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 899 } 900 901 if (ether_type == ETHERTYPE_IP) { 902 af = AF_INET; 903 hdrsize = 20; /* sizeof(struct ip) */ 904 #ifdef INET6 905 } else if (ether_type == ETHERTYPE_IPV6) { 906 af = AF_INET6; 907 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 908 #endif 909 } else 910 goto bad; 911 912 while (m->m_len <= hlen) { 913 hlen -= m->m_len; 914 m = m->m_next; 915 } 916 if (m->m_len < hlen + hdrsize) { 917 /* 918 * ip header is not in a single mbuf. this should not 919 * happen in the current code. 920 * (todo: use m_pulldown in the future) 921 */ 922 goto bad; 923 } 924 m->m_data += hlen; 925 m->m_len -= hlen; 926 ifq_classify(ifq, m, af, pktattr); 927 m->m_data -= hlen; 928 m->m_len += hlen; 929 930 return; 931 932 bad: 933 pktattr->pattr_class = NULL; 934 pktattr->pattr_hdr = NULL; 935 pktattr->pattr_af = AF_UNSPEC; 936 } 937 938 static void 939 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 940 const struct ether_header *save_eh) 941 { 942 struct mbuf *m = *m0; 943 944 ether_restore_hdr++; 945 946 /* 947 * Prepend the header, optimize for the common case of 948 * eh pointing into the mbuf. 949 */ 950 if ((const void *)(eh + 1) == (void *)m->m_data) { 951 m->m_data -= ETHER_HDR_LEN; 952 m->m_len += ETHER_HDR_LEN; 953 m->m_pkthdr.len += ETHER_HDR_LEN; 954 } else { 955 ether_prepend_hdr++; 956 957 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 958 if (m != NULL) { 959 bcopy(save_eh, mtod(m, struct ether_header *), 960 ETHER_HDR_LEN); 961 } 962 } 963 *m0 = m; 964 } 965 966 /* 967 * Upper layer processing for a received Ethernet packet. 968 */ 969 void 970 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 971 { 972 struct ether_header *eh; 973 int isr, discard = 0; 974 u_short ether_type; 975 struct ip_fw *rule = NULL; 976 977 M_ASSERTPKTHDR(m); 978 KASSERT(m->m_len >= ETHER_HDR_LEN, 979 ("ether header is not contiguous!")); 980 981 eh = mtod(m, struct ether_header *); 982 983 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 984 struct m_tag *mtag; 985 986 /* Extract info from dummynet tag */ 987 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 988 KKASSERT(mtag != NULL); 989 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 990 KKASSERT(rule != NULL); 991 992 m_tag_delete(m, mtag); 993 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 994 995 /* packet is passing the second time */ 996 goto post_stats; 997 } 998 999 /* 1000 * We got a packet which was unicast to a different Ethernet 1001 * address. If the driver is working properly, then this 1002 * situation can only happen when the interface is in 1003 * promiscuous mode. We defer the packet discarding until the 1004 * vlan processing is done, so that vlan/bridge or vlan/netgraph 1005 * could work. 1006 */ 1007 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 1008 !ETHER_IS_MULTICAST(eh->ether_dhost) && 1009 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 1010 if (ether_debug & 1) { 1011 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 1012 "%02x:%02x:%02x:%02x:%02x:%02x " 1013 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 1014 eh->ether_dhost[0], 1015 eh->ether_dhost[1], 1016 eh->ether_dhost[2], 1017 eh->ether_dhost[3], 1018 eh->ether_dhost[4], 1019 eh->ether_dhost[5], 1020 eh->ether_shost[0], 1021 eh->ether_shost[1], 1022 eh->ether_shost[2], 1023 eh->ether_shost[3], 1024 eh->ether_shost[4], 1025 eh->ether_shost[5], 1026 eh->ether_type, 1027 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1028 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1029 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1030 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1031 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1032 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1033 ); 1034 } 1035 if ((ether_debug & 2) == 0) 1036 discard = 1; 1037 } 1038 1039 post_stats: 1040 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1041 struct ether_header save_eh = *eh; 1042 1043 /* XXX old crufty stuff, needs to be removed */ 1044 m_adj(m, sizeof(struct ether_header)); 1045 1046 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1047 m_freem(m); 1048 return; 1049 } 1050 1051 ether_restore_header(&m, eh, &save_eh); 1052 if (m == NULL) 1053 return; 1054 eh = mtod(m, struct ether_header *); 1055 } 1056 1057 ether_type = ntohs(eh->ether_type); 1058 KKASSERT(ether_type != ETHERTYPE_VLAN); 1059 1060 if (m->m_flags & M_VLANTAG) { 1061 void (*vlan_input_func)(struct mbuf *); 1062 1063 vlan_input_func = vlan_input_p; 1064 if (vlan_input_func != NULL) { 1065 vlan_input_func(m); 1066 } else { 1067 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1068 m_freem(m); 1069 } 1070 return; 1071 } 1072 1073 /* 1074 * If we have been asked to discard this packet 1075 * (e.g. not for us), drop it before entering 1076 * the upper layer. 1077 */ 1078 if (discard) { 1079 m_freem(m); 1080 return; 1081 } 1082 1083 /* 1084 * Clear protocol specific flags, 1085 * before entering the upper layer. 1086 */ 1087 m->m_flags &= ~M_ETHER_FLAGS; 1088 1089 /* Strip ethernet header. */ 1090 m_adj(m, sizeof(struct ether_header)); 1091 1092 switch (ether_type) { 1093 #ifdef INET 1094 case ETHERTYPE_IP: 1095 if ((m->m_flags & M_LENCHECKED) == 0) { 1096 if (!ip_lengthcheck(&m, 0)) 1097 return; 1098 } 1099 if (ipflow_fastforward(m)) 1100 return; 1101 isr = NETISR_IP; 1102 break; 1103 1104 case ETHERTYPE_ARP: 1105 if (ifp->if_flags & IFF_NOARP) { 1106 /* Discard packet if ARP is disabled on interface */ 1107 m_freem(m); 1108 return; 1109 } 1110 isr = NETISR_ARP; 1111 break; 1112 #endif 1113 1114 #ifdef INET6 1115 case ETHERTYPE_IPV6: 1116 isr = NETISR_IPV6; 1117 break; 1118 #endif 1119 1120 #ifdef IPX 1121 case ETHERTYPE_IPX: 1122 if (ef_inputp) { 1123 /* 1124 * Hold BGL and recheck ef_inputp 1125 */ 1126 get_mplock(); 1127 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1128 rel_mplock(); 1129 return; 1130 } 1131 rel_mplock(); 1132 } 1133 isr = NETISR_IPX; 1134 break; 1135 #endif 1136 1137 #ifdef MPLS 1138 case ETHERTYPE_MPLS: 1139 case ETHERTYPE_MPLS_MCAST: 1140 /* Should have been set by ether_input_pkt(). */ 1141 KKASSERT(m->m_flags & M_MPLSLABELED); 1142 isr = NETISR_MPLS; 1143 break; 1144 #endif 1145 1146 default: 1147 /* 1148 * The accurate msgport is not determined before 1149 * we reach here, so recharacterize packet. 1150 */ 1151 m->m_flags &= ~M_HASH; 1152 #ifdef IPX 1153 if (ef_inputp) { 1154 /* 1155 * Hold BGL and recheck ef_inputp 1156 */ 1157 get_mplock(); 1158 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1159 rel_mplock(); 1160 return; 1161 } 1162 rel_mplock(); 1163 } 1164 #endif 1165 if (ng_ether_input_orphan_p != NULL) { 1166 /* 1167 * Put back the ethernet header so netgraph has a 1168 * consistent view of inbound packets. 1169 */ 1170 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1171 if (m == NULL) { 1172 /* 1173 * M_PREPEND frees the mbuf in case of failure. 1174 */ 1175 return; 1176 } 1177 /* 1178 * Hold BGL and recheck ng_ether_input_orphan_p 1179 */ 1180 get_mplock(); 1181 if (ng_ether_input_orphan_p != NULL) { 1182 ng_ether_input_orphan_p(ifp, m); 1183 rel_mplock(); 1184 return; 1185 } 1186 rel_mplock(); 1187 } 1188 m_freem(m); 1189 return; 1190 } 1191 1192 if (m->m_flags & M_HASH) { 1193 if (&curthread->td_msgport == 1194 netisr_hashport(m->m_pkthdr.hash)) { 1195 netisr_handle(isr, m); 1196 return; 1197 } else { 1198 /* 1199 * XXX Something is wrong, 1200 * we probably should panic here! 1201 */ 1202 m->m_flags &= ~M_HASH; 1203 atomic_add_long(ðer_input_wronghash, 1); 1204 } 1205 } 1206 #ifdef RSS_DEBUG 1207 atomic_add_long(ðer_input_requeue, 1); 1208 #endif 1209 netisr_queue(isr, m); 1210 } 1211 1212 /* 1213 * First we perform any link layer operations, then continue to the 1214 * upper layers with ether_demux_oncpu(). 1215 */ 1216 static void 1217 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1218 { 1219 #ifdef CARP 1220 void *carp; 1221 #endif 1222 1223 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1224 /* 1225 * Receiving interface's flags are changed, when this 1226 * packet is waiting for processing; discard it. 1227 */ 1228 m_freem(m); 1229 return; 1230 } 1231 1232 /* 1233 * Tap the packet off here for a bridge. bridge_input() 1234 * will return NULL if it has consumed the packet, otherwise 1235 * it gets processed as normal. Note that bridge_input() 1236 * will always return the original packet if we need to 1237 * process it locally. 1238 */ 1239 if (ifp->if_bridge) { 1240 KASSERT(bridge_input_p != NULL, 1241 ("%s: if_bridge not loaded!", __func__)); 1242 1243 if(m->m_flags & M_ETHER_BRIDGED) { 1244 m->m_flags &= ~M_ETHER_BRIDGED; 1245 } else { 1246 m = bridge_input_p(ifp, m); 1247 if (m == NULL) 1248 return; 1249 1250 KASSERT(ifp == m->m_pkthdr.rcvif, 1251 ("bridge_input_p changed rcvif")); 1252 } 1253 } 1254 1255 #ifdef CARP 1256 carp = ifp->if_carp; 1257 if (carp) { 1258 m = carp_input(carp, m); 1259 if (m == NULL) 1260 return; 1261 KASSERT(ifp == m->m_pkthdr.rcvif, 1262 ("carp_input changed rcvif")); 1263 } 1264 #endif 1265 1266 /* Handle ng_ether(4) processing, if any */ 1267 if (ng_ether_input_p != NULL) { 1268 /* 1269 * Hold BGL and recheck ng_ether_input_p 1270 */ 1271 get_mplock(); 1272 if (ng_ether_input_p != NULL) 1273 ng_ether_input_p(ifp, &m); 1274 rel_mplock(); 1275 1276 if (m == NULL) 1277 return; 1278 } 1279 1280 /* Continue with upper layer processing */ 1281 ether_demux_oncpu(ifp, m); 1282 } 1283 1284 /* 1285 * Perform certain functions of ether_input_pkt(): 1286 * - Test IFF_UP 1287 * - Update statistics 1288 * - Run bpf(4) tap if requested 1289 * Then pass the packet to ether_input_oncpu(). 1290 * 1291 * This function should be used by pseudo interface (e.g. vlan(4)), 1292 * when it tries to claim that the packet is received by it. 1293 * 1294 * REINPUT_KEEPRCVIF 1295 * REINPUT_RUNBPF 1296 */ 1297 void 1298 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1299 { 1300 /* Discard packet if interface is not up */ 1301 if (!(ifp->if_flags & IFF_UP)) { 1302 m_freem(m); 1303 return; 1304 } 1305 1306 /* 1307 * Change receiving interface. The bridge will often pass a flag to 1308 * ask that this not be done so ARPs get applied to the correct 1309 * side. 1310 */ 1311 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1312 m->m_pkthdr.rcvif == NULL) { 1313 m->m_pkthdr.rcvif = ifp; 1314 } 1315 1316 /* Update statistics */ 1317 IFNET_STAT_INC(ifp, ipackets, 1); 1318 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1319 if (m->m_flags & (M_MCAST | M_BCAST)) 1320 IFNET_STAT_INC(ifp, imcasts, 1); 1321 1322 if (reinput_flags & REINPUT_RUNBPF) 1323 BPF_MTAP(ifp, m); 1324 1325 ether_input_oncpu(ifp, m); 1326 } 1327 1328 static __inline boolean_t 1329 ether_vlancheck(struct mbuf **m0) 1330 { 1331 struct mbuf *m = *m0; 1332 struct ether_header *eh; 1333 uint16_t ether_type; 1334 1335 eh = mtod(m, struct ether_header *); 1336 ether_type = ntohs(eh->ether_type); 1337 1338 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1339 /* 1340 * Extract vlan tag if hardware does not do it for us 1341 */ 1342 vlan_ether_decap(&m); 1343 if (m == NULL) 1344 goto failed; 1345 1346 eh = mtod(m, struct ether_header *); 1347 ether_type = ntohs(eh->ether_type); 1348 } 1349 1350 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1351 /* 1352 * To prevent possible dangerous recursion, 1353 * we don't do vlan-in-vlan 1354 */ 1355 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1356 goto failed; 1357 } 1358 KKASSERT(ether_type != ETHERTYPE_VLAN); 1359 1360 m->m_flags |= M_ETHER_VLANCHECKED; 1361 *m0 = m; 1362 return TRUE; 1363 failed: 1364 if (m != NULL) 1365 m_freem(m); 1366 *m0 = NULL; 1367 return FALSE; 1368 } 1369 1370 static void 1371 ether_input_handler(netmsg_t nmsg) 1372 { 1373 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1374 struct ether_header *eh; 1375 struct ifnet *ifp; 1376 struct mbuf *m; 1377 1378 m = nmp->nm_packet; 1379 M_ASSERTPKTHDR(m); 1380 1381 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1382 if (!ether_vlancheck(&m)) { 1383 KKASSERT(m == NULL); 1384 return; 1385 } 1386 } 1387 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) || 1388 __predict_false(ether_input_ckhash)) { 1389 int isr; 1390 1391 /* 1392 * Need to verify the hash supplied by the hardware 1393 * which could be wrong. 1394 */ 1395 m->m_flags &= ~(M_HASH | M_CKHASH); 1396 isr = ether_characterize(&m); 1397 if (m == NULL) 1398 return; 1399 KKASSERT(m->m_flags & M_HASH); 1400 1401 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) { 1402 /* 1403 * Wrong hardware supplied hash; redispatch 1404 */ 1405 ether_dispatch(isr, m); 1406 if (__predict_false(ether_input_ckhash)) 1407 atomic_add_long(ðer_input_wronghwhash, 1); 1408 return; 1409 } 1410 } 1411 ifp = m->m_pkthdr.rcvif; 1412 1413 eh = mtod(m, struct ether_header *); 1414 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1415 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1416 ifp->if_addrlen) == 0) 1417 m->m_flags |= M_BCAST; 1418 else 1419 m->m_flags |= M_MCAST; 1420 IFNET_STAT_INC(ifp, imcasts, 1); 1421 } 1422 1423 ether_input_oncpu(ifp, m); 1424 } 1425 1426 /* 1427 * Send the packet to the target netisr msgport 1428 * 1429 * At this point the packet must be characterized (M_HASH set), 1430 * so we know which netisr to send it to. 1431 */ 1432 static void 1433 ether_dispatch(int isr, struct mbuf *m) 1434 { 1435 struct netmsg_packet *pmsg; 1436 1437 KKASSERT(m->m_flags & M_HASH); 1438 pmsg = &m->m_hdr.mh_netmsg; 1439 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1440 0, ether_input_handler); 1441 pmsg->nm_packet = m; 1442 pmsg->base.lmsg.u.ms_result = isr; 1443 1444 logether(disp_beg, NULL); 1445 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1446 logether(disp_end, NULL); 1447 } 1448 1449 /* 1450 * Process a received Ethernet packet. 1451 * 1452 * The ethernet header is assumed to be in the mbuf so the caller 1453 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1454 * bytes in the first mbuf. 1455 */ 1456 void 1457 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi) 1458 { 1459 int isr; 1460 1461 M_ASSERTPKTHDR(m); 1462 1463 /* Discard packet if interface is not up */ 1464 if (!(ifp->if_flags & IFF_UP)) { 1465 m_freem(m); 1466 return; 1467 } 1468 1469 if (m->m_len < sizeof(struct ether_header)) { 1470 /* XXX error in the caller. */ 1471 m_freem(m); 1472 return; 1473 } 1474 1475 m->m_pkthdr.rcvif = ifp; 1476 1477 logether(pkt_beg, ifp); 1478 1479 ETHER_BPF_MTAP(ifp, m); 1480 1481 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1482 1483 if (ifp->if_flags & IFF_MONITOR) { 1484 struct ether_header *eh; 1485 1486 eh = mtod(m, struct ether_header *); 1487 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1488 IFNET_STAT_INC(ifp, imcasts, 1); 1489 1490 /* 1491 * Interface marked for monitoring; discard packet. 1492 */ 1493 m_freem(m); 1494 1495 logether(pkt_end, ifp); 1496 return; 1497 } 1498 1499 /* 1500 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1501 * we can dispatch it immediately with trivial checks. 1502 */ 1503 if (pi != NULL && (m->m_flags & M_HASH)) { 1504 #ifdef RSS_DEBUG 1505 atomic_add_long(ðer_pktinfo_try, 1); 1506 #endif 1507 netisr_hashcheck(pi->pi_netisr, m, pi); 1508 if (m->m_flags & M_HASH) { 1509 ether_dispatch(pi->pi_netisr, m); 1510 #ifdef RSS_DEBUG 1511 atomic_add_long(ðer_pktinfo_hit, 1); 1512 #endif 1513 logether(pkt_end, ifp); 1514 return; 1515 } 1516 } 1517 #ifdef RSS_DEBUG 1518 else if (ifp->if_capenable & IFCAP_RSS) { 1519 if (pi == NULL) 1520 atomic_add_long(ðer_rss_nopi, 1); 1521 else 1522 atomic_add_long(ðer_rss_nohash, 1); 1523 } 1524 #endif 1525 1526 /* 1527 * Packet hash will be recalculated by software, so clear 1528 * the M_HASH and M_CKHASH flag set by the driver; the hash 1529 * value calculated by the hardware may not be exactly what 1530 * we want. 1531 */ 1532 m->m_flags &= ~(M_HASH | M_CKHASH); 1533 1534 if (!ether_vlancheck(&m)) { 1535 KKASSERT(m == NULL); 1536 logether(pkt_end, ifp); 1537 return; 1538 } 1539 1540 isr = ether_characterize(&m); 1541 if (m == NULL) { 1542 logether(pkt_end, ifp); 1543 return; 1544 } 1545 1546 /* 1547 * Finally dispatch it 1548 */ 1549 ether_dispatch(isr, m); 1550 1551 logether(pkt_end, ifp); 1552 } 1553 1554 static int 1555 ether_characterize(struct mbuf **m0) 1556 { 1557 struct mbuf *m = *m0; 1558 struct ether_header *eh; 1559 uint16_t ether_type; 1560 int isr; 1561 1562 eh = mtod(m, struct ether_header *); 1563 ether_type = ntohs(eh->ether_type); 1564 1565 /* 1566 * Map ether type to netisr id. 1567 */ 1568 switch (ether_type) { 1569 #ifdef INET 1570 case ETHERTYPE_IP: 1571 isr = NETISR_IP; 1572 break; 1573 1574 case ETHERTYPE_ARP: 1575 isr = NETISR_ARP; 1576 break; 1577 #endif 1578 1579 #ifdef INET6 1580 case ETHERTYPE_IPV6: 1581 isr = NETISR_IPV6; 1582 break; 1583 #endif 1584 1585 #ifdef IPX 1586 case ETHERTYPE_IPX: 1587 isr = NETISR_IPX; 1588 break; 1589 #endif 1590 1591 #ifdef MPLS 1592 case ETHERTYPE_MPLS: 1593 case ETHERTYPE_MPLS_MCAST: 1594 m->m_flags |= M_MPLSLABELED; 1595 isr = NETISR_MPLS; 1596 break; 1597 #endif 1598 1599 default: 1600 /* 1601 * NETISR_MAX is an invalid value; it is chosen to let 1602 * netisr_characterize() know that we have no clear 1603 * idea where this packet should go. 1604 */ 1605 isr = NETISR_MAX; 1606 break; 1607 } 1608 1609 /* 1610 * Ask the isr to characterize the packet since we couldn't. 1611 * This is an attempt to optimally get us onto the correct protocol 1612 * thread. 1613 */ 1614 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1615 1616 *m0 = m; 1617 return isr; 1618 } 1619 1620 static void 1621 ether_demux_handler(netmsg_t nmsg) 1622 { 1623 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1624 struct ifnet *ifp; 1625 struct mbuf *m; 1626 1627 m = nmp->nm_packet; 1628 M_ASSERTPKTHDR(m); 1629 ifp = m->m_pkthdr.rcvif; 1630 1631 ether_demux_oncpu(ifp, m); 1632 } 1633 1634 void 1635 ether_demux(struct mbuf *m) 1636 { 1637 struct netmsg_packet *pmsg; 1638 int isr; 1639 1640 isr = ether_characterize(&m); 1641 if (m == NULL) 1642 return; 1643 1644 KKASSERT(m->m_flags & M_HASH); 1645 pmsg = &m->m_hdr.mh_netmsg; 1646 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1647 0, ether_demux_handler); 1648 pmsg->nm_packet = m; 1649 pmsg->base.lmsg.u.ms_result = isr; 1650 1651 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1652 } 1653 1654 u_char * 1655 kether_aton(const char *macstr, u_char *addr) 1656 { 1657 unsigned int o0, o1, o2, o3, o4, o5; 1658 int n; 1659 1660 if (macstr == NULL || addr == NULL) 1661 return NULL; 1662 1663 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1664 &o3, &o4, &o5); 1665 if (n != 6) 1666 return NULL; 1667 1668 addr[0] = o0; 1669 addr[1] = o1; 1670 addr[2] = o2; 1671 addr[3] = o3; 1672 addr[4] = o4; 1673 addr[5] = o5; 1674 1675 return addr; 1676 } 1677 1678 char * 1679 kether_ntoa(const u_char *addr, char *buf) 1680 { 1681 int len = ETHER_ADDRSTRLEN + 1; 1682 int n; 1683 1684 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1685 addr[1], addr[2], addr[3], addr[4], addr[5]); 1686 1687 if (n < 17) 1688 return NULL; 1689 1690 return buf; 1691 } 1692 1693 MODULE_VERSION(ether, 1); 1694