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/ipfw3/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 MPLS 90 #include <netproto/mpls/mpls.h> 91 #endif 92 93 /* netgraph node hooks for ng_ether(4) */ 94 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 95 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m); 96 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 97 void (*ng_ether_attach_p)(struct ifnet *ifp); 98 void (*ng_ether_detach_p)(struct ifnet *ifp); 99 100 void (*vlan_input_p)(struct mbuf *); 101 102 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 103 struct rtentry *); 104 static void ether_restore_header(struct mbuf **, const struct ether_header *, 105 const struct ether_header *); 106 static int ether_characterize(struct mbuf **); 107 static void ether_dispatch(int, struct mbuf *, int); 108 109 /* 110 * if_bridge support 111 */ 112 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 113 int (*bridge_output_p)(struct ifnet *, struct mbuf *); 114 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 115 struct ifnet *(*bridge_interface_p)(void *if_bridge); 116 117 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 118 struct sockaddr *); 119 120 /* 121 * if_lagg(4) support 122 */ 123 void (*lagg_input_p)(struct ifnet *, struct mbuf *); 124 int (*lagg_output_p)(struct ifnet *, struct mbuf *); 125 126 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 127 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 128 }; 129 130 #define gotoerr(e) do { error = (e); goto bad; } while (0) 131 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 132 133 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 134 struct ip_fw **rule, 135 const struct ether_header *eh); 136 137 static int ether_ipfw; 138 static u_long ether_restore_hdr; 139 static u_long ether_prepend_hdr; 140 static u_long ether_input_wronghash; 141 static int ether_debug; 142 143 #ifdef RSS_DEBUG 144 static u_long ether_pktinfo_try; 145 static u_long ether_pktinfo_hit; 146 static u_long ether_rss_nopi; 147 static u_long ether_rss_nohash; 148 static u_long ether_input_requeue; 149 #endif 150 static u_long ether_input_wronghwhash; 151 static int ether_input_ckhash; 152 153 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU) 154 155 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 156 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default); 157 158 SYSCTL_DECL(_net_link); 159 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 160 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 161 ðer_debug, 0, "Ether debug"); 162 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 163 ðer_ipfw, 0, "Pass ether pkts through firewall"); 164 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 165 ðer_restore_hdr, 0, "# of ether header restoration"); 166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 167 ðer_prepend_hdr, 0, 168 "# of ether header restoration which prepends mbuf"); 169 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 170 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 171 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW, 172 ðer_tsolen_default, 0, "Default max TSO length"); 173 174 #ifdef RSS_DEBUG 175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 176 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 177 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 178 ðer_rss_nohash, 0, "# of packets do not have hash"); 179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 180 ðer_pktinfo_try, 0, 181 "# of tries to find packets' msgport using pktinfo"); 182 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 183 ðer_pktinfo_hit, 0, 184 "# of packets whose msgport are found using pktinfo"); 185 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 186 ðer_input_requeue, 0, "# of input packets gets requeued"); 187 #endif 188 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 189 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 190 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 191 ðer_input_ckhash, 0, "always check hash"); 192 193 #define ETHER_KTR_STR "ifp=%p" 194 #define ETHER_KTR_ARGS struct ifnet *ifp 195 #ifndef KTR_ETHERNET 196 #define KTR_ETHERNET KTR_ALL 197 #endif 198 KTR_INFO_MASTER(ether); 199 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 200 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 201 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 202 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 203 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 204 205 /* 206 * Ethernet output routine. 207 * Encapsulate a packet of type family for the local net. 208 * Use trailer local net encapsulation if enough data in first 209 * packet leaves a multiple of 512 bytes of data in remainder. 210 * Assumes that ifp is actually pointer to arpcom structure. 211 */ 212 static int 213 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 214 struct rtentry *rt) 215 { 216 struct ether_header *eh, *deh; 217 u_char *edst; 218 int loop_copy = 0; 219 int hlen = ETHER_HDR_LEN; /* link layer header length */ 220 struct arpcom *ac = IFP2AC(ifp); 221 int error; 222 223 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 224 225 if (ifp->if_flags & IFF_MONITOR) 226 gotoerr(ENETDOWN); 227 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 228 gotoerr(ENETDOWN); 229 230 M_PREPEND(m, sizeof(struct ether_header), M_NOWAIT); 231 if (m == NULL) 232 return (ENOBUFS); 233 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 234 eh = mtod(m, struct ether_header *); 235 edst = eh->ether_dhost; 236 237 /* 238 * Fill in the destination ethernet address and frame type. 239 */ 240 switch (dst->sa_family) { 241 #ifdef INET 242 case AF_INET: 243 if (!arpresolve(ifp, rt, m, dst, edst)) 244 return (0); /* if not yet resolved */ 245 #ifdef MPLS 246 if (m->m_flags & M_MPLSLABELED) 247 eh->ether_type = htons(ETHERTYPE_MPLS); 248 else 249 #endif 250 eh->ether_type = htons(ETHERTYPE_IP); 251 break; 252 #endif 253 #ifdef INET6 254 case AF_INET6: 255 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 256 return (0); /* Something bad happenned. */ 257 eh->ether_type = htons(ETHERTYPE_IPV6); 258 break; 259 #endif 260 case pseudo_AF_HDRCMPLT: 261 case AF_UNSPEC: 262 loop_copy = -1; /* if this is for us, don't do it */ 263 deh = (struct ether_header *)dst->sa_data; 264 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 265 eh->ether_type = deh->ether_type; 266 break; 267 268 default: 269 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 270 gotoerr(EAFNOSUPPORT); 271 } 272 273 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 274 memcpy(eh->ether_shost, 275 ((struct ether_header *)dst->sa_data)->ether_shost, 276 ETHER_ADDR_LEN); 277 else 278 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 279 280 /* 281 * Bridges require special output handling. 282 */ 283 if (ifp->if_bridge) { 284 KASSERT(bridge_output_p != NULL, 285 ("%s: if_bridge not loaded!", __func__)); 286 return bridge_output_p(ifp, m); 287 } 288 #if 0 /* XXX */ 289 if (ifp->if_lagg) { 290 KASSERT(lagg_output_p != NULL, 291 ("%s: if_lagg not loaded!", __func__)); 292 return lagg_output_p(ifp, m); 293 } 294 #endif 295 296 /* 297 * If a simplex interface, and the packet is being sent to our 298 * Ethernet address or a broadcast address, loopback a copy. 299 * XXX To make a simplex device behave exactly like a duplex 300 * device, we should copy in the case of sending to our own 301 * ethernet address (thus letting the original actually appear 302 * on the wire). However, we don't do that here for security 303 * reasons and compatibility with the original behavior. 304 */ 305 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 306 int csum_flags = 0; 307 308 if (m->m_pkthdr.csum_flags & CSUM_IP) 309 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 310 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 311 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 312 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 313 struct mbuf *n; 314 315 if ((n = m_copypacket(m, M_NOWAIT)) != NULL) { 316 n->m_pkthdr.csum_flags |= csum_flags; 317 if (csum_flags & CSUM_DATA_VALID) 318 n->m_pkthdr.csum_data = 0xffff; 319 if_simloop(ifp, n, dst->sa_family, hlen); 320 } else 321 IFNET_STAT_INC(ifp, iqdrops, 1); 322 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 323 ETHER_ADDR_LEN) == 0) { 324 m->m_pkthdr.csum_flags |= csum_flags; 325 if (csum_flags & CSUM_DATA_VALID) 326 m->m_pkthdr.csum_data = 0xffff; 327 if_simloop(ifp, m, dst->sa_family, hlen); 328 return (0); /* XXX */ 329 } 330 } 331 332 #ifdef CARP 333 if (ifp->if_type == IFT_CARP) { 334 ifp = carp_parent(ifp); 335 if (ifp == NULL) 336 gotoerr(ENETUNREACH); 337 338 ac = IFP2AC(ifp); 339 340 /* 341 * Check precondition again 342 */ 343 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 344 345 if (ifp->if_flags & IFF_MONITOR) 346 gotoerr(ENETDOWN); 347 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 348 (IFF_UP | IFF_RUNNING)) 349 gotoerr(ENETDOWN); 350 } 351 #endif 352 353 /* Handle ng_ether(4) processing, if any */ 354 if (ng_ether_output_p != NULL) { 355 /* 356 * Hold BGL and recheck ng_ether_output_p 357 */ 358 get_mplock(); 359 if (ng_ether_output_p != NULL) { 360 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 361 rel_mplock(); 362 goto bad; 363 } 364 if (m == NULL) { 365 rel_mplock(); 366 return (0); 367 } 368 } 369 rel_mplock(); 370 } 371 372 /* Continue with link-layer output */ 373 return ether_output_frame(ifp, m); 374 375 bad: 376 m_freem(m); 377 return (error); 378 } 379 380 /* 381 * Returns the bridge interface an ifp is associated 382 * with. 383 * 384 * Only call if ifp->if_bridge != NULL. 385 */ 386 struct ifnet * 387 ether_bridge_interface(struct ifnet *ifp) 388 { 389 if (bridge_interface_p) 390 return(bridge_interface_p(ifp->if_bridge)); 391 return (ifp); 392 } 393 394 /* 395 * Ethernet link layer output routine to send a raw frame to the device. 396 * 397 * This assumes that the 14 byte Ethernet header is present and contiguous 398 * in the first mbuf. 399 */ 400 int 401 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 402 { 403 struct ip_fw *rule = NULL; 404 int error = 0; 405 struct altq_pktattr pktattr; 406 407 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 408 409 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 410 struct m_tag *mtag; 411 412 /* Extract info from dummynet tag */ 413 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 414 KKASSERT(mtag != NULL); 415 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 416 KKASSERT(rule != NULL); 417 418 m_tag_delete(m, mtag); 419 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 420 } 421 422 if (ifq_is_enabled(&ifp->if_snd)) 423 altq_etherclassify(&ifp->if_snd, m, &pktattr); 424 crit_enter(); 425 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0) { 426 struct ether_header save_eh, *eh; 427 428 eh = mtod(m, struct ether_header *); 429 save_eh = *eh; 430 m_adj(m, ETHER_HDR_LEN); 431 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 432 crit_exit(); 433 if (m != NULL) { 434 m_freem(m); 435 return ENOBUFS; /* pkt dropped */ 436 } else 437 return 0; /* consumed e.g. in a pipe */ 438 } 439 440 /* packet was ok, restore the ethernet header */ 441 ether_restore_header(&m, eh, &save_eh); 442 if (m == NULL) { 443 crit_exit(); 444 return ENOBUFS; 445 } 446 } 447 crit_exit(); 448 449 /* 450 * Queue message on interface, update output statistics if 451 * successful, and start output if interface not yet active. 452 */ 453 error = ifq_dispatch(ifp, m, &pktattr); 454 return (error); 455 } 456 457 /* 458 * ipfw processing for ethernet packets (in and out). 459 * The second parameter is NULL from ether_demux(), and ifp from 460 * ether_output_frame(). 461 */ 462 static boolean_t 463 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 464 const struct ether_header *eh) 465 { 466 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 467 struct ip_fw_args args; 468 struct m_tag *mtag; 469 struct mbuf *m; 470 int i; 471 472 if (*rule != NULL && fw_one_pass) 473 return TRUE; /* dummynet packet, already partially processed */ 474 475 /* 476 * I need some amount of data to be contiguous. 477 */ 478 i = min((*m0)->m_pkthdr.len, max_protohdr); 479 if ((*m0)->m_len < i) { 480 *m0 = m_pullup(*m0, i); 481 if (*m0 == NULL) 482 return FALSE; 483 } 484 485 /* 486 * Clean up tags 487 */ 488 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 489 m_tag_delete(*m0, mtag); 490 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 491 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 492 KKASSERT(mtag != NULL); 493 m_tag_delete(*m0, mtag); 494 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 495 } 496 497 args.m = *m0; /* the packet we are looking at */ 498 args.oif = dst; /* destination, if any */ 499 args.rule = *rule; /* matching rule to restart */ 500 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 501 i = ip_fw_chk_ptr(&args); 502 *m0 = args.m; 503 *rule = args.rule; 504 505 if (*m0 == NULL) 506 return FALSE; 507 508 switch (i) { 509 case IP_FW_PASS: 510 return TRUE; 511 512 case IP_FW_DIVERT: 513 case IP_FW_TEE: 514 case IP_FW_DENY: 515 /* 516 * XXX at some point add support for divert/forward actions. 517 * If none of the above matches, we have to drop the pkt. 518 */ 519 return FALSE; 520 521 case IP_FW_DUMMYNET: 522 /* 523 * Pass the pkt to dummynet, which consumes it. 524 */ 525 m = *m0; /* pass the original to dummynet */ 526 *m0 = NULL; /* and nothing back to the caller */ 527 528 ether_restore_header(&m, eh, &save_eh); 529 if (m == NULL) 530 return FALSE; 531 532 ip_fw_dn_io_ptr(m, args.cookie, 533 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 534 ip_dn_queue(m); 535 return FALSE; 536 537 default: 538 panic("unknown ipfw return value: %d", i); 539 } 540 } 541 542 /* 543 * Perform common duties while attaching to interface list 544 */ 545 void 546 ether_ifattach(struct ifnet *ifp, const uint8_t *lla, 547 lwkt_serialize_t serializer) 548 { 549 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 550 serializer); 551 } 552 553 void 554 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla, 555 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer) 556 { 557 struct sockaddr_dl *sdl; 558 char ethstr[ETHER_ADDRSTRLEN + 1]; 559 struct ifaltq *ifq; 560 int i; 561 562 ifp->if_type = IFT_ETHER; 563 ifp->if_addrlen = ETHER_ADDR_LEN; 564 ifp->if_hdrlen = ETHER_HDR_LEN; 565 if_attach(ifp, serializer); 566 ifq = &ifp->if_snd; 567 for (i = 0; i < ifq->altq_subq_cnt; ++i) { 568 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i); 569 570 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen * 571 (ETHER_MAX_LEN - ETHER_CRC_LEN); 572 } 573 ifp->if_mtu = ETHERMTU; 574 if (ifp->if_tsolen <= 0) { 575 if ((ether_tsolen_default / ETHERMTU) < 2) { 576 kprintf("ether TSO maxlen %d -> %d\n", 577 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 578 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 579 } 580 ifp->if_tsolen = ether_tsolen_default; 581 } 582 if (ifp->if_baudrate == 0) 583 ifp->if_baudrate = 10000000; 584 ifp->if_output = ether_output; 585 ifp->if_input = ether_input; 586 ifp->if_resolvemulti = ether_resolvemulti; 587 ifp->if_broadcastaddr = etherbroadcastaddr; 588 sdl = IF_LLSOCKADDR(ifp); 589 sdl->sdl_type = IFT_ETHER; 590 sdl->sdl_alen = ifp->if_addrlen; 591 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 592 /* 593 * XXX Keep the current drivers happy. 594 * XXX Remove once all drivers have been cleaned up 595 */ 596 if (lla != IFP2AC(ifp)->ac_enaddr) 597 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 598 bpfattach(ifp, dlt, hdrlen); 599 if (ng_ether_attach_p != NULL) 600 (*ng_ether_attach_p)(ifp); 601 602 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 603 } 604 605 /* 606 * Perform common duties while detaching an Ethernet interface 607 */ 608 void 609 ether_ifdetach(struct ifnet *ifp) 610 { 611 if_down(ifp); 612 613 if (ng_ether_detach_p != NULL) 614 (*ng_ether_detach_p)(ifp); 615 bpfdetach(ifp); 616 if_detach(ifp); 617 } 618 619 int 620 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 621 { 622 struct ifaddr *ifa = (struct ifaddr *) data; 623 struct ifreq *ifr = (struct ifreq *) data; 624 int error = 0; 625 626 #define IF_INIT(ifp) \ 627 do { \ 628 if (((ifp)->if_flags & IFF_UP) == 0) { \ 629 (ifp)->if_flags |= IFF_UP; \ 630 (ifp)->if_init((ifp)->if_softc); \ 631 } \ 632 } while (0) 633 634 ASSERT_IFNET_SERIALIZED_ALL(ifp); 635 636 switch (command) { 637 case SIOCSIFADDR: 638 switch (ifa->ifa_addr->sa_family) { 639 #ifdef INET 640 case AF_INET: 641 IF_INIT(ifp); /* before arpwhohas */ 642 arp_ifinit(ifp, ifa); 643 break; 644 #endif 645 default: 646 IF_INIT(ifp); 647 break; 648 } 649 break; 650 651 case SIOCGIFADDR: 652 bcopy(IFP2AC(ifp)->ac_enaddr, 653 ((struct sockaddr *)ifr->ifr_data)->sa_data, 654 ETHER_ADDR_LEN); 655 break; 656 657 case SIOCSIFMTU: 658 /* 659 * Set the interface MTU. 660 */ 661 if (ifr->ifr_mtu > ETHERMTU) { 662 error = EINVAL; 663 } else { 664 ifp->if_mtu = ifr->ifr_mtu; 665 } 666 break; 667 default: 668 error = EINVAL; 669 break; 670 } 671 return (error); 672 673 #undef IF_INIT 674 } 675 676 int 677 ether_resolvemulti( 678 struct ifnet *ifp, 679 struct sockaddr **llsa, 680 struct sockaddr *sa) 681 { 682 struct sockaddr_dl *sdl; 683 #ifdef INET 684 struct sockaddr_in *sin; 685 #endif 686 #ifdef INET6 687 struct sockaddr_in6 *sin6; 688 #endif 689 u_char *e_addr; 690 691 switch(sa->sa_family) { 692 case AF_LINK: 693 /* 694 * No mapping needed. Just check that it's a valid MC address. 695 */ 696 sdl = (struct sockaddr_dl *)sa; 697 e_addr = LLADDR(sdl); 698 if ((e_addr[0] & 1) != 1) 699 return EADDRNOTAVAIL; 700 *llsa = NULL; 701 return 0; 702 703 #ifdef INET 704 case AF_INET: 705 sin = (struct sockaddr_in *)sa; 706 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 707 return EADDRNOTAVAIL; 708 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 709 sdl->sdl_len = sizeof *sdl; 710 sdl->sdl_family = AF_LINK; 711 sdl->sdl_index = ifp->if_index; 712 sdl->sdl_type = IFT_ETHER; 713 sdl->sdl_alen = ETHER_ADDR_LEN; 714 e_addr = LLADDR(sdl); 715 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 716 *llsa = (struct sockaddr *)sdl; 717 return 0; 718 #endif 719 #ifdef INET6 720 case AF_INET6: 721 sin6 = (struct sockaddr_in6 *)sa; 722 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 723 /* 724 * An IP6 address of 0 means listen to all 725 * of the Ethernet multicast address used for IP6. 726 * (This is used for multicast routers.) 727 */ 728 ifp->if_flags |= IFF_ALLMULTI; 729 *llsa = NULL; 730 return 0; 731 } 732 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 733 return EADDRNOTAVAIL; 734 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 735 sdl->sdl_len = sizeof *sdl; 736 sdl->sdl_family = AF_LINK; 737 sdl->sdl_index = ifp->if_index; 738 sdl->sdl_type = IFT_ETHER; 739 sdl->sdl_alen = ETHER_ADDR_LEN; 740 e_addr = LLADDR(sdl); 741 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 742 *llsa = (struct sockaddr *)sdl; 743 return 0; 744 #endif 745 746 default: 747 /* 748 * Well, the text isn't quite right, but it's the name 749 * that counts... 750 */ 751 return EAFNOSUPPORT; 752 } 753 } 754 755 #if 0 756 /* 757 * This is for reference. We have a table-driven version 758 * of the little-endian crc32 generator, which is faster 759 * than the double-loop. 760 */ 761 uint32_t 762 ether_crc32_le(const uint8_t *buf, size_t len) 763 { 764 uint32_t c, crc, carry; 765 size_t i, j; 766 767 crc = 0xffffffffU; /* initial value */ 768 769 for (i = 0; i < len; i++) { 770 c = buf[i]; 771 for (j = 0; j < 8; j++) { 772 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 773 crc >>= 1; 774 c >>= 1; 775 if (carry) 776 crc = (crc ^ ETHER_CRC_POLY_LE); 777 } 778 } 779 780 return (crc); 781 } 782 #else 783 uint32_t 784 ether_crc32_le(const uint8_t *buf, size_t len) 785 { 786 static const uint32_t crctab[] = { 787 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 788 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 789 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 790 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 791 }; 792 uint32_t crc; 793 size_t i; 794 795 crc = 0xffffffffU; /* initial value */ 796 797 for (i = 0; i < len; i++) { 798 crc ^= buf[i]; 799 crc = (crc >> 4) ^ crctab[crc & 0xf]; 800 crc = (crc >> 4) ^ crctab[crc & 0xf]; 801 } 802 803 return (crc); 804 } 805 #endif 806 807 uint32_t 808 ether_crc32_be(const uint8_t *buf, size_t len) 809 { 810 uint32_t c, crc, carry; 811 size_t i, j; 812 813 crc = 0xffffffffU; /* initial value */ 814 815 for (i = 0; i < len; i++) { 816 c = buf[i]; 817 for (j = 0; j < 8; j++) { 818 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 819 crc <<= 1; 820 c >>= 1; 821 if (carry) 822 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 823 } 824 } 825 826 return (crc); 827 } 828 829 /* 830 * find the size of ethernet header, and call classifier 831 */ 832 void 833 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 834 struct altq_pktattr *pktattr) 835 { 836 struct ether_header *eh; 837 uint16_t ether_type; 838 int hlen, af, hdrsize; 839 840 hlen = sizeof(struct ether_header); 841 eh = mtod(m, struct ether_header *); 842 843 ether_type = ntohs(eh->ether_type); 844 if (ether_type < ETHERMTU) { 845 /* ick! LLC/SNAP */ 846 struct llc *llc = (struct llc *)(eh + 1); 847 hlen += 8; 848 849 if (m->m_len < hlen || 850 llc->llc_dsap != LLC_SNAP_LSAP || 851 llc->llc_ssap != LLC_SNAP_LSAP || 852 llc->llc_control != LLC_UI) 853 goto bad; /* not snap! */ 854 855 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 856 } 857 858 if (ether_type == ETHERTYPE_IP) { 859 af = AF_INET; 860 hdrsize = 20; /* sizeof(struct ip) */ 861 #ifdef INET6 862 } else if (ether_type == ETHERTYPE_IPV6) { 863 af = AF_INET6; 864 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 865 #endif 866 } else 867 goto bad; 868 869 while (m->m_len <= hlen) { 870 hlen -= m->m_len; 871 m = m->m_next; 872 } 873 if (m->m_len < hlen + hdrsize) { 874 /* 875 * ip header is not in a single mbuf. this should not 876 * happen in the current code. 877 * (todo: use m_pulldown in the future) 878 */ 879 goto bad; 880 } 881 m->m_data += hlen; 882 m->m_len -= hlen; 883 ifq_classify(ifq, m, af, pktattr); 884 m->m_data -= hlen; 885 m->m_len += hlen; 886 887 return; 888 889 bad: 890 pktattr->pattr_class = NULL; 891 pktattr->pattr_hdr = NULL; 892 pktattr->pattr_af = AF_UNSPEC; 893 } 894 895 static void 896 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 897 const struct ether_header *save_eh) 898 { 899 struct mbuf *m = *m0; 900 901 ether_restore_hdr++; 902 903 /* 904 * Prepend the header, optimize for the common case of 905 * eh pointing into the mbuf. 906 */ 907 if ((const void *)(eh + 1) == (void *)m->m_data) { 908 m->m_data -= ETHER_HDR_LEN; 909 m->m_len += ETHER_HDR_LEN; 910 m->m_pkthdr.len += ETHER_HDR_LEN; 911 } else { 912 ether_prepend_hdr++; 913 914 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 915 if (m != NULL) { 916 bcopy(save_eh, mtod(m, struct ether_header *), 917 ETHER_HDR_LEN); 918 } 919 } 920 *m0 = m; 921 } 922 923 /* 924 * Upper layer processing for a received Ethernet packet. 925 */ 926 void 927 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 928 { 929 struct ether_header *eh; 930 int isr, discard = 0; 931 u_short ether_type; 932 struct ip_fw *rule = NULL; 933 934 M_ASSERTPKTHDR(m); 935 KASSERT(m->m_len >= ETHER_HDR_LEN, 936 ("ether header is not contiguous!")); 937 938 eh = mtod(m, struct ether_header *); 939 940 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 941 struct m_tag *mtag; 942 943 /* Extract info from dummynet tag */ 944 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 945 KKASSERT(mtag != NULL); 946 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 947 KKASSERT(rule != NULL); 948 949 m_tag_delete(m, mtag); 950 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 951 952 /* packet is passing the second time */ 953 goto post_stats; 954 } 955 956 /* 957 * We got a packet which was unicast to a different Ethernet 958 * address. If the driver is working properly, then this 959 * situation can only happen when the interface is in 960 * promiscuous mode. We defer the packet discarding until the 961 * vlan processing is done, so that vlan/bridge or vlan/netgraph 962 * could work. 963 */ 964 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 965 !ETHER_IS_MULTICAST(eh->ether_dhost) && 966 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 967 if (ether_debug & 1) { 968 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 969 "%02x:%02x:%02x:%02x:%02x:%02x " 970 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 971 eh->ether_dhost[0], 972 eh->ether_dhost[1], 973 eh->ether_dhost[2], 974 eh->ether_dhost[3], 975 eh->ether_dhost[4], 976 eh->ether_dhost[5], 977 eh->ether_shost[0], 978 eh->ether_shost[1], 979 eh->ether_shost[2], 980 eh->ether_shost[3], 981 eh->ether_shost[4], 982 eh->ether_shost[5], 983 eh->ether_type, 984 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 985 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 986 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 987 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 988 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 989 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 990 ); 991 } 992 if ((ether_debug & 2) == 0) 993 discard = 1; 994 } 995 996 post_stats: 997 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0 && !discard) { 998 struct ether_header save_eh = *eh; 999 1000 /* XXX old crufty stuff, needs to be removed */ 1001 m_adj(m, sizeof(struct ether_header)); 1002 1003 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1004 m_freem(m); 1005 return; 1006 } 1007 1008 ether_restore_header(&m, eh, &save_eh); 1009 if (m == NULL) 1010 return; 1011 eh = mtod(m, struct ether_header *); 1012 } 1013 1014 ether_type = ntohs(eh->ether_type); 1015 KKASSERT(ether_type != ETHERTYPE_VLAN); 1016 1017 /* Handle input from a lagg(4) port */ 1018 if (ifp->if_type == IFT_IEEE8023ADLAG) { 1019 KASSERT(lagg_input_p != NULL, 1020 ("%s: if_lagg not loaded!", __func__)); 1021 (*lagg_input_p)(ifp, m); 1022 return; 1023 } 1024 1025 if (m->m_flags & M_VLANTAG) { 1026 void (*vlan_input_func)(struct mbuf *); 1027 1028 vlan_input_func = vlan_input_p; 1029 if (vlan_input_func != NULL) { 1030 vlan_input_func(m); 1031 } else { 1032 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1033 m_freem(m); 1034 } 1035 return; 1036 } 1037 1038 /* 1039 * If we have been asked to discard this packet 1040 * (e.g. not for us), drop it before entering 1041 * the upper layer. 1042 */ 1043 if (discard) { 1044 m_freem(m); 1045 return; 1046 } 1047 1048 /* 1049 * Clear protocol specific flags, 1050 * before entering the upper layer. 1051 */ 1052 m->m_flags &= ~M_ETHER_FLAGS; 1053 1054 /* Strip ethernet header. */ 1055 m_adj(m, sizeof(struct ether_header)); 1056 1057 switch (ether_type) { 1058 #ifdef INET 1059 case ETHERTYPE_IP: 1060 if ((m->m_flags & M_LENCHECKED) == 0) { 1061 if (!ip_lengthcheck(&m, 0)) 1062 return; 1063 } 1064 if (ipflow_fastforward(m)) 1065 return; 1066 isr = NETISR_IP; 1067 break; 1068 1069 case ETHERTYPE_ARP: 1070 if (ifp->if_flags & IFF_NOARP) { 1071 /* Discard packet if ARP is disabled on interface */ 1072 m_freem(m); 1073 return; 1074 } 1075 isr = NETISR_ARP; 1076 break; 1077 #endif 1078 1079 #ifdef INET6 1080 case ETHERTYPE_IPV6: 1081 isr = NETISR_IPV6; 1082 break; 1083 #endif 1084 1085 #ifdef MPLS 1086 case ETHERTYPE_MPLS: 1087 case ETHERTYPE_MPLS_MCAST: 1088 /* Should have been set by ether_input(). */ 1089 KKASSERT(m->m_flags & M_MPLSLABELED); 1090 isr = NETISR_MPLS; 1091 break; 1092 #endif 1093 1094 default: 1095 /* 1096 * The accurate msgport is not determined before 1097 * we reach here, so recharacterize packet. 1098 */ 1099 m->m_flags &= ~M_HASH; 1100 if (ng_ether_input_orphan_p != NULL) { 1101 /* 1102 * Put back the ethernet header so netgraph has a 1103 * consistent view of inbound packets. 1104 */ 1105 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 1106 if (m == NULL) { 1107 /* 1108 * M_PREPEND frees the mbuf in case of failure. 1109 */ 1110 return; 1111 } 1112 /* 1113 * Hold BGL and recheck ng_ether_input_orphan_p 1114 */ 1115 get_mplock(); 1116 if (ng_ether_input_orphan_p != NULL) { 1117 ng_ether_input_orphan_p(ifp, m); 1118 rel_mplock(); 1119 return; 1120 } 1121 rel_mplock(); 1122 } 1123 m_freem(m); 1124 return; 1125 } 1126 1127 if (m->m_flags & M_HASH) { 1128 if (&curthread->td_msgport == 1129 netisr_hashport(m->m_pkthdr.hash)) { 1130 netisr_handle(isr, m); 1131 return; 1132 } else { 1133 /* 1134 * XXX Something is wrong, 1135 * we probably should panic here! 1136 */ 1137 m->m_flags &= ~M_HASH; 1138 atomic_add_long(ðer_input_wronghash, 1); 1139 } 1140 } 1141 #ifdef RSS_DEBUG 1142 atomic_add_long(ðer_input_requeue, 1); 1143 #endif 1144 netisr_queue(isr, m); 1145 } 1146 1147 /* 1148 * First we perform any link layer operations, then continue to the 1149 * upper layers with ether_demux_oncpu(). 1150 */ 1151 static void 1152 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1153 { 1154 #ifdef CARP 1155 void *carp; 1156 #endif 1157 1158 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1159 /* 1160 * Receiving interface's flags are changed, when this 1161 * packet is waiting for processing; discard it. 1162 */ 1163 m_freem(m); 1164 return; 1165 } 1166 1167 /* 1168 * Tap the packet off here for a bridge. bridge_input() 1169 * will return NULL if it has consumed the packet, otherwise 1170 * it gets processed as normal. Note that bridge_input() 1171 * will always return the original packet if we need to 1172 * process it locally. 1173 */ 1174 if (ifp->if_bridge) { 1175 KASSERT(bridge_input_p != NULL, 1176 ("%s: if_bridge not loaded!", __func__)); 1177 1178 if(m->m_flags & M_ETHER_BRIDGED) { 1179 m->m_flags &= ~M_ETHER_BRIDGED; 1180 } else { 1181 m = bridge_input_p(ifp, m); 1182 if (m == NULL) 1183 return; 1184 1185 KASSERT(ifp == m->m_pkthdr.rcvif, 1186 ("bridge_input_p changed rcvif")); 1187 } 1188 } 1189 1190 #ifdef CARP 1191 carp = ifp->if_carp; 1192 if (carp) { 1193 m = carp_input(carp, m); 1194 if (m == NULL) 1195 return; 1196 KASSERT(ifp == m->m_pkthdr.rcvif, 1197 ("carp_input changed rcvif")); 1198 } 1199 #endif 1200 1201 /* Handle ng_ether(4) processing, if any */ 1202 if (ng_ether_input_p != NULL) { 1203 /* 1204 * Hold BGL and recheck ng_ether_input_p 1205 */ 1206 get_mplock(); 1207 if (ng_ether_input_p != NULL) 1208 ng_ether_input_p(ifp, &m); 1209 rel_mplock(); 1210 1211 if (m == NULL) 1212 return; 1213 } 1214 1215 /* Continue with upper layer processing */ 1216 ether_demux_oncpu(ifp, m); 1217 } 1218 1219 /* 1220 * Perform certain functions of ether_input(): 1221 * - Test IFF_UP 1222 * - Update statistics 1223 * - Run bpf(4) tap if requested 1224 * Then pass the packet to ether_input_oncpu(). 1225 * 1226 * This function should be used by pseudo interface (e.g. vlan(4)), 1227 * when it tries to claim that the packet is received by it. 1228 * 1229 * REINPUT_KEEPRCVIF 1230 * REINPUT_RUNBPF 1231 */ 1232 void 1233 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1234 { 1235 /* Discard packet if interface is not up */ 1236 if (!(ifp->if_flags & IFF_UP)) { 1237 m_freem(m); 1238 return; 1239 } 1240 1241 /* 1242 * Change receiving interface. The bridge will often pass a flag to 1243 * ask that this not be done so ARPs get applied to the correct 1244 * side. 1245 */ 1246 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1247 m->m_pkthdr.rcvif == NULL) { 1248 m->m_pkthdr.rcvif = ifp; 1249 } 1250 1251 /* Update statistics */ 1252 IFNET_STAT_INC(ifp, ipackets, 1); 1253 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1254 if (m->m_flags & (M_MCAST | M_BCAST)) 1255 IFNET_STAT_INC(ifp, imcasts, 1); 1256 1257 if (reinput_flags & REINPUT_RUNBPF) 1258 BPF_MTAP(ifp, m); 1259 1260 ether_input_oncpu(ifp, m); 1261 } 1262 1263 static __inline boolean_t 1264 ether_vlancheck(struct mbuf **m0) 1265 { 1266 struct mbuf *m = *m0; 1267 struct ether_header *eh; 1268 uint16_t ether_type; 1269 1270 eh = mtod(m, struct ether_header *); 1271 ether_type = ntohs(eh->ether_type); 1272 1273 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1274 /* 1275 * Extract vlan tag if hardware does not do it for us 1276 */ 1277 vlan_ether_decap(&m); 1278 if (m == NULL) 1279 goto failed; 1280 1281 eh = mtod(m, struct ether_header *); 1282 ether_type = ntohs(eh->ether_type); 1283 } 1284 1285 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1286 /* 1287 * To prevent possible dangerous recursion, 1288 * we don't do vlan-in-vlan 1289 */ 1290 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1291 goto failed; 1292 } 1293 KKASSERT(ether_type != ETHERTYPE_VLAN); 1294 1295 m->m_flags |= M_ETHER_VLANCHECKED; 1296 *m0 = m; 1297 return TRUE; 1298 failed: 1299 if (m != NULL) 1300 m_freem(m); 1301 *m0 = NULL; 1302 return FALSE; 1303 } 1304 1305 static void 1306 ether_input_handler(netmsg_t nmsg) 1307 { 1308 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1309 struct ether_header *eh; 1310 struct ifnet *ifp; 1311 struct mbuf *m; 1312 1313 m = nmp->nm_packet; 1314 M_ASSERTPKTHDR(m); 1315 1316 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1317 if (!ether_vlancheck(&m)) { 1318 KKASSERT(m == NULL); 1319 return; 1320 } 1321 } 1322 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) || 1323 __predict_false(ether_input_ckhash)) { 1324 int isr; 1325 1326 /* 1327 * Need to verify the hash supplied by the hardware 1328 * which could be wrong. 1329 */ 1330 m->m_flags &= ~(M_HASH | M_CKHASH); 1331 isr = ether_characterize(&m); 1332 if (m == NULL) 1333 return; 1334 KKASSERT(m->m_flags & M_HASH); 1335 1336 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) { 1337 /* 1338 * Wrong hardware supplied hash; redispatch 1339 */ 1340 ether_dispatch(isr, m, -1); 1341 if (__predict_false(ether_input_ckhash)) 1342 atomic_add_long(ðer_input_wronghwhash, 1); 1343 return; 1344 } 1345 } 1346 ifp = m->m_pkthdr.rcvif; 1347 1348 eh = mtod(m, struct ether_header *); 1349 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1350 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1351 ifp->if_addrlen) == 0) 1352 m->m_flags |= M_BCAST; 1353 else 1354 m->m_flags |= M_MCAST; 1355 IFNET_STAT_INC(ifp, imcasts, 1); 1356 } 1357 1358 ether_input_oncpu(ifp, m); 1359 } 1360 1361 /* 1362 * Send the packet to the target netisr msgport 1363 * 1364 * At this point the packet must be characterized (M_HASH set), 1365 * so we know which netisr to send it to. 1366 */ 1367 static void 1368 ether_dispatch(int isr, struct mbuf *m, int cpuid) 1369 { 1370 struct netmsg_packet *pmsg; 1371 int target_cpuid; 1372 1373 KKASSERT(m->m_flags & M_HASH); 1374 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash); 1375 1376 pmsg = &m->m_hdr.mh_netmsg; 1377 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1378 0, ether_input_handler); 1379 pmsg->nm_packet = m; 1380 pmsg->base.lmsg.u.ms_result = isr; 1381 1382 logether(disp_beg, NULL); 1383 if (target_cpuid == cpuid) { 1384 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid), 1385 &pmsg->base.lmsg); 1386 } else { 1387 lwkt_sendmsg(netisr_cpuport(target_cpuid), 1388 &pmsg->base.lmsg); 1389 } 1390 logether(disp_end, NULL); 1391 } 1392 1393 /* 1394 * Process a received Ethernet packet. 1395 * 1396 * The ethernet header is assumed to be in the mbuf so the caller 1397 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1398 * bytes in the first mbuf. 1399 * 1400 * If the caller knows that the current thread is stick to the current 1401 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid 1402 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should 1403 * be passed as 'cpuid' argument. 1404 */ 1405 void 1406 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1407 int cpuid) 1408 { 1409 int isr; 1410 1411 M_ASSERTPKTHDR(m); 1412 1413 /* Discard packet if interface is not up */ 1414 if (!(ifp->if_flags & IFF_UP)) { 1415 m_freem(m); 1416 return; 1417 } 1418 1419 if (m->m_len < sizeof(struct ether_header)) { 1420 /* XXX error in the caller. */ 1421 m_freem(m); 1422 return; 1423 } 1424 1425 m->m_pkthdr.rcvif = ifp; 1426 1427 logether(pkt_beg, ifp); 1428 1429 ETHER_BPF_MTAP(ifp, m); 1430 1431 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1432 1433 if (ifp->if_flags & IFF_MONITOR) { 1434 struct ether_header *eh; 1435 1436 eh = mtod(m, struct ether_header *); 1437 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1438 IFNET_STAT_INC(ifp, imcasts, 1); 1439 1440 /* 1441 * Interface marked for monitoring; discard packet. 1442 */ 1443 m_freem(m); 1444 1445 logether(pkt_end, ifp); 1446 return; 1447 } 1448 1449 /* 1450 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1451 * we can dispatch it immediately with trivial checks. 1452 */ 1453 if (pi != NULL && (m->m_flags & M_HASH)) { 1454 #ifdef RSS_DEBUG 1455 atomic_add_long(ðer_pktinfo_try, 1); 1456 #endif 1457 netisr_hashcheck(pi->pi_netisr, m, pi); 1458 if (m->m_flags & M_HASH) { 1459 ether_dispatch(pi->pi_netisr, m, cpuid); 1460 #ifdef RSS_DEBUG 1461 atomic_add_long(ðer_pktinfo_hit, 1); 1462 #endif 1463 logether(pkt_end, ifp); 1464 return; 1465 } 1466 } 1467 #ifdef RSS_DEBUG 1468 else if (ifp->if_capenable & IFCAP_RSS) { 1469 if (pi == NULL) 1470 atomic_add_long(ðer_rss_nopi, 1); 1471 else 1472 atomic_add_long(ðer_rss_nohash, 1); 1473 } 1474 #endif 1475 1476 /* 1477 * Packet hash will be recalculated by software, so clear 1478 * the M_HASH and M_CKHASH flag set by the driver; the hash 1479 * value calculated by the hardware may not be exactly what 1480 * we want. 1481 */ 1482 m->m_flags &= ~(M_HASH | M_CKHASH); 1483 1484 if (!ether_vlancheck(&m)) { 1485 KKASSERT(m == NULL); 1486 logether(pkt_end, ifp); 1487 return; 1488 } 1489 1490 isr = ether_characterize(&m); 1491 if (m == NULL) { 1492 logether(pkt_end, ifp); 1493 return; 1494 } 1495 1496 /* 1497 * Finally dispatch it 1498 */ 1499 ether_dispatch(isr, m, cpuid); 1500 1501 logether(pkt_end, ifp); 1502 } 1503 1504 static int 1505 ether_characterize(struct mbuf **m0) 1506 { 1507 struct mbuf *m = *m0; 1508 struct ether_header *eh; 1509 uint16_t ether_type; 1510 int isr; 1511 1512 eh = mtod(m, struct ether_header *); 1513 ether_type = ntohs(eh->ether_type); 1514 1515 /* 1516 * Map ether type to netisr id. 1517 */ 1518 switch (ether_type) { 1519 #ifdef INET 1520 case ETHERTYPE_IP: 1521 isr = NETISR_IP; 1522 break; 1523 1524 case ETHERTYPE_ARP: 1525 isr = NETISR_ARP; 1526 break; 1527 #endif 1528 1529 #ifdef INET6 1530 case ETHERTYPE_IPV6: 1531 isr = NETISR_IPV6; 1532 break; 1533 #endif 1534 1535 #ifdef MPLS 1536 case ETHERTYPE_MPLS: 1537 case ETHERTYPE_MPLS_MCAST: 1538 m->m_flags |= M_MPLSLABELED; 1539 isr = NETISR_MPLS; 1540 break; 1541 #endif 1542 1543 default: 1544 /* 1545 * NETISR_MAX is an invalid value; it is chosen to let 1546 * netisr_characterize() know that we have no clear 1547 * idea where this packet should go. 1548 */ 1549 isr = NETISR_MAX; 1550 break; 1551 } 1552 1553 /* 1554 * Ask the isr to characterize the packet since we couldn't. 1555 * This is an attempt to optimally get us onto the correct protocol 1556 * thread. 1557 */ 1558 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1559 1560 *m0 = m; 1561 return isr; 1562 } 1563 1564 static void 1565 ether_demux_handler(netmsg_t nmsg) 1566 { 1567 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1568 struct ifnet *ifp; 1569 struct mbuf *m; 1570 1571 m = nmp->nm_packet; 1572 M_ASSERTPKTHDR(m); 1573 ifp = m->m_pkthdr.rcvif; 1574 1575 ether_demux_oncpu(ifp, m); 1576 } 1577 1578 void 1579 ether_demux(struct mbuf *m) 1580 { 1581 struct netmsg_packet *pmsg; 1582 int isr; 1583 1584 isr = ether_characterize(&m); 1585 if (m == NULL) 1586 return; 1587 1588 KKASSERT(m->m_flags & M_HASH); 1589 pmsg = &m->m_hdr.mh_netmsg; 1590 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1591 0, ether_demux_handler); 1592 pmsg->nm_packet = m; 1593 pmsg->base.lmsg.u.ms_result = isr; 1594 1595 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1596 } 1597 1598 u_char * 1599 kether_aton(const char *macstr, u_char *addr) 1600 { 1601 unsigned int o0, o1, o2, o3, o4, o5; 1602 int n; 1603 1604 if (macstr == NULL || addr == NULL) 1605 return NULL; 1606 1607 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1608 &o3, &o4, &o5); 1609 if (n != 6) 1610 return NULL; 1611 1612 addr[0] = o0; 1613 addr[1] = o1; 1614 addr[2] = o2; 1615 addr[3] = o3; 1616 addr[4] = o4; 1617 addr[5] = o5; 1618 1619 return addr; 1620 } 1621 1622 char * 1623 kether_ntoa(const u_char *addr, char *buf) 1624 { 1625 int len = ETHER_ADDRSTRLEN + 1; 1626 int n; 1627 1628 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1629 addr[1], addr[2], addr[3], addr[4], addr[5]); 1630 1631 if (n < 17) 1632 return NULL; 1633 1634 return buf; 1635 } 1636 1637 MODULE_VERSION(ether, 1); 1638