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 #define ETHER_NMBCLUSTERS_DEFMIN 32 156 #define ETHER_NMBCLUSTERS_DEFAULT 256 157 158 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 159 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default); 160 161 static int ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT; 162 TUNABLE_INT("net.link.ether.nmbclusters", ðer_nmbclusters_default); 163 164 SYSCTL_DECL(_net_link); 165 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 166 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW, 167 ðer_debug, 0, "Ether debug"); 168 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 169 ðer_ipfw, 0, "Pass ether pkts through firewall"); 170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 171 ðer_restore_hdr, 0, "# of ether header restoration"); 172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 173 ðer_prepend_hdr, 0, 174 "# of ether header restoration which prepends mbuf"); 175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW, 176 ðer_input_wronghash, 0, "# of input packets with wrong hash"); 177 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW, 178 ðer_tsolen_default, 0, "Default max TSO length"); 179 180 #ifdef RSS_DEBUG 181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 182 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 183 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 184 ðer_rss_nohash, 0, "# of packets do not have hash"); 185 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW, 186 ðer_pktinfo_try, 0, 187 "# of tries to find packets' msgport using pktinfo"); 188 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 189 ðer_pktinfo_hit, 0, 190 "# of packets whose msgport are found using pktinfo"); 191 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW, 192 ðer_input_requeue, 0, "# of input packets gets requeued"); 193 #endif 194 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW, 195 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash"); 196 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW, 197 ðer_input_ckhash, 0, "always check hash"); 198 199 #define ETHER_KTR_STR "ifp=%p" 200 #define ETHER_KTR_ARGS struct ifnet *ifp 201 #ifndef KTR_ETHERNET 202 #define KTR_ETHERNET KTR_ALL 203 #endif 204 KTR_INFO_MASTER(ether); 205 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS); 206 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS); 207 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS); 208 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS); 209 #define logether(name, arg) KTR_LOG(ether_ ## name, arg) 210 211 /* 212 * Ethernet output routine. 213 * Encapsulate a packet of type family for the local net. 214 * Use trailer local net encapsulation if enough data in first 215 * packet leaves a multiple of 512 bytes of data in remainder. 216 * Assumes that ifp is actually pointer to arpcom structure. 217 */ 218 static int 219 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 220 struct rtentry *rt) 221 { 222 struct ether_header *eh, *deh; 223 u_char *edst; 224 int loop_copy = 0; 225 int hlen = ETHER_HDR_LEN; /* link layer header length */ 226 struct arpcom *ac = IFP2AC(ifp); 227 int error; 228 229 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 230 231 if (ifp->if_flags & IFF_MONITOR) 232 gotoerr(ENETDOWN); 233 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 234 gotoerr(ENETDOWN); 235 236 M_PREPEND(m, sizeof(struct ether_header), M_NOWAIT); 237 if (m == NULL) 238 return (ENOBUFS); 239 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 240 eh = mtod(m, struct ether_header *); 241 edst = eh->ether_dhost; 242 243 /* 244 * Fill in the destination ethernet address and frame type. 245 */ 246 switch (dst->sa_family) { 247 #ifdef INET 248 case AF_INET: 249 if (!arpresolve(ifp, rt, m, dst, edst)) 250 return (0); /* if not yet resolved */ 251 #ifdef MPLS 252 if (m->m_flags & M_MPLSLABELED) 253 eh->ether_type = htons(ETHERTYPE_MPLS); 254 else 255 #endif 256 eh->ether_type = htons(ETHERTYPE_IP); 257 break; 258 #endif 259 #ifdef INET6 260 case AF_INET6: 261 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 262 return (0); /* Something bad happenned. */ 263 eh->ether_type = htons(ETHERTYPE_IPV6); 264 break; 265 #endif 266 case pseudo_AF_HDRCMPLT: 267 case AF_UNSPEC: 268 loop_copy = -1; /* if this is for us, don't do it */ 269 deh = (struct ether_header *)dst->sa_data; 270 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 271 eh->ether_type = deh->ether_type; 272 break; 273 274 default: 275 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 276 gotoerr(EAFNOSUPPORT); 277 } 278 279 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 280 memcpy(eh->ether_shost, 281 ((struct ether_header *)dst->sa_data)->ether_shost, 282 ETHER_ADDR_LEN); 283 else 284 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 285 286 /* 287 * Bridges require special output handling. 288 */ 289 if (ifp->if_bridge) { 290 KASSERT(bridge_output_p != NULL, 291 ("%s: if_bridge not loaded!", __func__)); 292 return bridge_output_p(ifp, m); 293 } 294 #if 0 /* XXX */ 295 if (ifp->if_lagg) { 296 KASSERT(lagg_output_p != NULL, 297 ("%s: if_lagg not loaded!", __func__)); 298 return lagg_output_p(ifp, m); 299 } 300 #endif 301 302 /* 303 * If a simplex interface, and the packet is being sent to our 304 * Ethernet address or a broadcast address, loopback a copy. 305 * XXX To make a simplex device behave exactly like a duplex 306 * device, we should copy in the case of sending to our own 307 * ethernet address (thus letting the original actually appear 308 * on the wire). However, we don't do that here for security 309 * reasons and compatibility with the original behavior. 310 */ 311 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 312 int csum_flags = 0; 313 314 if (m->m_pkthdr.csum_flags & CSUM_IP) 315 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 316 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 317 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 318 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 319 struct mbuf *n; 320 321 if ((n = m_copypacket(m, M_NOWAIT)) != NULL) { 322 n->m_pkthdr.csum_flags |= csum_flags; 323 if (csum_flags & CSUM_DATA_VALID) 324 n->m_pkthdr.csum_data = 0xffff; 325 if_simloop(ifp, n, dst->sa_family, hlen); 326 } else 327 IFNET_STAT_INC(ifp, iqdrops, 1); 328 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 329 ETHER_ADDR_LEN) == 0) { 330 m->m_pkthdr.csum_flags |= csum_flags; 331 if (csum_flags & CSUM_DATA_VALID) 332 m->m_pkthdr.csum_data = 0xffff; 333 if_simloop(ifp, m, dst->sa_family, hlen); 334 return (0); /* XXX */ 335 } 336 } 337 338 #ifdef CARP 339 if (ifp->if_type == IFT_CARP) { 340 ifp = carp_parent(ifp); 341 if (ifp == NULL) 342 gotoerr(ENETUNREACH); 343 344 ac = IFP2AC(ifp); 345 346 /* 347 * Check precondition again 348 */ 349 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 350 351 if (ifp->if_flags & IFF_MONITOR) 352 gotoerr(ENETDOWN); 353 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 354 (IFF_UP | IFF_RUNNING)) 355 gotoerr(ENETDOWN); 356 } 357 #endif 358 359 /* Handle ng_ether(4) processing, if any */ 360 if (ng_ether_output_p != NULL) { 361 /* 362 * Hold BGL and recheck ng_ether_output_p 363 */ 364 get_mplock(); 365 if (ng_ether_output_p != NULL) { 366 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 367 rel_mplock(); 368 goto bad; 369 } 370 if (m == NULL) { 371 rel_mplock(); 372 return (0); 373 } 374 } 375 rel_mplock(); 376 } 377 378 /* Continue with link-layer output */ 379 return ether_output_frame(ifp, m); 380 381 bad: 382 m_freem(m); 383 return (error); 384 } 385 386 /* 387 * Returns the bridge interface an ifp is associated 388 * with. 389 * 390 * Only call if ifp->if_bridge != NULL. 391 */ 392 struct ifnet * 393 ether_bridge_interface(struct ifnet *ifp) 394 { 395 if (bridge_interface_p) 396 return(bridge_interface_p(ifp->if_bridge)); 397 return (ifp); 398 } 399 400 /* 401 * Ethernet link layer output routine to send a raw frame to the device. 402 * 403 * This assumes that the 14 byte Ethernet header is present and contiguous 404 * in the first mbuf. 405 */ 406 int 407 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 408 { 409 struct ip_fw *rule = NULL; 410 int error = 0; 411 struct altq_pktattr pktattr; 412 413 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 414 415 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 416 struct m_tag *mtag; 417 418 /* Extract info from dummynet tag */ 419 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 420 KKASSERT(mtag != NULL); 421 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 422 KKASSERT(rule != NULL); 423 424 m_tag_delete(m, mtag); 425 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 426 } 427 428 if (ifq_is_enabled(&ifp->if_snd)) 429 altq_etherclassify(&ifp->if_snd, m, &pktattr); 430 crit_enter(); 431 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0) { 432 struct ether_header save_eh, *eh; 433 434 eh = mtod(m, struct ether_header *); 435 save_eh = *eh; 436 m_adj(m, ETHER_HDR_LEN); 437 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 438 crit_exit(); 439 if (m != NULL) { 440 m_freem(m); 441 return ENOBUFS; /* pkt dropped */ 442 } else 443 return 0; /* consumed e.g. in a pipe */ 444 } 445 446 /* packet was ok, restore the ethernet header */ 447 ether_restore_header(&m, eh, &save_eh); 448 if (m == NULL) { 449 crit_exit(); 450 return ENOBUFS; 451 } 452 } 453 crit_exit(); 454 455 /* 456 * Queue message on interface, update output statistics if 457 * successful, and start output if interface not yet active. 458 */ 459 error = ifq_dispatch(ifp, m, &pktattr); 460 return (error); 461 } 462 463 /* 464 * ipfw processing for ethernet packets (in and out). 465 * The second parameter is NULL from ether_demux(), and ifp from 466 * ether_output_frame(). 467 */ 468 static boolean_t 469 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 470 const struct ether_header *eh) 471 { 472 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 473 struct ip_fw_args args; 474 struct m_tag *mtag; 475 struct mbuf *m; 476 int i; 477 478 if (*rule != NULL && fw_one_pass) 479 return TRUE; /* dummynet packet, already partially processed */ 480 481 /* 482 * I need some amount of data to be contiguous. 483 */ 484 i = min((*m0)->m_pkthdr.len, max_protohdr); 485 if ((*m0)->m_len < i) { 486 *m0 = m_pullup(*m0, i); 487 if (*m0 == NULL) 488 return FALSE; 489 } 490 491 /* 492 * Clean up tags 493 */ 494 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 495 m_tag_delete(*m0, mtag); 496 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 497 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 498 KKASSERT(mtag != NULL); 499 m_tag_delete(*m0, mtag); 500 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 501 } 502 503 args.m = *m0; /* the packet we are looking at */ 504 args.oif = dst; /* destination, if any */ 505 args.rule = *rule; /* matching rule to restart */ 506 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 507 i = ip_fw_chk_ptr(&args); 508 *m0 = args.m; 509 *rule = args.rule; 510 511 if (*m0 == NULL) 512 return FALSE; 513 514 switch (i) { 515 case IP_FW_PASS: 516 return TRUE; 517 518 case IP_FW_DIVERT: 519 case IP_FW_TEE: 520 case IP_FW_DENY: 521 /* 522 * XXX at some point add support for divert/forward actions. 523 * If none of the above matches, we have to drop the pkt. 524 */ 525 return FALSE; 526 527 case IP_FW_DUMMYNET: 528 /* 529 * Pass the pkt to dummynet, which consumes it. 530 */ 531 m = *m0; /* pass the original to dummynet */ 532 *m0 = NULL; /* and nothing back to the caller */ 533 534 ether_restore_header(&m, eh, &save_eh); 535 if (m == NULL) 536 return FALSE; 537 538 ip_fw_dn_io_ptr(m, args.cookie, 539 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 540 ip_dn_queue(m); 541 return FALSE; 542 543 default: 544 panic("unknown ipfw return value: %d", i); 545 } 546 } 547 548 /* 549 * Perform common duties while attaching to interface list 550 */ 551 void 552 ether_ifattach(struct ifnet *ifp, const uint8_t *lla, 553 lwkt_serialize_t serializer) 554 { 555 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 556 serializer); 557 } 558 559 void 560 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla, 561 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer) 562 { 563 struct sockaddr_dl *sdl; 564 char ethstr[ETHER_ADDRSTRLEN + 1]; 565 struct ifaltq *ifq; 566 int i; 567 568 /* 569 * If driver does not configure # of mbuf clusters/jclusters 570 * that could sit on the device queues for quite some time, 571 * we then assume: 572 * - The device queues only consume mbuf clusters. 573 * - No more than ether_nmbclusters_default (by default 256) 574 * mbuf clusters will sit on the device queues for quite 575 * some time. 576 */ 577 if (ifp->if_nmbclusters <= 0 && ifp->if_nmbjclusters <= 0) { 578 if (ether_nmbclusters_default < ETHER_NMBCLUSTERS_DEFMIN) { 579 kprintf("ether nmbclusters %d -> %d\n", 580 ether_nmbclusters_default, 581 ETHER_NMBCLUSTERS_DEFAULT); 582 ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT; 583 } 584 ifp->if_nmbclusters = ether_nmbclusters_default; 585 } 586 587 ifp->if_type = IFT_ETHER; 588 ifp->if_addrlen = ETHER_ADDR_LEN; 589 ifp->if_hdrlen = ETHER_HDR_LEN; 590 if_attach(ifp, serializer); 591 ifq = &ifp->if_snd; 592 for (i = 0; i < ifq->altq_subq_cnt; ++i) { 593 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i); 594 595 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen * 596 (ETHER_MAX_LEN - ETHER_CRC_LEN); 597 } 598 ifp->if_mtu = ETHERMTU; 599 if (ifp->if_tsolen <= 0) { 600 if ((ether_tsolen_default / ETHERMTU) < 2) { 601 kprintf("ether TSO maxlen %d -> %d\n", 602 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 603 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 604 } 605 ifp->if_tsolen = ether_tsolen_default; 606 } 607 if (ifp->if_baudrate == 0) 608 ifp->if_baudrate = 10000000; 609 ifp->if_output = ether_output; 610 ifp->if_input = ether_input; 611 ifp->if_resolvemulti = ether_resolvemulti; 612 ifp->if_broadcastaddr = etherbroadcastaddr; 613 sdl = IF_LLSOCKADDR(ifp); 614 sdl->sdl_type = IFT_ETHER; 615 sdl->sdl_alen = ifp->if_addrlen; 616 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 617 /* 618 * XXX Keep the current drivers happy. 619 * XXX Remove once all drivers have been cleaned up 620 */ 621 if (lla != IFP2AC(ifp)->ac_enaddr) 622 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 623 bpfattach(ifp, dlt, hdrlen); 624 if (ng_ether_attach_p != NULL) 625 (*ng_ether_attach_p)(ifp); 626 627 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 628 } 629 630 /* 631 * Perform common duties while detaching an Ethernet interface 632 */ 633 void 634 ether_ifdetach(struct ifnet *ifp) 635 { 636 if_down(ifp); 637 638 if (ng_ether_detach_p != NULL) 639 (*ng_ether_detach_p)(ifp); 640 bpfdetach(ifp); 641 if_detach(ifp); 642 } 643 644 int 645 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 646 { 647 struct ifaddr *ifa = (struct ifaddr *) data; 648 struct ifreq *ifr = (struct ifreq *) data; 649 int error = 0; 650 651 #define IF_INIT(ifp) \ 652 do { \ 653 if (((ifp)->if_flags & IFF_UP) == 0) { \ 654 (ifp)->if_flags |= IFF_UP; \ 655 (ifp)->if_init((ifp)->if_softc); \ 656 } \ 657 } while (0) 658 659 ASSERT_IFNET_SERIALIZED_ALL(ifp); 660 661 switch (command) { 662 case SIOCSIFADDR: 663 switch (ifa->ifa_addr->sa_family) { 664 #ifdef INET 665 case AF_INET: 666 IF_INIT(ifp); /* before arpwhohas */ 667 arp_ifinit(ifp, ifa); 668 break; 669 #endif 670 default: 671 IF_INIT(ifp); 672 break; 673 } 674 break; 675 676 case SIOCGIFADDR: 677 bcopy(IFP2AC(ifp)->ac_enaddr, 678 ((struct sockaddr *)ifr->ifr_data)->sa_data, 679 ETHER_ADDR_LEN); 680 break; 681 682 case SIOCSIFMTU: 683 /* 684 * Set the interface MTU. 685 */ 686 if (ifr->ifr_mtu > ETHERMTU) { 687 error = EINVAL; 688 } else { 689 ifp->if_mtu = ifr->ifr_mtu; 690 } 691 break; 692 default: 693 error = EINVAL; 694 break; 695 } 696 return (error); 697 698 #undef IF_INIT 699 } 700 701 static int 702 ether_resolvemulti( 703 struct ifnet *ifp, 704 struct sockaddr **llsa, 705 struct sockaddr *sa) 706 { 707 struct sockaddr_dl *sdl; 708 #ifdef INET 709 struct sockaddr_in *sin; 710 #endif 711 #ifdef INET6 712 struct sockaddr_in6 *sin6; 713 #endif 714 u_char *e_addr; 715 716 switch(sa->sa_family) { 717 case AF_LINK: 718 /* 719 * No mapping needed. Just check that it's a valid MC address. 720 */ 721 sdl = (struct sockaddr_dl *)sa; 722 e_addr = LLADDR(sdl); 723 if ((e_addr[0] & 1) != 1) 724 return EADDRNOTAVAIL; 725 *llsa = NULL; 726 return 0; 727 728 #ifdef INET 729 case AF_INET: 730 sin = (struct sockaddr_in *)sa; 731 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 732 return EADDRNOTAVAIL; 733 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 734 sdl->sdl_len = sizeof *sdl; 735 sdl->sdl_family = AF_LINK; 736 sdl->sdl_index = ifp->if_index; 737 sdl->sdl_type = IFT_ETHER; 738 sdl->sdl_alen = ETHER_ADDR_LEN; 739 e_addr = LLADDR(sdl); 740 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 741 *llsa = (struct sockaddr *)sdl; 742 return 0; 743 #endif 744 #ifdef INET6 745 case AF_INET6: 746 sin6 = (struct sockaddr_in6 *)sa; 747 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 748 /* 749 * An IP6 address of 0 means listen to all 750 * of the Ethernet multicast address used for IP6. 751 * (This is used for multicast routers.) 752 */ 753 ifp->if_flags |= IFF_ALLMULTI; 754 *llsa = NULL; 755 return 0; 756 } 757 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 758 return EADDRNOTAVAIL; 759 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 760 sdl->sdl_len = sizeof *sdl; 761 sdl->sdl_family = AF_LINK; 762 sdl->sdl_index = ifp->if_index; 763 sdl->sdl_type = IFT_ETHER; 764 sdl->sdl_alen = ETHER_ADDR_LEN; 765 e_addr = LLADDR(sdl); 766 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 767 *llsa = (struct sockaddr *)sdl; 768 return 0; 769 #endif 770 771 default: 772 /* 773 * Well, the text isn't quite right, but it's the name 774 * that counts... 775 */ 776 return EAFNOSUPPORT; 777 } 778 } 779 780 #if 0 781 /* 782 * This is for reference. We have a table-driven version 783 * of the little-endian crc32 generator, which is faster 784 * than the double-loop. 785 */ 786 uint32_t 787 ether_crc32_le(const uint8_t *buf, size_t len) 788 { 789 uint32_t c, crc, carry; 790 size_t i, j; 791 792 crc = 0xffffffffU; /* initial value */ 793 794 for (i = 0; i < len; i++) { 795 c = buf[i]; 796 for (j = 0; j < 8; j++) { 797 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 798 crc >>= 1; 799 c >>= 1; 800 if (carry) 801 crc = (crc ^ ETHER_CRC_POLY_LE); 802 } 803 } 804 805 return (crc); 806 } 807 #else 808 uint32_t 809 ether_crc32_le(const uint8_t *buf, size_t len) 810 { 811 static const uint32_t crctab[] = { 812 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 813 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 814 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 815 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 816 }; 817 uint32_t crc; 818 size_t i; 819 820 crc = 0xffffffffU; /* initial value */ 821 822 for (i = 0; i < len; i++) { 823 crc ^= buf[i]; 824 crc = (crc >> 4) ^ crctab[crc & 0xf]; 825 crc = (crc >> 4) ^ crctab[crc & 0xf]; 826 } 827 828 return (crc); 829 } 830 #endif 831 832 uint32_t 833 ether_crc32_be(const uint8_t *buf, size_t len) 834 { 835 uint32_t c, crc, carry; 836 size_t i, j; 837 838 crc = 0xffffffffU; /* initial value */ 839 840 for (i = 0; i < len; i++) { 841 c = buf[i]; 842 for (j = 0; j < 8; j++) { 843 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 844 crc <<= 1; 845 c >>= 1; 846 if (carry) 847 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 848 } 849 } 850 851 return (crc); 852 } 853 854 /* 855 * find the size of ethernet header, and call classifier 856 */ 857 void 858 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 859 struct altq_pktattr *pktattr) 860 { 861 struct ether_header *eh; 862 uint16_t ether_type; 863 int hlen, af, hdrsize; 864 865 hlen = sizeof(struct ether_header); 866 eh = mtod(m, struct ether_header *); 867 868 ether_type = ntohs(eh->ether_type); 869 if (ether_type < ETHERMTU) { 870 /* ick! LLC/SNAP */ 871 struct llc *llc = (struct llc *)(eh + 1); 872 hlen += 8; 873 874 if (m->m_len < hlen || 875 llc->llc_dsap != LLC_SNAP_LSAP || 876 llc->llc_ssap != LLC_SNAP_LSAP || 877 llc->llc_control != LLC_UI) 878 goto bad; /* not snap! */ 879 880 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 881 } 882 883 if (ether_type == ETHERTYPE_IP) { 884 af = AF_INET; 885 hdrsize = 20; /* sizeof(struct ip) */ 886 #ifdef INET6 887 } else if (ether_type == ETHERTYPE_IPV6) { 888 af = AF_INET6; 889 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 890 #endif 891 } else 892 goto bad; 893 894 while (m->m_len <= hlen) { 895 hlen -= m->m_len; 896 m = m->m_next; 897 } 898 if (m->m_len < hlen + hdrsize) { 899 /* 900 * ip header is not in a single mbuf. this should not 901 * happen in the current code. 902 * (todo: use m_pulldown in the future) 903 */ 904 goto bad; 905 } 906 m->m_data += hlen; 907 m->m_len -= hlen; 908 ifq_classify(ifq, m, af, pktattr); 909 m->m_data -= hlen; 910 m->m_len += hlen; 911 912 return; 913 914 bad: 915 pktattr->pattr_class = NULL; 916 pktattr->pattr_hdr = NULL; 917 pktattr->pattr_af = AF_UNSPEC; 918 } 919 920 static void 921 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 922 const struct ether_header *save_eh) 923 { 924 struct mbuf *m = *m0; 925 926 ether_restore_hdr++; 927 928 /* 929 * Prepend the header, optimize for the common case of 930 * eh pointing into the mbuf. 931 */ 932 if ((const void *)(eh + 1) == (void *)m->m_data) { 933 m->m_data -= ETHER_HDR_LEN; 934 m->m_len += ETHER_HDR_LEN; 935 m->m_pkthdr.len += ETHER_HDR_LEN; 936 } else { 937 ether_prepend_hdr++; 938 939 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 940 if (m != NULL) { 941 bcopy(save_eh, mtod(m, struct ether_header *), 942 ETHER_HDR_LEN); 943 } 944 } 945 *m0 = m; 946 } 947 948 /* 949 * Upper layer processing for a received Ethernet packet. 950 */ 951 void 952 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 953 { 954 struct ether_header *eh; 955 int isr, discard = 0; 956 u_short ether_type; 957 struct ip_fw *rule = NULL; 958 959 M_ASSERTPKTHDR(m); 960 KASSERT(m->m_len >= ETHER_HDR_LEN, 961 ("ether header is not contiguous!")); 962 963 eh = mtod(m, struct ether_header *); 964 965 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 966 struct m_tag *mtag; 967 968 /* Extract info from dummynet tag */ 969 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 970 KKASSERT(mtag != NULL); 971 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 972 KKASSERT(rule != NULL); 973 974 m_tag_delete(m, mtag); 975 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 976 977 /* packet is passing the second time */ 978 goto post_stats; 979 } 980 981 /* 982 * We got a packet which was unicast to a different Ethernet 983 * address. If the driver is working properly, then this 984 * situation can only happen when the interface is in 985 * promiscuous mode. We defer the packet discarding until the 986 * vlan processing is done, so that vlan/bridge or vlan/netgraph 987 * could work. 988 */ 989 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 990 !ETHER_IS_MULTICAST(eh->ether_dhost) && 991 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 992 if (ether_debug & 1) { 993 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 994 "%02x:%02x:%02x:%02x:%02x:%02x " 995 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 996 eh->ether_dhost[0], 997 eh->ether_dhost[1], 998 eh->ether_dhost[2], 999 eh->ether_dhost[3], 1000 eh->ether_dhost[4], 1001 eh->ether_dhost[5], 1002 eh->ether_shost[0], 1003 eh->ether_shost[1], 1004 eh->ether_shost[2], 1005 eh->ether_shost[3], 1006 eh->ether_shost[4], 1007 eh->ether_shost[5], 1008 eh->ether_type, 1009 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1010 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1011 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1012 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1013 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1014 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1015 ); 1016 } 1017 if ((ether_debug & 2) == 0) 1018 discard = 1; 1019 } 1020 1021 post_stats: 1022 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0 && !discard) { 1023 struct ether_header save_eh = *eh; 1024 1025 /* XXX old crufty stuff, needs to be removed */ 1026 m_adj(m, sizeof(struct ether_header)); 1027 1028 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1029 m_freem(m); 1030 return; 1031 } 1032 1033 ether_restore_header(&m, eh, &save_eh); 1034 if (m == NULL) 1035 return; 1036 eh = mtod(m, struct ether_header *); 1037 } 1038 1039 ether_type = ntohs(eh->ether_type); 1040 KKASSERT(ether_type != ETHERTYPE_VLAN); 1041 1042 /* Handle input from a lagg(4) port */ 1043 if (ifp->if_type == IFT_IEEE8023ADLAG) { 1044 KASSERT(lagg_input_p != NULL, 1045 ("%s: if_lagg not loaded!", __func__)); 1046 (*lagg_input_p)(ifp, m); 1047 return; 1048 } 1049 1050 if (m->m_flags & M_VLANTAG) { 1051 void (*vlan_input_func)(struct mbuf *); 1052 1053 vlan_input_func = vlan_input_p; 1054 if (vlan_input_func != NULL) { 1055 vlan_input_func(m); 1056 } else { 1057 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1058 m_freem(m); 1059 } 1060 return; 1061 } 1062 1063 /* 1064 * If we have been asked to discard this packet 1065 * (e.g. not for us), drop it before entering 1066 * the upper layer. 1067 */ 1068 if (discard) { 1069 m_freem(m); 1070 return; 1071 } 1072 1073 /* 1074 * Clear protocol specific flags, 1075 * before entering the upper layer. 1076 */ 1077 m->m_flags &= ~M_ETHER_FLAGS; 1078 1079 /* Strip ethernet header. */ 1080 m_adj(m, sizeof(struct ether_header)); 1081 1082 switch (ether_type) { 1083 #ifdef INET 1084 case ETHERTYPE_IP: 1085 if ((m->m_flags & M_LENCHECKED) == 0) { 1086 if (!ip_lengthcheck(&m, 0)) 1087 return; 1088 } 1089 if (ipflow_fastforward(m)) 1090 return; 1091 isr = NETISR_IP; 1092 break; 1093 1094 case ETHERTYPE_ARP: 1095 if (ifp->if_flags & IFF_NOARP) { 1096 /* Discard packet if ARP is disabled on interface */ 1097 m_freem(m); 1098 return; 1099 } 1100 isr = NETISR_ARP; 1101 break; 1102 #endif 1103 1104 #ifdef INET6 1105 case ETHERTYPE_IPV6: 1106 isr = NETISR_IPV6; 1107 break; 1108 #endif 1109 1110 #ifdef MPLS 1111 case ETHERTYPE_MPLS: 1112 case ETHERTYPE_MPLS_MCAST: 1113 /* Should have been set by ether_input(). */ 1114 KKASSERT(m->m_flags & M_MPLSLABELED); 1115 isr = NETISR_MPLS; 1116 break; 1117 #endif 1118 1119 default: 1120 /* 1121 * The accurate msgport is not determined before 1122 * we reach here, so recharacterize packet. 1123 */ 1124 m->m_flags &= ~M_HASH; 1125 if (ng_ether_input_orphan_p != NULL) { 1126 /* 1127 * Put back the ethernet header so netgraph has a 1128 * consistent view of inbound packets. 1129 */ 1130 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 1131 if (m == NULL) { 1132 /* 1133 * M_PREPEND frees the mbuf in case of failure. 1134 */ 1135 return; 1136 } 1137 /* 1138 * Hold BGL and recheck ng_ether_input_orphan_p 1139 */ 1140 get_mplock(); 1141 if (ng_ether_input_orphan_p != NULL) { 1142 ng_ether_input_orphan_p(ifp, m); 1143 rel_mplock(); 1144 return; 1145 } 1146 rel_mplock(); 1147 } 1148 m_freem(m); 1149 return; 1150 } 1151 1152 if (m->m_flags & M_HASH) { 1153 if (&curthread->td_msgport == 1154 netisr_hashport(m->m_pkthdr.hash)) { 1155 netisr_handle(isr, m); 1156 return; 1157 } else { 1158 /* 1159 * XXX Something is wrong, 1160 * we probably should panic here! 1161 */ 1162 m->m_flags &= ~M_HASH; 1163 atomic_add_long(ðer_input_wronghash, 1); 1164 } 1165 } 1166 #ifdef RSS_DEBUG 1167 atomic_add_long(ðer_input_requeue, 1); 1168 #endif 1169 netisr_queue(isr, m); 1170 } 1171 1172 /* 1173 * First we perform any link layer operations, then continue to the 1174 * upper layers with ether_demux_oncpu(). 1175 */ 1176 static void 1177 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1178 { 1179 #ifdef CARP 1180 void *carp; 1181 #endif 1182 1183 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1184 /* 1185 * Receiving interface's flags are changed, when this 1186 * packet is waiting for processing; discard it. 1187 */ 1188 m_freem(m); 1189 return; 1190 } 1191 1192 /* 1193 * Tap the packet off here for a bridge. bridge_input() 1194 * will return NULL if it has consumed the packet, otherwise 1195 * it gets processed as normal. Note that bridge_input() 1196 * will always return the original packet if we need to 1197 * process it locally. 1198 */ 1199 if (ifp->if_bridge) { 1200 KASSERT(bridge_input_p != NULL, 1201 ("%s: if_bridge not loaded!", __func__)); 1202 1203 if(m->m_flags & M_ETHER_BRIDGED) { 1204 m->m_flags &= ~M_ETHER_BRIDGED; 1205 } else { 1206 m = bridge_input_p(ifp, m); 1207 if (m == NULL) 1208 return; 1209 1210 KASSERT(ifp == m->m_pkthdr.rcvif, 1211 ("bridge_input_p changed rcvif")); 1212 } 1213 } 1214 1215 #ifdef CARP 1216 carp = ifp->if_carp; 1217 if (carp) { 1218 m = carp_input(carp, m); 1219 if (m == NULL) 1220 return; 1221 KASSERT(ifp == m->m_pkthdr.rcvif, 1222 ("carp_input changed rcvif")); 1223 } 1224 #endif 1225 1226 /* Handle ng_ether(4) processing, if any */ 1227 if (ng_ether_input_p != NULL) { 1228 /* 1229 * Hold BGL and recheck ng_ether_input_p 1230 */ 1231 get_mplock(); 1232 if (ng_ether_input_p != NULL) 1233 ng_ether_input_p(ifp, &m); 1234 rel_mplock(); 1235 1236 if (m == NULL) 1237 return; 1238 } 1239 1240 /* Continue with upper layer processing */ 1241 ether_demux_oncpu(ifp, m); 1242 } 1243 1244 /* 1245 * Perform certain functions of ether_input(): 1246 * - Test IFF_UP 1247 * - Update statistics 1248 * - Run bpf(4) tap if requested 1249 * Then pass the packet to ether_input_oncpu(). 1250 * 1251 * This function should be used by pseudo interface (e.g. vlan(4)), 1252 * when it tries to claim that the packet is received by it. 1253 * 1254 * REINPUT_KEEPRCVIF 1255 * REINPUT_RUNBPF 1256 */ 1257 void 1258 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1259 { 1260 /* Discard packet if interface is not up */ 1261 if (!(ifp->if_flags & IFF_UP)) { 1262 m_freem(m); 1263 return; 1264 } 1265 1266 /* 1267 * Change receiving interface. The bridge will often pass a flag to 1268 * ask that this not be done so ARPs get applied to the correct 1269 * side. 1270 */ 1271 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1272 m->m_pkthdr.rcvif == NULL) { 1273 m->m_pkthdr.rcvif = ifp; 1274 } 1275 1276 /* Update statistics */ 1277 IFNET_STAT_INC(ifp, ipackets, 1); 1278 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1279 if (m->m_flags & (M_MCAST | M_BCAST)) 1280 IFNET_STAT_INC(ifp, imcasts, 1); 1281 1282 if (reinput_flags & REINPUT_RUNBPF) 1283 BPF_MTAP(ifp, m); 1284 1285 ether_input_oncpu(ifp, m); 1286 } 1287 1288 static __inline boolean_t 1289 ether_vlancheck(struct mbuf **m0) 1290 { 1291 struct mbuf *m = *m0; 1292 struct ether_header *eh; 1293 uint16_t ether_type; 1294 1295 eh = mtod(m, struct ether_header *); 1296 ether_type = ntohs(eh->ether_type); 1297 1298 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1299 /* 1300 * Extract vlan tag if hardware does not do it for us 1301 */ 1302 vlan_ether_decap(&m); 1303 if (m == NULL) 1304 goto failed; 1305 1306 eh = mtod(m, struct ether_header *); 1307 ether_type = ntohs(eh->ether_type); 1308 } 1309 1310 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1311 /* 1312 * To prevent possible dangerous recursion, 1313 * we don't do vlan-in-vlan 1314 */ 1315 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1316 goto failed; 1317 } 1318 KKASSERT(ether_type != ETHERTYPE_VLAN); 1319 1320 m->m_flags |= M_ETHER_VLANCHECKED; 1321 *m0 = m; 1322 return TRUE; 1323 failed: 1324 if (m != NULL) 1325 m_freem(m); 1326 *m0 = NULL; 1327 return FALSE; 1328 } 1329 1330 static void 1331 ether_input_handler(netmsg_t nmsg) 1332 { 1333 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1334 struct ether_header *eh; 1335 struct ifnet *ifp; 1336 struct mbuf *m; 1337 1338 m = nmp->nm_packet; 1339 M_ASSERTPKTHDR(m); 1340 1341 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1342 if (!ether_vlancheck(&m)) { 1343 KKASSERT(m == NULL); 1344 return; 1345 } 1346 } 1347 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) || 1348 __predict_false(ether_input_ckhash)) { 1349 int isr; 1350 1351 /* 1352 * Need to verify the hash supplied by the hardware 1353 * which could be wrong. 1354 */ 1355 m->m_flags &= ~(M_HASH | M_CKHASH); 1356 isr = ether_characterize(&m); 1357 if (m == NULL) 1358 return; 1359 KKASSERT(m->m_flags & M_HASH); 1360 1361 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) { 1362 /* 1363 * Wrong hardware supplied hash; redispatch 1364 */ 1365 ether_dispatch(isr, m, -1); 1366 if (__predict_false(ether_input_ckhash)) 1367 atomic_add_long(ðer_input_wronghwhash, 1); 1368 return; 1369 } 1370 } 1371 ifp = m->m_pkthdr.rcvif; 1372 1373 eh = mtod(m, struct ether_header *); 1374 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1375 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1376 ifp->if_addrlen) == 0) 1377 m->m_flags |= M_BCAST; 1378 else 1379 m->m_flags |= M_MCAST; 1380 IFNET_STAT_INC(ifp, imcasts, 1); 1381 } 1382 1383 ether_input_oncpu(ifp, m); 1384 } 1385 1386 /* 1387 * Send the packet to the target netisr msgport 1388 * 1389 * At this point the packet must be characterized (M_HASH set), 1390 * so we know which netisr to send it to. 1391 */ 1392 static void 1393 ether_dispatch(int isr, struct mbuf *m, int cpuid) 1394 { 1395 struct netmsg_packet *pmsg; 1396 int target_cpuid; 1397 1398 KKASSERT(m->m_flags & M_HASH); 1399 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash); 1400 1401 pmsg = &m->m_hdr.mh_netmsg; 1402 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1403 0, ether_input_handler); 1404 pmsg->nm_packet = m; 1405 pmsg->base.lmsg.u.ms_result = isr; 1406 1407 logether(disp_beg, NULL); 1408 if (target_cpuid == cpuid) { 1409 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid), 1410 &pmsg->base.lmsg); 1411 } else { 1412 lwkt_sendmsg(netisr_cpuport(target_cpuid), 1413 &pmsg->base.lmsg); 1414 } 1415 logether(disp_end, NULL); 1416 } 1417 1418 /* 1419 * Process a received Ethernet packet. 1420 * 1421 * The ethernet header is assumed to be in the mbuf so the caller 1422 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1423 * bytes in the first mbuf. 1424 * 1425 * If the caller knows that the current thread is stick to the current 1426 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid 1427 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should 1428 * be passed as 'cpuid' argument. 1429 */ 1430 void 1431 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1432 int cpuid) 1433 { 1434 int isr; 1435 1436 M_ASSERTPKTHDR(m); 1437 1438 /* Discard packet if interface is not up */ 1439 if (!(ifp->if_flags & IFF_UP)) { 1440 m_freem(m); 1441 return; 1442 } 1443 1444 if (m->m_len < sizeof(struct ether_header)) { 1445 /* XXX error in the caller. */ 1446 m_freem(m); 1447 return; 1448 } 1449 1450 m->m_pkthdr.rcvif = ifp; 1451 1452 logether(pkt_beg, ifp); 1453 1454 ETHER_BPF_MTAP(ifp, m); 1455 1456 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1457 1458 if (ifp->if_flags & IFF_MONITOR) { 1459 struct ether_header *eh; 1460 1461 eh = mtod(m, struct ether_header *); 1462 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1463 IFNET_STAT_INC(ifp, imcasts, 1); 1464 1465 /* 1466 * Interface marked for monitoring; discard packet. 1467 */ 1468 m_freem(m); 1469 1470 logether(pkt_end, ifp); 1471 return; 1472 } 1473 1474 /* 1475 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1476 * we can dispatch it immediately with trivial checks. 1477 */ 1478 if (pi != NULL && (m->m_flags & M_HASH)) { 1479 #ifdef RSS_DEBUG 1480 atomic_add_long(ðer_pktinfo_try, 1); 1481 #endif 1482 netisr_hashcheck(pi->pi_netisr, m, pi); 1483 if (m->m_flags & M_HASH) { 1484 ether_dispatch(pi->pi_netisr, m, cpuid); 1485 #ifdef RSS_DEBUG 1486 atomic_add_long(ðer_pktinfo_hit, 1); 1487 #endif 1488 logether(pkt_end, ifp); 1489 return; 1490 } 1491 } 1492 #ifdef RSS_DEBUG 1493 else if (ifp->if_capenable & IFCAP_RSS) { 1494 if (pi == NULL) 1495 atomic_add_long(ðer_rss_nopi, 1); 1496 else 1497 atomic_add_long(ðer_rss_nohash, 1); 1498 } 1499 #endif 1500 1501 /* 1502 * Packet hash will be recalculated by software, so clear 1503 * the M_HASH and M_CKHASH flag set by the driver; the hash 1504 * value calculated by the hardware may not be exactly what 1505 * we want. 1506 */ 1507 m->m_flags &= ~(M_HASH | M_CKHASH); 1508 1509 if (!ether_vlancheck(&m)) { 1510 KKASSERT(m == NULL); 1511 logether(pkt_end, ifp); 1512 return; 1513 } 1514 1515 isr = ether_characterize(&m); 1516 if (m == NULL) { 1517 logether(pkt_end, ifp); 1518 return; 1519 } 1520 1521 /* 1522 * Finally dispatch it 1523 */ 1524 ether_dispatch(isr, m, cpuid); 1525 1526 logether(pkt_end, ifp); 1527 } 1528 1529 static int 1530 ether_characterize(struct mbuf **m0) 1531 { 1532 struct mbuf *m = *m0; 1533 struct ether_header *eh; 1534 uint16_t ether_type; 1535 int isr; 1536 1537 eh = mtod(m, struct ether_header *); 1538 ether_type = ntohs(eh->ether_type); 1539 1540 /* 1541 * Map ether type to netisr id. 1542 */ 1543 switch (ether_type) { 1544 #ifdef INET 1545 case ETHERTYPE_IP: 1546 isr = NETISR_IP; 1547 break; 1548 1549 case ETHERTYPE_ARP: 1550 isr = NETISR_ARP; 1551 break; 1552 #endif 1553 1554 #ifdef INET6 1555 case ETHERTYPE_IPV6: 1556 isr = NETISR_IPV6; 1557 break; 1558 #endif 1559 1560 #ifdef MPLS 1561 case ETHERTYPE_MPLS: 1562 case ETHERTYPE_MPLS_MCAST: 1563 m->m_flags |= M_MPLSLABELED; 1564 isr = NETISR_MPLS; 1565 break; 1566 #endif 1567 1568 default: 1569 /* 1570 * NETISR_MAX is an invalid value; it is chosen to let 1571 * netisr_characterize() know that we have no clear 1572 * idea where this packet should go. 1573 */ 1574 isr = NETISR_MAX; 1575 break; 1576 } 1577 1578 /* 1579 * Ask the isr to characterize the packet since we couldn't. 1580 * This is an attempt to optimally get us onto the correct protocol 1581 * thread. 1582 */ 1583 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1584 1585 *m0 = m; 1586 return isr; 1587 } 1588 1589 static void 1590 ether_demux_handler(netmsg_t nmsg) 1591 { 1592 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1593 struct ifnet *ifp; 1594 struct mbuf *m; 1595 1596 m = nmp->nm_packet; 1597 M_ASSERTPKTHDR(m); 1598 ifp = m->m_pkthdr.rcvif; 1599 1600 ether_demux_oncpu(ifp, m); 1601 } 1602 1603 void 1604 ether_demux(struct mbuf *m) 1605 { 1606 struct netmsg_packet *pmsg; 1607 int isr; 1608 1609 isr = ether_characterize(&m); 1610 if (m == NULL) 1611 return; 1612 1613 KKASSERT(m->m_flags & M_HASH); 1614 pmsg = &m->m_hdr.mh_netmsg; 1615 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1616 0, ether_demux_handler); 1617 pmsg->nm_packet = m; 1618 pmsg->base.lmsg.u.ms_result = isr; 1619 1620 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1621 } 1622 1623 u_char * 1624 kether_aton(const char *macstr, u_char *addr) 1625 { 1626 unsigned int o0, o1, o2, o3, o4, o5; 1627 int n; 1628 1629 if (macstr == NULL || addr == NULL) 1630 return NULL; 1631 1632 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1633 &o3, &o4, &o5); 1634 if (n != 6) 1635 return NULL; 1636 1637 addr[0] = o0; 1638 addr[1] = o1; 1639 addr[2] = o2; 1640 addr[3] = o3; 1641 addr[4] = o4; 1642 addr[5] = o5; 1643 1644 return addr; 1645 } 1646 1647 char * 1648 kether_ntoa(const u_char *addr, char *buf) 1649 { 1650 int len = ETHER_ADDRSTRLEN + 1; 1651 int n; 1652 1653 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1654 addr[1], addr[2], addr[3], addr[4], addr[5]); 1655 1656 if (n < 17) 1657 return NULL; 1658 1659 return buf; 1660 } 1661 1662 MODULE_VERSION(ether, 1); 1663