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(struct ifnet *, 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_NETISR_NCPUS(mycpuid); 230 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 231 232 if (ifp->if_flags & IFF_MONITOR) 233 gotoerr(ENETDOWN); 234 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING)) 235 gotoerr(ENETDOWN); 236 237 M_PREPEND(m, sizeof(struct ether_header), M_NOWAIT); 238 if (m == NULL) 239 return (ENOBUFS); 240 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header); 241 eh = mtod(m, struct ether_header *); 242 edst = eh->ether_dhost; 243 244 /* 245 * Fill in the destination ethernet address and frame type. 246 */ 247 switch (dst->sa_family) { 248 #ifdef INET 249 case AF_INET: 250 error = arpresolve(ifp, rt, m, dst, edst); 251 if (error != 0) 252 return error == EWOULDBLOCK ? 0 : error; 253 #ifdef MPLS 254 if (m->m_flags & M_MPLSLABELED) 255 eh->ether_type = htons(ETHERTYPE_MPLS); 256 else 257 #endif 258 eh->ether_type = htons(ETHERTYPE_IP); 259 break; 260 #endif 261 #ifdef INET6 262 case AF_INET6: 263 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 264 return (0); /* Something bad happenned. */ 265 eh->ether_type = htons(ETHERTYPE_IPV6); 266 break; 267 #endif 268 case pseudo_AF_HDRCMPLT: 269 case AF_UNSPEC: 270 loop_copy = -1; /* if this is for us, don't do it */ 271 deh = (struct ether_header *)dst->sa_data; 272 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 273 eh->ether_type = deh->ether_type; 274 break; 275 276 default: 277 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 278 gotoerr(EAFNOSUPPORT); 279 } 280 281 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 282 memcpy(eh->ether_shost, 283 ((struct ether_header *)dst->sa_data)->ether_shost, 284 ETHER_ADDR_LEN); 285 else 286 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 287 288 /* 289 * Bridges require special output handling. 290 */ 291 if (ifp->if_bridge) { 292 KASSERT(bridge_output_p != NULL, 293 ("%s: if_bridge not loaded!", __func__)); 294 return bridge_output_p(ifp, m); 295 } 296 #if 0 /* XXX */ 297 if (ifp->if_lagg) { 298 KASSERT(lagg_output_p != NULL, 299 ("%s: if_lagg not loaded!", __func__)); 300 return lagg_output_p(ifp, m); 301 } 302 #endif 303 304 /* 305 * If a simplex interface, and the packet is being sent to our 306 * Ethernet address or a broadcast address, loopback a copy. 307 * XXX To make a simplex device behave exactly like a duplex 308 * device, we should copy in the case of sending to our own 309 * ethernet address (thus letting the original actually appear 310 * on the wire). However, we don't do that here for security 311 * reasons and compatibility with the original behavior. 312 */ 313 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 314 int csum_flags = 0; 315 316 if (m->m_pkthdr.csum_flags & CSUM_IP) 317 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 318 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 319 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 320 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 321 struct mbuf *n; 322 323 if ((n = m_copypacket(m, M_NOWAIT)) != NULL) { 324 n->m_pkthdr.csum_flags |= csum_flags; 325 if (csum_flags & CSUM_DATA_VALID) 326 n->m_pkthdr.csum_data = 0xffff; 327 if_simloop(ifp, n, dst->sa_family, hlen); 328 } else 329 IFNET_STAT_INC(ifp, iqdrops, 1); 330 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 331 ETHER_ADDR_LEN) == 0) { 332 m->m_pkthdr.csum_flags |= csum_flags; 333 if (csum_flags & CSUM_DATA_VALID) 334 m->m_pkthdr.csum_data = 0xffff; 335 if_simloop(ifp, m, dst->sa_family, hlen); 336 return (0); /* XXX */ 337 } 338 } 339 340 #ifdef CARP 341 if (ifp->if_type == IFT_CARP) { 342 ifp = carp_parent(ifp); 343 if (ifp == NULL) 344 gotoerr(ENETUNREACH); 345 346 ac = IFP2AC(ifp); 347 348 /* 349 * Check precondition again 350 */ 351 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 352 353 if (ifp->if_flags & IFF_MONITOR) 354 gotoerr(ENETDOWN); 355 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != 356 (IFF_UP | IFF_RUNNING)) 357 gotoerr(ENETDOWN); 358 } 359 #endif 360 361 /* Handle ng_ether(4) processing, if any */ 362 if (ng_ether_output_p != NULL) { 363 /* 364 * Hold BGL and recheck ng_ether_output_p 365 */ 366 get_mplock(); 367 if (ng_ether_output_p != NULL) { 368 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 369 rel_mplock(); 370 goto bad; 371 } 372 if (m == NULL) { 373 rel_mplock(); 374 return (0); 375 } 376 } 377 rel_mplock(); 378 } 379 380 /* Continue with link-layer output */ 381 return ether_output_frame(ifp, m); 382 383 bad: 384 m_freem(m); 385 return (error); 386 } 387 388 /* 389 * Returns the bridge interface an ifp is associated 390 * with. 391 * 392 * Only call if ifp->if_bridge != NULL. 393 */ 394 struct ifnet * 395 ether_bridge_interface(struct ifnet *ifp) 396 { 397 if (bridge_interface_p) 398 return(bridge_interface_p(ifp->if_bridge)); 399 return (ifp); 400 } 401 402 /* 403 * Ethernet link layer output routine to send a raw frame to the device. 404 * 405 * This assumes that the 14 byte Ethernet header is present and contiguous 406 * in the first mbuf. 407 */ 408 int 409 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 410 { 411 struct ip_fw *rule = NULL; 412 int error = 0; 413 struct altq_pktattr pktattr; 414 415 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 416 417 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 418 struct m_tag *mtag; 419 420 /* Extract info from dummynet tag */ 421 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 422 KKASSERT(mtag != NULL); 423 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 424 KKASSERT(rule != NULL); 425 426 m_tag_delete(m, mtag); 427 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 428 } 429 430 if (ifq_is_enabled(&ifp->if_snd)) 431 altq_etherclassify(&ifp->if_snd, m, &pktattr); 432 crit_enter(); 433 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0) { 434 struct ether_header save_eh, *eh; 435 436 eh = mtod(m, struct ether_header *); 437 save_eh = *eh; 438 m_adj(m, ETHER_HDR_LEN); 439 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 440 crit_exit(); 441 if (m != NULL) { 442 m_freem(m); 443 return ENOBUFS; /* pkt dropped */ 444 } else 445 return 0; /* consumed e.g. in a pipe */ 446 } 447 448 /* packet was ok, restore the ethernet header */ 449 ether_restore_header(&m, eh, &save_eh); 450 if (m == NULL) { 451 crit_exit(); 452 return ENOBUFS; 453 } 454 } 455 crit_exit(); 456 457 /* 458 * Queue message on interface, update output statistics if 459 * successful, and start output if interface not yet active. 460 */ 461 error = ifq_dispatch(ifp, m, &pktattr); 462 return (error); 463 } 464 465 /* 466 * ipfw processing for ethernet packets (in and out). 467 * The second parameter is NULL from ether_demux(), and ifp from 468 * ether_output_frame(). 469 */ 470 static boolean_t 471 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 472 const struct ether_header *eh) 473 { 474 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 475 struct ip_fw_args args; 476 struct m_tag *mtag; 477 struct mbuf *m; 478 int i; 479 480 if (*rule != NULL && fw_one_pass) 481 return TRUE; /* dummynet packet, already partially processed */ 482 483 /* 484 * I need some amount of data to be contiguous. 485 */ 486 i = min((*m0)->m_pkthdr.len, max_protohdr); 487 if ((*m0)->m_len < i) { 488 *m0 = m_pullup(*m0, i); 489 if (*m0 == NULL) 490 return FALSE; 491 } 492 493 /* 494 * Clean up tags 495 */ 496 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 497 m_tag_delete(*m0, mtag); 498 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 499 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 500 KKASSERT(mtag != NULL); 501 m_tag_delete(*m0, mtag); 502 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 503 } 504 505 args.flags = 0; 506 args.xlat = NULL; 507 args.m = *m0; /* the packet we are looking at */ 508 args.oif = dst; /* destination, if any */ 509 args.rule = *rule; /* matching rule to restart */ 510 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 511 i = ip_fw_chk_ptr(&args); 512 *m0 = args.m; 513 *rule = args.rule; 514 515 if (*m0 == NULL) 516 return FALSE; 517 518 switch (i) { 519 case IP_FW_PASS: 520 return TRUE; 521 522 case IP_FW_DIVERT: 523 case IP_FW_TEE: 524 case IP_FW_DENY: 525 /* 526 * XXX at some point add support for divert/forward actions. 527 * If none of the above matches, we have to drop the pkt. 528 */ 529 return FALSE; 530 531 case IP_FW_DUMMYNET: 532 /* 533 * Pass the pkt to dummynet, which consumes it. 534 */ 535 m = *m0; /* pass the original to dummynet */ 536 *m0 = NULL; /* and nothing back to the caller */ 537 538 ether_restore_header(&m, eh, &save_eh); 539 if (m == NULL) 540 return FALSE; 541 542 m = ip_fw_dn_io_ptr(m, args.cookie, 543 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 544 if (m != NULL) 545 ip_dn_queue(m); 546 return FALSE; 547 548 default: 549 panic("unknown ipfw return value: %d", i); 550 } 551 } 552 553 /* 554 * Perform common duties while attaching to interface list 555 */ 556 void 557 ether_ifattach(struct ifnet *ifp, const uint8_t *lla, 558 lwkt_serialize_t serializer) 559 { 560 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 561 serializer); 562 } 563 564 void 565 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla, 566 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer) 567 { 568 struct sockaddr_dl *sdl; 569 char ethstr[ETHER_ADDRSTRLEN + 1]; 570 struct ifaltq *ifq; 571 int i; 572 573 /* 574 * If driver does not configure # of mbuf clusters/jclusters 575 * that could sit on the device queues for quite some time, 576 * we then assume: 577 * - The device queues only consume mbuf clusters. 578 * - No more than ether_nmbclusters_default (by default 256) 579 * mbuf clusters will sit on the device queues for quite 580 * some time. 581 */ 582 if (ifp->if_nmbclusters <= 0 && ifp->if_nmbjclusters <= 0) { 583 if (ether_nmbclusters_default < ETHER_NMBCLUSTERS_DEFMIN) { 584 kprintf("ether nmbclusters %d -> %d\n", 585 ether_nmbclusters_default, 586 ETHER_NMBCLUSTERS_DEFAULT); 587 ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT; 588 } 589 ifp->if_nmbclusters = ether_nmbclusters_default; 590 } 591 592 ifp->if_type = IFT_ETHER; 593 ifp->if_addrlen = ETHER_ADDR_LEN; 594 ifp->if_hdrlen = ETHER_HDR_LEN; 595 if_attach(ifp, serializer); 596 ifq = &ifp->if_snd; 597 for (i = 0; i < ifq->altq_subq_cnt; ++i) { 598 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i); 599 600 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen * 601 (ETHER_MAX_LEN - ETHER_CRC_LEN); 602 } 603 ifp->if_mtu = ETHERMTU; 604 if (ifp->if_tsolen <= 0) { 605 if ((ether_tsolen_default / ETHERMTU) < 2) { 606 kprintf("ether TSO maxlen %d -> %d\n", 607 ether_tsolen_default, ETHER_TSOLEN_DEFAULT); 608 ether_tsolen_default = ETHER_TSOLEN_DEFAULT; 609 } 610 ifp->if_tsolen = ether_tsolen_default; 611 } 612 if (ifp->if_baudrate == 0) 613 ifp->if_baudrate = 10000000; 614 ifp->if_output = ether_output; 615 ifp->if_input = ether_input; 616 ifp->if_resolvemulti = ether_resolvemulti; 617 ifp->if_broadcastaddr = etherbroadcastaddr; 618 sdl = IF_LLSOCKADDR(ifp); 619 sdl->sdl_type = IFT_ETHER; 620 sdl->sdl_alen = ifp->if_addrlen; 621 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 622 /* 623 * XXX Keep the current drivers happy. 624 * XXX Remove once all drivers have been cleaned up 625 */ 626 if (lla != IFP2AC(ifp)->ac_enaddr) 627 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 628 bpfattach(ifp, dlt, hdrlen); 629 if (ng_ether_attach_p != NULL) 630 (*ng_ether_attach_p)(ifp); 631 632 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr)); 633 } 634 635 /* 636 * Perform common duties while detaching an Ethernet interface 637 */ 638 void 639 ether_ifdetach(struct ifnet *ifp) 640 { 641 if_down(ifp); 642 643 if (ng_ether_detach_p != NULL) 644 (*ng_ether_detach_p)(ifp); 645 bpfdetach(ifp); 646 if_detach(ifp); 647 } 648 649 int 650 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data) 651 { 652 struct ifaddr *ifa = (struct ifaddr *) data; 653 struct ifreq *ifr = (struct ifreq *) data; 654 int error = 0; 655 656 #define IF_INIT(ifp) \ 657 do { \ 658 if (((ifp)->if_flags & IFF_UP) == 0) { \ 659 (ifp)->if_flags |= IFF_UP; \ 660 (ifp)->if_init((ifp)->if_softc); \ 661 } \ 662 } while (0) 663 664 ASSERT_IFNET_SERIALIZED_ALL(ifp); 665 666 switch (command) { 667 case SIOCSIFADDR: 668 switch (ifa->ifa_addr->sa_family) { 669 #ifdef INET 670 case AF_INET: 671 IF_INIT(ifp); /* before arpwhohas */ 672 arp_ifinit(ifp, ifa); 673 break; 674 #endif 675 default: 676 IF_INIT(ifp); 677 break; 678 } 679 break; 680 681 case SIOCGIFADDR: 682 bcopy(IFP2AC(ifp)->ac_enaddr, 683 ((struct sockaddr *)ifr->ifr_data)->sa_data, 684 ETHER_ADDR_LEN); 685 break; 686 687 case SIOCSIFMTU: 688 /* 689 * Set the interface MTU. 690 */ 691 if (ifr->ifr_mtu > ETHERMTU) { 692 error = EINVAL; 693 } else { 694 ifp->if_mtu = ifr->ifr_mtu; 695 } 696 break; 697 default: 698 error = EINVAL; 699 break; 700 } 701 return (error); 702 703 #undef IF_INIT 704 } 705 706 static int 707 ether_resolvemulti( 708 struct ifnet *ifp, 709 struct sockaddr **llsa, 710 struct sockaddr *sa) 711 { 712 struct sockaddr_dl *sdl; 713 #ifdef INET 714 struct sockaddr_in *sin; 715 #endif 716 #ifdef INET6 717 struct sockaddr_in6 *sin6; 718 #endif 719 u_char *e_addr; 720 721 switch(sa->sa_family) { 722 case AF_LINK: 723 /* 724 * No mapping needed. Just check that it's a valid MC address. 725 */ 726 sdl = (struct sockaddr_dl *)sa; 727 e_addr = LLADDR(sdl); 728 if ((e_addr[0] & 1) != 1) 729 return EADDRNOTAVAIL; 730 *llsa = NULL; 731 return 0; 732 733 #ifdef INET 734 case AF_INET: 735 sin = (struct sockaddr_in *)sa; 736 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 737 return EADDRNOTAVAIL; 738 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 739 sdl->sdl_len = sizeof *sdl; 740 sdl->sdl_family = AF_LINK; 741 sdl->sdl_index = ifp->if_index; 742 sdl->sdl_type = IFT_ETHER; 743 sdl->sdl_alen = ETHER_ADDR_LEN; 744 e_addr = LLADDR(sdl); 745 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 746 *llsa = (struct sockaddr *)sdl; 747 return 0; 748 #endif 749 #ifdef INET6 750 case AF_INET6: 751 sin6 = (struct sockaddr_in6 *)sa; 752 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 753 /* 754 * An IP6 address of 0 means listen to all 755 * of the Ethernet multicast address used for IP6. 756 * (This is used for multicast routers.) 757 */ 758 ifp->if_flags |= IFF_ALLMULTI; 759 *llsa = NULL; 760 return 0; 761 } 762 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 763 return EADDRNOTAVAIL; 764 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO); 765 sdl->sdl_len = sizeof *sdl; 766 sdl->sdl_family = AF_LINK; 767 sdl->sdl_index = ifp->if_index; 768 sdl->sdl_type = IFT_ETHER; 769 sdl->sdl_alen = ETHER_ADDR_LEN; 770 e_addr = LLADDR(sdl); 771 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 772 *llsa = (struct sockaddr *)sdl; 773 return 0; 774 #endif 775 776 default: 777 /* 778 * Well, the text isn't quite right, but it's the name 779 * that counts... 780 */ 781 return EAFNOSUPPORT; 782 } 783 } 784 785 #if 0 786 /* 787 * This is for reference. We have a table-driven version 788 * of the little-endian crc32 generator, which is faster 789 * than the double-loop. 790 */ 791 uint32_t 792 ether_crc32_le(const uint8_t *buf, size_t len) 793 { 794 uint32_t c, crc, carry; 795 size_t i, j; 796 797 crc = 0xffffffffU; /* initial value */ 798 799 for (i = 0; i < len; i++) { 800 c = buf[i]; 801 for (j = 0; j < 8; j++) { 802 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 803 crc >>= 1; 804 c >>= 1; 805 if (carry) 806 crc = (crc ^ ETHER_CRC_POLY_LE); 807 } 808 } 809 810 return (crc); 811 } 812 #else 813 uint32_t 814 ether_crc32_le(const uint8_t *buf, size_t len) 815 { 816 static const uint32_t crctab[] = { 817 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 818 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 819 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 820 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 821 }; 822 uint32_t crc; 823 size_t i; 824 825 crc = 0xffffffffU; /* initial value */ 826 827 for (i = 0; i < len; i++) { 828 crc ^= buf[i]; 829 crc = (crc >> 4) ^ crctab[crc & 0xf]; 830 crc = (crc >> 4) ^ crctab[crc & 0xf]; 831 } 832 833 return (crc); 834 } 835 #endif 836 837 uint32_t 838 ether_crc32_be(const uint8_t *buf, size_t len) 839 { 840 uint32_t c, crc, carry; 841 size_t i, j; 842 843 crc = 0xffffffffU; /* initial value */ 844 845 for (i = 0; i < len; i++) { 846 c = buf[i]; 847 for (j = 0; j < 8; j++) { 848 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 849 crc <<= 1; 850 c >>= 1; 851 if (carry) 852 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 853 } 854 } 855 856 return (crc); 857 } 858 859 /* 860 * find the size of ethernet header, and call classifier 861 */ 862 void 863 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 864 struct altq_pktattr *pktattr) 865 { 866 struct ether_header *eh; 867 uint16_t ether_type; 868 int hlen, af, hdrsize; 869 870 hlen = sizeof(struct ether_header); 871 eh = mtod(m, struct ether_header *); 872 873 ether_type = ntohs(eh->ether_type); 874 if (ether_type < ETHERMTU) { 875 /* ick! LLC/SNAP */ 876 struct llc *llc = (struct llc *)(eh + 1); 877 hlen += 8; 878 879 if (m->m_len < hlen || 880 llc->llc_dsap != LLC_SNAP_LSAP || 881 llc->llc_ssap != LLC_SNAP_LSAP || 882 llc->llc_control != LLC_UI) 883 goto bad; /* not snap! */ 884 885 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 886 } 887 888 if (ether_type == ETHERTYPE_IP) { 889 af = AF_INET; 890 hdrsize = 20; /* sizeof(struct ip) */ 891 #ifdef INET6 892 } else if (ether_type == ETHERTYPE_IPV6) { 893 af = AF_INET6; 894 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 895 #endif 896 } else 897 goto bad; 898 899 while (m->m_len <= hlen) { 900 hlen -= m->m_len; 901 m = m->m_next; 902 } 903 if (m->m_len < hlen + hdrsize) { 904 /* 905 * ip header is not in a single mbuf. this should not 906 * happen in the current code. 907 * (todo: use m_pulldown in the future) 908 */ 909 goto bad; 910 } 911 m->m_data += hlen; 912 m->m_len -= hlen; 913 ifq_classify(ifq, m, af, pktattr); 914 m->m_data -= hlen; 915 m->m_len += hlen; 916 917 return; 918 919 bad: 920 pktattr->pattr_class = NULL; 921 pktattr->pattr_hdr = NULL; 922 pktattr->pattr_af = AF_UNSPEC; 923 } 924 925 static void 926 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 927 const struct ether_header *save_eh) 928 { 929 struct mbuf *m = *m0; 930 931 ether_restore_hdr++; 932 933 /* 934 * Prepend the header, optimize for the common case of 935 * eh pointing into the mbuf. 936 */ 937 if ((const void *)(eh + 1) == (void *)m->m_data) { 938 m->m_data -= ETHER_HDR_LEN; 939 m->m_len += ETHER_HDR_LEN; 940 m->m_pkthdr.len += ETHER_HDR_LEN; 941 } else { 942 ether_prepend_hdr++; 943 944 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 945 if (m != NULL) { 946 bcopy(save_eh, mtod(m, struct ether_header *), 947 ETHER_HDR_LEN); 948 } 949 } 950 *m0 = m; 951 } 952 953 /* 954 * Upper layer processing for a received Ethernet packet. 955 */ 956 void 957 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 958 { 959 struct ether_header *eh; 960 int isr, discard = 0; 961 u_short ether_type; 962 struct ip_fw *rule = NULL; 963 964 M_ASSERTPKTHDR(m); 965 KASSERT(m->m_len >= ETHER_HDR_LEN, 966 ("ether header is not contiguous!")); 967 968 eh = mtod(m, struct ether_header *); 969 970 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 971 struct m_tag *mtag; 972 973 /* Extract info from dummynet tag */ 974 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 975 KKASSERT(mtag != NULL); 976 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 977 KKASSERT(rule != NULL); 978 979 m_tag_delete(m, mtag); 980 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 981 982 /* packet is passing the second time */ 983 goto post_stats; 984 } 985 986 /* 987 * We got a packet which was unicast to a different Ethernet 988 * address. If the driver is working properly, then this 989 * situation can only happen when the interface is in 990 * promiscuous mode. We defer the packet discarding until the 991 * vlan processing is done, so that vlan/bridge or vlan/netgraph 992 * could work. 993 */ 994 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 995 !ETHER_IS_MULTICAST(eh->ether_dhost) && 996 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) { 997 if (ether_debug & 1) { 998 kprintf("%02x:%02x:%02x:%02x:%02x:%02x " 999 "%02x:%02x:%02x:%02x:%02x:%02x " 1000 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n", 1001 eh->ether_dhost[0], 1002 eh->ether_dhost[1], 1003 eh->ether_dhost[2], 1004 eh->ether_dhost[3], 1005 eh->ether_dhost[4], 1006 eh->ether_dhost[5], 1007 eh->ether_shost[0], 1008 eh->ether_shost[1], 1009 eh->ether_shost[2], 1010 eh->ether_shost[3], 1011 eh->ether_shost[4], 1012 eh->ether_shost[5], 1013 eh->ether_type, 1014 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0], 1015 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1], 1016 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2], 1017 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3], 1018 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4], 1019 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5] 1020 ); 1021 } 1022 if ((ether_debug & 2) == 0) 1023 discard = 1; 1024 } 1025 1026 post_stats: 1027 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0 && !discard) { 1028 struct ether_header save_eh = *eh; 1029 1030 /* XXX old crufty stuff, needs to be removed */ 1031 m_adj(m, sizeof(struct ether_header)); 1032 1033 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1034 m_freem(m); 1035 return; 1036 } 1037 1038 ether_restore_header(&m, eh, &save_eh); 1039 if (m == NULL) 1040 return; 1041 eh = mtod(m, struct ether_header *); 1042 } 1043 1044 ether_type = ntohs(eh->ether_type); 1045 KKASSERT(ether_type != ETHERTYPE_VLAN); 1046 1047 /* Handle input from a lagg(4) port */ 1048 if (ifp->if_type == IFT_IEEE8023ADLAG) { 1049 KASSERT(lagg_input_p != NULL, 1050 ("%s: if_lagg not loaded!", __func__)); 1051 (*lagg_input_p)(ifp, m); 1052 return; 1053 } 1054 1055 if (m->m_flags & M_VLANTAG) { 1056 void (*vlan_input_func)(struct mbuf *); 1057 1058 vlan_input_func = vlan_input_p; 1059 /* Make sure 'vlan_input_func' is really used. */ 1060 cpu_ccfence(); 1061 if (vlan_input_func != NULL) { 1062 vlan_input_func(m); 1063 } else { 1064 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1065 m_freem(m); 1066 } 1067 return; 1068 } 1069 1070 /* 1071 * If we have been asked to discard this packet 1072 * (e.g. not for us), drop it before entering 1073 * the upper layer. 1074 */ 1075 if (discard) { 1076 m_freem(m); 1077 return; 1078 } 1079 1080 /* 1081 * Clear protocol specific flags, 1082 * before entering the upper layer. 1083 */ 1084 m->m_flags &= ~M_ETHER_FLAGS; 1085 1086 /* Strip ethernet header. */ 1087 m_adj(m, sizeof(struct ether_header)); 1088 1089 switch (ether_type) { 1090 #ifdef INET 1091 case ETHERTYPE_IP: 1092 if ((m->m_flags & M_LENCHECKED) == 0) { 1093 if (!ip_lengthcheck(&m, 0)) 1094 return; 1095 } 1096 if (ipflow_fastforward(m)) 1097 return; 1098 isr = NETISR_IP; 1099 break; 1100 1101 case ETHERTYPE_ARP: 1102 if (ifp->if_flags & IFF_NOARP) { 1103 /* Discard packet if ARP is disabled on interface */ 1104 m_freem(m); 1105 return; 1106 } 1107 isr = NETISR_ARP; 1108 break; 1109 #endif 1110 1111 #ifdef INET6 1112 case ETHERTYPE_IPV6: 1113 isr = NETISR_IPV6; 1114 break; 1115 #endif 1116 1117 #ifdef MPLS 1118 case ETHERTYPE_MPLS: 1119 case ETHERTYPE_MPLS_MCAST: 1120 /* Should have been set by ether_input(). */ 1121 KKASSERT(m->m_flags & M_MPLSLABELED); 1122 isr = NETISR_MPLS; 1123 break; 1124 #endif 1125 1126 default: 1127 /* 1128 * The accurate msgport is not determined before 1129 * we reach here, so recharacterize packet. 1130 */ 1131 m->m_flags &= ~M_HASH; 1132 if (ng_ether_input_orphan_p != NULL) { 1133 /* 1134 * Put back the ethernet header so netgraph has a 1135 * consistent view of inbound packets. 1136 */ 1137 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); 1138 if (m == NULL) { 1139 /* 1140 * M_PREPEND frees the mbuf in case of failure. 1141 */ 1142 return; 1143 } 1144 /* 1145 * Hold BGL and recheck ng_ether_input_orphan_p 1146 */ 1147 get_mplock(); 1148 if (ng_ether_input_orphan_p != NULL) { 1149 ng_ether_input_orphan_p(ifp, m); 1150 rel_mplock(); 1151 return; 1152 } 1153 rel_mplock(); 1154 } 1155 m_freem(m); 1156 return; 1157 } 1158 1159 if (m->m_flags & M_HASH) { 1160 if (&curthread->td_msgport == 1161 netisr_hashport(m->m_pkthdr.hash)) { 1162 netisr_handle(isr, m); 1163 return; 1164 } else { 1165 /* 1166 * XXX Something is wrong, 1167 * we probably should panic here! 1168 */ 1169 m->m_flags &= ~M_HASH; 1170 atomic_add_long(ðer_input_wronghash, 1); 1171 } 1172 } 1173 #ifdef RSS_DEBUG 1174 atomic_add_long(ðer_input_requeue, 1); 1175 #endif 1176 netisr_queue(isr, m); 1177 } 1178 1179 /* 1180 * First we perform any link layer operations, then continue to the 1181 * upper layers with ether_demux_oncpu(). 1182 */ 1183 static void 1184 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1185 { 1186 #ifdef CARP 1187 void *carp; 1188 #endif 1189 1190 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1191 /* 1192 * Receiving interface's flags are changed, when this 1193 * packet is waiting for processing; discard it. 1194 */ 1195 m_freem(m); 1196 return; 1197 } 1198 1199 /* 1200 * Tap the packet off here for a bridge. bridge_input() 1201 * will return NULL if it has consumed the packet, otherwise 1202 * it gets processed as normal. Note that bridge_input() 1203 * will always return the original packet if we need to 1204 * process it locally. 1205 */ 1206 if (ifp->if_bridge) { 1207 KASSERT(bridge_input_p != NULL, 1208 ("%s: if_bridge not loaded!", __func__)); 1209 1210 if(m->m_flags & M_ETHER_BRIDGED) { 1211 m->m_flags &= ~M_ETHER_BRIDGED; 1212 } else { 1213 m = bridge_input_p(ifp, m); 1214 if (m == NULL) 1215 return; 1216 1217 KASSERT(ifp == m->m_pkthdr.rcvif, 1218 ("bridge_input_p changed rcvif")); 1219 } 1220 } 1221 1222 #ifdef CARP 1223 carp = ifp->if_carp; 1224 if (carp) { 1225 m = carp_input(carp, m); 1226 if (m == NULL) 1227 return; 1228 KASSERT(ifp == m->m_pkthdr.rcvif, 1229 ("carp_input changed rcvif")); 1230 } 1231 #endif 1232 1233 /* Handle ng_ether(4) processing, if any */ 1234 if (ng_ether_input_p != NULL) { 1235 /* 1236 * Hold BGL and recheck ng_ether_input_p 1237 */ 1238 get_mplock(); 1239 if (ng_ether_input_p != NULL) 1240 ng_ether_input_p(ifp, &m); 1241 rel_mplock(); 1242 1243 if (m == NULL) 1244 return; 1245 } 1246 1247 /* Continue with upper layer processing */ 1248 ether_demux_oncpu(ifp, m); 1249 } 1250 1251 /* 1252 * Perform certain functions of ether_input(): 1253 * - Test IFF_UP 1254 * - Update statistics 1255 * - Run bpf(4) tap if requested 1256 * Then pass the packet to ether_input_oncpu(). 1257 * 1258 * This function should be used by pseudo interface (e.g. vlan(4)), 1259 * when it tries to claim that the packet is received by it. 1260 * 1261 * REINPUT_KEEPRCVIF 1262 * REINPUT_RUNBPF 1263 */ 1264 void 1265 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags) 1266 { 1267 /* Discard packet if interface is not up */ 1268 if (!(ifp->if_flags & IFF_UP)) { 1269 m_freem(m); 1270 return; 1271 } 1272 1273 /* 1274 * Change receiving interface. The bridge will often pass a flag to 1275 * ask that this not be done so ARPs get applied to the correct 1276 * side. 1277 */ 1278 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 || 1279 m->m_pkthdr.rcvif == NULL) { 1280 m->m_pkthdr.rcvif = ifp; 1281 } 1282 1283 /* Update statistics */ 1284 IFNET_STAT_INC(ifp, ipackets, 1); 1285 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1286 if (m->m_flags & (M_MCAST | M_BCAST)) 1287 IFNET_STAT_INC(ifp, imcasts, 1); 1288 1289 if (reinput_flags & REINPUT_RUNBPF) 1290 BPF_MTAP(ifp, m); 1291 1292 ether_input_oncpu(ifp, m); 1293 } 1294 1295 static __inline boolean_t 1296 ether_vlancheck(struct mbuf **m0) 1297 { 1298 struct mbuf *m = *m0; 1299 struct ether_header *eh = mtod(m, struct ether_header *); 1300 uint16_t ether_type = ntohs(eh->ether_type); 1301 1302 if (ether_type == ETHERTYPE_VLAN) { 1303 if ((m->m_flags & M_VLANTAG) == 0) { 1304 /* 1305 * Extract vlan tag if hardware does not do 1306 * it for us. 1307 */ 1308 vlan_ether_decap(&m); 1309 if (m == NULL) 1310 goto failed; 1311 1312 eh = mtod(m, struct ether_header *); 1313 ether_type = ntohs(eh->ether_type); 1314 if (ether_type == ETHERTYPE_VLAN) { 1315 /* 1316 * To prevent possible dangerous recursion, 1317 * we don't do vlan-in-vlan. 1318 */ 1319 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1320 goto failed; 1321 } 1322 } else { 1323 /* 1324 * To prevent possible dangerous recursion, 1325 * we don't do vlan-in-vlan. 1326 */ 1327 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1); 1328 goto failed; 1329 } 1330 KKASSERT(ether_type != ETHERTYPE_VLAN); 1331 } 1332 1333 m->m_flags |= M_ETHER_VLANCHECKED; 1334 *m0 = m; 1335 return TRUE; 1336 failed: 1337 if (m != NULL) 1338 m_freem(m); 1339 *m0 = NULL; 1340 return FALSE; 1341 } 1342 1343 static void 1344 ether_input_handler(netmsg_t nmsg) 1345 { 1346 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1347 struct ether_header *eh; 1348 struct ifnet *ifp; 1349 struct mbuf *m; 1350 1351 m = nmp->nm_packet; 1352 M_ASSERTPKTHDR(m); 1353 1354 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1355 if (!ether_vlancheck(&m)) { 1356 KKASSERT(m == NULL); 1357 return; 1358 } 1359 } 1360 1361 ifp = m->m_pkthdr.rcvif; 1362 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) || 1363 __predict_false(ether_input_ckhash)) { 1364 int isr; 1365 1366 /* 1367 * Need to verify the hash supplied by the hardware 1368 * which could be wrong. 1369 */ 1370 m->m_flags &= ~(M_HASH | M_CKHASH); 1371 isr = ether_characterize(&m); 1372 if (m == NULL) 1373 return; 1374 KKASSERT(m->m_flags & M_HASH); 1375 1376 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) { 1377 /* 1378 * Wrong hardware supplied hash; redispatch 1379 */ 1380 ether_dispatch(ifp, isr, m, -1); 1381 if (__predict_false(ether_input_ckhash)) 1382 atomic_add_long(ðer_input_wronghwhash, 1); 1383 return; 1384 } 1385 } 1386 1387 eh = mtod(m, struct ether_header *); 1388 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1389 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1390 ifp->if_addrlen) == 0) 1391 m->m_flags |= M_BCAST; 1392 else 1393 m->m_flags |= M_MCAST; 1394 IFNET_STAT_INC(ifp, imcasts, 1); 1395 } 1396 1397 ether_input_oncpu(ifp, m); 1398 } 1399 1400 /* 1401 * Send the packet to the target netisr msgport 1402 * 1403 * At this point the packet must be characterized (M_HASH set), 1404 * so we know which netisr to send it to. 1405 */ 1406 static void 1407 ether_dispatch(struct ifnet *ifp, int isr, struct mbuf *m, int cpuid) 1408 { 1409 struct netmsg_packet *pmsg; 1410 int target_cpuid; 1411 1412 KKASSERT(m->m_flags & M_HASH); 1413 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash); 1414 1415 pmsg = &m->m_hdr.mh_netmsg; 1416 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1417 0, ether_input_handler); 1418 pmsg->nm_packet = m; 1419 pmsg->base.lmsg.u.ms_result = isr; 1420 1421 logether(disp_beg, NULL); 1422 if (target_cpuid == cpuid) { 1423 if ((ifp->if_flags & IFF_IDIRECT) && IN_NETISR_NCPUS(cpuid)) { 1424 ether_input_handler((netmsg_t)pmsg); 1425 } else { 1426 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid), 1427 &pmsg->base.lmsg); 1428 } 1429 } else { 1430 lwkt_sendmsg(netisr_cpuport(target_cpuid), 1431 &pmsg->base.lmsg); 1432 } 1433 logether(disp_end, NULL); 1434 } 1435 1436 /* 1437 * Process a received Ethernet packet. 1438 * 1439 * The ethernet header is assumed to be in the mbuf so the caller 1440 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1441 * bytes in the first mbuf. 1442 * 1443 * If the caller knows that the current thread is stick to the current 1444 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid 1445 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should 1446 * be passed as 'cpuid' argument. 1447 */ 1448 void 1449 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1450 int cpuid) 1451 { 1452 int isr; 1453 1454 M_ASSERTPKTHDR(m); 1455 1456 /* Discard packet if interface is not up */ 1457 if (!(ifp->if_flags & IFF_UP)) { 1458 m_freem(m); 1459 return; 1460 } 1461 1462 if (m->m_len < sizeof(struct ether_header)) { 1463 /* XXX error in the caller. */ 1464 m_freem(m); 1465 return; 1466 } 1467 1468 m->m_pkthdr.rcvif = ifp; 1469 1470 logether(pkt_beg, ifp); 1471 1472 ETHER_BPF_MTAP(ifp, m); 1473 1474 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len); 1475 1476 if (ifp->if_flags & IFF_MONITOR) { 1477 struct ether_header *eh; 1478 1479 eh = mtod(m, struct ether_header *); 1480 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1481 IFNET_STAT_INC(ifp, imcasts, 1); 1482 1483 /* 1484 * Interface marked for monitoring; discard packet. 1485 */ 1486 m_freem(m); 1487 1488 logether(pkt_end, ifp); 1489 return; 1490 } 1491 1492 /* 1493 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1494 * we can dispatch it immediately with trivial checks. 1495 */ 1496 if (pi != NULL && (m->m_flags & M_HASH)) { 1497 #ifdef RSS_DEBUG 1498 atomic_add_long(ðer_pktinfo_try, 1); 1499 #endif 1500 netisr_hashcheck(pi->pi_netisr, m, pi); 1501 if (m->m_flags & M_HASH) { 1502 ether_dispatch(ifp, pi->pi_netisr, m, cpuid); 1503 #ifdef RSS_DEBUG 1504 atomic_add_long(ðer_pktinfo_hit, 1); 1505 #endif 1506 logether(pkt_end, ifp); 1507 return; 1508 } 1509 } 1510 #ifdef RSS_DEBUG 1511 else if (ifp->if_capenable & IFCAP_RSS) { 1512 if (pi == NULL) 1513 atomic_add_long(ðer_rss_nopi, 1); 1514 else 1515 atomic_add_long(ðer_rss_nohash, 1); 1516 } 1517 #endif 1518 1519 /* 1520 * Packet hash will be recalculated by software, so clear 1521 * the M_HASH and M_CKHASH flag set by the driver; the hash 1522 * value calculated by the hardware may not be exactly what 1523 * we want. 1524 */ 1525 m->m_flags &= ~(M_HASH | M_CKHASH); 1526 1527 if (!ether_vlancheck(&m)) { 1528 KKASSERT(m == NULL); 1529 logether(pkt_end, ifp); 1530 return; 1531 } 1532 1533 isr = ether_characterize(&m); 1534 if (m == NULL) { 1535 logether(pkt_end, ifp); 1536 return; 1537 } 1538 1539 /* 1540 * Finally dispatch it 1541 */ 1542 ether_dispatch(ifp, isr, m, cpuid); 1543 1544 logether(pkt_end, ifp); 1545 } 1546 1547 static int 1548 ether_characterize(struct mbuf **m0) 1549 { 1550 struct mbuf *m = *m0; 1551 struct ether_header *eh; 1552 uint16_t ether_type; 1553 int isr; 1554 1555 eh = mtod(m, struct ether_header *); 1556 ether_type = ntohs(eh->ether_type); 1557 1558 /* 1559 * Map ether type to netisr id. 1560 */ 1561 switch (ether_type) { 1562 #ifdef INET 1563 case ETHERTYPE_IP: 1564 isr = NETISR_IP; 1565 break; 1566 1567 case ETHERTYPE_ARP: 1568 isr = NETISR_ARP; 1569 break; 1570 #endif 1571 1572 #ifdef INET6 1573 case ETHERTYPE_IPV6: 1574 isr = NETISR_IPV6; 1575 break; 1576 #endif 1577 1578 #ifdef MPLS 1579 case ETHERTYPE_MPLS: 1580 case ETHERTYPE_MPLS_MCAST: 1581 m->m_flags |= M_MPLSLABELED; 1582 isr = NETISR_MPLS; 1583 break; 1584 #endif 1585 1586 default: 1587 /* 1588 * NETISR_MAX is an invalid value; it is chosen to let 1589 * netisr_characterize() know that we have no clear 1590 * idea where this packet should go. 1591 */ 1592 isr = NETISR_MAX; 1593 break; 1594 } 1595 1596 /* 1597 * Ask the isr to characterize the packet since we couldn't. 1598 * This is an attempt to optimally get us onto the correct protocol 1599 * thread. 1600 */ 1601 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1602 1603 *m0 = m; 1604 return isr; 1605 } 1606 1607 static void 1608 ether_demux_handler(netmsg_t nmsg) 1609 { 1610 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1611 struct ifnet *ifp; 1612 struct mbuf *m; 1613 1614 m = nmp->nm_packet; 1615 M_ASSERTPKTHDR(m); 1616 ifp = m->m_pkthdr.rcvif; 1617 1618 ether_demux_oncpu(ifp, m); 1619 } 1620 1621 void 1622 ether_demux(struct mbuf *m) 1623 { 1624 struct netmsg_packet *pmsg; 1625 int isr; 1626 1627 isr = ether_characterize(&m); 1628 if (m == NULL) 1629 return; 1630 1631 KKASSERT(m->m_flags & M_HASH); 1632 pmsg = &m->m_hdr.mh_netmsg; 1633 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1634 0, ether_demux_handler); 1635 pmsg->nm_packet = m; 1636 pmsg->base.lmsg.u.ms_result = isr; 1637 1638 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg); 1639 } 1640 1641 u_char * 1642 kether_aton(const char *macstr, u_char *addr) 1643 { 1644 unsigned int o0, o1, o2, o3, o4, o5; 1645 int n; 1646 1647 if (macstr == NULL || addr == NULL) 1648 return NULL; 1649 1650 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2, 1651 &o3, &o4, &o5); 1652 if (n != 6) 1653 return NULL; 1654 1655 addr[0] = o0; 1656 addr[1] = o1; 1657 addr[2] = o2; 1658 addr[3] = o3; 1659 addr[4] = o4; 1660 addr[5] = o5; 1661 1662 return addr; 1663 } 1664 1665 char * 1666 kether_ntoa(const u_char *addr, char *buf) 1667 { 1668 int len = ETHER_ADDRSTRLEN + 1; 1669 int n; 1670 1671 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0], 1672 addr[1], addr[2], addr[3], addr[4], addr[5]); 1673 1674 if (n < 17) 1675 return NULL; 1676 1677 return buf; 1678 } 1679 1680 MODULE_VERSION(ether, 1); 1681