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. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93 34 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $ 35 * $DragonFly: src/sys/net/if_ethersubr.c,v 1.96 2008/11/22 04:00:53 sephe Exp $ 36 */ 37 38 #include "opt_atalk.h" 39 #include "opt_inet.h" 40 #include "opt_inet6.h" 41 #include "opt_ipx.h" 42 #include "opt_mpls.h" 43 #include "opt_netgraph.h" 44 #include "opt_carp.h" 45 #include "opt_rss.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/globaldata.h> 50 #include <sys/kernel.h> 51 #include <sys/ktr.h> 52 #include <sys/lock.h> 53 #include <sys/malloc.h> 54 #include <sys/mbuf.h> 55 #include <sys/msgport.h> 56 #include <sys/socket.h> 57 #include <sys/sockio.h> 58 #include <sys/sysctl.h> 59 #include <sys/thread.h> 60 61 #include <sys/thread2.h> 62 #include <sys/mplock2.h> 63 64 #include <net/if.h> 65 #include <net/netisr.h> 66 #include <net/route.h> 67 #include <net/if_llc.h> 68 #include <net/if_dl.h> 69 #include <net/if_types.h> 70 #include <net/ifq_var.h> 71 #include <net/bpf.h> 72 #include <net/ethernet.h> 73 #include <net/vlan/if_vlan_ether.h> 74 #include <net/netmsg2.h> 75 76 #if defined(INET) || defined(INET6) 77 #include <netinet/in.h> 78 #include <netinet/ip_var.h> 79 #include <netinet/if_ether.h> 80 #include <netinet/ip_flow.h> 81 #include <net/ipfw/ip_fw.h> 82 #include <net/dummynet/ip_dummynet.h> 83 #endif 84 #ifdef INET6 85 #include <netinet6/nd6.h> 86 #endif 87 88 #ifdef CARP 89 #include <netinet/ip_carp.h> 90 #endif 91 92 #ifdef IPX 93 #include <netproto/ipx/ipx.h> 94 #include <netproto/ipx/ipx_if.h> 95 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m); 96 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, 97 short *tp, int *hlen); 98 #endif 99 100 #ifdef NS 101 #include <netns/ns.h> 102 #include <netns/ns_if.h> 103 ushort ns_nettype; 104 int ether_outputdebug = 0; 105 int ether_inputdebug = 0; 106 #endif 107 108 #ifdef NETATALK 109 #include <netproto/atalk/at.h> 110 #include <netproto/atalk/at_var.h> 111 #include <netproto/atalk/at_extern.h> 112 113 #define llc_snap_org_code llc_un.type_snap.org_code 114 #define llc_snap_ether_type llc_un.type_snap.ether_type 115 116 extern u_char at_org_code[3]; 117 extern u_char aarp_org_code[3]; 118 #endif /* NETATALK */ 119 120 #ifdef MPLS 121 #include <netproto/mpls/mpls.h> 122 #endif 123 124 /* netgraph node hooks for ng_ether(4) */ 125 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp); 126 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, 127 struct mbuf *m, const struct ether_header *eh); 128 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp); 129 void (*ng_ether_attach_p)(struct ifnet *ifp); 130 void (*ng_ether_detach_p)(struct ifnet *ifp); 131 132 void (*vlan_input_p)(struct mbuf *); 133 134 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *, 135 struct rtentry *); 136 static void ether_restore_header(struct mbuf **, const struct ether_header *, 137 const struct ether_header *); 138 139 /* 140 * if_bridge support 141 */ 142 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *); 143 int (*bridge_output_p)(struct ifnet *, struct mbuf *); 144 void (*bridge_dn_p)(struct mbuf *, struct ifnet *); 145 146 static int ether_resolvemulti(struct ifnet *, struct sockaddr **, 147 struct sockaddr *); 148 149 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = { 150 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 151 }; 152 153 #define gotoerr(e) do { error = (e); goto bad; } while (0) 154 #define IFP2AC(ifp) ((struct arpcom *)(ifp)) 155 156 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, 157 struct ip_fw **rule, 158 const struct ether_header *eh); 159 160 static int ether_ipfw; 161 static u_int ether_restore_hdr; 162 static u_int ether_prepend_hdr; 163 164 #ifdef RSS_DEBUG 165 static u_int ether_pktinfo_try; 166 static u_int ether_pktinfo_hit; 167 static u_int ether_rss_nopi; 168 static u_int ether_rss_nohash; 169 #endif 170 171 SYSCTL_DECL(_net_link); 172 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet"); 173 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW, 174 ðer_ipfw, 0, "Pass ether pkts through firewall"); 175 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW, 176 ðer_restore_hdr, 0, "# of ether header restoration"); 177 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW, 178 ðer_prepend_hdr, 0, 179 "# of ether header restoration which prepends mbuf"); 180 #ifdef RSS_DEBUG 181 SYSCTL_UINT(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW, 182 ðer_rss_nopi, 0, "# of packets do not have pktinfo"); 183 SYSCTL_UINT(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW, 184 ðer_rss_nohash, 0, "# of packets do not have hash"); 185 SYSCTL_UINT(_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_UINT(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW, 189 ðer_pktinfo_hit, 0, 190 "# of packets whose msgport are found using pktinfo"); 191 #endif 192 193 #define ETHER_KTR_STR "ifp=%p" 194 #define ETHER_KTR_ARG_SIZE (sizeof(void *)) 195 #ifndef KTR_ETHERNET 196 #define KTR_ETHERNET KTR_ALL 197 #endif 198 KTR_INFO_MASTER(ether); 199 KTR_INFO(KTR_ETHERNET, ether, chain_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 200 KTR_INFO(KTR_ETHERNET, ether, chain_end, 1, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 201 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 202 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE); 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), MB_DONTWAIT); 231 if (m == NULL) 232 return (ENOBUFS); 233 eh = mtod(m, struct ether_header *); 234 edst = eh->ether_dhost; 235 236 /* 237 * Fill in the destination ethernet address and frame type. 238 */ 239 switch (dst->sa_family) { 240 #ifdef INET 241 case AF_INET: 242 if (!arpresolve(ifp, rt, m, dst, edst)) 243 return (0); /* if not yet resolved */ 244 #ifdef MPLS 245 if (m->m_flags & M_MPLSLABELED) 246 eh->ether_type = htons(ETHERTYPE_MPLS); 247 else 248 #endif 249 eh->ether_type = htons(ETHERTYPE_IP); 250 break; 251 #endif 252 #ifdef INET6 253 case AF_INET6: 254 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst)) 255 return (0); /* Something bad happenned. */ 256 eh->ether_type = htons(ETHERTYPE_IPV6); 257 break; 258 #endif 259 #ifdef IPX 260 case AF_IPX: 261 if (ef_outputp != NULL) { 262 /* 263 * Hold BGL and recheck ef_outputp 264 */ 265 get_mplock(); 266 if (ef_outputp != NULL) { 267 error = ef_outputp(ifp, &m, dst, 268 &eh->ether_type, &hlen); 269 rel_mplock(); 270 if (error) 271 goto bad; 272 else 273 break; 274 } 275 rel_mplock(); 276 } 277 eh->ether_type = htons(ETHERTYPE_IPX); 278 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host), 279 edst, ETHER_ADDR_LEN); 280 break; 281 #endif 282 #ifdef NETATALK 283 case AF_APPLETALK: { 284 struct at_ifaddr *aa; 285 286 /* 287 * Hold BGL 288 */ 289 get_mplock(); 290 291 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) { 292 error = 0; /* XXX */ 293 rel_mplock(); 294 goto bad; 295 } 296 /* 297 * In the phase 2 case, need to prepend an mbuf for 298 * the llc header. Since we must preserve the value 299 * of m, which is passed to us by value, we m_copy() 300 * the first mbuf, and use it for our llc header. 301 */ 302 if (aa->aa_flags & AFA_PHASE2) { 303 struct llc llc; 304 305 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT); 306 eh = mtod(m, struct ether_header *); 307 edst = eh->ether_dhost; 308 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; 309 llc.llc_control = LLC_UI; 310 bcopy(at_org_code, llc.llc_snap_org_code, 311 sizeof at_org_code); 312 llc.llc_snap_ether_type = htons(ETHERTYPE_AT); 313 bcopy(&llc, 314 mtod(m, caddr_t) + sizeof(struct ether_header), 315 sizeof(struct llc)); 316 eh->ether_type = htons(m->m_pkthdr.len); 317 hlen = sizeof(struct llc) + ETHER_HDR_LEN; 318 } else { 319 eh->ether_type = htons(ETHERTYPE_AT); 320 } 321 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) { 322 rel_mplock(); 323 return (0); 324 } 325 326 rel_mplock(); 327 break; 328 } 329 #endif 330 #ifdef NS 331 case AF_NS: 332 switch(ns_nettype) { 333 default: 334 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 335 eh->ether_type = 0x8137; 336 break; 337 case 0x0: /* Novell 802.3 */ 338 eh->ether_type = htons(m->m_pkthdr.len); 339 break; 340 case 0xe0e0: /* Novell 802.2 and Token-Ring */ 341 M_PREPEND(m, 3, MB_DONTWAIT); 342 eh = mtod(m, struct ether_header *); 343 edst = eh->ether_dhost; 344 eh->ether_type = htons(m->m_pkthdr.len); 345 cp = mtod(m, u_char *) + sizeof(struct ether_header); 346 *cp++ = 0xE0; 347 *cp++ = 0xE0; 348 *cp++ = 0x03; 349 break; 350 } 351 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst, 352 ETHER_ADDR_LEN); 353 /* 354 * XXX if ns_thishost is the same as the node's ethernet 355 * address then just the default code will catch this anyhow. 356 * So I'm not sure if this next clause should be here at all? 357 * [JRE] 358 */ 359 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) { 360 m->m_pkthdr.rcvif = ifp; 361 netisr_queue(NETISR_NS, m); 362 return (error); 363 } 364 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0) 365 m->m_flags |= M_BCAST; 366 break; 367 #endif 368 case pseudo_AF_HDRCMPLT: 369 case AF_UNSPEC: 370 loop_copy = -1; /* if this is for us, don't do it */ 371 deh = (struct ether_header *)dst->sa_data; 372 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN); 373 eh->ether_type = deh->ether_type; 374 break; 375 376 default: 377 if_printf(ifp, "can't handle af%d\n", dst->sa_family); 378 gotoerr(EAFNOSUPPORT); 379 } 380 381 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */ 382 memcpy(eh->ether_shost, 383 ((struct ether_header *)dst->sa_data)->ether_shost, 384 ETHER_ADDR_LEN); 385 else 386 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN); 387 388 /* 389 * Bridges require special output handling. 390 */ 391 if (ifp->if_bridge) { 392 KASSERT(bridge_output_p != NULL, 393 ("%s: if_bridge not loaded!", __func__)); 394 return bridge_output_p(ifp, m); 395 } 396 397 /* 398 * If a simplex interface, and the packet is being sent to our 399 * Ethernet address or a broadcast address, loopback a copy. 400 * XXX To make a simplex device behave exactly like a duplex 401 * device, we should copy in the case of sending to our own 402 * ethernet address (thus letting the original actually appear 403 * on the wire). However, we don't do that here for security 404 * reasons and compatibility with the original behavior. 405 */ 406 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) { 407 int csum_flags = 0; 408 409 if (m->m_pkthdr.csum_flags & CSUM_IP) 410 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID); 411 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) 412 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 413 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) { 414 struct mbuf *n; 415 416 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) { 417 n->m_pkthdr.csum_flags |= csum_flags; 418 if (csum_flags & CSUM_DATA_VALID) 419 n->m_pkthdr.csum_data = 0xffff; 420 if_simloop(ifp, n, dst->sa_family, hlen); 421 } else 422 ifp->if_iqdrops++; 423 } else if (bcmp(eh->ether_dhost, eh->ether_shost, 424 ETHER_ADDR_LEN) == 0) { 425 m->m_pkthdr.csum_flags |= csum_flags; 426 if (csum_flags & CSUM_DATA_VALID) 427 m->m_pkthdr.csum_data = 0xffff; 428 if_simloop(ifp, m, dst->sa_family, hlen); 429 return (0); /* XXX */ 430 } 431 } 432 433 #ifdef CARP 434 if (ifp->if_carp) { 435 /* 436 * Hold BGL and recheck ifp->if_carp 437 */ 438 get_mplock(); 439 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) { 440 rel_mplock(); 441 goto bad; 442 } 443 rel_mplock(); 444 } 445 #endif 446 447 448 /* Handle ng_ether(4) processing, if any */ 449 if (ng_ether_output_p != NULL) { 450 /* 451 * Hold BGL and recheck ng_ether_output_p 452 */ 453 get_mplock(); 454 if (ng_ether_output_p != NULL) { 455 if ((error = ng_ether_output_p(ifp, &m)) != 0) { 456 rel_mplock(); 457 goto bad; 458 } 459 if (m == NULL) { 460 rel_mplock(); 461 return (0); 462 } 463 } 464 rel_mplock(); 465 } 466 467 /* Continue with link-layer output */ 468 return ether_output_frame(ifp, m); 469 470 bad: 471 m_freem(m); 472 return (error); 473 } 474 475 /* 476 * Ethernet link layer output routine to send a raw frame to the device. 477 * 478 * This assumes that the 14 byte Ethernet header is present and contiguous 479 * in the first mbuf. 480 */ 481 int 482 ether_output_frame(struct ifnet *ifp, struct mbuf *m) 483 { 484 struct ip_fw *rule = NULL; 485 int error = 0; 486 struct altq_pktattr pktattr; 487 488 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp); 489 490 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 491 struct m_tag *mtag; 492 493 /* Extract info from dummynet tag */ 494 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 495 KKASSERT(mtag != NULL); 496 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 497 KKASSERT(rule != NULL); 498 499 m_tag_delete(m, mtag); 500 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 501 } 502 503 if (ifq_is_enabled(&ifp->if_snd)) 504 altq_etherclassify(&ifp->if_snd, m, &pktattr); 505 crit_enter(); 506 if (IPFW_LOADED && ether_ipfw != 0) { 507 struct ether_header save_eh, *eh; 508 509 eh = mtod(m, struct ether_header *); 510 save_eh = *eh; 511 m_adj(m, ETHER_HDR_LEN); 512 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) { 513 crit_exit(); 514 if (m != NULL) { 515 m_freem(m); 516 return ENOBUFS; /* pkt dropped */ 517 } else 518 return 0; /* consumed e.g. in a pipe */ 519 } 520 521 /* packet was ok, restore the ethernet header */ 522 ether_restore_header(&m, eh, &save_eh); 523 if (m == NULL) { 524 crit_exit(); 525 return ENOBUFS; 526 } 527 } 528 crit_exit(); 529 530 /* 531 * Queue message on interface, update output statistics if 532 * successful, and start output if interface not yet active. 533 */ 534 error = ifq_dispatch(ifp, m, &pktattr); 535 return (error); 536 } 537 538 /* 539 * ipfw processing for ethernet packets (in and out). 540 * The second parameter is NULL from ether_demux(), and ifp from 541 * ether_output_frame(). 542 */ 543 static boolean_t 544 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule, 545 const struct ether_header *eh) 546 { 547 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */ 548 struct ip_fw_args args; 549 struct m_tag *mtag; 550 struct mbuf *m; 551 int i; 552 553 if (*rule != NULL && fw_one_pass) 554 return TRUE; /* dummynet packet, already partially processed */ 555 556 /* 557 * I need some amount of data to be contiguous. 558 */ 559 i = min((*m0)->m_pkthdr.len, max_protohdr); 560 if ((*m0)->m_len < i) { 561 *m0 = m_pullup(*m0, i); 562 if (*m0 == NULL) 563 return FALSE; 564 } 565 566 /* 567 * Clean up tags 568 */ 569 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) 570 m_tag_delete(*m0, mtag); 571 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 572 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL); 573 KKASSERT(mtag != NULL); 574 m_tag_delete(*m0, mtag); 575 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED; 576 } 577 578 args.m = *m0; /* the packet we are looking at */ 579 args.oif = dst; /* destination, if any */ 580 args.rule = *rule; /* matching rule to restart */ 581 args.eh = &save_eh; /* MAC header for bridged/MAC packets */ 582 i = ip_fw_chk_ptr(&args); 583 *m0 = args.m; 584 *rule = args.rule; 585 586 if (*m0 == NULL) 587 return FALSE; 588 589 switch (i) { 590 case IP_FW_PASS: 591 return TRUE; 592 593 case IP_FW_DIVERT: 594 case IP_FW_TEE: 595 case IP_FW_DENY: 596 /* 597 * XXX at some point add support for divert/forward actions. 598 * If none of the above matches, we have to drop the pkt. 599 */ 600 return FALSE; 601 602 case IP_FW_DUMMYNET: 603 /* 604 * Pass the pkt to dummynet, which consumes it. 605 */ 606 m = *m0; /* pass the original to dummynet */ 607 *m0 = NULL; /* and nothing back to the caller */ 608 609 ether_restore_header(&m, eh, &save_eh); 610 if (m == NULL) 611 return FALSE; 612 613 ip_fw_dn_io_ptr(m, args.cookie, 614 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args); 615 ip_dn_queue(m); 616 return FALSE; 617 618 default: 619 panic("unknown ipfw return value: %d\n", i); 620 } 621 } 622 623 static void 624 ether_input(struct ifnet *ifp, struct mbuf *m) 625 { 626 ether_input_chain(ifp, m, NULL, NULL); 627 } 628 629 /* 630 * Perform common duties while attaching to interface list 631 */ 632 void 633 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer) 634 { 635 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header), 636 serializer); 637 } 638 639 void 640 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen, 641 lwkt_serialize_t serializer) 642 { 643 struct sockaddr_dl *sdl; 644 645 ifp->if_type = IFT_ETHER; 646 ifp->if_addrlen = ETHER_ADDR_LEN; 647 ifp->if_hdrlen = ETHER_HDR_LEN; 648 if_attach(ifp, serializer); 649 ifp->if_mtu = ETHERMTU; 650 if (ifp->if_baudrate == 0) 651 ifp->if_baudrate = 10000000; 652 ifp->if_output = ether_output; 653 ifp->if_input = ether_input; 654 ifp->if_resolvemulti = ether_resolvemulti; 655 ifp->if_broadcastaddr = etherbroadcastaddr; 656 sdl = IF_LLSOCKADDR(ifp); 657 sdl->sdl_type = IFT_ETHER; 658 sdl->sdl_alen = ifp->if_addrlen; 659 bcopy(lla, LLADDR(sdl), ifp->if_addrlen); 660 /* 661 * XXX Keep the current drivers happy. 662 * XXX Remove once all drivers have been cleaned up 663 */ 664 if (lla != IFP2AC(ifp)->ac_enaddr) 665 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen); 666 bpfattach(ifp, dlt, hdrlen); 667 if (ng_ether_attach_p != NULL) 668 (*ng_ether_attach_p)(ifp); 669 670 if_printf(ifp, "MAC address: %6D\n", lla, ":"); 671 } 672 673 /* 674 * Perform common duties while detaching an Ethernet interface 675 */ 676 void 677 ether_ifdetach(struct ifnet *ifp) 678 { 679 if_down(ifp); 680 681 if (ng_ether_detach_p != NULL) 682 (*ng_ether_detach_p)(ifp); 683 bpfdetach(ifp); 684 if_detach(ifp); 685 } 686 687 int 688 ether_ioctl(struct ifnet *ifp, int command, caddr_t data) 689 { 690 struct ifaddr *ifa = (struct ifaddr *) data; 691 struct ifreq *ifr = (struct ifreq *) data; 692 int error = 0; 693 694 #define IF_INIT(ifp) \ 695 do { \ 696 if (((ifp)->if_flags & IFF_UP) == 0) { \ 697 (ifp)->if_flags |= IFF_UP; \ 698 (ifp)->if_init((ifp)->if_softc); \ 699 } \ 700 } while (0) 701 702 ASSERT_IFNET_SERIALIZED_ALL(ifp); 703 704 switch (command) { 705 case SIOCSIFADDR: 706 switch (ifa->ifa_addr->sa_family) { 707 #ifdef INET 708 case AF_INET: 709 IF_INIT(ifp); /* before arpwhohas */ 710 arp_ifinit(ifp, ifa); 711 break; 712 #endif 713 #ifdef IPX 714 /* 715 * XXX - This code is probably wrong 716 */ 717 case AF_IPX: 718 { 719 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr; 720 struct arpcom *ac = IFP2AC(ifp); 721 722 if (ipx_nullhost(*ina)) 723 ina->x_host = *(union ipx_host *) ac->ac_enaddr; 724 else 725 bcopy(ina->x_host.c_host, ac->ac_enaddr, 726 sizeof ac->ac_enaddr); 727 728 IF_INIT(ifp); /* Set new address. */ 729 break; 730 } 731 #endif 732 #ifdef NS 733 /* 734 * XXX - This code is probably wrong 735 */ 736 case AF_NS: 737 { 738 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr); 739 struct arpcom *ac = IFP2AC(ifp); 740 741 if (ns_nullhost(*ina)) 742 ina->x_host = *(union ns_host *)(ac->ac_enaddr); 743 else 744 bcopy(ina->x_host.c_host, ac->ac_enaddr, 745 sizeof ac->ac_enaddr); 746 747 /* 748 * Set new address 749 */ 750 IF_INIT(ifp); 751 break; 752 } 753 #endif 754 default: 755 IF_INIT(ifp); 756 break; 757 } 758 break; 759 760 case SIOCGIFADDR: 761 bcopy(IFP2AC(ifp)->ac_enaddr, 762 ((struct sockaddr *)ifr->ifr_data)->sa_data, 763 ETHER_ADDR_LEN); 764 break; 765 766 case SIOCSIFMTU: 767 /* 768 * Set the interface MTU. 769 */ 770 if (ifr->ifr_mtu > ETHERMTU) { 771 error = EINVAL; 772 } else { 773 ifp->if_mtu = ifr->ifr_mtu; 774 } 775 break; 776 default: 777 error = EINVAL; 778 break; 779 } 780 return (error); 781 782 #undef IF_INIT 783 } 784 785 int 786 ether_resolvemulti( 787 struct ifnet *ifp, 788 struct sockaddr **llsa, 789 struct sockaddr *sa) 790 { 791 struct sockaddr_dl *sdl; 792 struct sockaddr_in *sin; 793 #ifdef INET6 794 struct sockaddr_in6 *sin6; 795 #endif 796 u_char *e_addr; 797 798 switch(sa->sa_family) { 799 case AF_LINK: 800 /* 801 * No mapping needed. Just check that it's a valid MC address. 802 */ 803 sdl = (struct sockaddr_dl *)sa; 804 e_addr = LLADDR(sdl); 805 if ((e_addr[0] & 1) != 1) 806 return EADDRNOTAVAIL; 807 *llsa = 0; 808 return 0; 809 810 #ifdef INET 811 case AF_INET: 812 sin = (struct sockaddr_in *)sa; 813 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) 814 return EADDRNOTAVAIL; 815 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 816 M_WAITOK | M_ZERO); 817 sdl->sdl_len = sizeof *sdl; 818 sdl->sdl_family = AF_LINK; 819 sdl->sdl_index = ifp->if_index; 820 sdl->sdl_type = IFT_ETHER; 821 sdl->sdl_alen = ETHER_ADDR_LEN; 822 e_addr = LLADDR(sdl); 823 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr); 824 *llsa = (struct sockaddr *)sdl; 825 return 0; 826 #endif 827 #ifdef INET6 828 case AF_INET6: 829 sin6 = (struct sockaddr_in6 *)sa; 830 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 831 /* 832 * An IP6 address of 0 means listen to all 833 * of the Ethernet multicast address used for IP6. 834 * (This is used for multicast routers.) 835 */ 836 ifp->if_flags |= IFF_ALLMULTI; 837 *llsa = 0; 838 return 0; 839 } 840 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 841 return EADDRNOTAVAIL; 842 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR, 843 M_WAITOK | M_ZERO); 844 sdl->sdl_len = sizeof *sdl; 845 sdl->sdl_family = AF_LINK; 846 sdl->sdl_index = ifp->if_index; 847 sdl->sdl_type = IFT_ETHER; 848 sdl->sdl_alen = ETHER_ADDR_LEN; 849 e_addr = LLADDR(sdl); 850 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr); 851 *llsa = (struct sockaddr *)sdl; 852 return 0; 853 #endif 854 855 default: 856 /* 857 * Well, the text isn't quite right, but it's the name 858 * that counts... 859 */ 860 return EAFNOSUPPORT; 861 } 862 } 863 864 #if 0 865 /* 866 * This is for reference. We have a table-driven version 867 * of the little-endian crc32 generator, which is faster 868 * than the double-loop. 869 */ 870 uint32_t 871 ether_crc32_le(const uint8_t *buf, size_t len) 872 { 873 uint32_t c, crc, carry; 874 size_t i, j; 875 876 crc = 0xffffffffU; /* initial value */ 877 878 for (i = 0; i < len; i++) { 879 c = buf[i]; 880 for (j = 0; j < 8; j++) { 881 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01); 882 crc >>= 1; 883 c >>= 1; 884 if (carry) 885 crc = (crc ^ ETHER_CRC_POLY_LE); 886 } 887 } 888 889 return (crc); 890 } 891 #else 892 uint32_t 893 ether_crc32_le(const uint8_t *buf, size_t len) 894 { 895 static const uint32_t crctab[] = { 896 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 897 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, 898 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 899 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c 900 }; 901 uint32_t crc; 902 size_t i; 903 904 crc = 0xffffffffU; /* initial value */ 905 906 for (i = 0; i < len; i++) { 907 crc ^= buf[i]; 908 crc = (crc >> 4) ^ crctab[crc & 0xf]; 909 crc = (crc >> 4) ^ crctab[crc & 0xf]; 910 } 911 912 return (crc); 913 } 914 #endif 915 916 uint32_t 917 ether_crc32_be(const uint8_t *buf, size_t len) 918 { 919 uint32_t c, crc, carry; 920 size_t i, j; 921 922 crc = 0xffffffffU; /* initial value */ 923 924 for (i = 0; i < len; i++) { 925 c = buf[i]; 926 for (j = 0; j < 8; j++) { 927 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01); 928 crc <<= 1; 929 c >>= 1; 930 if (carry) 931 crc = (crc ^ ETHER_CRC_POLY_BE) | carry; 932 } 933 } 934 935 return (crc); 936 } 937 938 /* 939 * find the size of ethernet header, and call classifier 940 */ 941 void 942 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m, 943 struct altq_pktattr *pktattr) 944 { 945 struct ether_header *eh; 946 uint16_t ether_type; 947 int hlen, af, hdrsize; 948 caddr_t hdr; 949 950 hlen = sizeof(struct ether_header); 951 eh = mtod(m, struct ether_header *); 952 953 ether_type = ntohs(eh->ether_type); 954 if (ether_type < ETHERMTU) { 955 /* ick! LLC/SNAP */ 956 struct llc *llc = (struct llc *)(eh + 1); 957 hlen += 8; 958 959 if (m->m_len < hlen || 960 llc->llc_dsap != LLC_SNAP_LSAP || 961 llc->llc_ssap != LLC_SNAP_LSAP || 962 llc->llc_control != LLC_UI) 963 goto bad; /* not snap! */ 964 965 ether_type = ntohs(llc->llc_un.type_snap.ether_type); 966 } 967 968 if (ether_type == ETHERTYPE_IP) { 969 af = AF_INET; 970 hdrsize = 20; /* sizeof(struct ip) */ 971 #ifdef INET6 972 } else if (ether_type == ETHERTYPE_IPV6) { 973 af = AF_INET6; 974 hdrsize = 40; /* sizeof(struct ip6_hdr) */ 975 #endif 976 } else 977 goto bad; 978 979 while (m->m_len <= hlen) { 980 hlen -= m->m_len; 981 m = m->m_next; 982 } 983 hdr = m->m_data + hlen; 984 if (m->m_len < hlen + hdrsize) { 985 /* 986 * ip header is not in a single mbuf. this should not 987 * happen in the current code. 988 * (todo: use m_pulldown in the future) 989 */ 990 goto bad; 991 } 992 m->m_data += hlen; 993 m->m_len -= hlen; 994 ifq_classify(ifq, m, af, pktattr); 995 m->m_data -= hlen; 996 m->m_len += hlen; 997 998 return; 999 1000 bad: 1001 pktattr->pattr_class = NULL; 1002 pktattr->pattr_hdr = NULL; 1003 pktattr->pattr_af = AF_UNSPEC; 1004 } 1005 1006 static void 1007 ether_restore_header(struct mbuf **m0, const struct ether_header *eh, 1008 const struct ether_header *save_eh) 1009 { 1010 struct mbuf *m = *m0; 1011 1012 ether_restore_hdr++; 1013 1014 /* 1015 * Prepend the header, optimize for the common case of 1016 * eh pointing into the mbuf. 1017 */ 1018 if ((const void *)(eh + 1) == (void *)m->m_data) { 1019 m->m_data -= ETHER_HDR_LEN; 1020 m->m_len += ETHER_HDR_LEN; 1021 m->m_pkthdr.len += ETHER_HDR_LEN; 1022 } else { 1023 ether_prepend_hdr++; 1024 1025 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT); 1026 if (m != NULL) { 1027 bcopy(save_eh, mtod(m, struct ether_header *), 1028 ETHER_HDR_LEN); 1029 } 1030 } 1031 *m0 = m; 1032 } 1033 1034 static void 1035 ether_input_ipifunc(void *arg) 1036 { 1037 struct mbuf *m, *next; 1038 lwkt_port_t port = cpu_portfn(mycpu->gd_cpuid); 1039 1040 m = arg; 1041 do { 1042 next = m->m_nextpkt; 1043 m->m_nextpkt = NULL; 1044 lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.base.lmsg); 1045 m = next; 1046 } while (m != NULL); 1047 } 1048 1049 void 1050 ether_input_dispatch(struct mbuf_chain *chain) 1051 { 1052 #ifdef SMP 1053 int i; 1054 1055 logether(disp_beg, NULL); 1056 for (i = 0; i < ncpus; ++i) { 1057 if (chain[i].mc_head != NULL) { 1058 lwkt_send_ipiq(globaldata_find(i), 1059 ether_input_ipifunc, chain[i].mc_head); 1060 } 1061 } 1062 #else 1063 logether(disp_beg, NULL); 1064 if (chain->mc_head != NULL) 1065 ether_input_ipifunc(chain->mc_head); 1066 #endif 1067 logether(disp_end, NULL); 1068 } 1069 1070 void 1071 ether_input_chain_init(struct mbuf_chain *chain) 1072 { 1073 #ifdef SMP 1074 int i; 1075 1076 for (i = 0; i < ncpus; ++i) 1077 chain[i].mc_head = chain[i].mc_tail = NULL; 1078 #else 1079 chain->mc_head = chain->mc_tail = NULL; 1080 #endif 1081 } 1082 1083 /* 1084 * Upper layer processing for a received Ethernet packet. 1085 */ 1086 void 1087 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m) 1088 { 1089 struct ether_header *eh; 1090 int isr, discard = 0; 1091 u_short ether_type; 1092 struct ip_fw *rule = NULL; 1093 #ifdef NETATALK 1094 struct llc *l; 1095 #endif 1096 1097 M_ASSERTPKTHDR(m); 1098 KASSERT(m->m_len >= ETHER_HDR_LEN, 1099 ("ether header is no contiguous!\n")); 1100 1101 eh = mtod(m, struct ether_header *); 1102 1103 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 1104 struct m_tag *mtag; 1105 1106 /* Extract info from dummynet tag */ 1107 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 1108 KKASSERT(mtag != NULL); 1109 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv; 1110 KKASSERT(rule != NULL); 1111 1112 m_tag_delete(m, mtag); 1113 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 1114 1115 /* packet is passing the second time */ 1116 goto post_stats; 1117 } 1118 1119 #ifdef CARP 1120 /* 1121 * XXX: Okay, we need to call carp_forus() and - if it is for 1122 * us jump over code that does the normal check 1123 * "ac_enaddr == ether_dhost". The check sequence is a bit 1124 * different from OpenBSD, so we jump over as few code as 1125 * possible, to catch _all_ sanity checks. This needs 1126 * evaluation, to see if the carp ether_dhost values break any 1127 * of these checks! 1128 */ 1129 if (ifp->if_carp) { 1130 /* 1131 * Hold BGL and recheck ifp->if_carp 1132 */ 1133 get_mplock(); 1134 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) { 1135 rel_mplock(); 1136 goto post_stats; 1137 } 1138 rel_mplock(); 1139 } 1140 #endif 1141 1142 /* 1143 * We got a packet which was unicast to a different Ethernet 1144 * address. If the driver is working properly, then this 1145 * situation can only happen when the interface is in 1146 * promiscuous mode. We defer the packet discarding until the 1147 * vlan processing is done, so that vlan/bridge or vlan/netgraph 1148 * could work. 1149 */ 1150 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) && 1151 !ETHER_IS_MULTICAST(eh->ether_dhost) && 1152 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) 1153 discard = 1; 1154 1155 post_stats: 1156 if (IPFW_LOADED && ether_ipfw != 0 && !discard) { 1157 struct ether_header save_eh = *eh; 1158 1159 /* XXX old crufty stuff, needs to be removed */ 1160 m_adj(m, sizeof(struct ether_header)); 1161 1162 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) { 1163 m_freem(m); 1164 return; 1165 } 1166 1167 ether_restore_header(&m, eh, &save_eh); 1168 if (m == NULL) 1169 return; 1170 eh = mtod(m, struct ether_header *); 1171 } 1172 1173 ether_type = ntohs(eh->ether_type); 1174 KKASSERT(ether_type != ETHERTYPE_VLAN); 1175 1176 if (m->m_flags & M_VLANTAG) { 1177 void (*vlan_input_func)(struct mbuf *); 1178 1179 vlan_input_func = vlan_input_p; 1180 if (vlan_input_func != NULL) { 1181 vlan_input_func(m); 1182 } else { 1183 m->m_pkthdr.rcvif->if_noproto++; 1184 m_freem(m); 1185 } 1186 return; 1187 } 1188 1189 /* 1190 * If we have been asked to discard this packet 1191 * (e.g. not for us), drop it before entering 1192 * the upper layer. 1193 */ 1194 if (discard) { 1195 m_freem(m); 1196 return; 1197 } 1198 1199 /* 1200 * Clear protocol specific flags, 1201 * before entering the upper layer. 1202 */ 1203 m->m_flags &= ~M_ETHER_FLAGS; 1204 1205 /* Strip ethernet header. */ 1206 m_adj(m, sizeof(struct ether_header)); 1207 1208 switch (ether_type) { 1209 #ifdef INET 1210 case ETHERTYPE_IP: 1211 if ((m->m_flags & M_LENCHECKED) == 0) { 1212 if (!ip_lengthcheck(&m, 0)) 1213 return; 1214 } 1215 if (ipflow_fastforward(m)) 1216 return; 1217 isr = NETISR_IP; 1218 break; 1219 1220 case ETHERTYPE_ARP: 1221 if (ifp->if_flags & IFF_NOARP) { 1222 /* Discard packet if ARP is disabled on interface */ 1223 m_freem(m); 1224 return; 1225 } 1226 isr = NETISR_ARP; 1227 break; 1228 #endif 1229 1230 #ifdef INET6 1231 case ETHERTYPE_IPV6: 1232 isr = NETISR_IPV6; 1233 break; 1234 #endif 1235 1236 #ifdef IPX 1237 case ETHERTYPE_IPX: 1238 if (ef_inputp) { 1239 /* 1240 * Hold BGL and recheck ef_inputp 1241 */ 1242 get_mplock(); 1243 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1244 rel_mplock(); 1245 return; 1246 } 1247 rel_mplock(); 1248 } 1249 isr = NETISR_IPX; 1250 break; 1251 #endif 1252 1253 #ifdef NS 1254 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 1255 isr = NETISR_NS; 1256 break; 1257 1258 #endif 1259 1260 #ifdef NETATALK 1261 case ETHERTYPE_AT: 1262 isr = NETISR_ATALK1; 1263 break; 1264 case ETHERTYPE_AARP: 1265 isr = NETISR_AARP; 1266 break; 1267 #endif 1268 1269 #ifdef MPLS 1270 case ETHERTYPE_MPLS: 1271 case ETHERTYPE_MPLS_MCAST: 1272 /* Should have been set by ether_input_chain(). */ 1273 KKASSERT(m->m_flags & M_MPLSLABELED); 1274 isr = NETISR_MPLS; 1275 break; 1276 #endif 1277 1278 default: 1279 /* 1280 * The accurate msgport is not determined before 1281 * we reach here, so redo the dispatching 1282 */ 1283 #ifdef IPX 1284 if (ef_inputp) { 1285 /* 1286 * Hold BGL and recheck ef_inputp 1287 */ 1288 get_mplock(); 1289 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) { 1290 rel_mplock(); 1291 return; 1292 } 1293 rel_mplock(); 1294 } 1295 #endif 1296 #ifdef NS 1297 checksum = mtod(m, ushort *); 1298 /* Novell 802.3 */ 1299 if ((ether_type <= ETHERMTU) && 1300 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) { 1301 if (*checksum == 0xE0E0) { 1302 m->m_pkthdr.len -= 3; 1303 m->m_len -= 3; 1304 m->m_data += 3; 1305 } 1306 isr = NETISR_NS; 1307 break; 1308 } 1309 #endif 1310 #ifdef NETATALK 1311 if (ether_type > ETHERMTU) 1312 goto dropanyway; 1313 l = mtod(m, struct llc *); 1314 if (l->llc_dsap == LLC_SNAP_LSAP && 1315 l->llc_ssap == LLC_SNAP_LSAP && 1316 l->llc_control == LLC_UI) { 1317 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code, 1318 sizeof at_org_code) == 0 && 1319 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) { 1320 m_adj(m, sizeof(struct llc)); 1321 isr = NETISR_ATALK2; 1322 break; 1323 } 1324 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code, 1325 sizeof aarp_org_code) == 0 && 1326 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) { 1327 m_adj(m, sizeof(struct llc)); 1328 isr = NETISR_AARP; 1329 break; 1330 } 1331 } 1332 dropanyway: 1333 #endif 1334 if (ng_ether_input_orphan_p != NULL) { 1335 /* 1336 * Hold BGL and recheck ng_ether_input_orphan_p 1337 */ 1338 get_mplock(); 1339 if (ng_ether_input_orphan_p != NULL) { 1340 ng_ether_input_orphan_p(ifp, m, eh); 1341 rel_mplock(); 1342 return; 1343 } 1344 rel_mplock(); 1345 } 1346 m_freem(m); 1347 return; 1348 } 1349 1350 netisr_queue(isr, m); 1351 } 1352 1353 /* 1354 * First we perform any link layer operations, then continue to the 1355 * upper layers with ether_demux_oncpu(). 1356 */ 1357 static void 1358 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m) 1359 { 1360 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) { 1361 /* 1362 * Receiving interface's flags are changed, when this 1363 * packet is waiting for processing; discard it. 1364 */ 1365 m_freem(m); 1366 return; 1367 } 1368 1369 /* 1370 * Tap the packet off here for a bridge. bridge_input() 1371 * will return NULL if it has consumed the packet, otherwise 1372 * it gets processed as normal. Note that bridge_input() 1373 * will always return the original packet if we need to 1374 * process it locally. 1375 */ 1376 if (ifp->if_bridge) { 1377 KASSERT(bridge_input_p != NULL, 1378 ("%s: if_bridge not loaded!", __func__)); 1379 1380 if(m->m_flags & M_ETHER_BRIDGED) { 1381 m->m_flags &= ~M_ETHER_BRIDGED; 1382 } else { 1383 m = bridge_input_p(ifp, m); 1384 if (m == NULL) 1385 return; 1386 1387 KASSERT(ifp == m->m_pkthdr.rcvif, 1388 ("bridge_input_p changed rcvif\n")); 1389 } 1390 } 1391 1392 /* Handle ng_ether(4) processing, if any */ 1393 if (ng_ether_input_p != NULL) { 1394 /* 1395 * Hold BGL and recheck ng_ether_input_p 1396 */ 1397 get_mplock(); 1398 if (ng_ether_input_p != NULL) 1399 ng_ether_input_p(ifp, &m); 1400 rel_mplock(); 1401 1402 if (m == NULL) 1403 return; 1404 } 1405 1406 /* Continue with upper layer processing */ 1407 ether_demux_oncpu(ifp, m); 1408 } 1409 1410 /* 1411 * Perform certain functions of ether_input_chain(): 1412 * - Test IFF_UP 1413 * - Update statistics 1414 * - Run bpf(4) tap if requested 1415 * Then pass the packet to ether_input_oncpu(). 1416 * 1417 * This function should be used by pseudo interface (e.g. vlan(4)), 1418 * when it tries to claim that the packet is received by it. 1419 */ 1420 void 1421 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int run_bpf) 1422 { 1423 /* Discard packet if interface is not up */ 1424 if (!(ifp->if_flags & IFF_UP)) { 1425 m_freem(m); 1426 return; 1427 } 1428 1429 /* Change receiving interface */ 1430 m->m_pkthdr.rcvif = ifp; 1431 1432 /* Update statistics */ 1433 ifp->if_ipackets++; 1434 ifp->if_ibytes += m->m_pkthdr.len; 1435 if (m->m_flags & (M_MCAST | M_BCAST)) 1436 ifp->if_imcasts++; 1437 1438 if (run_bpf) 1439 BPF_MTAP(ifp, m); 1440 1441 ether_input_oncpu(ifp, m); 1442 } 1443 1444 static __inline boolean_t 1445 ether_vlancheck(struct mbuf **m0) 1446 { 1447 struct mbuf *m = *m0; 1448 struct ether_header *eh; 1449 uint16_t ether_type; 1450 1451 eh = mtod(m, struct ether_header *); 1452 ether_type = ntohs(eh->ether_type); 1453 1454 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) { 1455 /* 1456 * Extract vlan tag if hardware does not do it for us 1457 */ 1458 vlan_ether_decap(&m); 1459 if (m == NULL) 1460 goto failed; 1461 1462 eh = mtod(m, struct ether_header *); 1463 ether_type = ntohs(eh->ether_type); 1464 } 1465 1466 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) { 1467 /* 1468 * To prevent possible dangerous recursion, 1469 * we don't do vlan-in-vlan 1470 */ 1471 m->m_pkthdr.rcvif->if_noproto++; 1472 goto failed; 1473 } 1474 KKASSERT(ether_type != ETHERTYPE_VLAN); 1475 1476 m->m_flags |= M_ETHER_VLANCHECKED; 1477 *m0 = m; 1478 return TRUE; 1479 failed: 1480 if (m != NULL) 1481 m_freem(m); 1482 *m0 = NULL; 1483 return FALSE; 1484 } 1485 1486 static void 1487 ether_input_handler(netmsg_t nmsg) 1488 { 1489 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */ 1490 struct ether_header *eh; 1491 struct ifnet *ifp; 1492 struct mbuf *m; 1493 1494 m = nmp->nm_packet; 1495 M_ASSERTPKTHDR(m); 1496 ifp = m->m_pkthdr.rcvif; 1497 1498 eh = mtod(m, struct ether_header *); 1499 if (ETHER_IS_MULTICAST(eh->ether_dhost)) { 1500 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost, 1501 ifp->if_addrlen) == 0) 1502 m->m_flags |= M_BCAST; 1503 else 1504 m->m_flags |= M_MCAST; 1505 ifp->if_imcasts++; 1506 } 1507 1508 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) { 1509 if (!ether_vlancheck(&m)) { 1510 KKASSERT(m == NULL); 1511 return; 1512 } 1513 } 1514 1515 ether_input_oncpu(ifp, m); 1516 } 1517 1518 /* 1519 * Send the packet to the target msgport or queue it into 'chain'. 1520 * 1521 * At this point the packet had better be characterized (M_HASH set), 1522 * so we know which cpu to send it to. 1523 */ 1524 static void 1525 ether_dispatch(int isr, struct mbuf *m, struct mbuf_chain *chain) 1526 { 1527 struct netmsg_packet *pmsg; 1528 1529 KKASSERT(m->m_flags & M_HASH); 1530 pmsg = &m->m_hdr.mh_netmsg; 1531 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1532 0, ether_input_handler); 1533 pmsg->nm_packet = m; 1534 pmsg->base.lmsg.u.ms_result = isr; 1535 1536 if (chain != NULL) { 1537 int cpuid = m->m_pkthdr.hash; 1538 struct mbuf_chain *c; 1539 1540 c = &chain[cpuid]; 1541 if (c->mc_head == NULL) { 1542 c->mc_head = c->mc_tail = m; 1543 } else { 1544 c->mc_tail->m_nextpkt = m; 1545 c->mc_tail = m; 1546 } 1547 m->m_nextpkt = NULL; 1548 } else { 1549 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg); 1550 } 1551 } 1552 1553 /* 1554 * Process a received Ethernet packet. 1555 * 1556 * The ethernet header is assumed to be in the mbuf so the caller 1557 * MUST MAKE SURE that there are at least sizeof(struct ether_header) 1558 * bytes in the first mbuf. 1559 * 1560 * - If 'chain' is NULL, this ether frame is sent to the target msgport 1561 * immediately. This situation happens when ether_input_chain is 1562 * accessed through ifnet.if_input. 1563 * 1564 * - If 'chain' is not NULL, this ether frame is queued to the 'chain' 1565 * bucket indexed by the target msgport's cpuid and the target msgport 1566 * is saved in mbuf's m_pkthdr.m_head. Caller of ether_input_chain 1567 * must initialize 'chain' by calling ether_input_chain_init(). 1568 * ether_input_dispatch must be called later to send ether frames 1569 * queued on 'chain' to their target msgport. 1570 */ 1571 void 1572 ether_input_chain(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi, 1573 struct mbuf_chain *chain) 1574 { 1575 struct ether_header *eh; 1576 uint16_t ether_type; 1577 int isr; 1578 1579 M_ASSERTPKTHDR(m); 1580 1581 /* Discard packet if interface is not up */ 1582 if (!(ifp->if_flags & IFF_UP)) { 1583 m_freem(m); 1584 return; 1585 } 1586 1587 if (m->m_len < sizeof(struct ether_header)) { 1588 /* XXX error in the caller. */ 1589 m_freem(m); 1590 return; 1591 } 1592 1593 m->m_pkthdr.rcvif = ifp; 1594 1595 logether(chain_beg, ifp); 1596 1597 ETHER_BPF_MTAP(ifp, m); 1598 1599 ifp->if_ibytes += m->m_pkthdr.len; 1600 1601 if (ifp->if_flags & IFF_MONITOR) { 1602 eh = mtod(m, struct ether_header *); 1603 if (ETHER_IS_MULTICAST(eh->ether_dhost)) 1604 ifp->if_imcasts++; 1605 1606 /* 1607 * Interface marked for monitoring; discard packet. 1608 */ 1609 m_freem(m); 1610 1611 logether(chain_end, ifp); 1612 return; 1613 } 1614 1615 /* 1616 * If the packet has been characterized (pi->pi_netisr / M_HASH) 1617 * we can dispatch it immediately without further inspection. 1618 */ 1619 if (pi != NULL && (m->m_flags & M_HASH)) { 1620 #ifdef RSS_DEBUG 1621 ether_pktinfo_try++; 1622 #endif 1623 ether_dispatch(pi->pi_netisr, m, chain); 1624 1625 #ifdef RSS_DEBUG 1626 ether_pktinfo_hit++; 1627 #endif 1628 logether(chain_end, ifp); 1629 return; 1630 } 1631 #ifdef RSS_DEBUG 1632 else if (ifp->if_capenable & IFCAP_RSS) { 1633 if (pi == NULL) 1634 ether_rss_nopi++; 1635 else 1636 ether_rss_nohash++; 1637 } 1638 #endif 1639 1640 /* 1641 * Packet hash will be recalculated by software, 1642 * so clear the M_HASH flag set by the driver; 1643 * the hash value calculated by the hardware may 1644 * not be exactly what we want. 1645 */ 1646 m->m_flags &= ~M_HASH; 1647 1648 if (!ether_vlancheck(&m)) { 1649 KKASSERT(m == NULL); 1650 logether(chain_end, ifp); 1651 return; 1652 } 1653 eh = mtod(m, struct ether_header *); 1654 ether_type = ntohs(eh->ether_type); 1655 1656 /* 1657 * Map ether type to netisr id. 1658 */ 1659 switch (ether_type) { 1660 #ifdef INET 1661 case ETHERTYPE_IP: 1662 isr = NETISR_IP; 1663 break; 1664 1665 case ETHERTYPE_ARP: 1666 isr = NETISR_ARP; 1667 break; 1668 #endif 1669 1670 #ifdef INET6 1671 case ETHERTYPE_IPV6: 1672 isr = NETISR_IPV6; 1673 break; 1674 #endif 1675 1676 #ifdef IPX 1677 case ETHERTYPE_IPX: 1678 isr = NETISR_IPX; 1679 break; 1680 #endif 1681 1682 #ifdef NS 1683 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */ 1684 isr = NETISR_NS; 1685 break; 1686 #endif 1687 1688 #ifdef NETATALK 1689 case ETHERTYPE_AT: 1690 isr = NETISR_ATALK1; 1691 break; 1692 case ETHERTYPE_AARP: 1693 isr = NETISR_AARP; 1694 break; 1695 #endif 1696 1697 #ifdef MPLS 1698 case ETHERTYPE_MPLS: 1699 case ETHERTYPE_MPLS_MCAST: 1700 m->m_flags |= M_MPLSLABELED; 1701 isr = NETISR_MPLS; 1702 break; 1703 #endif 1704 1705 default: 1706 /* 1707 * NETISR_MAX is an invalid value; it is chosen to let 1708 */ 1709 isr = NETISR_MAX; 1710 break; 1711 } 1712 1713 /* 1714 * Ask the isr to characterize the packet since we couldn't. 1715 * This is an attempt to optimally get us onto the correct protocol 1716 * thread. 1717 */ 1718 netisr_characterize(isr, &m, sizeof(struct ether_header)); 1719 if (m == NULL) { 1720 logether(chain_end, ifp); 1721 return; 1722 } 1723 1724 /* 1725 * Finally dispatch it 1726 */ 1727 ether_dispatch(isr, m, chain); 1728 1729 logether(chain_end, ifp); 1730 } 1731 1732 MODULE_VERSION(ether, 1); 1733