xref: /freebsd/sys/net/if_ethersubr.c (revision 0efd6615cd5f39b67cec82a7034e655f3b5801e3)
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  * 4. 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  * $FreeBSD$
31  */
32 
33 #include "opt_atalk.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_ipx.h"
37 #include "opt_mac.h"
38 #include "opt_netgraph.h"
39 #include "opt_carp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
45 #include <sys/module.h>
46 #include <sys/mbuf.h>
47 #include <sys/random.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/if_arp.h>
54 #include <net/netisr.h>
55 #include <net/route.h>
56 #include <net/if_llc.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/bpf.h>
60 #include <net/ethernet.h>
61 #include <net/if_bridgevar.h>
62 #include <net/if_vlan_var.h>
63 
64 #if defined(INET) || defined(INET6)
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/ip_fw.h>
69 #include <netinet/ip_dummynet.h>
70 #endif
71 #ifdef INET6
72 #include <netinet6/nd6.h>
73 #endif
74 
75 #ifdef DEV_CARP
76 #include <netinet/ip_carp.h>
77 #endif
78 
79 #ifdef IPX
80 #include <netipx/ipx.h>
81 #include <netipx/ipx_if.h>
82 #endif
83 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
84 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
85 		struct sockaddr *dst, short *tp, int *hlen);
86 
87 #ifdef NETATALK
88 #include <netatalk/at.h>
89 #include <netatalk/at_var.h>
90 #include <netatalk/at_extern.h>
91 
92 #define llc_snap_org_code llc_un.type_snap.org_code
93 #define llc_snap_ether_type llc_un.type_snap.ether_type
94 
95 extern u_char	at_org_code[3];
96 extern u_char	aarp_org_code[3];
97 #endif /* NETATALK */
98 
99 #include <security/mac/mac_framework.h>
100 
101 /* netgraph node hooks for ng_ether(4) */
102 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void	(*ng_ether_attach_p)(struct ifnet *ifp);
106 void	(*ng_ether_detach_p)(struct ifnet *ifp);
107 
108 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
109 
110 /* if_bridge(4) support */
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
113 		struct sockaddr *, struct rtentry *);
114 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 
116 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
117 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
118 
119 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
120 		struct sockaddr *);
121 
122 /* XXX: should be in an arp support file, not here */
123 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
124 
125 #define senderr(e) do { error = (e); goto bad;} while (0)
126 
127 #if defined(INET) || defined(INET6)
128 int
129 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
130 	struct ip_fw **rule, int shared);
131 static int ether_ipfw;
132 #endif
133 
134 /*
135  * Ethernet output routine.
136  * Encapsulate a packet of type family for the local net.
137  * Use trailer local net encapsulation if enough data in first
138  * packet leaves a multiple of 512 bytes of data in remainder.
139  */
140 int
141 ether_output(struct ifnet *ifp, struct mbuf *m,
142 	struct sockaddr *dst, struct rtentry *rt0)
143 {
144 	short type;
145 	int error, hdrcmplt = 0;
146 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
147 	struct ether_header *eh;
148 	int loop_copy = 1;
149 	int hlen;	/* link layer header length */
150 
151 #ifdef MAC
152 	error = mac_check_ifnet_transmit(ifp, m);
153 	if (error)
154 		senderr(error);
155 #endif
156 
157 	if (ifp->if_flags & IFF_MONITOR)
158 		senderr(ENETDOWN);
159 	if (!((ifp->if_flags & IFF_UP) &&
160 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
161 		senderr(ENETDOWN);
162 
163 	hlen = ETHER_HDR_LEN;
164 	switch (dst->sa_family) {
165 #ifdef INET
166 	case AF_INET:
167 		error = arpresolve(ifp, rt0, m, dst, edst);
168 		if (error)
169 			return (error == EWOULDBLOCK ? 0 : error);
170 		type = htons(ETHERTYPE_IP);
171 		break;
172 	case AF_ARP:
173 	{
174 		struct arphdr *ah;
175 		ah = mtod(m, struct arphdr *);
176 		ah->ar_hrd = htons(ARPHRD_ETHER);
177 
178 		loop_copy = 0; /* if this is for us, don't do it */
179 
180 		switch(ntohs(ah->ar_op)) {
181 		case ARPOP_REVREQUEST:
182 		case ARPOP_REVREPLY:
183 			type = htons(ETHERTYPE_REVARP);
184 			break;
185 		case ARPOP_REQUEST:
186 		case ARPOP_REPLY:
187 		default:
188 			type = htons(ETHERTYPE_ARP);
189 			break;
190 		}
191 
192 		if (m->m_flags & M_BCAST)
193 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
194 		else
195 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
196 
197 	}
198 	break;
199 #endif
200 #ifdef INET6
201 	case AF_INET6:
202 		error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst);
203 		if (error)
204 			return error;
205 		type = htons(ETHERTYPE_IPV6);
206 		break;
207 #endif
208 #ifdef IPX
209 	case AF_IPX:
210 		if (ef_outputp) {
211 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
212 		    if (error)
213 			goto bad;
214 		} else
215 		    type = htons(ETHERTYPE_IPX);
216 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
217 		    (caddr_t)edst, sizeof (edst));
218 		break;
219 #endif
220 #ifdef NETATALK
221 	case AF_APPLETALK:
222 	  {
223 	    struct at_ifaddr *aa;
224 
225 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
226 		    senderr(EHOSTUNREACH); /* XXX */
227 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst))
228 		    return (0);
229 	    /*
230 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
231 	     */
232 	    if ( aa->aa_flags & AFA_PHASE2 ) {
233 		struct llc llc;
234 
235 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
236 		if (m == NULL)
237 			senderr(ENOBUFS);
238 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
239 		llc.llc_control = LLC_UI;
240 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
241 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
242 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
243 		type = htons(m->m_pkthdr.len);
244 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
245 	    } else {
246 		type = htons(ETHERTYPE_AT);
247 	    }
248 	    break;
249 	  }
250 #endif /* NETATALK */
251 
252 	case pseudo_AF_HDRCMPLT:
253 		hdrcmplt = 1;
254 		eh = (struct ether_header *)dst->sa_data;
255 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
256 		/* FALLTHROUGH */
257 
258 	case AF_UNSPEC:
259 		loop_copy = 0; /* if this is for us, don't do it */
260 		eh = (struct ether_header *)dst->sa_data;
261 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
262 		type = eh->ether_type;
263 		break;
264 
265 	default:
266 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
267 		senderr(EAFNOSUPPORT);
268 	}
269 
270 	/*
271 	 * Add local net header.  If no space in first mbuf,
272 	 * allocate another.
273 	 */
274 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
275 	if (m == NULL)
276 		senderr(ENOBUFS);
277 	eh = mtod(m, struct ether_header *);
278 	(void)memcpy(&eh->ether_type, &type,
279 		sizeof(eh->ether_type));
280 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
281 	if (hdrcmplt)
282 		(void)memcpy(eh->ether_shost, esrc,
283 			sizeof(eh->ether_shost));
284 	else
285 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
286 			sizeof(eh->ether_shost));
287 
288 	/*
289 	 * If a simplex interface, and the packet is being sent to our
290 	 * Ethernet address or a broadcast address, loopback a copy.
291 	 * XXX To make a simplex device behave exactly like a duplex
292 	 * device, we should copy in the case of sending to our own
293 	 * ethernet address (thus letting the original actually appear
294 	 * on the wire). However, we don't do that here for security
295 	 * reasons and compatibility with the original behavior.
296 	 */
297 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
298 	    m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) {
299 		int csum_flags = 0;
300 
301 		if (m->m_pkthdr.csum_flags & CSUM_IP)
302 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
303 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
304 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
305 
306 		if (m->m_flags & M_BCAST) {
307 			struct mbuf *n;
308 
309 			/*
310 			 * Because if_simloop() modifies the packet, we need a
311 			 * writable copy through m_dup() instead of a readonly
312 			 * one as m_copy[m] would give us. The alternative would
313 			 * be to modify if_simloop() to handle the readonly mbuf,
314 			 * but performancewise it is mostly equivalent (trading
315 			 * extra data copying vs. extra locking).
316 			 */
317 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
318 				n->m_pkthdr.csum_flags |= csum_flags;
319 				if (csum_flags & CSUM_DATA_VALID)
320 					n->m_pkthdr.csum_data = 0xffff;
321 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
322 			} else
323 				ifp->if_iqdrops++;
324 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
325 				ETHER_ADDR_LEN) == 0) {
326 			m->m_pkthdr.csum_flags |= csum_flags;
327 			if (csum_flags & CSUM_DATA_VALID)
328 				m->m_pkthdr.csum_data = 0xffff;
329 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
330 			return (0);	/* XXX */
331 		}
332 	}
333 
334        /*
335 	* Bridges require special output handling.
336 	*/
337 	if (ifp->if_bridge) {
338 		BRIDGE_OUTPUT(ifp, m, error);
339 		return (error);
340 	}
341 
342 #ifdef DEV_CARP
343 	if (ifp->if_carp &&
344 	    (error = carp_output(ifp, m, dst, NULL)))
345 		goto bad;
346 #endif
347 
348 	/* Handle ng_ether(4) processing, if any */
349 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
350 		KASSERT(ng_ether_output_p != NULL,
351 		    ("ng_ether_output_p is NULL"));
352 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
353 bad:			if (m != NULL)
354 				m_freem(m);
355 			return (error);
356 		}
357 		if (m == NULL)
358 			return (0);
359 	}
360 
361 	/* Continue with link-layer output */
362 	return ether_output_frame(ifp, m);
363 }
364 
365 /*
366  * Ethernet link layer output routine to send a raw frame to the device.
367  *
368  * This assumes that the 14 byte Ethernet header is present and contiguous
369  * in the first mbuf (if BRIDGE'ing).
370  */
371 int
372 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
373 {
374 	int error;
375 #if defined(INET) || defined(INET6)
376 	struct ip_fw *rule = ip_dn_claim_rule(m);
377 
378 	if (IPFW_LOADED && ether_ipfw != 0) {
379 		if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) {
380 			if (m) {
381 				m_freem(m);
382 				return EACCES;	/* pkt dropped */
383 			} else
384 				return 0;	/* consumed e.g. in a pipe */
385 		}
386 	}
387 #endif
388 
389 	/*
390 	 * Queue message on interface, update output statistics if
391 	 * successful, and start output if interface not yet active.
392 	 */
393 	IFQ_HANDOFF(ifp, m, error);
394 	return (error);
395 }
396 
397 #if defined(INET) || defined(INET6)
398 /*
399  * ipfw processing for ethernet packets (in and out).
400  * The second parameter is NULL from ether_demux, and ifp from
401  * ether_output_frame.
402  */
403 int
404 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
405 	struct ip_fw **rule, int shared)
406 {
407 	struct ether_header *eh;
408 	struct ether_header save_eh;
409 	struct mbuf *m;
410 	int i;
411 	struct ip_fw_args args;
412 
413 	if (*rule != NULL && fw_one_pass)
414 		return 1; /* dummynet packet, already partially processed */
415 
416 	/*
417 	 * I need some amt of data to be contiguous, and in case others need
418 	 * the packet (shared==1) also better be in the first mbuf.
419 	 */
420 	m = *m0;
421 	i = min( m->m_pkthdr.len, max_protohdr);
422 	if ( shared || m->m_len < i) {
423 		m = m_pullup(m, i);
424 		if (m == NULL) {
425 			*m0 = m;
426 			return 0;
427 		}
428 	}
429 	eh = mtod(m, struct ether_header *);
430 	save_eh = *eh;			/* save copy for restore below */
431 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
432 
433 	args.m = m;		/* the packet we are looking at		*/
434 	args.oif = dst;		/* destination, if any			*/
435 	args.rule = *rule;	/* matching rule to restart		*/
436 	args.next_hop = NULL;	/* we do not support forward yet	*/
437 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
438 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
439 	i = ip_fw_chk_ptr(&args);
440 	m = args.m;
441 	if (m != NULL) {
442 		/*
443 		 * Restore Ethernet header, as needed, in case the
444 		 * mbuf chain was replaced by ipfw.
445 		 */
446 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
447 		if (m == NULL) {
448 			*m0 = m;
449 			return 0;
450 		}
451 		if (eh != mtod(m, struct ether_header *))
452 			bcopy(&save_eh, mtod(m, struct ether_header *),
453 				ETHER_HDR_LEN);
454 	}
455 	*m0 = m;
456 	*rule = args.rule;
457 
458 	if (i == IP_FW_DENY) /* drop */
459 		return 0;
460 
461 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
462 
463 	if (i == IP_FW_PASS) /* a PASS rule.  */
464 		return 1;
465 
466 	if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
467 		/*
468 		 * Pass the pkt to dummynet, which consumes it.
469 		 * If shared, make a copy and keep the original.
470 		 */
471 		if (shared) {
472 			m = m_copypacket(m, M_DONTWAIT);
473 			if (m == NULL)
474 				return 0;
475 		} else {
476 			/*
477 			 * Pass the original to dummynet and
478 			 * nothing back to the caller
479 			 */
480 			*m0 = NULL ;
481 		}
482 		ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
483 		return 0;
484 	}
485 	/*
486 	 * XXX at some point add support for divert/forward actions.
487 	 * If none of the above matches, we have to drop the pkt.
488 	 */
489 	return 0;
490 }
491 #endif
492 
493 /*
494  * Process a received Ethernet packet; the packet is in the
495  * mbuf chain m with the ethernet header at the front.
496  */
497 static void
498 ether_input(struct ifnet *ifp, struct mbuf *m)
499 {
500 	struct ether_header *eh;
501 	u_short etype;
502 
503 	/*
504 	 * Do consistency checks to verify assumptions
505 	 * made by code past this point.
506 	 */
507 	if ((m->m_flags & M_PKTHDR) == 0) {
508 		if_printf(ifp, "discard frame w/o packet header\n");
509 		ifp->if_ierrors++;
510 		m_freem(m);
511 		return;
512 	}
513 	if (m->m_len < ETHER_HDR_LEN) {
514 		/* XXX maybe should pullup? */
515 		if_printf(ifp, "discard frame w/o leading ethernet "
516 				"header (len %u pkt len %u)\n",
517 				m->m_len, m->m_pkthdr.len);
518 		ifp->if_ierrors++;
519 		m_freem(m);
520 		return;
521 	}
522 	eh = mtod(m, struct ether_header *);
523 	etype = ntohs(eh->ether_type);
524 	if (m->m_pkthdr.len >
525 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
526 		if_printf(ifp, "discard oversize frame "
527 				"(ether type %x flags %x len %u > max %lu)\n",
528 				etype, m->m_flags, m->m_pkthdr.len,
529 				ETHER_MAX_FRAME(ifp, etype,
530 						m->m_flags & M_HASFCS));
531 		ifp->if_ierrors++;
532 		m_freem(m);
533 		return;
534 	}
535 	if (m->m_pkthdr.rcvif == NULL) {
536 		if_printf(ifp, "discard frame w/o interface pointer\n");
537 		ifp->if_ierrors++;
538 		m_freem(m);
539 		return;
540 	}
541 #ifdef DIAGNOSTIC
542 	if (m->m_pkthdr.rcvif != ifp) {
543 		if_printf(ifp, "Warning, frame marked as received on %s\n",
544 			m->m_pkthdr.rcvif->if_xname);
545 	}
546 #endif
547 
548 #ifdef MAC
549 	/*
550 	 * Tag the mbuf with an appropriate MAC label before any other
551 	 * consumers can get to it.
552 	 */
553 	mac_create_mbuf_from_ifnet(ifp, m);
554 #endif
555 
556 	/*
557 	 * Give bpf a chance at the packet.
558 	 */
559 	BPF_MTAP(ifp, m);
560 
561 	/* If the CRC is still on the packet, trim it off. */
562 	if (m->m_flags & M_HASFCS) {
563 		m_adj(m, -ETHER_CRC_LEN);
564 		m->m_flags &= ~M_HASFCS;
565 	}
566 
567 	ifp->if_ibytes += m->m_pkthdr.len;
568 
569 	if (ifp->if_flags & IFF_MONITOR) {
570 		/*
571 		 * Interface marked for monitoring; discard packet.
572 		 */
573 		m_freem(m);
574 		return;
575 	}
576 
577 	/* Handle ng_ether(4) processing, if any */
578 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
579 		KASSERT(ng_ether_input_p != NULL,
580 		    ("ng_ether_input_p is NULL"));
581 		(*ng_ether_input_p)(ifp, &m);
582 		if (m == NULL)
583 			return;
584 	}
585 
586 	/*
587 	 * Tap the packet off here for a bridge.  bridge_input()
588 	 * will return NULL if it has consumed the packet, otherwise
589 	 * it gets processed as normal.  Note that bridge_input()
590 	 * will always return the original packet if we need to
591 	 * process it locally.
592 	 *
593 	 * XXX: When changing the below block, please also look
594 	 * at the src/sys/netgraph/ng_ether.c:ng_ether_rcv_upper()
595 	 */
596 	if (ifp->if_bridge) {
597 		BRIDGE_INPUT(ifp, m);
598 		if (m == NULL)
599 			return;
600 	}
601 
602 	/* First chunk of an mbuf contains good entropy */
603 	if (harvest.ethernet)
604 		random_harvest(m, 16, 3, 0, RANDOM_NET);
605 	ether_demux(ifp, m);
606 }
607 
608 /*
609  * Upper layer processing for a received Ethernet packet.
610  */
611 void
612 ether_demux(struct ifnet *ifp, struct mbuf *m)
613 {
614 	struct ether_header *eh;
615 	int isr;
616 	u_short ether_type;
617 #if defined(NETATALK)
618 	struct llc *l;
619 #endif
620 #if defined(INET) || defined(INET6)
621 	struct ip_fw *rule = ip_dn_claim_rule(m);
622 #endif
623 
624 	KASSERT(ifp != NULL, ("ether_demux: NULL interface pointer"));
625 
626 	eh = mtod(m, struct ether_header *);
627 	ether_type = ntohs(eh->ether_type);
628 
629 #if defined(INET) || defined(INET6)
630 	if (rule)	/* packet was already bridged */
631 		goto post_stats;
632 #endif
633 
634 	if (!(ifp->if_bridge) &&
635 	    !((ether_type == ETHERTYPE_VLAN || m->m_flags & M_VLANTAG) &&
636 	    ifp->if_vlantrunk != NULL)) {
637 #ifdef DEV_CARP
638 		/*
639 		 * XXX: Okay, we need to call carp_forus() and - if it is for
640 		 * us jump over code that does the normal check
641 		 * "IF_LLADDR(ifp) == ether_dhost". The check sequence is a bit
642 		 * different from OpenBSD, so we jump over as few code as
643 		 * possible, to catch _all_ sanity checks. This needs
644 		 * evaluation, to see if the carp ether_dhost values break any
645 		 * of these checks!
646 		 */
647 		if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
648 			goto pre_stats;
649 #endif
650 		/*
651 		 * Discard packet if upper layers shouldn't see it because it
652 		 * was unicast to a different Ethernet address. If the driver
653 		 * is working properly, then this situation can only happen
654 		 * when the interface is in promiscuous mode.
655 		 *
656 		 * If VLANs are active, and this packet has a VLAN tag, do
657 		 * not drop it here but pass it on to the VLAN layer, to
658 		 * give them a chance to consider it as well (e. g. in case
659 		 * bridging is only active on a VLAN).  They will drop it if
660 		 * it's undesired.
661 		 */
662 		if ((ifp->if_flags & IFF_PROMISC) != 0
663 		    && !ETHER_IS_MULTICAST(eh->ether_dhost)
664 		    && bcmp(eh->ether_dhost,
665 		      IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0
666 		    && (ifp->if_flags & IFF_PPROMISC) == 0) {
667 			    m_freem(m);
668 			    return;
669 		}
670 	}
671 
672 #ifdef DEV_CARP
673 pre_stats:
674 #endif
675 	/* Discard packet if interface is not up */
676 	if ((ifp->if_flags & IFF_UP) == 0) {
677 		m_freem(m);
678 		return;
679 	}
680 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
681 		if (bcmp(etherbroadcastaddr, eh->ether_dhost,
682 		    sizeof(etherbroadcastaddr)) == 0)
683 			m->m_flags |= M_BCAST;
684 		else
685 			m->m_flags |= M_MCAST;
686 	}
687 	if (m->m_flags & (M_BCAST|M_MCAST))
688 		ifp->if_imcasts++;
689 
690 #if defined(INET) || defined(INET6)
691 post_stats:
692 	if (IPFW_LOADED && ether_ipfw != 0) {
693 		if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) {
694 			if (m)
695 				m_freem(m);
696 			return;
697 		}
698 	}
699 #endif
700 
701 	/*
702 	 * Check to see if the device performed the VLAN decapsulation and
703 	 * provided us with the tag.
704 	 */
705 	if (m->m_flags & M_VLANTAG) {
706 		/*
707 		 * If no VLANs are configured, drop.
708 		 */
709 		if (ifp->if_vlantrunk == NULL) {
710 			ifp->if_noproto++;
711 			m_freem(m);
712 			return;
713 		}
714 		/*
715 		 * vlan_input() will either recursively call ether_input()
716 		 * or drop the packet.
717 		 */
718 		KASSERT(vlan_input_p != NULL,("ether_input: VLAN not loaded!"));
719 		(*vlan_input_p)(ifp, m);
720 		return;
721 	}
722 
723 	/*
724 	 * Handle protocols that expect to have the Ethernet header
725 	 * (and possibly FCS) intact.
726 	 */
727 	switch (ether_type) {
728 	case ETHERTYPE_VLAN:
729 		if (ifp->if_vlantrunk != NULL) {
730 			KASSERT(vlan_input_p,("ether_input: VLAN not loaded!"));
731 			(*vlan_input_p)(ifp, m);
732 		} else {
733 			ifp->if_noproto++;
734 			m_freem(m);
735 		}
736 		return;
737 	}
738 
739 	/* Strip off Ethernet header. */
740 	m_adj(m, ETHER_HDR_LEN);
741 
742 	/* If the CRC is still on the packet, trim it off. */
743 	if (m->m_flags & M_HASFCS) {
744 		m_adj(m, -ETHER_CRC_LEN);
745 		m->m_flags &= ~M_HASFCS;
746 	}
747 
748 	/* Reset layer specific mbuf flags to avoid confusing upper layers. */
749 	m->m_flags &= ~(M_PROTOFLAGS);
750 
751 	switch (ether_type) {
752 #ifdef INET
753 	case ETHERTYPE_IP:
754 		if ((m = ip_fastforward(m)) == NULL)
755 			return;
756 		isr = NETISR_IP;
757 		break;
758 
759 	case ETHERTYPE_ARP:
760 		if (ifp->if_flags & IFF_NOARP) {
761 			/* Discard packet if ARP is disabled on interface */
762 			m_freem(m);
763 			return;
764 		}
765 		isr = NETISR_ARP;
766 		break;
767 #endif
768 #ifdef IPX
769 	case ETHERTYPE_IPX:
770 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
771 			return;
772 		isr = NETISR_IPX;
773 		break;
774 #endif
775 #ifdef INET6
776 	case ETHERTYPE_IPV6:
777 		isr = NETISR_IPV6;
778 		break;
779 #endif
780 #ifdef NETATALK
781 	case ETHERTYPE_AT:
782 		isr = NETISR_ATALK1;
783 		break;
784 	case ETHERTYPE_AARP:
785 		isr = NETISR_AARP;
786 		break;
787 #endif /* NETATALK */
788 	default:
789 #ifdef IPX
790 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
791 			return;
792 #endif /* IPX */
793 #if defined(NETATALK)
794 		if (ether_type > ETHERMTU)
795 			goto discard;
796 		l = mtod(m, struct llc *);
797 		if (l->llc_dsap == LLC_SNAP_LSAP &&
798 		    l->llc_ssap == LLC_SNAP_LSAP &&
799 		    l->llc_control == LLC_UI) {
800 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
801 			    sizeof(at_org_code)) == 0 &&
802 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
803 				m_adj(m, LLC_SNAPFRAMELEN);
804 				isr = NETISR_ATALK2;
805 				break;
806 			}
807 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
808 			    sizeof(aarp_org_code)) == 0 &&
809 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
810 				m_adj(m, LLC_SNAPFRAMELEN);
811 				isr = NETISR_AARP;
812 				break;
813 			}
814 		}
815 #endif /* NETATALK */
816 		goto discard;
817 	}
818 	netisr_dispatch(isr, m);
819 	return;
820 
821 discard:
822 	/*
823 	 * Packet is to be discarded.  If netgraph is present,
824 	 * hand the packet to it for last chance processing;
825 	 * otherwise dispose of it.
826 	 */
827 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
828 		KASSERT(ng_ether_input_orphan_p != NULL,
829 		    ("ng_ether_input_orphan_p is NULL"));
830 		/*
831 		 * Put back the ethernet header so netgraph has a
832 		 * consistent view of inbound packets.
833 		 */
834 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
835 		(*ng_ether_input_orphan_p)(ifp, m);
836 		return;
837 	}
838 	m_freem(m);
839 }
840 
841 /*
842  * Convert Ethernet address to printable (loggable) representation.
843  * This routine is for compatibility; it's better to just use
844  *
845  *	printf("%6D", <pointer to address>, ":");
846  *
847  * since there's no static buffer involved.
848  */
849 char *
850 ether_sprintf(const u_char *ap)
851 {
852 	static char etherbuf[18];
853 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
854 	return (etherbuf);
855 }
856 
857 /*
858  * Perform common duties while attaching to interface list
859  */
860 void
861 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
862 {
863 	int i;
864 	struct ifaddr *ifa;
865 	struct sockaddr_dl *sdl;
866 
867 	ifp->if_addrlen = ETHER_ADDR_LEN;
868 	ifp->if_hdrlen = ETHER_HDR_LEN;
869 	if_attach(ifp);
870 	ifp->if_mtu = ETHERMTU;
871 	ifp->if_output = ether_output;
872 	ifp->if_input = ether_input;
873 	ifp->if_resolvemulti = ether_resolvemulti;
874 	if (ifp->if_baudrate == 0)
875 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
876 	ifp->if_broadcastaddr = etherbroadcastaddr;
877 
878 	ifa = ifp->if_addr;
879 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
880 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
881 	sdl->sdl_type = IFT_ETHER;
882 	sdl->sdl_alen = ifp->if_addrlen;
883 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
884 
885 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
886 	if (ng_ether_attach_p != NULL)
887 		(*ng_ether_attach_p)(ifp);
888 
889 	/* Announce Ethernet MAC address if non-zero. */
890 	for (i = 0; i < ifp->if_addrlen; i++)
891 		if (lla[i] != 0)
892 			break;
893 	if (i != ifp->if_addrlen)
894 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
895 	if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0)
896 		if_printf(ifp, "if_start running deferred for Giant\n");
897 }
898 
899 /*
900  * Perform common duties while detaching an Ethernet interface
901  */
902 void
903 ether_ifdetach(struct ifnet *ifp)
904 {
905 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
906 		KASSERT(ng_ether_detach_p != NULL,
907 		    ("ng_ether_detach_p is NULL"));
908 		(*ng_ether_detach_p)(ifp);
909 	}
910 
911 	bpfdetach(ifp);
912 	if_detach(ifp);
913 }
914 
915 SYSCTL_DECL(_net_link);
916 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
917 #if defined(INET) || defined(INET6)
918 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
919 	    &ether_ipfw,0,"Pass ether pkts through firewall");
920 #endif
921 
922 #if 0
923 /*
924  * This is for reference.  We have a table-driven version
925  * of the little-endian crc32 generator, which is faster
926  * than the double-loop.
927  */
928 uint32_t
929 ether_crc32_le(const uint8_t *buf, size_t len)
930 {
931 	size_t i;
932 	uint32_t crc;
933 	int bit;
934 	uint8_t data;
935 
936 	crc = 0xffffffff;	/* initial value */
937 
938 	for (i = 0; i < len; i++) {
939 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1)
940 			carry = (crc ^ data) & 1;
941 			crc >>= 1;
942 			if (carry)
943 				crc = (crc ^ ETHER_CRC_POLY_LE);
944 	}
945 
946 	return (crc);
947 }
948 #else
949 uint32_t
950 ether_crc32_le(const uint8_t *buf, size_t len)
951 {
952 	static const uint32_t crctab[] = {
953 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
954 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
955 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
956 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
957 	};
958 	size_t i;
959 	uint32_t crc;
960 
961 	crc = 0xffffffff;	/* initial value */
962 
963 	for (i = 0; i < len; i++) {
964 		crc ^= buf[i];
965 		crc = (crc >> 4) ^ crctab[crc & 0xf];
966 		crc = (crc >> 4) ^ crctab[crc & 0xf];
967 	}
968 
969 	return (crc);
970 }
971 #endif
972 
973 uint32_t
974 ether_crc32_be(const uint8_t *buf, size_t len)
975 {
976 	size_t i;
977 	uint32_t crc, carry;
978 	int bit;
979 	uint8_t data;
980 
981 	crc = 0xffffffff;	/* initial value */
982 
983 	for (i = 0; i < len; i++) {
984 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
985 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
986 			crc <<= 1;
987 			if (carry)
988 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
989 		}
990 	}
991 
992 	return (crc);
993 }
994 
995 int
996 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
997 {
998 	struct ifaddr *ifa = (struct ifaddr *) data;
999 	struct ifreq *ifr = (struct ifreq *) data;
1000 	int error = 0;
1001 
1002 	switch (command) {
1003 	case SIOCSIFADDR:
1004 		ifp->if_flags |= IFF_UP;
1005 
1006 		switch (ifa->ifa_addr->sa_family) {
1007 #ifdef INET
1008 		case AF_INET:
1009 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1010 			arp_ifinit(ifp, ifa);
1011 			break;
1012 #endif
1013 #ifdef IPX
1014 		/*
1015 		 * XXX - This code is probably wrong
1016 		 */
1017 		case AF_IPX:
1018 			{
1019 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1020 
1021 			if (ipx_nullhost(*ina))
1022 				ina->x_host =
1023 				    *(union ipx_host *)
1024 				    IF_LLADDR(ifp);
1025 			else {
1026 				bcopy((caddr_t) ina->x_host.c_host,
1027 				      (caddr_t) IF_LLADDR(ifp),
1028 				      ETHER_ADDR_LEN);
1029 			}
1030 
1031 			/*
1032 			 * Set new address
1033 			 */
1034 			ifp->if_init(ifp->if_softc);
1035 			break;
1036 			}
1037 #endif
1038 		default:
1039 			ifp->if_init(ifp->if_softc);
1040 			break;
1041 		}
1042 		break;
1043 
1044 	case SIOCGIFADDR:
1045 		{
1046 			struct sockaddr *sa;
1047 
1048 			sa = (struct sockaddr *) & ifr->ifr_data;
1049 			bcopy(IF_LLADDR(ifp),
1050 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1051 		}
1052 		break;
1053 
1054 	case SIOCSIFMTU:
1055 		/*
1056 		 * Set the interface MTU.
1057 		 */
1058 		if (ifr->ifr_mtu > ETHERMTU) {
1059 			error = EINVAL;
1060 		} else {
1061 			ifp->if_mtu = ifr->ifr_mtu;
1062 		}
1063 		break;
1064 	default:
1065 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1066 		break;
1067 	}
1068 	return (error);
1069 }
1070 
1071 static int
1072 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1073 	struct sockaddr *sa)
1074 {
1075 	struct sockaddr_dl *sdl;
1076 #ifdef INET
1077 	struct sockaddr_in *sin;
1078 #endif
1079 #ifdef INET6
1080 	struct sockaddr_in6 *sin6;
1081 #endif
1082 	u_char *e_addr;
1083 
1084 	switch(sa->sa_family) {
1085 	case AF_LINK:
1086 		/*
1087 		 * No mapping needed. Just check that it's a valid MC address.
1088 		 */
1089 		sdl = (struct sockaddr_dl *)sa;
1090 		e_addr = LLADDR(sdl);
1091 		if (!ETHER_IS_MULTICAST(e_addr))
1092 			return EADDRNOTAVAIL;
1093 		*llsa = 0;
1094 		return 0;
1095 
1096 #ifdef INET
1097 	case AF_INET:
1098 		sin = (struct sockaddr_in *)sa;
1099 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1100 			return EADDRNOTAVAIL;
1101 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1102 		       M_NOWAIT|M_ZERO);
1103 		if (sdl == NULL)
1104 			return ENOMEM;
1105 		sdl->sdl_len = sizeof *sdl;
1106 		sdl->sdl_family = AF_LINK;
1107 		sdl->sdl_index = ifp->if_index;
1108 		sdl->sdl_type = IFT_ETHER;
1109 		sdl->sdl_alen = ETHER_ADDR_LEN;
1110 		e_addr = LLADDR(sdl);
1111 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1112 		*llsa = (struct sockaddr *)sdl;
1113 		return 0;
1114 #endif
1115 #ifdef INET6
1116 	case AF_INET6:
1117 		sin6 = (struct sockaddr_in6 *)sa;
1118 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1119 			/*
1120 			 * An IP6 address of 0 means listen to all
1121 			 * of the Ethernet multicast address used for IP6.
1122 			 * (This is used for multicast routers.)
1123 			 */
1124 			ifp->if_flags |= IFF_ALLMULTI;
1125 			*llsa = 0;
1126 			return 0;
1127 		}
1128 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1129 			return EADDRNOTAVAIL;
1130 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1131 		       M_NOWAIT|M_ZERO);
1132 		if (sdl == NULL)
1133 			return (ENOMEM);
1134 		sdl->sdl_len = sizeof *sdl;
1135 		sdl->sdl_family = AF_LINK;
1136 		sdl->sdl_index = ifp->if_index;
1137 		sdl->sdl_type = IFT_ETHER;
1138 		sdl->sdl_alen = ETHER_ADDR_LEN;
1139 		e_addr = LLADDR(sdl);
1140 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1141 		*llsa = (struct sockaddr *)sdl;
1142 		return 0;
1143 #endif
1144 
1145 	default:
1146 		/*
1147 		 * Well, the text isn't quite right, but it's the name
1148 		 * that counts...
1149 		 */
1150 		return EAFNOSUPPORT;
1151 	}
1152 }
1153 
1154 static void*
1155 ether_alloc(u_char type, struct ifnet *ifp)
1156 {
1157 	struct arpcom	*ac;
1158 
1159 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1160 	ac->ac_ifp = ifp;
1161 
1162 	return (ac);
1163 }
1164 
1165 static void
1166 ether_free(void *com, u_char type)
1167 {
1168 
1169 	free(com, M_ARPCOM);
1170 }
1171 
1172 static int
1173 ether_modevent(module_t mod, int type, void *data)
1174 {
1175 
1176 	switch (type) {
1177 	case MOD_LOAD:
1178 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1179 		break;
1180 	case MOD_UNLOAD:
1181 		if_deregister_com_alloc(IFT_ETHER);
1182 		break;
1183 	default:
1184 		return EOPNOTSUPP;
1185 	}
1186 
1187 	return (0);
1188 }
1189 
1190 static moduledata_t ether_mod = {
1191 	"ether",
1192 	ether_modevent,
1193 	0
1194 };
1195 
1196 void
1197 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1198 {
1199 	struct ether_vlan_header vlan;
1200 	struct mbuf mv, mb;
1201 
1202 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1203 	    ("%s: vlan information not present", __func__));
1204 	KASSERT(m->m_len >= sizeof(struct ether_header),
1205 	    ("%s: mbuf not large enough for header", __func__));
1206 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1207 	vlan.evl_proto = vlan.evl_encap_proto;
1208 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1209 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1210 	m->m_len -= sizeof(struct ether_header);
1211 	m->m_data += sizeof(struct ether_header);
1212 	/*
1213 	 * If a data link has been supplied by the caller, then we will need to
1214 	 * re-create a stack allocated mbuf chain with the following structure:
1215 	 *
1216 	 * (1) mbuf #1 will contain the supplied data link
1217 	 * (2) mbuf #2 will contain the vlan header
1218 	 * (3) mbuf #3 will contain the original mbuf's packet data
1219 	 *
1220 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1221 	 */
1222 	if (data != NULL) {
1223 		mv.m_next = m;
1224 		mv.m_data = (caddr_t)&vlan;
1225 		mv.m_len = sizeof(vlan);
1226 		mb.m_next = &mv;
1227 		mb.m_data = data;
1228 		mb.m_len = dlen;
1229 		bpf_mtap(bp, &mb);
1230 	} else
1231 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1232 	m->m_len += sizeof(struct ether_header);
1233 	m->m_data -= sizeof(struct ether_header);
1234 }
1235 
1236 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1237 MODULE_VERSION(ether, 1);
1238