xref: /freebsd/sys/net/if_ethersubr.c (revision 3ee1a36e2ef343e0af2cf0389c268695aba14eac)
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_netgraph.h"
38 #include "opt_mbuf_profiling.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/module.h>
46 #include <sys/mbuf.h>
47 #include <sys/random.h>
48 #include <sys/rwlock.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 
53 #include <net/if.h>
54 #include <net/if_arp.h>
55 #include <net/netisr.h>
56 #include <net/route.h>
57 #include <net/if_llc.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/if_bridgevar.h>
63 #include <net/if_vlan_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/pf_mtag.h>
66 #include <net/vnet.h>
67 
68 #if defined(INET) || defined(INET6)
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/if_ether.h>
72 #include <netinet/ip_carp.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_fw.h>
75 #include <netinet/ipfw/ip_fw_private.h>
76 #endif
77 #ifdef INET6
78 #include <netinet6/nd6.h>
79 #endif
80 
81 #ifdef IPX
82 #include <netipx/ipx.h>
83 #include <netipx/ipx_if.h>
84 #endif
85 
86 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
87 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
88 		struct sockaddr *dst, short *tp, int *hlen);
89 
90 #ifdef NETATALK
91 #include <netatalk/at.h>
92 #include <netatalk/at_var.h>
93 #include <netatalk/at_extern.h>
94 
95 #define llc_snap_org_code llc_un.type_snap.org_code
96 #define llc_snap_ether_type llc_un.type_snap.ether_type
97 
98 extern u_char	at_org_code[3];
99 extern u_char	aarp_org_code[3];
100 #endif /* NETATALK */
101 
102 #include <security/mac/mac_framework.h>
103 
104 #ifdef CTASSERT
105 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
106 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
107 #endif
108 
109 /* netgraph node hooks for ng_ether(4) */
110 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
111 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
112 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
113 void	(*ng_ether_attach_p)(struct ifnet *ifp);
114 void	(*ng_ether_detach_p)(struct ifnet *ifp);
115 
116 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
117 
118 /* if_bridge(4) support */
119 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
120 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
121 		struct sockaddr *, struct rtentry *);
122 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
123 
124 /* if_lagg(4) support */
125 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
126 
127 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
128 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
129 
130 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
131 		struct sockaddr *);
132 #ifdef VIMAGE
133 static	void ether_reassign(struct ifnet *, struct vnet *, char *);
134 #endif
135 
136 /* XXX: should be in an arp support file, not here */
137 static MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
138 
139 #define	ETHER_IS_BROADCAST(addr) \
140 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
141 
142 #define senderr(e) do { error = (e); goto bad;} while (0)
143 
144 #if defined(INET) || defined(INET6)
145 int
146 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared);
147 static VNET_DEFINE(int, ether_ipfw);
148 #define	V_ether_ipfw	VNET(ether_ipfw)
149 #endif
150 
151 
152 /*
153  * Ethernet output routine.
154  * Encapsulate a packet of type family for the local net.
155  * Use trailer local net encapsulation if enough data in first
156  * packet leaves a multiple of 512 bytes of data in remainder.
157  */
158 int
159 ether_output(struct ifnet *ifp, struct mbuf *m,
160 	struct sockaddr *dst, struct route *ro)
161 {
162 	short type;
163 	int error = 0, hdrcmplt = 0;
164 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
165 	struct llentry *lle = NULL;
166 	struct rtentry *rt0 = NULL;
167 	struct ether_header *eh;
168 	struct pf_mtag *t;
169 	int loop_copy = 1;
170 	int hlen;	/* link layer header length */
171 
172 	if (ro != NULL) {
173 		if (!(m->m_flags & (M_BCAST | M_MCAST)))
174 			lle = ro->ro_lle;
175 		rt0 = ro->ro_rt;
176 	}
177 #ifdef MAC
178 	error = mac_ifnet_check_transmit(ifp, m);
179 	if (error)
180 		senderr(error);
181 #endif
182 
183 	M_PROFILE(m);
184 	if (ifp->if_flags & IFF_MONITOR)
185 		senderr(ENETDOWN);
186 	if (!((ifp->if_flags & IFF_UP) &&
187 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
188 		senderr(ENETDOWN);
189 
190 	hlen = ETHER_HDR_LEN;
191 	switch (dst->sa_family) {
192 #ifdef INET
193 	case AF_INET:
194 		if (lle != NULL && (lle->la_flags & LLE_VALID))
195 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
196 		else
197 			error = arpresolve(ifp, rt0, m, dst, edst, &lle);
198 		if (error)
199 			return (error == EWOULDBLOCK ? 0 : error);
200 		type = htons(ETHERTYPE_IP);
201 		break;
202 	case AF_ARP:
203 	{
204 		struct arphdr *ah;
205 		ah = mtod(m, struct arphdr *);
206 		ah->ar_hrd = htons(ARPHRD_ETHER);
207 
208 		loop_copy = 0; /* if this is for us, don't do it */
209 
210 		switch(ntohs(ah->ar_op)) {
211 		case ARPOP_REVREQUEST:
212 		case ARPOP_REVREPLY:
213 			type = htons(ETHERTYPE_REVARP);
214 			break;
215 		case ARPOP_REQUEST:
216 		case ARPOP_REPLY:
217 		default:
218 			type = htons(ETHERTYPE_ARP);
219 			break;
220 		}
221 
222 		if (m->m_flags & M_BCAST)
223 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
224 		else
225 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
226 
227 	}
228 	break;
229 #endif
230 #ifdef INET6
231 	case AF_INET6:
232 		if (lle != NULL && (lle->la_flags & LLE_VALID))
233 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
234 		else
235 			error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle);
236 		if (error)
237 			return error;
238 		type = htons(ETHERTYPE_IPV6);
239 		break;
240 #endif
241 #ifdef IPX
242 	case AF_IPX:
243 		if (ef_outputp) {
244 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
245 		    if (error)
246 			goto bad;
247 		} else
248 		    type = htons(ETHERTYPE_IPX);
249 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
250 		    (caddr_t)edst, sizeof (edst));
251 		break;
252 #endif
253 #ifdef NETATALK
254 	case AF_APPLETALK:
255 	  {
256 	    struct at_ifaddr *aa;
257 
258 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
259 		    senderr(EHOSTUNREACH); /* XXX */
260 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) {
261 		    ifa_free(&aa->aa_ifa);
262 		    return (0);
263 	    }
264 	    /*
265 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
266 	     */
267 	    if ( aa->aa_flags & AFA_PHASE2 ) {
268 		struct llc llc;
269 
270 		ifa_free(&aa->aa_ifa);
271 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
272 		if (m == NULL)
273 			senderr(ENOBUFS);
274 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
275 		llc.llc_control = LLC_UI;
276 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
277 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
278 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
279 		type = htons(m->m_pkthdr.len);
280 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
281 	    } else {
282 		ifa_free(&aa->aa_ifa);
283 		type = htons(ETHERTYPE_AT);
284 	    }
285 	    break;
286 	  }
287 #endif /* NETATALK */
288 
289 	case pseudo_AF_HDRCMPLT:
290 		hdrcmplt = 1;
291 		eh = (struct ether_header *)dst->sa_data;
292 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
293 		/* FALLTHROUGH */
294 
295 	case AF_UNSPEC:
296 		loop_copy = 0; /* if this is for us, don't do it */
297 		eh = (struct ether_header *)dst->sa_data;
298 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
299 		type = eh->ether_type;
300 		break;
301 
302 	default:
303 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
304 		senderr(EAFNOSUPPORT);
305 	}
306 
307 	if (lle != NULL && (lle->la_flags & LLE_IFADDR)) {
308 		int csum_flags = 0;
309 		if (m->m_pkthdr.csum_flags & CSUM_IP)
310 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
311 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
312 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
313 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
314 			csum_flags |= CSUM_SCTP_VALID;
315 		m->m_pkthdr.csum_flags |= csum_flags;
316 		m->m_pkthdr.csum_data = 0xffff;
317 		return (if_simloop(ifp, m, dst->sa_family, 0));
318 	}
319 
320 	/*
321 	 * Add local net header.  If no space in first mbuf,
322 	 * allocate another.
323 	 */
324 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
325 	if (m == NULL)
326 		senderr(ENOBUFS);
327 	eh = mtod(m, struct ether_header *);
328 	(void)memcpy(&eh->ether_type, &type,
329 		sizeof(eh->ether_type));
330 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
331 	if (hdrcmplt)
332 		(void)memcpy(eh->ether_shost, esrc,
333 			sizeof(eh->ether_shost));
334 	else
335 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
336 			sizeof(eh->ether_shost));
337 
338 	/*
339 	 * If a simplex interface, and the packet is being sent to our
340 	 * Ethernet address or a broadcast address, loopback a copy.
341 	 * XXX To make a simplex device behave exactly like a duplex
342 	 * device, we should copy in the case of sending to our own
343 	 * ethernet address (thus letting the original actually appear
344 	 * on the wire). However, we don't do that here for security
345 	 * reasons and compatibility with the original behavior.
346 	 */
347 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
348 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
349 		int csum_flags = 0;
350 
351 		if (m->m_pkthdr.csum_flags & CSUM_IP)
352 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
353 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
354 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
355 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
356 			csum_flags |= CSUM_SCTP_VALID;
357 
358 		if (m->m_flags & M_BCAST) {
359 			struct mbuf *n;
360 
361 			/*
362 			 * Because if_simloop() modifies the packet, we need a
363 			 * writable copy through m_dup() instead of a readonly
364 			 * one as m_copy[m] would give us. The alternative would
365 			 * be to modify if_simloop() to handle the readonly mbuf,
366 			 * but performancewise it is mostly equivalent (trading
367 			 * extra data copying vs. extra locking).
368 			 *
369 			 * XXX This is a local workaround.  A number of less
370 			 * often used kernel parts suffer from the same bug.
371 			 * See PR kern/105943 for a proposed general solution.
372 			 */
373 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
374 				n->m_pkthdr.csum_flags |= csum_flags;
375 				if (csum_flags & CSUM_DATA_VALID)
376 					n->m_pkthdr.csum_data = 0xffff;
377 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
378 			} else
379 				ifp->if_iqdrops++;
380 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
381 				ETHER_ADDR_LEN) == 0) {
382 			m->m_pkthdr.csum_flags |= csum_flags;
383 			if (csum_flags & CSUM_DATA_VALID)
384 				m->m_pkthdr.csum_data = 0xffff;
385 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
386 			return (0);	/* XXX */
387 		}
388 	}
389 
390        /*
391 	* Bridges require special output handling.
392 	*/
393 	if (ifp->if_bridge) {
394 		BRIDGE_OUTPUT(ifp, m, error);
395 		return (error);
396 	}
397 
398 #if defined(INET) || defined(INET6)
399 	if (ifp->if_carp &&
400 	    (error = (*carp_output_p)(ifp, m, dst, NULL)))
401 		goto bad;
402 #endif
403 
404 	/* Handle ng_ether(4) processing, if any */
405 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
406 		KASSERT(ng_ether_output_p != NULL,
407 		    ("ng_ether_output_p is NULL"));
408 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
409 bad:			if (m != NULL)
410 				m_freem(m);
411 			return (error);
412 		}
413 		if (m == NULL)
414 			return (0);
415 	}
416 
417 	/* Continue with link-layer output */
418 	return ether_output_frame(ifp, m);
419 }
420 
421 /*
422  * Ethernet link layer output routine to send a raw frame to the device.
423  *
424  * This assumes that the 14 byte Ethernet header is present and contiguous
425  * in the first mbuf (if BRIDGE'ing).
426  */
427 int
428 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
429 {
430 #if defined(INET) || defined(INET6)
431 
432 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0) {
433 		if (ether_ipfw_chk(&m, ifp, 0) == 0) {
434 			if (m) {
435 				m_freem(m);
436 				return EACCES;	/* pkt dropped */
437 			} else
438 				return 0;	/* consumed e.g. in a pipe */
439 		}
440 	}
441 #endif
442 
443 	/*
444 	 * Queue message on interface, update output statistics if
445 	 * successful, and start output if interface not yet active.
446 	 */
447 	return ((ifp->if_transmit)(ifp, m));
448 }
449 
450 #if defined(INET) || defined(INET6)
451 /*
452  * ipfw processing for ethernet packets (in and out).
453  * The second parameter is NULL from ether_demux, and ifp from
454  * ether_output_frame.
455  */
456 int
457 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared)
458 {
459 	struct ether_header *eh;
460 	struct ether_header save_eh;
461 	struct mbuf *m;
462 	int i;
463 	struct ip_fw_args args;
464 	struct m_tag *mtag;
465 
466 	/* fetch start point from rule, if any */
467 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
468 	if (mtag == NULL) {
469 		args.rule.slot = 0;
470 	} else {
471 		/* dummynet packet, already partially processed */
472 		struct ipfw_rule_ref *r;
473 
474 		/* XXX can we free it after use ? */
475 		mtag->m_tag_id = PACKET_TAG_NONE;
476 		r = (struct ipfw_rule_ref *)(mtag + 1);
477 		if (r->info & IPFW_ONEPASS)
478 			return (1);
479 		args.rule = *r;
480 	}
481 
482 	/*
483 	 * I need some amt of data to be contiguous, and in case others need
484 	 * the packet (shared==1) also better be in the first mbuf.
485 	 */
486 	m = *m0;
487 	i = min( m->m_pkthdr.len, max_protohdr);
488 	if ( shared || m->m_len < i) {
489 		m = m_pullup(m, i);
490 		if (m == NULL) {
491 			*m0 = m;
492 			return 0;
493 		}
494 	}
495 	eh = mtod(m, struct ether_header *);
496 	save_eh = *eh;			/* save copy for restore below */
497 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
498 
499 	args.m = m;		/* the packet we are looking at		*/
500 	args.oif = dst;		/* destination, if any			*/
501 	args.next_hop = NULL;	/* we do not support forward yet	*/
502 	args.next_hop6 = NULL;	/* we do not support forward yet	*/
503 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
504 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
505 	i = V_ip_fw_chk_ptr(&args);
506 	m = args.m;
507 	if (m != NULL) {
508 		/*
509 		 * Restore Ethernet header, as needed, in case the
510 		 * mbuf chain was replaced by ipfw.
511 		 */
512 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
513 		if (m == NULL) {
514 			*m0 = m;
515 			return 0;
516 		}
517 		if (eh != mtod(m, struct ether_header *))
518 			bcopy(&save_eh, mtod(m, struct ether_header *),
519 				ETHER_HDR_LEN);
520 	}
521 	*m0 = m;
522 
523 	if (i == IP_FW_DENY) /* drop */
524 		return 0;
525 
526 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
527 
528 	if (i == IP_FW_PASS) /* a PASS rule.  */
529 		return 1;
530 
531 	if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
532 		int dir;
533 		/*
534 		 * Pass the pkt to dummynet, which consumes it.
535 		 * If shared, make a copy and keep the original.
536 		 */
537 		if (shared) {
538 			m = m_copypacket(m, M_DONTWAIT);
539 			if (m == NULL)
540 				return 0;
541 		} else {
542 			/*
543 			 * Pass the original to dummynet and
544 			 * nothing back to the caller
545 			 */
546 			*m0 = NULL ;
547 		}
548 		dir = PROTO_LAYER2 | (dst ? DIR_OUT : DIR_IN);
549 		ip_dn_io_ptr(&m, dir, &args);
550 		return 0;
551 	}
552 	/*
553 	 * XXX at some point add support for divert/forward actions.
554 	 * If none of the above matches, we have to drop the pkt.
555 	 */
556 	return 0;
557 }
558 #endif
559 
560 /*
561  * Process a received Ethernet packet; the packet is in the
562  * mbuf chain m with the ethernet header at the front.
563  */
564 static void
565 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
566 {
567 	struct ether_header *eh;
568 	u_short etype;
569 
570 	if ((ifp->if_flags & IFF_UP) == 0) {
571 		m_freem(m);
572 		return;
573 	}
574 #ifdef DIAGNOSTIC
575 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
576 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
577 		m_freem(m);
578 		return;
579 	}
580 #endif
581 	/*
582 	 * Do consistency checks to verify assumptions
583 	 * made by code past this point.
584 	 */
585 	if ((m->m_flags & M_PKTHDR) == 0) {
586 		if_printf(ifp, "discard frame w/o packet header\n");
587 		ifp->if_ierrors++;
588 		m_freem(m);
589 		return;
590 	}
591 	if (m->m_len < ETHER_HDR_LEN) {
592 		/* XXX maybe should pullup? */
593 		if_printf(ifp, "discard frame w/o leading ethernet "
594 				"header (len %u pkt len %u)\n",
595 				m->m_len, m->m_pkthdr.len);
596 		ifp->if_ierrors++;
597 		m_freem(m);
598 		return;
599 	}
600 	eh = mtod(m, struct ether_header *);
601 	etype = ntohs(eh->ether_type);
602 	if (m->m_pkthdr.rcvif == NULL) {
603 		if_printf(ifp, "discard frame w/o interface pointer\n");
604 		ifp->if_ierrors++;
605 		m_freem(m);
606 		return;
607 	}
608 #ifdef DIAGNOSTIC
609 	if (m->m_pkthdr.rcvif != ifp) {
610 		if_printf(ifp, "Warning, frame marked as received on %s\n",
611 			m->m_pkthdr.rcvif->if_xname);
612 	}
613 #endif
614 
615 	CURVNET_SET_QUIET(ifp->if_vnet);
616 
617 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
618 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
619 			m->m_flags |= M_BCAST;
620 		else
621 			m->m_flags |= M_MCAST;
622 		ifp->if_imcasts++;
623 	}
624 
625 #ifdef MAC
626 	/*
627 	 * Tag the mbuf with an appropriate MAC label before any other
628 	 * consumers can get to it.
629 	 */
630 	mac_ifnet_create_mbuf(ifp, m);
631 #endif
632 
633 	/*
634 	 * Give bpf a chance at the packet.
635 	 */
636 	ETHER_BPF_MTAP(ifp, m);
637 
638 	/*
639 	 * If the CRC is still on the packet, trim it off. We do this once
640 	 * and once only in case we are re-entered. Nothing else on the
641 	 * Ethernet receive path expects to see the FCS.
642 	 */
643 	if (m->m_flags & M_HASFCS) {
644 		m_adj(m, -ETHER_CRC_LEN);
645 		m->m_flags &= ~M_HASFCS;
646 	}
647 
648 	ifp->if_ibytes += m->m_pkthdr.len;
649 
650 	/* Allow monitor mode to claim this frame, after stats are updated. */
651 	if (ifp->if_flags & IFF_MONITOR) {
652 		m_freem(m);
653 		CURVNET_RESTORE();
654 		return;
655 	}
656 
657 	/* Handle input from a lagg(4) port */
658 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
659 		KASSERT(lagg_input_p != NULL,
660 		    ("%s: if_lagg not loaded!", __func__));
661 		m = (*lagg_input_p)(ifp, m);
662 		if (m != NULL)
663 			ifp = m->m_pkthdr.rcvif;
664 		else
665 			return;
666 	}
667 
668 	/*
669 	 * If the hardware did not process an 802.1Q tag, do this now,
670 	 * to allow 802.1P priority frames to be passed to the main input
671 	 * path correctly.
672 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
673 	 */
674 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
675 		struct ether_vlan_header *evl;
676 
677 		if (m->m_len < sizeof(*evl) &&
678 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
679 #ifdef DIAGNOSTIC
680 			if_printf(ifp, "cannot pullup VLAN header\n");
681 #endif
682 			ifp->if_ierrors++;
683 			m_freem(m);
684 			return;
685 		}
686 
687 		evl = mtod(m, struct ether_vlan_header *);
688 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
689 		m->m_flags |= M_VLANTAG;
690 
691 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
692 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
693 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
694 	}
695 
696 	M_SETFIB(m, ifp->if_fib);
697 
698 	/* Allow ng_ether(4) to claim this frame. */
699 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
700 		KASSERT(ng_ether_input_p != NULL,
701 		    ("%s: ng_ether_input_p is NULL", __func__));
702 		m->m_flags &= ~M_PROMISC;
703 		(*ng_ether_input_p)(ifp, &m);
704 		if (m == NULL) {
705 			CURVNET_RESTORE();
706 			return;
707 		}
708 	}
709 
710 	/*
711 	 * Allow if_bridge(4) to claim this frame.
712 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
713 	 * and the frame should be delivered locally.
714 	 */
715 	if (ifp->if_bridge != NULL) {
716 		m->m_flags &= ~M_PROMISC;
717 		BRIDGE_INPUT(ifp, m);
718 		if (m == NULL) {
719 			CURVNET_RESTORE();
720 			return;
721 		}
722 	}
723 
724 #if defined(INET) || defined(INET6)
725 	/*
726 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
727 	 * mbuf flows up to Layer 3.
728 	 * FreeBSD's implementation of carp(4) uses the inprotosw
729 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
730 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
731 	 * is outside the scope of the M_PROMISC test below.
732 	 * TODO: Maintain a hash table of ethernet addresses other than
733 	 * ether_dhost which may be active on this ifp.
734 	 */
735 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
736 		m->m_flags &= ~M_PROMISC;
737 	} else
738 #endif
739 	{
740 		/*
741 		 * If the frame received was not for our MAC address, set the
742 		 * M_PROMISC flag on the mbuf chain. The frame may need to
743 		 * be seen by the rest of the Ethernet input path in case of
744 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
745 		 * seen by upper protocol layers.
746 		 */
747 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
748 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
749 			m->m_flags |= M_PROMISC;
750 	}
751 
752 	/* First chunk of an mbuf contains good entropy */
753 	if (harvest.ethernet)
754 		random_harvest(m, 16, 3, 0, RANDOM_NET);
755 
756 	ether_demux(ifp, m);
757 	CURVNET_RESTORE();
758 }
759 
760 /*
761  * Ethernet input dispatch; by default, direct dispatch here regardless of
762  * global configuration.
763  */
764 static void
765 ether_nh_input(struct mbuf *m)
766 {
767 
768 	ether_input_internal(m->m_pkthdr.rcvif, m);
769 }
770 
771 static struct netisr_handler	ether_nh = {
772 	.nh_name = "ether",
773 	.nh_handler = ether_nh_input,
774 	.nh_proto = NETISR_ETHER,
775 	.nh_policy = NETISR_POLICY_SOURCE,
776 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
777 };
778 
779 static void
780 ether_init(__unused void *arg)
781 {
782 
783 	netisr_register(&ether_nh);
784 }
785 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
786 
787 static void
788 ether_input(struct ifnet *ifp, struct mbuf *m)
789 {
790 
791 	/*
792 	 * We will rely on rcvif being set properly in the deferred context,
793 	 * so assert it is correct here.
794 	 */
795 	KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__));
796 
797 	netisr_dispatch(NETISR_ETHER, m);
798 }
799 
800 /*
801  * Upper layer processing for a received Ethernet packet.
802  */
803 void
804 ether_demux(struct ifnet *ifp, struct mbuf *m)
805 {
806 	struct ether_header *eh;
807 	int isr;
808 	u_short ether_type;
809 #if defined(NETATALK)
810 	struct llc *l;
811 #endif
812 
813 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
814 
815 #if defined(INET) || defined(INET6)
816 	/*
817 	 * Allow dummynet and/or ipfw to claim the frame.
818 	 * Do not do this for PROMISC frames in case we are re-entered.
819 	 */
820 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
821 		if (ether_ipfw_chk(&m, NULL, 0) == 0) {
822 			if (m)
823 				m_freem(m);	/* dropped; free mbuf chain */
824 			return;			/* consumed */
825 		}
826 	}
827 #endif
828 	eh = mtod(m, struct ether_header *);
829 	ether_type = ntohs(eh->ether_type);
830 
831 	/*
832 	 * If this frame has a VLAN tag other than 0, call vlan_input()
833 	 * if its module is loaded. Otherwise, drop.
834 	 */
835 	if ((m->m_flags & M_VLANTAG) &&
836 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
837 		if (ifp->if_vlantrunk == NULL) {
838 			ifp->if_noproto++;
839 			m_freem(m);
840 			return;
841 		}
842 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
843 		    __func__));
844 		/* Clear before possibly re-entering ether_input(). */
845 		m->m_flags &= ~M_PROMISC;
846 		(*vlan_input_p)(ifp, m);
847 		return;
848 	}
849 
850 	/*
851 	 * Pass promiscuously received frames to the upper layer if the user
852 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
853 	 */
854 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
855 		m_freem(m);
856 		return;
857 	}
858 
859 	/*
860 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
861 	 * Strip off Ethernet header.
862 	 */
863 	m->m_flags &= ~M_VLANTAG;
864 	m->m_flags &= ~(M_PROTOFLAGS);
865 	m_adj(m, ETHER_HDR_LEN);
866 
867 	/*
868 	 * Dispatch frame to upper layer.
869 	 */
870 	switch (ether_type) {
871 #ifdef INET
872 	case ETHERTYPE_IP:
873 		if ((m = ip_fastforward(m)) == NULL)
874 			return;
875 		isr = NETISR_IP;
876 		break;
877 
878 	case ETHERTYPE_ARP:
879 		if (ifp->if_flags & IFF_NOARP) {
880 			/* Discard packet if ARP is disabled on interface */
881 			m_freem(m);
882 			return;
883 		}
884 		isr = NETISR_ARP;
885 		break;
886 #endif
887 #ifdef IPX
888 	case ETHERTYPE_IPX:
889 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
890 			return;
891 		isr = NETISR_IPX;
892 		break;
893 #endif
894 #ifdef INET6
895 	case ETHERTYPE_IPV6:
896 		isr = NETISR_IPV6;
897 		break;
898 #endif
899 #ifdef NETATALK
900 	case ETHERTYPE_AT:
901 		isr = NETISR_ATALK1;
902 		break;
903 	case ETHERTYPE_AARP:
904 		isr = NETISR_AARP;
905 		break;
906 #endif /* NETATALK */
907 	default:
908 #ifdef IPX
909 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
910 			return;
911 #endif /* IPX */
912 #if defined(NETATALK)
913 		if (ether_type > ETHERMTU)
914 			goto discard;
915 		l = mtod(m, struct llc *);
916 		if (l->llc_dsap == LLC_SNAP_LSAP &&
917 		    l->llc_ssap == LLC_SNAP_LSAP &&
918 		    l->llc_control == LLC_UI) {
919 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
920 			    sizeof(at_org_code)) == 0 &&
921 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
922 				m_adj(m, LLC_SNAPFRAMELEN);
923 				isr = NETISR_ATALK2;
924 				break;
925 			}
926 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
927 			    sizeof(aarp_org_code)) == 0 &&
928 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
929 				m_adj(m, LLC_SNAPFRAMELEN);
930 				isr = NETISR_AARP;
931 				break;
932 			}
933 		}
934 #endif /* NETATALK */
935 		goto discard;
936 	}
937 	netisr_dispatch(isr, m);
938 	return;
939 
940 discard:
941 	/*
942 	 * Packet is to be discarded.  If netgraph is present,
943 	 * hand the packet to it for last chance processing;
944 	 * otherwise dispose of it.
945 	 */
946 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
947 		KASSERT(ng_ether_input_orphan_p != NULL,
948 		    ("ng_ether_input_orphan_p is NULL"));
949 		/*
950 		 * Put back the ethernet header so netgraph has a
951 		 * consistent view of inbound packets.
952 		 */
953 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
954 		(*ng_ether_input_orphan_p)(ifp, m);
955 		return;
956 	}
957 	m_freem(m);
958 }
959 
960 /*
961  * Convert Ethernet address to printable (loggable) representation.
962  * This routine is for compatibility; it's better to just use
963  *
964  *	printf("%6D", <pointer to address>, ":");
965  *
966  * since there's no static buffer involved.
967  */
968 char *
969 ether_sprintf(const u_char *ap)
970 {
971 	static char etherbuf[18];
972 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
973 	return (etherbuf);
974 }
975 
976 /*
977  * Perform common duties while attaching to interface list
978  */
979 void
980 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
981 {
982 	int i;
983 	struct ifaddr *ifa;
984 	struct sockaddr_dl *sdl;
985 
986 	ifp->if_addrlen = ETHER_ADDR_LEN;
987 	ifp->if_hdrlen = ETHER_HDR_LEN;
988 	if_attach(ifp);
989 	ifp->if_mtu = ETHERMTU;
990 	ifp->if_output = ether_output;
991 	ifp->if_input = ether_input;
992 	ifp->if_resolvemulti = ether_resolvemulti;
993 #ifdef VIMAGE
994 	ifp->if_reassign = ether_reassign;
995 #endif
996 	if (ifp->if_baudrate == 0)
997 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
998 	ifp->if_broadcastaddr = etherbroadcastaddr;
999 
1000 	ifa = ifp->if_addr;
1001 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
1002 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1003 	sdl->sdl_type = IFT_ETHER;
1004 	sdl->sdl_alen = ifp->if_addrlen;
1005 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1006 
1007 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1008 	if (ng_ether_attach_p != NULL)
1009 		(*ng_ether_attach_p)(ifp);
1010 
1011 	/* Announce Ethernet MAC address if non-zero. */
1012 	for (i = 0; i < ifp->if_addrlen; i++)
1013 		if (lla[i] != 0)
1014 			break;
1015 	if (i != ifp->if_addrlen)
1016 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1017 }
1018 
1019 /*
1020  * Perform common duties while detaching an Ethernet interface
1021  */
1022 void
1023 ether_ifdetach(struct ifnet *ifp)
1024 {
1025 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
1026 		KASSERT(ng_ether_detach_p != NULL,
1027 		    ("ng_ether_detach_p is NULL"));
1028 		(*ng_ether_detach_p)(ifp);
1029 	}
1030 
1031 	bpfdetach(ifp);
1032 	if_detach(ifp);
1033 }
1034 
1035 #ifdef VIMAGE
1036 void
1037 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1038 {
1039 
1040 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
1041 		KASSERT(ng_ether_detach_p != NULL,
1042 		    ("ng_ether_detach_p is NULL"));
1043 		(*ng_ether_detach_p)(ifp);
1044 	}
1045 
1046 	if (ng_ether_attach_p != NULL) {
1047 		CURVNET_SET_QUIET(new_vnet);
1048 		(*ng_ether_attach_p)(ifp);
1049 		CURVNET_RESTORE();
1050 	}
1051 }
1052 #endif
1053 
1054 SYSCTL_DECL(_net_link);
1055 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1056 #if defined(INET) || defined(INET6)
1057 SYSCTL_VNET_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
1058 	     &VNET_NAME(ether_ipfw), 0, "Pass ether pkts through firewall");
1059 #endif
1060 
1061 #if 0
1062 /*
1063  * This is for reference.  We have a table-driven version
1064  * of the little-endian crc32 generator, which is faster
1065  * than the double-loop.
1066  */
1067 uint32_t
1068 ether_crc32_le(const uint8_t *buf, size_t len)
1069 {
1070 	size_t i;
1071 	uint32_t crc;
1072 	int bit;
1073 	uint8_t data;
1074 
1075 	crc = 0xffffffff;	/* initial value */
1076 
1077 	for (i = 0; i < len; i++) {
1078 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1079 			carry = (crc ^ data) & 1;
1080 			crc >>= 1;
1081 			if (carry)
1082 				crc = (crc ^ ETHER_CRC_POLY_LE);
1083 		}
1084 	}
1085 
1086 	return (crc);
1087 }
1088 #else
1089 uint32_t
1090 ether_crc32_le(const uint8_t *buf, size_t len)
1091 {
1092 	static const uint32_t crctab[] = {
1093 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1094 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1095 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1096 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1097 	};
1098 	size_t i;
1099 	uint32_t crc;
1100 
1101 	crc = 0xffffffff;	/* initial value */
1102 
1103 	for (i = 0; i < len; i++) {
1104 		crc ^= buf[i];
1105 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1106 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1107 	}
1108 
1109 	return (crc);
1110 }
1111 #endif
1112 
1113 uint32_t
1114 ether_crc32_be(const uint8_t *buf, size_t len)
1115 {
1116 	size_t i;
1117 	uint32_t crc, carry;
1118 	int bit;
1119 	uint8_t data;
1120 
1121 	crc = 0xffffffff;	/* initial value */
1122 
1123 	for (i = 0; i < len; i++) {
1124 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1125 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1126 			crc <<= 1;
1127 			if (carry)
1128 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1129 		}
1130 	}
1131 
1132 	return (crc);
1133 }
1134 
1135 int
1136 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1137 {
1138 	struct ifaddr *ifa = (struct ifaddr *) data;
1139 	struct ifreq *ifr = (struct ifreq *) data;
1140 	int error = 0;
1141 
1142 	switch (command) {
1143 	case SIOCSIFADDR:
1144 		ifp->if_flags |= IFF_UP;
1145 
1146 		switch (ifa->ifa_addr->sa_family) {
1147 #ifdef INET
1148 		case AF_INET:
1149 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1150 			arp_ifinit(ifp, ifa);
1151 			break;
1152 #endif
1153 #ifdef IPX
1154 		/*
1155 		 * XXX - This code is probably wrong
1156 		 */
1157 		case AF_IPX:
1158 			{
1159 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1160 
1161 			if (ipx_nullhost(*ina))
1162 				ina->x_host =
1163 				    *(union ipx_host *)
1164 				    IF_LLADDR(ifp);
1165 			else {
1166 				bcopy((caddr_t) ina->x_host.c_host,
1167 				      (caddr_t) IF_LLADDR(ifp),
1168 				      ETHER_ADDR_LEN);
1169 			}
1170 
1171 			/*
1172 			 * Set new address
1173 			 */
1174 			ifp->if_init(ifp->if_softc);
1175 			break;
1176 			}
1177 #endif
1178 		default:
1179 			ifp->if_init(ifp->if_softc);
1180 			break;
1181 		}
1182 		break;
1183 
1184 	case SIOCGIFADDR:
1185 		{
1186 			struct sockaddr *sa;
1187 
1188 			sa = (struct sockaddr *) & ifr->ifr_data;
1189 			bcopy(IF_LLADDR(ifp),
1190 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1191 		}
1192 		break;
1193 
1194 	case SIOCSIFMTU:
1195 		/*
1196 		 * Set the interface MTU.
1197 		 */
1198 		if (ifr->ifr_mtu > ETHERMTU) {
1199 			error = EINVAL;
1200 		} else {
1201 			ifp->if_mtu = ifr->ifr_mtu;
1202 		}
1203 		break;
1204 	default:
1205 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1206 		break;
1207 	}
1208 	return (error);
1209 }
1210 
1211 static int
1212 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1213 	struct sockaddr *sa)
1214 {
1215 	struct sockaddr_dl *sdl;
1216 #ifdef INET
1217 	struct sockaddr_in *sin;
1218 #endif
1219 #ifdef INET6
1220 	struct sockaddr_in6 *sin6;
1221 #endif
1222 	u_char *e_addr;
1223 
1224 	switch(sa->sa_family) {
1225 	case AF_LINK:
1226 		/*
1227 		 * No mapping needed. Just check that it's a valid MC address.
1228 		 */
1229 		sdl = (struct sockaddr_dl *)sa;
1230 		e_addr = LLADDR(sdl);
1231 		if (!ETHER_IS_MULTICAST(e_addr))
1232 			return EADDRNOTAVAIL;
1233 		*llsa = 0;
1234 		return 0;
1235 
1236 #ifdef INET
1237 	case AF_INET:
1238 		sin = (struct sockaddr_in *)sa;
1239 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1240 			return EADDRNOTAVAIL;
1241 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1242 		       M_NOWAIT|M_ZERO);
1243 		if (sdl == NULL)
1244 			return ENOMEM;
1245 		sdl->sdl_len = sizeof *sdl;
1246 		sdl->sdl_family = AF_LINK;
1247 		sdl->sdl_index = ifp->if_index;
1248 		sdl->sdl_type = IFT_ETHER;
1249 		sdl->sdl_alen = ETHER_ADDR_LEN;
1250 		e_addr = LLADDR(sdl);
1251 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1252 		*llsa = (struct sockaddr *)sdl;
1253 		return 0;
1254 #endif
1255 #ifdef INET6
1256 	case AF_INET6:
1257 		sin6 = (struct sockaddr_in6 *)sa;
1258 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1259 			/*
1260 			 * An IP6 address of 0 means listen to all
1261 			 * of the Ethernet multicast address used for IP6.
1262 			 * (This is used for multicast routers.)
1263 			 */
1264 			ifp->if_flags |= IFF_ALLMULTI;
1265 			*llsa = 0;
1266 			return 0;
1267 		}
1268 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1269 			return EADDRNOTAVAIL;
1270 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1271 		       M_NOWAIT|M_ZERO);
1272 		if (sdl == NULL)
1273 			return (ENOMEM);
1274 		sdl->sdl_len = sizeof *sdl;
1275 		sdl->sdl_family = AF_LINK;
1276 		sdl->sdl_index = ifp->if_index;
1277 		sdl->sdl_type = IFT_ETHER;
1278 		sdl->sdl_alen = ETHER_ADDR_LEN;
1279 		e_addr = LLADDR(sdl);
1280 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1281 		*llsa = (struct sockaddr *)sdl;
1282 		return 0;
1283 #endif
1284 
1285 	default:
1286 		/*
1287 		 * Well, the text isn't quite right, but it's the name
1288 		 * that counts...
1289 		 */
1290 		return EAFNOSUPPORT;
1291 	}
1292 }
1293 
1294 static void*
1295 ether_alloc(u_char type, struct ifnet *ifp)
1296 {
1297 	struct arpcom	*ac;
1298 
1299 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1300 	ac->ac_ifp = ifp;
1301 
1302 	return (ac);
1303 }
1304 
1305 static void
1306 ether_free(void *com, u_char type)
1307 {
1308 
1309 	free(com, M_ARPCOM);
1310 }
1311 
1312 static int
1313 ether_modevent(module_t mod, int type, void *data)
1314 {
1315 
1316 	switch (type) {
1317 	case MOD_LOAD:
1318 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1319 		break;
1320 	case MOD_UNLOAD:
1321 		if_deregister_com_alloc(IFT_ETHER);
1322 		break;
1323 	default:
1324 		return EOPNOTSUPP;
1325 	}
1326 
1327 	return (0);
1328 }
1329 
1330 static moduledata_t ether_mod = {
1331 	"ether",
1332 	ether_modevent,
1333 	0
1334 };
1335 
1336 void
1337 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1338 {
1339 	struct ether_vlan_header vlan;
1340 	struct mbuf mv, mb;
1341 
1342 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1343 	    ("%s: vlan information not present", __func__));
1344 	KASSERT(m->m_len >= sizeof(struct ether_header),
1345 	    ("%s: mbuf not large enough for header", __func__));
1346 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1347 	vlan.evl_proto = vlan.evl_encap_proto;
1348 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1349 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1350 	m->m_len -= sizeof(struct ether_header);
1351 	m->m_data += sizeof(struct ether_header);
1352 	/*
1353 	 * If a data link has been supplied by the caller, then we will need to
1354 	 * re-create a stack allocated mbuf chain with the following structure:
1355 	 *
1356 	 * (1) mbuf #1 will contain the supplied data link
1357 	 * (2) mbuf #2 will contain the vlan header
1358 	 * (3) mbuf #3 will contain the original mbuf's packet data
1359 	 *
1360 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1361 	 */
1362 	if (data != NULL) {
1363 		mv.m_next = m;
1364 		mv.m_data = (caddr_t)&vlan;
1365 		mv.m_len = sizeof(vlan);
1366 		mb.m_next = &mv;
1367 		mb.m_data = data;
1368 		mb.m_len = dlen;
1369 		bpf_mtap(bp, &mb);
1370 	} else
1371 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1372 	m->m_len += sizeof(struct ether_header);
1373 	m->m_data -= sizeof(struct ether_header);
1374 }
1375 
1376 struct mbuf *
1377 ether_vlanencap(struct mbuf *m, uint16_t tag)
1378 {
1379 	struct ether_vlan_header *evl;
1380 
1381 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
1382 	if (m == NULL)
1383 		return (NULL);
1384 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1385 
1386 	if (m->m_len < sizeof(*evl)) {
1387 		m = m_pullup(m, sizeof(*evl));
1388 		if (m == NULL)
1389 			return (NULL);
1390 	}
1391 
1392 	/*
1393 	 * Transform the Ethernet header into an Ethernet header
1394 	 * with 802.1Q encapsulation.
1395 	 */
1396 	evl = mtod(m, struct ether_vlan_header *);
1397 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1398 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1399 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1400 	evl->evl_tag = htons(tag);
1401 	return (m);
1402 }
1403 
1404 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1405 MODULE_VERSION(ether, 1);
1406