xref: /freebsd/sys/net/if_ethersubr.c (revision 7aa383846770374466b1dcb2cefd71bde9acf463)
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_carp.h"
39 #include "opt_mbuf_profiling.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/module.h>
47 #include <sys/mbuf.h>
48 #include <sys/random.h>
49 #include <sys/rwlock.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 
54 #include <net/if.h>
55 #include <net/if_arp.h>
56 #include <net/netisr.h>
57 #include <net/route.h>
58 #include <net/if_llc.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/bpf.h>
62 #include <net/ethernet.h>
63 #include <net/if_bridgevar.h>
64 #include <net/if_vlan_var.h>
65 #include <net/if_llatbl.h>
66 #include <net/pf_mtag.h>
67 #include <net/vnet.h>
68 
69 #if defined(INET) || defined(INET6)
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.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 #if defined(INET) || defined(INET6)
82 #ifdef DEV_CARP
83 #include <netinet/ip_carp.h>
84 #endif
85 #endif
86 
87 #ifdef IPX
88 #include <netipx/ipx.h>
89 #include <netipx/ipx_if.h>
90 #endif
91 
92 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
93 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
94 		struct sockaddr *dst, short *tp, int *hlen);
95 
96 #ifdef NETATALK
97 #include <netatalk/at.h>
98 #include <netatalk/at_var.h>
99 #include <netatalk/at_extern.h>
100 
101 #define llc_snap_org_code llc_un.type_snap.org_code
102 #define llc_snap_ether_type llc_un.type_snap.ether_type
103 
104 extern u_char	at_org_code[3];
105 extern u_char	aarp_org_code[3];
106 #endif /* NETATALK */
107 
108 #include <security/mac/mac_framework.h>
109 
110 #ifdef CTASSERT
111 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
112 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
113 #endif
114 
115 /* netgraph node hooks for ng_ether(4) */
116 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
117 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
118 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
119 void	(*ng_ether_attach_p)(struct ifnet *ifp);
120 void	(*ng_ether_detach_p)(struct ifnet *ifp);
121 
122 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
123 
124 /* if_bridge(4) support */
125 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
126 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
127 		struct sockaddr *, struct rtentry *);
128 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
129 
130 /* if_lagg(4) support */
131 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
132 
133 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
134 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
135 
136 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
137 		struct sockaddr *);
138 
139 /* XXX: should be in an arp support file, not here */
140 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
141 
142 #define	ETHER_IS_BROADCAST(addr) \
143 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
144 
145 #define senderr(e) do { error = (e); goto bad;} while (0)
146 
147 #if defined(INET) || defined(INET6)
148 int
149 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared);
150 static VNET_DEFINE(int, ether_ipfw);
151 #define	V_ether_ipfw	VNET(ether_ipfw)
152 #endif
153 
154 
155 /*
156  * Ethernet output routine.
157  * Encapsulate a packet of type family for the local net.
158  * Use trailer local net encapsulation if enough data in first
159  * packet leaves a multiple of 512 bytes of data in remainder.
160  */
161 int
162 ether_output(struct ifnet *ifp, struct mbuf *m,
163 	struct sockaddr *dst, struct route *ro)
164 {
165 	short type;
166 	int error = 0, hdrcmplt = 0;
167 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
168 	struct llentry *lle = NULL;
169 	struct rtentry *rt0 = NULL;
170 	struct ether_header *eh;
171 	struct pf_mtag *t;
172 	int loop_copy = 1;
173 	int hlen;	/* link layer header length */
174 
175 	if (ro != NULL) {
176 		if (!(m->m_flags & (M_BCAST | M_MCAST)))
177 			lle = ro->ro_lle;
178 		rt0 = ro->ro_rt;
179 	}
180 #ifdef MAC
181 	error = mac_ifnet_check_transmit(ifp, m);
182 	if (error)
183 		senderr(error);
184 #endif
185 
186 	M_PROFILE(m);
187 	if (ifp->if_flags & IFF_MONITOR)
188 		senderr(ENETDOWN);
189 	if (!((ifp->if_flags & IFF_UP) &&
190 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
191 		senderr(ENETDOWN);
192 
193 	hlen = ETHER_HDR_LEN;
194 	switch (dst->sa_family) {
195 #ifdef INET
196 	case AF_INET:
197 		if (lle != NULL && (lle->la_flags & LLE_VALID))
198 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
199 		else
200 			error = arpresolve(ifp, rt0, m, dst, edst, &lle);
201 		if (error)
202 			return (error == EWOULDBLOCK ? 0 : error);
203 		type = htons(ETHERTYPE_IP);
204 		break;
205 	case AF_ARP:
206 	{
207 		struct arphdr *ah;
208 		ah = mtod(m, struct arphdr *);
209 		ah->ar_hrd = htons(ARPHRD_ETHER);
210 
211 		loop_copy = 0; /* if this is for us, don't do it */
212 
213 		switch(ntohs(ah->ar_op)) {
214 		case ARPOP_REVREQUEST:
215 		case ARPOP_REVREPLY:
216 			type = htons(ETHERTYPE_REVARP);
217 			break;
218 		case ARPOP_REQUEST:
219 		case ARPOP_REPLY:
220 		default:
221 			type = htons(ETHERTYPE_ARP);
222 			break;
223 		}
224 
225 		if (m->m_flags & M_BCAST)
226 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
227 		else
228 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
229 
230 	}
231 	break;
232 #endif
233 #ifdef INET6
234 	case AF_INET6:
235 		if (lle != NULL && (lle->la_flags & LLE_VALID))
236 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
237 		else
238 			error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle);
239 		if (error)
240 			return error;
241 		type = htons(ETHERTYPE_IPV6);
242 		break;
243 #endif
244 #ifdef IPX
245 	case AF_IPX:
246 		if (ef_outputp) {
247 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
248 		    if (error)
249 			goto bad;
250 		} else
251 		    type = htons(ETHERTYPE_IPX);
252 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
253 		    (caddr_t)edst, sizeof (edst));
254 		break;
255 #endif
256 #ifdef NETATALK
257 	case AF_APPLETALK:
258 	  {
259 	    struct at_ifaddr *aa;
260 
261 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
262 		    senderr(EHOSTUNREACH); /* XXX */
263 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) {
264 		    ifa_free(&aa->aa_ifa);
265 		    return (0);
266 	    }
267 	    /*
268 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
269 	     */
270 	    if ( aa->aa_flags & AFA_PHASE2 ) {
271 		struct llc llc;
272 
273 		ifa_free(&aa->aa_ifa);
274 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
275 		if (m == NULL)
276 			senderr(ENOBUFS);
277 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
278 		llc.llc_control = LLC_UI;
279 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
280 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
281 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
282 		type = htons(m->m_pkthdr.len);
283 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
284 	    } else {
285 		ifa_free(&aa->aa_ifa);
286 		type = htons(ETHERTYPE_AT);
287 	    }
288 	    break;
289 	  }
290 #endif /* NETATALK */
291 
292 	case pseudo_AF_HDRCMPLT:
293 		hdrcmplt = 1;
294 		eh = (struct ether_header *)dst->sa_data;
295 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
296 		/* FALLTHROUGH */
297 
298 	case AF_UNSPEC:
299 		loop_copy = 0; /* if this is for us, don't do it */
300 		eh = (struct ether_header *)dst->sa_data;
301 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
302 		type = eh->ether_type;
303 		break;
304 
305 	default:
306 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
307 		senderr(EAFNOSUPPORT);
308 	}
309 
310 	if (lle != NULL && (lle->la_flags & LLE_IFADDR)) {
311 		int csum_flags = 0;
312 		if (m->m_pkthdr.csum_flags & CSUM_IP)
313 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
314 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
315 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
316 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
317 			csum_flags |= CSUM_SCTP_VALID;
318 		m->m_pkthdr.csum_flags |= csum_flags;
319 		m->m_pkthdr.csum_data = 0xffff;
320 		return (if_simloop(ifp, m, dst->sa_family, 0));
321 	}
322 
323 	/*
324 	 * Add local net header.  If no space in first mbuf,
325 	 * allocate another.
326 	 */
327 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
328 	if (m == NULL)
329 		senderr(ENOBUFS);
330 	eh = mtod(m, struct ether_header *);
331 	(void)memcpy(&eh->ether_type, &type,
332 		sizeof(eh->ether_type));
333 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
334 	if (hdrcmplt)
335 		(void)memcpy(eh->ether_shost, esrc,
336 			sizeof(eh->ether_shost));
337 	else
338 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
339 			sizeof(eh->ether_shost));
340 
341 	/*
342 	 * If a simplex interface, and the packet is being sent to our
343 	 * Ethernet address or a broadcast address, loopback a copy.
344 	 * XXX To make a simplex device behave exactly like a duplex
345 	 * device, we should copy in the case of sending to our own
346 	 * ethernet address (thus letting the original actually appear
347 	 * on the wire). However, we don't do that here for security
348 	 * reasons and compatibility with the original behavior.
349 	 */
350 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
351 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
352 		int csum_flags = 0;
353 
354 		if (m->m_pkthdr.csum_flags & CSUM_IP)
355 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
356 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
357 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
358 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
359 			csum_flags |= CSUM_SCTP_VALID;
360 
361 		if (m->m_flags & M_BCAST) {
362 			struct mbuf *n;
363 
364 			/*
365 			 * Because if_simloop() modifies the packet, we need a
366 			 * writable copy through m_dup() instead of a readonly
367 			 * one as m_copy[m] would give us. The alternative would
368 			 * be to modify if_simloop() to handle the readonly mbuf,
369 			 * but performancewise it is mostly equivalent (trading
370 			 * extra data copying vs. extra locking).
371 			 *
372 			 * XXX This is a local workaround.  A number of less
373 			 * often used kernel parts suffer from the same bug.
374 			 * See PR kern/105943 for a proposed general solution.
375 			 */
376 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
377 				n->m_pkthdr.csum_flags |= csum_flags;
378 				if (csum_flags & CSUM_DATA_VALID)
379 					n->m_pkthdr.csum_data = 0xffff;
380 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
381 			} else
382 				ifp->if_iqdrops++;
383 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
384 				ETHER_ADDR_LEN) == 0) {
385 			m->m_pkthdr.csum_flags |= csum_flags;
386 			if (csum_flags & CSUM_DATA_VALID)
387 				m->m_pkthdr.csum_data = 0xffff;
388 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
389 			return (0);	/* XXX */
390 		}
391 	}
392 
393        /*
394 	* Bridges require special output handling.
395 	*/
396 	if (ifp->if_bridge) {
397 		BRIDGE_OUTPUT(ifp, m, error);
398 		return (error);
399 	}
400 
401 #if defined(INET) || defined(INET6)
402 #ifdef DEV_CARP
403 	if (ifp->if_carp &&
404 	    (error = carp_output(ifp, m, dst, NULL)))
405 		goto bad;
406 #endif
407 #endif
408 
409 	/* Handle ng_ether(4) processing, if any */
410 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
411 		KASSERT(ng_ether_output_p != NULL,
412 		    ("ng_ether_output_p is NULL"));
413 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
414 bad:			if (m != NULL)
415 				m_freem(m);
416 			return (error);
417 		}
418 		if (m == NULL)
419 			return (0);
420 	}
421 
422 	/* Continue with link-layer output */
423 	return ether_output_frame(ifp, m);
424 }
425 
426 /*
427  * Ethernet link layer output routine to send a raw frame to the device.
428  *
429  * This assumes that the 14 byte Ethernet header is present and contiguous
430  * in the first mbuf (if BRIDGE'ing).
431  */
432 int
433 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
434 {
435 #if defined(INET) || defined(INET6)
436 
437 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0) {
438 		if (ether_ipfw_chk(&m, ifp, 0) == 0) {
439 			if (m) {
440 				m_freem(m);
441 				return EACCES;	/* pkt dropped */
442 			} else
443 				return 0;	/* consumed e.g. in a pipe */
444 		}
445 	}
446 #endif
447 
448 	/*
449 	 * Queue message on interface, update output statistics if
450 	 * successful, and start output if interface not yet active.
451 	 */
452 	return ((ifp->if_transmit)(ifp, m));
453 }
454 
455 #if defined(INET) || defined(INET6)
456 /*
457  * ipfw processing for ethernet packets (in and out).
458  * The second parameter is NULL from ether_demux, and ifp from
459  * ether_output_frame.
460  */
461 int
462 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared)
463 {
464 	struct ether_header *eh;
465 	struct ether_header save_eh;
466 	struct mbuf *m;
467 	int i;
468 	struct ip_fw_args args;
469 	struct m_tag *mtag;
470 
471 	/* fetch start point from rule, if any */
472 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
473 	if (mtag == NULL) {
474 		args.rule.slot = 0;
475 	} else {
476 		/* dummynet packet, already partially processed */
477 		struct ipfw_rule_ref *r;
478 
479 		/* XXX can we free it after use ? */
480 		mtag->m_tag_id = PACKET_TAG_NONE;
481 		r = (struct ipfw_rule_ref *)(mtag + 1);
482 		if (r->info & IPFW_ONEPASS)
483 			return (1);
484 		args.rule = *r;
485 	}
486 
487 	/*
488 	 * I need some amt of data to be contiguous, and in case others need
489 	 * the packet (shared==1) also better be in the first mbuf.
490 	 */
491 	m = *m0;
492 	i = min( m->m_pkthdr.len, max_protohdr);
493 	if ( shared || m->m_len < i) {
494 		m = m_pullup(m, i);
495 		if (m == NULL) {
496 			*m0 = m;
497 			return 0;
498 		}
499 	}
500 	eh = mtod(m, struct ether_header *);
501 	save_eh = *eh;			/* save copy for restore below */
502 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
503 
504 	args.m = m;		/* the packet we are looking at		*/
505 	args.oif = dst;		/* destination, if any			*/
506 	args.next_hop = NULL;	/* we do not support forward yet	*/
507 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
508 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
509 	i = V_ip_fw_chk_ptr(&args);
510 	m = args.m;
511 	if (m != NULL) {
512 		/*
513 		 * Restore Ethernet header, as needed, in case the
514 		 * mbuf chain was replaced by ipfw.
515 		 */
516 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
517 		if (m == NULL) {
518 			*m0 = m;
519 			return 0;
520 		}
521 		if (eh != mtod(m, struct ether_header *))
522 			bcopy(&save_eh, mtod(m, struct ether_header *),
523 				ETHER_HDR_LEN);
524 	}
525 	*m0 = m;
526 
527 	if (i == IP_FW_DENY) /* drop */
528 		return 0;
529 
530 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
531 
532 	if (i == IP_FW_PASS) /* a PASS rule.  */
533 		return 1;
534 
535 	if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
536 		int dir;
537 		/*
538 		 * Pass the pkt to dummynet, which consumes it.
539 		 * If shared, make a copy and keep the original.
540 		 */
541 		if (shared) {
542 			m = m_copypacket(m, M_DONTWAIT);
543 			if (m == NULL)
544 				return 0;
545 		} else {
546 			/*
547 			 * Pass the original to dummynet and
548 			 * nothing back to the caller
549 			 */
550 			*m0 = NULL ;
551 		}
552 		dir = PROTO_LAYER2 | (dst ? DIR_OUT : DIR_IN);
553 		ip_dn_io_ptr(&m, dir, &args);
554 		return 0;
555 	}
556 	/*
557 	 * XXX at some point add support for divert/forward actions.
558 	 * If none of the above matches, we have to drop the pkt.
559 	 */
560 	return 0;
561 }
562 #endif
563 
564 /*
565  * Process a received Ethernet packet; the packet is in the
566  * mbuf chain m with the ethernet header at the front.
567  */
568 static void
569 ether_input(struct ifnet *ifp, struct mbuf *m)
570 {
571 	struct ether_header *eh;
572 	u_short etype;
573 
574 	if ((ifp->if_flags & IFF_UP) == 0) {
575 		m_freem(m);
576 		return;
577 	}
578 #ifdef DIAGNOSTIC
579 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
580 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
581 		m_freem(m);
582 		return;
583 	}
584 #endif
585 	/*
586 	 * Do consistency checks to verify assumptions
587 	 * made by code past this point.
588 	 */
589 	if ((m->m_flags & M_PKTHDR) == 0) {
590 		if_printf(ifp, "discard frame w/o packet header\n");
591 		ifp->if_ierrors++;
592 		m_freem(m);
593 		return;
594 	}
595 	if (m->m_len < ETHER_HDR_LEN) {
596 		/* XXX maybe should pullup? */
597 		if_printf(ifp, "discard frame w/o leading ethernet "
598 				"header (len %u pkt len %u)\n",
599 				m->m_len, m->m_pkthdr.len);
600 		ifp->if_ierrors++;
601 		m_freem(m);
602 		return;
603 	}
604 	eh = mtod(m, struct ether_header *);
605 	etype = ntohs(eh->ether_type);
606 	if (m->m_pkthdr.rcvif == NULL) {
607 		if_printf(ifp, "discard frame w/o interface pointer\n");
608 		ifp->if_ierrors++;
609 		m_freem(m);
610 		return;
611 	}
612 #ifdef DIAGNOSTIC
613 	if (m->m_pkthdr.rcvif != ifp) {
614 		if_printf(ifp, "Warning, frame marked as received on %s\n",
615 			m->m_pkthdr.rcvif->if_xname);
616 	}
617 #endif
618 
619 	CURVNET_SET_QUIET(ifp->if_vnet);
620 
621 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
622 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
623 			m->m_flags |= M_BCAST;
624 		else
625 			m->m_flags |= M_MCAST;
626 		ifp->if_imcasts++;
627 	}
628 
629 #ifdef MAC
630 	/*
631 	 * Tag the mbuf with an appropriate MAC label before any other
632 	 * consumers can get to it.
633 	 */
634 	mac_ifnet_create_mbuf(ifp, m);
635 #endif
636 
637 	/*
638 	 * Give bpf a chance at the packet.
639 	 */
640 	ETHER_BPF_MTAP(ifp, m);
641 
642 	/*
643 	 * If the CRC is still on the packet, trim it off. We do this once
644 	 * and once only in case we are re-entered. Nothing else on the
645 	 * Ethernet receive path expects to see the FCS.
646 	 */
647 	if (m->m_flags & M_HASFCS) {
648 		m_adj(m, -ETHER_CRC_LEN);
649 		m->m_flags &= ~M_HASFCS;
650 	}
651 
652 	ifp->if_ibytes += m->m_pkthdr.len;
653 
654 	/* Allow monitor mode to claim this frame, after stats are updated. */
655 	if (ifp->if_flags & IFF_MONITOR) {
656 		m_freem(m);
657 		CURVNET_RESTORE();
658 		return;
659 	}
660 
661 	/* Handle input from a lagg(4) port */
662 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
663 		KASSERT(lagg_input_p != NULL,
664 		    ("%s: if_lagg not loaded!", __func__));
665 		m = (*lagg_input_p)(ifp, m);
666 		if (m != NULL)
667 			ifp = m->m_pkthdr.rcvif;
668 		else
669 			return;
670 	}
671 
672 	/*
673 	 * If the hardware did not process an 802.1Q tag, do this now,
674 	 * to allow 802.1P priority frames to be passed to the main input
675 	 * path correctly.
676 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
677 	 */
678 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
679 		struct ether_vlan_header *evl;
680 
681 		if (m->m_len < sizeof(*evl) &&
682 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
683 #ifdef DIAGNOSTIC
684 			if_printf(ifp, "cannot pullup VLAN header\n");
685 #endif
686 			ifp->if_ierrors++;
687 			m_freem(m);
688 			return;
689 		}
690 
691 		evl = mtod(m, struct ether_vlan_header *);
692 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
693 		m->m_flags |= M_VLANTAG;
694 
695 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
696 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
697 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
698 	}
699 
700 	/* Allow ng_ether(4) to claim this frame. */
701 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
702 		KASSERT(ng_ether_input_p != NULL,
703 		    ("%s: ng_ether_input_p is NULL", __func__));
704 		m->m_flags &= ~M_PROMISC;
705 		(*ng_ether_input_p)(ifp, &m);
706 		if (m == NULL) {
707 			CURVNET_RESTORE();
708 			return;
709 		}
710 	}
711 
712 	/*
713 	 * Allow if_bridge(4) to claim this frame.
714 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
715 	 * and the frame should be delivered locally.
716 	 */
717 	if (ifp->if_bridge != NULL) {
718 		m->m_flags &= ~M_PROMISC;
719 		BRIDGE_INPUT(ifp, m);
720 		if (m == NULL) {
721 			CURVNET_RESTORE();
722 			return;
723 		}
724 	}
725 
726 #if defined(INET) || defined(INET6)
727 #ifdef DEV_CARP
728 	/*
729 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
730 	 * mbuf flows up to Layer 3.
731 	 * FreeBSD's implementation of carp(4) uses the inprotosw
732 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
733 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
734 	 * is outside the scope of the M_PROMISC test below.
735 	 * TODO: Maintain a hash table of ethernet addresses other than
736 	 * ether_dhost which may be active on this ifp.
737 	 */
738 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) {
739 		m->m_flags &= ~M_PROMISC;
740 	} else
741 #endif
742 #endif
743 	{
744 		/*
745 		 * If the frame received was not for our MAC address, set the
746 		 * M_PROMISC flag on the mbuf chain. The frame may need to
747 		 * be seen by the rest of the Ethernet input path in case of
748 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
749 		 * seen by upper protocol layers.
750 		 */
751 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
752 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
753 			m->m_flags |= M_PROMISC;
754 	}
755 
756 	/* First chunk of an mbuf contains good entropy */
757 	if (harvest.ethernet)
758 		random_harvest(m, 16, 3, 0, RANDOM_NET);
759 
760 	ether_demux(ifp, m);
761 	CURVNET_RESTORE();
762 }
763 
764 /*
765  * Upper layer processing for a received Ethernet packet.
766  */
767 void
768 ether_demux(struct ifnet *ifp, struct mbuf *m)
769 {
770 	struct ether_header *eh;
771 	int isr;
772 	u_short ether_type;
773 #if defined(NETATALK)
774 	struct llc *l;
775 #endif
776 
777 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
778 
779 #if defined(INET) || defined(INET6)
780 	/*
781 	 * Allow dummynet and/or ipfw to claim the frame.
782 	 * Do not do this for PROMISC frames in case we are re-entered.
783 	 */
784 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
785 		if (ether_ipfw_chk(&m, NULL, 0) == 0) {
786 			if (m)
787 				m_freem(m);	/* dropped; free mbuf chain */
788 			return;			/* consumed */
789 		}
790 	}
791 #endif
792 	eh = mtod(m, struct ether_header *);
793 	ether_type = ntohs(eh->ether_type);
794 
795 	/*
796 	 * If this frame has a VLAN tag other than 0, call vlan_input()
797 	 * if its module is loaded. Otherwise, drop.
798 	 */
799 	if ((m->m_flags & M_VLANTAG) &&
800 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
801 		if (ifp->if_vlantrunk == NULL) {
802 			ifp->if_noproto++;
803 			m_freem(m);
804 			return;
805 		}
806 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
807 		    __func__));
808 		/* Clear before possibly re-entering ether_input(). */
809 		m->m_flags &= ~M_PROMISC;
810 		(*vlan_input_p)(ifp, m);
811 		return;
812 	}
813 
814 	/*
815 	 * Pass promiscuously received frames to the upper layer if the user
816 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
817 	 */
818 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
819 		m_freem(m);
820 		return;
821 	}
822 
823 	/*
824 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
825 	 * Strip off Ethernet header.
826 	 */
827 	m->m_flags &= ~M_VLANTAG;
828 	m->m_flags &= ~(M_PROTOFLAGS);
829 	m_adj(m, ETHER_HDR_LEN);
830 
831 	/*
832 	 * Dispatch frame to upper layer.
833 	 */
834 	switch (ether_type) {
835 #ifdef INET
836 	case ETHERTYPE_IP:
837 		if ((m = ip_fastforward(m)) == NULL)
838 			return;
839 		isr = NETISR_IP;
840 		break;
841 
842 	case ETHERTYPE_ARP:
843 		if (ifp->if_flags & IFF_NOARP) {
844 			/* Discard packet if ARP is disabled on interface */
845 			m_freem(m);
846 			return;
847 		}
848 		isr = NETISR_ARP;
849 		break;
850 #endif
851 #ifdef IPX
852 	case ETHERTYPE_IPX:
853 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
854 			return;
855 		isr = NETISR_IPX;
856 		break;
857 #endif
858 #ifdef INET6
859 	case ETHERTYPE_IPV6:
860 		isr = NETISR_IPV6;
861 		break;
862 #endif
863 #ifdef NETATALK
864 	case ETHERTYPE_AT:
865 		isr = NETISR_ATALK1;
866 		break;
867 	case ETHERTYPE_AARP:
868 		isr = NETISR_AARP;
869 		break;
870 #endif /* NETATALK */
871 	default:
872 #ifdef IPX
873 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
874 			return;
875 #endif /* IPX */
876 #if defined(NETATALK)
877 		if (ether_type > ETHERMTU)
878 			goto discard;
879 		l = mtod(m, struct llc *);
880 		if (l->llc_dsap == LLC_SNAP_LSAP &&
881 		    l->llc_ssap == LLC_SNAP_LSAP &&
882 		    l->llc_control == LLC_UI) {
883 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
884 			    sizeof(at_org_code)) == 0 &&
885 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
886 				m_adj(m, LLC_SNAPFRAMELEN);
887 				isr = NETISR_ATALK2;
888 				break;
889 			}
890 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
891 			    sizeof(aarp_org_code)) == 0 &&
892 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
893 				m_adj(m, LLC_SNAPFRAMELEN);
894 				isr = NETISR_AARP;
895 				break;
896 			}
897 		}
898 #endif /* NETATALK */
899 		goto discard;
900 	}
901 	netisr_dispatch(isr, m);
902 	return;
903 
904 discard:
905 	/*
906 	 * Packet is to be discarded.  If netgraph is present,
907 	 * hand the packet to it for last chance processing;
908 	 * otherwise dispose of it.
909 	 */
910 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
911 		KASSERT(ng_ether_input_orphan_p != NULL,
912 		    ("ng_ether_input_orphan_p is NULL"));
913 		/*
914 		 * Put back the ethernet header so netgraph has a
915 		 * consistent view of inbound packets.
916 		 */
917 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
918 		(*ng_ether_input_orphan_p)(ifp, m);
919 		return;
920 	}
921 	m_freem(m);
922 }
923 
924 /*
925  * Convert Ethernet address to printable (loggable) representation.
926  * This routine is for compatibility; it's better to just use
927  *
928  *	printf("%6D", <pointer to address>, ":");
929  *
930  * since there's no static buffer involved.
931  */
932 char *
933 ether_sprintf(const u_char *ap)
934 {
935 	static char etherbuf[18];
936 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
937 	return (etherbuf);
938 }
939 
940 /*
941  * Perform common duties while attaching to interface list
942  */
943 void
944 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
945 {
946 	int i;
947 	struct ifaddr *ifa;
948 	struct sockaddr_dl *sdl;
949 
950 	ifp->if_addrlen = ETHER_ADDR_LEN;
951 	ifp->if_hdrlen = ETHER_HDR_LEN;
952 	if_attach(ifp);
953 	ifp->if_mtu = ETHERMTU;
954 	ifp->if_output = ether_output;
955 	ifp->if_input = ether_input;
956 	ifp->if_resolvemulti = ether_resolvemulti;
957 	if (ifp->if_baudrate == 0)
958 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
959 	ifp->if_broadcastaddr = etherbroadcastaddr;
960 
961 	ifa = ifp->if_addr;
962 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
963 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
964 	sdl->sdl_type = IFT_ETHER;
965 	sdl->sdl_alen = ifp->if_addrlen;
966 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
967 
968 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
969 	if (ng_ether_attach_p != NULL)
970 		(*ng_ether_attach_p)(ifp);
971 
972 	/* Announce Ethernet MAC address if non-zero. */
973 	for (i = 0; i < ifp->if_addrlen; i++)
974 		if (lla[i] != 0)
975 			break;
976 	if (i != ifp->if_addrlen)
977 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
978 }
979 
980 /*
981  * Perform common duties while detaching an Ethernet interface
982  */
983 void
984 ether_ifdetach(struct ifnet *ifp)
985 {
986 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
987 		KASSERT(ng_ether_detach_p != NULL,
988 		    ("ng_ether_detach_p is NULL"));
989 		(*ng_ether_detach_p)(ifp);
990 	}
991 
992 	bpfdetach(ifp);
993 	if_detach(ifp);
994 }
995 
996 SYSCTL_DECL(_net_link);
997 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
998 #if defined(INET) || defined(INET6)
999 SYSCTL_VNET_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
1000 	     &VNET_NAME(ether_ipfw), 0, "Pass ether pkts through firewall");
1001 #endif
1002 
1003 #if 0
1004 /*
1005  * This is for reference.  We have a table-driven version
1006  * of the little-endian crc32 generator, which is faster
1007  * than the double-loop.
1008  */
1009 uint32_t
1010 ether_crc32_le(const uint8_t *buf, size_t len)
1011 {
1012 	size_t i;
1013 	uint32_t crc;
1014 	int bit;
1015 	uint8_t data;
1016 
1017 	crc = 0xffffffff;	/* initial value */
1018 
1019 	for (i = 0; i < len; i++) {
1020 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1021 			carry = (crc ^ data) & 1;
1022 			crc >>= 1;
1023 			if (carry)
1024 				crc = (crc ^ ETHER_CRC_POLY_LE);
1025 		}
1026 	}
1027 
1028 	return (crc);
1029 }
1030 #else
1031 uint32_t
1032 ether_crc32_le(const uint8_t *buf, size_t len)
1033 {
1034 	static const uint32_t crctab[] = {
1035 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1036 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1037 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1038 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1039 	};
1040 	size_t i;
1041 	uint32_t crc;
1042 
1043 	crc = 0xffffffff;	/* initial value */
1044 
1045 	for (i = 0; i < len; i++) {
1046 		crc ^= buf[i];
1047 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1048 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1049 	}
1050 
1051 	return (crc);
1052 }
1053 #endif
1054 
1055 uint32_t
1056 ether_crc32_be(const uint8_t *buf, size_t len)
1057 {
1058 	size_t i;
1059 	uint32_t crc, carry;
1060 	int bit;
1061 	uint8_t data;
1062 
1063 	crc = 0xffffffff;	/* initial value */
1064 
1065 	for (i = 0; i < len; i++) {
1066 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1067 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1068 			crc <<= 1;
1069 			if (carry)
1070 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1071 		}
1072 	}
1073 
1074 	return (crc);
1075 }
1076 
1077 int
1078 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1079 {
1080 	struct ifaddr *ifa = (struct ifaddr *) data;
1081 	struct ifreq *ifr = (struct ifreq *) data;
1082 	int error = 0;
1083 
1084 	switch (command) {
1085 	case SIOCSIFADDR:
1086 		ifp->if_flags |= IFF_UP;
1087 
1088 		switch (ifa->ifa_addr->sa_family) {
1089 #ifdef INET
1090 		case AF_INET:
1091 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1092 			arp_ifinit(ifp, ifa);
1093 			break;
1094 #endif
1095 #ifdef IPX
1096 		/*
1097 		 * XXX - This code is probably wrong
1098 		 */
1099 		case AF_IPX:
1100 			{
1101 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1102 
1103 			if (ipx_nullhost(*ina))
1104 				ina->x_host =
1105 				    *(union ipx_host *)
1106 				    IF_LLADDR(ifp);
1107 			else {
1108 				bcopy((caddr_t) ina->x_host.c_host,
1109 				      (caddr_t) IF_LLADDR(ifp),
1110 				      ETHER_ADDR_LEN);
1111 			}
1112 
1113 			/*
1114 			 * Set new address
1115 			 */
1116 			ifp->if_init(ifp->if_softc);
1117 			break;
1118 			}
1119 #endif
1120 		default:
1121 			ifp->if_init(ifp->if_softc);
1122 			break;
1123 		}
1124 		break;
1125 
1126 	case SIOCGIFADDR:
1127 		{
1128 			struct sockaddr *sa;
1129 
1130 			sa = (struct sockaddr *) & ifr->ifr_data;
1131 			bcopy(IF_LLADDR(ifp),
1132 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1133 		}
1134 		break;
1135 
1136 	case SIOCSIFMTU:
1137 		/*
1138 		 * Set the interface MTU.
1139 		 */
1140 		if (ifr->ifr_mtu > ETHERMTU) {
1141 			error = EINVAL;
1142 		} else {
1143 			ifp->if_mtu = ifr->ifr_mtu;
1144 		}
1145 		break;
1146 	default:
1147 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1148 		break;
1149 	}
1150 	return (error);
1151 }
1152 
1153 static int
1154 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1155 	struct sockaddr *sa)
1156 {
1157 	struct sockaddr_dl *sdl;
1158 #ifdef INET
1159 	struct sockaddr_in *sin;
1160 #endif
1161 #ifdef INET6
1162 	struct sockaddr_in6 *sin6;
1163 #endif
1164 	u_char *e_addr;
1165 
1166 	switch(sa->sa_family) {
1167 	case AF_LINK:
1168 		/*
1169 		 * No mapping needed. Just check that it's a valid MC address.
1170 		 */
1171 		sdl = (struct sockaddr_dl *)sa;
1172 		e_addr = LLADDR(sdl);
1173 		if (!ETHER_IS_MULTICAST(e_addr))
1174 			return EADDRNOTAVAIL;
1175 		*llsa = 0;
1176 		return 0;
1177 
1178 #ifdef INET
1179 	case AF_INET:
1180 		sin = (struct sockaddr_in *)sa;
1181 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1182 			return EADDRNOTAVAIL;
1183 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1184 		       M_NOWAIT|M_ZERO);
1185 		if (sdl == NULL)
1186 			return ENOMEM;
1187 		sdl->sdl_len = sizeof *sdl;
1188 		sdl->sdl_family = AF_LINK;
1189 		sdl->sdl_index = ifp->if_index;
1190 		sdl->sdl_type = IFT_ETHER;
1191 		sdl->sdl_alen = ETHER_ADDR_LEN;
1192 		e_addr = LLADDR(sdl);
1193 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1194 		*llsa = (struct sockaddr *)sdl;
1195 		return 0;
1196 #endif
1197 #ifdef INET6
1198 	case AF_INET6:
1199 		sin6 = (struct sockaddr_in6 *)sa;
1200 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1201 			/*
1202 			 * An IP6 address of 0 means listen to all
1203 			 * of the Ethernet multicast address used for IP6.
1204 			 * (This is used for multicast routers.)
1205 			 */
1206 			ifp->if_flags |= IFF_ALLMULTI;
1207 			*llsa = 0;
1208 			return 0;
1209 		}
1210 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1211 			return EADDRNOTAVAIL;
1212 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1213 		       M_NOWAIT|M_ZERO);
1214 		if (sdl == NULL)
1215 			return (ENOMEM);
1216 		sdl->sdl_len = sizeof *sdl;
1217 		sdl->sdl_family = AF_LINK;
1218 		sdl->sdl_index = ifp->if_index;
1219 		sdl->sdl_type = IFT_ETHER;
1220 		sdl->sdl_alen = ETHER_ADDR_LEN;
1221 		e_addr = LLADDR(sdl);
1222 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1223 		*llsa = (struct sockaddr *)sdl;
1224 		return 0;
1225 #endif
1226 
1227 	default:
1228 		/*
1229 		 * Well, the text isn't quite right, but it's the name
1230 		 * that counts...
1231 		 */
1232 		return EAFNOSUPPORT;
1233 	}
1234 }
1235 
1236 static void*
1237 ether_alloc(u_char type, struct ifnet *ifp)
1238 {
1239 	struct arpcom	*ac;
1240 
1241 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1242 	ac->ac_ifp = ifp;
1243 
1244 	return (ac);
1245 }
1246 
1247 static void
1248 ether_free(void *com, u_char type)
1249 {
1250 
1251 	free(com, M_ARPCOM);
1252 }
1253 
1254 static int
1255 ether_modevent(module_t mod, int type, void *data)
1256 {
1257 
1258 	switch (type) {
1259 	case MOD_LOAD:
1260 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1261 		break;
1262 	case MOD_UNLOAD:
1263 		if_deregister_com_alloc(IFT_ETHER);
1264 		break;
1265 	default:
1266 		return EOPNOTSUPP;
1267 	}
1268 
1269 	return (0);
1270 }
1271 
1272 static moduledata_t ether_mod = {
1273 	"ether",
1274 	ether_modevent,
1275 	0
1276 };
1277 
1278 void
1279 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1280 {
1281 	struct ether_vlan_header vlan;
1282 	struct mbuf mv, mb;
1283 
1284 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1285 	    ("%s: vlan information not present", __func__));
1286 	KASSERT(m->m_len >= sizeof(struct ether_header),
1287 	    ("%s: mbuf not large enough for header", __func__));
1288 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1289 	vlan.evl_proto = vlan.evl_encap_proto;
1290 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1291 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1292 	m->m_len -= sizeof(struct ether_header);
1293 	m->m_data += sizeof(struct ether_header);
1294 	/*
1295 	 * If a data link has been supplied by the caller, then we will need to
1296 	 * re-create a stack allocated mbuf chain with the following structure:
1297 	 *
1298 	 * (1) mbuf #1 will contain the supplied data link
1299 	 * (2) mbuf #2 will contain the vlan header
1300 	 * (3) mbuf #3 will contain the original mbuf's packet data
1301 	 *
1302 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1303 	 */
1304 	if (data != NULL) {
1305 		mv.m_next = m;
1306 		mv.m_data = (caddr_t)&vlan;
1307 		mv.m_len = sizeof(vlan);
1308 		mb.m_next = &mv;
1309 		mb.m_data = data;
1310 		mb.m_len = dlen;
1311 		bpf_mtap(bp, &mb);
1312 	} else
1313 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1314 	m->m_len += sizeof(struct ether_header);
1315 	m->m_data -= sizeof(struct ether_header);
1316 }
1317 
1318 struct mbuf *
1319 ether_vlanencap(struct mbuf *m, uint16_t tag)
1320 {
1321 	struct ether_vlan_header *evl;
1322 
1323 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
1324 	if (m == NULL)
1325 		return (NULL);
1326 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1327 
1328 	if (m->m_len < sizeof(*evl)) {
1329 		m = m_pullup(m, sizeof(*evl));
1330 		if (m == NULL)
1331 			return (NULL);
1332 	}
1333 
1334 	/*
1335 	 * Transform the Ethernet header into an Ethernet header
1336 	 * with 802.1Q encapsulation.
1337 	 */
1338 	evl = mtod(m, struct ether_vlan_header *);
1339 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1340 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1341 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1342 	evl->evl_tag = htons(tag);
1343 	return (m);
1344 }
1345 
1346 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1347 MODULE_VERSION(ether, 1);
1348