xref: /freebsd/sys/net/if_ethersubr.c (revision 5861f9665471e98e544f6fa3ce73c4912229ff82)
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 #include <sys/vimage.h>
54 
55 #include <net/if.h>
56 #include <net/if_arp.h>
57 #include <net/netisr.h>
58 #include <net/route.h>
59 #include <net/if_llc.h>
60 #include <net/if_dl.h>
61 #include <net/if_types.h>
62 #include <net/bpf.h>
63 #include <net/ethernet.h>
64 #include <net/if_bridgevar.h>
65 #include <net/if_vlan_var.h>
66 #include <net/if_llatbl.h>
67 #include <net/pf_mtag.h>
68 #include <net/vnet.h>
69 
70 #if defined(INET) || defined(INET6)
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/if_ether.h>
74 #include <netinet/ip_fw.h>
75 #include <netinet/ip_dummynet.h>
76 #include <netinet/vinet.h>
77 #endif
78 #ifdef INET6
79 #include <netinet6/nd6.h>
80 #endif
81 
82 #if defined(INET) || defined(INET6)
83 #ifdef DEV_CARP
84 #include <netinet/ip_carp.h>
85 #endif
86 #endif
87 
88 #ifdef IPX
89 #include <netipx/ipx.h>
90 #include <netipx/ipx_if.h>
91 #endif
92 
93 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
94 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
95 		struct sockaddr *dst, short *tp, int *hlen);
96 
97 #ifdef NETATALK
98 #include <netatalk/at.h>
99 #include <netatalk/at_var.h>
100 #include <netatalk/at_extern.h>
101 
102 #define llc_snap_org_code llc_un.type_snap.org_code
103 #define llc_snap_ether_type llc_un.type_snap.ether_type
104 
105 extern u_char	at_org_code[3];
106 extern u_char	aarp_org_code[3];
107 #endif /* NETATALK */
108 
109 #include <security/mac/mac_framework.h>
110 
111 #ifdef CTASSERT
112 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
113 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
114 #endif
115 
116 /* netgraph node hooks for ng_ether(4) */
117 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
118 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
119 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
120 void	(*ng_ether_attach_p)(struct ifnet *ifp);
121 void	(*ng_ether_detach_p)(struct ifnet *ifp);
122 
123 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
124 
125 /* if_bridge(4) support */
126 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
127 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
128 		struct sockaddr *, struct rtentry *);
129 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
130 
131 /* if_lagg(4) support */
132 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
133 
134 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
135 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
136 
137 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
138 		struct sockaddr *);
139 
140 /* XXX: should be in an arp support file, not here */
141 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
142 
143 #define	ETHER_IS_BROADCAST(addr) \
144 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
145 
146 #define senderr(e) do { error = (e); goto bad;} while (0)
147 
148 #if defined(INET) || defined(INET6)
149 int
150 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared);
151 #ifdef VIMAGE_GLOBALS
152 static int ether_ipfw;
153 #endif
154 #endif
155 
156 
157 /*
158  * Ethernet output routine.
159  * Encapsulate a packet of type family for the local net.
160  * Use trailer local net encapsulation if enough data in first
161  * packet leaves a multiple of 512 bytes of data in remainder.
162  */
163 int
164 ether_output(struct ifnet *ifp, struct mbuf *m,
165 	struct sockaddr *dst, struct route *ro)
166 {
167 	short type;
168 	int error = 0, hdrcmplt = 0;
169 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
170 	struct llentry *lle = NULL;
171 	struct rtentry *rt0 = NULL;
172 	struct ether_header *eh;
173 	struct pf_mtag *t;
174 	int loop_copy = 1;
175 	int hlen;	/* link layer header length */
176 
177 	if (ro != NULL) {
178 		lle = ro->ro_lle;
179 		rt0 = ro->ro_rt;
180 	}
181 #ifdef MAC
182 	error = mac_ifnet_check_transmit(ifp, m);
183 	if (error)
184 		senderr(error);
185 #endif
186 
187 	M_PROFILE(m);
188 	if (ifp->if_flags & IFF_MONITOR)
189 		senderr(ENETDOWN);
190 	if (!((ifp->if_flags & IFF_UP) &&
191 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
192 		senderr(ENETDOWN);
193 
194 	hlen = ETHER_HDR_LEN;
195 	switch (dst->sa_family) {
196 #ifdef INET
197 	case AF_INET:
198 		if (lle != NULL && (lle->la_flags & LLE_VALID))
199 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
200 		else
201 			error = arpresolve(ifp, rt0, m, dst, edst, &lle);
202 		if (error)
203 			return (error == EWOULDBLOCK ? 0 : error);
204 		type = htons(ETHERTYPE_IP);
205 		break;
206 	case AF_ARP:
207 	{
208 		struct arphdr *ah;
209 		ah = mtod(m, struct arphdr *);
210 		ah->ar_hrd = htons(ARPHRD_ETHER);
211 
212 		loop_copy = 0; /* if this is for us, don't do it */
213 
214 		switch(ntohs(ah->ar_op)) {
215 		case ARPOP_REVREQUEST:
216 		case ARPOP_REVREPLY:
217 			type = htons(ETHERTYPE_REVARP);
218 			break;
219 		case ARPOP_REQUEST:
220 		case ARPOP_REPLY:
221 		default:
222 			type = htons(ETHERTYPE_ARP);
223 			break;
224 		}
225 
226 		if (m->m_flags & M_BCAST)
227 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
228 		else
229 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
230 
231 	}
232 	break;
233 #endif
234 #ifdef INET6
235 	case AF_INET6:
236 		if (lle != NULL && (lle->la_flags & LLE_VALID))
237 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
238 		else
239 			error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle);
240 		if (error)
241 			return error;
242 		type = htons(ETHERTYPE_IPV6);
243 		break;
244 #endif
245 #ifdef IPX
246 	case AF_IPX:
247 		if (ef_outputp) {
248 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
249 		    if (error)
250 			goto bad;
251 		} else
252 		    type = htons(ETHERTYPE_IPX);
253 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
254 		    (caddr_t)edst, sizeof (edst));
255 		break;
256 #endif
257 #ifdef NETATALK
258 	case AF_APPLETALK:
259 	  {
260 	    struct at_ifaddr *aa;
261 
262 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
263 		    senderr(EHOSTUNREACH); /* XXX */
264 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) {
265 		    ifa_free(&aa->aa_ifa);
266 		    return (0);
267 	    }
268 	    /*
269 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
270 	     */
271 	    if ( aa->aa_flags & AFA_PHASE2 ) {
272 		struct llc llc;
273 
274 		ifa_free(&aa->aa_ifa);
275 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
276 		if (m == NULL)
277 			senderr(ENOBUFS);
278 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
279 		llc.llc_control = LLC_UI;
280 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
281 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
282 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
283 		type = htons(m->m_pkthdr.len);
284 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
285 	    } else {
286 		ifa_free(&aa->aa_ifa);
287 		type = htons(ETHERTYPE_AT);
288 	    }
289 	    break;
290 	  }
291 #endif /* NETATALK */
292 
293 	case pseudo_AF_HDRCMPLT:
294 		hdrcmplt = 1;
295 		eh = (struct ether_header *)dst->sa_data;
296 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
297 		/* FALLTHROUGH */
298 
299 	case AF_UNSPEC:
300 		loop_copy = 0; /* if this is for us, don't do it */
301 		eh = (struct ether_header *)dst->sa_data;
302 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
303 		type = eh->ether_type;
304 		break;
305 
306 	default:
307 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
308 		senderr(EAFNOSUPPORT);
309 	}
310 
311 	if (lle != NULL && (lle->la_flags & LLE_IFADDR)) {
312 		int csum_flags = 0;
313 		if (m->m_pkthdr.csum_flags & CSUM_IP)
314 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
315 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
316 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
317 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
318 			csum_flags |= CSUM_SCTP_VALID;
319 		m->m_pkthdr.csum_flags |= csum_flags;
320 		m->m_pkthdr.csum_data = 0xffff;
321 		return (if_simloop(ifp, m, dst->sa_family, 0));
322 	}
323 
324 	/*
325 	 * Add local net header.  If no space in first mbuf,
326 	 * allocate another.
327 	 */
328 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
329 	if (m == NULL)
330 		senderr(ENOBUFS);
331 	eh = mtod(m, struct ether_header *);
332 	(void)memcpy(&eh->ether_type, &type,
333 		sizeof(eh->ether_type));
334 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
335 	if (hdrcmplt)
336 		(void)memcpy(eh->ether_shost, esrc,
337 			sizeof(eh->ether_shost));
338 	else
339 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
340 			sizeof(eh->ether_shost));
341 
342 	/*
343 	 * If a simplex interface, and the packet is being sent to our
344 	 * Ethernet address or a broadcast address, loopback a copy.
345 	 * XXX To make a simplex device behave exactly like a duplex
346 	 * device, we should copy in the case of sending to our own
347 	 * ethernet address (thus letting the original actually appear
348 	 * on the wire). However, we don't do that here for security
349 	 * reasons and compatibility with the original behavior.
350 	 */
351 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
352 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
353 		int csum_flags = 0;
354 
355 		if (m->m_pkthdr.csum_flags & CSUM_IP)
356 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
357 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
358 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
359 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
360 			csum_flags |= CSUM_SCTP_VALID;
361 
362 		if (m->m_flags & M_BCAST) {
363 			struct mbuf *n;
364 
365 			/*
366 			 * Because if_simloop() modifies the packet, we need a
367 			 * writable copy through m_dup() instead of a readonly
368 			 * one as m_copy[m] would give us. The alternative would
369 			 * be to modify if_simloop() to handle the readonly mbuf,
370 			 * but performancewise it is mostly equivalent (trading
371 			 * extra data copying vs. extra locking).
372 			 *
373 			 * XXX This is a local workaround.  A number of less
374 			 * often used kernel parts suffer from the same bug.
375 			 * See PR kern/105943 for a proposed general solution.
376 			 */
377 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
378 				n->m_pkthdr.csum_flags |= csum_flags;
379 				if (csum_flags & CSUM_DATA_VALID)
380 					n->m_pkthdr.csum_data = 0xffff;
381 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
382 			} else
383 				ifp->if_iqdrops++;
384 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
385 				ETHER_ADDR_LEN) == 0) {
386 			m->m_pkthdr.csum_flags |= csum_flags;
387 			if (csum_flags & CSUM_DATA_VALID)
388 				m->m_pkthdr.csum_data = 0xffff;
389 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
390 			return (0);	/* XXX */
391 		}
392 	}
393 
394        /*
395 	* Bridges require special output handling.
396 	*/
397 	if (ifp->if_bridge) {
398 		BRIDGE_OUTPUT(ifp, m, error);
399 		return (error);
400 	}
401 
402 #if defined(INET) || defined(INET6)
403 #ifdef DEV_CARP
404 	if (ifp->if_carp &&
405 	    (error = carp_output(ifp, m, dst, NULL)))
406 		goto bad;
407 #endif
408 #endif
409 
410 	/* Handle ng_ether(4) processing, if any */
411 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
412 		KASSERT(ng_ether_output_p != NULL,
413 		    ("ng_ether_output_p is NULL"));
414 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
415 bad:			if (m != NULL)
416 				m_freem(m);
417 			return (error);
418 		}
419 		if (m == NULL)
420 			return (0);
421 	}
422 
423 	/* Continue with link-layer output */
424 	return ether_output_frame(ifp, m);
425 }
426 
427 /*
428  * Ethernet link layer output routine to send a raw frame to the device.
429  *
430  * This assumes that the 14 byte Ethernet header is present and contiguous
431  * in the first mbuf (if BRIDGE'ing).
432  */
433 int
434 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
435 {
436 #if defined(INET) || defined(INET6)
437 	INIT_VNET_NET(ifp->if_vnet);
438 
439 	if (ip_fw_chk_ptr && V_ether_ipfw != 0) {
440 		if (ether_ipfw_chk(&m, ifp, 0) == 0) {
441 			if (m) {
442 				m_freem(m);
443 				return EACCES;	/* pkt dropped */
444 			} else
445 				return 0;	/* consumed e.g. in a pipe */
446 		}
447 	}
448 #endif
449 
450 	/*
451 	 * Queue message on interface, update output statistics if
452 	 * successful, and start output if interface not yet active.
453 	 */
454 	return ((ifp->if_transmit)(ifp, m));
455 }
456 
457 #if defined(INET) || defined(INET6)
458 /*
459  * ipfw processing for ethernet packets (in and out).
460  * The second parameter is NULL from ether_demux, and ifp from
461  * ether_output_frame.
462  */
463 int
464 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared)
465 {
466 	INIT_VNET_INET(dst->if_vnet);
467 	struct ether_header *eh;
468 	struct ether_header save_eh;
469 	struct mbuf *m;
470 	int i;
471 	struct ip_fw_args args;
472 	struct dn_pkt_tag *dn_tag;
473 
474 	dn_tag = ip_dn_claim_tag(*m0);
475 
476 	if (dn_tag != NULL) {
477 		if (dn_tag->rule != NULL && V_fw_one_pass)
478 			/* dummynet packet, already partially processed */
479 			return (1);
480 		args.rule = dn_tag->rule;	/* matching rule to restart */
481 		args.rule_id = dn_tag->rule_id;
482 		args.chain_id = dn_tag->chain_id;
483 	} else
484 		args.rule = NULL;
485 
486 	/*
487 	 * I need some amt of data to be contiguous, and in case others need
488 	 * the packet (shared==1) also better be in the first mbuf.
489 	 */
490 	m = *m0;
491 	i = min( m->m_pkthdr.len, max_protohdr);
492 	if ( shared || m->m_len < i) {
493 		m = m_pullup(m, i);
494 		if (m == NULL) {
495 			*m0 = m;
496 			return 0;
497 		}
498 	}
499 	eh = mtod(m, struct ether_header *);
500 	save_eh = *eh;			/* save copy for restore below */
501 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
502 
503 	args.m = m;		/* the packet we are looking at		*/
504 	args.oif = dst;		/* destination, if any			*/
505 	args.next_hop = NULL;	/* we do not support forward yet	*/
506 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
507 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
508 	i = ip_fw_chk_ptr(&args);
509 	m = args.m;
510 	if (m != NULL) {
511 		/*
512 		 * Restore Ethernet header, as needed, in case the
513 		 * mbuf chain was replaced by ipfw.
514 		 */
515 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
516 		if (m == NULL) {
517 			*m0 = m;
518 			return 0;
519 		}
520 		if (eh != mtod(m, struct ether_header *))
521 			bcopy(&save_eh, mtod(m, struct ether_header *),
522 				ETHER_HDR_LEN);
523 	}
524 	*m0 = m;
525 
526 	if (i == IP_FW_DENY) /* drop */
527 		return 0;
528 
529 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
530 
531 	if (i == IP_FW_PASS) /* a PASS rule.  */
532 		return 1;
533 
534 	if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
535 		/*
536 		 * Pass the pkt to dummynet, which consumes it.
537 		 * If shared, make a copy and keep the original.
538 		 */
539 		if (shared) {
540 			m = m_copypacket(m, M_DONTWAIT);
541 			if (m == NULL)
542 				return 0;
543 		} else {
544 			/*
545 			 * Pass the original to dummynet and
546 			 * nothing back to the caller
547 			 */
548 			*m0 = NULL ;
549 		}
550 		ip_dn_io_ptr(&m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
551 		return 0;
552 	}
553 	/*
554 	 * XXX at some point add support for divert/forward actions.
555 	 * If none of the above matches, we have to drop the pkt.
556 	 */
557 	return 0;
558 }
559 #endif
560 
561 /*
562  * Process a received Ethernet packet; the packet is in the
563  * mbuf chain m with the ethernet header at the front.
564  */
565 static void
566 ether_input(struct ifnet *ifp, struct mbuf *m)
567 {
568 	struct ether_header *eh;
569 	u_short etype;
570 
571 	if ((ifp->if_flags & IFF_UP) == 0) {
572 		m_freem(m);
573 		return;
574 	}
575 #ifdef DIAGNOSTIC
576 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
577 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
578 		m_freem(m);
579 		return;
580 	}
581 #endif
582 	/*
583 	 * Do consistency checks to verify assumptions
584 	 * made by code past this point.
585 	 */
586 	if ((m->m_flags & M_PKTHDR) == 0) {
587 		if_printf(ifp, "discard frame w/o packet header\n");
588 		ifp->if_ierrors++;
589 		m_freem(m);
590 		return;
591 	}
592 	if (m->m_len < ETHER_HDR_LEN) {
593 		/* XXX maybe should pullup? */
594 		if_printf(ifp, "discard frame w/o leading ethernet "
595 				"header (len %u pkt len %u)\n",
596 				m->m_len, m->m_pkthdr.len);
597 		ifp->if_ierrors++;
598 		m_freem(m);
599 		return;
600 	}
601 	eh = mtod(m, struct ether_header *);
602 	etype = ntohs(eh->ether_type);
603 	if (m->m_pkthdr.rcvif == NULL) {
604 		if_printf(ifp, "discard frame w/o interface pointer\n");
605 		ifp->if_ierrors++;
606 		m_freem(m);
607 		return;
608 	}
609 #ifdef DIAGNOSTIC
610 	if (m->m_pkthdr.rcvif != ifp) {
611 		if_printf(ifp, "Warning, frame marked as received on %s\n",
612 			m->m_pkthdr.rcvif->if_xname);
613 	}
614 #endif
615 
616 	CURVNET_SET_QUIET(ifp->if_vnet);
617 
618 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
619 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
620 			m->m_flags |= M_BCAST;
621 		else
622 			m->m_flags |= M_MCAST;
623 		ifp->if_imcasts++;
624 	}
625 
626 #ifdef MAC
627 	/*
628 	 * Tag the mbuf with an appropriate MAC label before any other
629 	 * consumers can get to it.
630 	 */
631 	mac_ifnet_create_mbuf(ifp, m);
632 #endif
633 
634 	/*
635 	 * Give bpf a chance at the packet.
636 	 */
637 	ETHER_BPF_MTAP(ifp, m);
638 
639 	/*
640 	 * If the CRC is still on the packet, trim it off. We do this once
641 	 * and once only in case we are re-entered. Nothing else on the
642 	 * Ethernet receive path expects to see the FCS.
643 	 */
644 	if (m->m_flags & M_HASFCS) {
645 		m_adj(m, -ETHER_CRC_LEN);
646 		m->m_flags &= ~M_HASFCS;
647 	}
648 
649 	ifp->if_ibytes += m->m_pkthdr.len;
650 
651 	/* Allow monitor mode to claim this frame, after stats are updated. */
652 	if (ifp->if_flags & IFF_MONITOR) {
653 		m_freem(m);
654 		CURVNET_RESTORE();
655 		return;
656 	}
657 
658 	/* Handle input from a lagg(4) port */
659 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
660 		KASSERT(lagg_input_p != NULL,
661 		    ("%s: if_lagg not loaded!", __func__));
662 		m = (*lagg_input_p)(ifp, m);
663 		if (m != NULL)
664 			ifp = m->m_pkthdr.rcvif;
665 		else
666 			return;
667 	}
668 
669 	/*
670 	 * If the hardware did not process an 802.1Q tag, do this now,
671 	 * to allow 802.1P priority frames to be passed to the main input
672 	 * path correctly.
673 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
674 	 */
675 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
676 		struct ether_vlan_header *evl;
677 
678 		if (m->m_len < sizeof(*evl) &&
679 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
680 #ifdef DIAGNOSTIC
681 			if_printf(ifp, "cannot pullup VLAN header\n");
682 #endif
683 			ifp->if_ierrors++;
684 			m_freem(m);
685 			return;
686 		}
687 
688 		evl = mtod(m, struct ether_vlan_header *);
689 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
690 		m->m_flags |= M_VLANTAG;
691 
692 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
693 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
694 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
695 	}
696 
697 	/* Allow ng_ether(4) to claim this frame. */
698 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
699 		KASSERT(ng_ether_input_p != NULL,
700 		    ("%s: ng_ether_input_p is NULL", __func__));
701 		m->m_flags &= ~M_PROMISC;
702 		(*ng_ether_input_p)(ifp, &m);
703 		if (m == NULL) {
704 			CURVNET_RESTORE();
705 			return;
706 		}
707 	}
708 
709 	/*
710 	 * Allow if_bridge(4) to claim this frame.
711 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
712 	 * and the frame should be delivered locally.
713 	 */
714 	if (ifp->if_bridge != NULL) {
715 		m->m_flags &= ~M_PROMISC;
716 		BRIDGE_INPUT(ifp, m);
717 		if (m == NULL) {
718 			CURVNET_RESTORE();
719 			return;
720 		}
721 	}
722 
723 #if defined(INET) || defined(INET6)
724 #ifdef DEV_CARP
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(ifp->if_carp, eh->ether_dhost)) {
736 		m->m_flags &= ~M_PROMISC;
737 	} else
738 #endif
739 #endif
740 	{
741 		/*
742 		 * If the frame received was not for our MAC address, set the
743 		 * M_PROMISC flag on the mbuf chain. The frame may need to
744 		 * be seen by the rest of the Ethernet input path in case of
745 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
746 		 * seen by upper protocol layers.
747 		 */
748 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
749 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
750 			m->m_flags |= M_PROMISC;
751 	}
752 
753 	/* First chunk of an mbuf contains good entropy */
754 	if (harvest.ethernet)
755 		random_harvest(m, 16, 3, 0, RANDOM_NET);
756 
757 	ether_demux(ifp, m);
758 	CURVNET_RESTORE();
759 }
760 
761 /*
762  * Upper layer processing for a received Ethernet packet.
763  */
764 void
765 ether_demux(struct ifnet *ifp, struct mbuf *m)
766 {
767 	struct ether_header *eh;
768 	int isr;
769 	u_short ether_type;
770 #if defined(NETATALK)
771 	struct llc *l;
772 #endif
773 
774 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
775 
776 #if defined(INET) || defined(INET6)
777 	INIT_VNET_NET(ifp->if_vnet);
778 	/*
779 	 * Allow dummynet and/or ipfw to claim the frame.
780 	 * Do not do this for PROMISC frames in case we are re-entered.
781 	 */
782 	if (ip_fw_chk_ptr && V_ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
783 		if (ether_ipfw_chk(&m, NULL, 0) == 0) {
784 			if (m)
785 				m_freem(m);	/* dropped; free mbuf chain */
786 			return;			/* consumed */
787 		}
788 	}
789 #endif
790 	eh = mtod(m, struct ether_header *);
791 	ether_type = ntohs(eh->ether_type);
792 
793 	/*
794 	 * If this frame has a VLAN tag other than 0, call vlan_input()
795 	 * if its module is loaded. Otherwise, drop.
796 	 */
797 	if ((m->m_flags & M_VLANTAG) &&
798 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
799 		if (ifp->if_vlantrunk == NULL) {
800 			ifp->if_noproto++;
801 			m_freem(m);
802 			return;
803 		}
804 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
805 		    __func__));
806 		/* Clear before possibly re-entering ether_input(). */
807 		m->m_flags &= ~M_PROMISC;
808 		(*vlan_input_p)(ifp, m);
809 		return;
810 	}
811 
812 	/*
813 	 * Pass promiscuously received frames to the upper layer if the user
814 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
815 	 */
816 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
817 		m_freem(m);
818 		return;
819 	}
820 
821 	/*
822 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
823 	 * Strip off Ethernet header.
824 	 */
825 	m->m_flags &= ~M_VLANTAG;
826 	m->m_flags &= ~(M_PROTOFLAGS);
827 	m_adj(m, ETHER_HDR_LEN);
828 
829 	/*
830 	 * Dispatch frame to upper layer.
831 	 */
832 	switch (ether_type) {
833 #ifdef INET
834 	case ETHERTYPE_IP:
835 		if ((m = ip_fastforward(m)) == NULL)
836 			return;
837 		isr = NETISR_IP;
838 		break;
839 
840 	case ETHERTYPE_ARP:
841 		if (ifp->if_flags & IFF_NOARP) {
842 			/* Discard packet if ARP is disabled on interface */
843 			m_freem(m);
844 			return;
845 		}
846 		isr = NETISR_ARP;
847 		break;
848 #endif
849 #ifdef IPX
850 	case ETHERTYPE_IPX:
851 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
852 			return;
853 		isr = NETISR_IPX;
854 		break;
855 #endif
856 #ifdef INET6
857 	case ETHERTYPE_IPV6:
858 		isr = NETISR_IPV6;
859 		break;
860 #endif
861 #ifdef NETATALK
862 	case ETHERTYPE_AT:
863 		isr = NETISR_ATALK1;
864 		break;
865 	case ETHERTYPE_AARP:
866 		isr = NETISR_AARP;
867 		break;
868 #endif /* NETATALK */
869 	default:
870 #ifdef IPX
871 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
872 			return;
873 #endif /* IPX */
874 #if defined(NETATALK)
875 		if (ether_type > ETHERMTU)
876 			goto discard;
877 		l = mtod(m, struct llc *);
878 		if (l->llc_dsap == LLC_SNAP_LSAP &&
879 		    l->llc_ssap == LLC_SNAP_LSAP &&
880 		    l->llc_control == LLC_UI) {
881 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
882 			    sizeof(at_org_code)) == 0 &&
883 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
884 				m_adj(m, LLC_SNAPFRAMELEN);
885 				isr = NETISR_ATALK2;
886 				break;
887 			}
888 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
889 			    sizeof(aarp_org_code)) == 0 &&
890 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
891 				m_adj(m, LLC_SNAPFRAMELEN);
892 				isr = NETISR_AARP;
893 				break;
894 			}
895 		}
896 #endif /* NETATALK */
897 		goto discard;
898 	}
899 	netisr_dispatch(isr, m);
900 	return;
901 
902 discard:
903 	/*
904 	 * Packet is to be discarded.  If netgraph is present,
905 	 * hand the packet to it for last chance processing;
906 	 * otherwise dispose of it.
907 	 */
908 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
909 		KASSERT(ng_ether_input_orphan_p != NULL,
910 		    ("ng_ether_input_orphan_p is NULL"));
911 		/*
912 		 * Put back the ethernet header so netgraph has a
913 		 * consistent view of inbound packets.
914 		 */
915 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
916 		(*ng_ether_input_orphan_p)(ifp, m);
917 		return;
918 	}
919 	m_freem(m);
920 }
921 
922 /*
923  * Convert Ethernet address to printable (loggable) representation.
924  * This routine is for compatibility; it's better to just use
925  *
926  *	printf("%6D", <pointer to address>, ":");
927  *
928  * since there's no static buffer involved.
929  */
930 char *
931 ether_sprintf(const u_char *ap)
932 {
933 	static char etherbuf[18];
934 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
935 	return (etherbuf);
936 }
937 
938 /*
939  * Perform common duties while attaching to interface list
940  */
941 void
942 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
943 {
944 	int i;
945 	struct ifaddr *ifa;
946 	struct sockaddr_dl *sdl;
947 
948 	ifp->if_addrlen = ETHER_ADDR_LEN;
949 	ifp->if_hdrlen = ETHER_HDR_LEN;
950 	if_attach(ifp);
951 	ifp->if_mtu = ETHERMTU;
952 	ifp->if_output = ether_output;
953 	ifp->if_input = ether_input;
954 	ifp->if_resolvemulti = ether_resolvemulti;
955 	if (ifp->if_baudrate == 0)
956 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
957 	ifp->if_broadcastaddr = etherbroadcastaddr;
958 
959 	ifa = ifp->if_addr;
960 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
961 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
962 	sdl->sdl_type = IFT_ETHER;
963 	sdl->sdl_alen = ifp->if_addrlen;
964 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
965 
966 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
967 	if (ng_ether_attach_p != NULL)
968 		(*ng_ether_attach_p)(ifp);
969 
970 	/* Announce Ethernet MAC address if non-zero. */
971 	for (i = 0; i < ifp->if_addrlen; i++)
972 		if (lla[i] != 0)
973 			break;
974 	if (i != ifp->if_addrlen)
975 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
976 }
977 
978 /*
979  * Perform common duties while detaching an Ethernet interface
980  */
981 void
982 ether_ifdetach(struct ifnet *ifp)
983 {
984 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
985 		KASSERT(ng_ether_detach_p != NULL,
986 		    ("ng_ether_detach_p is NULL"));
987 		(*ng_ether_detach_p)(ifp);
988 	}
989 
990 	bpfdetach(ifp);
991 	if_detach(ifp);
992 }
993 
994 SYSCTL_DECL(_net_link);
995 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
996 #if defined(INET) || defined(INET6)
997 SYSCTL_V_INT(V_NET, vnet_net, _net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
998 	     ether_ipfw, 0, "Pass ether pkts through firewall");
999 #endif
1000 
1001 #if 0
1002 /*
1003  * This is for reference.  We have a table-driven version
1004  * of the little-endian crc32 generator, which is faster
1005  * than the double-loop.
1006  */
1007 uint32_t
1008 ether_crc32_le(const uint8_t *buf, size_t len)
1009 {
1010 	size_t i;
1011 	uint32_t crc;
1012 	int bit;
1013 	uint8_t data;
1014 
1015 	crc = 0xffffffff;	/* initial value */
1016 
1017 	for (i = 0; i < len; i++) {
1018 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1019 			carry = (crc ^ data) & 1;
1020 			crc >>= 1;
1021 			if (carry)
1022 				crc = (crc ^ ETHER_CRC_POLY_LE);
1023 		}
1024 	}
1025 
1026 	return (crc);
1027 }
1028 #else
1029 uint32_t
1030 ether_crc32_le(const uint8_t *buf, size_t len)
1031 {
1032 	static const uint32_t crctab[] = {
1033 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1034 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1035 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1036 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1037 	};
1038 	size_t i;
1039 	uint32_t crc;
1040 
1041 	crc = 0xffffffff;	/* initial value */
1042 
1043 	for (i = 0; i < len; i++) {
1044 		crc ^= buf[i];
1045 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1046 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1047 	}
1048 
1049 	return (crc);
1050 }
1051 #endif
1052 
1053 uint32_t
1054 ether_crc32_be(const uint8_t *buf, size_t len)
1055 {
1056 	size_t i;
1057 	uint32_t crc, carry;
1058 	int bit;
1059 	uint8_t data;
1060 
1061 	crc = 0xffffffff;	/* initial value */
1062 
1063 	for (i = 0; i < len; i++) {
1064 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1065 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1066 			crc <<= 1;
1067 			if (carry)
1068 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1069 		}
1070 	}
1071 
1072 	return (crc);
1073 }
1074 
1075 int
1076 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1077 {
1078 	struct ifaddr *ifa = (struct ifaddr *) data;
1079 	struct ifreq *ifr = (struct ifreq *) data;
1080 	int error = 0;
1081 
1082 	switch (command) {
1083 	case SIOCSIFADDR:
1084 		ifp->if_flags |= IFF_UP;
1085 
1086 		switch (ifa->ifa_addr->sa_family) {
1087 #ifdef INET
1088 		case AF_INET:
1089 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1090 			arp_ifinit(ifp, ifa);
1091 			break;
1092 #endif
1093 #ifdef IPX
1094 		/*
1095 		 * XXX - This code is probably wrong
1096 		 */
1097 		case AF_IPX:
1098 			{
1099 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1100 
1101 			if (ipx_nullhost(*ina))
1102 				ina->x_host =
1103 				    *(union ipx_host *)
1104 				    IF_LLADDR(ifp);
1105 			else {
1106 				bcopy((caddr_t) ina->x_host.c_host,
1107 				      (caddr_t) IF_LLADDR(ifp),
1108 				      ETHER_ADDR_LEN);
1109 			}
1110 
1111 			/*
1112 			 * Set new address
1113 			 */
1114 			ifp->if_init(ifp->if_softc);
1115 			break;
1116 			}
1117 #endif
1118 		default:
1119 			ifp->if_init(ifp->if_softc);
1120 			break;
1121 		}
1122 		break;
1123 
1124 	case SIOCGIFADDR:
1125 		{
1126 			struct sockaddr *sa;
1127 
1128 			sa = (struct sockaddr *) & ifr->ifr_data;
1129 			bcopy(IF_LLADDR(ifp),
1130 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1131 		}
1132 		break;
1133 
1134 	case SIOCSIFMTU:
1135 		/*
1136 		 * Set the interface MTU.
1137 		 */
1138 		if (ifr->ifr_mtu > ETHERMTU) {
1139 			error = EINVAL;
1140 		} else {
1141 			ifp->if_mtu = ifr->ifr_mtu;
1142 		}
1143 		break;
1144 	default:
1145 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1146 		break;
1147 	}
1148 	return (error);
1149 }
1150 
1151 static int
1152 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1153 	struct sockaddr *sa)
1154 {
1155 	struct sockaddr_dl *sdl;
1156 #ifdef INET
1157 	struct sockaddr_in *sin;
1158 #endif
1159 #ifdef INET6
1160 	struct sockaddr_in6 *sin6;
1161 #endif
1162 	u_char *e_addr;
1163 
1164 	switch(sa->sa_family) {
1165 	case AF_LINK:
1166 		/*
1167 		 * No mapping needed. Just check that it's a valid MC address.
1168 		 */
1169 		sdl = (struct sockaddr_dl *)sa;
1170 		e_addr = LLADDR(sdl);
1171 		if (!ETHER_IS_MULTICAST(e_addr))
1172 			return EADDRNOTAVAIL;
1173 		*llsa = 0;
1174 		return 0;
1175 
1176 #ifdef INET
1177 	case AF_INET:
1178 		sin = (struct sockaddr_in *)sa;
1179 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1180 			return EADDRNOTAVAIL;
1181 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1182 		       M_NOWAIT|M_ZERO);
1183 		if (sdl == NULL)
1184 			return ENOMEM;
1185 		sdl->sdl_len = sizeof *sdl;
1186 		sdl->sdl_family = AF_LINK;
1187 		sdl->sdl_index = ifp->if_index;
1188 		sdl->sdl_type = IFT_ETHER;
1189 		sdl->sdl_alen = ETHER_ADDR_LEN;
1190 		e_addr = LLADDR(sdl);
1191 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1192 		*llsa = (struct sockaddr *)sdl;
1193 		return 0;
1194 #endif
1195 #ifdef INET6
1196 	case AF_INET6:
1197 		sin6 = (struct sockaddr_in6 *)sa;
1198 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1199 			/*
1200 			 * An IP6 address of 0 means listen to all
1201 			 * of the Ethernet multicast address used for IP6.
1202 			 * (This is used for multicast routers.)
1203 			 */
1204 			ifp->if_flags |= IFF_ALLMULTI;
1205 			*llsa = 0;
1206 			return 0;
1207 		}
1208 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1209 			return EADDRNOTAVAIL;
1210 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1211 		       M_NOWAIT|M_ZERO);
1212 		if (sdl == NULL)
1213 			return (ENOMEM);
1214 		sdl->sdl_len = sizeof *sdl;
1215 		sdl->sdl_family = AF_LINK;
1216 		sdl->sdl_index = ifp->if_index;
1217 		sdl->sdl_type = IFT_ETHER;
1218 		sdl->sdl_alen = ETHER_ADDR_LEN;
1219 		e_addr = LLADDR(sdl);
1220 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1221 		*llsa = (struct sockaddr *)sdl;
1222 		return 0;
1223 #endif
1224 
1225 	default:
1226 		/*
1227 		 * Well, the text isn't quite right, but it's the name
1228 		 * that counts...
1229 		 */
1230 		return EAFNOSUPPORT;
1231 	}
1232 }
1233 
1234 static void*
1235 ether_alloc(u_char type, struct ifnet *ifp)
1236 {
1237 	struct arpcom	*ac;
1238 
1239 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1240 	ac->ac_ifp = ifp;
1241 
1242 	return (ac);
1243 }
1244 
1245 static void
1246 ether_free(void *com, u_char type)
1247 {
1248 
1249 	free(com, M_ARPCOM);
1250 }
1251 
1252 static int
1253 ether_modevent(module_t mod, int type, void *data)
1254 {
1255 
1256 	switch (type) {
1257 	case MOD_LOAD:
1258 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1259 		break;
1260 	case MOD_UNLOAD:
1261 		if_deregister_com_alloc(IFT_ETHER);
1262 		break;
1263 	default:
1264 		return EOPNOTSUPP;
1265 	}
1266 
1267 	return (0);
1268 }
1269 
1270 static moduledata_t ether_mod = {
1271 	"ether",
1272 	ether_modevent,
1273 	0
1274 };
1275 
1276 void
1277 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1278 {
1279 	struct ether_vlan_header vlan;
1280 	struct mbuf mv, mb;
1281 
1282 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1283 	    ("%s: vlan information not present", __func__));
1284 	KASSERT(m->m_len >= sizeof(struct ether_header),
1285 	    ("%s: mbuf not large enough for header", __func__));
1286 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1287 	vlan.evl_proto = vlan.evl_encap_proto;
1288 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1289 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1290 	m->m_len -= sizeof(struct ether_header);
1291 	m->m_data += sizeof(struct ether_header);
1292 	/*
1293 	 * If a data link has been supplied by the caller, then we will need to
1294 	 * re-create a stack allocated mbuf chain with the following structure:
1295 	 *
1296 	 * (1) mbuf #1 will contain the supplied data link
1297 	 * (2) mbuf #2 will contain the vlan header
1298 	 * (3) mbuf #3 will contain the original mbuf's packet data
1299 	 *
1300 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1301 	 */
1302 	if (data != NULL) {
1303 		mv.m_next = m;
1304 		mv.m_data = (caddr_t)&vlan;
1305 		mv.m_len = sizeof(vlan);
1306 		mb.m_next = &mv;
1307 		mb.m_data = data;
1308 		mb.m_len = dlen;
1309 		bpf_mtap(bp, &mb);
1310 	} else
1311 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1312 	m->m_len += sizeof(struct ether_header);
1313 	m->m_data -= sizeof(struct ether_header);
1314 }
1315 
1316 struct mbuf *
1317 ether_vlanencap(struct mbuf *m, uint16_t tag)
1318 {
1319 	struct ether_vlan_header *evl;
1320 
1321 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
1322 	if (m == NULL)
1323 		return (NULL);
1324 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1325 
1326 	if (m->m_len < sizeof(*evl)) {
1327 		m = m_pullup(m, sizeof(*evl));
1328 		if (m == NULL)
1329 			return (NULL);
1330 	}
1331 
1332 	/*
1333 	 * Transform the Ethernet header into an Ethernet header
1334 	 * with 802.1Q encapsulation.
1335 	 */
1336 	evl = mtod(m, struct ether_vlan_header *);
1337 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1338 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1339 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1340 	evl->evl_tag = htons(tag);
1341 	return (m);
1342 }
1343 
1344 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1345 MODULE_VERSION(ether, 1);
1346