xref: /freebsd/sys/net/if_ethersubr.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
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 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.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
503 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
504 	i = V_ip_fw_chk_ptr(&args);
505 	m = args.m;
506 	if (m != NULL) {
507 		/*
508 		 * Restore Ethernet header, as needed, in case the
509 		 * mbuf chain was replaced by ipfw.
510 		 */
511 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
512 		if (m == NULL) {
513 			*m0 = m;
514 			return 0;
515 		}
516 		if (eh != mtod(m, struct ether_header *))
517 			bcopy(&save_eh, mtod(m, struct ether_header *),
518 				ETHER_HDR_LEN);
519 	}
520 	*m0 = m;
521 
522 	if (i == IP_FW_DENY) /* drop */
523 		return 0;
524 
525 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
526 
527 	if (i == IP_FW_PASS) /* a PASS rule.  */
528 		return 1;
529 
530 	if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
531 		int dir;
532 		/*
533 		 * Pass the pkt to dummynet, which consumes it.
534 		 * If shared, make a copy and keep the original.
535 		 */
536 		if (shared) {
537 			m = m_copypacket(m, M_DONTWAIT);
538 			if (m == NULL)
539 				return 0;
540 		} else {
541 			/*
542 			 * Pass the original to dummynet and
543 			 * nothing back to the caller
544 			 */
545 			*m0 = NULL ;
546 		}
547 		dir = PROTO_LAYER2 | (dst ? DIR_OUT : DIR_IN);
548 		ip_dn_io_ptr(&m, dir, &args);
549 		return 0;
550 	}
551 	/*
552 	 * XXX at some point add support for divert/forward actions.
553 	 * If none of the above matches, we have to drop the pkt.
554 	 */
555 	return 0;
556 }
557 #endif
558 
559 /*
560  * Process a received Ethernet packet; the packet is in the
561  * mbuf chain m with the ethernet header at the front.
562  */
563 static void
564 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
565 {
566 	struct ether_header *eh;
567 	u_short etype;
568 
569 	if ((ifp->if_flags & IFF_UP) == 0) {
570 		m_freem(m);
571 		return;
572 	}
573 #ifdef DIAGNOSTIC
574 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
575 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
576 		m_freem(m);
577 		return;
578 	}
579 #endif
580 	/*
581 	 * Do consistency checks to verify assumptions
582 	 * made by code past this point.
583 	 */
584 	if ((m->m_flags & M_PKTHDR) == 0) {
585 		if_printf(ifp, "discard frame w/o packet header\n");
586 		ifp->if_ierrors++;
587 		m_freem(m);
588 		return;
589 	}
590 	if (m->m_len < ETHER_HDR_LEN) {
591 		/* XXX maybe should pullup? */
592 		if_printf(ifp, "discard frame w/o leading ethernet "
593 				"header (len %u pkt len %u)\n",
594 				m->m_len, m->m_pkthdr.len);
595 		ifp->if_ierrors++;
596 		m_freem(m);
597 		return;
598 	}
599 	eh = mtod(m, struct ether_header *);
600 	etype = ntohs(eh->ether_type);
601 	if (m->m_pkthdr.rcvif == NULL) {
602 		if_printf(ifp, "discard frame w/o interface pointer\n");
603 		ifp->if_ierrors++;
604 		m_freem(m);
605 		return;
606 	}
607 #ifdef DIAGNOSTIC
608 	if (m->m_pkthdr.rcvif != ifp) {
609 		if_printf(ifp, "Warning, frame marked as received on %s\n",
610 			m->m_pkthdr.rcvif->if_xname);
611 	}
612 #endif
613 
614 	CURVNET_SET_QUIET(ifp->if_vnet);
615 
616 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
617 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
618 			m->m_flags |= M_BCAST;
619 		else
620 			m->m_flags |= M_MCAST;
621 		ifp->if_imcasts++;
622 	}
623 
624 #ifdef MAC
625 	/*
626 	 * Tag the mbuf with an appropriate MAC label before any other
627 	 * consumers can get to it.
628 	 */
629 	mac_ifnet_create_mbuf(ifp, m);
630 #endif
631 
632 	/*
633 	 * Give bpf a chance at the packet.
634 	 */
635 	ETHER_BPF_MTAP(ifp, m);
636 
637 	/*
638 	 * If the CRC is still on the packet, trim it off. We do this once
639 	 * and once only in case we are re-entered. Nothing else on the
640 	 * Ethernet receive path expects to see the FCS.
641 	 */
642 	if (m->m_flags & M_HASFCS) {
643 		m_adj(m, -ETHER_CRC_LEN);
644 		m->m_flags &= ~M_HASFCS;
645 	}
646 
647 	ifp->if_ibytes += m->m_pkthdr.len;
648 
649 	/* Allow monitor mode to claim this frame, after stats are updated. */
650 	if (ifp->if_flags & IFF_MONITOR) {
651 		m_freem(m);
652 		CURVNET_RESTORE();
653 		return;
654 	}
655 
656 	/* Handle input from a lagg(4) port */
657 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
658 		KASSERT(lagg_input_p != NULL,
659 		    ("%s: if_lagg not loaded!", __func__));
660 		m = (*lagg_input_p)(ifp, m);
661 		if (m != NULL)
662 			ifp = m->m_pkthdr.rcvif;
663 		else
664 			return;
665 	}
666 
667 	/*
668 	 * If the hardware did not process an 802.1Q tag, do this now,
669 	 * to allow 802.1P priority frames to be passed to the main input
670 	 * path correctly.
671 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
672 	 */
673 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
674 		struct ether_vlan_header *evl;
675 
676 		if (m->m_len < sizeof(*evl) &&
677 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
678 #ifdef DIAGNOSTIC
679 			if_printf(ifp, "cannot pullup VLAN header\n");
680 #endif
681 			ifp->if_ierrors++;
682 			m_freem(m);
683 			return;
684 		}
685 
686 		evl = mtod(m, struct ether_vlan_header *);
687 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
688 		m->m_flags |= M_VLANTAG;
689 
690 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
691 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
692 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
693 	}
694 
695 	/* Allow ng_ether(4) to claim this frame. */
696 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
697 		KASSERT(ng_ether_input_p != NULL,
698 		    ("%s: ng_ether_input_p is NULL", __func__));
699 		m->m_flags &= ~M_PROMISC;
700 		(*ng_ether_input_p)(ifp, &m);
701 		if (m == NULL) {
702 			CURVNET_RESTORE();
703 			return;
704 		}
705 	}
706 
707 	/*
708 	 * Allow if_bridge(4) to claim this frame.
709 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
710 	 * and the frame should be delivered locally.
711 	 */
712 	if (ifp->if_bridge != NULL) {
713 		m->m_flags &= ~M_PROMISC;
714 		BRIDGE_INPUT(ifp, m);
715 		if (m == NULL) {
716 			CURVNET_RESTORE();
717 			return;
718 		}
719 	}
720 
721 #if defined(INET) || defined(INET6)
722 	/*
723 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
724 	 * mbuf flows up to Layer 3.
725 	 * FreeBSD's implementation of carp(4) uses the inprotosw
726 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
727 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
728 	 * is outside the scope of the M_PROMISC test below.
729 	 * TODO: Maintain a hash table of ethernet addresses other than
730 	 * ether_dhost which may be active on this ifp.
731 	 */
732 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
733 		m->m_flags &= ~M_PROMISC;
734 	} else
735 #endif
736 	{
737 		/*
738 		 * If the frame received was not for our MAC address, set the
739 		 * M_PROMISC flag on the mbuf chain. The frame may need to
740 		 * be seen by the rest of the Ethernet input path in case of
741 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
742 		 * seen by upper protocol layers.
743 		 */
744 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
745 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
746 			m->m_flags |= M_PROMISC;
747 	}
748 
749 	/* First chunk of an mbuf contains good entropy */
750 	if (harvest.ethernet)
751 		random_harvest(m, 16, 3, 0, RANDOM_NET);
752 
753 	ether_demux(ifp, m);
754 	CURVNET_RESTORE();
755 }
756 
757 /*
758  * Ethernet input dispatch; by default, direct dispatch here regardless of
759  * global configuration.
760  */
761 static void
762 ether_nh_input(struct mbuf *m)
763 {
764 
765 	ether_input_internal(m->m_pkthdr.rcvif, m);
766 }
767 
768 static struct netisr_handler	ether_nh = {
769 	.nh_name = "ether",
770 	.nh_handler = ether_nh_input,
771 	.nh_proto = NETISR_ETHER,
772 	.nh_policy = NETISR_POLICY_SOURCE,
773 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
774 };
775 
776 static void
777 ether_init(__unused void *arg)
778 {
779 
780 	netisr_register(&ether_nh);
781 }
782 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
783 
784 static void
785 ether_input(struct ifnet *ifp, struct mbuf *m)
786 {
787 
788 	/*
789 	 * We will rely on rcvif being set properly in the deferred context,
790 	 * so assert it is correct here.
791 	 */
792 	KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__));
793 
794 	netisr_dispatch(NETISR_ETHER, m);
795 }
796 
797 /*
798  * Upper layer processing for a received Ethernet packet.
799  */
800 void
801 ether_demux(struct ifnet *ifp, struct mbuf *m)
802 {
803 	struct ether_header *eh;
804 	int isr;
805 	u_short ether_type;
806 #if defined(NETATALK)
807 	struct llc *l;
808 #endif
809 
810 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
811 
812 #if defined(INET) || defined(INET6)
813 	/*
814 	 * Allow dummynet and/or ipfw to claim the frame.
815 	 * Do not do this for PROMISC frames in case we are re-entered.
816 	 */
817 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
818 		if (ether_ipfw_chk(&m, NULL, 0) == 0) {
819 			if (m)
820 				m_freem(m);	/* dropped; free mbuf chain */
821 			return;			/* consumed */
822 		}
823 	}
824 #endif
825 	eh = mtod(m, struct ether_header *);
826 	ether_type = ntohs(eh->ether_type);
827 
828 	/*
829 	 * If this frame has a VLAN tag other than 0, call vlan_input()
830 	 * if its module is loaded. Otherwise, drop.
831 	 */
832 	if ((m->m_flags & M_VLANTAG) &&
833 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
834 		if (ifp->if_vlantrunk == NULL) {
835 			ifp->if_noproto++;
836 			m_freem(m);
837 			return;
838 		}
839 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
840 		    __func__));
841 		/* Clear before possibly re-entering ether_input(). */
842 		m->m_flags &= ~M_PROMISC;
843 		(*vlan_input_p)(ifp, m);
844 		return;
845 	}
846 
847 	/*
848 	 * Pass promiscuously received frames to the upper layer if the user
849 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
850 	 */
851 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
852 		m_freem(m);
853 		return;
854 	}
855 
856 	/*
857 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
858 	 * Strip off Ethernet header.
859 	 */
860 	m->m_flags &= ~M_VLANTAG;
861 	m->m_flags &= ~(M_PROTOFLAGS);
862 	m_adj(m, ETHER_HDR_LEN);
863 
864 	/*
865 	 * Dispatch frame to upper layer.
866 	 */
867 	switch (ether_type) {
868 #ifdef INET
869 	case ETHERTYPE_IP:
870 		if ((m = ip_fastforward(m)) == NULL)
871 			return;
872 		isr = NETISR_IP;
873 		break;
874 
875 	case ETHERTYPE_ARP:
876 		if (ifp->if_flags & IFF_NOARP) {
877 			/* Discard packet if ARP is disabled on interface */
878 			m_freem(m);
879 			return;
880 		}
881 		isr = NETISR_ARP;
882 		break;
883 #endif
884 #ifdef IPX
885 	case ETHERTYPE_IPX:
886 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
887 			return;
888 		isr = NETISR_IPX;
889 		break;
890 #endif
891 #ifdef INET6
892 	case ETHERTYPE_IPV6:
893 		isr = NETISR_IPV6;
894 		break;
895 #endif
896 #ifdef NETATALK
897 	case ETHERTYPE_AT:
898 		isr = NETISR_ATALK1;
899 		break;
900 	case ETHERTYPE_AARP:
901 		isr = NETISR_AARP;
902 		break;
903 #endif /* NETATALK */
904 	default:
905 #ifdef IPX
906 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
907 			return;
908 #endif /* IPX */
909 #if defined(NETATALK)
910 		if (ether_type > ETHERMTU)
911 			goto discard;
912 		l = mtod(m, struct llc *);
913 		if (l->llc_dsap == LLC_SNAP_LSAP &&
914 		    l->llc_ssap == LLC_SNAP_LSAP &&
915 		    l->llc_control == LLC_UI) {
916 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
917 			    sizeof(at_org_code)) == 0 &&
918 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
919 				m_adj(m, LLC_SNAPFRAMELEN);
920 				isr = NETISR_ATALK2;
921 				break;
922 			}
923 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
924 			    sizeof(aarp_org_code)) == 0 &&
925 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
926 				m_adj(m, LLC_SNAPFRAMELEN);
927 				isr = NETISR_AARP;
928 				break;
929 			}
930 		}
931 #endif /* NETATALK */
932 		goto discard;
933 	}
934 	netisr_dispatch(isr, m);
935 	return;
936 
937 discard:
938 	/*
939 	 * Packet is to be discarded.  If netgraph is present,
940 	 * hand the packet to it for last chance processing;
941 	 * otherwise dispose of it.
942 	 */
943 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
944 		KASSERT(ng_ether_input_orphan_p != NULL,
945 		    ("ng_ether_input_orphan_p is NULL"));
946 		/*
947 		 * Put back the ethernet header so netgraph has a
948 		 * consistent view of inbound packets.
949 		 */
950 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
951 		(*ng_ether_input_orphan_p)(ifp, m);
952 		return;
953 	}
954 	m_freem(m);
955 }
956 
957 /*
958  * Convert Ethernet address to printable (loggable) representation.
959  * This routine is for compatibility; it's better to just use
960  *
961  *	printf("%6D", <pointer to address>, ":");
962  *
963  * since there's no static buffer involved.
964  */
965 char *
966 ether_sprintf(const u_char *ap)
967 {
968 	static char etherbuf[18];
969 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
970 	return (etherbuf);
971 }
972 
973 /*
974  * Perform common duties while attaching to interface list
975  */
976 void
977 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
978 {
979 	int i;
980 	struct ifaddr *ifa;
981 	struct sockaddr_dl *sdl;
982 
983 	ifp->if_addrlen = ETHER_ADDR_LEN;
984 	ifp->if_hdrlen = ETHER_HDR_LEN;
985 	if_attach(ifp);
986 	ifp->if_mtu = ETHERMTU;
987 	ifp->if_output = ether_output;
988 	ifp->if_input = ether_input;
989 	ifp->if_resolvemulti = ether_resolvemulti;
990 #ifdef VIMAGE
991 	ifp->if_reassign = ether_reassign;
992 #endif
993 	if (ifp->if_baudrate == 0)
994 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
995 	ifp->if_broadcastaddr = etherbroadcastaddr;
996 
997 	ifa = ifp->if_addr;
998 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
999 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1000 	sdl->sdl_type = IFT_ETHER;
1001 	sdl->sdl_alen = ifp->if_addrlen;
1002 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1003 
1004 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1005 	if (ng_ether_attach_p != NULL)
1006 		(*ng_ether_attach_p)(ifp);
1007 
1008 	/* Announce Ethernet MAC address if non-zero. */
1009 	for (i = 0; i < ifp->if_addrlen; i++)
1010 		if (lla[i] != 0)
1011 			break;
1012 	if (i != ifp->if_addrlen)
1013 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1014 }
1015 
1016 /*
1017  * Perform common duties while detaching an Ethernet interface
1018  */
1019 void
1020 ether_ifdetach(struct ifnet *ifp)
1021 {
1022 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
1023 		KASSERT(ng_ether_detach_p != NULL,
1024 		    ("ng_ether_detach_p is NULL"));
1025 		(*ng_ether_detach_p)(ifp);
1026 	}
1027 
1028 	bpfdetach(ifp);
1029 	if_detach(ifp);
1030 }
1031 
1032 #ifdef VIMAGE
1033 void
1034 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1035 {
1036 
1037 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
1038 		KASSERT(ng_ether_detach_p != NULL,
1039 		    ("ng_ether_detach_p is NULL"));
1040 		(*ng_ether_detach_p)(ifp);
1041 	}
1042 
1043 	if (ng_ether_attach_p != NULL) {
1044 		CURVNET_SET_QUIET(new_vnet);
1045 		(*ng_ether_attach_p)(ifp);
1046 		CURVNET_RESTORE();
1047 	}
1048 }
1049 #endif
1050 
1051 SYSCTL_DECL(_net_link);
1052 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1053 #if defined(INET) || defined(INET6)
1054 SYSCTL_VNET_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
1055 	     &VNET_NAME(ether_ipfw), 0, "Pass ether pkts through firewall");
1056 #endif
1057 
1058 #if 0
1059 /*
1060  * This is for reference.  We have a table-driven version
1061  * of the little-endian crc32 generator, which is faster
1062  * than the double-loop.
1063  */
1064 uint32_t
1065 ether_crc32_le(const uint8_t *buf, size_t len)
1066 {
1067 	size_t i;
1068 	uint32_t crc;
1069 	int bit;
1070 	uint8_t data;
1071 
1072 	crc = 0xffffffff;	/* initial value */
1073 
1074 	for (i = 0; i < len; i++) {
1075 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1076 			carry = (crc ^ data) & 1;
1077 			crc >>= 1;
1078 			if (carry)
1079 				crc = (crc ^ ETHER_CRC_POLY_LE);
1080 		}
1081 	}
1082 
1083 	return (crc);
1084 }
1085 #else
1086 uint32_t
1087 ether_crc32_le(const uint8_t *buf, size_t len)
1088 {
1089 	static const uint32_t crctab[] = {
1090 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1091 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1092 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1093 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1094 	};
1095 	size_t i;
1096 	uint32_t crc;
1097 
1098 	crc = 0xffffffff;	/* initial value */
1099 
1100 	for (i = 0; i < len; i++) {
1101 		crc ^= buf[i];
1102 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1103 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1104 	}
1105 
1106 	return (crc);
1107 }
1108 #endif
1109 
1110 uint32_t
1111 ether_crc32_be(const uint8_t *buf, size_t len)
1112 {
1113 	size_t i;
1114 	uint32_t crc, carry;
1115 	int bit;
1116 	uint8_t data;
1117 
1118 	crc = 0xffffffff;	/* initial value */
1119 
1120 	for (i = 0; i < len; i++) {
1121 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1122 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1123 			crc <<= 1;
1124 			if (carry)
1125 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1126 		}
1127 	}
1128 
1129 	return (crc);
1130 }
1131 
1132 int
1133 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1134 {
1135 	struct ifaddr *ifa = (struct ifaddr *) data;
1136 	struct ifreq *ifr = (struct ifreq *) data;
1137 	int error = 0;
1138 
1139 	switch (command) {
1140 	case SIOCSIFADDR:
1141 		ifp->if_flags |= IFF_UP;
1142 
1143 		switch (ifa->ifa_addr->sa_family) {
1144 #ifdef INET
1145 		case AF_INET:
1146 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1147 			arp_ifinit(ifp, ifa);
1148 			break;
1149 #endif
1150 #ifdef IPX
1151 		/*
1152 		 * XXX - This code is probably wrong
1153 		 */
1154 		case AF_IPX:
1155 			{
1156 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1157 
1158 			if (ipx_nullhost(*ina))
1159 				ina->x_host =
1160 				    *(union ipx_host *)
1161 				    IF_LLADDR(ifp);
1162 			else {
1163 				bcopy((caddr_t) ina->x_host.c_host,
1164 				      (caddr_t) IF_LLADDR(ifp),
1165 				      ETHER_ADDR_LEN);
1166 			}
1167 
1168 			/*
1169 			 * Set new address
1170 			 */
1171 			ifp->if_init(ifp->if_softc);
1172 			break;
1173 			}
1174 #endif
1175 		default:
1176 			ifp->if_init(ifp->if_softc);
1177 			break;
1178 		}
1179 		break;
1180 
1181 	case SIOCGIFADDR:
1182 		{
1183 			struct sockaddr *sa;
1184 
1185 			sa = (struct sockaddr *) & ifr->ifr_data;
1186 			bcopy(IF_LLADDR(ifp),
1187 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1188 		}
1189 		break;
1190 
1191 	case SIOCSIFMTU:
1192 		/*
1193 		 * Set the interface MTU.
1194 		 */
1195 		if (ifr->ifr_mtu > ETHERMTU) {
1196 			error = EINVAL;
1197 		} else {
1198 			ifp->if_mtu = ifr->ifr_mtu;
1199 		}
1200 		break;
1201 	default:
1202 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1203 		break;
1204 	}
1205 	return (error);
1206 }
1207 
1208 static int
1209 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1210 	struct sockaddr *sa)
1211 {
1212 	struct sockaddr_dl *sdl;
1213 #ifdef INET
1214 	struct sockaddr_in *sin;
1215 #endif
1216 #ifdef INET6
1217 	struct sockaddr_in6 *sin6;
1218 #endif
1219 	u_char *e_addr;
1220 
1221 	switch(sa->sa_family) {
1222 	case AF_LINK:
1223 		/*
1224 		 * No mapping needed. Just check that it's a valid MC address.
1225 		 */
1226 		sdl = (struct sockaddr_dl *)sa;
1227 		e_addr = LLADDR(sdl);
1228 		if (!ETHER_IS_MULTICAST(e_addr))
1229 			return EADDRNOTAVAIL;
1230 		*llsa = 0;
1231 		return 0;
1232 
1233 #ifdef INET
1234 	case AF_INET:
1235 		sin = (struct sockaddr_in *)sa;
1236 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1237 			return EADDRNOTAVAIL;
1238 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1239 		       M_NOWAIT|M_ZERO);
1240 		if (sdl == NULL)
1241 			return ENOMEM;
1242 		sdl->sdl_len = sizeof *sdl;
1243 		sdl->sdl_family = AF_LINK;
1244 		sdl->sdl_index = ifp->if_index;
1245 		sdl->sdl_type = IFT_ETHER;
1246 		sdl->sdl_alen = ETHER_ADDR_LEN;
1247 		e_addr = LLADDR(sdl);
1248 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1249 		*llsa = (struct sockaddr *)sdl;
1250 		return 0;
1251 #endif
1252 #ifdef INET6
1253 	case AF_INET6:
1254 		sin6 = (struct sockaddr_in6 *)sa;
1255 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1256 			/*
1257 			 * An IP6 address of 0 means listen to all
1258 			 * of the Ethernet multicast address used for IP6.
1259 			 * (This is used for multicast routers.)
1260 			 */
1261 			ifp->if_flags |= IFF_ALLMULTI;
1262 			*llsa = 0;
1263 			return 0;
1264 		}
1265 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1266 			return EADDRNOTAVAIL;
1267 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1268 		       M_NOWAIT|M_ZERO);
1269 		if (sdl == NULL)
1270 			return (ENOMEM);
1271 		sdl->sdl_len = sizeof *sdl;
1272 		sdl->sdl_family = AF_LINK;
1273 		sdl->sdl_index = ifp->if_index;
1274 		sdl->sdl_type = IFT_ETHER;
1275 		sdl->sdl_alen = ETHER_ADDR_LEN;
1276 		e_addr = LLADDR(sdl);
1277 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1278 		*llsa = (struct sockaddr *)sdl;
1279 		return 0;
1280 #endif
1281 
1282 	default:
1283 		/*
1284 		 * Well, the text isn't quite right, but it's the name
1285 		 * that counts...
1286 		 */
1287 		return EAFNOSUPPORT;
1288 	}
1289 }
1290 
1291 static void*
1292 ether_alloc(u_char type, struct ifnet *ifp)
1293 {
1294 	struct arpcom	*ac;
1295 
1296 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1297 	ac->ac_ifp = ifp;
1298 
1299 	return (ac);
1300 }
1301 
1302 static void
1303 ether_free(void *com, u_char type)
1304 {
1305 
1306 	free(com, M_ARPCOM);
1307 }
1308 
1309 static int
1310 ether_modevent(module_t mod, int type, void *data)
1311 {
1312 
1313 	switch (type) {
1314 	case MOD_LOAD:
1315 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1316 		break;
1317 	case MOD_UNLOAD:
1318 		if_deregister_com_alloc(IFT_ETHER);
1319 		break;
1320 	default:
1321 		return EOPNOTSUPP;
1322 	}
1323 
1324 	return (0);
1325 }
1326 
1327 static moduledata_t ether_mod = {
1328 	"ether",
1329 	ether_modevent,
1330 	0
1331 };
1332 
1333 void
1334 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1335 {
1336 	struct ether_vlan_header vlan;
1337 	struct mbuf mv, mb;
1338 
1339 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1340 	    ("%s: vlan information not present", __func__));
1341 	KASSERT(m->m_len >= sizeof(struct ether_header),
1342 	    ("%s: mbuf not large enough for header", __func__));
1343 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1344 	vlan.evl_proto = vlan.evl_encap_proto;
1345 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1346 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1347 	m->m_len -= sizeof(struct ether_header);
1348 	m->m_data += sizeof(struct ether_header);
1349 	/*
1350 	 * If a data link has been supplied by the caller, then we will need to
1351 	 * re-create a stack allocated mbuf chain with the following structure:
1352 	 *
1353 	 * (1) mbuf #1 will contain the supplied data link
1354 	 * (2) mbuf #2 will contain the vlan header
1355 	 * (3) mbuf #3 will contain the original mbuf's packet data
1356 	 *
1357 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1358 	 */
1359 	if (data != NULL) {
1360 		mv.m_next = m;
1361 		mv.m_data = (caddr_t)&vlan;
1362 		mv.m_len = sizeof(vlan);
1363 		mb.m_next = &mv;
1364 		mb.m_data = data;
1365 		mb.m_len = dlen;
1366 		bpf_mtap(bp, &mb);
1367 	} else
1368 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1369 	m->m_len += sizeof(struct ether_header);
1370 	m->m_data -= sizeof(struct ether_header);
1371 }
1372 
1373 struct mbuf *
1374 ether_vlanencap(struct mbuf *m, uint16_t tag)
1375 {
1376 	struct ether_vlan_header *evl;
1377 
1378 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
1379 	if (m == NULL)
1380 		return (NULL);
1381 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1382 
1383 	if (m->m_len < sizeof(*evl)) {
1384 		m = m_pullup(m, sizeof(*evl));
1385 		if (m == NULL)
1386 			return (NULL);
1387 	}
1388 
1389 	/*
1390 	 * Transform the Ethernet header into an Ethernet header
1391 	 * with 802.1Q encapsulation.
1392 	 */
1393 	evl = mtod(m, struct ether_vlan_header *);
1394 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1395 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1396 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1397 	evl->evl_tag = htons(tag);
1398 	return (m);
1399 }
1400 
1401 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1402 MODULE_VERSION(ether, 1);
1403