xref: /freebsd/sys/net/if_ethersubr.c (revision 8be96e101f2691b80ff9562b72f874da82e735aa)
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 #if defined(INET) || defined(INET6)
118 int (*carp_forus_p)(struct ifnet *, u_char *);
119 int (*carp_output_p)(struct ifnet *, struct mbuf *, struct sockaddr *,
120     struct rtentry *);
121 #endif
122 
123 /* if_bridge(4) support */
124 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
125 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
126 		struct sockaddr *, struct rtentry *);
127 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
128 
129 /* if_lagg(4) support */
130 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
131 
132 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
133 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
134 
135 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
136 		struct sockaddr *);
137 
138 /* XXX: should be in an arp support file, not here */
139 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
140 
141 #define	ETHER_IS_BROADCAST(addr) \
142 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
143 
144 #define senderr(e) do { error = (e); goto bad;} while (0)
145 
146 #if defined(INET) || defined(INET6)
147 int
148 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared);
149 static VNET_DEFINE(int, ether_ipfw);
150 #define	V_ether_ipfw	VNET(ether_ipfw)
151 #endif
152 
153 
154 /*
155  * Ethernet output routine.
156  * Encapsulate a packet of type family for the local net.
157  * Use trailer local net encapsulation if enough data in first
158  * packet leaves a multiple of 512 bytes of data in remainder.
159  */
160 int
161 ether_output(struct ifnet *ifp, struct mbuf *m,
162 	struct sockaddr *dst, struct route *ro)
163 {
164 	short type;
165 	int error = 0, hdrcmplt = 0;
166 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
167 	struct llentry *lle = NULL;
168 	struct rtentry *rt0 = NULL;
169 	struct ether_header *eh;
170 	struct pf_mtag *t;
171 	int loop_copy = 1;
172 	int hlen;	/* link layer header length */
173 
174 	if (ro != NULL) {
175 		if (!(m->m_flags & (M_BCAST | M_MCAST)))
176 			lle = ro->ro_lle;
177 		rt0 = ro->ro_rt;
178 	}
179 #ifdef MAC
180 	error = mac_ifnet_check_transmit(ifp, m);
181 	if (error)
182 		senderr(error);
183 #endif
184 
185 	M_PROFILE(m);
186 	if (ifp->if_flags & IFF_MONITOR)
187 		senderr(ENETDOWN);
188 	if (!((ifp->if_flags & IFF_UP) &&
189 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
190 		senderr(ENETDOWN);
191 
192 	hlen = ETHER_HDR_LEN;
193 	switch (dst->sa_family) {
194 #ifdef INET
195 	case AF_INET:
196 		if (lle != NULL && (lle->la_flags & LLE_VALID))
197 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
198 		else
199 			error = arpresolve(ifp, rt0, m, dst, edst, &lle);
200 		if (error)
201 			return (error == EWOULDBLOCK ? 0 : error);
202 		type = htons(ETHERTYPE_IP);
203 		break;
204 	case AF_ARP:
205 	{
206 		struct arphdr *ah;
207 		ah = mtod(m, struct arphdr *);
208 		ah->ar_hrd = htons(ARPHRD_ETHER);
209 
210 		loop_copy = 0; /* if this is for us, don't do it */
211 
212 		switch(ntohs(ah->ar_op)) {
213 		case ARPOP_REVREQUEST:
214 		case ARPOP_REVREPLY:
215 			type = htons(ETHERTYPE_REVARP);
216 			break;
217 		case ARPOP_REQUEST:
218 		case ARPOP_REPLY:
219 		default:
220 			type = htons(ETHERTYPE_ARP);
221 			break;
222 		}
223 
224 		if (m->m_flags & M_BCAST)
225 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
226 		else
227 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
228 
229 	}
230 	break;
231 #endif
232 #ifdef INET6
233 	case AF_INET6:
234 		if (lle != NULL && (lle->la_flags & LLE_VALID))
235 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
236 		else
237 			error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle);
238 		if (error)
239 			return error;
240 		type = htons(ETHERTYPE_IPV6);
241 		break;
242 #endif
243 #ifdef IPX
244 	case AF_IPX:
245 		if (ef_outputp) {
246 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
247 		    if (error)
248 			goto bad;
249 		} else
250 		    type = htons(ETHERTYPE_IPX);
251 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
252 		    (caddr_t)edst, sizeof (edst));
253 		break;
254 #endif
255 #ifdef NETATALK
256 	case AF_APPLETALK:
257 	  {
258 	    struct at_ifaddr *aa;
259 
260 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
261 		    senderr(EHOSTUNREACH); /* XXX */
262 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) {
263 		    ifa_free(&aa->aa_ifa);
264 		    return (0);
265 	    }
266 	    /*
267 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
268 	     */
269 	    if ( aa->aa_flags & AFA_PHASE2 ) {
270 		struct llc llc;
271 
272 		ifa_free(&aa->aa_ifa);
273 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
274 		if (m == NULL)
275 			senderr(ENOBUFS);
276 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
277 		llc.llc_control = LLC_UI;
278 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
279 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
280 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
281 		type = htons(m->m_pkthdr.len);
282 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
283 	    } else {
284 		ifa_free(&aa->aa_ifa);
285 		type = htons(ETHERTYPE_AT);
286 	    }
287 	    break;
288 	  }
289 #endif /* NETATALK */
290 
291 	case pseudo_AF_HDRCMPLT:
292 		hdrcmplt = 1;
293 		eh = (struct ether_header *)dst->sa_data;
294 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
295 		/* FALLTHROUGH */
296 
297 	case AF_UNSPEC:
298 		loop_copy = 0; /* if this is for us, don't do it */
299 		eh = (struct ether_header *)dst->sa_data;
300 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
301 		type = eh->ether_type;
302 		break;
303 
304 	default:
305 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
306 		senderr(EAFNOSUPPORT);
307 	}
308 
309 	if (lle != NULL && (lle->la_flags & LLE_IFADDR)) {
310 		int csum_flags = 0;
311 		if (m->m_pkthdr.csum_flags & CSUM_IP)
312 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
313 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
314 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
315 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
316 			csum_flags |= CSUM_SCTP_VALID;
317 		m->m_pkthdr.csum_flags |= csum_flags;
318 		m->m_pkthdr.csum_data = 0xffff;
319 		return (if_simloop(ifp, m, dst->sa_family, 0));
320 	}
321 
322 	/*
323 	 * Add local net header.  If no space in first mbuf,
324 	 * allocate another.
325 	 */
326 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
327 	if (m == NULL)
328 		senderr(ENOBUFS);
329 	eh = mtod(m, struct ether_header *);
330 	(void)memcpy(&eh->ether_type, &type,
331 		sizeof(eh->ether_type));
332 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
333 	if (hdrcmplt)
334 		(void)memcpy(eh->ether_shost, esrc,
335 			sizeof(eh->ether_shost));
336 	else
337 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
338 			sizeof(eh->ether_shost));
339 
340 	/*
341 	 * If a simplex interface, and the packet is being sent to our
342 	 * Ethernet address or a broadcast address, loopback a copy.
343 	 * XXX To make a simplex device behave exactly like a duplex
344 	 * device, we should copy in the case of sending to our own
345 	 * ethernet address (thus letting the original actually appear
346 	 * on the wire). However, we don't do that here for security
347 	 * reasons and compatibility with the original behavior.
348 	 */
349 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
350 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
351 		int csum_flags = 0;
352 
353 		if (m->m_pkthdr.csum_flags & CSUM_IP)
354 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
355 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
356 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
357 		if (m->m_pkthdr.csum_flags & CSUM_SCTP)
358 			csum_flags |= CSUM_SCTP_VALID;
359 
360 		if (m->m_flags & M_BCAST) {
361 			struct mbuf *n;
362 
363 			/*
364 			 * Because if_simloop() modifies the packet, we need a
365 			 * writable copy through m_dup() instead of a readonly
366 			 * one as m_copy[m] would give us. The alternative would
367 			 * be to modify if_simloop() to handle the readonly mbuf,
368 			 * but performancewise it is mostly equivalent (trading
369 			 * extra data copying vs. extra locking).
370 			 *
371 			 * XXX This is a local workaround.  A number of less
372 			 * often used kernel parts suffer from the same bug.
373 			 * See PR kern/105943 for a proposed general solution.
374 			 */
375 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
376 				n->m_pkthdr.csum_flags |= csum_flags;
377 				if (csum_flags & CSUM_DATA_VALID)
378 					n->m_pkthdr.csum_data = 0xffff;
379 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
380 			} else
381 				ifp->if_iqdrops++;
382 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
383 				ETHER_ADDR_LEN) == 0) {
384 			m->m_pkthdr.csum_flags |= csum_flags;
385 			if (csum_flags & CSUM_DATA_VALID)
386 				m->m_pkthdr.csum_data = 0xffff;
387 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
388 			return (0);	/* XXX */
389 		}
390 	}
391 
392        /*
393 	* Bridges require special output handling.
394 	*/
395 	if (ifp->if_bridge) {
396 		BRIDGE_OUTPUT(ifp, m, error);
397 		return (error);
398 	}
399 
400 #if defined(INET) || defined(INET6)
401 	if (ifp->if_carp &&
402 	    (error = (*carp_output_p)(ifp, m, dst, NULL)))
403 		goto bad;
404 #endif
405 
406 	/* Handle ng_ether(4) processing, if any */
407 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
408 		KASSERT(ng_ether_output_p != NULL,
409 		    ("ng_ether_output_p is NULL"));
410 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
411 bad:			if (m != NULL)
412 				m_freem(m);
413 			return (error);
414 		}
415 		if (m == NULL)
416 			return (0);
417 	}
418 
419 	/* Continue with link-layer output */
420 	return ether_output_frame(ifp, m);
421 }
422 
423 /*
424  * Ethernet link layer output routine to send a raw frame to the device.
425  *
426  * This assumes that the 14 byte Ethernet header is present and contiguous
427  * in the first mbuf (if BRIDGE'ing).
428  */
429 int
430 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
431 {
432 #if defined(INET) || defined(INET6)
433 
434 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0) {
435 		if (ether_ipfw_chk(&m, ifp, 0) == 0) {
436 			if (m) {
437 				m_freem(m);
438 				return EACCES;	/* pkt dropped */
439 			} else
440 				return 0;	/* consumed e.g. in a pipe */
441 		}
442 	}
443 #endif
444 
445 	/*
446 	 * Queue message on interface, update output statistics if
447 	 * successful, and start output if interface not yet active.
448 	 */
449 	return ((ifp->if_transmit)(ifp, m));
450 }
451 
452 #if defined(INET) || defined(INET6)
453 /*
454  * ipfw processing for ethernet packets (in and out).
455  * The second parameter is NULL from ether_demux, and ifp from
456  * ether_output_frame.
457  */
458 int
459 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, int shared)
460 {
461 	struct ether_header *eh;
462 	struct ether_header save_eh;
463 	struct mbuf *m;
464 	int i;
465 	struct ip_fw_args args;
466 	struct m_tag *mtag;
467 
468 	/* fetch start point from rule, if any */
469 	mtag = m_tag_locate(*m0, MTAG_IPFW_RULE, 0, NULL);
470 	if (mtag == NULL) {
471 		args.rule.slot = 0;
472 	} else {
473 		/* dummynet packet, already partially processed */
474 		struct ipfw_rule_ref *r;
475 
476 		/* XXX can we free it after use ? */
477 		mtag->m_tag_id = PACKET_TAG_NONE;
478 		r = (struct ipfw_rule_ref *)(mtag + 1);
479 		if (r->info & IPFW_ONEPASS)
480 			return (1);
481 		args.rule = *r;
482 	}
483 
484 	/*
485 	 * I need some amt of data to be contiguous, and in case others need
486 	 * the packet (shared==1) also better be in the first mbuf.
487 	 */
488 	m = *m0;
489 	i = min( m->m_pkthdr.len, max_protohdr);
490 	if ( shared || m->m_len < i) {
491 		m = m_pullup(m, i);
492 		if (m == NULL) {
493 			*m0 = m;
494 			return 0;
495 		}
496 	}
497 	eh = mtod(m, struct ether_header *);
498 	save_eh = *eh;			/* save copy for restore below */
499 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
500 
501 	args.m = m;		/* the packet we are looking at		*/
502 	args.oif = dst;		/* destination, if any			*/
503 	args.next_hop = NULL;	/* we do not support forward yet	*/
504 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
505 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
506 	i = V_ip_fw_chk_ptr(&args);
507 	m = args.m;
508 	if (m != NULL) {
509 		/*
510 		 * Restore Ethernet header, as needed, in case the
511 		 * mbuf chain was replaced by ipfw.
512 		 */
513 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
514 		if (m == NULL) {
515 			*m0 = m;
516 			return 0;
517 		}
518 		if (eh != mtod(m, struct ether_header *))
519 			bcopy(&save_eh, mtod(m, struct ether_header *),
520 				ETHER_HDR_LEN);
521 	}
522 	*m0 = m;
523 
524 	if (i == IP_FW_DENY) /* drop */
525 		return 0;
526 
527 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
528 
529 	if (i == IP_FW_PASS) /* a PASS rule.  */
530 		return 1;
531 
532 	if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
533 		int dir;
534 		/*
535 		 * Pass the pkt to dummynet, which consumes it.
536 		 * If shared, make a copy and keep the original.
537 		 */
538 		if (shared) {
539 			m = m_copypacket(m, M_DONTWAIT);
540 			if (m == NULL)
541 				return 0;
542 		} else {
543 			/*
544 			 * Pass the original to dummynet and
545 			 * nothing back to the caller
546 			 */
547 			*m0 = NULL ;
548 		}
549 		dir = PROTO_LAYER2 | (dst ? DIR_OUT : DIR_IN);
550 		ip_dn_io_ptr(&m, dir, &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 	/*
725 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
726 	 * mbuf flows up to Layer 3.
727 	 * FreeBSD's implementation of carp(4) uses the inprotosw
728 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
729 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
730 	 * is outside the scope of the M_PROMISC test below.
731 	 * TODO: Maintain a hash table of ethernet addresses other than
732 	 * ether_dhost which may be active on this ifp.
733 	 */
734 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
735 		m->m_flags &= ~M_PROMISC;
736 	} else
737 #endif
738 	{
739 		/*
740 		 * If the frame received was not for our MAC address, set the
741 		 * M_PROMISC flag on the mbuf chain. The frame may need to
742 		 * be seen by the rest of the Ethernet input path in case of
743 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
744 		 * seen by upper protocol layers.
745 		 */
746 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
747 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
748 			m->m_flags |= M_PROMISC;
749 	}
750 
751 	/* First chunk of an mbuf contains good entropy */
752 	if (harvest.ethernet)
753 		random_harvest(m, 16, 3, 0, RANDOM_NET);
754 
755 	ether_demux(ifp, m);
756 	CURVNET_RESTORE();
757 }
758 
759 /*
760  * Upper layer processing for a received Ethernet packet.
761  */
762 void
763 ether_demux(struct ifnet *ifp, struct mbuf *m)
764 {
765 	struct ether_header *eh;
766 	int isr;
767 	u_short ether_type;
768 #if defined(NETATALK)
769 	struct llc *l;
770 #endif
771 
772 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
773 
774 #if defined(INET) || defined(INET6)
775 	/*
776 	 * Allow dummynet and/or ipfw to claim the frame.
777 	 * Do not do this for PROMISC frames in case we are re-entered.
778 	 */
779 	if (V_ip_fw_chk_ptr && V_ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
780 		if (ether_ipfw_chk(&m, NULL, 0) == 0) {
781 			if (m)
782 				m_freem(m);	/* dropped; free mbuf chain */
783 			return;			/* consumed */
784 		}
785 	}
786 #endif
787 	eh = mtod(m, struct ether_header *);
788 	ether_type = ntohs(eh->ether_type);
789 
790 	/*
791 	 * If this frame has a VLAN tag other than 0, call vlan_input()
792 	 * if its module is loaded. Otherwise, drop.
793 	 */
794 	if ((m->m_flags & M_VLANTAG) &&
795 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
796 		if (ifp->if_vlantrunk == NULL) {
797 			ifp->if_noproto++;
798 			m_freem(m);
799 			return;
800 		}
801 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
802 		    __func__));
803 		/* Clear before possibly re-entering ether_input(). */
804 		m->m_flags &= ~M_PROMISC;
805 		(*vlan_input_p)(ifp, m);
806 		return;
807 	}
808 
809 	/*
810 	 * Pass promiscuously received frames to the upper layer if the user
811 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
812 	 */
813 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
814 		m_freem(m);
815 		return;
816 	}
817 
818 	/*
819 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
820 	 * Strip off Ethernet header.
821 	 */
822 	m->m_flags &= ~M_VLANTAG;
823 	m->m_flags &= ~(M_PROTOFLAGS);
824 	m_adj(m, ETHER_HDR_LEN);
825 
826 	/*
827 	 * Dispatch frame to upper layer.
828 	 */
829 	switch (ether_type) {
830 #ifdef INET
831 	case ETHERTYPE_IP:
832 		if ((m = ip_fastforward(m)) == NULL)
833 			return;
834 		isr = NETISR_IP;
835 		break;
836 
837 	case ETHERTYPE_ARP:
838 		if (ifp->if_flags & IFF_NOARP) {
839 			/* Discard packet if ARP is disabled on interface */
840 			m_freem(m);
841 			return;
842 		}
843 		isr = NETISR_ARP;
844 		break;
845 #endif
846 #ifdef IPX
847 	case ETHERTYPE_IPX:
848 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
849 			return;
850 		isr = NETISR_IPX;
851 		break;
852 #endif
853 #ifdef INET6
854 	case ETHERTYPE_IPV6:
855 		isr = NETISR_IPV6;
856 		break;
857 #endif
858 #ifdef NETATALK
859 	case ETHERTYPE_AT:
860 		isr = NETISR_ATALK1;
861 		break;
862 	case ETHERTYPE_AARP:
863 		isr = NETISR_AARP;
864 		break;
865 #endif /* NETATALK */
866 	default:
867 #ifdef IPX
868 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
869 			return;
870 #endif /* IPX */
871 #if defined(NETATALK)
872 		if (ether_type > ETHERMTU)
873 			goto discard;
874 		l = mtod(m, struct llc *);
875 		if (l->llc_dsap == LLC_SNAP_LSAP &&
876 		    l->llc_ssap == LLC_SNAP_LSAP &&
877 		    l->llc_control == LLC_UI) {
878 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
879 			    sizeof(at_org_code)) == 0 &&
880 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
881 				m_adj(m, LLC_SNAPFRAMELEN);
882 				isr = NETISR_ATALK2;
883 				break;
884 			}
885 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
886 			    sizeof(aarp_org_code)) == 0 &&
887 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
888 				m_adj(m, LLC_SNAPFRAMELEN);
889 				isr = NETISR_AARP;
890 				break;
891 			}
892 		}
893 #endif /* NETATALK */
894 		goto discard;
895 	}
896 	netisr_dispatch(isr, m);
897 	return;
898 
899 discard:
900 	/*
901 	 * Packet is to be discarded.  If netgraph is present,
902 	 * hand the packet to it for last chance processing;
903 	 * otherwise dispose of it.
904 	 */
905 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
906 		KASSERT(ng_ether_input_orphan_p != NULL,
907 		    ("ng_ether_input_orphan_p is NULL"));
908 		/*
909 		 * Put back the ethernet header so netgraph has a
910 		 * consistent view of inbound packets.
911 		 */
912 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
913 		(*ng_ether_input_orphan_p)(ifp, m);
914 		return;
915 	}
916 	m_freem(m);
917 }
918 
919 /*
920  * Convert Ethernet address to printable (loggable) representation.
921  * This routine is for compatibility; it's better to just use
922  *
923  *	printf("%6D", <pointer to address>, ":");
924  *
925  * since there's no static buffer involved.
926  */
927 char *
928 ether_sprintf(const u_char *ap)
929 {
930 	static char etherbuf[18];
931 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
932 	return (etherbuf);
933 }
934 
935 /*
936  * Perform common duties while attaching to interface list
937  */
938 void
939 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
940 {
941 	int i;
942 	struct ifaddr *ifa;
943 	struct sockaddr_dl *sdl;
944 
945 	ifp->if_addrlen = ETHER_ADDR_LEN;
946 	ifp->if_hdrlen = ETHER_HDR_LEN;
947 	if_attach(ifp);
948 	ifp->if_mtu = ETHERMTU;
949 	ifp->if_output = ether_output;
950 	ifp->if_input = ether_input;
951 	ifp->if_resolvemulti = ether_resolvemulti;
952 	if (ifp->if_baudrate == 0)
953 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
954 	ifp->if_broadcastaddr = etherbroadcastaddr;
955 
956 	ifa = ifp->if_addr;
957 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
958 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
959 	sdl->sdl_type = IFT_ETHER;
960 	sdl->sdl_alen = ifp->if_addrlen;
961 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
962 
963 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
964 	if (ng_ether_attach_p != NULL)
965 		(*ng_ether_attach_p)(ifp);
966 
967 	/* Announce Ethernet MAC address if non-zero. */
968 	for (i = 0; i < ifp->if_addrlen; i++)
969 		if (lla[i] != 0)
970 			break;
971 	if (i != ifp->if_addrlen)
972 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
973 }
974 
975 /*
976  * Perform common duties while detaching an Ethernet interface
977  */
978 void
979 ether_ifdetach(struct ifnet *ifp)
980 {
981 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
982 		KASSERT(ng_ether_detach_p != NULL,
983 		    ("ng_ether_detach_p is NULL"));
984 		(*ng_ether_detach_p)(ifp);
985 	}
986 
987 	bpfdetach(ifp);
988 	if_detach(ifp);
989 }
990 
991 SYSCTL_DECL(_net_link);
992 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
993 #if defined(INET) || defined(INET6)
994 SYSCTL_VNET_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
995 	     &VNET_NAME(ether_ipfw), 0, "Pass ether pkts through firewall");
996 #endif
997 
998 #if 0
999 /*
1000  * This is for reference.  We have a table-driven version
1001  * of the little-endian crc32 generator, which is faster
1002  * than the double-loop.
1003  */
1004 uint32_t
1005 ether_crc32_le(const uint8_t *buf, size_t len)
1006 {
1007 	size_t i;
1008 	uint32_t crc;
1009 	int bit;
1010 	uint8_t data;
1011 
1012 	crc = 0xffffffff;	/* initial value */
1013 
1014 	for (i = 0; i < len; i++) {
1015 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1016 			carry = (crc ^ data) & 1;
1017 			crc >>= 1;
1018 			if (carry)
1019 				crc = (crc ^ ETHER_CRC_POLY_LE);
1020 		}
1021 	}
1022 
1023 	return (crc);
1024 }
1025 #else
1026 uint32_t
1027 ether_crc32_le(const uint8_t *buf, size_t len)
1028 {
1029 	static const uint32_t crctab[] = {
1030 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1031 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1032 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1033 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1034 	};
1035 	size_t i;
1036 	uint32_t crc;
1037 
1038 	crc = 0xffffffff;	/* initial value */
1039 
1040 	for (i = 0; i < len; i++) {
1041 		crc ^= buf[i];
1042 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1043 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1044 	}
1045 
1046 	return (crc);
1047 }
1048 #endif
1049 
1050 uint32_t
1051 ether_crc32_be(const uint8_t *buf, size_t len)
1052 {
1053 	size_t i;
1054 	uint32_t crc, carry;
1055 	int bit;
1056 	uint8_t data;
1057 
1058 	crc = 0xffffffff;	/* initial value */
1059 
1060 	for (i = 0; i < len; i++) {
1061 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1062 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1063 			crc <<= 1;
1064 			if (carry)
1065 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1066 		}
1067 	}
1068 
1069 	return (crc);
1070 }
1071 
1072 int
1073 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1074 {
1075 	struct ifaddr *ifa = (struct ifaddr *) data;
1076 	struct ifreq *ifr = (struct ifreq *) data;
1077 	int error = 0;
1078 
1079 	switch (command) {
1080 	case SIOCSIFADDR:
1081 		ifp->if_flags |= IFF_UP;
1082 
1083 		switch (ifa->ifa_addr->sa_family) {
1084 #ifdef INET
1085 		case AF_INET:
1086 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1087 			arp_ifinit(ifp, ifa);
1088 			break;
1089 #endif
1090 #ifdef IPX
1091 		/*
1092 		 * XXX - This code is probably wrong
1093 		 */
1094 		case AF_IPX:
1095 			{
1096 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1097 
1098 			if (ipx_nullhost(*ina))
1099 				ina->x_host =
1100 				    *(union ipx_host *)
1101 				    IF_LLADDR(ifp);
1102 			else {
1103 				bcopy((caddr_t) ina->x_host.c_host,
1104 				      (caddr_t) IF_LLADDR(ifp),
1105 				      ETHER_ADDR_LEN);
1106 			}
1107 
1108 			/*
1109 			 * Set new address
1110 			 */
1111 			ifp->if_init(ifp->if_softc);
1112 			break;
1113 			}
1114 #endif
1115 		default:
1116 			ifp->if_init(ifp->if_softc);
1117 			break;
1118 		}
1119 		break;
1120 
1121 	case SIOCGIFADDR:
1122 		{
1123 			struct sockaddr *sa;
1124 
1125 			sa = (struct sockaddr *) & ifr->ifr_data;
1126 			bcopy(IF_LLADDR(ifp),
1127 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1128 		}
1129 		break;
1130 
1131 	case SIOCSIFMTU:
1132 		/*
1133 		 * Set the interface MTU.
1134 		 */
1135 		if (ifr->ifr_mtu > ETHERMTU) {
1136 			error = EINVAL;
1137 		} else {
1138 			ifp->if_mtu = ifr->ifr_mtu;
1139 		}
1140 		break;
1141 	default:
1142 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1143 		break;
1144 	}
1145 	return (error);
1146 }
1147 
1148 static int
1149 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1150 	struct sockaddr *sa)
1151 {
1152 	struct sockaddr_dl *sdl;
1153 #ifdef INET
1154 	struct sockaddr_in *sin;
1155 #endif
1156 #ifdef INET6
1157 	struct sockaddr_in6 *sin6;
1158 #endif
1159 	u_char *e_addr;
1160 
1161 	switch(sa->sa_family) {
1162 	case AF_LINK:
1163 		/*
1164 		 * No mapping needed. Just check that it's a valid MC address.
1165 		 */
1166 		sdl = (struct sockaddr_dl *)sa;
1167 		e_addr = LLADDR(sdl);
1168 		if (!ETHER_IS_MULTICAST(e_addr))
1169 			return EADDRNOTAVAIL;
1170 		*llsa = 0;
1171 		return 0;
1172 
1173 #ifdef INET
1174 	case AF_INET:
1175 		sin = (struct sockaddr_in *)sa;
1176 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1177 			return EADDRNOTAVAIL;
1178 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1179 		       M_NOWAIT|M_ZERO);
1180 		if (sdl == NULL)
1181 			return ENOMEM;
1182 		sdl->sdl_len = sizeof *sdl;
1183 		sdl->sdl_family = AF_LINK;
1184 		sdl->sdl_index = ifp->if_index;
1185 		sdl->sdl_type = IFT_ETHER;
1186 		sdl->sdl_alen = ETHER_ADDR_LEN;
1187 		e_addr = LLADDR(sdl);
1188 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1189 		*llsa = (struct sockaddr *)sdl;
1190 		return 0;
1191 #endif
1192 #ifdef INET6
1193 	case AF_INET6:
1194 		sin6 = (struct sockaddr_in6 *)sa;
1195 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1196 			/*
1197 			 * An IP6 address of 0 means listen to all
1198 			 * of the Ethernet multicast address used for IP6.
1199 			 * (This is used for multicast routers.)
1200 			 */
1201 			ifp->if_flags |= IFF_ALLMULTI;
1202 			*llsa = 0;
1203 			return 0;
1204 		}
1205 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1206 			return EADDRNOTAVAIL;
1207 		sdl = malloc(sizeof *sdl, M_IFMADDR,
1208 		       M_NOWAIT|M_ZERO);
1209 		if (sdl == NULL)
1210 			return (ENOMEM);
1211 		sdl->sdl_len = sizeof *sdl;
1212 		sdl->sdl_family = AF_LINK;
1213 		sdl->sdl_index = ifp->if_index;
1214 		sdl->sdl_type = IFT_ETHER;
1215 		sdl->sdl_alen = ETHER_ADDR_LEN;
1216 		e_addr = LLADDR(sdl);
1217 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1218 		*llsa = (struct sockaddr *)sdl;
1219 		return 0;
1220 #endif
1221 
1222 	default:
1223 		/*
1224 		 * Well, the text isn't quite right, but it's the name
1225 		 * that counts...
1226 		 */
1227 		return EAFNOSUPPORT;
1228 	}
1229 }
1230 
1231 static void*
1232 ether_alloc(u_char type, struct ifnet *ifp)
1233 {
1234 	struct arpcom	*ac;
1235 
1236 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1237 	ac->ac_ifp = ifp;
1238 
1239 	return (ac);
1240 }
1241 
1242 static void
1243 ether_free(void *com, u_char type)
1244 {
1245 
1246 	free(com, M_ARPCOM);
1247 }
1248 
1249 static int
1250 ether_modevent(module_t mod, int type, void *data)
1251 {
1252 
1253 	switch (type) {
1254 	case MOD_LOAD:
1255 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1256 		break;
1257 	case MOD_UNLOAD:
1258 		if_deregister_com_alloc(IFT_ETHER);
1259 		break;
1260 	default:
1261 		return EOPNOTSUPP;
1262 	}
1263 
1264 	return (0);
1265 }
1266 
1267 static moduledata_t ether_mod = {
1268 	"ether",
1269 	ether_modevent,
1270 	0
1271 };
1272 
1273 void
1274 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1275 {
1276 	struct ether_vlan_header vlan;
1277 	struct mbuf mv, mb;
1278 
1279 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1280 	    ("%s: vlan information not present", __func__));
1281 	KASSERT(m->m_len >= sizeof(struct ether_header),
1282 	    ("%s: mbuf not large enough for header", __func__));
1283 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1284 	vlan.evl_proto = vlan.evl_encap_proto;
1285 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1286 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1287 	m->m_len -= sizeof(struct ether_header);
1288 	m->m_data += sizeof(struct ether_header);
1289 	/*
1290 	 * If a data link has been supplied by the caller, then we will need to
1291 	 * re-create a stack allocated mbuf chain with the following structure:
1292 	 *
1293 	 * (1) mbuf #1 will contain the supplied data link
1294 	 * (2) mbuf #2 will contain the vlan header
1295 	 * (3) mbuf #3 will contain the original mbuf's packet data
1296 	 *
1297 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1298 	 */
1299 	if (data != NULL) {
1300 		mv.m_next = m;
1301 		mv.m_data = (caddr_t)&vlan;
1302 		mv.m_len = sizeof(vlan);
1303 		mb.m_next = &mv;
1304 		mb.m_data = data;
1305 		mb.m_len = dlen;
1306 		bpf_mtap(bp, &mb);
1307 	} else
1308 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1309 	m->m_len += sizeof(struct ether_header);
1310 	m->m_data -= sizeof(struct ether_header);
1311 }
1312 
1313 struct mbuf *
1314 ether_vlanencap(struct mbuf *m, uint16_t tag)
1315 {
1316 	struct ether_vlan_header *evl;
1317 
1318 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
1319 	if (m == NULL)
1320 		return (NULL);
1321 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1322 
1323 	if (m->m_len < sizeof(*evl)) {
1324 		m = m_pullup(m, sizeof(*evl));
1325 		if (m == NULL)
1326 			return (NULL);
1327 	}
1328 
1329 	/*
1330 	 * Transform the Ethernet header into an Ethernet header
1331 	 * with 802.1Q encapsulation.
1332 	 */
1333 	evl = mtod(m, struct ether_vlan_header *);
1334 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1335 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1336 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1337 	evl->evl_tag = htons(tag);
1338 	return (m);
1339 }
1340 
1341 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1342 MODULE_VERSION(ether, 1);
1343