xref: /freebsd/sys/net/if_ethersubr.c (revision f0a75d274af375d15b97b830966b99a02b7db911)
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_mac.h"
38 #include "opt_netgraph.h"
39 #include "opt_carp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
45 #include <sys/module.h>
46 #include <sys/mbuf.h>
47 #include <sys/random.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/if_arp.h>
54 #include <net/netisr.h>
55 #include <net/route.h>
56 #include <net/if_llc.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/bpf.h>
60 #include <net/ethernet.h>
61 #include <net/if_bridgevar.h>
62 #include <net/if_vlan_var.h>
63 
64 #if defined(INET) || defined(INET6)
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/ip_fw.h>
69 #include <netinet/ip_dummynet.h>
70 #endif
71 #ifdef INET6
72 #include <netinet6/nd6.h>
73 #endif
74 
75 #ifdef DEV_CARP
76 #include <netinet/ip_carp.h>
77 #endif
78 
79 #ifdef IPX
80 #include <netipx/ipx.h>
81 #include <netipx/ipx_if.h>
82 #endif
83 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
84 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
85 		struct sockaddr *dst, short *tp, int *hlen);
86 
87 #ifdef NETATALK
88 #include <netatalk/at.h>
89 #include <netatalk/at_var.h>
90 #include <netatalk/at_extern.h>
91 
92 #define llc_snap_org_code llc_un.type_snap.org_code
93 #define llc_snap_ether_type llc_un.type_snap.ether_type
94 
95 extern u_char	at_org_code[3];
96 extern u_char	aarp_org_code[3];
97 #endif /* NETATALK */
98 
99 #include <security/mac/mac_framework.h>
100 
101 /* netgraph node hooks for ng_ether(4) */
102 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void	(*ng_ether_attach_p)(struct ifnet *ifp);
106 void	(*ng_ether_detach_p)(struct ifnet *ifp);
107 
108 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
109 
110 /* if_bridge(4) support */
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
113 		struct sockaddr *, struct rtentry *);
114 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 
116 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
117 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
118 
119 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
120 		struct sockaddr *);
121 
122 /* XXX: should be in an arp support file, not here */
123 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
124 
125 #define	ETHER_IS_BROADCAST(addr) \
126 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
127 
128 #define senderr(e) do { error = (e); goto bad;} while (0)
129 
130 #if defined(INET) || defined(INET6)
131 int
132 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
133 	struct ip_fw **rule, int shared);
134 static int ether_ipfw;
135 #endif
136 
137 /*
138  * Ethernet output routine.
139  * Encapsulate a packet of type family for the local net.
140  * Use trailer local net encapsulation if enough data in first
141  * packet leaves a multiple of 512 bytes of data in remainder.
142  */
143 int
144 ether_output(struct ifnet *ifp, struct mbuf *m,
145 	struct sockaddr *dst, struct rtentry *rt0)
146 {
147 	short type;
148 	int error, hdrcmplt = 0;
149 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
150 	struct ether_header *eh;
151 	int loop_copy = 1;
152 	int hlen;	/* link layer header length */
153 
154 #ifdef MAC
155 	error = mac_check_ifnet_transmit(ifp, m);
156 	if (error)
157 		senderr(error);
158 #endif
159 
160 	if (ifp->if_flags & IFF_MONITOR)
161 		senderr(ENETDOWN);
162 	if (!((ifp->if_flags & IFF_UP) &&
163 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
164 		senderr(ENETDOWN);
165 
166 	hlen = ETHER_HDR_LEN;
167 	switch (dst->sa_family) {
168 #ifdef INET
169 	case AF_INET:
170 		error = arpresolve(ifp, rt0, m, dst, edst);
171 		if (error)
172 			return (error == EWOULDBLOCK ? 0 : error);
173 		type = htons(ETHERTYPE_IP);
174 		break;
175 	case AF_ARP:
176 	{
177 		struct arphdr *ah;
178 		ah = mtod(m, struct arphdr *);
179 		ah->ar_hrd = htons(ARPHRD_ETHER);
180 
181 		loop_copy = 0; /* if this is for us, don't do it */
182 
183 		switch(ntohs(ah->ar_op)) {
184 		case ARPOP_REVREQUEST:
185 		case ARPOP_REVREPLY:
186 			type = htons(ETHERTYPE_REVARP);
187 			break;
188 		case ARPOP_REQUEST:
189 		case ARPOP_REPLY:
190 		default:
191 			type = htons(ETHERTYPE_ARP);
192 			break;
193 		}
194 
195 		if (m->m_flags & M_BCAST)
196 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
197 		else
198 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
199 
200 	}
201 	break;
202 #endif
203 #ifdef INET6
204 	case AF_INET6:
205 		error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst);
206 		if (error)
207 			return error;
208 		type = htons(ETHERTYPE_IPV6);
209 		break;
210 #endif
211 #ifdef IPX
212 	case AF_IPX:
213 		if (ef_outputp) {
214 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
215 		    if (error)
216 			goto bad;
217 		} else
218 		    type = htons(ETHERTYPE_IPX);
219 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
220 		    (caddr_t)edst, sizeof (edst));
221 		break;
222 #endif
223 #ifdef NETATALK
224 	case AF_APPLETALK:
225 	  {
226 	    struct at_ifaddr *aa;
227 
228 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
229 		    senderr(EHOSTUNREACH); /* XXX */
230 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst))
231 		    return (0);
232 	    /*
233 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
234 	     */
235 	    if ( aa->aa_flags & AFA_PHASE2 ) {
236 		struct llc llc;
237 
238 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
239 		if (m == NULL)
240 			senderr(ENOBUFS);
241 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
242 		llc.llc_control = LLC_UI;
243 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
244 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
245 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
246 		type = htons(m->m_pkthdr.len);
247 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
248 	    } else {
249 		type = htons(ETHERTYPE_AT);
250 	    }
251 	    break;
252 	  }
253 #endif /* NETATALK */
254 
255 	case pseudo_AF_HDRCMPLT:
256 		hdrcmplt = 1;
257 		eh = (struct ether_header *)dst->sa_data;
258 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
259 		/* FALLTHROUGH */
260 
261 	case AF_UNSPEC:
262 		loop_copy = 0; /* if this is for us, don't do it */
263 		eh = (struct ether_header *)dst->sa_data;
264 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
265 		type = eh->ether_type;
266 		break;
267 
268 	default:
269 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
270 		senderr(EAFNOSUPPORT);
271 	}
272 
273 	/*
274 	 * Add local net header.  If no space in first mbuf,
275 	 * allocate another.
276 	 */
277 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
278 	if (m == NULL)
279 		senderr(ENOBUFS);
280 	eh = mtod(m, struct ether_header *);
281 	(void)memcpy(&eh->ether_type, &type,
282 		sizeof(eh->ether_type));
283 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
284 	if (hdrcmplt)
285 		(void)memcpy(eh->ether_shost, esrc,
286 			sizeof(eh->ether_shost));
287 	else
288 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
289 			sizeof(eh->ether_shost));
290 
291 	/*
292 	 * If a simplex interface, and the packet is being sent to our
293 	 * Ethernet address or a broadcast address, loopback a copy.
294 	 * XXX To make a simplex device behave exactly like a duplex
295 	 * device, we should copy in the case of sending to our own
296 	 * ethernet address (thus letting the original actually appear
297 	 * on the wire). However, we don't do that here for security
298 	 * reasons and compatibility with the original behavior.
299 	 */
300 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
301 	    m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) {
302 		int csum_flags = 0;
303 
304 		if (m->m_pkthdr.csum_flags & CSUM_IP)
305 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
306 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
307 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
308 
309 		if (m->m_flags & M_BCAST) {
310 			struct mbuf *n;
311 
312 			/*
313 			 * Because if_simloop() modifies the packet, we need a
314 			 * writable copy through m_dup() instead of a readonly
315 			 * one as m_copy[m] would give us. The alternative would
316 			 * be to modify if_simloop() to handle the readonly mbuf,
317 			 * but performancewise it is mostly equivalent (trading
318 			 * extra data copying vs. extra locking).
319 			 *
320 			 * XXX This is a local workaround.  A number of less
321 			 * often used kernel parts suffer from the same bug.
322 			 * See PR kern/105943 for a proposed general solution.
323 			 */
324 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
325 				n->m_pkthdr.csum_flags |= csum_flags;
326 				if (csum_flags & CSUM_DATA_VALID)
327 					n->m_pkthdr.csum_data = 0xffff;
328 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
329 			} else
330 				ifp->if_iqdrops++;
331 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
332 				ETHER_ADDR_LEN) == 0) {
333 			m->m_pkthdr.csum_flags |= csum_flags;
334 			if (csum_flags & CSUM_DATA_VALID)
335 				m->m_pkthdr.csum_data = 0xffff;
336 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
337 			return (0);	/* XXX */
338 		}
339 	}
340 
341        /*
342 	* Bridges require special output handling.
343 	*/
344 	if (ifp->if_bridge) {
345 		BRIDGE_OUTPUT(ifp, m, error);
346 		return (error);
347 	}
348 
349 #ifdef DEV_CARP
350 	if (ifp->if_carp &&
351 	    (error = carp_output(ifp, m, dst, NULL)))
352 		goto bad;
353 #endif
354 
355 	/* Handle ng_ether(4) processing, if any */
356 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
357 		KASSERT(ng_ether_output_p != NULL,
358 		    ("ng_ether_output_p is NULL"));
359 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
360 bad:			if (m != NULL)
361 				m_freem(m);
362 			return (error);
363 		}
364 		if (m == NULL)
365 			return (0);
366 	}
367 
368 	/* Continue with link-layer output */
369 	return ether_output_frame(ifp, m);
370 }
371 
372 /*
373  * Ethernet link layer output routine to send a raw frame to the device.
374  *
375  * This assumes that the 14 byte Ethernet header is present and contiguous
376  * in the first mbuf (if BRIDGE'ing).
377  */
378 int
379 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
380 {
381 	int error;
382 #if defined(INET) || defined(INET6)
383 	struct ip_fw *rule = ip_dn_claim_rule(m);
384 
385 	if (IPFW_LOADED && ether_ipfw != 0) {
386 		if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) {
387 			if (m) {
388 				m_freem(m);
389 				return EACCES;	/* pkt dropped */
390 			} else
391 				return 0;	/* consumed e.g. in a pipe */
392 		}
393 	}
394 #endif
395 
396 	/*
397 	 * Queue message on interface, update output statistics if
398 	 * successful, and start output if interface not yet active.
399 	 */
400 	IFQ_HANDOFF(ifp, m, error);
401 	return (error);
402 }
403 
404 #if defined(INET) || defined(INET6)
405 /*
406  * ipfw processing for ethernet packets (in and out).
407  * The second parameter is NULL from ether_demux, and ifp from
408  * ether_output_frame.
409  */
410 int
411 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
412 	struct ip_fw **rule, int shared)
413 {
414 	struct ether_header *eh;
415 	struct ether_header save_eh;
416 	struct mbuf *m;
417 	int i;
418 	struct ip_fw_args args;
419 
420 	if (*rule != NULL && fw_one_pass)
421 		return 1; /* dummynet packet, already partially processed */
422 
423 	/*
424 	 * I need some amt of data to be contiguous, and in case others need
425 	 * the packet (shared==1) also better be in the first mbuf.
426 	 */
427 	m = *m0;
428 	i = min( m->m_pkthdr.len, max_protohdr);
429 	if ( shared || m->m_len < i) {
430 		m = m_pullup(m, i);
431 		if (m == NULL) {
432 			*m0 = m;
433 			return 0;
434 		}
435 	}
436 	eh = mtod(m, struct ether_header *);
437 	save_eh = *eh;			/* save copy for restore below */
438 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
439 
440 	args.m = m;		/* the packet we are looking at		*/
441 	args.oif = dst;		/* destination, if any			*/
442 	args.rule = *rule;	/* matching rule to restart		*/
443 	args.next_hop = NULL;	/* we do not support forward yet	*/
444 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
445 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
446 	i = ip_fw_chk_ptr(&args);
447 	m = args.m;
448 	if (m != NULL) {
449 		/*
450 		 * Restore Ethernet header, as needed, in case the
451 		 * mbuf chain was replaced by ipfw.
452 		 */
453 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
454 		if (m == NULL) {
455 			*m0 = m;
456 			return 0;
457 		}
458 		if (eh != mtod(m, struct ether_header *))
459 			bcopy(&save_eh, mtod(m, struct ether_header *),
460 				ETHER_HDR_LEN);
461 	}
462 	*m0 = m;
463 	*rule = args.rule;
464 
465 	if (i == IP_FW_DENY) /* drop */
466 		return 0;
467 
468 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
469 
470 	if (i == IP_FW_PASS) /* a PASS rule.  */
471 		return 1;
472 
473 	if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
474 		/*
475 		 * Pass the pkt to dummynet, which consumes it.
476 		 * If shared, make a copy and keep the original.
477 		 */
478 		if (shared) {
479 			m = m_copypacket(m, M_DONTWAIT);
480 			if (m == NULL)
481 				return 0;
482 		} else {
483 			/*
484 			 * Pass the original to dummynet and
485 			 * nothing back to the caller
486 			 */
487 			*m0 = NULL ;
488 		}
489 		ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
490 		return 0;
491 	}
492 	/*
493 	 * XXX at some point add support for divert/forward actions.
494 	 * If none of the above matches, we have to drop the pkt.
495 	 */
496 	return 0;
497 }
498 #endif
499 
500 /*
501  * Process a received Ethernet packet; the packet is in the
502  * mbuf chain m with the ethernet header at the front.
503  */
504 static void
505 ether_input(struct ifnet *ifp, struct mbuf *m)
506 {
507 	struct ether_header *eh;
508 	u_short etype;
509 
510 	if ((ifp->if_flags & IFF_UP) == 0) {
511 		m_freem(m);
512 		return;
513 	}
514 #ifdef DIAGNOSTIC
515 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
516 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
517 		m_freem(m);
518 		return;
519 	}
520 #endif
521 	/*
522 	 * Do consistency checks to verify assumptions
523 	 * made by code past this point.
524 	 */
525 	if ((m->m_flags & M_PKTHDR) == 0) {
526 		if_printf(ifp, "discard frame w/o packet header\n");
527 		ifp->if_ierrors++;
528 		m_freem(m);
529 		return;
530 	}
531 	if (m->m_len < ETHER_HDR_LEN) {
532 		/* XXX maybe should pullup? */
533 		if_printf(ifp, "discard frame w/o leading ethernet "
534 				"header (len %u pkt len %u)\n",
535 				m->m_len, m->m_pkthdr.len);
536 		ifp->if_ierrors++;
537 		m_freem(m);
538 		return;
539 	}
540 	eh = mtod(m, struct ether_header *);
541 	etype = ntohs(eh->ether_type);
542 	if (m->m_pkthdr.len >
543 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
544 		if_printf(ifp, "discard oversize frame "
545 				"(ether type %x flags %x len %u > max %lu)\n",
546 				etype, m->m_flags, m->m_pkthdr.len,
547 				ETHER_MAX_FRAME(ifp, etype,
548 						m->m_flags & M_HASFCS));
549 		ifp->if_ierrors++;
550 		m_freem(m);
551 		return;
552 	}
553 	if (m->m_pkthdr.rcvif == NULL) {
554 		if_printf(ifp, "discard frame w/o interface pointer\n");
555 		ifp->if_ierrors++;
556 		m_freem(m);
557 		return;
558 	}
559 #ifdef DIAGNOSTIC
560 	if (m->m_pkthdr.rcvif != ifp) {
561 		if_printf(ifp, "Warning, frame marked as received on %s\n",
562 			m->m_pkthdr.rcvif->if_xname);
563 	}
564 #endif
565 
566 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
567 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
568 			m->m_flags |= M_BCAST;
569 		else
570 			m->m_flags |= M_MCAST;
571 		ifp->if_imcasts++;
572 	}
573 
574 #ifdef MAC
575 	/*
576 	 * Tag the mbuf with an appropriate MAC label before any other
577 	 * consumers can get to it.
578 	 */
579 	mac_create_mbuf_from_ifnet(ifp, m);
580 #endif
581 
582 	/*
583 	 * Give bpf a chance at the packet.
584 	 */
585 	ETHER_BPF_MTAP(ifp, m);
586 
587 	/*
588 	 * If the CRC is still on the packet, trim it off. We do this once
589 	 * and once only in case we are re-entered. Nothing else on the
590 	 * Ethernet receive path expects to see the FCS.
591 	 */
592 	if (m->m_flags & M_HASFCS) {
593 		m_adj(m, -ETHER_CRC_LEN);
594 		m->m_flags &= ~M_HASFCS;
595 	}
596 
597 	ifp->if_ibytes += m->m_pkthdr.len;
598 
599 	/* Allow monitor mode to claim this frame, after stats are updated. */
600 	if (ifp->if_flags & IFF_MONITOR) {
601 		m_freem(m);
602 		return;
603 	}
604 
605 	/*
606 	 * If the hardware did not process an 802.1Q tag, do this now,
607 	 * to allow 802.1P priority frames to be passed to the main input
608 	 * path correctly.
609 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
610 	 */
611 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
612 		struct ether_vlan_header *evl;
613 
614 		if (m->m_len < sizeof(*evl) &&
615 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
616 #ifdef DIAGNOSTIC
617 			if_printf(ifp, "cannot pullup VLAN header\n");
618 #endif
619 			ifp->if_ierrors++;
620 			m_freem(m);
621 			return;
622 		}
623 
624 		evl = mtod(m, struct ether_vlan_header *);
625 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
626 		m->m_flags |= M_VLANTAG;
627 
628 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
629 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
630 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
631 	}
632 
633 	/* Allow ng_ether(4) to claim this frame. */
634 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
635 		KASSERT(ng_ether_input_p != NULL,
636 		    ("%s: ng_ether_input_p is NULL", __func__));
637 		m->m_flags &= ~M_PROMISC;
638 		(*ng_ether_input_p)(ifp, &m);
639 		if (m == NULL)
640 			return;
641 	}
642 
643 	/*
644 	 * Allow if_bridge(4) to claim this frame.
645 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
646 	 * and the frame should be delivered locally.
647 	 */
648 	if (ifp->if_bridge != NULL) {
649 		m->m_flags &= ~M_PROMISC;
650 		BRIDGE_INPUT(ifp, m);
651 		if (m == NULL)
652 			return;
653 	}
654 
655 #ifdef DEV_CARP
656 	/*
657 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
658 	 * mbuf flows up to Layer 3.
659 	 * FreeBSD's implementation of carp(4) uses the inprotosw
660 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
661 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
662 	 * is outside the scope of the M_PROMISC test below.
663 	 * TODO: Maintain a hash table of ethernet addresses other than
664 	 * ether_dhost which may be active on this ifp.
665 	 */
666 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) {
667 		m->m_flags &= ~M_PROMISC;
668 	} else
669 #endif
670 	{
671 		/*
672 		 * If the frame received was not for our MAC address, set the
673 		 * M_PROMISC flag on the mbuf chain. The frame may need to
674 		 * be seen by the rest of the Ethernet input path in case of
675 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
676 		 * seen by upper protocol layers.
677 		 */
678 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
679 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
680 			m->m_flags |= M_PROMISC;
681 	}
682 
683 	/* First chunk of an mbuf contains good entropy */
684 	if (harvest.ethernet)
685 		random_harvest(m, 16, 3, 0, RANDOM_NET);
686 
687 	ether_demux(ifp, m);
688 }
689 
690 /*
691  * Upper layer processing for a received Ethernet packet.
692  */
693 void
694 ether_demux(struct ifnet *ifp, struct mbuf *m)
695 {
696 	struct ether_header *eh;
697 	int isr;
698 	u_short ether_type;
699 #if defined(NETATALK)
700 	struct llc *l;
701 #endif
702 
703 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
704 
705 #if defined(INET) || defined(INET6)
706 	/*
707 	 * Allow dummynet and/or ipfw to claim the frame.
708 	 * Do not do this for PROMISC frames in case we are re-entered.
709 	 */
710 	if (IPFW_LOADED && ether_ipfw != 0 && !(m->m_flags & M_PROMISC)) {
711 		struct ip_fw *rule = ip_dn_claim_rule(m);
712 
713 		if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) {
714 			if (m)
715 				m_freem(m);	/* dropped; free mbuf chain */
716 			return;			/* consumed */
717 		}
718 	}
719 #endif
720 	eh = mtod(m, struct ether_header *);
721 	ether_type = ntohs(eh->ether_type);
722 
723 	/*
724 	 * If this frame has a VLAN tag other than 0, call vlan_input()
725 	 * if its module is loaded. Otherwise, drop.
726 	 */
727 	if ((m->m_flags & M_VLANTAG) &&
728 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
729 		if (ifp->if_vlantrunk == NULL) {
730 			ifp->if_noproto++;
731 			m_freem(m);
732 			return;
733 		}
734 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
735 		    __func__));
736 		/* Clear before possibly re-entering ether_input(). */
737 		m->m_flags &= ~M_PROMISC;
738 		(*vlan_input_p)(ifp, m);
739 		return;
740 	}
741 
742 	/*
743 	 * Pass promiscuously received frames to the upper layer if the user
744 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
745 	 */
746 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
747 		m_freem(m);
748 		return;
749 	}
750 
751 	/*
752 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
753 	 * Strip off Ethernet header.
754 	 */
755 	m->m_flags &= ~M_VLANTAG;
756 	m->m_flags &= ~(M_PROTOFLAGS);
757 	m_adj(m, ETHER_HDR_LEN);
758 
759 	/*
760 	 * Dispatch frame to upper layer.
761 	 */
762 	switch (ether_type) {
763 #ifdef INET
764 	case ETHERTYPE_IP:
765 		if ((m = ip_fastforward(m)) == NULL)
766 			return;
767 		isr = NETISR_IP;
768 		break;
769 
770 	case ETHERTYPE_ARP:
771 		if (ifp->if_flags & IFF_NOARP) {
772 			/* Discard packet if ARP is disabled on interface */
773 			m_freem(m);
774 			return;
775 		}
776 		isr = NETISR_ARP;
777 		break;
778 #endif
779 #ifdef IPX
780 	case ETHERTYPE_IPX:
781 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
782 			return;
783 		isr = NETISR_IPX;
784 		break;
785 #endif
786 #ifdef INET6
787 	case ETHERTYPE_IPV6:
788 		isr = NETISR_IPV6;
789 		break;
790 #endif
791 #ifdef NETATALK
792 	case ETHERTYPE_AT:
793 		isr = NETISR_ATALK1;
794 		break;
795 	case ETHERTYPE_AARP:
796 		isr = NETISR_AARP;
797 		break;
798 #endif /* NETATALK */
799 	default:
800 #ifdef IPX
801 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
802 			return;
803 #endif /* IPX */
804 #if defined(NETATALK)
805 		if (ether_type > ETHERMTU)
806 			goto discard;
807 		l = mtod(m, struct llc *);
808 		if (l->llc_dsap == LLC_SNAP_LSAP &&
809 		    l->llc_ssap == LLC_SNAP_LSAP &&
810 		    l->llc_control == LLC_UI) {
811 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
812 			    sizeof(at_org_code)) == 0 &&
813 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
814 				m_adj(m, LLC_SNAPFRAMELEN);
815 				isr = NETISR_ATALK2;
816 				break;
817 			}
818 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
819 			    sizeof(aarp_org_code)) == 0 &&
820 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
821 				m_adj(m, LLC_SNAPFRAMELEN);
822 				isr = NETISR_AARP;
823 				break;
824 			}
825 		}
826 #endif /* NETATALK */
827 		goto discard;
828 	}
829 	netisr_dispatch(isr, m);
830 	return;
831 
832 discard:
833 	/*
834 	 * Packet is to be discarded.  If netgraph is present,
835 	 * hand the packet to it for last chance processing;
836 	 * otherwise dispose of it.
837 	 */
838 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
839 		KASSERT(ng_ether_input_orphan_p != NULL,
840 		    ("ng_ether_input_orphan_p is NULL"));
841 		/*
842 		 * Put back the ethernet header so netgraph has a
843 		 * consistent view of inbound packets.
844 		 */
845 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
846 		(*ng_ether_input_orphan_p)(ifp, m);
847 		return;
848 	}
849 	m_freem(m);
850 }
851 
852 /*
853  * Convert Ethernet address to printable (loggable) representation.
854  * This routine is for compatibility; it's better to just use
855  *
856  *	printf("%6D", <pointer to address>, ":");
857  *
858  * since there's no static buffer involved.
859  */
860 char *
861 ether_sprintf(const u_char *ap)
862 {
863 	static char etherbuf[18];
864 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
865 	return (etherbuf);
866 }
867 
868 /*
869  * Perform common duties while attaching to interface list
870  */
871 void
872 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
873 {
874 	int i;
875 	struct ifaddr *ifa;
876 	struct sockaddr_dl *sdl;
877 
878 	ifp->if_addrlen = ETHER_ADDR_LEN;
879 	ifp->if_hdrlen = ETHER_HDR_LEN;
880 	if_attach(ifp);
881 	ifp->if_mtu = ETHERMTU;
882 	ifp->if_output = ether_output;
883 	ifp->if_input = ether_input;
884 	ifp->if_resolvemulti = ether_resolvemulti;
885 	if (ifp->if_baudrate == 0)
886 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
887 	ifp->if_broadcastaddr = etherbroadcastaddr;
888 
889 	ifa = ifp->if_addr;
890 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
891 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
892 	sdl->sdl_type = IFT_ETHER;
893 	sdl->sdl_alen = ifp->if_addrlen;
894 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
895 
896 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
897 	if (ng_ether_attach_p != NULL)
898 		(*ng_ether_attach_p)(ifp);
899 
900 	/* Announce Ethernet MAC address if non-zero. */
901 	for (i = 0; i < ifp->if_addrlen; i++)
902 		if (lla[i] != 0)
903 			break;
904 	if (i != ifp->if_addrlen)
905 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
906 	if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0)
907 		if_printf(ifp, "if_start running deferred for Giant\n");
908 }
909 
910 /*
911  * Perform common duties while detaching an Ethernet interface
912  */
913 void
914 ether_ifdetach(struct ifnet *ifp)
915 {
916 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
917 		KASSERT(ng_ether_detach_p != NULL,
918 		    ("ng_ether_detach_p is NULL"));
919 		(*ng_ether_detach_p)(ifp);
920 	}
921 
922 	bpfdetach(ifp);
923 	if_detach(ifp);
924 }
925 
926 SYSCTL_DECL(_net_link);
927 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
928 #if defined(INET) || defined(INET6)
929 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
930 	    &ether_ipfw,0,"Pass ether pkts through firewall");
931 #endif
932 
933 #if 0
934 /*
935  * This is for reference.  We have a table-driven version
936  * of the little-endian crc32 generator, which is faster
937  * than the double-loop.
938  */
939 uint32_t
940 ether_crc32_le(const uint8_t *buf, size_t len)
941 {
942 	size_t i;
943 	uint32_t crc;
944 	int bit;
945 	uint8_t data;
946 
947 	crc = 0xffffffff;	/* initial value */
948 
949 	for (i = 0; i < len; i++) {
950 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1)
951 			carry = (crc ^ data) & 1;
952 			crc >>= 1;
953 			if (carry)
954 				crc = (crc ^ ETHER_CRC_POLY_LE);
955 	}
956 
957 	return (crc);
958 }
959 #else
960 uint32_t
961 ether_crc32_le(const uint8_t *buf, size_t len)
962 {
963 	static const uint32_t crctab[] = {
964 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
965 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
966 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
967 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
968 	};
969 	size_t i;
970 	uint32_t crc;
971 
972 	crc = 0xffffffff;	/* initial value */
973 
974 	for (i = 0; i < len; i++) {
975 		crc ^= buf[i];
976 		crc = (crc >> 4) ^ crctab[crc & 0xf];
977 		crc = (crc >> 4) ^ crctab[crc & 0xf];
978 	}
979 
980 	return (crc);
981 }
982 #endif
983 
984 uint32_t
985 ether_crc32_be(const uint8_t *buf, size_t len)
986 {
987 	size_t i;
988 	uint32_t crc, carry;
989 	int bit;
990 	uint8_t data;
991 
992 	crc = 0xffffffff;	/* initial value */
993 
994 	for (i = 0; i < len; i++) {
995 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
996 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
997 			crc <<= 1;
998 			if (carry)
999 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1000 		}
1001 	}
1002 
1003 	return (crc);
1004 }
1005 
1006 int
1007 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
1008 {
1009 	struct ifaddr *ifa = (struct ifaddr *) data;
1010 	struct ifreq *ifr = (struct ifreq *) data;
1011 	int error = 0;
1012 
1013 	switch (command) {
1014 	case SIOCSIFADDR:
1015 		ifp->if_flags |= IFF_UP;
1016 
1017 		switch (ifa->ifa_addr->sa_family) {
1018 #ifdef INET
1019 		case AF_INET:
1020 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1021 			arp_ifinit(ifp, ifa);
1022 			break;
1023 #endif
1024 #ifdef IPX
1025 		/*
1026 		 * XXX - This code is probably wrong
1027 		 */
1028 		case AF_IPX:
1029 			{
1030 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1031 
1032 			if (ipx_nullhost(*ina))
1033 				ina->x_host =
1034 				    *(union ipx_host *)
1035 				    IF_LLADDR(ifp);
1036 			else {
1037 				bcopy((caddr_t) ina->x_host.c_host,
1038 				      (caddr_t) IF_LLADDR(ifp),
1039 				      ETHER_ADDR_LEN);
1040 			}
1041 
1042 			/*
1043 			 * Set new address
1044 			 */
1045 			ifp->if_init(ifp->if_softc);
1046 			break;
1047 			}
1048 #endif
1049 		default:
1050 			ifp->if_init(ifp->if_softc);
1051 			break;
1052 		}
1053 		break;
1054 
1055 	case SIOCGIFADDR:
1056 		{
1057 			struct sockaddr *sa;
1058 
1059 			sa = (struct sockaddr *) & ifr->ifr_data;
1060 			bcopy(IF_LLADDR(ifp),
1061 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1062 		}
1063 		break;
1064 
1065 	case SIOCSIFMTU:
1066 		/*
1067 		 * Set the interface MTU.
1068 		 */
1069 		if (ifr->ifr_mtu > ETHERMTU) {
1070 			error = EINVAL;
1071 		} else {
1072 			ifp->if_mtu = ifr->ifr_mtu;
1073 		}
1074 		break;
1075 	default:
1076 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1077 		break;
1078 	}
1079 	return (error);
1080 }
1081 
1082 static int
1083 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1084 	struct sockaddr *sa)
1085 {
1086 	struct sockaddr_dl *sdl;
1087 #ifdef INET
1088 	struct sockaddr_in *sin;
1089 #endif
1090 #ifdef INET6
1091 	struct sockaddr_in6 *sin6;
1092 #endif
1093 	u_char *e_addr;
1094 
1095 	switch(sa->sa_family) {
1096 	case AF_LINK:
1097 		/*
1098 		 * No mapping needed. Just check that it's a valid MC address.
1099 		 */
1100 		sdl = (struct sockaddr_dl *)sa;
1101 		e_addr = LLADDR(sdl);
1102 		if (!ETHER_IS_MULTICAST(e_addr))
1103 			return EADDRNOTAVAIL;
1104 		*llsa = 0;
1105 		return 0;
1106 
1107 #ifdef INET
1108 	case AF_INET:
1109 		sin = (struct sockaddr_in *)sa;
1110 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1111 			return EADDRNOTAVAIL;
1112 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1113 		       M_NOWAIT|M_ZERO);
1114 		if (sdl == NULL)
1115 			return ENOMEM;
1116 		sdl->sdl_len = sizeof *sdl;
1117 		sdl->sdl_family = AF_LINK;
1118 		sdl->sdl_index = ifp->if_index;
1119 		sdl->sdl_type = IFT_ETHER;
1120 		sdl->sdl_alen = ETHER_ADDR_LEN;
1121 		e_addr = LLADDR(sdl);
1122 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1123 		*llsa = (struct sockaddr *)sdl;
1124 		return 0;
1125 #endif
1126 #ifdef INET6
1127 	case AF_INET6:
1128 		sin6 = (struct sockaddr_in6 *)sa;
1129 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1130 			/*
1131 			 * An IP6 address of 0 means listen to all
1132 			 * of the Ethernet multicast address used for IP6.
1133 			 * (This is used for multicast routers.)
1134 			 */
1135 			ifp->if_flags |= IFF_ALLMULTI;
1136 			*llsa = 0;
1137 			return 0;
1138 		}
1139 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1140 			return EADDRNOTAVAIL;
1141 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1142 		       M_NOWAIT|M_ZERO);
1143 		if (sdl == NULL)
1144 			return (ENOMEM);
1145 		sdl->sdl_len = sizeof *sdl;
1146 		sdl->sdl_family = AF_LINK;
1147 		sdl->sdl_index = ifp->if_index;
1148 		sdl->sdl_type = IFT_ETHER;
1149 		sdl->sdl_alen = ETHER_ADDR_LEN;
1150 		e_addr = LLADDR(sdl);
1151 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1152 		*llsa = (struct sockaddr *)sdl;
1153 		return 0;
1154 #endif
1155 
1156 	default:
1157 		/*
1158 		 * Well, the text isn't quite right, but it's the name
1159 		 * that counts...
1160 		 */
1161 		return EAFNOSUPPORT;
1162 	}
1163 }
1164 
1165 static void*
1166 ether_alloc(u_char type, struct ifnet *ifp)
1167 {
1168 	struct arpcom	*ac;
1169 
1170 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1171 	ac->ac_ifp = ifp;
1172 
1173 	return (ac);
1174 }
1175 
1176 static void
1177 ether_free(void *com, u_char type)
1178 {
1179 
1180 	free(com, M_ARPCOM);
1181 }
1182 
1183 static int
1184 ether_modevent(module_t mod, int type, void *data)
1185 {
1186 
1187 	switch (type) {
1188 	case MOD_LOAD:
1189 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1190 		break;
1191 	case MOD_UNLOAD:
1192 		if_deregister_com_alloc(IFT_ETHER);
1193 		break;
1194 	default:
1195 		return EOPNOTSUPP;
1196 	}
1197 
1198 	return (0);
1199 }
1200 
1201 static moduledata_t ether_mod = {
1202 	"ether",
1203 	ether_modevent,
1204 	0
1205 };
1206 
1207 void
1208 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1209 {
1210 	struct ether_vlan_header vlan;
1211 	struct mbuf mv, mb;
1212 
1213 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1214 	    ("%s: vlan information not present", __func__));
1215 	KASSERT(m->m_len >= sizeof(struct ether_header),
1216 	    ("%s: mbuf not large enough for header", __func__));
1217 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1218 	vlan.evl_proto = vlan.evl_encap_proto;
1219 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1220 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1221 	m->m_len -= sizeof(struct ether_header);
1222 	m->m_data += sizeof(struct ether_header);
1223 	/*
1224 	 * If a data link has been supplied by the caller, then we will need to
1225 	 * re-create a stack allocated mbuf chain with the following structure:
1226 	 *
1227 	 * (1) mbuf #1 will contain the supplied data link
1228 	 * (2) mbuf #2 will contain the vlan header
1229 	 * (3) mbuf #3 will contain the original mbuf's packet data
1230 	 *
1231 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1232 	 */
1233 	if (data != NULL) {
1234 		mv.m_next = m;
1235 		mv.m_data = (caddr_t)&vlan;
1236 		mv.m_len = sizeof(vlan);
1237 		mb.m_next = &mv;
1238 		mb.m_data = data;
1239 		mb.m_len = dlen;
1240 		bpf_mtap(bp, &mb);
1241 	} else
1242 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1243 	m->m_len += sizeof(struct ether_header);
1244 	m->m_data -= sizeof(struct ether_header);
1245 }
1246 
1247 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1248 MODULE_VERSION(ether, 1);
1249