xref: /freebsd/sys/net/if_ethersubr.c (revision 3642298923e528d795e3a30ec165d2b469e28b40)
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_bdg.h"
38 #include "opt_mac.h"
39 #include "opt_netgraph.h"
40 #include "opt_carp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/mac.h>
46 #include <sys/malloc.h>
47 #include <sys/module.h>
48 #include <sys/mbuf.h>
49 #include <sys/random.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 
54 #include <net/if.h>
55 #include <net/if_arp.h>
56 #include <net/netisr.h>
57 #include <net/route.h>
58 #include <net/if_llc.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/bpf.h>
62 #include <net/ethernet.h>
63 #include <net/bridge.h>
64 #include <net/if_vlan_var.h>
65 
66 #if defined(INET) || defined(INET6)
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/if_ether.h>
70 #include <netinet/ip_fw.h>
71 #include <netinet/ip_dummynet.h>
72 #endif
73 #ifdef INET6
74 #include <netinet6/nd6.h>
75 #endif
76 
77 #ifdef DEV_CARP
78 #include <netinet/ip_carp.h>
79 #endif
80 
81 #ifdef IPX
82 #include <netipx/ipx.h>
83 #include <netipx/ipx_if.h>
84 #endif
85 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
86 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
87 		struct sockaddr *dst, short *tp, int *hlen);
88 
89 #ifdef NETATALK
90 #include <netatalk/at.h>
91 #include <netatalk/at_var.h>
92 #include <netatalk/at_extern.h>
93 
94 #define llc_snap_org_code llc_un.type_snap.org_code
95 #define llc_snap_ether_type llc_un.type_snap.ether_type
96 
97 extern u_char	at_org_code[3];
98 extern u_char	aarp_org_code[3];
99 #endif /* NETATALK */
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 /* bridge support */
111 int do_bridge;
112 bridge_in_t *bridge_in_ptr;
113 bdg_forward_t *bdg_forward_ptr;
114 bdgtakeifaces_t *bdgtakeifaces_ptr;
115 struct bdg_softc *ifp2sc;
116 
117 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
118 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
119 		struct sockaddr *, struct rtentry *);
120 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
121 
122 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
123 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
124 
125 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
126 		struct sockaddr *);
127 
128 /* XXX: should be in an arp support file, not here */
129 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
130 
131 #define senderr(e) do { error = (e); goto bad;} while (0)
132 
133 #if defined(INET) || defined(INET6)
134 int
135 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
136 	struct ip_fw **rule, int shared);
137 static int ether_ipfw;
138 #endif
139 
140 /*
141  * Ethernet output routine.
142  * Encapsulate a packet of type family for the local net.
143  * Use trailer local net encapsulation if enough data in first
144  * packet leaves a multiple of 512 bytes of data in remainder.
145  * Assumes that ifp is actually pointer to arpcom structure.
146  */
147 int
148 ether_output(struct ifnet *ifp, struct mbuf *m,
149 	struct sockaddr *dst, struct rtentry *rt0)
150 {
151 	short type;
152 	int error, hdrcmplt = 0;
153 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
154 	struct ether_header *eh;
155 	int loop_copy = 0;
156 	int hlen;	/* link layer header length */
157 
158 #ifdef MAC
159 	error = mac_check_ifnet_transmit(ifp, m);
160 	if (error)
161 		senderr(error);
162 #endif
163 
164 	if (ifp->if_flags & IFF_MONITOR)
165 		senderr(ENETDOWN);
166 	if (!((ifp->if_flags & IFF_UP) &&
167 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
168 		senderr(ENETDOWN);
169 
170 	hlen = ETHER_HDR_LEN;
171 	switch (dst->sa_family) {
172 #ifdef INET
173 	case AF_INET:
174 		error = arpresolve(ifp, rt0, m, dst, edst);
175 		if (error)
176 			return (error == EWOULDBLOCK ? 0 : error);
177 		type = htons(ETHERTYPE_IP);
178 		break;
179 	case AF_ARP:
180 	{
181 		struct arphdr *ah;
182 		ah = mtod(m, struct arphdr *);
183 		ah->ar_hrd = htons(ARPHRD_ETHER);
184 
185 		loop_copy = -1; /* if this is for us, don't do it */
186 
187 		switch(ntohs(ah->ar_op)) {
188 		case ARPOP_REVREQUEST:
189 		case ARPOP_REVREPLY:
190 			type = htons(ETHERTYPE_REVARP);
191 			break;
192 		case ARPOP_REQUEST:
193 		case ARPOP_REPLY:
194 		default:
195 			type = htons(ETHERTYPE_ARP);
196 			break;
197 		}
198 
199 		if (m->m_flags & M_BCAST)
200 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
201 		else
202 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
203 
204 	}
205 	break;
206 #endif
207 #ifdef INET6
208 	case AF_INET6:
209 		error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst);
210 		if (error)
211 			return error;
212 		type = htons(ETHERTYPE_IPV6);
213 		break;
214 #endif
215 #ifdef IPX
216 	case AF_IPX:
217 		if (ef_outputp) {
218 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
219 		    if (error)
220 			goto bad;
221 		} else
222 		    type = htons(ETHERTYPE_IPX);
223 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
224 		    (caddr_t)edst, sizeof (edst));
225 		break;
226 #endif
227 #ifdef NETATALK
228 	case AF_APPLETALK:
229 	  {
230 	    struct at_ifaddr *aa;
231 
232 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
233 		    senderr(EHOSTUNREACH); /* XXX */
234 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst))
235 		    return (0);
236 	    /*
237 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
238 	     */
239 	    if ( aa->aa_flags & AFA_PHASE2 ) {
240 		struct llc llc;
241 
242 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
243 		if (m == NULL)
244 			senderr(ENOBUFS);
245 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
246 		llc.llc_control = LLC_UI;
247 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
248 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
249 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
250 		type = htons(m->m_pkthdr.len);
251 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
252 	    } else {
253 		type = htons(ETHERTYPE_AT);
254 	    }
255 	    break;
256 	  }
257 #endif /* NETATALK */
258 
259 	case pseudo_AF_HDRCMPLT:
260 		hdrcmplt = 1;
261 		eh = (struct ether_header *)dst->sa_data;
262 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
263 		/* FALLTHROUGH */
264 
265 	case AF_UNSPEC:
266 		loop_copy = -1; /* if this is for us, don't do it */
267 		eh = (struct ether_header *)dst->sa_data;
268 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
269 		type = eh->ether_type;
270 		break;
271 
272 	default:
273 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
274 		senderr(EAFNOSUPPORT);
275 	}
276 
277 	/*
278 	 * Add local net header.  If no space in first mbuf,
279 	 * allocate another.
280 	 */
281 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
282 	if (m == NULL)
283 		senderr(ENOBUFS);
284 	eh = mtod(m, struct ether_header *);
285 	(void)memcpy(&eh->ether_type, &type,
286 		sizeof(eh->ether_type));
287 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
288 	if (hdrcmplt)
289 		(void)memcpy(eh->ether_shost, esrc,
290 			sizeof(eh->ether_shost));
291 	else
292 		(void)memcpy(eh->ether_shost, IFP2ENADDR(ifp),
293 			sizeof(eh->ether_shost));
294 
295        /*
296 	* Bridges require special output handling.
297 	*/
298 	if (ifp->if_bridge) {
299 		KASSERT(bridge_output_p != NULL,
300 		    ("ether_input: if_bridge not loaded!"));
301 		return ((*bridge_output_p)(ifp, m, NULL, NULL));
302 	}
303 
304 	/*
305 	 * If a simplex interface, and the packet is being sent to our
306 	 * Ethernet address or a broadcast address, loopback a copy.
307 	 * XXX To make a simplex device behave exactly like a duplex
308 	 * device, we should copy in the case of sending to our own
309 	 * ethernet address (thus letting the original actually appear
310 	 * on the wire). However, we don't do that here for security
311 	 * reasons and compatibility with the original behavior.
312 	 */
313 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1) &&
314 	    m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) {
315 		int csum_flags = 0;
316 
317 		if (m->m_pkthdr.csum_flags & CSUM_IP)
318 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
319 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
320 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
321 
322 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
323 			struct mbuf *n;
324 
325 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
326 				n->m_pkthdr.csum_flags |= csum_flags;
327 				if (csum_flags & CSUM_DATA_VALID)
328 					n->m_pkthdr.csum_data = 0xffff;
329 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
330 			} else
331 				ifp->if_iqdrops++;
332 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
333 				ETHER_ADDR_LEN) == 0) {
334 			m->m_pkthdr.csum_flags |= csum_flags;
335 			if (csum_flags & CSUM_DATA_VALID)
336 				m->m_pkthdr.csum_data = 0xffff;
337 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
338 			return (0);	/* XXX */
339 		}
340 	}
341 
342 #ifdef DEV_CARP
343 	if (ifp->if_carp &&
344 	    (error = carp_output(ifp, m, dst, NULL)))
345 		goto bad;
346 #endif
347 
348 	/* Handle ng_ether(4) processing, if any */
349 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
350 		KASSERT(ng_ether_output_p != NULL,
351 		    ("ng_ether_output_p is NULL"));
352 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
353 bad:			if (m != NULL)
354 				m_freem(m);
355 			return (error);
356 		}
357 		if (m == NULL)
358 			return (0);
359 	}
360 
361 	/* Continue with link-layer output */
362 	return ether_output_frame(ifp, m);
363 }
364 
365 /*
366  * Ethernet link layer output routine to send a raw frame to the device.
367  *
368  * This assumes that the 14 byte Ethernet header is present and contiguous
369  * in the first mbuf (if BRIDGE'ing).
370  */
371 int
372 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
373 {
374 #if defined(INET) || defined(INET6)
375 	struct ip_fw *rule = ip_dn_claim_rule(m);
376 #else
377 	void *rule = NULL;
378 #endif
379 	int error;
380 
381 	if (rule == NULL && BDG_ACTIVE(ifp)) {
382 		/*
383 		 * Beware, the bridge code notices the null rcvif and
384 		 * uses that identify that it's being called from
385 		 * ether_output as opposd to ether_input.  Yech.
386 		 */
387 		m->m_pkthdr.rcvif = NULL;
388 		m = bdg_forward_ptr(m, ifp);
389 		if (m != NULL)
390 			m_freem(m);
391 		return (0);
392 	}
393 #if defined(INET) || defined(INET6)
394 	if (IPFW_LOADED && ether_ipfw != 0) {
395 		if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) {
396 			if (m) {
397 				m_freem(m);
398 				return EACCES;	/* pkt dropped */
399 			} else
400 				return 0;	/* consumed e.g. in a pipe */
401 		}
402 	}
403 #endif
404 
405 	/*
406 	 * Queue message on interface, update output statistics if
407 	 * successful, and start output if interface not yet active.
408 	 */
409 	IFQ_HANDOFF(ifp, m, error);
410 	return (error);
411 }
412 
413 #if defined(INET) || defined(INET6)
414 /*
415  * ipfw processing for ethernet packets (in and out).
416  * The second parameter is NULL from ether_demux, and ifp from
417  * ether_output_frame. This section of code could be used from
418  * bridge.c as well as long as we use some extra info
419  * to distinguish that case from ether_output_frame();
420  */
421 int
422 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
423 	struct ip_fw **rule, int shared)
424 {
425 	struct ether_header *eh;
426 	struct ether_header save_eh;
427 	struct mbuf *m;
428 	int i;
429 	struct ip_fw_args args;
430 
431 	if (*rule != NULL && fw_one_pass)
432 		return 1; /* dummynet packet, already partially processed */
433 
434 	/*
435 	 * I need some amt of data to be contiguous, and in case others need
436 	 * the packet (shared==1) also better be in the first mbuf.
437 	 */
438 	m = *m0;
439 	i = min( m->m_pkthdr.len, max_protohdr);
440 	if ( shared || m->m_len < i) {
441 		m = m_pullup(m, i);
442 		if (m == NULL) {
443 			*m0 = m;
444 			return 0;
445 		}
446 	}
447 	eh = mtod(m, struct ether_header *);
448 	save_eh = *eh;			/* save copy for restore below */
449 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
450 
451 	args.m = m;		/* the packet we are looking at		*/
452 	args.oif = dst;		/* destination, if any			*/
453 	args.rule = *rule;	/* matching rule to restart		*/
454 	args.next_hop = NULL;	/* we do not support forward yet	*/
455 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
456 	i = ip_fw_chk_ptr(&args);
457 	m = args.m;
458 	if (m != NULL) {
459 		/*
460 		 * Restore Ethernet header, as needed, in case the
461 		 * mbuf chain was replaced by ipfw.
462 		 */
463 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
464 		if (m == NULL) {
465 			*m0 = m;
466 			return 0;
467 		}
468 		if (eh != mtod(m, struct ether_header *))
469 			bcopy(&save_eh, mtod(m, struct ether_header *),
470 				ETHER_HDR_LEN);
471 	}
472 	*m0 = m;
473 	*rule = args.rule;
474 
475 	if (i == IP_FW_DENY) /* drop */
476 		return 0;
477 
478 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
479 
480 	if (i == IP_FW_PASS) /* a PASS rule.  */
481 		return 1;
482 
483 	if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
484 		/*
485 		 * Pass the pkt to dummynet, which consumes it.
486 		 * If shared, make a copy and keep the original.
487 		 */
488 		if (shared) {
489 			m = m_copypacket(m, M_DONTWAIT);
490 			if (m == NULL)
491 				return 0;
492 		} else {
493 			/*
494 			 * Pass the original to dummynet and
495 			 * nothing back to the caller
496 			 */
497 			*m0 = NULL ;
498 		}
499 		ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
500 		return 0;
501 	}
502 	/*
503 	 * XXX at some point add support for divert/forward actions.
504 	 * If none of the above matches, we have to drop the pkt.
505 	 */
506 	return 0;
507 }
508 #endif
509 
510 /*
511  * Process a received Ethernet packet; the packet is in the
512  * mbuf chain m with the ethernet header at the front.
513  */
514 static void
515 ether_input(struct ifnet *ifp, struct mbuf *m)
516 {
517 	struct ether_header *eh;
518 	u_short etype;
519 
520 	/*
521 	 * Do consistency checks to verify assumptions
522 	 * made by code past this point.
523 	 */
524 	if ((m->m_flags & M_PKTHDR) == 0) {
525 		if_printf(ifp, "discard frame w/o packet header\n");
526 		ifp->if_ierrors++;
527 		m_freem(m);
528 		return;
529 	}
530 	if (m->m_len < ETHER_HDR_LEN) {
531 		/* XXX maybe should pullup? */
532 		if_printf(ifp, "discard frame w/o leading ethernet "
533 				"header (len %u pkt len %u)\n",
534 				m->m_len, m->m_pkthdr.len);
535 		ifp->if_ierrors++;
536 		m_freem(m);
537 		return;
538 	}
539 	eh = mtod(m, struct ether_header *);
540 	etype = ntohs(eh->ether_type);
541 	if (m->m_pkthdr.len >
542 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
543 		if_printf(ifp, "discard oversize frame "
544 				"(ether type %x flags %x len %u > max %lu)\n",
545 				etype, m->m_flags, m->m_pkthdr.len,
546 				ETHER_MAX_FRAME(ifp, etype,
547 						m->m_flags & M_HASFCS));
548 		ifp->if_ierrors++;
549 		m_freem(m);
550 		return;
551 	}
552 	if (m->m_pkthdr.rcvif == NULL) {
553 		if_printf(ifp, "discard frame w/o interface pointer\n");
554 		ifp->if_ierrors++;
555 		m_freem(m);
556 		return;
557 	}
558 #ifdef DIAGNOSTIC
559 	if (m->m_pkthdr.rcvif != ifp) {
560 		if_printf(ifp, "Warning, frame marked as received on %s\n",
561 			m->m_pkthdr.rcvif->if_xname);
562 	}
563 #endif
564 
565 #ifdef MAC
566 	/*
567 	 * Tag the mbuf with an appropriate MAC label before any other
568 	 * consumers can get to it.
569 	 */
570 	mac_create_mbuf_from_ifnet(ifp, m);
571 #endif
572 
573 	/*
574 	 * Give bpf a chance at the packet.
575 	 */
576 	BPF_MTAP(ifp, m);
577 
578 	if (ifp->if_flags & IFF_MONITOR) {
579 		/*
580 		 * Interface marked for monitoring; discard packet.
581 		 */
582 		m_freem(m);
583 		return;
584 	}
585 
586 	/* If the CRC is still on the packet, trim it off. */
587 	if (m->m_flags & M_HASFCS) {
588 		m_adj(m, -ETHER_CRC_LEN);
589 		m->m_flags &= ~M_HASFCS;
590 	}
591 
592 	ifp->if_ibytes += m->m_pkthdr.len;
593 
594 	/* Handle ng_ether(4) processing, if any */
595 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
596 		KASSERT(ng_ether_input_p != NULL,
597 		    ("ng_ether_input_p is NULL"));
598 		(*ng_ether_input_p)(ifp, &m);
599 		if (m == NULL)
600 			return;
601 	}
602 
603 	/*
604 	 * Tap the packet off here for a bridge.  bridge_input()
605 	 * will return NULL if it has consumed the packet, otherwise
606 	 * it gets processed as normal.  Note that bridge_input()
607 	 * will always return the original packet if we need to
608 	 * process it locally.
609 	 */
610 	if (ifp->if_bridge) {
611 		KASSERT(bridge_input_p != NULL,
612 		    ("ether_input: if_bridge not loaded!"));
613 
614 		/* Mark the packet as broadcast or multicast. This is also set
615 		 * further down the code in ether_demux() but since the bridge
616 		 * input routine rarely returns a mbuf for further processing,
617 		 * it is an acceptable duplication.
618 		 */
619 		if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
620 			if (bcmp(etherbroadcastaddr, eh->ether_dhost,
621 				sizeof(etherbroadcastaddr)) == 0)
622 				m->m_flags |= M_BCAST;
623 			else
624 				m->m_flags |= M_MCAST;
625 		}
626 
627 		m = (*bridge_input_p)(ifp, m);
628 		if (m == NULL)
629 			return;
630 		/*
631 		 * Bridge has determined that the packet is for us.
632 		 * Update our interface pointer -- we may have had
633 		 * to "bridge" the packet locally.
634 		 */
635 		ifp = m->m_pkthdr.rcvif;
636 	}
637 
638 	/* Check for bridging mode */
639 	if (BDG_ACTIVE(ifp) )
640 		if ((m = bridge_in_ptr(ifp, m)) == NULL)
641 			return;
642 
643 	/* First chunk of an mbuf contains good entropy */
644 	if (harvest.ethernet)
645 		random_harvest(m, 16, 3, 0, RANDOM_NET);
646 	ether_demux(ifp, m);
647 }
648 
649 /*
650  * Upper layer processing for a received Ethernet packet.
651  */
652 void
653 ether_demux(struct ifnet *ifp, struct mbuf *m)
654 {
655 	struct ether_header *eh;
656 	int isr;
657 	u_short ether_type;
658 #if defined(NETATALK)
659 	struct llc *l;
660 #endif
661 #if defined(INET) || defined(INET6)
662 	struct ip_fw *rule = ip_dn_claim_rule(m);
663 #endif
664 
665 	KASSERT(ifp != NULL, ("ether_demux: NULL interface pointer"));
666 
667 	eh = mtod(m, struct ether_header *);
668 	ether_type = ntohs(eh->ether_type);
669 
670 #if defined(INET) || defined(INET6)
671 	if (rule)	/* packet was already bridged */
672 		goto post_stats;
673 #endif
674 
675 	if (!(BDG_ACTIVE(ifp)) && !(ifp->if_bridge) &&
676 	    !((ether_type == ETHERTYPE_VLAN || m->m_flags & M_VLANTAG) &&
677 	    ifp->if_nvlans > 0)) {
678 #ifdef DEV_CARP
679 		/*
680 		 * XXX: Okay, we need to call carp_forus() and - if it is for
681 		 * us jump over code that does the normal check
682 		 * "IFP2ENADDR(ifp) == ether_dhost". The check sequence is a bit
683 		 * different from OpenBSD, so we jump over as few code as
684 		 * possible, to catch _all_ sanity checks. This needs
685 		 * evaluation, to see if the carp ether_dhost values break any
686 		 * of these checks!
687 		 */
688 		if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
689 			goto pre_stats;
690 #endif
691 		/*
692 		 * Discard packet if upper layers shouldn't see it because it
693 		 * was unicast to a different Ethernet address. If the driver
694 		 * is working properly, then this situation can only happen
695 		 * when the interface is in promiscuous mode.
696 		 *
697 		 * If VLANs are active, and this packet has a VLAN tag, do
698 		 * not drop it here but pass it on to the VLAN layer, to
699 		 * give them a chance to consider it as well (e. g. in case
700 		 * bridging is only active on a VLAN).  They will drop it if
701 		 * it's undesired.
702 		 */
703 		if ((ifp->if_flags & IFF_PROMISC) != 0
704 		    && !ETHER_IS_MULTICAST(eh->ether_dhost)
705 		    && bcmp(eh->ether_dhost,
706 		      IFP2ENADDR(ifp), ETHER_ADDR_LEN) != 0
707 		    && (ifp->if_flags & IFF_PPROMISC) == 0) {
708 			    m_freem(m);
709 			    return;
710 		}
711 	}
712 
713 #ifdef DEV_CARP
714 pre_stats:
715 #endif
716 	/* Discard packet if interface is not up */
717 	if ((ifp->if_flags & IFF_UP) == 0) {
718 		m_freem(m);
719 		return;
720 	}
721 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
722 		if (bcmp(etherbroadcastaddr, eh->ether_dhost,
723 		    sizeof(etherbroadcastaddr)) == 0)
724 			m->m_flags |= M_BCAST;
725 		else
726 			m->m_flags |= M_MCAST;
727 	}
728 	if (m->m_flags & (M_BCAST|M_MCAST))
729 		ifp->if_imcasts++;
730 
731 #if defined(INET) || defined(INET6)
732 post_stats:
733 	if (IPFW_LOADED && ether_ipfw != 0) {
734 		if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) {
735 			if (m)
736 				m_freem(m);
737 			return;
738 		}
739 	}
740 #endif
741 
742 	/*
743 	 * Check to see if the device performed the VLAN decapsulation and
744 	 * provided us with the tag.
745 	 */
746 	if (m->m_flags & M_VLANTAG) {
747 		/*
748 		 * If no VLANs are configured, drop.
749 		 */
750 		if (ifp->if_nvlans == 0) {
751 			ifp->if_noproto++;
752 			m_freem(m);
753 			return;
754 		}
755 		/*
756 		 * vlan_input() will either recursively call ether_input()
757 		 * or drop the packet.
758 		 */
759 		KASSERT(vlan_input_p != NULL,("ether_input: VLAN not loaded!"));
760 		(*vlan_input_p)(ifp, m);
761 		return;
762 	}
763 
764 	/*
765 	 * Handle protocols that expect to have the Ethernet header
766 	 * (and possibly FCS) intact.
767 	 */
768 	switch (ether_type) {
769 	case ETHERTYPE_VLAN:
770 		if (ifp->if_nvlans != 0) {
771 			KASSERT(vlan_input_p,("ether_input: VLAN not loaded!"));
772 			(*vlan_input_p)(ifp, m);
773 		} else {
774 			ifp->if_noproto++;
775 			m_freem(m);
776 		}
777 		return;
778 	}
779 
780 	/* Strip off Ethernet header. */
781 	m_adj(m, ETHER_HDR_LEN);
782 
783 	/* If the CRC is still on the packet, trim it off. */
784 	if (m->m_flags & M_HASFCS) {
785 		m_adj(m, -ETHER_CRC_LEN);
786 		m->m_flags &= ~M_HASFCS;
787 	}
788 
789 	switch (ether_type) {
790 #ifdef INET
791 	case ETHERTYPE_IP:
792 		if (ip_fastforward(m))
793 			return;
794 		isr = NETISR_IP;
795 		break;
796 
797 	case ETHERTYPE_ARP:
798 		if (ifp->if_flags & IFF_NOARP) {
799 			/* Discard packet if ARP is disabled on interface */
800 			m_freem(m);
801 			return;
802 		}
803 		isr = NETISR_ARP;
804 		break;
805 #endif
806 #ifdef IPX
807 	case ETHERTYPE_IPX:
808 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
809 			return;
810 		isr = NETISR_IPX;
811 		break;
812 #endif
813 #ifdef INET6
814 	case ETHERTYPE_IPV6:
815 		isr = NETISR_IPV6;
816 		break;
817 #endif
818 #ifdef NETATALK
819 	case ETHERTYPE_AT:
820 		isr = NETISR_ATALK1;
821 		break;
822 	case ETHERTYPE_AARP:
823 		isr = NETISR_AARP;
824 		break;
825 #endif /* NETATALK */
826 	default:
827 #ifdef IPX
828 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
829 			return;
830 #endif /* IPX */
831 #if defined(NETATALK)
832 		if (ether_type > ETHERMTU)
833 			goto discard;
834 		l = mtod(m, struct llc *);
835 		if (l->llc_dsap == LLC_SNAP_LSAP &&
836 		    l->llc_ssap == LLC_SNAP_LSAP &&
837 		    l->llc_control == LLC_UI) {
838 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
839 			    sizeof(at_org_code)) == 0 &&
840 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
841 				m_adj(m, LLC_SNAPFRAMELEN);
842 				isr = NETISR_ATALK2;
843 				break;
844 			}
845 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
846 			    sizeof(aarp_org_code)) == 0 &&
847 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
848 				m_adj(m, LLC_SNAPFRAMELEN);
849 				isr = NETISR_AARP;
850 				break;
851 			}
852 		}
853 #endif /* NETATALK */
854 		goto discard;
855 	}
856 	netisr_dispatch(isr, m);
857 	return;
858 
859 discard:
860 	/*
861 	 * Packet is to be discarded.  If netgraph is present,
862 	 * hand the packet to it for last chance processing;
863 	 * otherwise dispose of it.
864 	 */
865 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
866 		KASSERT(ng_ether_input_orphan_p != NULL,
867 		    ("ng_ether_input_orphan_p is NULL"));
868 		/*
869 		 * Put back the ethernet header so netgraph has a
870 		 * consistent view of inbound packets.
871 		 */
872 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
873 		(*ng_ether_input_orphan_p)(ifp, m);
874 		return;
875 	}
876 	m_freem(m);
877 }
878 
879 /*
880  * Convert Ethernet address to printable (loggable) representation.
881  * This routine is for compatibility; it's better to just use
882  *
883  *	printf("%6D", <pointer to address>, ":");
884  *
885  * since there's no static buffer involved.
886  */
887 char *
888 ether_sprintf(const u_char *ap)
889 {
890 	static char etherbuf[18];
891 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
892 	return (etherbuf);
893 }
894 
895 /*
896  * Perform common duties while attaching to interface list
897  */
898 void
899 ether_ifattach(struct ifnet *ifp, const u_int8_t *llc)
900 {
901 	int i;
902 	struct ifaddr *ifa;
903 	struct sockaddr_dl *sdl;
904 
905 	ifp->if_addrlen = ETHER_ADDR_LEN;
906 	ifp->if_hdrlen = ETHER_HDR_LEN;
907 	if_attach(ifp);
908 	ifp->if_mtu = ETHERMTU;
909 	ifp->if_output = ether_output;
910 	ifp->if_input = ether_input;
911 	ifp->if_resolvemulti = ether_resolvemulti;
912 	if (ifp->if_baudrate == 0)
913 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
914 	ifp->if_broadcastaddr = etherbroadcastaddr;
915 
916 	ifa = ifaddr_byindex(ifp->if_index);
917 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
918 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
919 	sdl->sdl_type = IFT_ETHER;
920 	sdl->sdl_alen = ifp->if_addrlen;
921 	bcopy(llc, LLADDR(sdl), ifp->if_addrlen);
922 	/*
923 	 * XXX: This doesn't belong here; we do it until
924 	 * XXX:  all drivers are cleaned up
925 	 */
926 	if (llc != IFP2ENADDR(ifp))
927 		bcopy(llc, IFP2ENADDR(ifp), ifp->if_addrlen);
928 
929 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
930 	if (ng_ether_attach_p != NULL)
931 		(*ng_ether_attach_p)(ifp);
932 	if (BDG_LOADED)
933 		bdgtakeifaces_ptr();
934 
935 	/* Announce Ethernet MAC address if non-zero. */
936 	for (i = 0; i < ifp->if_addrlen; i++)
937 		if (llc[i] != 0)
938 			break;
939 	if (i != ifp->if_addrlen)
940 		if_printf(ifp, "Ethernet address: %6D\n", llc, ":");
941 	if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0)
942 		if_printf(ifp, "if_start running deferred for Giant\n");
943 }
944 
945 /*
946  * Perform common duties while detaching an Ethernet interface
947  */
948 void
949 ether_ifdetach(struct ifnet *ifp)
950 {
951 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
952 		KASSERT(ng_ether_detach_p != NULL,
953 		    ("ng_ether_detach_p is NULL"));
954 		(*ng_ether_detach_p)(ifp);
955 	}
956 	bpfdetach(ifp);
957 	if_detach(ifp);
958 	if (BDG_LOADED)
959 		bdgtakeifaces_ptr();
960 }
961 
962 SYSCTL_DECL(_net_link);
963 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
964 #if defined(INET) || defined(INET6)
965 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
966 	    &ether_ipfw,0,"Pass ether pkts through firewall");
967 #endif
968 
969 #if 0
970 /*
971  * This is for reference.  We have a table-driven version
972  * of the little-endian crc32 generator, which is faster
973  * than the double-loop.
974  */
975 uint32_t
976 ether_crc32_le(const uint8_t *buf, size_t len)
977 {
978 	size_t i;
979 	uint32_t crc;
980 	int bit;
981 	uint8_t data;
982 
983 	crc = 0xffffffff;	/* initial value */
984 
985 	for (i = 0; i < len; i++) {
986 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1)
987 			carry = (crc ^ data) & 1;
988 			crc >>= 1;
989 			if (carry)
990 				crc = (crc ^ ETHER_CRC_POLY_LE);
991 	}
992 
993 	return (crc);
994 }
995 #else
996 uint32_t
997 ether_crc32_le(const uint8_t *buf, size_t len)
998 {
999 	static const uint32_t crctab[] = {
1000 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1001 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1002 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1003 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1004 	};
1005 	size_t i;
1006 	uint32_t crc;
1007 
1008 	crc = 0xffffffff;	/* initial value */
1009 
1010 	for (i = 0; i < len; i++) {
1011 		crc ^= buf[i];
1012 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1013 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1014 	}
1015 
1016 	return (crc);
1017 }
1018 #endif
1019 
1020 uint32_t
1021 ether_crc32_be(const uint8_t *buf, size_t len)
1022 {
1023 	size_t i;
1024 	uint32_t crc, carry;
1025 	int bit;
1026 	uint8_t data;
1027 
1028 	crc = 0xffffffff;	/* initial value */
1029 
1030 	for (i = 0; i < len; i++) {
1031 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1032 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1033 			crc <<= 1;
1034 			if (carry)
1035 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1036 		}
1037 	}
1038 
1039 	return (crc);
1040 }
1041 
1042 int
1043 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
1044 {
1045 	struct ifaddr *ifa = (struct ifaddr *) data;
1046 	struct ifreq *ifr = (struct ifreq *) data;
1047 	int error = 0;
1048 
1049 	switch (command) {
1050 	case SIOCSIFADDR:
1051 		ifp->if_flags |= IFF_UP;
1052 
1053 		switch (ifa->ifa_addr->sa_family) {
1054 #ifdef INET
1055 		case AF_INET:
1056 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1057 			arp_ifinit(ifp, ifa);
1058 			break;
1059 #endif
1060 #ifdef IPX
1061 		/*
1062 		 * XXX - This code is probably wrong
1063 		 */
1064 		case AF_IPX:
1065 			{
1066 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1067 
1068 			if (ipx_nullhost(*ina))
1069 				ina->x_host =
1070 				    *(union ipx_host *)
1071 				    IFP2ENADDR(ifp);
1072 			else {
1073 				bcopy((caddr_t) ina->x_host.c_host,
1074 				      (caddr_t) IFP2ENADDR(ifp),
1075 				      ETHER_ADDR_LEN);
1076 			}
1077 
1078 			/*
1079 			 * Set new address
1080 			 */
1081 			ifp->if_init(ifp->if_softc);
1082 			break;
1083 			}
1084 #endif
1085 		default:
1086 			ifp->if_init(ifp->if_softc);
1087 			break;
1088 		}
1089 		break;
1090 
1091 	case SIOCGIFADDR:
1092 		{
1093 			struct sockaddr *sa;
1094 
1095 			sa = (struct sockaddr *) & ifr->ifr_data;
1096 			bcopy(IFP2ENADDR(ifp),
1097 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1098 		}
1099 		break;
1100 
1101 	case SIOCSIFMTU:
1102 		/*
1103 		 * Set the interface MTU.
1104 		 */
1105 		if (ifr->ifr_mtu > ETHERMTU) {
1106 			error = EINVAL;
1107 		} else {
1108 			ifp->if_mtu = ifr->ifr_mtu;
1109 		}
1110 		break;
1111 	default:
1112 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1113 		break;
1114 	}
1115 	return (error);
1116 }
1117 
1118 static int
1119 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1120 	struct sockaddr *sa)
1121 {
1122 	struct sockaddr_dl *sdl;
1123 #ifdef INET
1124 	struct sockaddr_in *sin;
1125 #endif
1126 #ifdef INET6
1127 	struct sockaddr_in6 *sin6;
1128 #endif
1129 	u_char *e_addr;
1130 
1131 	switch(sa->sa_family) {
1132 	case AF_LINK:
1133 		/*
1134 		 * No mapping needed. Just check that it's a valid MC address.
1135 		 */
1136 		sdl = (struct sockaddr_dl *)sa;
1137 		e_addr = LLADDR(sdl);
1138 		if (!ETHER_IS_MULTICAST(e_addr))
1139 			return EADDRNOTAVAIL;
1140 		*llsa = 0;
1141 		return 0;
1142 
1143 #ifdef INET
1144 	case AF_INET:
1145 		sin = (struct sockaddr_in *)sa;
1146 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1147 			return EADDRNOTAVAIL;
1148 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1149 		       M_NOWAIT|M_ZERO);
1150 		if (sdl == NULL)
1151 			return ENOMEM;
1152 		sdl->sdl_len = sizeof *sdl;
1153 		sdl->sdl_family = AF_LINK;
1154 		sdl->sdl_index = ifp->if_index;
1155 		sdl->sdl_type = IFT_ETHER;
1156 		sdl->sdl_alen = ETHER_ADDR_LEN;
1157 		e_addr = LLADDR(sdl);
1158 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1159 		*llsa = (struct sockaddr *)sdl;
1160 		return 0;
1161 #endif
1162 #ifdef INET6
1163 	case AF_INET6:
1164 		sin6 = (struct sockaddr_in6 *)sa;
1165 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1166 			/*
1167 			 * An IP6 address of 0 means listen to all
1168 			 * of the Ethernet multicast address used for IP6.
1169 			 * (This is used for multicast routers.)
1170 			 */
1171 			ifp->if_flags |= IFF_ALLMULTI;
1172 			*llsa = 0;
1173 			return 0;
1174 		}
1175 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1176 			return EADDRNOTAVAIL;
1177 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1178 		       M_NOWAIT|M_ZERO);
1179 		if (sdl == NULL)
1180 			return (ENOMEM);
1181 		sdl->sdl_len = sizeof *sdl;
1182 		sdl->sdl_family = AF_LINK;
1183 		sdl->sdl_index = ifp->if_index;
1184 		sdl->sdl_type = IFT_ETHER;
1185 		sdl->sdl_alen = ETHER_ADDR_LEN;
1186 		e_addr = LLADDR(sdl);
1187 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1188 		*llsa = (struct sockaddr *)sdl;
1189 		return 0;
1190 #endif
1191 
1192 	default:
1193 		/*
1194 		 * Well, the text isn't quite right, but it's the name
1195 		 * that counts...
1196 		 */
1197 		return EAFNOSUPPORT;
1198 	}
1199 }
1200 
1201 static void*
1202 ether_alloc(u_char type, struct ifnet *ifp)
1203 {
1204 	struct arpcom	*ac;
1205 
1206 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1207 	ac->ac_ifp = ifp;
1208 
1209 	return (ac);
1210 }
1211 
1212 static void
1213 ether_free(void *com, u_char type)
1214 {
1215 
1216 	free(com, M_ARPCOM);
1217 }
1218 
1219 static int
1220 ether_modevent(module_t mod, int type, void *data)
1221 {
1222 
1223 	switch (type) {
1224 	case MOD_LOAD:
1225 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1226 		break;
1227 	case MOD_UNLOAD:
1228 		if_deregister_com_alloc(IFT_ETHER);
1229 		break;
1230 	default:
1231 		return EOPNOTSUPP;
1232 	}
1233 
1234 	return (0);
1235 }
1236 
1237 static moduledata_t ether_mod = {
1238 	"ether",
1239 	ether_modevent,
1240 	0
1241 };
1242 
1243 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1244 MODULE_VERSION(ether, 1);
1245