xref: /freebsd/sys/net/if_ethersubr.c (revision f856af0466c076beef4ea9b15d088e1119a945b8)
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 senderr(e) do { error = (e); goto bad;} while (0)
126 
127 #if defined(INET) || defined(INET6)
128 int
129 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
130 	struct ip_fw **rule, int shared);
131 static int ether_ipfw;
132 #endif
133 
134 /*
135  * Ethernet output routine.
136  * Encapsulate a packet of type family for the local net.
137  * Use trailer local net encapsulation if enough data in first
138  * packet leaves a multiple of 512 bytes of data in remainder.
139  */
140 int
141 ether_output(struct ifnet *ifp, struct mbuf *m,
142 	struct sockaddr *dst, struct rtentry *rt0)
143 {
144 	short type;
145 	int error, hdrcmplt = 0;
146 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
147 	struct ether_header *eh;
148 	int loop_copy = 1;
149 	int hlen;	/* link layer header length */
150 
151 #ifdef MAC
152 	error = mac_check_ifnet_transmit(ifp, m);
153 	if (error)
154 		senderr(error);
155 #endif
156 
157 	if (ifp->if_flags & IFF_MONITOR)
158 		senderr(ENETDOWN);
159 	if (!((ifp->if_flags & IFF_UP) &&
160 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
161 		senderr(ENETDOWN);
162 
163 	hlen = ETHER_HDR_LEN;
164 	switch (dst->sa_family) {
165 #ifdef INET
166 	case AF_INET:
167 		error = arpresolve(ifp, rt0, m, dst, edst);
168 		if (error)
169 			return (error == EWOULDBLOCK ? 0 : error);
170 		type = htons(ETHERTYPE_IP);
171 		break;
172 	case AF_ARP:
173 	{
174 		struct arphdr *ah;
175 		ah = mtod(m, struct arphdr *);
176 		ah->ar_hrd = htons(ARPHRD_ETHER);
177 
178 		loop_copy = 0; /* if this is for us, don't do it */
179 
180 		switch(ntohs(ah->ar_op)) {
181 		case ARPOP_REVREQUEST:
182 		case ARPOP_REVREPLY:
183 			type = htons(ETHERTYPE_REVARP);
184 			break;
185 		case ARPOP_REQUEST:
186 		case ARPOP_REPLY:
187 		default:
188 			type = htons(ETHERTYPE_ARP);
189 			break;
190 		}
191 
192 		if (m->m_flags & M_BCAST)
193 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
194 		else
195 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
196 
197 	}
198 	break;
199 #endif
200 #ifdef INET6
201 	case AF_INET6:
202 		error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst);
203 		if (error)
204 			return error;
205 		type = htons(ETHERTYPE_IPV6);
206 		break;
207 #endif
208 #ifdef IPX
209 	case AF_IPX:
210 		if (ef_outputp) {
211 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
212 		    if (error)
213 			goto bad;
214 		} else
215 		    type = htons(ETHERTYPE_IPX);
216 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
217 		    (caddr_t)edst, sizeof (edst));
218 		break;
219 #endif
220 #ifdef NETATALK
221 	case AF_APPLETALK:
222 	  {
223 	    struct at_ifaddr *aa;
224 
225 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
226 		    senderr(EHOSTUNREACH); /* XXX */
227 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst))
228 		    return (0);
229 	    /*
230 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
231 	     */
232 	    if ( aa->aa_flags & AFA_PHASE2 ) {
233 		struct llc llc;
234 
235 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
236 		if (m == NULL)
237 			senderr(ENOBUFS);
238 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
239 		llc.llc_control = LLC_UI;
240 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
241 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
242 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
243 		type = htons(m->m_pkthdr.len);
244 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
245 	    } else {
246 		type = htons(ETHERTYPE_AT);
247 	    }
248 	    break;
249 	  }
250 #endif /* NETATALK */
251 
252 	case pseudo_AF_HDRCMPLT:
253 		hdrcmplt = 1;
254 		eh = (struct ether_header *)dst->sa_data;
255 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
256 		/* FALLTHROUGH */
257 
258 	case AF_UNSPEC:
259 		loop_copy = 0; /* if this is for us, don't do it */
260 		eh = (struct ether_header *)dst->sa_data;
261 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
262 		type = eh->ether_type;
263 		break;
264 
265 	default:
266 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
267 		senderr(EAFNOSUPPORT);
268 	}
269 
270 	/*
271 	 * Add local net header.  If no space in first mbuf,
272 	 * allocate another.
273 	 */
274 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
275 	if (m == NULL)
276 		senderr(ENOBUFS);
277 	eh = mtod(m, struct ether_header *);
278 	(void)memcpy(&eh->ether_type, &type,
279 		sizeof(eh->ether_type));
280 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
281 	if (hdrcmplt)
282 		(void)memcpy(eh->ether_shost, esrc,
283 			sizeof(eh->ether_shost));
284 	else
285 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
286 			sizeof(eh->ether_shost));
287 
288 	/*
289 	 * If a simplex interface, and the packet is being sent to our
290 	 * Ethernet address or a broadcast address, loopback a copy.
291 	 * XXX To make a simplex device behave exactly like a duplex
292 	 * device, we should copy in the case of sending to our own
293 	 * ethernet address (thus letting the original actually appear
294 	 * on the wire). However, we don't do that here for security
295 	 * reasons and compatibility with the original behavior.
296 	 */
297 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
298 	    m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) {
299 		int csum_flags = 0;
300 
301 		if (m->m_pkthdr.csum_flags & CSUM_IP)
302 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
303 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
304 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
305 
306 		if (m->m_flags & M_BCAST) {
307 			struct mbuf *n;
308 
309 			/*
310 			 * Because if_simloop() modifies the packet, we need a
311 			 * writable copy through m_dup() instead of a readonly
312 			 * one as m_copy[m] would give us. The alternative would
313 			 * be to modify if_simloop() to handle the readonly mbuf,
314 			 * but performancewise it is mostly equivalent (trading
315 			 * extra data copying vs. extra locking).
316 			 *
317 			 * XXX This is a local workaround.  A number of less
318 			 * often used kernel parts suffer from the same bug.
319 			 * See PR kern/105943 for a proposed general solution.
320 			 */
321 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
322 				n->m_pkthdr.csum_flags |= csum_flags;
323 				if (csum_flags & CSUM_DATA_VALID)
324 					n->m_pkthdr.csum_data = 0xffff;
325 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
326 			} else
327 				ifp->if_iqdrops++;
328 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
329 				ETHER_ADDR_LEN) == 0) {
330 			m->m_pkthdr.csum_flags |= csum_flags;
331 			if (csum_flags & CSUM_DATA_VALID)
332 				m->m_pkthdr.csum_data = 0xffff;
333 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
334 			return (0);	/* XXX */
335 		}
336 	}
337 
338        /*
339 	* Bridges require special output handling.
340 	*/
341 	if (ifp->if_bridge) {
342 		BRIDGE_OUTPUT(ifp, m, error);
343 		return (error);
344 	}
345 
346 #ifdef DEV_CARP
347 	if (ifp->if_carp &&
348 	    (error = carp_output(ifp, m, dst, NULL)))
349 		goto bad;
350 #endif
351 
352 	/* Handle ng_ether(4) processing, if any */
353 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
354 		KASSERT(ng_ether_output_p != NULL,
355 		    ("ng_ether_output_p is NULL"));
356 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
357 bad:			if (m != NULL)
358 				m_freem(m);
359 			return (error);
360 		}
361 		if (m == NULL)
362 			return (0);
363 	}
364 
365 	/* Continue with link-layer output */
366 	return ether_output_frame(ifp, m);
367 }
368 
369 /*
370  * Ethernet link layer output routine to send a raw frame to the device.
371  *
372  * This assumes that the 14 byte Ethernet header is present and contiguous
373  * in the first mbuf (if BRIDGE'ing).
374  */
375 int
376 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
377 {
378 	int error;
379 #if defined(INET) || defined(INET6)
380 	struct ip_fw *rule = ip_dn_claim_rule(m);
381 
382 	if (IPFW_LOADED && ether_ipfw != 0) {
383 		if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) {
384 			if (m) {
385 				m_freem(m);
386 				return EACCES;	/* pkt dropped */
387 			} else
388 				return 0;	/* consumed e.g. in a pipe */
389 		}
390 	}
391 #endif
392 
393 	/*
394 	 * Queue message on interface, update output statistics if
395 	 * successful, and start output if interface not yet active.
396 	 */
397 	IFQ_HANDOFF(ifp, m, error);
398 	return (error);
399 }
400 
401 #if defined(INET) || defined(INET6)
402 /*
403  * ipfw processing for ethernet packets (in and out).
404  * The second parameter is NULL from ether_demux, and ifp from
405  * ether_output_frame.
406  */
407 int
408 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
409 	struct ip_fw **rule, int shared)
410 {
411 	struct ether_header *eh;
412 	struct ether_header save_eh;
413 	struct mbuf *m;
414 	int i;
415 	struct ip_fw_args args;
416 
417 	if (*rule != NULL && fw_one_pass)
418 		return 1; /* dummynet packet, already partially processed */
419 
420 	/*
421 	 * I need some amt of data to be contiguous, and in case others need
422 	 * the packet (shared==1) also better be in the first mbuf.
423 	 */
424 	m = *m0;
425 	i = min( m->m_pkthdr.len, max_protohdr);
426 	if ( shared || m->m_len < i) {
427 		m = m_pullup(m, i);
428 		if (m == NULL) {
429 			*m0 = m;
430 			return 0;
431 		}
432 	}
433 	eh = mtod(m, struct ether_header *);
434 	save_eh = *eh;			/* save copy for restore below */
435 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
436 
437 	args.m = m;		/* the packet we are looking at		*/
438 	args.oif = dst;		/* destination, if any			*/
439 	args.rule = *rule;	/* matching rule to restart		*/
440 	args.next_hop = NULL;	/* we do not support forward yet	*/
441 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
442 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
443 	i = ip_fw_chk_ptr(&args);
444 	m = args.m;
445 	if (m != NULL) {
446 		/*
447 		 * Restore Ethernet header, as needed, in case the
448 		 * mbuf chain was replaced by ipfw.
449 		 */
450 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
451 		if (m == NULL) {
452 			*m0 = m;
453 			return 0;
454 		}
455 		if (eh != mtod(m, struct ether_header *))
456 			bcopy(&save_eh, mtod(m, struct ether_header *),
457 				ETHER_HDR_LEN);
458 	}
459 	*m0 = m;
460 	*rule = args.rule;
461 
462 	if (i == IP_FW_DENY) /* drop */
463 		return 0;
464 
465 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
466 
467 	if (i == IP_FW_PASS) /* a PASS rule.  */
468 		return 1;
469 
470 	if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
471 		/*
472 		 * Pass the pkt to dummynet, which consumes it.
473 		 * If shared, make a copy and keep the original.
474 		 */
475 		if (shared) {
476 			m = m_copypacket(m, M_DONTWAIT);
477 			if (m == NULL)
478 				return 0;
479 		} else {
480 			/*
481 			 * Pass the original to dummynet and
482 			 * nothing back to the caller
483 			 */
484 			*m0 = NULL ;
485 		}
486 		ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
487 		return 0;
488 	}
489 	/*
490 	 * XXX at some point add support for divert/forward actions.
491 	 * If none of the above matches, we have to drop the pkt.
492 	 */
493 	return 0;
494 }
495 #endif
496 
497 /*
498  * Process a received Ethernet packet; the packet is in the
499  * mbuf chain m with the ethernet header at the front.
500  */
501 static void
502 ether_input(struct ifnet *ifp, struct mbuf *m)
503 {
504 	struct ether_header *eh;
505 	u_short etype;
506 
507 	/*
508 	 * Do consistency checks to verify assumptions
509 	 * made by code past this point.
510 	 */
511 	if ((m->m_flags & M_PKTHDR) == 0) {
512 		if_printf(ifp, "discard frame w/o packet header\n");
513 		ifp->if_ierrors++;
514 		m_freem(m);
515 		return;
516 	}
517 	if (m->m_len < ETHER_HDR_LEN) {
518 		/* XXX maybe should pullup? */
519 		if_printf(ifp, "discard frame w/o leading ethernet "
520 				"header (len %u pkt len %u)\n",
521 				m->m_len, m->m_pkthdr.len);
522 		ifp->if_ierrors++;
523 		m_freem(m);
524 		return;
525 	}
526 	eh = mtod(m, struct ether_header *);
527 	etype = ntohs(eh->ether_type);
528 	if (m->m_pkthdr.len >
529 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
530 		if_printf(ifp, "discard oversize frame "
531 				"(ether type %x flags %x len %u > max %lu)\n",
532 				etype, m->m_flags, m->m_pkthdr.len,
533 				ETHER_MAX_FRAME(ifp, etype,
534 						m->m_flags & M_HASFCS));
535 		ifp->if_ierrors++;
536 		m_freem(m);
537 		return;
538 	}
539 	if (m->m_pkthdr.rcvif == NULL) {
540 		if_printf(ifp, "discard frame w/o interface pointer\n");
541 		ifp->if_ierrors++;
542 		m_freem(m);
543 		return;
544 	}
545 #ifdef DIAGNOSTIC
546 	if (m->m_pkthdr.rcvif != ifp) {
547 		if_printf(ifp, "Warning, frame marked as received on %s\n",
548 			m->m_pkthdr.rcvif->if_xname);
549 	}
550 #endif
551 
552 #ifdef MAC
553 	/*
554 	 * Tag the mbuf with an appropriate MAC label before any other
555 	 * consumers can get to it.
556 	 */
557 	mac_create_mbuf_from_ifnet(ifp, m);
558 #endif
559 
560 	/*
561 	 * Give bpf a chance at the packet.
562 	 */
563 	BPF_MTAP(ifp, m);
564 
565 	/* If the CRC is still on the packet, trim it off. */
566 	if (m->m_flags & M_HASFCS) {
567 		m_adj(m, -ETHER_CRC_LEN);
568 		m->m_flags &= ~M_HASFCS;
569 	}
570 
571 	ifp->if_ibytes += m->m_pkthdr.len;
572 
573 	if (ifp->if_flags & IFF_MONITOR) {
574 		/*
575 		 * Interface marked for monitoring; discard packet.
576 		 */
577 		m_freem(m);
578 		return;
579 	}
580 
581 	/* Handle ng_ether(4) processing, if any */
582 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
583 		KASSERT(ng_ether_input_p != NULL,
584 		    ("ng_ether_input_p is NULL"));
585 		(*ng_ether_input_p)(ifp, &m);
586 		if (m == NULL)
587 			return;
588 	}
589 
590 	/*
591 	 * Tap the packet off here for a bridge.  bridge_input()
592 	 * will return NULL if it has consumed the packet, otherwise
593 	 * it gets processed as normal.  Note that bridge_input()
594 	 * will always return the original packet if we need to
595 	 * process it locally.
596 	 *
597 	 * XXX: When changing the below block, please also look
598 	 * at the src/sys/netgraph/ng_ether.c:ng_ether_rcv_upper()
599 	 */
600 	if (ifp->if_bridge) {
601 		BRIDGE_INPUT(ifp, m);
602 		if (m == NULL)
603 			return;
604 	}
605 
606 	/* First chunk of an mbuf contains good entropy */
607 	if (harvest.ethernet)
608 		random_harvest(m, 16, 3, 0, RANDOM_NET);
609 	ether_demux(ifp, m);
610 }
611 
612 /*
613  * Upper layer processing for a received Ethernet packet.
614  */
615 void
616 ether_demux(struct ifnet *ifp, struct mbuf *m)
617 {
618 	struct ether_header *eh;
619 	int isr;
620 	u_short ether_type;
621 #if defined(NETATALK)
622 	struct llc *l;
623 #endif
624 #if defined(INET) || defined(INET6)
625 	struct ip_fw *rule = ip_dn_claim_rule(m);
626 #endif
627 
628 	KASSERT(ifp != NULL, ("ether_demux: NULL interface pointer"));
629 
630 	eh = mtod(m, struct ether_header *);
631 	ether_type = ntohs(eh->ether_type);
632 
633 #if defined(INET) || defined(INET6)
634 	if (rule)	/* packet was already bridged */
635 		goto post_stats;
636 #endif
637 
638 	if (!(ifp->if_bridge) &&
639 	    !((ether_type == ETHERTYPE_VLAN || m->m_flags & M_VLANTAG) &&
640 	    ifp->if_vlantrunk != NULL)) {
641 #ifdef DEV_CARP
642 		/*
643 		 * XXX: Okay, we need to call carp_forus() and - if it is for
644 		 * us jump over code that does the normal check
645 		 * "IF_LLADDR(ifp) == ether_dhost". The check sequence is a bit
646 		 * different from OpenBSD, so we jump over as few code as
647 		 * possible, to catch _all_ sanity checks. This needs
648 		 * evaluation, to see if the carp ether_dhost values break any
649 		 * of these checks!
650 		 */
651 		if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
652 			goto pre_stats;
653 #endif
654 		/*
655 		 * Discard packet if upper layers shouldn't see it because it
656 		 * was unicast to a different Ethernet address. If the driver
657 		 * is working properly, then this situation can only happen
658 		 * when the interface is in promiscuous mode.
659 		 *
660 		 * If VLANs are active, and this packet has a VLAN tag, do
661 		 * not drop it here but pass it on to the VLAN layer, to
662 		 * give them a chance to consider it as well (e. g. in case
663 		 * bridging is only active on a VLAN).  They will drop it if
664 		 * it's undesired.
665 		 */
666 		if ((ifp->if_flags & IFF_PROMISC) != 0
667 		    && !ETHER_IS_MULTICAST(eh->ether_dhost)
668 		    && bcmp(eh->ether_dhost,
669 		      IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0
670 		    && (ifp->if_flags & IFF_PPROMISC) == 0) {
671 			    m_freem(m);
672 			    return;
673 		}
674 	}
675 
676 #ifdef DEV_CARP
677 pre_stats:
678 #endif
679 	/* Discard packet if interface is not up */
680 	if ((ifp->if_flags & IFF_UP) == 0) {
681 		m_freem(m);
682 		return;
683 	}
684 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
685 		if (bcmp(etherbroadcastaddr, eh->ether_dhost,
686 		    sizeof(etherbroadcastaddr)) == 0)
687 			m->m_flags |= M_BCAST;
688 		else
689 			m->m_flags |= M_MCAST;
690 	}
691 	if (m->m_flags & (M_BCAST|M_MCAST))
692 		ifp->if_imcasts++;
693 
694 #if defined(INET) || defined(INET6)
695 post_stats:
696 	if (IPFW_LOADED && ether_ipfw != 0) {
697 		if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) {
698 			if (m)
699 				m_freem(m);
700 			return;
701 		}
702 	}
703 #endif
704 
705 	/*
706 	 * Check to see if the device performed the VLAN decapsulation and
707 	 * provided us with the tag.
708 	 */
709 	if (m->m_flags & M_VLANTAG) {
710 		/*
711 		 * If no VLANs are configured, drop.
712 		 */
713 		if (ifp->if_vlantrunk == NULL) {
714 			ifp->if_noproto++;
715 			m_freem(m);
716 			return;
717 		}
718 		/*
719 		 * vlan_input() will either recursively call ether_input()
720 		 * or drop the packet.
721 		 */
722 		KASSERT(vlan_input_p != NULL,("ether_input: VLAN not loaded!"));
723 		(*vlan_input_p)(ifp, m);
724 		return;
725 	}
726 
727 	/*
728 	 * Handle protocols that expect to have the Ethernet header
729 	 * (and possibly FCS) intact.
730 	 */
731 	switch (ether_type) {
732 	case ETHERTYPE_VLAN:
733 		if (ifp->if_vlantrunk != NULL) {
734 			KASSERT(vlan_input_p,("ether_input: VLAN not loaded!"));
735 			(*vlan_input_p)(ifp, m);
736 		} else {
737 			ifp->if_noproto++;
738 			m_freem(m);
739 		}
740 		return;
741 	}
742 
743 	/* Strip off Ethernet header. */
744 	m_adj(m, ETHER_HDR_LEN);
745 
746 	/* If the CRC is still on the packet, trim it off. */
747 	if (m->m_flags & M_HASFCS) {
748 		m_adj(m, -ETHER_CRC_LEN);
749 		m->m_flags &= ~M_HASFCS;
750 	}
751 
752 	/* Reset layer specific mbuf flags to avoid confusing upper layers. */
753 	m->m_flags &= ~(M_PROTOFLAGS);
754 
755 	switch (ether_type) {
756 #ifdef INET
757 	case ETHERTYPE_IP:
758 		if ((m = ip_fastforward(m)) == NULL)
759 			return;
760 		isr = NETISR_IP;
761 		break;
762 
763 	case ETHERTYPE_ARP:
764 		if (ifp->if_flags & IFF_NOARP) {
765 			/* Discard packet if ARP is disabled on interface */
766 			m_freem(m);
767 			return;
768 		}
769 		isr = NETISR_ARP;
770 		break;
771 #endif
772 #ifdef IPX
773 	case ETHERTYPE_IPX:
774 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
775 			return;
776 		isr = NETISR_IPX;
777 		break;
778 #endif
779 #ifdef INET6
780 	case ETHERTYPE_IPV6:
781 		isr = NETISR_IPV6;
782 		break;
783 #endif
784 #ifdef NETATALK
785 	case ETHERTYPE_AT:
786 		isr = NETISR_ATALK1;
787 		break;
788 	case ETHERTYPE_AARP:
789 		isr = NETISR_AARP;
790 		break;
791 #endif /* NETATALK */
792 	default:
793 #ifdef IPX
794 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
795 			return;
796 #endif /* IPX */
797 #if defined(NETATALK)
798 		if (ether_type > ETHERMTU)
799 			goto discard;
800 		l = mtod(m, struct llc *);
801 		if (l->llc_dsap == LLC_SNAP_LSAP &&
802 		    l->llc_ssap == LLC_SNAP_LSAP &&
803 		    l->llc_control == LLC_UI) {
804 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
805 			    sizeof(at_org_code)) == 0 &&
806 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
807 				m_adj(m, LLC_SNAPFRAMELEN);
808 				isr = NETISR_ATALK2;
809 				break;
810 			}
811 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
812 			    sizeof(aarp_org_code)) == 0 &&
813 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
814 				m_adj(m, LLC_SNAPFRAMELEN);
815 				isr = NETISR_AARP;
816 				break;
817 			}
818 		}
819 #endif /* NETATALK */
820 		goto discard;
821 	}
822 	netisr_dispatch(isr, m);
823 	return;
824 
825 discard:
826 	/*
827 	 * Packet is to be discarded.  If netgraph is present,
828 	 * hand the packet to it for last chance processing;
829 	 * otherwise dispose of it.
830 	 */
831 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
832 		KASSERT(ng_ether_input_orphan_p != NULL,
833 		    ("ng_ether_input_orphan_p is NULL"));
834 		/*
835 		 * Put back the ethernet header so netgraph has a
836 		 * consistent view of inbound packets.
837 		 */
838 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
839 		(*ng_ether_input_orphan_p)(ifp, m);
840 		return;
841 	}
842 	m_freem(m);
843 }
844 
845 /*
846  * Convert Ethernet address to printable (loggable) representation.
847  * This routine is for compatibility; it's better to just use
848  *
849  *	printf("%6D", <pointer to address>, ":");
850  *
851  * since there's no static buffer involved.
852  */
853 char *
854 ether_sprintf(const u_char *ap)
855 {
856 	static char etherbuf[18];
857 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
858 	return (etherbuf);
859 }
860 
861 /*
862  * Perform common duties while attaching to interface list
863  */
864 void
865 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
866 {
867 	int i;
868 	struct ifaddr *ifa;
869 	struct sockaddr_dl *sdl;
870 
871 	ifp->if_addrlen = ETHER_ADDR_LEN;
872 	ifp->if_hdrlen = ETHER_HDR_LEN;
873 	if_attach(ifp);
874 	ifp->if_mtu = ETHERMTU;
875 	ifp->if_output = ether_output;
876 	ifp->if_input = ether_input;
877 	ifp->if_resolvemulti = ether_resolvemulti;
878 	if (ifp->if_baudrate == 0)
879 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
880 	ifp->if_broadcastaddr = etherbroadcastaddr;
881 
882 	ifa = ifp->if_addr;
883 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
884 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
885 	sdl->sdl_type = IFT_ETHER;
886 	sdl->sdl_alen = ifp->if_addrlen;
887 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
888 
889 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
890 	if (ng_ether_attach_p != NULL)
891 		(*ng_ether_attach_p)(ifp);
892 
893 	/* Announce Ethernet MAC address if non-zero. */
894 	for (i = 0; i < ifp->if_addrlen; i++)
895 		if (lla[i] != 0)
896 			break;
897 	if (i != ifp->if_addrlen)
898 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
899 	if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0)
900 		if_printf(ifp, "if_start running deferred for Giant\n");
901 }
902 
903 /*
904  * Perform common duties while detaching an Ethernet interface
905  */
906 void
907 ether_ifdetach(struct ifnet *ifp)
908 {
909 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
910 		KASSERT(ng_ether_detach_p != NULL,
911 		    ("ng_ether_detach_p is NULL"));
912 		(*ng_ether_detach_p)(ifp);
913 	}
914 
915 	bpfdetach(ifp);
916 	if_detach(ifp);
917 }
918 
919 SYSCTL_DECL(_net_link);
920 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
921 #if defined(INET) || defined(INET6)
922 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
923 	    &ether_ipfw,0,"Pass ether pkts through firewall");
924 #endif
925 
926 #if 0
927 /*
928  * This is for reference.  We have a table-driven version
929  * of the little-endian crc32 generator, which is faster
930  * than the double-loop.
931  */
932 uint32_t
933 ether_crc32_le(const uint8_t *buf, size_t len)
934 {
935 	size_t i;
936 	uint32_t crc;
937 	int bit;
938 	uint8_t data;
939 
940 	crc = 0xffffffff;	/* initial value */
941 
942 	for (i = 0; i < len; i++) {
943 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1)
944 			carry = (crc ^ data) & 1;
945 			crc >>= 1;
946 			if (carry)
947 				crc = (crc ^ ETHER_CRC_POLY_LE);
948 	}
949 
950 	return (crc);
951 }
952 #else
953 uint32_t
954 ether_crc32_le(const uint8_t *buf, size_t len)
955 {
956 	static const uint32_t crctab[] = {
957 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
958 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
959 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
960 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
961 	};
962 	size_t i;
963 	uint32_t crc;
964 
965 	crc = 0xffffffff;	/* initial value */
966 
967 	for (i = 0; i < len; i++) {
968 		crc ^= buf[i];
969 		crc = (crc >> 4) ^ crctab[crc & 0xf];
970 		crc = (crc >> 4) ^ crctab[crc & 0xf];
971 	}
972 
973 	return (crc);
974 }
975 #endif
976 
977 uint32_t
978 ether_crc32_be(const uint8_t *buf, size_t len)
979 {
980 	size_t i;
981 	uint32_t crc, carry;
982 	int bit;
983 	uint8_t data;
984 
985 	crc = 0xffffffff;	/* initial value */
986 
987 	for (i = 0; i < len; i++) {
988 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
989 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
990 			crc <<= 1;
991 			if (carry)
992 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
993 		}
994 	}
995 
996 	return (crc);
997 }
998 
999 int
1000 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
1001 {
1002 	struct ifaddr *ifa = (struct ifaddr *) data;
1003 	struct ifreq *ifr = (struct ifreq *) data;
1004 	int error = 0;
1005 
1006 	switch (command) {
1007 	case SIOCSIFADDR:
1008 		ifp->if_flags |= IFF_UP;
1009 
1010 		switch (ifa->ifa_addr->sa_family) {
1011 #ifdef INET
1012 		case AF_INET:
1013 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1014 			arp_ifinit(ifp, ifa);
1015 			break;
1016 #endif
1017 #ifdef IPX
1018 		/*
1019 		 * XXX - This code is probably wrong
1020 		 */
1021 		case AF_IPX:
1022 			{
1023 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1024 
1025 			if (ipx_nullhost(*ina))
1026 				ina->x_host =
1027 				    *(union ipx_host *)
1028 				    IF_LLADDR(ifp);
1029 			else {
1030 				bcopy((caddr_t) ina->x_host.c_host,
1031 				      (caddr_t) IF_LLADDR(ifp),
1032 				      ETHER_ADDR_LEN);
1033 			}
1034 
1035 			/*
1036 			 * Set new address
1037 			 */
1038 			ifp->if_init(ifp->if_softc);
1039 			break;
1040 			}
1041 #endif
1042 		default:
1043 			ifp->if_init(ifp->if_softc);
1044 			break;
1045 		}
1046 		break;
1047 
1048 	case SIOCGIFADDR:
1049 		{
1050 			struct sockaddr *sa;
1051 
1052 			sa = (struct sockaddr *) & ifr->ifr_data;
1053 			bcopy(IF_LLADDR(ifp),
1054 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1055 		}
1056 		break;
1057 
1058 	case SIOCSIFMTU:
1059 		/*
1060 		 * Set the interface MTU.
1061 		 */
1062 		if (ifr->ifr_mtu > ETHERMTU) {
1063 			error = EINVAL;
1064 		} else {
1065 			ifp->if_mtu = ifr->ifr_mtu;
1066 		}
1067 		break;
1068 	default:
1069 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1070 		break;
1071 	}
1072 	return (error);
1073 }
1074 
1075 static int
1076 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1077 	struct sockaddr *sa)
1078 {
1079 	struct sockaddr_dl *sdl;
1080 #ifdef INET
1081 	struct sockaddr_in *sin;
1082 #endif
1083 #ifdef INET6
1084 	struct sockaddr_in6 *sin6;
1085 #endif
1086 	u_char *e_addr;
1087 
1088 	switch(sa->sa_family) {
1089 	case AF_LINK:
1090 		/*
1091 		 * No mapping needed. Just check that it's a valid MC address.
1092 		 */
1093 		sdl = (struct sockaddr_dl *)sa;
1094 		e_addr = LLADDR(sdl);
1095 		if (!ETHER_IS_MULTICAST(e_addr))
1096 			return EADDRNOTAVAIL;
1097 		*llsa = 0;
1098 		return 0;
1099 
1100 #ifdef INET
1101 	case AF_INET:
1102 		sin = (struct sockaddr_in *)sa;
1103 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1104 			return EADDRNOTAVAIL;
1105 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1106 		       M_NOWAIT|M_ZERO);
1107 		if (sdl == NULL)
1108 			return ENOMEM;
1109 		sdl->sdl_len = sizeof *sdl;
1110 		sdl->sdl_family = AF_LINK;
1111 		sdl->sdl_index = ifp->if_index;
1112 		sdl->sdl_type = IFT_ETHER;
1113 		sdl->sdl_alen = ETHER_ADDR_LEN;
1114 		e_addr = LLADDR(sdl);
1115 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1116 		*llsa = (struct sockaddr *)sdl;
1117 		return 0;
1118 #endif
1119 #ifdef INET6
1120 	case AF_INET6:
1121 		sin6 = (struct sockaddr_in6 *)sa;
1122 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1123 			/*
1124 			 * An IP6 address of 0 means listen to all
1125 			 * of the Ethernet multicast address used for IP6.
1126 			 * (This is used for multicast routers.)
1127 			 */
1128 			ifp->if_flags |= IFF_ALLMULTI;
1129 			*llsa = 0;
1130 			return 0;
1131 		}
1132 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1133 			return EADDRNOTAVAIL;
1134 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1135 		       M_NOWAIT|M_ZERO);
1136 		if (sdl == NULL)
1137 			return (ENOMEM);
1138 		sdl->sdl_len = sizeof *sdl;
1139 		sdl->sdl_family = AF_LINK;
1140 		sdl->sdl_index = ifp->if_index;
1141 		sdl->sdl_type = IFT_ETHER;
1142 		sdl->sdl_alen = ETHER_ADDR_LEN;
1143 		e_addr = LLADDR(sdl);
1144 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1145 		*llsa = (struct sockaddr *)sdl;
1146 		return 0;
1147 #endif
1148 
1149 	default:
1150 		/*
1151 		 * Well, the text isn't quite right, but it's the name
1152 		 * that counts...
1153 		 */
1154 		return EAFNOSUPPORT;
1155 	}
1156 }
1157 
1158 static void*
1159 ether_alloc(u_char type, struct ifnet *ifp)
1160 {
1161 	struct arpcom	*ac;
1162 
1163 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1164 	ac->ac_ifp = ifp;
1165 
1166 	return (ac);
1167 }
1168 
1169 static void
1170 ether_free(void *com, u_char type)
1171 {
1172 
1173 	free(com, M_ARPCOM);
1174 }
1175 
1176 static int
1177 ether_modevent(module_t mod, int type, void *data)
1178 {
1179 
1180 	switch (type) {
1181 	case MOD_LOAD:
1182 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1183 		break;
1184 	case MOD_UNLOAD:
1185 		if_deregister_com_alloc(IFT_ETHER);
1186 		break;
1187 	default:
1188 		return EOPNOTSUPP;
1189 	}
1190 
1191 	return (0);
1192 }
1193 
1194 static moduledata_t ether_mod = {
1195 	"ether",
1196 	ether_modevent,
1197 	0
1198 };
1199 
1200 void
1201 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1202 {
1203 	struct ether_vlan_header vlan;
1204 	struct mbuf mv, mb;
1205 
1206 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1207 	    ("%s: vlan information not present", __func__));
1208 	KASSERT(m->m_len >= sizeof(struct ether_header),
1209 	    ("%s: mbuf not large enough for header", __func__));
1210 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1211 	vlan.evl_proto = vlan.evl_encap_proto;
1212 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1213 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1214 	m->m_len -= sizeof(struct ether_header);
1215 	m->m_data += sizeof(struct ether_header);
1216 	/*
1217 	 * If a data link has been supplied by the caller, then we will need to
1218 	 * re-create a stack allocated mbuf chain with the following structure:
1219 	 *
1220 	 * (1) mbuf #1 will contain the supplied data link
1221 	 * (2) mbuf #2 will contain the vlan header
1222 	 * (3) mbuf #3 will contain the original mbuf's packet data
1223 	 *
1224 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1225 	 */
1226 	if (data != NULL) {
1227 		mv.m_next = m;
1228 		mv.m_data = (caddr_t)&vlan;
1229 		mv.m_len = sizeof(vlan);
1230 		mb.m_next = &mv;
1231 		mb.m_data = data;
1232 		mb.m_len = dlen;
1233 		bpf_mtap(bp, &mb);
1234 	} else
1235 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1236 	m->m_len += sizeof(struct ether_header);
1237 	m->m_data -= sizeof(struct ether_header);
1238 }
1239 
1240 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1241 MODULE_VERSION(ether, 1);
1242