xref: /dflybsd-src/sys/net/if_ethersubr.c (revision f7e25d559127833cbbb89b74fcf0cc036406459b)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
34  * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
35  * $DragonFly: src/sys/net/if_ethersubr.c,v 1.23 2004/12/24 04:54:49 dillon Exp $
36  */
37 
38 #include "opt_atalk.h"
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_ipx.h"
42 #include "opt_bdg.h"
43 #include "opt_netgraph.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.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/netisr.h>
56 #include <net/route.h>
57 #include <net/if_llc.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/bridge/bridge.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 <net/ipfw/ip_fw.h>
69 #include <net/dummynet/ip_dummynet.h>
70 #endif
71 #ifdef INET6
72 #include <netinet6/nd6.h>
73 #endif
74 
75 #ifdef IPX
76 #include <netproto/ipx/ipx.h>
77 #include <netproto/ipx/ipx_if.h>
78 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
79 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
80 		  short *tp, int *hlen);
81 #endif
82 
83 #ifdef NS
84 #include <netns/ns.h>
85 #include <netns/ns_if.h>
86 ushort ns_nettype;
87 int ether_outputdebug = 0;
88 int ether_inputdebug = 0;
89 #endif
90 
91 #ifdef NETATALK
92 #include <netproto/atalk/at.h>
93 #include <netproto/atalk/at_var.h>
94 #include <netproto/atalk/at_extern.h>
95 
96 #define	llc_snap_org_code	llc_un.type_snap.org_code
97 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
98 
99 extern u_char	at_org_code[3];
100 extern u_char	aarp_org_code[3];
101 #endif /* NETATALK */
102 
103 /* netgraph node hooks for ng_ether(4) */
104 void	(*ng_ether_input_p)(struct ifnet *ifp,
105 		struct mbuf **mp, struct ether_header *eh);
106 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
107 		struct mbuf *m, struct ether_header *eh);
108 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
109 void	(*ng_ether_attach_p)(struct ifnet *ifp);
110 void	(*ng_ether_detach_p)(struct ifnet *ifp);
111 
112 int	(*vlan_input_p)(struct ether_header *eh, struct mbuf *m);
113 int	(*vlan_input_tag_p)(struct mbuf *m, uint16_t t);
114 
115 static int	ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
116 			     struct rtentry *);
117 
118 /*
119  * bridge support
120  */
121 int do_bridge;
122 bridge_in_t *bridge_in_ptr;
123 bdg_forward_t *bdg_forward_ptr;
124 bdgtakeifaces_t *bdgtakeifaces_ptr;
125 struct bdg_softc *ifp2sc;
126 
127 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
128 		struct sockaddr *);
129 const uint8_t	etherbroadcastaddr[ETHER_ADDR_LEN] = {
130 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
131 };
132 
133 #define gotoerr(e) do { error = (e); goto bad;} while (0)
134 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
135 
136 int
137 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
138 	struct ip_fw **rule, struct ether_header *eh, int shared);
139 static int ether_ipfw;
140 
141 /*
142  * Ethernet output routine.
143  * Encapsulate a packet of type family for the local net.
144  * Use trailer local net encapsulation if enough data in first
145  * packet leaves a multiple of 512 bytes of data in remainder.
146  * Assumes that ifp is actually pointer to arpcom structure.
147  */
148 static int
149 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
150 	     struct rtentry *rt)
151 {
152 	struct ether_header *eh, *deh;
153 	u_char *edst;
154 	int loop_copy = 0;
155 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
156 	struct arpcom *ac = IFP2AC(ifp);
157 	int error;
158 
159 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
160 		gotoerr(ENETDOWN);
161 
162 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
163 	if (m == NULL)
164 		gotoerr(ENOBUFS);
165 	eh = mtod(m, struct ether_header *);
166 	edst = eh->ether_dhost;
167 
168 	/* Fill in the destination ethernet address and frame type. */
169 	switch (dst->sa_family) {
170 #ifdef INET
171 	case AF_INET:
172 		if (!arpresolve(ifp, rt, m, dst, edst))
173 			return (0);	/* if not yet resolved */
174 		eh->ether_type = htons(ETHERTYPE_IP);
175 		break;
176 #endif
177 #ifdef INET6
178 	case AF_INET6:
179 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
180 			return (0);		/* Something bad happened. */
181 		eh->ether_type = htons(ETHERTYPE_IPV6);
182 		break;
183 #endif
184 #ifdef IPX
185 	case AF_IPX:
186 		if (ef_outputp != NULL) {
187 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
188 					   &hlen);
189 			if (error)
190 				goto bad;
191 		} else {
192 			eh->ether_type = htons(ETHERTYPE_IPX);
193 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
194 			      edst, ETHER_ADDR_LEN);
195 		}
196 		break;
197 #endif
198 #ifdef NETATALK
199 	case AF_APPLETALK: {
200 		struct at_ifaddr *aa;
201 
202 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
203 			error = 0;	/* XXX */
204 			goto bad;
205 		}
206 		/*
207 		 * In the phase 2 case, need to prepend an mbuf for
208 		 * the llc header.  Since we must preserve the value
209 		 * of m, which is passed to us by value, we m_copy()
210 		 * the first mbuf, and use it for our llc header.
211 		 */
212 		if (aa->aa_flags & AFA_PHASE2) {
213 			struct llc llc;
214 
215 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
216 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
217 			llc.llc_control = LLC_UI;
218 			bcopy(at_org_code, llc.llc_snap_org_code,
219 			      sizeof at_org_code);
220 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
221 			bcopy(&llc, mtod(m, caddr_t), sizeof(struct llc));
222 			eh->ether_type = htons(m->m_pkthdr.len);
223 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
224 		} else {
225 			eh->ether_type = htons(ETHERTYPE_AT);
226 		}
227 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
228 			return (0);
229 		break;
230 	  }
231 #endif /* NETATALK */
232 #ifdef NS
233 	case AF_NS:
234 		switch(ns_nettype) {
235 		default:
236 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
237 			eh->ether_type = 0x8137;
238 			break;
239 		case 0x0:	/* Novell 802.3 */
240 			eh->ether_type = htons(m->m_pkthdr.len);
241 			break;
242 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
243 			M_PREPEND(m, 3, MB_DONTWAIT);
244 			eh->ether_type = htons(m->m_pkthdr.len);
245 			cp = mtod(m, u_char *);
246 			*cp++ = 0xE0;
247 			*cp++ = 0xE0;
248 			*cp++ = 0x03;
249 			break;
250 		}
251 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
252 		      ETHER_ADDR_LEN);
253 		/*
254 		 * XXX if ns_thishost is the same as the node's ethernet
255 		 * address then just the default code will catch this anyhow.
256 		 * So I'm not sure if this next clause should be here at all?
257 		 * [JRE]
258 		 */
259 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
260 			m->m_pkthdr.rcvif = ifp;
261 			netisr_dispatch(NETISR_NS, m);
262 			return (error);
263 		}
264 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
265 			m->m_flags |= M_BCAST;
266 		break;
267 #endif /* NS */
268 	case pseudo_AF_HDRCMPLT:
269 	case AF_UNSPEC:
270 		loop_copy = -1; /* if this is for us, don't do it */
271 		deh = (struct ether_header *)dst->sa_data;
272 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
273 		eh->ether_type = deh->ether_type;
274 		break;
275 
276 	default:
277 		printf("%s: can't handle af%d\n", ifp->if_xname,
278 			dst->sa_family);
279 		gotoerr(EAFNOSUPPORT);
280 	}
281 
282 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
283 		memcpy(eh->ether_shost,
284 		       ((struct ether_header *)dst->sa_data)->ether_shost,
285 		       ETHER_ADDR_LEN);
286 	else
287 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
288 
289 	/*
290 	 * If a simplex interface, and the packet is being sent to our
291 	 * Ethernet address or a broadcast address, loopback a copy.
292 	 * XXX To make a simplex device behave exactly like a duplex
293 	 * device, we should copy in the case of sending to our own
294 	 * ethernet address (thus letting the original actually appear
295 	 * on the wire). However, we don't do that here for security
296 	 * reasons and compatibility with the original behavior.
297 	 */
298 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
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 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
306 			struct mbuf *n;
307 
308 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
309 				n->m_pkthdr.csum_flags |= csum_flags;
310 				if (csum_flags & CSUM_DATA_VALID)
311 					n->m_pkthdr.csum_data = 0xffff;
312 				if_simloop(ifp, n, dst->sa_family, hlen);
313 			} else
314 				ifp->if_iqdrops++;
315 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
316 				ETHER_ADDR_LEN) == 0) {
317 			m->m_pkthdr.csum_flags |= csum_flags;
318 			if (csum_flags & CSUM_DATA_VALID)
319 				m->m_pkthdr.csum_data = 0xffff;
320 			if_simloop(ifp, m, dst->sa_family, hlen);
321 			return (0);	/* XXX */
322 		}
323 	}
324 
325 	/* Handle ng_ether(4) processing, if any */
326 	if (ng_ether_output_p != NULL) {
327 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
328 bad:			if (m != NULL)
329 				m_freem(m);
330 			return (error);
331 		}
332 		if (m == NULL)
333 			return (0);
334 	}
335 
336 	/* Continue with link-layer output */
337 	return ether_output_frame(ifp, m);
338 }
339 
340 /*
341  * Ethernet link layer output routine to send a raw frame to the device.
342  *
343  * This assumes that the 14 byte Ethernet header is present and contiguous
344  * in the first mbuf (if BRIDGE'ing).
345  */
346 int
347 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
348 {
349 	struct ip_fw *rule = NULL;
350 	int error = 0;
351 	int s;
352 
353 	/* Extract info from dummynet tag, ignore others */
354 	for (; m->m_type == MT_TAG; m = m->m_next)
355 		if (m->m_flags == PACKET_TAG_DUMMYNET)
356 			rule = ((struct dn_pkt *)m)->rule;
357 
358 	if (rule)	/* packet was already bridged */
359 		goto no_bridge;
360 
361 	if (BDG_ACTIVE(ifp) ) {
362 		struct ether_header *eh; /* a ptr suffices */
363 
364 		m->m_pkthdr.rcvif = NULL;
365 		eh = mtod(m, struct ether_header *);
366 		m_adj(m, ETHER_HDR_LEN);
367 		m = bdg_forward_ptr(m, eh, ifp);
368 		if (m != NULL)
369 			m_freem(m);
370 		return (0);
371 	}
372 
373 no_bridge:
374 	s = splimp();
375 	if (IPFW_LOADED && ether_ipfw != 0) {
376 		struct ether_header save_eh, *eh;
377 
378 		eh = mtod(m, struct ether_header *);
379 		save_eh = *eh;
380 		m_adj(m, ETHER_HDR_LEN);
381 		if (ether_ipfw_chk(&m, ifp, &rule, eh, 0) == 0) {
382 			if (m) {
383 				m_freem(m);
384 				return ENOBUFS; /* pkt dropped */
385 			} else
386 				return 0;	/* consumed e.g. in a pipe */
387 		}
388 		/* packet was ok, restore the ethernet header */
389 		if ((void *)(eh + 1) == (void *)m->m_data) {
390 			m->m_data -= ETHER_HDR_LEN ;
391 			m->m_len += ETHER_HDR_LEN ;
392 			m->m_pkthdr.len += ETHER_HDR_LEN ;
393 		} else {
394 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
395 			if (m == NULL) /* nope... */
396 				return ENOBUFS;
397 			bcopy(&save_eh, mtod(m, struct ether_header *),
398 			    ETHER_HDR_LEN);
399 		}
400 	}
401 
402 	/*
403 	 * Queue message on interface, update output statistics if
404 	 * successful, and start output if interface not yet active.
405 	 */
406 	if (!IF_HANDOFF(&ifp->if_snd, m, ifp))
407 		error = ENOBUFS;
408 	splx(s);
409 	return (error);
410 }
411 
412 /*
413  * ipfw processing for ethernet packets (in and out).
414  * The second parameter is NULL from ether_demux, and ifp from
415  * ether_output_frame. This section of code could be used from
416  * bridge.c as well as long as we use some extra info
417  * to distinguish that case from ether_output_frame();
418  */
419 int
420 ether_ipfw_chk(
421 	struct mbuf **m0,
422 	struct ifnet *dst,
423 	struct ip_fw **rule,
424 	struct ether_header *eh,
425 	int shared)
426 {
427 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
428 	struct ip_fw_args args;
429 	int i;
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 	i = min((*m0)->m_pkthdr.len, max_protohdr);
439 	if (shared || (*m0)->m_len < i) {
440 		*m0 = m_pullup(*m0, i);
441 		if (*m0 == NULL)
442 			return 0;
443 	}
444 
445 	args.m = *m0;		/* the packet we are looking at		*/
446 	args.oif = dst;		/* destination, if any			*/
447 	args.divert_rule = 0;	/* we do not support divert yet		*/
448 	args.rule = *rule;	/* matching rule to restart		*/
449 	args.next_hop = NULL;	/* we do not support forward yet	*/
450 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
451 	i = ip_fw_chk_ptr(&args);
452 	*m0 = args.m;
453 	*rule = args.rule;
454 
455 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */
456 		return 0;
457 
458 	if (i == 0) /* a PASS rule.  */
459 		return 1;
460 
461 	if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
462 		/*
463 		 * Pass the pkt to dummynet, which consumes it.
464 		 * If shared, make a copy and keep the original.
465 		 */
466 		struct mbuf *m ;
467 
468 		if (shared) {
469 			m = m_copypacket(*m0, MB_DONTWAIT);
470 			if (m == NULL)
471 				return 0;
472 		} else {
473 			m = *m0 ; /* pass the original to dummynet */
474 			*m0 = NULL ; /* and nothing back to the caller */
475 		}
476 		/*
477 		 * Prepend the header, optimize for the common case of
478 		 * eh pointing into the mbuf.
479 		 */
480 		if ((void *)(eh + 1) == (void *)m->m_data) {
481 			m->m_data -= ETHER_HDR_LEN ;
482 			m->m_len += ETHER_HDR_LEN ;
483 			m->m_pkthdr.len += ETHER_HDR_LEN ;
484 		} else {
485 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
486 			if (m == NULL) /* nope... */
487 				return 0;
488 			bcopy(&save_eh, mtod(m, struct ether_header *),
489 			    ETHER_HDR_LEN);
490 		}
491 		ip_dn_io_ptr(m, (i & 0xffff),
492 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
493 		return 0;
494 	}
495 	/*
496 	 * XXX at some point add support for divert/forward actions.
497 	 * If none of the above matches, we have to drop the pkt.
498 	 */
499 	return 0;
500 }
501 
502 /*
503  * XXX merge this function with ether_input.
504  */
505 static void
506 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
507 {
508 	ether_input(ifp, NULL, m);
509 }
510 
511 /*
512  * Process a received Ethernet packet. We have two different interfaces:
513  * one (conventional) assumes the packet in the mbuf, with the ethernet
514  * header provided separately in *eh. The second one (new) has everything
515  * in the mbuf, and we can tell it because eh == NULL.
516  * The caller MUST MAKE SURE that there are at least
517  * sizeof(struct ether_header) bytes in the first mbuf.
518  *
519  * This allows us to concentrate in one place a bunch of code which
520  * is replicated in all device drivers. Also, many functions called
521  * from ether_input() try to put the eh back into the mbuf, so we
522  * can later propagate the 'contiguous packet' interface to them,
523  * and handle the old interface just here.
524  *
525  * NOTA BENE: for many drivers "eh" is a pointer into the first mbuf or
526  * cluster, right before m_data. So be very careful when working on m,
527  * as you could destroy *eh !!
528  *
529  * First we perform any link layer operations, then continue
530  * to the upper layers with ether_demux().
531  */
532 void
533 ether_input(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
534 {
535 	struct ether_header save_eh;
536 
537 	if (eh == NULL) {
538 		if (m->m_len < sizeof(struct ether_header)) {
539 			/* XXX error in the caller. */
540 			m_freem(m);
541 			return;
542 		}
543 		m->m_pkthdr.rcvif = ifp;
544 		eh = mtod(m, struct ether_header *);
545 		m_adj(m, sizeof(struct ether_header));
546 		/* XXX */
547 		/* m->m_pkthdr.len = m->m_len; */
548 	}
549 
550 	/* Check for a BPF tap */
551 	if (ifp->if_bpf != NULL) {
552 		struct m_hdr mh;
553 
554 		/* This kludge is OK; BPF treats the "mbuf" as read-only */
555 		mh.mh_next = m;
556 		mh.mh_data = (char *)eh;
557 		mh.mh_len = ETHER_HDR_LEN;
558 		bpf_mtap(ifp, (struct mbuf *)&mh);
559 	}
560 
561 	ifp->if_ibytes += m->m_pkthdr.len + sizeof (*eh);
562 
563 	/* Handle ng_ether(4) processing, if any */
564 	if (ng_ether_input_p != NULL) {
565 		(*ng_ether_input_p)(ifp, &m, eh);
566 		if (m == NULL)
567 			return;
568 	}
569 
570 	/* Check for bridging mode */
571 	if (BDG_ACTIVE(ifp) ) {
572 		struct ifnet *bif;
573 
574 		/* Check with bridging code */
575 		if ((bif = bridge_in_ptr(ifp, eh)) == BDG_DROP) {
576 			m_freem(m);
577 			return;
578 		}
579 		if (bif != BDG_LOCAL) {
580 			save_eh = *eh ; /* because it might change */
581 			m = bdg_forward_ptr(m, eh, bif); /* needs forwarding */
582 			/*
583 			 * Do not continue if bdg_forward_ptr() processed our
584 			 * packet (and cleared the mbuf pointer m) or if
585 			 * it dropped (m_free'd) the packet itself.
586 			 */
587 			if (m == NULL) {
588 			    if (bif == BDG_BCAST || bif == BDG_MCAST)
589 				printf("bdg_forward drop MULTICAST PKT\n");
590 			    return;
591 			}
592 			eh = &save_eh ;
593 		}
594 		if (bif == BDG_LOCAL || bif == BDG_BCAST || bif == BDG_MCAST)
595 			goto recvLocal;		/* receive locally */
596 
597 		/* If not local and not multicast, just drop it */
598 		if (m != NULL)
599 			m_freem(m);
600 		return;
601 	}
602 
603 recvLocal:
604 	/* Continue with upper layer processing */
605 	ether_demux(ifp, eh, m);
606 }
607 
608 /*
609  * Upper layer processing for a received Ethernet packet.
610  */
611 void
612 ether_demux(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
613 {
614 	int isr;
615 	u_short ether_type;
616 #if defined(NETATALK)
617 	struct llc *l;
618 #endif
619 	struct ip_fw *rule = NULL;
620 
621 	/* Extract info from dummynet tag, ignore others */
622 	for (;m->m_type == MT_TAG; m = m->m_next)
623 		if (m->m_flags == PACKET_TAG_DUMMYNET) {
624 			rule = ((struct dn_pkt *)m)->rule;
625 			ifp = m->m_next->m_pkthdr.rcvif;
626 		}
627 
628 	if (rule)	/* packet was already bridged */
629 		goto post_stats;
630 
631 	/*
632 	 * Discard packet if upper layers shouldn't see it because
633 	 * it was unicast to a different Ethernet address.  If the
634 	 * driver is working properly, then this situation can only
635 	 * happen when the interface is in promiscuous mode.
636 	 */
637 	if (!BDG_ACTIVE(ifp) &&
638 	    ((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
639 	    (eh->ether_dhost[0] & 1) == 0 &&
640 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
641 		m_freem(m);
642 		return;
643 	}
644 	/* Discard packet if interface is not up */
645 	if (!(ifp->if_flags & IFF_UP)) {
646 		m_freem(m);
647 		return;
648 	}
649 	if (eh->ether_dhost[0] & 1) {
650 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
651 			 ifp->if_addrlen) == 0)
652 			m->m_flags |= M_BCAST;
653 		else
654 			m->m_flags |= M_MCAST;
655 	}
656 	if (m->m_flags & (M_BCAST|M_MCAST))
657 		ifp->if_imcasts++;
658 
659 post_stats:
660 	if (IPFW_LOADED && ether_ipfw != 0) {
661 		if (ether_ipfw_chk(&m, NULL, &rule, eh, 0 ) == 0) {
662 			if (m)
663 				m_freem(m);
664 			return;
665 		}
666 	}
667 
668 	ether_type = ntohs(eh->ether_type);
669 
670 	switch (ether_type) {
671 #ifdef INET
672 	case ETHERTYPE_IP:
673 		if (ipflow_fastforward(m))
674 			return;
675 		isr = NETISR_IP;
676 		break;
677 
678 	case ETHERTYPE_ARP:
679 		if (ifp->if_flags & IFF_NOARP) {
680 			/* Discard packet if ARP is disabled on interface */
681 			m_freem(m);
682 			return;
683 		}
684 		isr = NETISR_ARP;
685 		break;
686 #endif
687 #ifdef IPX
688 	case ETHERTYPE_IPX:
689 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
690 			return;
691 		isr = NETISR_IPX;
692 		break;
693 #endif
694 #ifdef INET6
695 	case ETHERTYPE_IPV6:
696 		isr = NETISR_IPV6;
697 		break;
698 #endif
699 #ifdef NS
700 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
701 		isr = NETISR_NS;
702 		break;
703 
704 #endif /* NS */
705 #ifdef NETATALK
706 	case ETHERTYPE_AT:
707 		isr = NETISR_ATALK1;
708 		break;
709 	case ETHERTYPE_AARP:
710 		isr = NETISR_AARP;
711 		break;
712 #endif /* NETATALK */
713 	case ETHERTYPE_VLAN:
714 		/* XXX lock ? */
715 		if (vlan_input_p != NULL)
716 			(*vlan_input_p)(eh, m);
717 		else {
718 			m->m_pkthdr.rcvif->if_noproto++;
719 			m_freem(m);
720 		}
721 		/* XXX unlock ? */
722 		return;
723 	default:
724 #ifdef IPX
725 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
726 			return;
727 #endif /* IPX */
728 #ifdef NS
729 		checksum = mtod(m, ushort *);
730 		/* Novell 802.3 */
731 		if ((ether_type <= ETHERMTU) &&
732 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
733 			if (*checksum == 0xE0E0) {
734 				m->m_pkthdr.len -= 3;
735 				m->m_len -= 3;
736 				m->m_data += 3;
737 			}
738 			isr = NETISR_NS;
739 			break;
740 		}
741 #endif /* NS */
742 #ifdef NETATALK
743 		if (ether_type > ETHERMTU)
744 			goto dropanyway;
745 		l = mtod(m, struct llc *);
746 		if (l->llc_dsap == LLC_SNAP_LSAP &&
747 		    l->llc_ssap == LLC_SNAP_LSAP &&
748 		    l->llc_control == LLC_UI) {
749 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
750 			    sizeof(at_org_code)) == 0 &&
751 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
752 				m_adj(m, sizeof(struct llc));
753 				isr = NETISR_ATALK2;
754 				break;
755 			}
756 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
757 			    sizeof(aarp_org_code)) == 0 &&
758 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
759 				m_adj(m, sizeof(struct llc));
760 				isr = NETISR_AARP;
761 				break;
762 			}
763 		}
764 dropanyway:
765 #endif /* NETATALK */
766 		if (ng_ether_input_orphan_p != NULL)
767 			(*ng_ether_input_orphan_p)(ifp, m, eh);
768 		else
769 			m_freem(m);
770 		return;
771 	}
772 	netisr_dispatch(isr, m);
773 }
774 
775 /*
776  * Perform common duties while attaching to interface list
777  */
778 
779 void
780 ether_ifattach(struct ifnet *ifp, uint8_t *lla)
781 {
782 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header));
783 }
784 
785 void
786 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen)
787 {
788 	struct ifaddr *ifa;
789 	struct sockaddr_dl *sdl;
790 
791 	ifp->if_output = ether_output;
792 	ifp->if_input = ether_input_internal;
793 	ifp->if_type = IFT_ETHER;
794 	ifp->if_addrlen = ETHER_ADDR_LEN;
795 	ifp->if_broadcastaddr = etherbroadcastaddr;
796 	ifp->if_hdrlen = 14;
797 	if_attach(ifp);
798 	ifp->if_mtu = ETHERMTU;
799 	ifp->if_resolvemulti = ether_resolvemulti;
800 	if (ifp->if_baudrate == 0)
801 		ifp->if_baudrate = 10000000;
802 	ifa = ifnet_addrs[ifp->if_index - 1];
803 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __FUNCTION__));
804 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
805 	sdl->sdl_type = IFT_ETHER;
806 	sdl->sdl_alen = ifp->if_addrlen;
807 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
808 	/*
809 	 * XXX Keep the current drivers happy.
810 	 * XXX Remove once all drivers have been cleaned up
811 	 */
812 	if (lla != IFP2AC(ifp)->ac_enaddr)
813 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
814 	bpfattach(ifp, dlt, hdrlen);
815 	if (ng_ether_attach_p != NULL)
816 		(*ng_ether_attach_p)(ifp);
817 	if (BDG_LOADED)
818 		bdgtakeifaces_ptr();
819 
820 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
821 }
822 
823 /*
824  * Perform common duties while detaching an Ethernet interface
825  */
826 void
827 ether_ifdetach(struct ifnet *ifp)
828 {
829 	int s;
830 
831 	s = splnet();
832 	if_down(ifp);
833 	splx(s);
834 
835 	if (ng_ether_detach_p != NULL)
836 		(*ng_ether_detach_p)(ifp);
837 	bpfdetach(ifp);
838 	if_detach(ifp);
839 	if (BDG_LOADED)
840 		bdgtakeifaces_ptr();
841 }
842 
843 SYSCTL_DECL(_net_link);
844 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
845 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
846 	    &ether_ipfw,0,"Pass ether pkts through firewall");
847 
848 int
849 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
850 {
851 	struct ifaddr *ifa = (struct ifaddr *) data;
852 	struct ifreq *ifr = (struct ifreq *) data;
853 	int error = 0;
854 
855 	switch (command) {
856 	case SIOCSIFADDR:
857 		ifp->if_flags |= IFF_UP;
858 
859 		switch (ifa->ifa_addr->sa_family) {
860 #ifdef INET
861 		case AF_INET:
862 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
863 			arp_ifinit(ifp, ifa);
864 			break;
865 #endif
866 #ifdef IPX
867 		/*
868 		 * XXX - This code is probably wrong
869 		 */
870 		case AF_IPX:
871 			{
872 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
873 			struct arpcom *ac = IFP2AC(ifp);
874 
875 			if (ipx_nullhost(*ina))
876 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
877 			else
878 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
879 				      sizeof ac->ac_enaddr);
880 
881 			ifp->if_init(ifp->if_softc);	/* Set new address. */
882 			break;
883 			}
884 #endif
885 #ifdef NS
886 		/*
887 		 * XXX - This code is probably wrong
888 		 */
889 		case AF_NS:
890 		{
891 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
892 			struct arpcom *ac = IFP2AC(ifp);
893 
894 			if (ns_nullhost(*ina))
895 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
896 			else
897 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
898 				      sizeof ac->ac_enaddr);
899 
900 			/*
901 			 * Set new address
902 			 */
903 			ifp->if_init(ifp->if_softc);
904 			break;
905 		}
906 #endif
907 		default:
908 			ifp->if_init(ifp->if_softc);
909 			break;
910 		}
911 		break;
912 
913 	case SIOCGIFADDR:
914 		bcopy(IFP2AC(ifp)->ac_enaddr,
915 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
916 		      ETHER_ADDR_LEN);
917 		break;
918 
919 	case SIOCSIFMTU:
920 		/*
921 		 * Set the interface MTU.
922 		 */
923 		if (ifr->ifr_mtu > ETHERMTU) {
924 			error = EINVAL;
925 		} else {
926 			ifp->if_mtu = ifr->ifr_mtu;
927 		}
928 		break;
929 	default:
930 		error = EINVAL;
931 		break;
932 	}
933 	return (error);
934 }
935 
936 int
937 ether_resolvemulti(
938 	struct ifnet *ifp,
939 	struct sockaddr **llsa,
940 	struct sockaddr *sa)
941 {
942 	struct sockaddr_dl *sdl;
943 	struct sockaddr_in *sin;
944 #ifdef INET6
945 	struct sockaddr_in6 *sin6;
946 #endif
947 	u_char *e_addr;
948 
949 	switch(sa->sa_family) {
950 	case AF_LINK:
951 		/*
952 		 * No mapping needed. Just check that it's a valid MC address.
953 		 */
954 		sdl = (struct sockaddr_dl *)sa;
955 		e_addr = LLADDR(sdl);
956 		if ((e_addr[0] & 1) != 1)
957 			return EADDRNOTAVAIL;
958 		*llsa = 0;
959 		return 0;
960 
961 #ifdef INET
962 	case AF_INET:
963 		sin = (struct sockaddr_in *)sa;
964 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
965 			return EADDRNOTAVAIL;
966 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
967 		       M_WAITOK|M_ZERO);
968 		sdl->sdl_len = sizeof *sdl;
969 		sdl->sdl_family = AF_LINK;
970 		sdl->sdl_index = ifp->if_index;
971 		sdl->sdl_type = IFT_ETHER;
972 		sdl->sdl_alen = ETHER_ADDR_LEN;
973 		e_addr = LLADDR(sdl);
974 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
975 		*llsa = (struct sockaddr *)sdl;
976 		return 0;
977 #endif
978 #ifdef INET6
979 	case AF_INET6:
980 		sin6 = (struct sockaddr_in6 *)sa;
981 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
982 			/*
983 			 * An IP6 address of 0 means listen to all
984 			 * of the Ethernet multicast address used for IP6.
985 			 * (This is used for multicast routers.)
986 			 */
987 			ifp->if_flags |= IFF_ALLMULTI;
988 			*llsa = 0;
989 			return 0;
990 		}
991 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
992 			return EADDRNOTAVAIL;
993 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
994 		       M_WAITOK|M_ZERO);
995 		sdl->sdl_len = sizeof *sdl;
996 		sdl->sdl_family = AF_LINK;
997 		sdl->sdl_index = ifp->if_index;
998 		sdl->sdl_type = IFT_ETHER;
999 		sdl->sdl_alen = ETHER_ADDR_LEN;
1000 		e_addr = LLADDR(sdl);
1001 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1002 		*llsa = (struct sockaddr *)sdl;
1003 		return 0;
1004 #endif
1005 
1006 	default:
1007 		/*
1008 		 * Well, the text isn't quite right, but it's the name
1009 		 * that counts...
1010 		 */
1011 		return EAFNOSUPPORT;
1012 	}
1013 }
1014 
1015 #if 0
1016 /*
1017  * This is for reference.  We have a table-driven version
1018  * of the little-endian crc32 generator, which is faster
1019  * than the double-loop.
1020  */
1021 uint32_t
1022 ether_crc32_le(const uint8_t *buf, size_t len)
1023 {
1024 	uint32_t c, crc, carry;
1025 	size_t i, j;
1026 
1027 	crc = 0xffffffffU;	/* initial value */
1028 
1029 	for (i = 0; i < len; i++) {
1030 		c = buf[i];
1031 		for (j = 0; j < 8; j++) {
1032 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1033 			crc >>= 1;
1034 			c >>= 1;
1035 			if (carry)
1036 				crc = (crc ^ ETHER_CRC_POLY_LE);
1037 		}
1038 	}
1039 
1040 	return (crc);
1041 }
1042 #else
1043 uint32_t
1044 ether_crc32_le(const uint8_t *buf, size_t len)
1045 {
1046 	static const uint32_t crctab[] = {
1047 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1048 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1049 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1050 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1051 	};
1052 	uint32_t crc;
1053 	size_t i;
1054 
1055 	crc = 0xffffffffU;	/* initial value */
1056 
1057 	for (i = 0; i < len; i++) {
1058 		crc ^= buf[i];
1059 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1060 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1061 	}
1062 
1063 	return (crc);
1064 }
1065 #endif
1066 
1067 uint32_t
1068 ether_crc32_be(const uint8_t *buf, size_t len)
1069 {
1070 	uint32_t c, crc, carry;
1071 	size_t i, j;
1072 
1073 	crc = 0xffffffffU;	/* initial value */
1074 
1075 	for (i = 0; i < len; i++) {
1076 		c = buf[i];
1077 		for (j = 0; j < 8; j++) {
1078 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1079 			crc <<= 1;
1080 			c >>= 1;
1081 			if (carry)
1082 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1083 		}
1084 	}
1085 
1086 	return (crc);
1087 }
1088