xref: /dragonfly/sys/net/if_ethersubr.c (revision d5f516c3)
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 1.19 2004/11/08 10:28:48 joerg 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,
80 		struct sockaddr *dst, 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 /* bridge support */
119 int do_bridge;
120 bridge_in_t *bridge_in_ptr;
121 bdg_forward_t *bdg_forward_ptr;
122 bdgtakeifaces_t *bdgtakeifaces_ptr;
123 struct bdg_softc *ifp2sc;
124 
125 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
126 		struct sockaddr *);
127 const uint8_t	etherbroadcastaddr[ETHER_ADDR_LEN] = {
128 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
129 };
130 
131 #define senderr(e) do { error = (e); goto bad;} while (0)
132 #define IFP2AC(IFP) ((struct arpcom *)IFP)
133 
134 int
135 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
136 	struct ip_fw **rule, struct ether_header *eh, int shared);
137 static int ether_ipfw;
138 
139 /*
140  * Ethernet output routine.
141  * Encapsulate a packet of type family for the local net.
142  * Use trailer local net encapsulation if enough data in first
143  * packet leaves a multiple of 512 bytes of data in remainder.
144  * Assumes that ifp is actually pointer to arpcom structure.
145  */
146 static int
147 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
148 	     struct rtentry *rt0)
149 {
150 	short type;
151 	int error = 0, hdrcmplt = 0;
152  	u_char esrc[6], edst[6];
153 	struct rtentry *rt;
154 	struct ether_header *eh;
155 	int loop_copy = 0;
156 	int hlen;	/* link layer header lenght */
157 	struct arpcom *ac = IFP2AC(ifp);
158 
159 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
160 		senderr(ENETDOWN);
161 	rt = rt0;
162 	if (rt) {
163 		if ((rt->rt_flags & RTF_UP) == 0) {
164 			rt0 = rt = rtalloc1(dst, 1, 0UL);
165 			if (rt0)
166 				rt->rt_refcnt--;
167 			else
168 				senderr(EHOSTUNREACH);
169 		}
170 		if (rt->rt_flags & RTF_GATEWAY) {
171 			if (rt->rt_gwroute == 0)
172 				goto lookup;
173 			if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
174 				rtfree(rt); rt = rt0;
175 			lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1,
176 							  0UL);
177 				if ((rt = rt->rt_gwroute) == 0)
178 					senderr(EHOSTUNREACH);
179 			}
180 		}
181 		if (rt->rt_flags & RTF_REJECT)
182 			if (rt->rt_rmx.rmx_expire == 0 ||
183 			    time_second < rt->rt_rmx.rmx_expire)
184 				senderr(rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
185 	}
186 	hlen = ETHER_HDR_LEN;
187 	switch (dst->sa_family) {
188 #ifdef INET
189 	case AF_INET:
190 		if (!arpresolve(ifp, rt, m, dst, edst, rt0))
191 			return (0);	/* if not yet resolved */
192 		type = htons(ETHERTYPE_IP);
193 		break;
194 #endif
195 #ifdef INET6
196 	case AF_INET6:
197 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, (u_char *)edst)) {
198 			/* Something bad happened */
199 			return(0);
200 		}
201 		type = htons(ETHERTYPE_IPV6);
202 		break;
203 #endif
204 #ifdef IPX
205 	case AF_IPX:
206 		if (ef_outputp) {
207 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
208 		    if (error)
209 			goto bad;
210 		} else
211 		    type = htons(ETHERTYPE_IPX);
212  		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
213 		    (caddr_t)edst, sizeof (edst));
214 		break;
215 #endif
216 #ifdef NETATALK
217 	case AF_APPLETALK:
218 	  {
219 	    struct at_ifaddr *aa;
220 
221 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
222 		    goto bad;
223 	    }
224 	    if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
225 		    return (0);
226 	    /*
227 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
228 	     * Since we must preserve the value of m, which is passed to us by
229 	     * value, we m_copy() the first mbuf, and use it for our llc header.
230 	     */
231 	    if ( aa->aa_flags & AFA_PHASE2 ) {
232 		struct llc llc;
233 
234 		M_PREPEND(m, sizeof(struct llc), MB_WAIT);
235 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
236 		llc.llc_control = LLC_UI;
237 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
238 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
239 		bcopy(&llc, mtod(m, caddr_t), sizeof(struct llc));
240 		type = htons(m->m_pkthdr.len);
241 		hlen = sizeof(struct llc) + ETHER_HDR_LEN;
242 	    } else {
243 		type = htons(ETHERTYPE_AT);
244 	    }
245 	    break;
246 	  }
247 #endif /* NETATALK */
248 #ifdef NS
249 	case AF_NS:
250 		switch(ns_nettype){
251 		default:
252 		case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
253 			type = 0x8137;
254 			break;
255 		case 0x0: /* Novell 802.3 */
256 			type = htons( m->m_pkthdr.len);
257 			break;
258 		case 0xe0e0: /* Novell 802.2 and Token-Ring */
259 			M_PREPEND(m, 3, MB_WAIT);
260 			type = htons( m->m_pkthdr.len);
261 			cp = mtod(m, u_char *);
262 			*cp++ = 0xE0;
263 			*cp++ = 0xE0;
264 			*cp++ = 0x03;
265 			break;
266 		}
267  		bcopy((caddr_t)&(((struct sockaddr_ns *)dst)->sns_addr.x_host),
268 		    (caddr_t)edst, sizeof (edst));
269 		/*
270 		 * XXX if ns_thishost is the same as the node's ethernet
271 		 * address then just the default code will catch this anyhow.
272 		 * So I'm not sure if this next clause should be here at all?
273 		 * [JRE]
274 		 */
275 		if (!bcmp((caddr_t)edst, (caddr_t)&ns_thishost, sizeof(edst))){
276 			m->m_pkthdr.rcvif = ifp;
277 			netisr_dispatch(NETISR_NS, m);
278 			return (error);
279 		}
280 		if (!bcmp((caddr_t)edst, (caddr_t)&ns_broadhost, sizeof(edst))){
281 			m->m_flags |= M_BCAST;
282 		}
283 		break;
284 #endif /* NS */
285 
286 	case pseudo_AF_HDRCMPLT:
287 		hdrcmplt = 1;
288 		eh = (struct ether_header *)dst->sa_data;
289 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
290 		/* FALLTHROUGH */
291 
292 	case AF_UNSPEC:
293 		loop_copy = -1; /* if this is for us, don't do it */
294 		eh = (struct ether_header *)dst->sa_data;
295  		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
296 		type = eh->ether_type;
297 		break;
298 
299 	default:
300 		printf("%s: can't handle af%d\n", ifp->if_xname,
301 			dst->sa_family);
302 		senderr(EAFNOSUPPORT);
303 	}
304 
305 	/*
306 	 * Add local net header.  If no space in first mbuf,
307 	 * allocate another.
308 	 */
309 	M_PREPEND(m, sizeof (struct ether_header), MB_DONTWAIT);
310 	if (m == 0)
311 		senderr(ENOBUFS);
312 	eh = mtod(m, struct ether_header *);
313 	(void)memcpy(&eh->ether_type, &type,
314 		sizeof(eh->ether_type));
315  	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
316 	if (hdrcmplt)
317 		(void)memcpy(eh->ether_shost, esrc,
318 			sizeof(eh->ether_shost));
319 	else
320 		(void)memcpy(eh->ether_shost, ac->ac_enaddr,
321 			sizeof(eh->ether_shost));
322 
323 	/*
324 	 * If a simplex interface, and the packet is being sent to our
325 	 * Ethernet address or a broadcast address, loopback a copy.
326 	 * XXX To make a simplex device behave exactly like a duplex
327 	 * device, we should copy in the case of sending to our own
328 	 * ethernet address (thus letting the original actually appear
329 	 * on the wire). However, we don't do that here for security
330 	 * reasons and compatibility with the original behavior.
331 	 */
332 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
333 		int csum_flags = 0;
334 
335 		if (m->m_pkthdr.csum_flags & CSUM_IP)
336 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
337 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
338 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
339 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
340 			struct mbuf *n;
341 
342 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
343 				n->m_pkthdr.csum_flags |= csum_flags;
344 				if (csum_flags & CSUM_DATA_VALID)
345 					n->m_pkthdr.csum_data = 0xffff;
346 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
347 			} else
348 				ifp->if_iqdrops++;
349 		} else if (bcmp(eh->ether_dhost,
350 		    eh->ether_shost, ETHER_ADDR_LEN) == 0) {
351 			m->m_pkthdr.csum_flags |= csum_flags;
352 			if (csum_flags & CSUM_DATA_VALID)
353 				m->m_pkthdr.csum_data = 0xffff;
354 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
355 			return (0);	/* XXX */
356 		}
357 	}
358 
359 	/* Handle ng_ether(4) processing, if any */
360 	if (ng_ether_output_p != NULL) {
361 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
362 bad:			if (m != NULL)
363 				m_freem(m);
364 			return (error);
365 		}
366 		if (m == NULL)
367 			return (0);
368 	}
369 
370 	/* Continue with link-layer output */
371 	return ether_output_frame(ifp, m);
372 }
373 
374 /*
375  * Ethernet link layer output routine to send a raw frame to the device.
376  *
377  * This assumes that the 14 byte Ethernet header is present and contiguous
378  * in the first mbuf (if BRIDGE'ing).
379  */
380 int
381 ether_output_frame(ifp, m)
382 	struct ifnet *ifp;
383 	struct mbuf *m;
384 {
385 	int error = 0;
386 	int s;
387 	struct ip_fw *rule = NULL;
388 
389 	/* Extract info from dummynet tag, ignore others */
390 	for (; m->m_type == MT_TAG; m = m->m_next)
391 		if (m->m_flags == PACKET_TAG_DUMMYNET)
392 			rule = ((struct dn_pkt *)m)->rule;
393 
394 	if (rule)	/* packet was already bridged */
395 		goto no_bridge;
396 
397 	if (BDG_ACTIVE(ifp) ) {
398 		struct ether_header *eh; /* a ptr suffices */
399 
400 		m->m_pkthdr.rcvif = NULL;
401 		eh = mtod(m, struct ether_header *);
402 		m_adj(m, ETHER_HDR_LEN);
403 		m = bdg_forward_ptr(m, eh, ifp);
404 		if (m != NULL)
405 			m_freem(m);
406 		return (0);
407 	}
408 
409 no_bridge:
410 	s = splimp();
411 	if (IPFW_LOADED && ether_ipfw != 0) {
412 		struct ether_header save_eh, *eh;
413 
414 		eh = mtod(m, struct ether_header *);
415 		save_eh = *eh;
416 		m_adj(m, ETHER_HDR_LEN);
417 		if (ether_ipfw_chk(&m, ifp, &rule, eh, 0) == 0) {
418 			if (m) {
419 				m_freem(m);
420 				return ENOBUFS;	/* pkt dropped */
421 			} else
422 				return 0;	/* consumed e.g. in a pipe */
423 		}
424 		/* packet was ok, restore the ethernet header */
425 		if ( (void *)(eh + 1) == (void *)m->m_data) {
426 			m->m_data -= ETHER_HDR_LEN ;
427 			m->m_len += ETHER_HDR_LEN ;
428 			m->m_pkthdr.len += ETHER_HDR_LEN ;
429 		} else {
430 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
431 			if (m == NULL) /* nope... */
432 				return ENOBUFS;
433 			bcopy(&save_eh, mtod(m, struct ether_header *),
434 			    ETHER_HDR_LEN);
435 		}
436 	}
437 
438 	/*
439 	 * Queue message on interface, update output statistics if
440 	 * successful, and start output if interface not yet active.
441 	 */
442 	if (!IF_HANDOFF(&ifp->if_snd, m, ifp))
443 		error = ENOBUFS;
444 	splx(s);
445 	return (error);
446 }
447 
448 /*
449  * ipfw processing for ethernet packets (in and out).
450  * The second parameter is NULL from ether_demux, and ifp from
451  * ether_output_frame. This section of code could be used from
452  * bridge.c as well as long as we use some extra info
453  * to distinguish that case from ether_output_frame();
454  */
455 int
456 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
457 	struct ip_fw **rule, struct ether_header *eh, int shared)
458 {
459 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
460 	int i;
461 	struct ip_fw_args args;
462 
463 	if (*rule != NULL && fw_one_pass)
464 		return 1; /* dummynet packet, already partially processed */
465 
466 	/*
467 	 * I need some amt of data to be contiguous, and in case others need
468 	 * the packet (shared==1) also better be in the first mbuf.
469 	 */
470 	i = min( (*m0)->m_pkthdr.len, max_protohdr);
471 	if ( shared || (*m0)->m_len < i) {
472 		*m0 = m_pullup(*m0, i);
473 		if (*m0 == NULL)
474 			return 0;
475 	}
476 
477 	args.m = *m0;		/* the packet we are looking at		*/
478 	args.oif = dst;		/* destination, if any			*/
479 	args.divert_rule = 0;	/* we do not support divert yet		*/
480 	args.rule = *rule;	/* matching rule to restart		*/
481 	args.next_hop = NULL;	/* we do not support forward yet	*/
482 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
483 	i = ip_fw_chk_ptr(&args);
484 	*m0 = args.m;
485 	*rule = args.rule;
486 
487 	if ( (i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL) /* drop */
488 		return 0;
489 
490 	if (i == 0) /* a PASS rule.  */
491 		return 1;
492 
493 	if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
494 		/*
495 		 * Pass the pkt to dummynet, which consumes it.
496 		 * If shared, make a copy and keep the original.
497 		 */
498 		struct mbuf *m ;
499 
500 		if (shared) {
501 			m = m_copypacket(*m0, MB_DONTWAIT);
502 			if (m == NULL)
503 				return 0;
504 		} else {
505 			m = *m0 ; /* pass the original to dummynet */
506 			*m0 = NULL ; /* and nothing back to the caller */
507 		}
508 		/*
509 		 * Prepend the header, optimize for the common case of
510 		 * eh pointing into the mbuf.
511 		 */
512 		if ( (void *)(eh + 1) == (void *)m->m_data) {
513 			m->m_data -= ETHER_HDR_LEN ;
514 			m->m_len += ETHER_HDR_LEN ;
515 			m->m_pkthdr.len += ETHER_HDR_LEN ;
516 		} else {
517 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
518 			if (m == NULL) /* nope... */
519 				return 0;
520 			bcopy(&save_eh, mtod(m, struct ether_header *),
521 			    ETHER_HDR_LEN);
522 		}
523 		ip_dn_io_ptr(m, (i & 0xffff),
524 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
525 		return 0;
526 	}
527 	/*
528 	 * XXX at some point add support for divert/forward actions.
529 	 * If none of the above matches, we have to drop the pkt.
530 	 */
531 	return 0;
532 }
533 
534 /*
535  * XXX merge this function with ether_input.
536  */
537 static void
538 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
539 {
540 	ether_input(ifp, NULL, m);
541 }
542 
543 /*
544  * Process a received Ethernet packet. We have two different interfaces:
545  * one (conventional) assumes the packet in the mbuf, with the ethernet
546  * header provided separately in *eh. The second one (new) has everything
547  * in the mbuf, and we can tell it because eh == NULL.
548  * The caller MUST MAKE SURE that there are at least
549  * sizeof(struct ether_header) bytes in the first mbuf.
550  *
551  * This allows us to concentrate in one place a bunch of code which
552  * is replicated in all device drivers. Also, many functions called
553  * from ether_input() try to put the eh back into the mbuf, so we
554  * can later propagate the 'contiguous packet' interface to them,
555  * and handle the old interface just here.
556  *
557  * NOTA BENE: for many drivers "eh" is a pointer into the first mbuf or
558  * cluster, right before m_data. So be very careful when working on m,
559  * as you could destroy *eh !!
560  *
561  * First we perform any link layer operations, then continue
562  * to the upper layers with ether_demux().
563  */
564 void
565 ether_input(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
566 {
567 	struct ether_header save_eh;
568 
569 	if (eh == NULL) {
570 		if (m->m_len < sizeof(struct ether_header)) {
571 			/* XXX error in the caller. */
572 			m_freem(m);
573 			return;
574 		}
575 		m->m_pkthdr.rcvif = ifp;
576 		eh = mtod(m, struct ether_header *);
577 		m_adj(m, sizeof(struct ether_header));
578 		/* XXX */
579 		/* m->m_pkthdr.len = m->m_len; */
580 	}
581 
582 	/* Check for a BPF tap */
583 	if (ifp->if_bpf != NULL) {
584 		struct m_hdr mh;
585 
586 		/* This kludge is OK; BPF treats the "mbuf" as read-only */
587 		mh.mh_next = m;
588 		mh.mh_data = (char *)eh;
589 		mh.mh_len = ETHER_HDR_LEN;
590 		bpf_mtap(ifp, (struct mbuf *)&mh);
591 	}
592 
593 	ifp->if_ibytes += m->m_pkthdr.len + sizeof (*eh);
594 
595 	/* Handle ng_ether(4) processing, if any */
596 	if (ng_ether_input_p != NULL) {
597 		(*ng_ether_input_p)(ifp, &m, eh);
598 		if (m == NULL)
599 			return;
600 	}
601 
602 	/* Check for bridging mode */
603 	if (BDG_ACTIVE(ifp) ) {
604 		struct ifnet *bif;
605 
606 		/* Check with bridging code */
607 		if ((bif = bridge_in_ptr(ifp, eh)) == BDG_DROP) {
608 			m_freem(m);
609 			return;
610 		}
611 		if (bif != BDG_LOCAL) {
612 			save_eh = *eh ; /* because it might change */
613 			m = bdg_forward_ptr(m, eh, bif); /* needs forwarding */
614 			/*
615 			 * Do not continue if bdg_forward_ptr() processed our
616 			 * packet (and cleared the mbuf pointer m) or if
617 			 * it dropped (m_free'd) the packet itself.
618 			 */
619 			if (m == NULL) {
620 			    if (bif == BDG_BCAST || bif == BDG_MCAST)
621 				printf("bdg_forward drop MULTICAST PKT\n");
622 			    return;
623 			}
624 			eh = &save_eh ;
625 		}
626 		if (bif == BDG_LOCAL
627 		    || bif == BDG_BCAST
628 		    || bif == BDG_MCAST)
629 			goto recvLocal;			/* receive locally */
630 
631 		/* If not local and not multicast, just drop it */
632 		if (m != NULL)
633 			m_freem(m);
634 		return;
635        }
636 
637 recvLocal:
638 	/* Continue with upper layer processing */
639 	ether_demux(ifp, eh, m);
640 }
641 
642 /*
643  * Upper layer processing for a received Ethernet packet.
644  */
645 void
646 ether_demux(ifp, eh, m)
647 	struct ifnet *ifp;
648 	struct ether_header *eh;
649 	struct mbuf *m;
650 {
651 	int isr;
652 	u_short ether_type;
653 #if defined(NETATALK)
654 	struct llc *l;
655 #endif
656 	struct ip_fw *rule = NULL;
657 
658 	/* Extract info from dummynet tag, ignore others */
659 	for (;m->m_type == MT_TAG; m = m->m_next)
660 		if (m->m_flags == PACKET_TAG_DUMMYNET) {
661 			rule = ((struct dn_pkt *)m)->rule;
662 			ifp = m->m_next->m_pkthdr.rcvif;
663 		}
664 
665 	if (rule)	/* packet was already bridged */
666 		goto post_stats;
667 
668     if (! (BDG_ACTIVE(ifp) ) )
669 	/* Discard packet if upper layers shouldn't see it because it was
670 	   unicast to a different Ethernet address. If the driver is working
671 	   properly, then this situation can only happen when the interface
672 	   is in promiscuous mode. */
673 	if ((ifp->if_flags & IFF_PROMISC) != 0
674 	    && (eh->ether_dhost[0] & 1) == 0
675 	    && bcmp(eh->ether_dhost,
676 	      IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN) != 0
677 	    && (ifp->if_flags & IFF_PPROMISC) == 0) {
678 		m_freem(m);
679 		return;
680 	}
681 
682 	/* Discard packet if interface is not up */
683 	if ((ifp->if_flags & IFF_UP) == 0) {
684 		m_freem(m);
685 		return;
686 	}
687 	if (eh->ether_dhost[0] & 1) {
688 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
689 			 ifp->if_addrlen) == 0)
690 			m->m_flags |= M_BCAST;
691 		else
692 			m->m_flags |= M_MCAST;
693 	}
694 	if (m->m_flags & (M_BCAST|M_MCAST))
695 		ifp->if_imcasts++;
696 
697 post_stats:
698 	if (IPFW_LOADED && ether_ipfw != 0) {
699 		if (ether_ipfw_chk(&m, NULL, &rule, eh, 0 ) == 0) {
700 			if (m)
701 				m_freem(m);
702 			return;
703 		}
704 	}
705 
706 	ether_type = ntohs(eh->ether_type);
707 
708 	switch (ether_type) {
709 #ifdef INET
710 	case ETHERTYPE_IP:
711 		if (ipflow_fastforward(m))
712 			return;
713 		isr = NETISR_IP;
714 		break;
715 
716 	case ETHERTYPE_ARP:
717 		if (ifp->if_flags & IFF_NOARP) {
718 			/* Discard packet if ARP is disabled on interface */
719 			m_freem(m);
720 			return;
721 		}
722 		isr = NETISR_ARP;
723 		break;
724 #endif
725 #ifdef IPX
726 	case ETHERTYPE_IPX:
727 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
728 			return;
729 		isr = NETISR_IPX;
730 		break;
731 #endif
732 #ifdef INET6
733 	case ETHERTYPE_IPV6:
734 		isr = NETISR_IPV6;
735 		break;
736 #endif
737 #ifdef NS
738 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
739 		isr = NETISR_NS;
740 		break;
741 
742 #endif /* NS */
743 #ifdef NETATALK
744         case ETHERTYPE_AT:
745 		isr = NETISR_ATALK1;
746                 break;
747         case ETHERTYPE_AARP:
748 		isr = NETISR_AARP;
749                 break;
750 #endif /* NETATALK */
751 	case ETHERTYPE_VLAN:
752 		/* XXX lock ? */
753 		if (vlan_input_p != NULL)
754 			(*vlan_input_p)(eh, m);
755 		else {
756 			m->m_pkthdr.rcvif->if_noproto++;
757 			m_freem(m);
758 		}
759 		/* XXX unlock ? */
760 		return;
761 	default:
762 #ifdef IPX
763 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
764 			return;
765 #endif /* IPX */
766 #ifdef NS
767 		checksum = mtod(m, ushort *);
768 		/* Novell 802.3 */
769 		if ((ether_type <= ETHERMTU) &&
770 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
771 			if (*checksum == 0xE0E0) {
772 				m->m_pkthdr.len -= 3;
773 				m->m_len -= 3;
774 				m->m_data += 3;
775 			}
776 			isr = NETISR_NS;
777 			break;
778 		}
779 #endif /* NS */
780 #ifdef NETATALK
781 		if (ether_type > ETHERMTU)
782 			goto dropanyway;
783 		l = mtod(m, struct llc *);
784 		if (l->llc_dsap == LLC_SNAP_LSAP &&
785 		    l->llc_ssap == LLC_SNAP_LSAP &&
786 		    l->llc_control == LLC_UI) {
787                         if (Bcmp(&(l->llc_snap_org_code)[0], at_org_code,
788 			    sizeof(at_org_code)) == 0 &&
789 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
790 				m_adj(m, sizeof(struct llc));
791 				isr = NETISR_ATALK2;
792 				break;
793 			}
794 			if (Bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
795 			    sizeof(aarp_org_code)) == 0 &&
796 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
797 				m_adj(m, sizeof(struct llc));
798 				isr = NETISR_AARP;
799 				break;
800 			}
801 		}
802 dropanyway:
803 #endif /* NETATALK */
804 		if (ng_ether_input_orphan_p != NULL)
805 			(*ng_ether_input_orphan_p)(ifp, m, eh);
806 		else
807 			m_freem(m);
808 		return;
809 	}
810 	netisr_dispatch(isr, m);
811 }
812 
813 /*
814  * Perform common duties while attaching to interface list
815  */
816 
817 void
818 ether_ifattach(struct ifnet *ifp, uint8_t *lla)
819 {
820 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header));
821 }
822 
823 void
824 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen)
825 {
826 	struct ifaddr *ifa;
827 	struct sockaddr_dl *sdl;
828 
829 	ifp->if_output = ether_output;
830 	ifp->if_input = ether_input_internal;
831 	ifp->if_type = IFT_ETHER;
832 	ifp->if_addrlen = ETHER_ADDR_LEN;
833 	ifp->if_broadcastaddr = etherbroadcastaddr;
834 	ifp->if_hdrlen = 14;
835 	if_attach(ifp);
836 	ifp->if_mtu = ETHERMTU;
837 	ifp->if_resolvemulti = ether_resolvemulti;
838 	if (ifp->if_baudrate == 0)
839 	    ifp->if_baudrate = 10000000;
840 	ifa = ifnet_addrs[ifp->if_index - 1];
841 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __FUNCTION__));
842 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
843 	sdl->sdl_type = IFT_ETHER;
844 	sdl->sdl_alen = ifp->if_addrlen;
845 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
846 	/*
847 	 * XXX Keep the current drivers happy.
848 	 * XXX Remove once all drivers have been cleaned up
849 	 */
850 	if (lla != IFP2AC(ifp)->ac_enaddr)
851 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
852 	bpfattach(ifp, dlt, hdrlen);
853 	if (ng_ether_attach_p != NULL)
854 		(*ng_ether_attach_p)(ifp);
855 	if (BDG_LOADED)
856 		bdgtakeifaces_ptr();
857 
858 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
859 }
860 
861 /*
862  * Perform common duties while detaching an Ethernet interface
863  */
864 void
865 ether_ifdetach(struct ifnet *ifp)
866 {
867 	int s;
868 
869 	s = splnet();
870 	if_down(ifp);
871 	splx(s);
872 
873 	if (ng_ether_detach_p != NULL)
874 		(*ng_ether_detach_p)(ifp);
875 	bpfdetach(ifp);
876 	if_detach(ifp);
877 	if (BDG_LOADED)
878 		bdgtakeifaces_ptr();
879 }
880 
881 SYSCTL_DECL(_net_link);
882 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
883 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
884 	    &ether_ipfw,0,"Pass ether pkts through firewall");
885 
886 int
887 ether_ioctl(ifp, command, data)
888 	struct ifnet *ifp;
889 	int command;
890 	caddr_t data;
891 {
892 	struct ifaddr *ifa = (struct ifaddr *) data;
893 	struct ifreq *ifr = (struct ifreq *) data;
894 	int error = 0;
895 
896 	switch (command) {
897 	case SIOCSIFADDR:
898 		ifp->if_flags |= IFF_UP;
899 
900 		switch (ifa->ifa_addr->sa_family) {
901 #ifdef INET
902 		case AF_INET:
903 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
904 			arp_ifinit(ifp, ifa);
905 			break;
906 #endif
907 #ifdef IPX
908 		/*
909 		 * XXX - This code is probably wrong
910 		 */
911 		case AF_IPX:
912 			{
913 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
914 			struct arpcom *ac = IFP2AC(ifp);
915 
916 			if (ipx_nullhost(*ina))
917 				ina->x_host =
918 				    *(union ipx_host *)
919 			            ac->ac_enaddr;
920 			else {
921 				bcopy((caddr_t) ina->x_host.c_host,
922 				      (caddr_t) ac->ac_enaddr,
923 				      sizeof(ac->ac_enaddr));
924 			}
925 
926 			/*
927 			 * Set new address
928 			 */
929 			ifp->if_init(ifp->if_softc);
930 			break;
931 			}
932 #endif
933 #ifdef NS
934 		/*
935 		 * XXX - This code is probably wrong
936 		 */
937 		case AF_NS:
938 		{
939 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
940 			struct arpcom *ac = IFP2AC(ifp);
941 
942 			if (ns_nullhost(*ina))
943 				ina->x_host =
944 				    *(union ns_host *) (ac->ac_enaddr);
945 			else {
946 				bcopy((caddr_t) ina->x_host.c_host,
947 				      (caddr_t) ac->ac_enaddr,
948 				      sizeof(ac->ac_enaddr));
949 			}
950 
951 			/*
952 			 * Set new address
953 			 */
954 			ifp->if_init(ifp->if_softc);
955 			break;
956 		}
957 #endif
958 		default:
959 			ifp->if_init(ifp->if_softc);
960 			break;
961 		}
962 		break;
963 
964 	case SIOCGIFADDR:
965 		{
966 			struct sockaddr *sa;
967 
968 			sa = (struct sockaddr *) & ifr->ifr_data;
969 			bcopy(IFP2AC(ifp)->ac_enaddr,
970 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
971 		}
972 		break;
973 
974 	case SIOCSIFMTU:
975 		/*
976 		 * Set the interface MTU.
977 		 */
978 		if (ifr->ifr_mtu > ETHERMTU) {
979 			error = EINVAL;
980 		} else {
981 			ifp->if_mtu = ifr->ifr_mtu;
982 		}
983 		break;
984 	default:
985 		error = EINVAL;
986 		break;
987 	}
988 	return (error);
989 }
990 
991 int
992 ether_resolvemulti(ifp, llsa, sa)
993 	struct ifnet *ifp;
994 	struct sockaddr **llsa;
995 	struct sockaddr *sa;
996 {
997 	struct sockaddr_dl *sdl;
998 	struct sockaddr_in *sin;
999 #ifdef INET6
1000 	struct sockaddr_in6 *sin6;
1001 #endif
1002 	u_char *e_addr;
1003 
1004 	switch(sa->sa_family) {
1005 	case AF_LINK:
1006 		/*
1007 		 * No mapping needed. Just check that it's a valid MC address.
1008 		 */
1009 		sdl = (struct sockaddr_dl *)sa;
1010 		e_addr = LLADDR(sdl);
1011 		if ((e_addr[0] & 1) != 1)
1012 			return EADDRNOTAVAIL;
1013 		*llsa = 0;
1014 		return 0;
1015 
1016 #ifdef INET
1017 	case AF_INET:
1018 		sin = (struct sockaddr_in *)sa;
1019 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1020 			return EADDRNOTAVAIL;
1021 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1022 		       M_WAITOK|M_ZERO);
1023 		sdl->sdl_len = sizeof *sdl;
1024 		sdl->sdl_family = AF_LINK;
1025 		sdl->sdl_index = ifp->if_index;
1026 		sdl->sdl_type = IFT_ETHER;
1027 		sdl->sdl_alen = ETHER_ADDR_LEN;
1028 		e_addr = LLADDR(sdl);
1029 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1030 		*llsa = (struct sockaddr *)sdl;
1031 		return 0;
1032 #endif
1033 #ifdef INET6
1034 	case AF_INET6:
1035 		sin6 = (struct sockaddr_in6 *)sa;
1036 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1037 			/*
1038 			 * An IP6 address of 0 means listen to all
1039 			 * of the Ethernet multicast address used for IP6.
1040 			 * (This is used for multicast routers.)
1041 			 */
1042 			ifp->if_flags |= IFF_ALLMULTI;
1043 			*llsa = 0;
1044 			return 0;
1045 		}
1046 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1047 			return EADDRNOTAVAIL;
1048 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1049 		       M_WAITOK|M_ZERO);
1050 		sdl->sdl_len = sizeof *sdl;
1051 		sdl->sdl_family = AF_LINK;
1052 		sdl->sdl_index = ifp->if_index;
1053 		sdl->sdl_type = IFT_ETHER;
1054 		sdl->sdl_alen = ETHER_ADDR_LEN;
1055 		e_addr = LLADDR(sdl);
1056 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1057 		*llsa = (struct sockaddr *)sdl;
1058 		return 0;
1059 #endif
1060 
1061 	default:
1062 		/*
1063 		 * Well, the text isn't quite right, but it's the name
1064 		 * that counts...
1065 		 */
1066 		return EAFNOSUPPORT;
1067 	}
1068 }
1069 
1070 #if 0
1071 /*
1072  * This is for reference.  We have a table-driven version
1073  * of the little-endian crc32 generator, which is faster
1074  * than the double-loop.
1075  */
1076 uint32_t
1077 ether_crc32_le(const uint8_t *buf, size_t len)
1078 {
1079 	uint32_t c, crc, carry;
1080 	size_t i, j;
1081 
1082 	crc = 0xffffffffU;	/* initial value */
1083 
1084 	for (i = 0; i < len; i++) {
1085 		c = buf[i];
1086 		for (j = 0; j < 8; j++) {
1087 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1088 			crc >>= 1;
1089 			c >>= 1;
1090 			if (carry)
1091 				crc = (crc ^ ETHER_CRC_POLY_LE);
1092 		}
1093 	}
1094 
1095 	return(crc);
1096 }
1097 #else
1098 uint32_t
1099 ether_crc32_le(const uint8_t *buf, size_t len)
1100 {
1101 	static const uint32_t crctab[] = {
1102 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1103 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1104 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1105 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1106 	};
1107 	uint32_t crc;
1108 	size_t i;
1109 
1110 	crc = 0xffffffffU;	/* initial value */
1111 
1112 	for (i = 0; i < len; i++) {
1113 		crc ^= buf[i];
1114 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1115 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1116 	}
1117 
1118 	return(crc);
1119 }
1120 #endif
1121 
1122 uint32_t
1123 ether_crc32_be(const uint8_t *buf, size_t len)
1124 {
1125 	uint32_t c, crc, carry;
1126 	size_t i, j;
1127 
1128 	crc = 0xffffffffU;	/* initial value */
1129 
1130 	for (i = 0; i < len; i++) {
1131 		c = buf[i];
1132 		for (j = 0; j < 8; j++) {
1133 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1134 			crc <<= 1;
1135 			c >>= 1;
1136 			if (carry)
1137 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1138 		}
1139 	}
1140 
1141 	return(crc);
1142 }
1143