xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 15a56cb3807bfc9539b6ac36cb59d42bd9af9659)
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 1.33 2005/11/28 17:13:45 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/ifq_var.h>
61 #include <net/bpf.h>
62 #include <net/ethernet.h>
63 #include <net/bridge/bridge.h>
64 
65 #if defined(INET) || defined(INET6)
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/if_ether.h>
69 #include <net/ipfw/ip_fw.h>
70 #include <net/dummynet/ip_dummynet.h>
71 #endif
72 #ifdef INET6
73 #include <netinet6/nd6.h>
74 #endif
75 
76 #ifdef IPX
77 #include <netproto/ipx/ipx.h>
78 #include <netproto/ipx/ipx_if.h>
79 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
80 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
81 		  short *tp, int *hlen);
82 #endif
83 
84 #ifdef NS
85 #include <netns/ns.h>
86 #include <netns/ns_if.h>
87 ushort ns_nettype;
88 int ether_outputdebug = 0;
89 int ether_inputdebug = 0;
90 #endif
91 
92 #ifdef NETATALK
93 #include <netproto/atalk/at.h>
94 #include <netproto/atalk/at_var.h>
95 #include <netproto/atalk/at_extern.h>
96 
97 #define	llc_snap_org_code	llc_un.type_snap.org_code
98 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
99 
100 extern u_char	at_org_code[3];
101 extern u_char	aarp_org_code[3];
102 #endif /* NETATALK */
103 
104 /* netgraph node hooks for ng_ether(4) */
105 void	(*ng_ether_input_p)(struct ifnet *ifp,
106 		struct mbuf **mp, struct ether_header *eh);
107 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
108 		struct mbuf *m, struct ether_header *eh);
109 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
110 void	(*ng_ether_attach_p)(struct ifnet *ifp);
111 void	(*ng_ether_detach_p)(struct ifnet *ifp);
112 
113 int	(*vlan_input_p)(struct ether_header *eh, struct mbuf *m);
114 int	(*vlan_input_tag_p)(struct mbuf *m, uint16_t t);
115 
116 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
117 			struct rtentry *);
118 
119 /*
120  * bridge support
121  */
122 int do_bridge;
123 bridge_in_t *bridge_in_ptr;
124 bdg_forward_t *bdg_forward_ptr;
125 bdgtakeifaces_t *bdgtakeifaces_ptr;
126 struct bdg_softc *ifp2sc;
127 
128 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
129 			      struct sockaddr *);
130 
131 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
132 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
133 };
134 
135 #define gotoerr(e) do { error = (e); goto bad; } while (0)
136 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
137 
138 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
139 				struct ip_fw **rule, struct ether_header *eh,
140 				boolean_t shared);
141 
142 static int ether_ipfw;
143 SYSCTL_DECL(_net_link);
144 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
145 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
146 	   &ether_ipfw, 0, "Pass ether pkts through firewall");
147 
148 /*
149  * Ethernet output routine.
150  * Encapsulate a packet of type family for the local net.
151  * Use trailer local net encapsulation if enough data in first
152  * packet leaves a multiple of 512 bytes of data in remainder.
153  * Assumes that ifp is actually pointer to arpcom structure.
154  */
155 static int
156 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
157 	     struct rtentry *rt)
158 {
159 	struct ether_header *eh, *deh;
160 	u_char *edst;
161 	int loop_copy = 0;
162 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
163 	struct arpcom *ac = IFP2AC(ifp);
164 	int error;
165 
166 	ASSERT_SERIALIZED(ifp->if_serializer);
167 
168 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
169 		gotoerr(ENETDOWN);
170 
171 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
172 	if (m == NULL)
173 		return (ENOBUFS);
174 	eh = mtod(m, struct ether_header *);
175 	edst = eh->ether_dhost;
176 
177 	/*
178 	 * Fill in the destination ethernet address and frame type.
179 	 */
180 	switch (dst->sa_family) {
181 #ifdef INET
182 	case AF_INET:
183 		if (!arpresolve(ifp, rt, m, dst, edst))
184 			return (0);	/* if not yet resolved */
185 		eh->ether_type = htons(ETHERTYPE_IP);
186 		break;
187 #endif
188 #ifdef INET6
189 	case AF_INET6:
190 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
191 			return (0);		/* Something bad happenned. */
192 		eh->ether_type = htons(ETHERTYPE_IPV6);
193 		break;
194 #endif
195 #ifdef IPX
196 	case AF_IPX:
197 		if (ef_outputp != NULL) {
198 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
199 					   &hlen);
200 			if (error)
201 				goto bad;
202 		} else {
203 			eh->ether_type = htons(ETHERTYPE_IPX);
204 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
205 			      edst, ETHER_ADDR_LEN);
206 		}
207 		break;
208 #endif
209 #ifdef NETATALK
210 	case AF_APPLETALK: {
211 		struct at_ifaddr *aa;
212 
213 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
214 			error = 0;	/* XXX */
215 			goto bad;
216 		}
217 		/*
218 		 * In the phase 2 case, need to prepend an mbuf for
219 		 * the llc header.  Since we must preserve the value
220 		 * of m, which is passed to us by value, we m_copy()
221 		 * the first mbuf, and use it for our llc header.
222 		 */
223 		if (aa->aa_flags & AFA_PHASE2) {
224 			struct llc llc;
225 
226 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
227 			eh = mtod(m, struct ether_header *);
228 			edst = eh->ether_dhost;
229 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
230 			llc.llc_control = LLC_UI;
231 			bcopy(at_org_code, llc.llc_snap_org_code,
232 			      sizeof at_org_code);
233 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
234 			bcopy(&llc,
235 			      mtod(m, caddr_t) + sizeof(struct ether_header),
236 			      sizeof(struct llc));
237 			eh->ether_type = htons(m->m_pkthdr.len);
238 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
239 		} else {
240 			eh->ether_type = htons(ETHERTYPE_AT);
241 		}
242 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
243 			return (0);
244 		break;
245 	  }
246 #endif
247 #ifdef NS
248 	case AF_NS:
249 		switch(ns_nettype) {
250 		default:
251 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
252 			eh->ether_type = 0x8137;
253 			break;
254 		case 0x0:	/* Novell 802.3 */
255 			eh->ether_type = htons(m->m_pkthdr.len);
256 			break;
257 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
258 			M_PREPEND(m, 3, MB_DONTWAIT);
259 			eh = mtod(m, struct ether_header *);
260 			edst = eh->ether_dhost;
261 			eh->ether_type = htons(m->m_pkthdr.len);
262 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
263 			*cp++ = 0xE0;
264 			*cp++ = 0xE0;
265 			*cp++ = 0x03;
266 			break;
267 		}
268 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
269 		      ETHER_ADDR_LEN);
270 		/*
271 		 * XXX if ns_thishost is the same as the node's ethernet
272 		 * address then just the default code will catch this anyhow.
273 		 * So I'm not sure if this next clause should be here at all?
274 		 * [JRE]
275 		 */
276 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
277 			m->m_pkthdr.rcvif = ifp;
278 			netisr_dispatch(NETISR_NS, m);
279 			return (error);
280 		}
281 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
282 			m->m_flags |= M_BCAST;
283 		break;
284 #endif
285 	case pseudo_AF_HDRCMPLT:
286 	case AF_UNSPEC:
287 		loop_copy = -1; /* if this is for us, don't do it */
288 		deh = (struct ether_header *)dst->sa_data;
289 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
290 		eh->ether_type = deh->ether_type;
291 		break;
292 
293 	default:
294 		printf("%s: can't handle af%d\n", ifp->if_xname,
295 			dst->sa_family);
296 		gotoerr(EAFNOSUPPORT);
297 	}
298 
299 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
300 		memcpy(eh->ether_shost,
301 		       ((struct ether_header *)dst->sa_data)->ether_shost,
302 		       ETHER_ADDR_LEN);
303 	else
304 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
305 
306 	/*
307 	 * If a simplex interface, and the packet is being sent to our
308 	 * Ethernet address or a broadcast address, loopback a copy.
309 	 * XXX To make a simplex device behave exactly like a duplex
310 	 * device, we should copy in the case of sending to our own
311 	 * ethernet address (thus letting the original actually appear
312 	 * on the wire). However, we don't do that here for security
313 	 * reasons and compatibility with the original behavior.
314 	 */
315 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
316 		int csum_flags = 0;
317 
318 		if (m->m_pkthdr.csum_flags & CSUM_IP)
319 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
320 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
321 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
322 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
323 			struct mbuf *n;
324 
325 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
326 				n->m_pkthdr.csum_flags |= csum_flags;
327 				if (csum_flags & CSUM_DATA_VALID)
328 					n->m_pkthdr.csum_data = 0xffff;
329 				if_simloop(ifp, n, dst->sa_family, hlen);
330 			} else
331 				ifp->if_iqdrops++;
332 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
333 				ETHER_ADDR_LEN) == 0) {
334 			m->m_pkthdr.csum_flags |= csum_flags;
335 			if (csum_flags & CSUM_DATA_VALID)
336 				m->m_pkthdr.csum_data = 0xffff;
337 			if_simloop(ifp, m, dst->sa_family, hlen);
338 			return (0);	/* XXX */
339 		}
340 	}
341 
342 	/* Handle ng_ether(4) processing, if any */
343 	if (ng_ether_output_p != NULL) {
344 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
345 			goto bad;
346 		if (m == NULL)
347 			return (0);
348 	}
349 
350 	/* Continue with link-layer output */
351 	return ether_output_frame(ifp, m);
352 
353 bad:
354 	m_freem(m);
355 	return (error);
356 }
357 
358 /*
359  * Ethernet link layer output routine to send a raw frame to the device.
360  *
361  * This assumes that the 14 byte Ethernet header is present and contiguous
362  * in the first mbuf (if BRIDGE'ing).
363  */
364 int
365 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
366 {
367 	struct ip_fw *rule = NULL;
368 	int error = 0;
369 	struct altq_pktattr pktattr;
370 
371 	ASSERT_SERIALIZED(ifp->if_serializer);
372 
373 	/* Extract info from dummynet tag, ignore others */
374 	while (m->m_type == MT_TAG) {
375 		if (m->m_flags == PACKET_TAG_DUMMYNET) {
376 			rule = ((struct dn_pkt *)m)->rule;
377 			break;
378 		}
379 		m = m->m_next;
380 	}
381 	if (rule != NULL)		/* packet was already bridged */
382 		goto no_bridge;
383 
384 	if (BDG_ACTIVE(ifp)) {
385 		struct ether_header *eh;	/* a pointer suffices */
386 
387 		m->m_pkthdr.rcvif = NULL;
388 		eh = mtod(m, struct ether_header *);
389 		m_adj(m, ETHER_HDR_LEN);
390 		lwkt_serialize_exit(ifp->if_serializer);
391 		m = bdg_forward_ptr(m, eh, ifp);
392 		lwkt_serialize_enter(ifp->if_serializer);
393 		m_freem(m);
394 		return (0);
395 	}
396 
397 no_bridge:
398 	if (ifq_is_enabled(&ifp->if_snd))
399 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
400 	crit_enter();
401 	if (IPFW_LOADED && ether_ipfw != 0) {
402 		struct ether_header save_eh, *eh;
403 
404 		eh = mtod(m, struct ether_header *);
405 		save_eh = *eh;
406 		m_adj(m, ETHER_HDR_LEN);
407 		if (!ether_ipfw_chk(&m, ifp, &rule, eh, FALSE)) {
408 			crit_exit();
409 			if (m != NULL) {
410 				m_freem(m);
411 				return ENOBUFS; /* pkt dropped */
412 			} else
413 				return 0;	/* consumed e.g. in a pipe */
414 		}
415 		eh = mtod(m, struct ether_header *);
416 		/* packet was ok, restore the ethernet header */
417 		if ((void *)(eh + 1) == (void *)m->m_data) {
418 			m->m_data -= ETHER_HDR_LEN ;
419 			m->m_len += ETHER_HDR_LEN ;
420 			m->m_pkthdr.len += ETHER_HDR_LEN ;
421 		} else {
422 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
423 			if (m == NULL) /* nope... */ {
424 				crit_exit();
425 				return ENOBUFS;
426 			}
427 			bcopy(&save_eh, mtod(m, struct ether_header *),
428 			      ETHER_HDR_LEN);
429 		}
430 	}
431 	crit_exit();
432 
433 	/*
434 	 * Queue message on interface, update output statistics if
435 	 * successful, and start output if interface not yet active.
436 	 */
437 	error = ifq_handoff(ifp, m, &pktattr);
438 	return (error);
439 }
440 
441 /*
442  * ipfw processing for ethernet packets (in and out).
443  * The second parameter is NULL from ether_demux(), and ifp from
444  * ether_output_frame(). This section of code could be used from
445  * bridge.c as well as long as we use some extra info
446  * to distinguish that case from ether_output_frame().
447  */
448 static boolean_t
449 ether_ipfw_chk(
450 	struct mbuf **m0,
451 	struct ifnet *dst,
452 	struct ip_fw **rule,
453 	struct ether_header *eh,
454 	boolean_t shared)
455 {
456 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
457 	struct ip_fw_args args;
458 	struct m_tag *mtag;
459 	int i;
460 
461 	if (*rule != NULL && fw_one_pass)
462 		return TRUE; /* dummynet packet, already partially processed */
463 
464 	/*
465 	 * I need some amount of data to be contiguous, and in case others
466 	 * need the packet (shared==TRUE), it also better be in the first mbuf.
467 	 */
468 	i = min((*m0)->m_pkthdr.len, max_protohdr);
469 	if (shared || (*m0)->m_len < i) {
470 		*m0 = m_pullup(*m0, i);
471 		if (*m0 == NULL)
472 			return FALSE;
473 	}
474 
475 	args.m = *m0;		/* the packet we are looking at		*/
476 	args.oif = dst;		/* destination, if any			*/
477 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
478 		m_tag_delete(*m0, mtag);
479 	args.rule = *rule;	/* matching rule to restart		*/
480 	args.next_hop = NULL;	/* we do not support forward yet	*/
481 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
482 	i = ip_fw_chk_ptr(&args);
483 	*m0 = args.m;
484 	*rule = args.rule;
485 
486 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL)	/* drop */
487 		return FALSE;
488 
489 	if (i == 0)					/* a PASS rule.  */
490 		return TRUE;
491 
492 	if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
493 		/*
494 		 * Pass the pkt to dummynet, which consumes it.
495 		 * If shared, make a copy and keep the original.
496 		 */
497 		struct mbuf *m ;
498 
499 		if (shared) {
500 			m = m_copypacket(*m0, MB_DONTWAIT);
501 			if (m == NULL)
502 				return FALSE;
503 		} else {
504 			m = *m0 ;	/* pass the original to dummynet */
505 			*m0 = NULL ;	/* and nothing back to the caller */
506 		}
507 		/*
508 		 * Prepend the header, optimize for the common case of
509 		 * eh pointing into the mbuf.
510 		 */
511 		if ((void *)(eh + 1) == (void *)m->m_data) {
512 			m->m_data -= ETHER_HDR_LEN ;
513 			m->m_len += ETHER_HDR_LEN ;
514 			m->m_pkthdr.len += ETHER_HDR_LEN ;
515 		} else {
516 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
517 			if (m == NULL)
518 				return FALSE;
519 			bcopy(&save_eh, mtod(m, struct ether_header *),
520 			      ETHER_HDR_LEN);
521 		}
522 		ip_dn_io_ptr(m, (i & 0xffff),
523 			     dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
524 		return FALSE;
525 	}
526 	/*
527 	 * XXX at some point add support for divert/forward actions.
528 	 * If none of the above matches, we have to drop the pkt.
529 	 */
530 	return FALSE;
531 }
532 
533 /*
534  * XXX merge this function with ether_input.
535  */
536 static void
537 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
538 {
539 	ether_input(ifp, NULL, m);
540 }
541 
542 /*
543  * Process a received Ethernet packet. We have two different interfaces:
544  * one (conventional) assumes the packet in the mbuf, with the ethernet
545  * header provided separately in *eh. The second one (new) has everything
546  * in the mbuf, and we can tell it because eh == NULL.
547  * The caller MUST MAKE SURE that there are at least
548  * sizeof(struct ether_header) bytes in the first mbuf.
549  *
550  * This allows us to concentrate in one place a bunch of code which
551  * is replicated in all device drivers. Also, many functions called
552  * from ether_input() try to put the eh back into the mbuf, so we
553  * can later propagate the 'contiguous packet' interface to them,
554  * and handle the old interface just here.
555  *
556  * NOTA BENE: for many drivers "eh" is a pointer into the first mbuf or
557  * cluster, right before m_data. So be very careful when working on m,
558  * as you could destroy *eh !!
559  *
560  * First we perform any link layer operations, then continue
561  * to the upper layers with ether_demux().
562  */
563 void
564 ether_input(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
565 {
566 	struct ether_header save_eh;
567 
568 	ASSERT_SERIALIZED(ifp->if_serializer);
569 
570 	if (eh == NULL) {
571 		if (m->m_len < sizeof(struct ether_header)) {
572 			/* XXX error in the caller. */
573 			m_freem(m);
574 			return;
575 		}
576 		m->m_pkthdr.rcvif = ifp;
577 		eh = mtod(m, struct ether_header *);
578 		m_adj(m, sizeof(struct ether_header));
579 		/* XXX */
580 		/* m->m_pkthdr.len = m->m_len; */
581 	}
582 
583 	if (ifp->if_bpf)
584 		bpf_ptap(ifp->if_bpf, m, eh, ETHER_HDR_LEN);
585 
586 	ifp->if_ibytes += m->m_pkthdr.len + (sizeof *eh);
587 
588 	/* Handle ng_ether(4) processing, if any */
589 	if (ng_ether_input_p != NULL) {
590 		lwkt_serialize_exit(ifp->if_serializer);
591 		(*ng_ether_input_p)(ifp, &m, eh);
592 		lwkt_serialize_enter(ifp->if_serializer);
593 		if (m == NULL)
594 			return;
595 	}
596 
597 	/* Check for bridging mode */
598 	if (BDG_ACTIVE(ifp)) {
599 		struct ifnet *bif;
600 
601 		/* Check with bridging code */
602 		if ((bif = bridge_in_ptr(ifp, eh)) == BDG_DROP) {
603 			m_freem(m);
604 			return;
605 		}
606 		if (bif != BDG_LOCAL) {
607 			save_eh = *eh ; /* because it might change */
608 			lwkt_serialize_exit(ifp->if_serializer);
609 			m = bdg_forward_ptr(m, eh, bif); /* needs forwarding */
610 			lwkt_serialize_enter(ifp->if_serializer);
611 			/*
612 			 * Do not continue if bdg_forward_ptr() processed our
613 			 * packet (and cleared the mbuf pointer m) or if
614 			 * it dropped (m_free'd) the packet itself.
615 			 */
616 			if (m == NULL) {
617 			    if (bif == BDG_BCAST || bif == BDG_MCAST)
618 				printf("bdg_forward drop MULTICAST PKT\n");
619 			    return;
620 			}
621 			eh = &save_eh ;
622 		}
623 		if (bif == BDG_LOCAL || bif == BDG_BCAST || bif == BDG_MCAST)
624 			goto recvLocal;		/* receive locally */
625 
626 		/* If not local and not multicast, just drop it */
627 		m_freem(m);
628 		return;
629 	}
630 
631 recvLocal:
632 	/* Continue with upper layer processing */
633 	ether_demux(ifp, eh, m);
634 }
635 
636 /*
637  * Upper layer processing for a received Ethernet packet.
638  */
639 void
640 ether_demux(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
641 {
642 	int isr;
643 	u_short ether_type;
644 	struct ip_fw *rule = NULL;
645 #ifdef NETATALK
646 	struct llc *l;
647 #endif
648 
649 	/* Extract info from dummynet tag, ignore others */
650 	while (m->m_type == MT_TAG) {
651 		if (m->m_flags == PACKET_TAG_DUMMYNET) {
652 			rule = ((struct dn_pkt *)m)->rule;
653 			ifp = m->m_next->m_pkthdr.rcvif;
654 			break;
655 		}
656 		m = m->m_next;
657 	}
658 	if (rule)	/* packet was already bridged */
659 		goto post_stats;
660 
661 	/*
662 	 * Discard packet if upper layers shouldn't see it because
663 	 * it was unicast to a different Ethernet address.  If the
664 	 * driver is working properly, then this situation can only
665 	 * happen when the interface is in promiscuous mode.
666 	 */
667 	if (!BDG_ACTIVE(ifp) &&
668 	    ((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
669 	    (eh->ether_dhost[0] & 1) == 0 &&
670 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
671 		m_freem(m);
672 		return;
673 	}
674 	/* Discard packet if interface is not up */
675 	if (!(ifp->if_flags & IFF_UP)) {
676 		m_freem(m);
677 		return;
678 	}
679 	if (eh->ether_dhost[0] & 1) {
680 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
681 			 ifp->if_addrlen) == 0)
682 			m->m_flags |= M_BCAST;
683 		else
684 			m->m_flags |= M_MCAST;
685 		ifp->if_imcasts++;
686 	}
687 
688 post_stats:
689 	if (IPFW_LOADED && ether_ipfw != 0) {
690 		if (!ether_ipfw_chk(&m, NULL, &rule, eh, FALSE)) {
691 			m_freem(m);
692 			return;
693 		}
694 		eh = mtod(m, struct ether_header *);
695 	}
696 
697 	ether_type = ntohs(eh->ether_type);
698 
699 	switch (ether_type) {
700 #ifdef INET
701 	case ETHERTYPE_IP:
702 		if (ipflow_fastforward(m, ifp->if_serializer))
703 			return;
704 		isr = NETISR_IP;
705 		break;
706 
707 	case ETHERTYPE_ARP:
708 		if (ifp->if_flags & IFF_NOARP) {
709 			/* Discard packet if ARP is disabled on interface */
710 			m_freem(m);
711 			return;
712 		}
713 		isr = NETISR_ARP;
714 		break;
715 #endif
716 
717 #ifdef INET6
718 	case ETHERTYPE_IPV6:
719 		isr = NETISR_IPV6;
720 		break;
721 #endif
722 
723 #ifdef IPX
724 	case ETHERTYPE_IPX:
725 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
726 			return;
727 		isr = NETISR_IPX;
728 		break;
729 #endif
730 
731 #ifdef NS
732 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
733 		isr = NETISR_NS;
734 		break;
735 
736 #endif
737 
738 #ifdef NETATALK
739 	case ETHERTYPE_AT:
740 		isr = NETISR_ATALK1;
741 		break;
742 	case ETHERTYPE_AARP:
743 		isr = NETISR_AARP;
744 		break;
745 #endif
746 
747 	case ETHERTYPE_VLAN:
748 		if (vlan_input_p != NULL)
749 			(*vlan_input_p)(eh, m);
750 		else {
751 			m->m_pkthdr.rcvif->if_noproto++;
752 			m_freem(m);
753 		}
754 		return;
755 
756 	default:
757 #ifdef IPX
758 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
759 			return;
760 #endif
761 #ifdef NS
762 		checksum = mtod(m, ushort *);
763 		/* Novell 802.3 */
764 		if ((ether_type <= ETHERMTU) &&
765 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
766 			if (*checksum == 0xE0E0) {
767 				m->m_pkthdr.len -= 3;
768 				m->m_len -= 3;
769 				m->m_data += 3;
770 			}
771 			isr = NETISR_NS;
772 			break;
773 		}
774 #endif
775 #ifdef NETATALK
776 		if (ether_type > ETHERMTU)
777 			goto dropanyway;
778 		l = mtod(m, struct llc *);
779 		if (l->llc_dsap == LLC_SNAP_LSAP &&
780 		    l->llc_ssap == LLC_SNAP_LSAP &&
781 		    l->llc_control == LLC_UI) {
782 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
783 				 sizeof at_org_code) == 0 &&
784 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
785 				m_adj(m, sizeof(struct llc));
786 				isr = NETISR_ATALK2;
787 				break;
788 			}
789 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
790 				 sizeof aarp_org_code) == 0 &&
791 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
792 				m_adj(m, sizeof(struct llc));
793 				isr = NETISR_AARP;
794 				break;
795 			}
796 		}
797 dropanyway:
798 #endif
799 		if (ng_ether_input_orphan_p != NULL)
800 			(*ng_ether_input_orphan_p)(ifp, m, eh);
801 		else
802 			m_freem(m);
803 		return;
804 	}
805 	netisr_dispatch(isr, m);
806 }
807 
808 /*
809  * Perform common duties while attaching to interface list
810  */
811 
812 void
813 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
814 {
815 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
816 			   serializer);
817 }
818 
819 void
820 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
821 		   lwkt_serialize_t serializer)
822 {
823 	struct sockaddr_dl *sdl;
824 
825 	ifp->if_type = IFT_ETHER;
826 	ifp->if_addrlen = ETHER_ADDR_LEN;
827 	ifp->if_hdrlen = ETHER_HDR_LEN;
828 	if_attach(ifp, serializer);
829 	ifp->if_mtu = ETHERMTU;
830 	if (ifp->if_baudrate == 0)
831 		ifp->if_baudrate = 10000000;
832 	ifp->if_output = ether_output;
833 	ifp->if_input = ether_input_internal;
834 	ifp->if_resolvemulti = ether_resolvemulti;
835 	ifp->if_broadcastaddr = etherbroadcastaddr;
836 	sdl = IF_LLSOCKADDR(ifp);
837 	sdl->sdl_type = IFT_ETHER;
838 	sdl->sdl_alen = ifp->if_addrlen;
839 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
840 	/*
841 	 * XXX Keep the current drivers happy.
842 	 * XXX Remove once all drivers have been cleaned up
843 	 */
844 	if (lla != IFP2AC(ifp)->ac_enaddr)
845 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
846 	bpfattach(ifp, dlt, hdrlen);
847 	if (ng_ether_attach_p != NULL)
848 		(*ng_ether_attach_p)(ifp);
849 	if (BDG_LOADED)
850 		bdgtakeifaces_ptr();
851 
852 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
853 }
854 
855 /*
856  * Perform common duties while detaching an Ethernet interface
857  */
858 void
859 ether_ifdetach(struct ifnet *ifp)
860 {
861 	if_down(ifp);
862 
863 	if (ng_ether_detach_p != NULL)
864 		(*ng_ether_detach_p)(ifp);
865 	bpfdetach(ifp);
866 	if_detach(ifp);
867 	if (BDG_LOADED)
868 		bdgtakeifaces_ptr();
869 }
870 
871 int
872 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
873 {
874 	struct ifaddr *ifa = (struct ifaddr *) data;
875 	struct ifreq *ifr = (struct ifreq *) data;
876 	int error = 0;
877 
878 	ASSERT_SERIALIZED(ifp->if_serializer);
879 
880 	switch (command) {
881 	case SIOCSIFADDR:
882 		ifp->if_flags |= IFF_UP;
883 
884 		switch (ifa->ifa_addr->sa_family) {
885 #ifdef INET
886 		case AF_INET:
887 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
888 			arp_ifinit(ifp, ifa);
889 			break;
890 #endif
891 #ifdef IPX
892 		/*
893 		 * XXX - This code is probably wrong
894 		 */
895 		case AF_IPX:
896 			{
897 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
898 			struct arpcom *ac = IFP2AC(ifp);
899 
900 			if (ipx_nullhost(*ina))
901 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
902 			else
903 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
904 				      sizeof ac->ac_enaddr);
905 
906 			ifp->if_init(ifp->if_softc);	/* Set new address. */
907 			break;
908 			}
909 #endif
910 #ifdef NS
911 		/*
912 		 * XXX - This code is probably wrong
913 		 */
914 		case AF_NS:
915 		{
916 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
917 			struct arpcom *ac = IFP2AC(ifp);
918 
919 			if (ns_nullhost(*ina))
920 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
921 			else
922 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
923 				      sizeof ac->ac_enaddr);
924 
925 			/*
926 			 * Set new address
927 			 */
928 			ifp->if_init(ifp->if_softc);
929 			break;
930 		}
931 #endif
932 		default:
933 			ifp->if_init(ifp->if_softc);
934 			break;
935 		}
936 		break;
937 
938 	case SIOCGIFADDR:
939 		bcopy(IFP2AC(ifp)->ac_enaddr,
940 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
941 		      ETHER_ADDR_LEN);
942 		break;
943 
944 	case SIOCSIFMTU:
945 		/*
946 		 * Set the interface MTU.
947 		 */
948 		if (ifr->ifr_mtu > ETHERMTU) {
949 			error = EINVAL;
950 		} else {
951 			ifp->if_mtu = ifr->ifr_mtu;
952 		}
953 		break;
954 	default:
955 		error = EINVAL;
956 		break;
957 	}
958 	return (error);
959 }
960 
961 int
962 ether_resolvemulti(
963 	struct ifnet *ifp,
964 	struct sockaddr **llsa,
965 	struct sockaddr *sa)
966 {
967 	struct sockaddr_dl *sdl;
968 	struct sockaddr_in *sin;
969 #ifdef INET6
970 	struct sockaddr_in6 *sin6;
971 #endif
972 	u_char *e_addr;
973 
974 	switch(sa->sa_family) {
975 	case AF_LINK:
976 		/*
977 		 * No mapping needed. Just check that it's a valid MC address.
978 		 */
979 		sdl = (struct sockaddr_dl *)sa;
980 		e_addr = LLADDR(sdl);
981 		if ((e_addr[0] & 1) != 1)
982 			return EADDRNOTAVAIL;
983 		*llsa = 0;
984 		return 0;
985 
986 #ifdef INET
987 	case AF_INET:
988 		sin = (struct sockaddr_in *)sa;
989 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
990 			return EADDRNOTAVAIL;
991 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
992 		       M_WAITOK | M_ZERO);
993 		sdl->sdl_len = sizeof *sdl;
994 		sdl->sdl_family = AF_LINK;
995 		sdl->sdl_index = ifp->if_index;
996 		sdl->sdl_type = IFT_ETHER;
997 		sdl->sdl_alen = ETHER_ADDR_LEN;
998 		e_addr = LLADDR(sdl);
999 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1000 		*llsa = (struct sockaddr *)sdl;
1001 		return 0;
1002 #endif
1003 #ifdef INET6
1004 	case AF_INET6:
1005 		sin6 = (struct sockaddr_in6 *)sa;
1006 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1007 			/*
1008 			 * An IP6 address of 0 means listen to all
1009 			 * of the Ethernet multicast address used for IP6.
1010 			 * (This is used for multicast routers.)
1011 			 */
1012 			ifp->if_flags |= IFF_ALLMULTI;
1013 			*llsa = 0;
1014 			return 0;
1015 		}
1016 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1017 			return EADDRNOTAVAIL;
1018 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1019 		       M_WAITOK | M_ZERO);
1020 		sdl->sdl_len = sizeof *sdl;
1021 		sdl->sdl_family = AF_LINK;
1022 		sdl->sdl_index = ifp->if_index;
1023 		sdl->sdl_type = IFT_ETHER;
1024 		sdl->sdl_alen = ETHER_ADDR_LEN;
1025 		e_addr = LLADDR(sdl);
1026 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1027 		*llsa = (struct sockaddr *)sdl;
1028 		return 0;
1029 #endif
1030 
1031 	default:
1032 		/*
1033 		 * Well, the text isn't quite right, but it's the name
1034 		 * that counts...
1035 		 */
1036 		return EAFNOSUPPORT;
1037 	}
1038 }
1039 
1040 #if 0
1041 /*
1042  * This is for reference.  We have a table-driven version
1043  * of the little-endian crc32 generator, which is faster
1044  * than the double-loop.
1045  */
1046 uint32_t
1047 ether_crc32_le(const uint8_t *buf, size_t len)
1048 {
1049 	uint32_t c, crc, carry;
1050 	size_t i, j;
1051 
1052 	crc = 0xffffffffU;	/* initial value */
1053 
1054 	for (i = 0; i < len; i++) {
1055 		c = buf[i];
1056 		for (j = 0; j < 8; j++) {
1057 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1058 			crc >>= 1;
1059 			c >>= 1;
1060 			if (carry)
1061 				crc = (crc ^ ETHER_CRC_POLY_LE);
1062 		}
1063 	}
1064 
1065 	return (crc);
1066 }
1067 #else
1068 uint32_t
1069 ether_crc32_le(const uint8_t *buf, size_t len)
1070 {
1071 	static const uint32_t crctab[] = {
1072 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1073 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1074 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1075 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1076 	};
1077 	uint32_t crc;
1078 	size_t i;
1079 
1080 	crc = 0xffffffffU;	/* initial value */
1081 
1082 	for (i = 0; i < len; i++) {
1083 		crc ^= buf[i];
1084 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1085 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1086 	}
1087 
1088 	return (crc);
1089 }
1090 #endif
1091 
1092 uint32_t
1093 ether_crc32_be(const uint8_t *buf, size_t len)
1094 {
1095 	uint32_t c, crc, carry;
1096 	size_t i, j;
1097 
1098 	crc = 0xffffffffU;	/* initial value */
1099 
1100 	for (i = 0; i < len; i++) {
1101 		c = buf[i];
1102 		for (j = 0; j < 8; j++) {
1103 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1104 			crc <<= 1;
1105 			c >>= 1;
1106 			if (carry)
1107 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1108 		}
1109 	}
1110 
1111 	return (crc);
1112 }
1113 
1114 /*
1115  * find the size of ethernet header, and call classifier
1116  */
1117 void
1118 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
1119 		   struct altq_pktattr *pktattr)
1120 {
1121 	struct ether_header *eh;
1122 	uint16_t ether_type;
1123 	int hlen, af, hdrsize;
1124 	caddr_t hdr;
1125 
1126 	hlen = sizeof(struct ether_header);
1127 	eh = mtod(m, struct ether_header *);
1128 
1129 	ether_type = ntohs(eh->ether_type);
1130 	if (ether_type < ETHERMTU) {
1131 		/* ick! LLC/SNAP */
1132 		struct llc *llc = (struct llc *)(eh + 1);
1133 		hlen += 8;
1134 
1135 		if (m->m_len < hlen ||
1136 		    llc->llc_dsap != LLC_SNAP_LSAP ||
1137 		    llc->llc_ssap != LLC_SNAP_LSAP ||
1138 		    llc->llc_control != LLC_UI)
1139 			goto bad;  /* not snap! */
1140 
1141 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
1142 	}
1143 
1144 	if (ether_type == ETHERTYPE_IP) {
1145 		af = AF_INET;
1146 		hdrsize = 20;  /* sizeof(struct ip) */
1147 #ifdef INET6
1148 	} else if (ether_type == ETHERTYPE_IPV6) {
1149 		af = AF_INET6;
1150 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
1151 #endif
1152 	} else
1153 		goto bad;
1154 
1155 	while (m->m_len <= hlen) {
1156 		hlen -= m->m_len;
1157 		m = m->m_next;
1158 	}
1159 	hdr = m->m_data + hlen;
1160 	if (m->m_len < hlen + hdrsize) {
1161 		/*
1162 		 * ip header is not in a single mbuf.  this should not
1163 		 * happen in the current code.
1164 		 * (todo: use m_pulldown in the future)
1165 		 */
1166 		goto bad;
1167 	}
1168 	m->m_data += hlen;
1169 	m->m_len -= hlen;
1170 	ifq_classify(ifq, m, af, pktattr);
1171 	m->m_data -= hlen;
1172 	m->m_len += hlen;
1173 
1174 	return;
1175 
1176 bad:
1177 	pktattr->pattr_class = NULL;
1178 	pktattr->pattr_hdr = NULL;
1179 	pktattr->pattr_af = AF_UNSPEC;
1180 }
1181